STP1012 Eb.1415051 1 PDF

Copyright by ASTM Int'l (all rights reserved); Mon Jan 16 20:12:47 EST 2012
Downloaded/printed by
(PDVSA Los Teques) pursuant to License Agreement. No further reproductions authorized.
STP 1012
Composite Materials:
Fatigue and Fracture,
Second Volume
Paul A. Lagace, editor
1916 Race Street

Philadelphia, PA 19103
ASTM Publication Code Number (PCN): 04-010120-33
ISBN: 0-8031-1190-8
ISSN: 1040-3086
Copyright 9 by AMERICAN SOCIETY FOR TESTING AND MATERIALS 1989
NOTE
The Society is not responsible, as a body,
for the statements and opinions
advanced in this publication.
Peer Review Policy
Each paper published in this volume was evaluated by three peer reviewers. The authors
addressed all of the reviewers' comments to the satisfaction of both the technical editor(s)
and the ASTM Committee on Publications.
The quality of the papers in this publication reflects not only the obvious efforts of the
authors and the technical editor(s), but also the work of these peer reviewers. The ASTM
Committee on Publications acknowledges with appreciation their dedication and contribution
of time and effort on behalf of ASTM.
Printed in Ann Arbor. MI

April 1989
Foreword
This publication, Composite Materials: Fatigue and Fracture, Second Volume, contains
papers presented at the Second Symposium on Composite Materials: Fatigue and Fracture,
which was held in Cincinnati, Ohio, 27-28 April 1987. The symposium was sponsored by
ASTM Committee D-30 on High Modulus Fibers and Their Composites and Committee E-
24 on Fracture Testing. Paul A. Lagace, Massachusetts Institute of Technology, presided
as symposium chairman and was the editor of this publication.
Contents
Overview
FATIGUE AND DAMAGE GROWTH
Effects of Matrix Toughness on Fatigue Response of Graphite Fiber Composite

Laminates--R. A. SIMONDS, C. E. BAKIS, A N D W. W. S T I N C H C O M B
Fatigue Life Prediction of Cross-Ply Composite Laminates--J. w. LEE,

I. M. D A N I E L , AND G . Y A N I V 19
Consideration of Environmental Conditions for the Fatigue Evaluation o f

Composite Airframe S t r u c t u r e - - M . BERG, J. J. GERHARZ, AND O. GOKGOL 29
Fatigue Damage Development in Notched (0,/---45), Laminates--A. POURSARTIP

AND N. CHINATAMBI 45
Damage Initiation and Growth in Notched Laminates Under Reversed Cyclic

Loading--c. E. BAKIS, H. R. YIH, W. W. STINCHCOMB, AND K. L. REIFSNIDER 66
MODELS AND ANALYSIS
Fatigue of Composite Materials--Damage Model and Life Predictionmw. HWANG

A N D K. S. H A N 87
The Influence of Fiber, Matrix, and Interface on Transverse Cracking in Carbon

Fiber-Reinforced Plastic Cross-Ply Laminates--P. w. M. PETERS 103
Micromechanics of Compression Failures in Open Hole Composite Laminates--

E. G. G U Y N N , W. L. B R A D L E Y , A N D W. E L B E R 118
Dynamic Delamination Buckling in Composite Laminates Under Impact Loading:

Computational SimulationmJ. E. G R A D Y , C. C. C H A M I S , AND R. A. A I E L L O 137
Edge Stresses in Woven Laminates at Low TemperaturesmR. D. KRIZ 150
Predicting lnterlaminar Fatigue Crack Growth Rates in Compressively Loaded

Laminates---A. J. RUSSELL AND K. N. STREET 162
DELAMINATION
Mode I Interlaminar Fracture Toughness of Unidirectional Carbon Fiber

Composites Using a Novel Wedge-Driven Delamination Design--
A. L. G L E S S N E R , M. T. T A K E M O R I , M. A . V A L L A N C E , AND S. K. G I F F O R D 181
Mode II Delamination Fracture Toughness of Unidirectional Graphite/Epoxy

Composites--c. R. C O R L E T O A N D W. L. B R A D L E Y 201
Interlaminar Shear Fracture Toughness and Fatigue Thresholds for Composite

Materials--T. K. O'BRIEN, G. B. MURRI, AND S. A. SALPEKAR 222
Mode ! and Mode II Delamination of Thermosetting and Thermoplastic

Composites--Y. J. PREL, P. DAVIES, M. L. BENZEGGAGH,AND
F.-X. DE CHARENTENAY 251
Free-Edge Delamination Characteristics in $2/CE9000 Glass/Epoxy Laminates

Under Static and Fatigue Loads--w. s. CHAN AND A. S. D. WANG 270
Characterization of Matrix Toughness Effect on Cyclic Delamination Growth in

Graphite Fiber Composites--s. MALL, K. T. YUN, AND N. K. KOCHHAR 296
STRUCTURAL ASPECTS
Delamination Failure Modes in Filament-Wound Composite T u b e s - - R . F. FORAL

AND D. R. G I L B R E A T H 313
Fracture of Pressurized Composite Cylinders with a High Strain-to-Failure Matrix

System--K. J. S A E G E R AND P. A. L A G A C E 326
Fiber Composite Structural Durability and Damage Tolerance: Simplified

Predictive Methods--c. c. CHAMIS A N D C. A. GINTY 338
Impact Damage Characteristics of Bismaleimides and Thermoplastic Composite

Laminates--E. D A N - J U M B O , A. R. L E E W O O D , A N D C. T. SUN 356
Lateral Impact of CompositeCylinders--A. P. C H R I S T O F O R O U , S. R. S W A N S O N , A N D

S. W. B E C K W I T H 373
Oelamination Damage in Central Impacts at Subperforation Speeds on Laminated

Kevlar/Epoxy Plates--L. E. M A L V E R N , C. T. SUN, A N D D. LIU 387
Indexes 407
STP1012-EB/Apr. 1989
Overview
This volume is a collection of peer-reviewed papers based on presentations made at the

Second Symposium on Composite Materials: Fatigue and Fracture, held in Cincinnati, Ohio,
27-28 April 1987. This is the second such symposium sponsored by ASTM Committee D-
30 on High Modulus Fibers and their Composites and devoted to the topics of fatigue and
fracture (the first was held in Dallas/Ft. Worth, Texas, 24-25 October 1984 and is published
as Composite Materials Fatigue and Fracture, ASTM STP 907).
Although this is only the second ASTM symposium dedicated to the important topics of
fatigue and fracture of composite materials, nearly 40% of the papers contained in previous
ASTM STP volumes of Committee D-30 symposia deal with these two topics. This represents
over 250 papers in the last 20 years. Despite all the work conducted, the topics of fatigue
and fracture of composite materials remain as challenges to researchers and practitioners
alike.
In order to use these materials in demanding structural applications, it is necessary to
fully understand the intricacies of their failure process. As we have gained knowledge and
experience over the years, our emphasis has shifted from empirical correlative techniques,
such as the S-N diagram, to characterizing and modelling the specifics of the damage growth
which occur prior to final failure. This is the motivation for the sections on Fatigue and
Damage Growth and on Models and Analysis. New fracture phenomena, such as delami-
nation, have been discovered and great effort placed on their study. Nearly half the papers
in this volume deal with various aspects of delamination, and one entire section is devoted
to this topic. Finally, we have begun to apply our knowledge of fracture and fatigue of
composite materials to structures made of these materials. This is demonstrated in the section
entitled Structural Aspects.
Both researchers and designers in the field of composite materials will find, in this volume,
important and useful information on state of the art work concerning fatigue and fracture
of composite materials. The great majority of papers deal with composite systems of ther-
mosetting epoxies. However, some attention is paid to the more recent thermoplastic sys-
tems. This volume thus contains information of interest to those employing the traditional
epoxy systems as well as those utilizing the emerging thermoplastic systems.
The hard work of the authors, reviewers, and session chairmen enabled 26 of the 27
papers presented at the symposium to be included in this volume. Special thanks are extended
to these session chairmen who aided in the important review process: Lee Gause, Steve
Johnson, Larry Rehfield, and Sam Garbo. Grateful appreciation is also extended to the
authors, reviewers, and the ASTM staff for making this volume an excellent and important
contribution to the composites literature.
Paul A. Lagace
Associate Professor of Aeronautics and
Astronautics, Technology Laboratory for
Advanced Composites, Massachusetts In-
stitute of Technology; symposium chair-
man and editor
1
Downloaded/printed
Copyright9 bybyASTM International www.astm.org
Fatigue and Damage Growth
Robert A. Simonds, 1 Charles E. Bakis, 1 and Wayne W. Stinchcomb 1
Effects of Matrix Toughness on Fatigue

Response of Graphite Fiber Composite
Laminates
REFERENCE: Simonds, R. A., Bakis, C. E., and Stinchcomb, W. W., "Effects of Matrix
Toughness on Fatigue Response of Graphite Fiber Composite Laminates" Composite Mate-
rials: Fatigue and Fracture, Second Volume, A S T M STP 1012, Paul A. Lagace, Ed., American
Society for Testing and Materials, Philadelphia, 1989, pp. 5-18.
ABSTRACT: To evaluate the possible advantages of using a tough resin, polyetheretherketone

(PEEK), as a matrix in a composite material, the fatigue response of (0/+ 45/90/-45),4 AS4/
PEEK graphite-epoxy specimens with drilled center-holes subjected to constant amplitude,
fully reversed cyclic loading was investigated. The results were compared with those of a
similar investigation using T300-5208 graphite-epoxy specimens to determine the effects of
matrix toughness on fully reversed fatigue.
Several load levels provided for lives between l& and 107 cycles to establish baseline (R =
- 1) S-N curves for the two materials. Additionally, damage evaluation methods such as
stiffness monitoring, penetrant-enhanced radiography, and residual strength measurements
were used to identify damage mechanisms; to monitor damage growth process; and to establish
relationships between damage and strength, stiffness, and life of the specimens.
Damage initiated at the hole in all cases, but the damage modes, their subsequent growth
and interaction, and their effects on fatigue response were dependent on cyclic load levels and
material. Damage in the brittle 5208 matrix specimens consisted of matrix cracks followed by
delamination. Over the range of fatigue lives studied, the residual tensile strengths of the 5208
matrix specimens were greater than the initial notched tensile strength. During loading the
compressive strength degraded to values less than initial notched compressive strength, and
the fatigue failure modes were compressive. Similarly, at the lower cyclic load levels, corre-
sponding to lives between 105 and 107 cycles, the PEEK specimens also suffered matrix cracking,
delamination, and an attendant compressive stiffness loss. The specimens failed under the
compressive portion of the cyclic loading. However, at higher cyclic loads, 0~ fiber damage in
the PEEK specimens was observed in addition to matrix damage, and the tensile stiffness
degraded. The failure modes were tensile in these cases.
The results of this study show that matrix toughness influences the long-term behavior of
graphite fiber composites. Although similar matrix damage modes were observed in the two
material systems, the consequences of the damage, as measured by strength, life, and failure
mode, were not similar.
KEY WORDS: composite materials, thermoplastics, polyetheretherketone, fatigue, damage,

strength, life, stiffness
Many current and proposed specifications for composite structures require that materials
maintain certain m i n i m u m properties throughout the service life of the structure. Interest
in so-called damage tolerant materials has motivated the d e v e l o p m e n t of n u m e r o u s new
1 Laboratory engineer, research associate, and professor, respectively, Department of Engineering

Science and Mechanics, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061.
5
Downloaded/printed
Copyright9 by
by ASTM International www.astm.org
6 SECOND COMPOSITE MATERIALS
composite material systems which can be classified in the general category of "tough com-
posites." The toughness of a composite is often measured as interlaminar fracture toughness,
that is, the ability of a laminate to resist initiation and growth of delaminations due to various
loading conditions such as monotonic tensile or compressive forces, cyclic forces, and impact
forces.
The interlaminar fracture toughness values, expressed as the critical strain energy release
rate (Go), for graphite-epoxy composites are typically 200 to 900 J / m 2 (1 to 5 in-lb/in 2) for
mode I and mode II opening, respectively [1]. Small variations in measured values may be
due to different epoxy matrix materials and different test methods (for example, double
cantilever beam test or edge delamination test for mode I). Values of critical strain energy
release rate for graphite polyetheretherketone (PEEK) laminates, determined using similar
test methods and several versions of PEEK matrix, are on the order of five to ten times
greater for mode I and two times greater for mode II than corresponding values for epoxy
matrix composites [2].
By interlaminar fracture toughness criteria alone, graphite PEEK composites are attractive
candidate materials for structures where damage requirements for structural performance
must be satisfied. However, the commonly accepted measure of "toughness" is obtained
under monotonic loading of specimens designed to provide a delamination driven response.
Actual structures and loading conditions may be much different than those used in the
laboratory to obtain basic material property data. For example, stress fields associated with
matrix cracks cause delaminations to initiate at lower values of strain energy release rate
than if matrix cracks were not present [3-5]. O'Brien [6] has observed that threshold values
of Gc for edge delamination driven response under cyclic loading conditions are much lower
than those under monotonic loading conditions. Furthermore, differences between values
of threshold mechanical Gc (neglecting hygrothermal effects) for T300/5208 and AS4/PEEK
graphite composites are much less than differences between corresponding values of G, for
monotonic tests.
The behavior o f " tough" composite materials is most often characterized through carefully
designed, specialized tests in which delamination is the major damage mode governing the
monotonic or cyclic response of the laboratory test specimen. There is, therefore, a need
to evaluate the long-term response of composite materials under loading conditions other
than monotonic and with damage modes in addition to delamination present.
We conducted completely reversed cyclic loading tests on center-notched, graphite PEEK
specimens to meet the following objectives:
1. Identify damage modes and failure modes occurring in tough, thermoplastic matrix
composites due to tension-compression cyclic loading.
2. Determine the fatigue response (including life, stiffness change, and residual tensile
and compressive strength) of tough matrix composites.
3. Compare results with those from similar tests on brittle matrix (epoxy) composites.
Material and Specimens

Quasi-isotropic panels of AS4/PEEK having a stacking sequence of (0/+ 4 5 / 9 0 / - 45)s 4
were manufactured by Imperial Chemical Industries (ICI), England, using a 380~ (716~
cure temperature [7]. The 4-mm (0.162 in.) (nominal)-thick panels were cut at NASA
Langley Research Center into specimens 121 mm (4.75 in.) long by 38.1 mm (1.5 in.) wide
(Fig. 1). A 9.5-mm (0.375 in.) diameter hole was machined into the center of each specimen
using a diamond core drill.
Specimen length was determined according to specifications of a test method to measure
SIMONDS ET AL. ON EFFECTS OF MATRIX TOUGHNESS ON FATIGUE RESPONSE 7
(38J)
1.50
/~ ExtensometI-~er
J
FIG. 1--Specimen geometry.
the response of composite materials under reversed cyclic loads [8]. Under cyclic loads,
damage initiates and grows, affecting the ability of a composite specimen to carry the imposed
loads. The test method allows the test specimen to respond freely to cyclic loads in a natural
mode rather than in a mode controlled by support fixtures. The test method was used
successfully in an earlier study on the fatigue response of center-notched T300/5208 graphite-
epoxy composites.
Experimental Procedure
Mechanical tests were carried out in accordance with the method for reversed cyclic loading
described in Ref 8. The two key features of the method are that the entire specimen gage-
length is unconstrained and the instability-related compressive failure modes are not arti-
ficially suppressed. Constant amplitude, fully reversed (R = - 1 ) sinusoidal loading was
applied at 10 Hz with servohydraulic test frames equipped with hydraulically actuated wedge
grips. One layer of 320-grit utility cloth was placed between the specimen and the wedges
to prevent damage to the top plies of the specimen caused by serrations on each wedge
face. Flat plates with rectangular cutouts to match the dimensions of the specimen were
placed inside the grip housing to ensure optimal alignment of the specimen relative to the
loading axis (Fig. 2).
During the fatigue tests, maximum and minimum strain excursions were continuously
monitored with a 25.4-mm (1 in.) gage-length extensometer centered on the notch. The
extensometer knife-edges were fixed in V-notches engraved in thin aluminum tabs bonded
to the specimen with a compliant silicone adhesive (Fig. 1). Static stiffnesses were periodically
determined by interrupting the test, manually ramping the load through the programmed
load excursions, and measuring the extensometer strain at zero, at maximum, and at min-
imum loads. An overall tension-compression stiffness could be calculated or tension and
compression stiffnesses could be calculated separately. To compare the specimens, stiffnesses
were normalized to their respective values on the first cycle of the test. In this manner, a
characteristic stiffness behavior during the fatigue lifetime could be identified for a particular
load amplitude, and subsequently used in other tests as a basis for estimating the fraction
of life consumed prior to the application of a destructive material evaluation method [8].
Nondestructive and destructive evaluation techniques were performed at several states of
damage throughout the fatigue life as outlined in Refs 8 and 9. The techniques used were
penetrant enhanced x-ray radiography, ultrasonic pulse-echo C-scan, specimen sectioning,
and specimen deply. In addition, residual tensile and compressive strengths were measured
8 SECONDCOMPOSITE MATERIALS
FIG. 2--Specimen gripping arrangement.
at several points in the fatigue lifetime to determine the influence of fatigue damage on the
integrity of the laminate. All specimens subjected to a destructive evaluation technique were
nondestructively evaluated beforehand.
Results of Static Tension and Compression Tests

As reported in Ref 8, the monotonic tensile strength of the T300/5208 notched specimens
was 265 MPa (38.4 ksi) based on the average of seven tests. AS4/PEEK notched specimens
had a significantly higher tensile strength of 351 MPa (50.9 ksi), although that number is
based on only three tests.
Also as reported in Ref 8, compression tests were performed on specimens with several
unsupported test section lengths to determine the effect of unsupported length on specimen
strength and failure mode for the T300/5208 and AS4/PEEK materials. As a result of these
compression tests, an unsupported length of 61.0 mm (2.4 in.) was chosen as the "standard,"
and it was used for all subsequent tests. A compressive strength of 280 MPa (40.7 ksi) was
used as a reference strength for T300/5208, and a slightly higher strength of 290 MPa (42.0
ksi) was determined for the AS4/PEEK using the 61.0 mm (2.4 in.) unsupported length.
Fatigue Test Results

Results of the fatigue tests are presented in Fig. 3 as normalized stress versus log cycles.
Normalized stress is the cyclic stress amplitude divided by the monotonic compressive strength
of the notched laminates described above. Data for specimens that were not cycled to failure
(such as those used for residual strength tests, for example) are denoted with an arrow (--~).
Various cyclic stress levels were used which produced fatigue lives ranging from 2410 to
6 457 100 cycles for the AS4/PEEK material and from 39 330 cycles to more than 2 600 000
cycles for the T300/5208. Results of these tests are presented in Table 1 and also in Fig. 4.
Note that two of the T300/5208 specimens did not actually fail in fatigue although their data
S t M O N D S ET AL. ON EFFECTS O F MATRIX T O U G H N E S S ON FATIGUE R E S P O N S E
I.O
.8
0
, r 0 R~,O=-O0,. 0
4 or o o
NORMALIZED "6r
CYCLIC
STRESS § 4,.o.,
.4
o AS4/PEEK
.2
+ T500/5208
0 I I I
103 10 4 I0 5 I0 6 107
CYCLES
FIG. 3--Normalized cyclic stress versus cycles for T300/5208 and AS4/ P E E K specimens
considered in this study.
TABLE 1--Fatigue test results.
Material Cyclic Stress, Percentage of Monotonic

System MPa (ksi) Comp. Strength Life, cycles
PEEK 207 (30.0) 71 2 410

5208 188 (27.2) 67 39 400
PEEK 186 (27.0) 64 9 350
PEEK 185 (26.8) 64 5 760
PEEK 183 (26.6) 63 5 470
5208 168 (24.4) 60 52 000
5208 168 (24.4) 60 53 000
5208 168 (24.4) 60 80 000
PEEK 170 (24.6) 59 232 260
PEEK 170 (24.6) 58 66 540
PEEK 169 (24.5) 58 373 260
5208 154 (22.4) 55 92 570
5208 154 (22.4) 55 106 360
5208 154 (22.4) 55 108 000
PEEK 160 (23.2) 55 181 380
PEEK 159 (23.1) 55 215 500
5208 154 (22.4) 55 184 000
5208 154 (22.4) 55 195 000
5208 154 (22.4) 55 315 190
PEEK 155 (22.5) 54 166 210
PEEK 155 (22.5) 54 882 660
5208 140 (20.3) 50 158 000
5208 140 (20.3) 50 514 000
PEEK 144 (20.9) 50 6 475 100
5208 140 (20.3) 50 899 000
5208 140 (20.3) 50 925 000
5208 140 (20.3) 50 960 880
5208 133 (19.3) 48 456 000
5208 133 (19.3) 48 850 000
5208 126 (18.3) 45 970 000
5208 126 (18.3) 45 2 100 00IY
5208 126 (18.3) 45 2 600 000~
a Load cycling halted before fatigue failure.
ik
.E~ ' ~ . ~ . ~ . ~ . ~ -I-
~+ o,l:.~ o
NORMALIZED ~ ~r162
CYCLIC ~ %* - ~ _ ~
STRESS t- - .Ir.,'-~
.zl
o DATA, A S 4 / P E E K
-I- DATA, TSOO/520e
.2 -- LS FIT, PEEK
- - LS. FIT, 5208
I I I
jO 3 ~ io ~ io6 io 7
LIFE, CYCLES
FIG. 4--Normalized cyclic stress versus life for specimens cycled to failure.
are included here because of the high numbers of cycles these specimens experienced without
failure. Linear regression straight line fits were applied to the normalized stress versus log
cycles to failure data included in Table 1 (results also shown in Fig. 4). The line fits are not
intended to suggest that the S - N data fit a straight line but, rather, to illustrate an observed
trend: at higher normalized cyclic stresses, the 5208 material appears to have a longer life
than the PEEK material, whereas at lower normalized cyclic stresses, the PEEK material
has a longer life.
Secant stiffness measurements taken during the fatigue tests indicate that the specimens
tend to go through three distinct phases, or stages, during their lives. Two typical tension-
compression stiffness-versus-life relationships, plotted as normalized stiffness versus nor-
malized life, are presented in Fig. 5. Stage I is characterized by a rapid loss of stiffness early
in the fatigue history. Typical stiffness degradation during stage I is 10%, and the stage lasts
for about the first 5 to 15% of the specimen's total life. After the rapid degradation of
stiffness of stage I, stage II stiffness degradation is much more gradual and takes place over
LO
.8
.6
NORMALIZED
STIFFNESS
.4
---- 5 2 0 8 , o-f/o'u,==.55, LIFE=9?.,600 CYC.
.2 PEEK, o-f/%,r LIFE=I66,200 CYC.
0 i i i i i i i i i
0 .2 .4 .6 .8 1.0
NORMALIZED LIFE
FIG. 5--Typical normalized tension--compression stiffness versus life.
a much longer proportion of the specimen's life. At about 90% of the specimen's life, its
stiffness has been reduced to about 80% of its unfatigued stiffness--so stage II has reduced
specimen stiffness by about the same amount as stage I, although it has taken much longer
to do so. At this point, the beginning of stage III, the stiffness degradation accelerates once
again. Stage III is characterized by a rapid loss of stiffness until the specimen eventually
fails.
There are some deviations from the clear stage I, stage II, and stage III stiffness degra-
dation histories of most of the specimens. This can occur if a fatigue test is a particularly
long one as illustrated in Fig. 6. There can be one or even more false stage IIIs where the
stiffness degrades rapidly over a relatively short period of time after a long period of little
or no stiffness change characteristic of stage II. We believe that this phenomenon is caused
by damage propagating past the 1-in. gage length over which the stiffness is being measured
for the long-life tests. As the damage approaches the tabs that locate the extensometer, the
rate of stiffness change increases. Once the damage has grown outside the extensometer
gage length, the extensometer is less sensitive to further damage development.
Another deviation from the clear stage I, stage II, and stage III-type tension--compression
stiffness history is demonstrated by those AS4/PEEK specimens whose lives are short,
typically less than 10 000 cycles. There is not a clear distinction between stage I and stage
II, as can be seen in Fig. 7. There is, instead, a fairly constant tension-compression stiffness
degradation over the first 80% of life followed by a more rapid stiffness degradation close
to failure. The more rapid stiffness degradation is like the stage III experienced by all the
specimens tested to failure.
Because the strains on the specimen were measured at both the maximum tensile stress
and at the minimum compressive stress, it was possible to determine whether the stiffness
changes took place equally in tension and in compression or whether the stiffness change
favored one or the other. A stiffness ratio was calculated by dividing the tensile stiffness by
the compressive stiffness: a value of one indicates that they are equal; a value greater than
one indicates that the tensile stiffness is greater than the compressive stiffness; and a value
less than one indicates that the compressive stiffness is greater than the te:,sile stiffness. For
all 5208 specimens, and for most PEEK specimens, the stiffness ratio started at a value
slightly higher than one and remained there throughout most of the fatigue life of the
specimen as can be seen in Fig. 8. Toward the end of the fatigue lives of these specimens,
1.0
.6
NORMALIZED
STIFFNESS
.4
o'f/~'~,r LIFE=6,475,100 CYC.
.2
I I I I I I I I I
00 .2 .4 .6 .8 IO
NORMALIZED LIFE
FIG. 6---Normalized stiffness versus life f o r a A S 4 / P E E K specimen cycled at a low stress
and with a correspondingly long life.
12 SECONDCOMPOSITEMATERIALS
I.C
.8
.6
NORMALIZED
STIFFNESS
.4 o-f/o-~,9 =.64, LIFE=9400 CYC.
.2
0 I 1 I I I I I I I
0 .2 .4 .6 B 1.0
NORMALIZEDLIFE
FIG. 7--Normalized tension-compression stiffness versus life for a AS4/ P E E K specimen
cycled at a high stress and with a correspondingly short life.
the stiffness ratio increased--indicating that the compression stiffness was degrading more
quickly than the tension stiffness.
The stiffness ratio versus normalized life plots for the AS4/PEEK specimens tested at
high cyclic stresses showed a markedly different behavior: the stiffness ratio started at a
value slightly higher than one as in the other tests, but as the test proceeded, the stiffness
ratio decreased to a value less than one--indicating that the tension stiffness was degrading
more rapidly than the compression stiffness as can also be seen in Fig. 8. The results of the
stiffness ratio versus normalized life plots show an interesting correlation with the observed
failure modes: those specimens whose stiffness ratio was increasing at the end of life had
compressive failures, whereas those specimens whose stiffness ratio was decreasing at the
end of life had tensile failures.
1.2
Il t
STIFFNESS
TENSILE Iii - - - ~
COMRRESSIVE
STIFFNESS
.8 I I I I I I I I I
0 .2 .4 .6 B 1.0
NORMALIZEDLIFE
FIG. 8--Stiffness ratio (tension stiffness~compression stiffness) for A S 4 / P E E K specimens
at two stress levels.
TABLE 2--Residual compressive strength test results.
Material Normalized Normalized Normalized

System Cyclic Stress a Cycles Stiffnessb Strength a
PEEK 0.440 1 000 000 0.910 0.89

PEEK 0.443 1 000 000 0.905 0.90
PEEK 0.500 1 000 000 0.880 0.91
PEEK 0.486 1 000 000 0.870 0.82
PEEK 0.523 6 000 0.953 0.93
PEEK 0.523 130 000 0.855 0.89
PEEK 0.523 243 000 0.690 0.79
PEEK 0.642 2 500 0.944 0.98
PEEK 0.642 2 750 0.833 0.91
PEEK 0.642 2 880 0.610 0.85
" Normalized to 290 MPa (42.0 ksi) unfatigued compressive strength.

b Tension-compression stiffness measured on last cycle normalized to that measured on the first load
cycle.
Results of Residual Strength Tests

The cyclic loading of several of the specimens was halted prior to fatigue failure so that
the specimens could be tested for their residual strength. The portion of the fatigue life the
specimens had e x p e n d e d prior to halting the cyclic loading was estimated by comparing their
stiffness curves with stiffness curves for specimens that had been fatigued to failure at the
same normalized cyclic stress level. Results of the residual strength tests appear in Table 2,
Table 3, and in Fig. 9, which plots residual tensile strength (normalized by m o n o t o n i c tensile
TABLE 3--Residual tensile strength test results.
Material Normalized Normalized Normalized

System Cyclic Stress a Cycles Stiffnessh Strength'
PEEK 0.445 1 000 000 0.920 1.14

PEEK 0.491 1 000 000 0.850 1.24
PEEK 0.523 6 000 0.949 0.98
PEEK 0.523 120 000 0.848 1.18
PEEK 0.523 719 090 0.700 1.14
PEEK 0.642 1 500 0.948 0.97
PEEK 0.642 2 000 0.851 0.88
PEEK 0.642 4 000 0.710 0.75
5208 0.450 300 000 0.915 1.10
5208 0.450 500 000 0.845 1.40
5208 0.500 961 000 0.855 1.20
5208 0.550 25 000 0.900 1.30
5208 0.550 250 000 0.835 1.30
5208 0.550 315 000 0.765 1.20
a Normalized to 290 MPa (42.0 ksi) unfatigued compression strength for PEEK and 280 (40.7 ksi)
unfatigued compression strength for 5208.
b Tension-compression stiffness measured on last cycle normalized to that measured on the first load
cycle.
c Normalized to 351 MPa (50.9 ksi) unfatigued tensile strength for PEEK and 265 MPa (38.4 ksi)
unfatigued compression strength tensile strength for 5208.
t4 SECOND COMPOSITE MATERIALS
1.4 d
j
1'2t
1.0
NORMALIZED .8
RESIDUAL
STRENGTH .6
~' " =.64PEE]<, RESIDUAL COMP.
.4 o " =52 PEEK, RESIDUAL COMR
+ " =.555208, RESIDUALTENS.
x 9 =.455208, RESIDUALTENS.
.2
O0 I I I I I I I I I
.2 .4 6 .8 IO
ESTIMATED NORMALIZED LIFE
FIG. 9--Residual strength test results.
strength) and residual compressive strength (normalized by monotonic compressive strength).

The data are plotted assuming that the specimens had a normalized residual strength of 1.0
at the beginning of life. Specimens were fatigued at specific stress level so that comparisons
could be made of the residual strength for different fractions of life at the same stress level
as well as for different stress levels for the same fraction of life. The figure separates the
data according to stress level. T300/5208 specimens were tested for residual strength in
tension only; the AS4/PEEK specimens were tested for residual strength in both tension
and compression.
All T300/5208 specimens showed an increase in residual tensile strength at the three
fatigue stress levels through a normalized life of 0.95. Similarly, AS4/PEEK specimens
fatigued at a normalized stress of 0.523 showed increasing residual tensile strengths through
at least 95% of life. However, the AS4/PEEK specimens fatigued at a higher normalized
stress, 0.642, showed decreases in residual tensile strength throughout their lives. The AS4/
PEEK specimens tested for residual compressive strength showed losses of strength as a
result of the cyclic loading.
Discussion
The results point to several significant differences in the response of the T300/5208 and
the AS4/PEEK materials to tension-compression cyclic loading, particularly at the higher
cyclic stresses. These differences have not been observed previously in either monotonic or
cyclic delamination-dominated tests, a special case. The first difference is indicated by the
S - N data: the 5208 specimens show superior performance in terms of life at the higher
normalized cyclic stresses and correspondingly shorter lives. This situation is reversed at the
lower normalized cyclic stresses with the PEEK material showing the superior performance.
Another difference appears in the stiffness data taken over the lives of the specimens as
they were tested. The stiffness of the 5208 material degrades in a similar manner regardless
of the cyclic stress level: the compression stiffness degrades more rapidly over the life of
the specimen than the tension stiffness. In contrast, there is a marked and important dif-
ference between the PEEK material stiffness curves recorded at low cyclic stresses and those
recorded at high cyclic stresses. At low cyclic stresses, the compression stiffness degrades
more rapidly than the tension stiffness in a manner similar to the stiffness changes of the
5208 material. At higher cyclic stresses, however, there is a change in the response of the
material so that the tension stiffness degrades more rapidly than the compression stiffness.
This change is particularly apparent if one compares the stiffness ratio taken over the life
of a PEEK specimen fatigued at a high cyclic stress with that of a PEEK specimen fatigued
at a lower cyclic stress, or with that of a 5208 specimen fatigued at any cyclic stress.
A third difference is evident from the results of the residual tensile strength tests. The
T300/5208 specimens and the AS4/PEEK specimens fatigued at lower cyclic stresses showed
increases in residual tensile strength, relative to the initial tensile strength, up to a normalized
life of 0.95. The AS4/PEEK specimens fatigued at higher cyclic stresses showed decreases
in residual tensile strength relative to the initial tensile strength.
Differences in the cyclic tension-compression response observed in the two material
systems are associated with differences in the progressive development of damage during
life. Figure 10 shows zinc-iodide enhanced x-ray radiographs of damage in the T300/5208
and AS4/PEEK notched specimens at the end of their fatigue lives. The damage development
sequence in the 5208 matrix specimens is similar for the high and low cyclic stress levels
when compared on the basis of percentage of life.
In the 5208 matrix specimens, matrix cracks initiate at the hole very early in the life in
the 0 ~ and off-axis plies at random locations through the thickness. By 1000 cycles (stage I
stiffness change), the matrix cracks are uniformly distributed through the thickness of the
specimens. Later in stage I, interlaminar cracks develop on interfaces between the off-axis
plies nearest the surface. Interlaminar cracks also initiate on the first 0/45 interface in regions
of high-density matrix cracks in the 0 ~ and 45 ~ plies. As the reversed cyclic loading continues,
interlaminar cracks develop at similar interfaces through the thickness.
During stage II, matrix cracks continue to extend in length and increase in density.
Delamination growth is slow and is enhanced by the coalescence of microdelaminations in
the regions of high-density crossing cracks in adjacent plies. Damage at the hole during
stage II alters the effective geometry of the hole [10] resulting in a decrease in local strain
concentration and an increase in the residual tensile strength to 140% of the initial notched
tensile strength [9,11]. At the end of stage II, the decrease in compressive stiffness is greater
than that of the tensile stiffness.
In the T300/5208 laminate stage III begins at the onset of more rapid delamination growth,
especially on the interfaces near the surfaces, to form several sublaminates. The tensile
strength decreases during stage III but remains greater than the initial notched tensile
strength throughout life. Delamination growth during the last stage of life reduces the
compressive strength and stiffness of the 5208 matrix laminate, and the specimens fail in
compression. As the delaminated subgroups of plies near the surface deflect out of the plane
of the laminate during the compressive portion of the loading cycle, the compressive load
is carried by the core sublaminate. The actual fatigue failure mode of the T300/5208 spec-
imens is compressive crushing of the core sublaminate.
The process of damage growth in the notched AS4/PEEK specimens tested at the lower
stress levels is similar to that observed in the T300/5208 specimens. Matrix cracks initiate
early in life and are followed by delaminations. Matrix cracks in the 0 ~ plies extend away
from the hole in the direction of loading, giving an elongated appearance to the damage
pattern. Tensile and compressive stiffnesses reduce throughout the life due to the damage.
The decrease in compressive stiffness is greater than the decrease in tensile stiffness, and
the AS4/PEEK specimens cycled at low loads to long lives fail in compression in a manner
similar to that reported for the T300/5208 specimens. The major difference between the
response of AS4/PEEK notched specimens and T300/5208 notched laminates cycled at low
loads is their fatigue lives. At normalized, completely reversed cyclic stresses less than
approximately 0.6, the fatigue lives of the PEEK specimens are greater than those of the
epoxy matrix specimens.
FIG. lO--X-ray radiographs of T300/5208 specimens (top) and AS4/ PEEK specimens
(bottom) cycled to impending failure at high and low normalized stress levels.
The damage process in AS4/PEEK specimens cycled at higher stress levels was much
different than that at low stress levels and that in the epoxy matrix material. Matrix cracks
and delaminations initiate and grow early in the fatigue life; however, the intense damage
zone is confined to a local region at the hole and does not grow longitudinally away from
the hole. The direction of damage growth during fatigue life is transverse to the loading
direction. The damage field in the PEEK specimens cycled at high stress levels contains
much more numerous and extensive macroscopically observable fiber fractures than either
of the other three cases described in the paper. The degradation of tensile stiffness is greater
than that of compressive stiffness, and residual tensile strengths throughout the fatigue lives
are less than the initial notched tensile strength. The fatigue lives of the PEEK specimens
cycled at normalized stresses greater than 0.6 are less than those of T300/5208 specimens
cycled at similar normalized cyclic stress levels.
The most significant effects of matrix material and load level on the fatigue response of
the two composite material systems are the failure modes and the sensitivity of damage,
residual strength, life, and fatigue failure mode to the level of cyclic stress. The PEEK
specimens cycled at high stress levels failed in tension; the laminates described in the other
three cases failed in compression in spite of the fact that the PEEK laminate has a significantly
higher initial tensile strength. Matrix cracks and delaminations alter the local geometry and
continuity of the laminate, thereby changing internal load paths and redistributing stress.
When delaminations are confined to the immediate region around the notch, as observed
in the "tough" PEEK laminate, the local stresses around the notch are not reduced by
changes in the load path and remain high, as in a monotonic tensile test. Consequences of
high cyclic stresses near the notch include broken fibers and damage growth perpendicular
to the loading direction. The generally transverse fatigue fracture shown in Fig. 10 closely
resembles the tensile fracture of monotonically, loaded graphite/PEEK specimens, although
the radiograph of the fatigue fracture shows more matrix cracks and delaminations than are
associated with monotonic tensile fractures.
The second, and equally important, effect is the change of failure mode of the PEEK
specimens with increasing level of cyclic stress. At the lower cyclic stresses, the PEEK
specimens fail in compression and have longer lives than the epoxy matrix specimens.
However, as the amplitude of the cyclic stress increases, the failure mode changes from
compression to tension. The corresponding fatigue lives of the PEEK laminates are less
than the epoxy matrix material cycled at the high stresses but which failed in compression.
Conclusions
Toughened matrix composites are attractive materials because of their resistance to de-
lamination-driven fracture. The "toughness" of such materials is usually measured and
quantified using specially designed monotonic loading tests. Under long-term cyclic loading
conditions, the greater toughness of thermoplastic matrix materials, compared with the
toughness of epoxy matrix materials, does not necessarily translate directly into improved
performance.
The long-term behavior of notched composite laminates is controlled by a number of
interacting factors including material properties, load level, intensity of stress, damage
modes, distribution of damage, and intensity of damage. At the lower cyclic stress levels
used in this investigation, the factors combine in such a way as to give the PEEK matrix
material longer lives than those of the epoxy matrix material. On an equal normalized cyclic
stress basis, the long-term lives of AS4/PEEK specimens were as much as an order of
magnitude greater than corresponding lives of T300/5208 specimens. However, at high load
levels, the damage factors interact in a different way to transition the failure mode of the
PEEK matrix specimens from compression to tension, resulting in fatigue lives shorter than
those for 5208 matrix specimens.
Results of the study demonstrate the need for a more complete understanding of inter-
actions between fiber and matrix constituents of composite materials, damage modes, and
stress fields. An understanding of the factors that control long-term behavior, not just short-
term behavior, must be achieved to satisfy damage requirements for structural performance.
There is also a need for test methods to accurately evaluate the long-term performance of
composite materials to be used in damage-tolerant structures.
Acknowledgments
This work was performed under N A S A Grant NAG-I-343, monitored by Dr. T. K.
O'Brien of the Structures Laboratory, U. S. Army Research and Technology Laboratories,
N A S A Langley Research Center. The authors sincerely appreciate the support of this pro-
gram by these organizations.
References
[1] O'Brien, T. K., Johnston, N. J., Raju, I. S., Morris, D. H., and Simonds, R. A., "Comparisons
of Various Configurations of the Edge Delamination Test For Interlaminar Fracture Toughness,"
in Toughened Composites, ASTM STP 937, American Society for Testing and Materials, Phila-
delphia, 1987.
[2] O'Brien, T. K., "Fatigue Delamination Behavior of PEEK Thermoplastic Composite Laminates,"
in Proceedings, First Technical Conference of the American Society for Composites, Dayton, OH,
1986.
[3] Talug, A. and Reifsnider, K. L., " Analysis of Stress Fields in Composite Laminates with Interior
Cracks," Fibre Science and Technology, Vol. 12. 1979, pp. 201-215.
[4] Wang, A. S. D., Kishore, N. N., and Li, C. A., "Crack Development in Graphite Epoxy Cross-
Ply Laminates Under Uniaxial Tension," Composites Science and Technology, Vol. 24, 1985, pp.
33-46.
[5] O'Brien, T. K., "Analysis of Local Delaminations and Their Influence on Composite Laminate
Behavior," in Delamination and Debonding of Materials, ASTM STP 876, W. S. Johnson, Ed.,
American Society for Testing and Materials, Philadelphia, 1985, pp. 282-297.
[6] O'Brien, T. K. and Murri, G. B., "Interlaminar Shear Fracture Toughness and Fatigue Thresholds
for Composite Materials," this publication.
[7] Dickson, R. F., Jones, C. J., Harris, B., Leach, D. C., and Moore, D. R., "The Environmental
Behaviour of Carbon Fibre Reinforced Polyether Ether Ketone," Journal of Materials Science,
Vol. 20, 1985, pp. 60-70.
[8] Bakis, C. E., Simonds, R. A., and Stinchcomb, W. W., "A Test Method to Measure the Response
of Composite Materials Under Reversed Cyclic Loads," Test Methods & Design Allowables ]•or
Fibrous Composites: 2nd Volume, STP 1003, American Society for Testing and Materials, Phila-
delphia, 1989.
[9] Bakis, C. E. and Stinchcomb, W. W., "Response of Thick, Notched Laminates Subjected to
Tension-Compression Cyclic Loads," in Composite Materials: Fatigue and Fracture, ASTM STP
907, H. T. Hahn, Ed., American Society for Testing and Materials, Philadelphia, 1986, pp. 314-
334.
[10] Sendeckyj, G. P., Stalnaker, H. D., and Kleismit, R. A. in Fatigue of Filamentary Composite
Materials, ASTM STP 636, K. L. Reifsnider and K. N. Lauraitis, Ed., American Society for
Testing and Materials, Philadelphia, 1977, pp. 123-140.
[11] Kress, G. R. and Stinchcomb, W. W., "Fatigue Response of Notched Graphite Epoxy Laminates,"
in Recent Advances in Composites in the United States and Japan, ASTM STP 864, J. R. Vinson
and M. Taya, Eds., American Society for Testing and Materials, Philadelphia, 1985, pp. 173-196.
J a e - W o n Lee, i Isaac M. Daniel, I a n d G e r s h o n Yaniv I
Fatigue Life Prediction of Cross-Ply

Composite Laminates
REFERENCE: Lee, J. W., Daniel, I. M., and Yaniv, G., "Fatigue Life Prediction of Cross-
Ply Composite Laminates," Composite Materials: Fatigueand Fracture, Second Volume, ASTM
STP 1012, Paul A. Lagace, Ed., American Society for Testing and Materials, Philadelphia,
1989, pp. 19-28.
ABSTRACT: Fatigue damage development in cross-ply laminates consists of transverse matrix

cracking followed by longitudinal matrix cracking, local delaminations at crack intersections,
and ultimately longitudinal fiber breakage of the load-carrying plies. Ultimate failure is de-
termined by the manner in which the load is redistributed and transferred into the load-carrying
plies. A model was developed for predicting fatigue life of cross-ply laminates which develop
transverse cracking up to the characteristic damage state (CDS) level before fatigue failure.
The fatigue life consists of two portions: the portion necessary to reach the CDS and the
residual life after attainment of the CDS. A procedure for estimating the CDS life is described
based on the relationship between residual stiffness and crack density and the stress-life
(S-N) curve of the 90~ lamina. The residual life after CDS was obtained by calculating the
stress carried by the 0~ plies and assuming that these plies within the damaged laminate behave
like a 0~ lamina under cyclic loading. Fatigue life predictions were in good agreement with
experimental results for three graphite/epoxy cross-ply laminates.
KEY WORDS: graphite/expoxy, cross-ply laminates, matrix cracking, fatigue, stiffness deg-
radation, fatigue life, damage accumulation
Stress-life properties (S-N curves) are essential in the development of fatigue failure
criteria and cumulative damage models. Because it is not practical to generate extensive
data for the S - N curves of various laminates, it is desirable to have a means for predicting
fatigue life for classes of laminates. The predictive model ideally should be based entirely
on lamina properties, static and fatigue, obtained during material characterization.
Damage in composite laminates consists of the development and accumulation of nu-
merous defects. The basic failure mechanisms (that is, intralaminar and interlaminar matrix
failures and fiber failures) have been observed and identified [1-3]. In the case of cross-ply
laminates, fatigue damage development consists of transverse matrix cracking, followed by
longitudinal matrix cracking, local delaminations at crack intersections, and ultimately lon-
gitudinal fiber breakage in the load-carrying plies [4]. The state of damage is intimately
related to the three most important properties of the material: stiffness, strength, and life.
Of these three, stiffness is related to damage in a more deterministic way. Some analytical
procedures have been developed for relating stiffness reduction to damage state for some
forms of damage, such as matrix cracking and delaminations [5-10].
The first stage of damage consists of development and multiplication of transverse matrix
Research assistant, professor, and visiting assistant professor, respectively, Department of Civil
Engineering, Northwestern University, Evanston, IL 60208.
19
Downloaded/printed
Copyright9 byby
ASTM International www.astm.org
cracks which depend on the current state of stress and the fatigue characteristics (S-N curve)
of the 90 ~ lamina. The crack density increases with number of cycles up to a saturation
point, or the characteristic damage state (CDS). As cracking progresses, stress redistribution
takes place with the 0 ~ plies carrying a bigger share of the load. It is then assumed that the
0 ~ plies within the damaged laminate behave like a 0 ~ lamina under cyclic loading. Using
the experimentally determined stress-life (S-N) curve for the 0~ lamina, it is possible to
predict failure of the 0 ~ plies and, therefore, the life of the laminate.
In this work a procedure is described for estimating the CDS life (that is, the number of
cycles necessary to reach the saturation level of transverse matrix cracking) and, then, the
residual life to failure of the 0~ plies. Predictions are compared with experimental data on
cross-ply graphite/epoxy laminates.
Prediction of CDS Life

Under monotonic loading, the stress-strain curve to failure of a typical cross-ply laminate
consists of three characteristic regions: (1) a linear elastic region without any discernible
damage; (2) a region of decreasing stiffness corresponding to transverse matrix crack ini-
tiation and multiplication; and (3) a nearly linear region of stabilized stiffness corresponding
to the limiting crack density or characteristic damage state (CDS).
Damage under cyclic tensile loading consists of transverse matrix cracking, longitudinal
matrix cracking, delaminations at the intersections of matrix cracks, and fiber fracture.
Damage development depends on the cyclic stress level and thereby on the state of damage
produced during the first loading cycle. Figure 1 shows the variation of transverse crack
density with number of cycles for five cyclic stress levels for a [0/902], graphite/epoxy
laminate. The stress level is expressed as a percentage of static strength. All crack density
curves approach and reach the CDS plateau at some time. At that point, longitudinal matrix
cracking develops with areas of delamination at the intersections of transverse and longi-
50,
0 00-0 0 0 85Z
0 66X
40 ~ 53X
T 0 35z
.E X 2ez
,~ ao-
F-
~ 20-
hi
Q
~ 10-
0
CYCLES (Log n)
FIG. 1--Variation of transverse crack density with number of cycles at various cyclic stress
levels .for a [0/902], graphite/epoxy laminate. (Stress level is expressed as a percentage of the
static strength.)
LEE ET AL. ON FATIGUE LIFE PREDICTION 21
FIG. 2--Master curve of effective normalized transverse modulus of cracked 90~ layer.
tudinal cracks, followed by fiber fractures. It has been observed that this last stage of damage
development begins at approximately 80% of the logarithmic lifetime of the specimen (log
n/log N) for all cyclic stress levels [1l].
One empirical way of estimating the CDS life is by extrapolation of the crack density
curves of Fig. 1 to the maximum density of 18.1 cracks/cm (46 cracks/in.). Another way is
to calculate the stress in the 90 ~ plies at a given stage of damage development and, based
on the fatigue characteristics of the 90~ lamina, predict growth to the next stage of damage
development.
The average stress in the 90 ~ plies of a [0m/90,], laminate for a given transverse matrix
crack density h is obtained as
~2(x) = E2(x)
E,(h) ~" (1)
where
tr2(h) = axial stress in 90 ~ layer at crack density h,
% = applied cyclic stress,
E2(h) = in situ transverse modulus of 90~ plies with crack density h, and
Ex(h) = axial modulus of laminate at crack density h.
It has been shown that the variation with crack density of the in situ transverse modulus
of the 90 ~ layer in normalized form can be represented by a master curve for all cross-ply
layups [5,10]. Figure 2 shows the variation of E2(h)/E~(O) with normalized crack density
~.h, where h is the thickness of the 90~ layer. If this curve is approximated by a straight line
of slope 13, then the effective transverse modulus of the damaged layer is given by
E2(k) = E2(0)[1 + 13~h] (2)
where
E2(0) = transverse modulus of undamaged lamina,
h = transverse layer thickness, and
13 = slope of curve of normalized transverse modulus versus normalized crack density.
The axial modulus of the laminate at a transverse crack density h is
1
E,(h) - 1 + k [El + kE2(h)] (3)
where
E1 = longitudinal lamina modulus and
k = ratio of 90 ~ to 0~ plies ( n / m ) .
From the equations above it follows that
cr2(h) = (1 + k) 1 + 13hh
[~E "/ff k ( l "1- 13Xh) O'a (4)
where
E,
OE = Ez(O)"
It is assumed that the stress above peaks between cracks and will cause a second set of
cracks, thus doubling the crack density to 2h, after an additional number of cycles that can
be obtained from the S - N curve of the 90 ~ lamina. This incremental approach for transverse
cracking predicts damage growth in discrete geometrically progressive steps. For a deter-
ministic definition of transverse strength, the incremental approach is not practical and may
not be accurate enough to predict the exact life at CDS. For this reason, plus the fact that
the CDS life of a cross-ply laminate is a small fraction of the total life, a one-step analytical
estimate of the CDS life was chosen.
A typical S - N curve for the 90~ lamina can best be described by a second degree polynomial
on a log-log scale (Fig. 3).
log ~r2 = a(log N) 2 + b(log N) + log F2re (5)

where
N = number of cycles to failure (or number of cycles to produce an additional set of
transverse cracks in the laminate),
a, b = curve fitting parameters, and
F2r~ = equivalent transverse tensile static strength of lamina corresponding to N = 1.
For a given laminate layup [0m/90,],, the crack density at the CDS level is known or can
be estimated or calculated. For the type of laminate and material studied here, the normalized
crack density at CDS was hh = 0.92. For this value of crack density, the average transverse
stress in the 90~ layer is calculated by Eq 4 and then substituted in Eq 5 to obtain the
corresponding number of cycles necessary to reach the CDS level
1
log Nct, s -~ ~a [ - b - ~ - 4ac] (6)
ii
la. N 1
ol
,Ul,
, .8
E 0
hi
N
J
0
Z
CYCLES (Lo9 N)
FIG. 3--Stress-life curve of graphite~epoxy 90~ lamina.
where
F2r~
c = log tyz(hcDs).
Residual Life After CDS

It was assumed that the damage induced in the 0 ~ plies is negligible up to the point of
the characteristic damage state. It is then assumed that the 0 ~ plies within the d a m a g e d
laminate behave under cyclic loading like a unidirectional 0 ~ laminate. The axial stress carried
by the 0 ~ plies is calculated as
tr,(CDS) - (1 + k ) E , (7)
E~ + kE~ <r~
where
El = longitudinal modulus of lamina, and
E~ = residual effective modulus of 90 ~ layer after CDS.
Failure of the 0 ~ plies and, therefore, of the entire laminate is predicted by using the
experimentally d e t e r m i n e d S - N curve for the 0 ~ lamina (Fig. 4). The latter can be described
by a second degree polynomial on a log-log scale:
log tr~ = a'(log N ) 2 + b'(log N ) + log F~re (8)

where
a', b' = curve fitting parameters, and
F~r, = equivalent tensile longitudinal static strength of lamina.
B.
.9 84
t~
.8
I.iJ
O .7
o-.
UI
N
.J
i ~ ~ ,i ~
CYCLES (Log N)
FIG. 4--Stress-life curve o f graphite~epoxy 0 ~ lamina.
Substituting ~rl from Eq 7 we obtain
log cr~ = a'(log N ) 2 + b'(log N ) + log F ' r (9)

where
cra= applied cyclic stress in laminate,
F'r = F~r,/f = calculated static tensile strength of laminate, and
f = (1 + k)E,/(E, + kE~).
The value F'r in most cases is slightly higher than the equivalent tensile strength of the
laminate, F~r,, obtained from the S - N curve of the laminate. Because the applied cyclic
stress, cro, cannot exceed the static or the equivalent static strength, the equation for residual
life of the laminate takes the form
log r = a'(log N) ~ + b'(log N) + log F'r for ~ ~ FxT e
or (10)
O'a : Fx Te
In the special case of the 0 ~ unidirectional lamina
F~r, = F ' r = F,~e
we obtain Eq 8.
Comparison with Experimental Results

The predicted life for a [0~/90.], laminate cycled at a stress cr. is obtained as
N =NcDs + N,~ (zl)

where
NcDs= the CDS life obtained by Eq 6 and
N~, = the residual life after attainment of the CDS obtained by Eq 10.
Three graphite/epoxy cross-ply laminates of [0/902],, [02/902], and [0/904]~ layups were
investigated. The graphite/epoxy material (AS4/3501-6) was fully characterized physically
and mechanically. Stress-strain curves to failure were obtained under monotonic loading,
and stress-life curves were obtained under cyclic loading for the 0 ~ and 90 ~ laminae. Moduli
and strengths were obtained from the first set of curves, and the curve-fitting parameters
a, b, a', b' were obtained from the second set.
Figure 5 shows the predicted and experimental results for the stress-life curve of the
[0/902]s laminate. Experimental results were fitted by a second order curve. The predicted
curve consists of two branches, a horizontal line and a curved branch as described by Eq
10. The prediction based on linear lamination theory appears to be an upper bound of the
experimental data, but the agreement is considered g o o d . ' A further refinement was intro-
duced in the analysis by taking into account that the stress-strain curve of the 0 ~ lamina is
nonlinear and shows increasing stiffness with stress. An incremental nonlinear lamination
theory was used to calculate the value F'r in Eq 10. In this case the value of F'r is not a
constant for a given laminate but is also a function of the applied cyclic stress. Using the
nonlinear lamination theory, the predicted S - N curve is in better agreement with the ex-
perimental results (Fig. 5).
Figures 6 and 7 show predicted and experimental results for stress-life ( S - N ) curves for
the [02/902], and [0/904], laminates. In the case of the [02/90z], laminate, the prediction based
v
.g
to
to .8'
hi
ne
F--
to
n .7 84 0 EXPERIMENTAL O~
hi
N - - 2rid ORDER CURVE F I T T I N G
M
...... LINEAR LAMINATION THEORY
. . . . NON-LINEAR L A M I N A T I O N THEORY
Oc
0
z
0 i 2 3 4 5 6 '7
CYCLES (Lo9 N)
FIG. 5--Experimental and predicted fatigue life for [0/902], graphite~epoxy laminate.
~a 0
0 ~---~.~..~
.8
r~ .7
0 EXPERIHENTAL
l.i
N - - 2 r i d ORDER CURVE F I T T I N G
.J ...... LINEAR LAHINATION THEORY
~r ----- N O N - L I N E A R L A H J N A T I O N THEORY
n, 96
o i ~ w ~ ~ 6 7
CYCLES (Log N)
FIG. 6--Experimental and predicted fatigue life for [02/90z], graphite~epoxy laminate.
on linear lamination theory is in very good agreement with experimental results and their
curve fit. In the case of the [0/904], laminate, the predicted results obtained by linear
lamination theory represent an upper envelope of the experimental data. In both cases the
predictions based on incremental nonlinear lamination theory were in better agreement with
experimental data.
Because all fatigue data were obtained for the first million cycles only, comparisons
between theory and experiment shown in Figs. 5 through 7 are limited to this range. In all
cases above, this range corresponds to cyclic stress levels high enough to produce transverse
cracking up to the CDS during the first loading cycle (that is, NcDs -~ 1). Calculations were
made for the [0/902], laminate loaded under cyclic stress levels of 66% and 53% of the static
~CU~--O-" "~
.g
0,"
In .8
t.U
n-
l--
U}
o EXPERIMENTAL
D .7
UJ - - 2 r i d ORDER CURVE F I T T I N G
N
..J ...... LINEAR LARINATIDN THEORY
X (5 . . . . NON-LINEAR LAMINATION THEORY

n, 9
Q
Z
CYCLES CLog N)
FIG. 7--Experimental and predicted fatigue life for [0/90s], graphite~epoxy laminate.
o1~.
Ul
tit
,,, .7'
r,,-
F-
In
0 .8' O EXPER1NENTAL
bJ
N
\
- - g r i d ~ROER CURVE F I T T I N G
,-I
<: ...... L I N E A R L A M I N A T I O N THEORY
9. r .5" \
1]r . . . . NON-LINEAR LANINATION THEORY
t-i
Z
i ~ w ~ ~ d ~ a ~ Ib 1"I ~
CYCLES (Log N)
FIG. 8--Experimental and predicted fatigue life for [0/902], graphite~epoxy laminate in-
cluding long-term prediction.
strength of the laminate. Figure 8 shows that these predictions fall in line with the extrap-
olated experimental results.
Summary and Conclusions

A procedure was described for predicting the fatigue life of cross-ply laminates based on
the static and fatigue properties of the unidirectional lamina. The predicted fatigue life
consists of two parts.
The first part corresponds to the first stage of damage development consisting of transverse
matrix cracking and multiplication up to the limiting characteristic damage state (CDS).
This part is based entirely on the state of stress in the 90 ~ layer as it is being damaged and
on the S - N curve of the 90 ~ lamina.
The second part of the fatigue life following attainment of CDS is based on the state of
stress in the 0 ~ plies and the S - N curve of the 0 ~ lamina. It is assumed that the 0 ~ plies
within the damaged laminate behave under cyclic loading like a 0 ~ unidirectional lamina.
The total life prediced is taken as the sum of the two parts.
Predicted results were in good agreement with experimental results. The agreement be-
came even better when the nonlinear behavior of the 0 ~ lamina was taken into account by
using an incremental nonlinear lamination theory.
Acknowledgment
The work described here was sponsored by the Office of Naval Research (ONR). We are
grateful to Y. Rajapakse of ONR for his encouragement and cooperation, and to Jan Boudart
for typing the manuscript.
References
[1] Reifsnider, K. L., Henneke, E. G. II, and Stinchcomb, W. W., "Defect Property Relationships
in Composite Materials," AFML-TR-76-81 Part IV, Air Force Wright Aeronautical Laboratories,
Ohio, 1979.
[2] Wang, A. S. D. and Crossman, E W., "Initiation and Growth of Transverse Cracks and Edge
Delamination in Composite Laminates--Part I: An Energy Method," Journal of Composite Ma-
terials, Special Issue, June 1980.
[3] Crossman, E W., Warren, W. J., Wang, A. S. D., and Law, G. L., "'Initiation and Growth of
Transverse Cracks and Edge Delamination in Composite Laminates--Part II: Experimental Cor-
relation," Journal of Composite Materials, Special Issue, June 1980.
[4] Charewicz, A. and Daniel, I. M., "Damage Mechanisms and Accumulation in Graphite/Epoxy
Laminates," in Composite Materials: Fatigue and Fracture, ASTM STP 907, H. T. Hahn, Ed.,
[5] Ryder, J. T. and Crossman, E W., "A Study of Stiffness, Residual Strength and Fatigue Life
Relationships for Composite Laminates," NASA CR-172211, Oct. 1983.
[6] Talreja, R., "Transverse Cracking and Stiffness Reduction in Composite Laminates," Journal of
Composite Materials, Vol. 19, 1985, pp. 355-375.
[7] Dvorak, G. J-, Laws, N., and Hejazi, M., "Analysis of Progressive Matrix Cracking in Composite
Laminates--I. Thermoelastic Properties of a Ply with Cracks," Journal of Composite Materials,
Vol. 19, May 1985, pp. 216-234.
[8] Hashin, Z., "Analysis of Cracked Laminates: A Variational Approach," Mechanics of Materials,
Vol. 4, 1985, pp. 121-136.
[9] Ogin, S. L., Smith, P. A., and Beaumont, P. W. R., "Matrix Cracking and Stiffness Reduction
During the Fatigue of a [0/90Is GFRP Laminate," Composites Science and Technology, Vol. 22,
1985, pp. 23-31.
[10] Yaniv, G. and Daniel, I. M., "Progressive Damage Analysis of Crossply Composite Laminates,"
in review.
[11] Daniel, I. M., Lee, J. W., and Yaniv, G., "Damage Mechanisms and Stiffness Degradation in
Graphite/Epoxy Composites," ICCM and ECCM (Sixth International Conference on Composite
Materials and Second European Conference on Composite Materials), E L. Matthews, N. C. R.
Buskell, J. M. Hodgkinson, and J. Morton, Eds., Elsevier Applied Science, London, 1987, Vol.
4, pp. 4.129-4.138.
Matthias Berg, i Johann J. Gerharz, ~ and Oguz GOkg6l 2
Consideration of Environmental Conditions

for the Fatigue Evaluation of Composite
Airframe Structure
REFERENCE: Berg, M., Gerharz, J. J., and GOkg61, O., "Consideration of Environmental
Conditions for the Fatigue Evaluation of Composite Airframe Structure," Composite Materials:
Fatigue and Fracture, Second Volume, ASTM STP 1012, Paul A. Lagace, Ed., American
ABSTRACT: This paper reports on experimentally derived compensation factors to account

for missing environmental history (temperature and humidity) during fatigue testing of large
composite structures (components). Compensation factors are either factors by which the
fatigue load level of component fatigue tests is increased or by which the required residual
strength has to be increased if preloading did not include environmental history.
Preconditioned coupon specimens were subjected to accelerated quasirealistic mechanical
and environmental flight-by-flight loading, and other coupon specimens were loaded exactly
like the composite airframe structure (that is, with less complete simulated environment).
Temperature and moisture content in residual strength testing were the same for surviving
coupon specimens and the composite airframe structure. Corresponding to composite airframe
structures of the Tornado and the Airbus fin box, the materials involved were the 175~ curing
systems Ciba 914C and Fiberite 1076 E, and the 125~ curing systems Ciba 913C and Hexcel
F550 with T300 carbon fibers in all resins. Different types of laminates and specimens were
used to generate predominantly matrix shear stresses or fiber normal stresses within the coupon
specimens.
At present design load levels, the fatigue life to be proven was reached without failures of
the coupon specimens subjected to the flight-by-flight load conditions of the component test
or to quasirealistic combined mechanical and environmental loading. Only small differences
in residual strength were noticed among the surviving specimens. At higher fatigue load levels
the superimposed more complete environmental test conditions decreased the life to failure
significantly, by factors between 1.5 and 9.1.
KEY WORDS: composite structure certification, graphite epoxy, thermal cycling, environ-
mental effects, fatigue life, residual strength
A b s o r b e d moisture and high temperatures soften the epoxy resin matrix material. This
affects the compressive strength of the composite, because in compression the matrix sup-
ports the small-diameter carbon fibers against buckling (without the matrix the carbon fibers
would not be able to withstand compression). Many investigators have shown this influence
(see, for example, R e f 1). Low temperatures and large temperature cycles lead to matrix
cracks in off-axis layers due to internal stresses and internal stress cycles, respectively [2].
Fatigue research engineers, Fraunhofer-Institut fiir Betriebsfestigkeit (LBF), Darmstadt, Germany.

: Chief of Stress Analysis Department, Messerschmitt-B61kow-Blohm (MBB-UT), Hamburg, Ger-
many.
29
Downloaded/printed
Copyright9 by
by ASTM lntcrnational www.astm.org
These thermal cracks stimulate further matrix cracks (delaminations) to initiate when the
laminate is mechanically fatigue loaded [3]. In consequence, certification of composite
airplane structure requires the consideration of the effects of environmental history expected
in service [4,5]. But fatigue testing of components with realistic simultaneously occurring
temperature and load cycles of the wet composite structure is very expensive and time
consuming. However, it is practicable to do this with smaller specimens [6, 7]. Then the
"compensation factor" approach may be used to meet the certification requirements. The
objective of this paper is to show this approach, the testing and test results involved, and
the compensation factors derived.
Concept of Compensation Factor

The objective is to derive residual strength factors KR or fatigue load factors KL as defined
in Fig, 1, where
O'r.Strap, Nstmp = residual strength, life to failure, respectively, of coupon specimens loaded
with a simplistic program exactly like the composite structure for the fatigue
proof
O'r.Real, NReal = residual strength, life to failure, respectively, of coupon specimens loaded
with a quasirealistic mechanical and environmental flight-by-flight loading
program
The life ratio J~simp/~Rcal m u s t be transformed to the load ratio
KL = tr(N.=.,)/~(Ns.mp)
on the S - N curve for simplistic loading. The power law relationship is used.
G=B.N-S
End Test Compensation Application of CF's

Loading: of Factors (CF)" in Component Testing
Test: Results: (Fatigue Proof of
Structure):
J Simplistic I
j _ Residual Strength to J
Required
~ KR = Or.si~p. be Proven=KRxResidual J
~ Strength after Simplistic
Life Load ng
or
//~ Life to
Failure KL=~RR'aI"] J
c ____
ot,gue
I simp,,sticLoo
KLLoodX
..v., forI
Loading--
DesignLevelFatigue
[ Quasi Realistic I
a Ratio of Residual Strength t>Ratio of Life to Failure c S-Slope of S-N-Curve Corresponding

to 0 = B N "l
FIG. 1--The compensation factor concept.
BERG ET AL. ON FATIGUE EVALUATION OF COMPOSITE AIRFRAME STRUCTURE 31
where
= average fatigue strength,
B = equivalent static strength,
N = average life to fracture, and
S = slope of S - N curve.
Inserting the life and load ratio in the equation results in
KL = (Ns,mp/NR=~)s
The compensation factor KR or KL is then applied in the fatigue proof testing program for
the composite structure. To compensate for the incomplete or missing environmental con-
ditions in the proof testing of the composite structure, either the fatigue load level is increased
by the factor KL or the required residual strength is increased by the factor KR. Thus the
effect of the different loading (simplistic and quasirealistic) on the laminate property is
transferred from composite coupon specimens to a composite component.
As already mentioned, the presence of compressive stresses in fiber direction is essential
to the environmental effect. Therefore the compensation factor concept (cfc) should work
as long as both specimen and component have laminates with layers loaded in compression
parallel to the fibers. The structure to which the cfc is applied will experience compressive
and torsional loading during service. Therefore the investigated laminates of the thin-walled
structure contain 0 ~ and -+45~ plies. The specimens are axially loaded parallel to the fibers
of the 0 ~ plies. The 45 ~ plies react to the in-plane shear loading in the composite component
by tensile and compressive stresses in fiber direction. Thus, in the composite component
more layers were loaded in compression than in the specimen. An influence on strength
behavior due to this, if any, was not taken into account. It is assumed that if the component
would have been subjected to the simplistic and quasirealistic loading, the same factors KL
and KR would result. A component comprises various critical details, some of which were
represented by the different types of coupon specimens applied here (see Fig. 2).
Loading Programs
The load spectra of the mechanical loading are shown in Fig. 3. The simplistic and realistic
loading did not differ in mechanical loading but in their environmental part.
Within the Airbus fin box program the simplistic loading consisted of
9 specimen moisture condition "wet," moisture content corresponding to saturation level

for 75% relative humidity,
9 constant room temperature (RT) during mechanical loading with
9 the flight-by-flight loading sequence for the Airbus fin, and
9 test temperature +70~ and wet condition during residual strength tests.
Within the Tornado program the simplistic loading consisted of
9 specimen moisture condition "dry," moisture content corresponding to normal atmo-

sphere humidity,
9 constant RT ( + 25~ during mechanical loading with
9 the wing loading sequence "The FAtigue Loading STAndard For Fighter Aircraft'"
(FALSTAFF) described in Ref 8, and
9 test temperature + 120~ and dry condition during residual strength tests.
c0
m
C~
thick !.1 mm t ~1.5 mm thick 0
Z
CJ
13/0/0], 0z/+4c.
g
E
"13
0
:astener m
teat CT-
E
m
>
r-
~/,.3mm
othiCk
[o/[_* 4s)]~
-2.1 mm thick
[or,.-~s/o~/_.4s~].
Specimen;
Type la: CFba 913C/T300 Type 20: Cibo 913C/T300 Type 3o: Cibo 91l.C/T300 type 40: Ciba 91/,C/T300
Type lb: HexcelF550/T300 rype2b: HexcelF550/T300 Type 3b: Fiberite1076E/T300 type 4b: Fiberite 1076E/T300
Progrom: Airbus. Fin Box Airbus.Fin Box Tornodo Tornodo

FIG. 2--Form of coupon specimens.
These were the environmental conditions applied in fatigue and residual strength testing
of the full-scale composite structures. The plain coupon specimens (types 3a and 3b) were
subjected to predominant compressive fatigue loading of the wing upper surface, and the
jointed specimens (types 4a and 4b) were subjected to predominant tensile fatigue loading
of the wing lower surface; correspondingly residual compression strength and residual tension
strength were determined.
For the quasi-realistic loading of the Airbus fin box, 20 flight temperature profiles shown
in Table I were defined? Each of the 1000 defined flight load sequences was connected with
one of the 20 flight temperature profiles. A typical connection is shown in Fig. 3. As within
the simplistic loading program, the specimens were moisturized to the saturation level
corresponding to 75% relative humidity.
For the quasi-realistic loading of Tornado composite structures, the specimens of types 3
and 4 were first moisturized to the saturation level corresponding to 85% relative humidity.
The temperature profiles shown on the right half of Table 1 were defined2 Each of the 200
flight load sequences of the FALSTAFF cycle block was connected with a temperature
profile.
The sequence of FALSTAFF flight numbers shown in Table 1 has the following meaning3:
9 the flight load sequence of the second flight in the cycle block is connected to the
temperature profile + 25~ - 43~ + ll0~ + 25~
9 the flight load sequence of the third flight is combined with + 25~
9 the flight load sequence of the fourth flight is connected with the temperature profile
+ 25~ -43~ + l l0~ + 25~ like the second flight of the cycle block;
9 the flight load sequence of the flights between the fourth and the twenty-fourth flight
are connected with +25~
9 and so forth.
This adds up to 69 flight load sequences being connected to the temperature profile with
-43~ and + ll0~ and the remaining 131 flight load sequences of the FALSTAFF cycle
block being combined with + 25~ Once in 1000 flights the temperature profile connected
to FALSTAFF-flight No. 4 had a maximum temperature of + 130~ instead of + 110~ (see
Table 1). A typical connection of the temperature profile and a flight load sequence is also
shown in Fig. 3 for the Tornado program. 3
The quasi-realistic programs for Airbus and Tornado programs include moisture periods
to keep the average moisture content of the specimens within prescribed tolerance limits.
During these periods and during temperature changes the loads were kept constant near
zero. The maximum loads of the spectra used in the Airbus fin box and Tornado programs
did occur at temperatures of - 5 ~ and + 25~ respectively. The load is zero at the high
temperature of +65~ simulating solar heating of the fin box while the airplane is on the
ground, and at + 110~ the load is equal to or less than 67% of the maximum spectrum
load simulating aerodynamic heating during a fighter mission with high speed at low altitude
on a hot day.
Material and Specimens

All specimens were made from carbon/epoxy prepregs consisting of 60% by volume of
conventional T300 fiber and epoxy resin systems of different manufacturers. From the
3 More details on the loading programs may be obtained directly from the authors.
Airbus Fin Box P r o g r a m
1.0 n = 1]
18
t
3/*6
O.S-
10175
Load Spectra:
d
0
1 10i Cumulative Frequency
p e r "Block" (1000 Rights)
-O.S
- t.O
70~ RH
$5eC .6Sot 9 -. . . .
II r"l i~ I#' ITwice
No,stvre Recovery ,
; I W ' ~ 1000 Fl~ghtsl
i / Temperotuee J I
,: \/
e \p,of,le TS i /
Typical
Connection of
Temperature
and Load
Sequence:
FIG. 3--Fatigue loads and sequences of temperature and loads.
prepreg material, laminate plates were manufactured by MBB and Dornier in autoclaves
with curing cycles recommended by the prepreg manufacturers. The specimens were cut
from the cured laminate plates. The types of specimens, the laminate structure, and the
candidate material systems for the Airbus fin box and the Tornado composite structures are
shown in Fig. 2. For the specimens of types 1 and 2, fabric prepregs (satin weave) were
used for the -+45~ layers and tape prepregs were used for the 0 ~ layers. The laminates of
specimen types 3 and 4 were uniformly made of tape prepregs. Curing temperatures were
+ 125~ for the Airbus fin box resins Ciba 913C and Hexcel F550 and + 175~ for the
Tornado composite structure resins Ciba 914C and Fiberite 1076E. The high temperature
curing systems have potentially higher glass transition temperatures, therefore the difference
in curing temperature corresponds to the difference in the maximum temperatures expected
during service of the transport and fighter airplane, as reflected in the loading program
described above.
The type 1 specimens were axially loaded and the type 2 specimens were loaded in a short
beam interlaminar shear test fixture similar to ASTM Test Method for Apparent Interlaminar
Shear Strength of Parallel Fiber Composites by Short Beam Method (D 2344-76); they failed
when 0 ~ layers broke and interlaminar shear strength was reached, respectively. Specimen
type 2 was included to represent the critical interlaminar stress areas (for example, ply drop
offs). The specimen types 3 and 4 were applied in earlier investigations [9] with the same
Tornado Program
.03.
.O2
9o1 tension I
9
ol
I,
-02 Comp,ess,o.
-03
FALSTAFF[81
~ -O.L
-05-
twin9 UpperSurloce o s
-OG
-O7.
-08.
O9
-10 ]
&
'~ T .130oC [Once Within1000 Flights)
~9I ~ , , - - .700C 95%RH
" i -25"C / ~\ , Moisture R. . . . . . y
~ ~ ~ (Twice Within 200 Flights)
I , -~3~ / I 0.,,.0,.~.,
~ - - ~ P~ i,m;--
FIG. 3--continued.
laminate and material (Ciba 914C/T300) and were chosen for reasons of comparability. The
jointed specimen (type 4) had two 4-mm Hi-Lok bolts installed with a clamping torque of
2 Nm. The hole was sized to receive a neat fit (no clearance, no interference). Before
installation, the contact areas of the bolt and the laying surfaces of the joint were covered
with a sealing compound. For the unnotched laminates, the critical axial loading in the 0 ~
direction was compression. Therefore the plain specimen axial loading was compression or
tension-compression, which required the use of antibuckling guides. The support took the
form of a single longitudinal line (width = 4 mm) running along the axis of symmetry of
the specimen leaving the free edges of the specimens uncovered, as shown with specimen
type 1 in Fig. 2. To avoid fretting, the contact areas of the antibuckling guides were covered
with 1-mm-thick Teflon sheets.
Test Setup
The combined mechanical and thermal loading of the specimens was performed on a
testing apparatus shown in Fig. 4. 3 The part for mechanical loading consists of four hydraulic
cylinders with a loading capacity of 20 kN per specimen.
The construction of the four load frames is such that two specimens are loaded in parallel
by one hydraulic cylinder. In this way eight specimens are loaded at the same time. During
o)
T A B L E 1--Connection of temperature profiles and flight load sequences. m

(3
Airbus Fin Box Program O
z
o
Flight Load Sequence, (3
N u m b e r of Occurrence Tornado Program 0
"o
Flight Load Sequence of 0
N u m b e r of Go
Temperature Profiles, ~ A B C D E F Total T e m p e r a t u r e Profiles, ~ F A L S T A F F Flight No.: Flights .-I
m
T 1 +65--*+40---,+15 1 2 6 9 +25--->-43--->+ 110--->+25 2, 4, 24, 30, 34, 39, 41,45
T 2 -20 1 5 6 46, 47, 51, 55, 59, 63, 67
T3 +65--~ + 15 1 1 7 30 39 m
71, 75, 77, 78, 80, 83, 84
T4 -~-20 1 4 20 75 100 86, 87, 89, 90, 91, 92, 94
T5 + 65--* + 15---, - 5-~ + 15 2 6 24 32 i-"-
96, 97, 98, 100, 102, 103 03
T 6 + 15---, - 2 0 1 1 5 7 104, 106, 108, 1(19, 111
T7 + 65--> + 15--, - 5--, + 15 2 10 36 48 113, 116, 119, 122, 127
T8 + 15---~- 5r--, - 20 2 1 5 22 30 128, 131, 132, 135, 139
T9 +40-->+15 2 7 25 34 158, 166, 188, 170, 172
T 10 +15-->-20-->-5 1 1 2 175, 176, 178, 179, 180
Tll + 40--* + 15--* - 5---~+ 15 1 2 8 39 144 194 182, 183, 186, 187, 188
T 12 + 15---,- 5---,- 2 0 - ~ - 5 1 2 5 8 190, 194, 195, 196 69
T 13 + 40---~+ 15--, - 5--~ + 15 1 1 6 24 32
T 14 + 15--*-20--*-5 1 1
T 15 +40---,-5--*+15 2 4 19 70 95 + 25--* - 43--->+ 130--->+ 25 4 (once in 1000 flights)
T 16 + 15--*-5---,-20--*-5 1 1 4 12 58 214 290
T 17 +40--->-5 1 1 2
T 18 + 40--~ - 5---, - 20-~ - 5 1 5 18 24 + 25 R e m a i n i n g flights of the
T 19 + 40--~ + 15--, - 20---~- 5 1 1 5 17 24 200 F A L S T A F F flights 131
T20 +40---,-5-->-20--,-5 1 5 17 23
Total N u m b e r of Flights: 1000 Total N u m b e r of Flights: 200
FIG. 4--General view of test setup.
the test of the plain specimens only the test sections of the specimens were located in an
insulated duct. Through this duct preconditioned air was blown by a fan with a velocity of
about 10 m/s. The air can be heated up to + 130~ by a high-performance electric heater
and moisturized by a humidifier in parallel, or it can be cooled down to - 5 0 ~ in a cold
chamber. The circulation of hot and cold air is controlled by pneumatically actuated flaps.
Depending on the type of specimen the mean change rate of the specimen temperature is
between 30~ and 50~
A computer controls the functions of the testing apparatus and provides the command
signal for flight-by-flight loading. The testing setup includes several recording units to register
load, temperature, deformation, and so forth [10].
Test Procedure
The static and the fatigue tests were performed with the testing equipment described
above. The loading speed was 1 mm/min during the static and between 10 and 20 Hz during
the fatigue test. The axially loaded specimens were clamped between plates with smooth
contact surfaces. Load introduction was by shear through friction. The preconditioned air
heated, cooled, and moisturized the loaded sections of the specimens.
The specimen temperature was controlled by continuous measurements of the air flow
temperature after a relationship between the internal temperature of the specimens and the
air temperature had been established.
For the moisture content control, travelers accompanying the wet specimens were weighed
frequently and the weights were recorded.
During deicing periods of the cold chambers, humidified air was blown over the specimens
so they could absorb the amount of water they had lost during heating periods. As signified
in Fig. 3 the deicing was done twice during each cycle block; the simultaneous moisture
(~
m
0
0
Tornado Program Airbus Fin Box P r o g r a m z
!:3 Simplistic Loading / Test Temperature +70~
Test Temperature ,120~ o
Wet Condition
O Ouosi-Realislic Loading 0
"0
0
(/)
Type 3a Type 3b Type ha Type l,b Type 2a Type 2b
o
2,0 91/,C/T300 1076E/T300 91/,C/T300 1076E/T300 913C/T300 F550/T300 -t
o m
I
Jointed Specimens I 2.8 E
II - Plain Specimens - )
"~ 1.8 I 2.6
J~
I m
r
2J. --n
e-
~ 1.6 L. 2,2 N
r-
df r
-~§ 90 a 2.0
(/I b
t.g 1.8
t-
1.6
~ t.2 .c_ 1.4
E 1.2
Fatique Design Load Level o Fatigue Design Load Level
1,0 p. a- 1.0
E
I I I 0,8 I 9"-
-----.._Life to b e . ~ " ~ Life to be ~ 1 1 h-.... Life to be
0.8
{ "0 0.6 I
Proven I I Proven I I Proven -
I t I I 0.4 i I
I I I I
0.6-
a Percent Probability of Survival

FIG. 5--Effect of environment on residual strength.
D0
m
m
--I
)>
.t-
O
Z
T o r n o d o Progrom Airbus Fin Box Program -11
2,0.
n Simplistic Loading J C
I Specimen Type lo. 913C/T300 m
L.._ 1 : 8.52 =i I O OuasJ-Realistic Loading 1.4 m
1.8
Specimen Type 30. 914C/T300 r-
16
I E
Specimen Type lb. 1:1./d,

6
t./, n J~' ~1.2 Z
O
11
1.2.
O
I 1.t O
Fatigue Design Load Level I -l~ Failure. Residual L!:z78 " 1
10- "13
I Strength see Fig. 5 O
Fatigue Oesi_.qnLoad Level
1.0 . . . . . . . I
I I" --t
0.8- Life to be I m
Proven Life to be
I Proven
0.6 , , " 1. l ,. 0.9 r J ' L' "1"1
0.05 0.1 0.2 05 t0 20 0.1 0.2 0.5 1.0 1.5 0 2.5
Life Life
m
FIG. 6--Effect of environment on life (S-N curves).
---I
21J
C
O
-t
C
m
CO
CO
recovery took 2 to 4 h. The total testing time for one cycle block was about 30 h for the
Airbus fin box program with 1000 flights per cycle block and about 15 h for the Tornado
program with 200 flights per cycle block.
Results and Discussion

The residual strength and fatigue life results are shown in Figs. 5 and 6. Compensation
factors were derived from these test results, as shown in Table 2, employing the concept
proposed in Fig. 1.
All residual strength results including the mean values are presented in the form of scatter
bands; all life-to-fracture results are presented by their mean values. They were determined
by the maximum-likelihood method assuming normal distribution for the residual strength
and Weibull distribution for the life to fracture. By this estimation procedure, run-out data
as well as failure data (regarding life to fracture) were used in the statistical analysis [11].
For reasons of clarity, the illustrations do not show individual test results. As examples,
residual strength and life-to-fracture test results of the type 3a plain specimens are presented
in Tables 3 and 4, respectively, including results of the statistical analysis conducted?
Residual Strength
Residual strength tests were carried out after preloading at the fatigue design load level
(FDLL) to the life to be proven. The mechanical preloading as well as test temperature and
laminate moisture content at residual strength tests of the axially loaded coupon specimens
were the same as for the fatigue proof testing of the composite structures. To reduce duration
of preloading from 120 000 to 60 000 flights, the FDLL for the Airbus fin box was increased
from the very beginning by a factor of 1.15. Otherwise the FDLL was generally about 50%
of the ultimate load of the coupon specimens at room temperature in dry condition. Pre-
loading duration consequently was 60 000 flights for the coupon specimens of the Airbus
fin box program and 16 000 flights for the specimens of the Tornado program.
The ratios of residual strength after quasirealistic to the residual strength after simplistic
fatigue loading are compared for different types of specimens in Fig. 5, and the corresponding
compensation factors KR are summarized in Table 2. Except for specimen type 3a, the ratios
and compensation factors indicate insignificant influence of the environment simulated dur-
ing the preloading.
Furthermore, preloading at the FDLL to the life to be proven did not show an effect on
the residual strength at high temperatures. This is concluded from a comparison between
residual and ultimate strength shown in Fig. 7 [12]. Despite the preloading the strength was
not significantly different from that of virgin specimens when loaded at the same conditions.
Therefore, the reduction of the residual compression strength found for specimen type 3a
must be attributed to the effect of the moisture content. Based on this outcome, it may be
speculated that particular to the material and testing conditions of this investigation, static
testing alone would have led to the results received so far.
Life to Fracture
Present design load levels for composite airframe structures are far below ultimate loads
because of the detrimental effects of nondetectable impact damage, notches, and 90~
off-axis loads on strength of the presently applied carbon/epoxy systems. In view of com-
Additional tables with individual test results may be obtained directly from the authors.
m
~0
6")
m
>
TABLE 2-Derivation o f compensation factors. m
0
Z
Residual Strength after Life to Failure, "TI
Simplistic Realistic Simplistic Realistic Compensation
Loading Ratio of Loading Ratio of Slope of Factors
Specimen Load Level ~r,Slmp, ~r,Real, Residual Strength, Ns,.m NR~= Life to Failure, S-N Curve, a c
N/mm z N/mm 2 F,.S~p/F,.ae~l Number of Flights m
Types (see Figs. 5, 6) N slmp/ N Real S KR KL
m
Airbus Fin Box Program r-"
la 1.33 b 49888 34 740 1.44 0.081 1.030 c
lb 1.17b 51 523 18 540 2.78 0.041 1.043
2a 1.0 c 44.62 d 42.68 a 1.05 1.05
Z
2b 1.0c 44.13 d 44.91 a 0.98 1.0
Tornado Program 0
-rl
3a 1.6" 8 861 1 041 8.51 0.144 1.361 t")
3a 1.42" 20 320 2 230 9.11 0.144 1.375 O
3a 1.0e 1.41 E
-645.9 -458.1 1.41
3b 1.0" -697.1 -689.1 1.01 1.01 0
6o
4a 1.Of 298.4 311.9 0.96 1.0
4b 1.OJ 309.5 325.4 0.95 1.o m
"n
~ Slope of S - N curve for simplistic loading. "D
b F a t i g u e D e s i g n L o a d L e v e l S = ---320 N / m m L >
c F a t i g u e D e s i g n L o a d L e v e l "r = 20 N / m m L E
Residual interlaminar shear strength. 17"1
" F a t i g u e D e s i g n L o a d L e v e l S = - 4 0 0 N / m m z. 6o
-I
r F a t i g u e D e s i g n L o a d L e v e l S = + 190 N / m m 2.
C
O
---I
C
m
4~
TABLE 3--Example of residual strength test data and results of analysis: type 3a (plain specimen)
914C/T300, FALSTAFF-preloading (wing, upper surface), load level S = -400 N/mm 2,
16 000 flights.
Residual Strength ~ Test Data,

Simplistic Realistic
Loading
O~r,Stmp Or,Real
N/mm z Analysis Results
692 457 Mean residual strength, b N/mm2:

604 417 ~,.simp = 645.9, ~,.R,t = 458.1
618 526
Ratio of residual strength:
643 417
652 476 ffr.Slmp/(~r.Reat = 1.41
618 464 Residual strength at 10% and 90% failure probability, N/mmZ:
660 O'r.S,mp.l0 = 602.8, tr,.s,~,p.90= 689.0
684 cr,.a~.~0 = 401.4, cr,.ReaLg0= 514.8
~ Compressive loading at 120~

Based on normal distribution.
posite materials allowing higher design loads, plain coupon specimens were also loaded to
levels above the F D L L of the preloading to required life. A t the higher load levels the
testing ended with the specimen fracture. The results are plotted in the form of S - N curves
in Fig. 6. Their slope was used in the calculation of the compensation factors KL presented
in Table 2. The more complete simulation of the environment in the quasi-realistic loading
program resulted generally in earlier fracture than the less complete simulation of the
e n v i r o n m e n t in the simplistic loading. The corresponding life ratios are shown in Fig. 6 for
the different specimens used in the Tornado and Airbus fin box programs.
The life reduction factors in the Airbus fin box program were only 1.4 and 2.8 c o m p a r e d
to 8.5 and 9.1 in the T o r n a d o program. This is most likely because within the Airbus program
the e n v i r o n m e n t simulated in c o m p o n e n t proof t e s t i n g - - a n d thus also in simplistic l o a d i n g - -
was more complete than in the Tornado program. Whereas the Airbus fin box and all
pertinent coupon specimens were preconditioned to the required saturation level, in the
TABLE 4--Example of fatigue test data and results of analysis: type 3a (plain specimen), 914C/T300,
FALSTAFF-loading (wing, upper surface), load level S = -640 N/mm:.
Life to Fracture
Simplistic Realistic
Loading
Ns,mp NReal
Number of Flights Analysis Results
2 432 329 Mean fatigue life and standard deviation (lg N):
> 2 432 > 329 Ns,mp = 8861 flights, NR~ = 1041 flights
5 973 573
> 5 973 > 573 SS,mp = 0.239, SR~, = 0.354
10 432 58O Ratio of life to failure:
>10 432 > 580 Ns,mplNR~al = 8.51
10 760 1 899
>10 760 >1 899
Plain Specimens
Type 3a Type 3b
800 - 914C/T300 1 076E/T300
E 689.1 16)
Z 650 (2)
J
600 -
c- 458.1 (6) L65 (2)
400
-Su
Eo
u~ -~'~ .
L
o [~
c
o c r
200
.--: 9
0
-~1--
o ~ ~
(...) Number of S p e c i m e n s
FIG. 7--Comparison between the residual and the ultimate compressive strength.
Tornado program the simplistic loading was carried out with dry and the quasirealistic loading
with wet specimens as already pointed out in the section entitled "Loading Programs."
Conclusions
The "compensation factor concept," which was accepted by and large by the certification
agencies, was shown to be practicable. However, to substantiate the concept, additional
component testing with quasirealistic mechanical and environmental fatigue loading is
needed.
Testing of coupon specimens has shown that at present allowable fatigue design load
levels, which are at 50% of ultimate strain or less for reasons other than environmental
effects, specimens do not fail within the life to be proven with or without quasirealistic
simultaneous environmental loading. Furthermore, the residual strength under "worst con-
dition" (that is, hot/wet) determined after the fatigue loading did not reveal an influence
of the preloading with or without the simultaneous environmental history.
However, at higher load levels than presently necessary for the applied carbon/epoxy
systems, fatigue loading with simultaneous quasirealistic environmental loading reduced the
life to fracture compared to the loading totally without or with less complete simulated
environment. This points out that carbon/epoxy systems allowing higher design load levels
should be carefully evaluated for application in structures subjected to simultaneously oc-
curring mechanical and environmental fatigue loading. For this the environmental loading
standard ENSTAFF (ENvironmental FALSTAFF) was established and published by Fraun-
hofer-Institut for Betriebsfestigkeit (LBF), Darmstadt, Germany [13]and the other members
of the ENSTAFF working party, including the Royal Aircraft Establishment (RAE), Struc-
tures Department, UK; National Aerospace Laboratory NLR, Amsterdam, Netherlands;
Industrie-Anlagen-Betriebsgesellschaft mbH ( I A B G ) Abt. TFB, Ottobrunn, Germany;

Flugzeugwerke Emmen, (F + W) Switzerland.
Acknowledgment
The coupon specimen tests were conducted at Fraunhofer-Institut ftir Betriebsfestigkeit
under contract to the German Ministry of Defense (LBF), Darmstadt, Research Branch
RtiFo4 and to MBB-UT, Hamburg. The "Fatigue of Carbon/Epoxy Composites" working
party was entrusted with planning and monitoring the fighter program; the working party
was brought into being by the German Defense Ministry and consisted of the members of
Dornier, MBB-UF Munich, D F V L R e. V. Stuttgart, BWB-ML and LBF, Darmstadt.
References
[1] Shyprykevich, P. and Wolter, W., in Composites for Extreme Environments, ASTM STP 768, A.
Adsit, Ed., American Society for Testing and Materials, Philadelphia, 1982, pp. 118-134.
[2] Givler, R. C., Gillespie, J. W., Jr., and Pipes, R. B., in Composites for Extreme Environments,
ASTM STP 768, A. Adsit, Ed., American Society for Testing and Materials, Philadelphia, 1982,
pp. 137-147.
[3] Reifsnider, K. L. and Stinchcomb, W. W., in Composite Materials: Fatigue and Fracture, ASTM
STP 907, H. T. Hahn, Ed., American Society for Testing and Materials, Philadelphia, 1986, pp.
298-313.
[4] Advisory Circular, Composite Aircraft Structure, AC No: 20-107 A, U. S. Department of Trans-
portation, FAA, 1984.
[5] Military Standard, Aircraft Structural Integrity Program, Airplane Requirements, MIL-STD-
1530A (11), ASD/ENFS, Wright Patterson Airforce Base, OH, 1975.
[6] Konishi, D. Y. and Johnston, W. R., in Composite Materials: Testing and Design (Fifth Conference),
ASTM STP 674, S. Tsai, Ed., American Society for Testing and Materials, Philadelphia, 1979,
pp. 597-619.
[7] Haskins, J. E, Wilkins, D. J., and Stein, B. A., in Environmental Effects on Advanced Composite
Materials, ASTM STP 602, American Society for Testing and Materials, Philadelphia, 1976, pp.
23-36.
[8] Flugzeugwerke Emmen, (F + W) Switzerland; Laboratorium fuer Betriebsfestigkeit (LBF), Darm-
stadt, Germany; National Aerospace Laboratory (NLR), Amsterdam, Netherlands; Industrie-
Anlagen-Betriebsgesellschaft mbH (IABG), Ottobrunn, Germany, "FALSTAFF, Description of
a Fighter Aircraft Loading STAndard For Fatigue Evaluation," Laboratorium for Betriebsfestig-
keit, Darmstadt, Germany, March 1976.
[9] Schlitz, D., Gerharz, J. J., and Alschweig, E., in Fatigue of Fibrous Composite Materials, ASTM
STP 723, American Society for Testing and Materials, Philadelphia, 1981, pp. 31-47.
[10] Gerharz, J. J. and Schtitz, D., "Fatigue Strength of CFRP Under Combined Flight-by-Flight
Loading and Flight-by-FlightTemperature Changes," AGARD-CP-288, Advisory Group for Aero-
space Research & Development, 1980, pp. 6/1-6/24.
[11] Butler, J. P., in International Conference on Structural Safety and Reliability, A. M. Freudenthal,
Ed., Pergamon Press, Elmsford, NY, 1972, pp. 181-211.
[12] Berg, M., "KlimaeinfluB auf die Festigkeitseigenschaften von Kohlefaserverst~irktem Kunststoff
(CFK) im Flugzeugbau," VFI (Der Versuchs- und Forschungs-lngenieur), Vol. 19, No. 1, Feb.
1986, pp. 45-48.
[13] Gerharz, J. J., "Standardized Environmental Fatigue Sequence for the Evaluation of Composite
Components in Combat Aircraft (ENSTAFF = ENvironmental FalSTAFF)," LBF Report No.
FB-179, Fraunhofer-Institut for Betriebsfestigkeit (ISSN 0721-5320), Darmstadt, Germany. 1987.
A n o u s h Poursartip ~ a n d N a r i n e C h i n a t a m b i 1
Fatigue Damage Development in Notched

(02/-+ 45)s Laminates
REFERENCE: Poursartip, A. and Chinatambi, N., "Fatigue Damage Development in Notched
(02/-45), Laminates," Composite Materials: Fatigue and Fracture, Second Volume, ASTM
1989, pp. 45-65.
ABSTRACT: This study investigates delamination growth from holes in carbon fibre reinforced
epoxy (CFRE) AS/3501-6 (02/-+45), laminates under tensile static and fatigue loading. De-
laminations grow in a direction parallel to the loading at the 0/45 and 4 5 / - 4 5 interfaces.
Attempts to control the location of the delamination by using inserts were unsuccessful. The
delamination growth rates were fitted by a power function of the stress amplitude and mean
stress. The stress amplitude is the dominant parameter. Tests were performed on dry and
saturated specimens at room temperature, and there was no measurable difference in growth
rates. However, block loading tests, where an overload was applied at regular intervals, resulted
in growth rates five times higher than a linear sum of the individual growth rates. Compliance
changes were measured and correlated with the extent of delamination. The associated matrix
cracking contributes substantially to the compliance changes. The amount of matrix cracking
for a given delamination length is a function of the type and level of loading. The results are
interpreted in terms of the total strain energy release rate, G. Once the presence of matrix
cracking is allowed for, there is reasonable agreement with results in the literature.
KEY WORDS: fatigue, delamination, matrix cracking, graphite fiber, strain energy release
rate, overload, stiffness, moisture
Nomenclature
2a Tip-to-tip delamination length, excluding hole diameter, mm
A Curve-fit parameter for d(2a)/dN versus total G
A, Curve-fit parameter for d(2a)/dN versus G~
An Curve-fit parameter for d(2a)/dN versus G..
B Crack width, m
C Compliance, m N -~
E Modulus, G N m -2
E0 U n d a m a g e d modulus across a hole, G N m -2
En Laminate modulus in axial direction, G N m 2
Ec~ Stiffness as measured by clip-gauge across hole, G N m -2
G Total strain energy release rate, Jm -2
GI Mode I strain energy release rate, Jm -2
G. M o d e II strain energy release rate, Jm -2
G~,ax M a x i m u m total strain energy release rate in a fatigue cycle, J m -2
Gin., Minimum total strain energy release rate in a fatigue cycle, J m -2
' Assistant professor and senior research technician, respectively, Department of Metals and Materials
Engineering, The University of British Columbia, Vancouver, B.C., Canada.
45
Downloaded/printed
Copyright9 by
AG Gmax - Gmi., J m -2
Gc Critical total strain energy release rate, Jm -2
k Curve-fit parameter for E versus (2a) equation, mm-
Clip-gauge gauge length, m
n Curve-fit exponent for d(2a)/dN versus total G
nl Curve-fit exponent for d(2a)/dN versus G~
nH Curve-fit exponent for d(2a)/dN versus G ,
N Number of cycles
P Load, N
R ratio Load ratio = trm~,/Crm~
R Hole radius, m
t Laminate thickness, m
W Laminate width, m
(3" Gross-section stress, MNm -2
AGr Gross-section fatigue stress range, M N m - :
O'mean Gross-section mean stress, MNm -2
or; Gross-section far-field applied stress, MNm -2
(5 Strain, mm/mm
Ecg Strain measured by clip-gauge, mm/mm
Introduction
It is now common practice to analyze delamination propagation using the fracture me-
chanics concept of the strain energy release rate [1-8]. Most work has concentrated on
model geometries such as double-cantilever-beam (DCB), end-notched flexure (ENF), and
cracked-lap-shear (CLS), because they can be analyzed accurately, the delamination growth
is generally well behaved, and there are no other damage modes present.
In practice, delaminations will grow from internal flaws or from the free edges of holes
or other cutouts. The delaminations may cross several interfaces, and there may be associated
matrix cracking and other damage. It is more difficult to monitor and analyze such a system,
although there has been some success [1].
This paper is a further attempt to apply fracture mechanics to more complex geometries.
Results of an experimental program of tensile static and fatigue tests on notched (02/-+45),
carbon fiber reinforced epoxy (CFRE) laminates are quantified as a function of the applied
loads, and then analyzed in terms of the energy release rate. The resultant equations are
then compared with the data from model geometries found in the literature.
Experiment Details
The material that was tested is AS/3501-6 carbon fiber reinforced epoxy. The lay-up of
(02/+-45), was chosen so the only delamination growth would be from the hole, and there
should be no delamination growth from the free edges.
The specimen geometry and dimensions are shown in Fig. 1. Given that low-load fatigue
tests are very time consuming, a specimen with three holes was used. So long as delamination
growth is of interest, rather than failure, this specimen geometry works reasonably well.
The first batches of specimens were manufactured with 12.7-mm diameter Teflon inserts,
concentric with the 6.35-mm holes. Each hole had an insert placed at a different interface.
The aim was to initiate and propagate the delamination at the interface of choice. As will
be shown later, this was not successful, and the delaminations grew at the same interfaces
regardless of insert positioning. Therefore, later batches of specimens were manufactured
without inserts.
POURSARTIP AND CHINATAMBI ON FATIGUE DAMAGE DEVELOPMENT 47
40 mm j ~ 220 mm
26 mm
insert
hole - 6.35 mm diameter
insert - 12.7 mm diameter

FIG. l--Specimen geometry and dimensions.
The majority of the specimens were kept in a dessicator after manufacturing, until they
were tested at ambient room temperature and humidity. However, one batch of specimens
was soaked in water at 75~ for approximately 2 months, at which stage they were saturated.
They were then stored in water at ambient temperature until needed. The fatigue tests were
conducted with the specimens fully immersed in water, at ambient temperature, using a
simple specially designed holder. In subsequent sections, these tests will be referred to as
"wet."
Most tests were run using a constant amplitude sine wave at a frequency of 10 Hz.
However, a waveform consisting of two different sine waves was used to determine the load
interaction effects under variable amplitude loading. This waveform (called "block loading"
hereafter) consisted of a defined number of cycles of a lower amplitude sine wave, punctuated
by a single overload (Fig. 2). The frequency of the overload was lower, such that the rate
of loading in both low- and high-amplitude cycles was the same.
Stiffness was measured by interrupting the test at predetermined intervals and ramping
the load between previously defined limits, while monitoring both load and clip-gauge
outputs for all three holes.
Delamination lengths were determined directly using dye-penetrant enhanced x-ray ra-
diography. The specimens were removed from the grips, the hole boundaries were infiltrated
with zinc iodide dye penetrant under load (typically 2 kN), and then the specimens were x-
rayed using a 30-kV beam at 10 mA for 60 seconds, onto Polaroid Type 55 positive-negative
film. All delamination lengths in this paper are direct measurements from x-ray photographs.
Results and DiscussionmGrowth Characterization

Static Baseline Properties
Axial and transverse strain gauges were bonded onto a specimen halfway between two
holes. The axial modulus, Eta, was measured as 76.8 GPa, and the major Poisson's ratio,
repeating block
FIG. 2--Load-time trace for a typical block-loading sequence.
v~:, was measured as 0.434. Predicted values using standard AS/3501-6 unidirectional prop-
erties as input to a laminated plate theory (LPT) analysis are 73.5 GPa and 0.694, respec-
tively. The discrepancy between observed and expected values for v~2 may be due to the
use of transformed properties for the +-45 sublaminate, which leads to a rather high laminate
Poisson's ratio.
The average undamaged stiffness measured across the holes using 25.4-mm gauge-length
clip-gauges is 50.1 GPa. The stiffness across the hole, Ecg is defined as follows (Fig. 3):
Ecg _ ~ tr;l~ = tr;l~

~ - ~c~ (~,o,o + ~ ....... ,) (1)
where the strain is the sum of the hole extension and the material extension divided by the
clip-gauge gauge length.
Position and Direction of Delamination Growth

Fatigue Loading--Delaminations initiate and propagate around the holes in a direction
parallel to the applied load.
A typical sequence of radiographs is shown in Fig. 4. In this example, the outline of the
insert at the hole can be seen, as well as a significant amount of matrix cracking in the +-45~
plies.
The mechanism of delamination formation consists of the growth of splits in the surface
0~ plies, at the edges of the hole. The splits are then connected by a delamination front,
forming a tongue of material which can be observed to lift off the central layer under an
applied load. The delamination front is not uniform, and therefore the values reported are
always the average value for the tip-to-tip distance between the fronts, less the hole diameter.
Static Loading--A similar damage pattern occurs under static loading, except that the
maximum amount of delamination is much smaller. This is due to the fact that the delam-
inations initiate at a load which is just lower than the static strength of the specimens. The
delaminations are stable under constant load and require an increasing load for propagation.
ITTTTTI
material
clip-gauge hole
length
material
1111111
FIG. 3--Schematic of clip-gauge gauge length.
FIG. 4 - - X - r a y radiographs showing extent o f delamination and matrix cracking under fatigue
loading at A~r = 22l MPa, R = 0.12 at (a) 75 000 cycles, (b) 150 000 cycles, (c) 200 000
cycles, (d) 275 000 cycles, (e) 350 000 cycles, (f) 425 000 cycles.
FIG. 5--Longitudinal cross-section of a fatigued specimen hole showing insert at midplane,

delaminations at the O~ + 45 interfaces, and extensive matrix cracking.
35 CFRP
(02/.45) s
F: 30
At7=340 MPa, R=0.15
dry, RT J
9 linear best fit
E bottomhole
25 9~ middle hole
O,I 9 top hole
.t::"
r 20 f
t-
./ /
r- 15
O
. m
t-
"~ 10
/,
Y
~ 5
0 5000 10000 15000 20000

cycles
FIG. 6--Typical plot of delamination length (2a) as a function of cycles N. Both axes are
linear.
Effect oflnserts--The aim of implanting inserts was to encourage the preferential initiation
and propagation of delaminations. This was unsuccessful, as can be seen in the axial cross-
section in Fig. 5. Although delamination initiation might be influenced by the presence of
the insert, propagation occurs at the interface with the highest interlaminar stresses [I]. In
Fig. 5, the insert is at the midplane and yet the dominant delaminations, which extend well
beyond the hole, are at the 0/45 interfaces. A delamination down the midplane is arrested
a short distance from the insert. With hindsight, this might have been expected because the
influence of the insert can only be short range, and beyond a certain distance the laminate
alone will determine the stress field. Note also the extensive cracking in the -+45~ plies,
which is a characteristic feature of the fatigue specimens. (This can also be seen in Fig. 4.)
As a result of the above observations, later batches of specimens were manufactured
without inserts, and the behavior of the delamination initiation and propagation was observed
to be the same as with the inserts, at least to the level of resolution of this work.
Crack Growth Rate as a Function o f Stress Range

All the tests were run under load control. A typical delamination length versus cycles
plot is shown in Fig. 6. Although there is some deviation, a straight line relationship fits
the results reasonably well for all the tests conducted in this work. For each hole, a straight
line was fitted to the results, and the slope of the line, d(2a)/dN, was recorded.
The logarithm of the values of d(2a)/dN are plotted as a function of the logarithm of the
applied stress range, A~ (which is constant throughout a given test), in Fig. 7, for all tests
where the R ratio ( = Crm,,/trm~x)is less than 0.2. The straight line superposed on the results
in Fig. 7 is a line of best fit to all results where R = 0. l. Thus,
d(2a)dN - 2 X 10-32(atl)1158 (2)
describes the results at constant R ~ 0.1 with a coefficient of correlation of 0.95.
Crack Growth Rate as a Function o f Mean Stress

The effect of the mean stress is quantified in a similar fashion. The mean stress, cr..... is
( l + R ) a~r (3)
cr. . . . - 2(1 - R )
For a fatigue test at an R ratio other than 0.1, the stress range effect is already accounted
for by Eq 2, and so a plot of ( d ( 2 a ) / d N ) / ( A x Air") as a function of (1 + R)/(1 - R)
will show the effect of the mean stress on the delamination growth rate, independent of the
magnitude of the stress range (Fig. 8). A line of best fit to this result leads to
d(2a)/dN (1 + R~ 3"-
A(Acr)" - 0.425 \1 - R / (4)
with a coefficient of correlation of 0.84, where A and n are defined in Eq 2. Combining Eq
-2 CFRP
//~ (02/94s)s
R <0.2
dry, RT
tD =
-~ -3
g
Z -4
8
"0
o -5 |
-6
200 300 400 500
stress range (MPa), log scale
FIG. 7--Delamination growth rate d(2a)/dN as a function of the stress range Air, for R <--
0.2. Both axes are logarithmic. Data are from dry specimens at room temperature. Each point
represents a full fatigue test.
2 CFRP
(02/~ 45)s
R~0.5
dry, RT
I /
t---
la
Y
/
Y,
Z |
t~
t'Xl /I
v
v
0
/.i
O I.
-1
1 1.5 2 2.5 3
(I +R)/(1-R), log scale
FIG. 8--Component of delamination growth rate due to mean stress, ( d(2a) / dN) / ( A( A(r )")
as a function of (1 + R)/(I - R), which is a measure of the mean stress. Both axes are
logarithmic. Data are from dry specimens at room temperature. Each point represents a full
fatigue test.
2 and Eq 4, we have, for the general case of any A~r and R,
d(Za) 1 + R] 38z (A~r)tl 58

aN = 8.5 • 10_33 \a- - - S ~ / (5)
Crack Growth Rate f o r Saturated Specimens

Similar equations can be derived for the tests on the wet specimens. The results for the
delamination growth rate as a function of A(r at R -< 0.2 is shown in Fig. 9. Note that Figs.
7 and 9 have the same scales, and that the results from dry and wet specimens overlap,
indicating little or no change in behavior as a result of saturating the specimens and testing
at room temperature. A line of best fit to the results in Fig. 9 leads to
d(2a) = 1.13 • 10-Zg(Atr) '~ (6)

dN
with a coefficient of correlation of 0.93. Note that the difference between Eq 2 and Eq 6
over the range of interest is small. The effect of the mean stress is shown in Fig. 10. The
line of best fit is now
d(2a)/dN (1 + R ] 5"38
A (A(r)" 0.243 \1- - - - - ~ ] (7)
-2 CFRP
(02/,45)s
m R ':0.2
wet, RT
II
m
/
-3
E
E
Z -4
,/ |
-5
-6
200 300 400 500
FIG. 9--Delamination growth rate d(2a)/dN as a function of the stress range Act. Both axes
are logarithmic. Data are from wet specimens at room temperature. Each point represents a
futt fatigue test.
with a coefficient of correlation of 0.95, and where A and n are taken from Eq 6. Combining
Eq 6 and Eq 7, we have, for the general case of any Atr and R in the wet condition,
38
d(2a) 10_30 (1 + R ] ' (A(~),o.,3 (8)
d----~ = 2.75 x \1 - R /
Crack Growth Rate for Block Loading

The block loading components were chosen such that the expected contribution from the
more numerous low-amplitude cycles would be negligible, and that the growth rate per
repeating block, given no load interaction effects, would be equal to the growth rate of the
one high-amplitude cycle in each block.
A number of different combinations of amplitudes and ratios of number of cycles were
tested, as shown in Table 1.
In general, if two components of a waveform, which would separately have growth rates
(d(2a)/dN)~ and (d(2a)/dN)2, were mixed in the ratio of N, and N2 cycles, respectively,
then we would expect an average growth rate of
(d(2a)~ NI (d(2a)~ N~ (d(2a)~

- x + ~ x (9)
\ d N / ..... ,~ N, + N~ \-'~], N, + N: \ d N 12
Such a prediction is a linear sum of the contribution of the two components and assumes
that there is no interaction between the two components.
CFRP
(02/, 45)s
R <0.5
/
wet, RT
t-
" b' " 1
<
Z
|
oJ
v 0
"0
-1
1 1.5 2 2.5 3
(I +R)/(1-R), log scale
FIG. lO--Component o f delamination growth rate due to mean stress, (d(2a)/dN)/( A ( Atr)")
as a function o f (1 + R)/(I - R), which is a measure o f the mean stress. Both axes are
logarithmic. Data are from wet specimens at room temperature. Each point represents a full
faligue tesl.
In Fig. 11, the expected delamination growth rate, using Eq 5 for the constant amplitude
dry specimens, Eq 5 and Eq 9 for the block loading dry specimens, and Eq 8 for the constant
amplitude wet specimens, is compared with the observed delamination growth rates. Al-
though the constant amplitude results show good agreement between the predictions of Eq
5 and Eq 8 and the observed behavior, the block loading results are such that the observed
growth rates are consistently higher than what would be expected from Eq 5 and Eq 9.
Given that the low-amplitude cycles would have negligible growth rates on their own (Eq
5), it appears that there is a load interaction effect that leads to the higher observed growth
rates.
This acceleration effect, defined as the observed growth rate divided by predicted growth
rate assuming no interaction effects (Eq 9), can be seen better in Fig. 12--a bar chart
TABLE 1--Block loading parameters.
AO'h,gh R ratioh,~h Atrlo~ R ratiolo. NhJgh Nh,,.

246 0.42 63 0.74 1 30
288 0.15 127 0.29 1 15
232 0.46 52 0.79 1 15
285 0.17 187 0.24 1 10
286 0.16 97 0.35 1 15
277 0.18 90 0.40 1 30
56 S E C O N D C O M P O S I T E MATERIALS
-2 CFRP
t- (02/~'45)s
O
,D R <0.5
.s 9 dry
"o
v wet
/
o. -3 9 block -O ~ 00
Q)
._.R
lib
-~
>, -4
E
E 9 VV
Z
v v
A -5
(0
"o
-6
-6 -5 -z -3 -2
log d(2a)/dN (mm/cycle), observed
FIG. 11--Comparison of observed delamination growth rates with predicted growth rates,
not allowing for load interaction.
12
r CFRP
(02/, 45)s
.s 11 R<0.5
._o
"o 10
block dry wet
~ 9 ..91- -ID..- -,91- --lb.-
Z
7
~"o 6
-~ 5
i11
P 4
o a
Z
N 2
"o
0
FIG. 12--Bar chart representation of the acceleration effect for block loading. Constant
amplitude dry and wet results are shown as a comparison.
representation of the results in Fig. 11, with the ordinate showing the magnitude of the
acceleration effect. The constant amplitude dry and wet specimen results are shown as a
reference and are equally distributed around a value of 1 (the average value of the accel-
eration effect is 1.17 for the dry specimens and 1.13 for the wet specimens). The block
loading tests consistently show an acceleration effect, with an average value of 5.49, and a
maximum of 12. Thus the use of a linear summation of the individual components of a
variable amplitude waveform could lead to a significant underestimate of the average growth
rate.
Stiffness Reduction as a Function of Delamination Size and Loading

Expected Form of Relationship--The current damage geometry is very similar to the
cracked-lap-shear specimen geometry used by other workers (see, for example, Ref 2).
There are two simple relationships to choose from: a linear reduction in stiffness with
delamination length, or an inverse relationship between stiffness and delamination length.
The length of the specimen covered by the clip-gauge is modeled in two parts: a central
section of length (2a) (section I), and two outer sections of total length 1c8 - (2a) (section
II).
In section I, it is assumed that the delaminated area carries no load. Thus the stiffness
El
In section II, the original properties are assumed. Thus, by analogy with Eq 1, the system
acts as two springs in series, and
E 1
Eu 2R (11)
1 + (2a)
(W - 2R)lcg
In deriving Eq 11, it is assumed that the delaminated area carries no load, and that the
undelaminated area sees a uniform stress not affected by the hole and delamination. This
is obviously not correct, and the virtue of Eq 11 is to suggest the form of the relationship,
rather than the exact value of the constant. The following relationship would be expected:
E 1
(12)
E00 = 1 + k x (2a)
where the constant k is determined experimentally.
Experimental Results
Tests were conducted in which the stiffness was measured continuously as a function of
loading, and the specimen was x-rayed at regular intervals. The results were then cross-
plotted as the normalized stiffness E/Eo against the delamination length, 2a (Fig. 13). Results
are presented for delaminations grown under fatigue loading (Act = 310 MPa, R = 0.1)
and quasistatic loading. Also shown are lines of best fit to the results, where the fitted
parameter is the constant k in Eq 12.
1.1 CFRP
---. 1
\ 102/.4s)s
dry, RT
* static
9 fatigue
o
LU (310 MPa, R=0.1)
t.U
v mUm
0.9
t- 9
,~ 0.8
"o
.N
-~ 0.7 mm~
E
0
Z 0.6
0.5
0 4 8 12 16 20 24 28
delamination length (2a), mm
FIG. 13--Stiffness reduction as a function of delamination length under static and fatigue
loading. The smooth lines are lines of best fit for an inverse stiffness-delamination relation.
There is a greater stiffness reduction for a given delamination length under fatigue loading
than under static loading. Although a simplistic analysis such as Eq 12 does not allow for
this effect, careful examination of x-ray radiographs of delaminations grown under fatigue
and static loading, respectively, provides the explanation. Figure 14a shows a typical de-
lamination that has grown under static loading; Fig. 14b shows a delamination of similar
length grown very slowly under fatigue loading (Act = 205 MPa, R = 0.1, for two million
cycles). Although the delaminations are of similar length, there is considerably more matrix
cracking in the +45 and - 4 5 plies in the fatigue case. These matrix cracks extend well
beyond the delamination limits, and given that the 45 plies make an appreciable contribution
to the stiffness of the laminate, it is not surprising that the fatigue results are very different
from the static results. As discussed in the next section, a higher applied fatigue stress will
lead to a smaller stiffness reduction for a given delamination length.
It is notable that the fatigue results in Fig. 13 would be fitted much better by an equation
of a different form than Eq 12. However, a primary reason for quantifying the stiffness
reduction results is to apply fracture mechanics concepts. In that case, the form of the
equation matters because it is used in the derivation of the strain energy release rate.
The predicted stiffness reductions according to the line of best fit in Fig. 13 are compared
with the observed values in Fig. 15.
Figure 13 emphasizes the danger in assuming that a given stiffness reduction corresponds
to a given damage state. Although that may be true for some systems, in general it is possible
to create the same stiffness reduction with different mixes of damage.
Stiffness was monitored for a number of fatigue tests at different loads, but these tests
were not interrupted to take radiographs. As a result, a direct comparison as in Fig. L3
cannot be made, but an indirect comparison is possible. A plot of the inverse of the nor-
malized stiffness (Eo/E) as a function of cycles (see, for example, Fig. 16) reveals a linear
FIG. 14--Relative amounts of matrix cracking for a given delamination length (a) static
loading and (b) fatigue loading at ~ r = 205 MPa, R = 0.1 for 2 000 000 cycles.
1.1 CFRP
(02/'45)S
fatigue
310 MPa, R=0.1
dPJ, RT
,.=F"_,_ ; V " "

~" m m
0.7
0.5
0.5 0.7 0.9 .1
(E/Eo) observed
FIG. 15--Comparison of observed normalized stiffness and predicted normalized stiffness,
using an inverse relation between delamination and stiffness.
CFRP
(02/=~45)s
/'r =360 MPa, R=0.1
dry, RT
1.8 observed
linear best fit
LU
"8
1.6 J
LU
v
O
t-
t~ 1.4
O.
E
O
O 1.2
0 500 1000 1500 2000 2500
cycles
FIG. 16--Typical plot of increase in normalized compliance with cycles. Both axes are
linear.
relation, albeit with scatter. This linear relation is to be expected from Eq 12, given that
d(2a)/dN is constant during a constant load test. The slopes of the (Eo/E) versus cycles
curves are plotted as a function of the stress range Atr in Fig. 17, in similar fashion to Fig.
7. A best fit line to the results leads to
d(Eo/E)
dN - 4.35 x 10 29(Act)946 (13)
with a coefficient of correlation of 0.90. Combining Eq 2 and Eq 13,
d(Eo/E)
d(2a) - 2.17 • 103(Act) -'-'2 (14)
Integration of Eq 14, noting that E = E0 when 2a = 0, will lead to Eq 12 where
k = 2.17 x 103(Act) -212 (15)
The results of the direct and indirect determination of k can be compared in Fig. 18.
There is qualitative agreement, showing that as the applied stress range increases in mag-
nitude, k decreases. The curve in Fig. 18 must be bounded at the lower end by a value
corresponding to the threshold stress below which matrix cracks do not initiate or propagate,
and at the upper end by the failure strength of the specimens. Equation 15 has only been
verified for R -< 0.2, and although a similar equation might be expected for the general
loading case, this is not available.
Figure 18 shows that, under fatigue loading, a substantial amount of stiffness reduction
is due to the secondary damage mechanism of matrix cracking, rather than the delamination
damage which is of primary interest.
-3 CFRP
., (o2/*4S)s
R <0.2
= dry, RT
-4
(1)
to
r
~tl:l
r
/
I
IE -5
O
o
-6 / I
LU
O
/,
LU
v
"O -7
C~
O
-8
200 300 400 500
FIG. 17--Rate o f increase in compliance as a function o f the stress range, Air, for R <- 0.2.
Both axes are logarithmic. Each point represents a full fatigue test.
0.04 l CFRP
(02/+45) s
R <0.2
dry, RT
0.03
\ from (E/Eo) versus (2a) curves
E
'\ 9
J\
E
'T"
0.02
-"k,
/#\. '\
,,,.,,,..,..
0.01 "....
from d(E0/E)/dN "...,..
and d(2a)/dNcurves
0 200 400 600

stress range (MPa)
FIG. 1B--Variation in the curve-fit parameter k in the stiffness-delamination relation (E/Eo)
= 1/(1 + k x (2a)) as a function o f the stress range Atr.
Analysis of Results Using Fracture Mechanics

The compliance derivation for the strain energy release rate G is [3]
P2dC
G- (16)
2Bda
where for our specimen geometry
C = tJ(WtE)
P = trWt
B = 2R
da --~ d(2a)
Substituting for E from Eq 12, we have
G=\4R/ (17)
Thus given an inverse relation for E, such as Eq 12, then G is proportional to cr2, while a
linear reduction in E with 2a would have led to G being proportional to c 2.
Equation 17 is further justification for fitting Eq 12 to the results of Fig. 13. In this work,
all the results confirm that d(2a)/dN is constant during a constant stress amplitude test.
However, other researchers, using other geometries, have shown that delamination growth
correlates well with G. If Eq 12 is held to apply, then there is no inconsistency.
Static Results--Stable Growth

Classic linear elastic fracture mechanics states that once G -> Gc a crack will grow spon-
taneously, provided that the propagation of the crack does not lead to a decrease in the
energy available to the system. According to Eq 17, G is independent of the crack length
for the current geometry. Thus it would be expected that once the delaminations have
initiated at a given static load, they would propagate catastrophically. This is not the case,
and the delaminations are stable and require an increasing load to grow under static loading.
A value for Gc can be estimated by inserting the observed static stress required to initiate
delaminations into Eq 17. The value for k is 0.0027 r a m - ' , as measured for the static tests.
The value estimated in this fashion is Gc = 510 J/m 2. This compares favorably with values
in the literature, although it is difficult to make a direct comparison because the delamination
front is seeing a mixed mode loading. Russell and Street [4] report Gtw = 460 J/m-' from
ENF tests where G,/G = 1, G~c = 110 J/m E for DCB tests where Gtl/G = 0. Although
the GII/G ratio is not known for the current geometry, CLS geometries in the literature are
predominantly mode II dominated.
Derivation of Fatigue Growth Equation in Terms of G

Fatigue data from model geometries such as DCB, ENF, and CLS are normally reported
in terms of the strain energy release rate range, AG. Although sometimes AG is expressed
as being proportional to Aorz, strictly AG = Gmax - G .... and thus AG is proportional to
(1 + R/1 - R)(A~)2. Although it is possible to combine Eq 15 and Eq 17 to evaluate AG,
Eq 15 is not accurate. Therefore, Eq 17 is quantified for the two loading conditions where
k has been measured directly, k = 0.0234 mm -~ under fatigue loading, at A~r = 283 MPa,
R = 0.1, while k = 0.0027 mm -I under static loading, cr = 400 MPa.
Thus under static loading G = Gc = 510 J/m 2, while under fatigue loading at 283 MPa,
AG = 2670 J / m z. It appears that the energy release rate under fatigue loading is higher
than Gc, and yet the delamination growth rate is lower. The explanation must lie in the
fact that the strain energy release rate is calculated by the compliance method. With the
compliance method, the energy supplied to the system is considered. As there are two
damage mechanisms capable of absorbing energy, it follows that the relative amounts of
energy absorbed vary as the ratio of the volumes of damage changes. Given the evidence
of Fig. 14, at lower stress levels the increasing relative amount of matrix cracking absorbs
a larger share of the energy supplied to the system.
This implies that as the applied stress decreases, more of the energy supplied to the system
is diverted from propagating the delamination front into creating matrix cracks. The lower
the applied stress levels, the more inefficient the process of delamination growth.
It is possible to evaluate the delamination growth rate in isolation from the matrix cracking
by considering only the energy absorbed by the delamination process. To a first approxi-
mation, it is assumed that all the stiffness changes under static loading are due to delamination
growth. Then k = ks,.... = 0.0027 mm -I quantifies the total contribution of delamination
on compliance changes. Thus, for all loading conditions
(Wtl~gt(ks""ct(1 + Rt (A~I)2 (18)

AG = \ 4 R / \ E0 / \ 1 - R/
Substituting Eq 18 into Eq 5 and Eq 8, respectively, leads to
d(2a) - 2.6 x 10 -'s (1 + RI-"97 (AG) 5-79 (19)

dN \1 - R/
for the dry specimens, and
d (2a) 10 -26 1 + R'~ ~ (A G) 527 (20)

dN = 6.4 x \]--S"-R]
for the wet specimens. Note that in allowing for mean stress effects in AG a large part of
the apparent mean stress effect in Eq 5 and Eq 8 disappears, especially for the wet specimens.
A possible contradictory result is that under static loading there is an increasing resistance
to crack growth, which is not reflected in the fatigue results. Previous work on similar edge
delamination growth with associated matrix cracking showed a decreasing delamination
growth rate under constant AG [5]. This was attributed to the increasing resistance to crack
growth. In the present case, this refinement is not necessary.
Comparison with Other Data

To compare Eq 19 and Eq 20 with other data requires a knowledge of the mixed mode
conditions seen by the delaminations in the present geometry, which is currently not avail-
able. Even a numerical analysis will be approximate, because the delamination front is badly
behaved and shifts between interfaces, and the geometry is truly three dimensional.
We may make a simplistic comparison with other data by making two approximations.
First, most CLS geometries reported in the literature have G,/G = 0.65 ---, 0.8 [2,4,6].
Second, an estimate of the growth rate under mixed mode conditions as a function of pure
mode conditions is needed. The simplest condition [2] would be
(-EN/m,xo.
d(2a)~ = AI(AGI)"~ + An(AG,,)"; (21)
The other possible simple relation, that the mixed mode delamination growth rate is a
function of the total energy release rate, implies that nt = nu and A~ = A~t. This is not the
case for the data of Russell and Street for AS/3501-6 material [7,8]. Substituting for the
constants from their data,
d(2a)]
- " d N ] ~ , ~ = 1.1 x 10-25(AG) 94 + 1.8 x 10-'8(AG) 58 for G,/G = 0.65
(22)
= 5.6 x 10-zS(AG)94 + 6.0 • 10 18(AG)58 for G,/G = 0.8
For GJG = 0.65 the mode I contribution to the growth rate will exceed the mode II
contribution for G -> 110 J / m 2, which corresponds to Crm,~--> 190 MPa. For G./G = 0.8 the
mode I contribution to the growth rate will exceed the mode II contribution for G - 610
J/m 2, which exceeds Gc and corresponds to crmax-> 440 MPa.
Comparison of Eq 22 with Eq 19 shows that the exponent in Eq 19 agrees remarkably
with the exponent n, from the data of Russell and Street. The constant in Eq 19, evaluated
for R -- 0, is also in reasonable agreement with the mode II constants in Eq 22. Although
the results of Eq 20 are not as easy to compare, note that Eq 19 would have also been a
reasonable fit to the wet results. Furthermore, the exponent in Eq 20 is much closer to n.
than to n~.
Given that Eq 22 suggests that mode I growth would dominate in the region of interest
for G./G = 0.65, and given that the present results agree better with the mode II values,
it appears that the current geometry must have a mixed mode ratio closer to 0.8 than to
0.65.
The above comparison has a number of uncertainties associated with it. However, the
discussion serves to show that although realistic geometries may appear to behave in a
unpredictable fashion, it is possible, by use of suitable approximations and empirical cor-
relations, to correlate the results with fracture mechanics results from well understood
geometries.
Conclusions
1. Delaminations from holes in an A5/3501-6 (0.,/-+45), laminate initiate and propagate
in a direction parallel to the loading direction. The delaminated interfaces are primarily the
0/45 interfaces, as well as the 4 5 / - 45 interfaces.
2. It was not possible to initiate and propagate delaminations at a desired interface by
means of placing inserts. Delaminations shifted to the interfaces of their choice within a
short distance.
3. The delamination growth rate, d(2a)/dN is constant during a constant load amplitude
test. The growth rate correlates well with a power function of the stress amplitude and the
mean stress. The stress amplitude is the dominant parameter.
4. Results from tests on dry and water-saturated specimens at room temperature overlap,
and there is no evidence of accelerated growth.
5. Results from block loading tests, where low-amplitude cycles are punctuated with a
large-amplitude cycle show an acceleration in measured growth rate over and above what
would be expected from a linear summation of the respective components. On average, the
observed growth rate was 5.5 times faster than the linear prediction.
6. There is a stiffness reduction associated with delamination growth. There is a further

stiffness reduction as a result of matrix cracking in the 45 plies. The amount of matrix
cracking for a given delamination length is a function of the level of loading. The lower the
fatigue loads the larger the relative amount of matrix cracking.
7. The present results can be interpreted in terms of the total energy release rate, G, by
using the compliance formulation. Care must be taken to allow for the energy absorbed by
the matrix cracking. There is reasonable agreement for both static and fatigue loading with
results from other geometries for the same material system.
Acknowledgments
This work was carried out under a contract with the Canadian Department of National
Defence, Defence Research Establishment Pacific. The authors are grateful for many useful
discussions with K. N. Street and A. J. Russell. The technical help of R. Bennett in testing,
and E. Jensen (DREP) in making the specimens is much appreciated.
References
[1] Mohlin, T., Blom, A. E, Carlsson, L. A., Gustavsson, A. I., "Delamination Growth in Notched
Graphite/Epoxy Laminate Under Compression Fatigue Loading," in Delamination and Debonding
of Materials, ASTM STP 876, W. S. Johnson, Ed., American Society for Testing and Materials,
Philadelphia, 1985, pp. 168-188.
[2] Ramkumar, R. L. and Whitcomb, J. D., "Characterization of Mode I and Mixed-Mode Delami-
nation Growth in T300/5208 Graphite/Epoxy," in Delamination and Debonding of Materials, ASTM
STP 876. W. S. Johnson, Ed., American Society for Testing and Materials, Philadelphia, 1985, pp.
315-335.
[3] Broek, D., Elementary Engineering Fracture Mechanics, 3rd rev. ed., Martinus Nijhoff, The Hague,
1982, Chaps. 5 and 10.
[4] Russell, A. J. and Street, K. N., "Moisture and Temperature Effects On The Mixed-Mode De-
lamination Fracture of Unidirectional Graphite/Epoxy," in Delamination and Debonding of Ma-
terials, ASTM STP 876, W. S. Johnson, Ed., American Society for Testing and Materials, Phila-
delphia, 1985, pp. 349-370.
[5] Poursartip, A., "The Characterization of Edge Delamination Growth in Laminates Under Tensile
Loading," in Toughened Composites, ASTM STP 937, N. J. Johnston, Ed., American Society for
[6] Armanios, E. A., Rehfield, L. W., and Reddy, A. D. "Design Analysis and Testing for Mixed-
Mode and Mode II lnterlaminar Fracture of Composites," in Composite Materials: Testing and
Design (Seventh Conference), ASTM STP 893, J. M. Whitney, Ed., American Society for Testing
and Materials, Philadelphia, 1986, pp. 232-255.
[71 Russell, A. J. and Street, K. N., "The Effect of Matrix Toughness on Delamination: Static and
Fatigue Fracture Under Mode II Shear Loading of Graphite Fiber Composites," in Toughened
Composites, ASTM STP 937, N. J. Johnston, Ed., American Society for Testing and Materials,
[8] Russell, A. J. and Street, K. N., "A Constant AG Test for Measuring Mode I lnterlaminar Fatigue
Crack Growth Rates," presented at the Eighth ASTM Symposium on Composite Materials: Testing
and Design, Charleston, SC, 19 April-I May 1986. in Composite Materials: Testing and Design
(18th Conference), ASTM, Philadelphia, 1988.
C. E. Bakis, x H. R. Yih, 1 W, W. Stinchcomb, 1
and K. L. Reifsnider ~
Damage Initiation and Growth in Notched

Laminates Under Reversed Cyclic Loading
REFERENCE: Bakis, C. E., Yih, H. R., Stinchcomb, W. W., and Reifsnider, K. L.. "Damage
Initiation and Growth in Notched Laminates Under Reversed Cyclic Loading," Composite
Materials: Fatigue and Fracture, Second Volume, ASTM STP 1012, Paul A. Lagace, Ed.,
American Society for Testing and Materials, Philadelphia, 1989. pp. 66-83.
ABSTRACT: The initiation and growth of fatigue damage in notched Cycom AS4/1808 graph-
ite/epoxy laminates were investigated experimentally. Two layups, (0/45/90/- 45),, and (0/
45/0/-45),4, provided contrasting responses to the fully reversed fatigue loads. Full-field
strains and stresses surrounding the circular notch were evaluated at several stages of fatigue
life with the photoelastic coating and adiabatic thermal emission measurement techniques and
were related to the corresponding damage state and residual strength, stiffness, and life.
KEY WORDS: composite material, graphite/epoxy, notched laminate, fatigue, damage, stiff-
ness, strength, strain redistribution, nondestructive evaluation, photoelasticity, thermography
Because fiber-reinforced composite materials with cutouts are being used more frequently
as primary load-bearing structural components under long-term cyclic loading, there is a
need for an understanding of fatigue damage development within stress gradients and its
effect on the residual strength, stiffness, and life of the components. Recent experimental
and analytical advances explicate the significant effects of localized subcritical damage on
load (or strain or stress) redistribution in unnotched laminates which, in turn, influences
the global fracture event [1,2]. The current understanding of load redistribution in notched
laminates is not as complete as that for unnotched laminates, due primarily to the additional
complexity of in-plane damage and stress gradients. Attempts to characterize strain and
displacement fields near a notch in the presence of damage have successfully used photo-
elasticity and moir6 interferometry (see, for example, Ref 3 for monotonic loads, and Ref
4 for tensile cyclic loads). Recent analytical efforts have used two- or quasi-three-dimensional
finite-element models to predict incremental (element-wise) damage growth near notches
during monotonic loads [5,6], but a unified treatment of the problem that accurately models
the changes in stress distribution and corresponding changes in residual properties during
fatigue damage development is still not available. The present paper addresses the problem
through an experimental investigation of the interaction of fatigue damage, the global state
of deformation, and the residual strength, stiffness, and life in two contrasting center-notched
graphite/epoxy laminates under fully reversed cyclic loading. Two methods are employed
to characterize full-field, in-plane deformations at intermittent points in the fatigue lifetime.
The first method involves measurement of surface strains with a photoelastic coating applied
to the specimen. The second method employs the measurement of the dynamic, adiabatic
Research associate, research associate, professor, and Reynolds Metals professor, respectively,
Materials Response Group, Engineering Science and Mechanics, Virginia Polytechnic Institute and
State University, Blacksburg, VA 24061-0219.
66
Downloaded/printed
BAKIS ET AL. ON DAMAGE INITIATION AND GROWTH IN NOTCHED LAMINATES 67
temperature change during cyclic loading of the specimen, which is related to the dilatation
of the microconstituents.
Mechanical Testing
The (0/45/90/-45),4 and (0/45/0/-45)s 4 graphite/epoxy laminates were fabricated in an
autoclave with Cycom AS4/1808 prepreg tape as per the manufacturer's specifications. 2
Fiber volume fraction was nominally 60% in both laminates. Specimens measuring 152 mm
long and 38.1 mm wide were cut from the cured panels with a diamond-impregnated wheel. 3
Average gross (unnotched) cross-sectional area for both specimen types was 174 mm 2. A
9.53-mm-diameter hole was drilled through the center of each specimen with a diamond-
impregnated core drill.
Mechanical tests were performed on a servohydraulic test frame equipped with hydraul-
ically actuated wedge grips. Using the test method for reversed cyclic loads described in
Ref 7, the specimens were designed with 63.5 mm of unsupported length between the grips
to allow damage to develop without the influence of antibuckling supports. Quasistatic
strength tests were performed in displacement control with a strain rate of 0.0004 s -~ mea-
sured over the gauge length. Fully reversed (R = - 1) cyclic loading was carried out with
a load-controlled 10-Hz sinusoidal form. Load levels for the two laminates were chosen such
that the specimens failed between approximately 5000 and 20 000 cycles.
An effective secant stiffness of the specimen was monitored with a 25.4-mm extensometer
centered on the notch, as previously described in Ref 8. During the fatigue lifetime of several
specimens cycled at a given load level, periodic monitoring of stiffness and the examination
of penetrant-enhanced X-ray radiographs enabled the identification of an approximate re-
lationship between stiffness and damage state (and, thereby, the fraction of life consumed),
both of which follow generally reproducible patterns [7,8]. For consistency, radiographs
included in this paper are all of the "front side" of the specimens, as per the ply-angle
notation described earlier.-'
The approach taken to document the influence of damage on the strain and stress dis-
tributions in and residual strength of notched laminates at a particular load level is: (1) cycle
four specimens to failure, with periodic interruptions for stiffness measurement and X-ray
radiography, to establish typical stiffness response and damage accumulation data throughout
life; (2) using stiffness measurement and radiography, identify three points in the fatigue
lifetime where stress redistribution and residual strength data are desired (these points turned
out to be approximately 15, 50, and 90% of life for the quasi-isotropic laminate, and 10,
50, and 90% of life for the orthotropic laminate); (3) cycle two additional specimens to each
selected point in the fatigue lifetime, based on stiffness change and radiographs; (4) with
the photoelastic coating and adiabatic thermal emission techniques, evaluate the full-field
strains for each specimen described in step 3; (5) measure-residual tensile and compressive
strength at each selected point in the fatigue lifetime.
Full-Field Strain Measurement Techniques

Photoelastic Coating
The photoelastic coating technique is well known for its capability of revealing true surface
strains of metallic components and has been used successfully to measure the static and
dynamic response of notched composite structures [3,9,10]. Detailed principles and pro-
cedures for the photoelastic coating technique are given in Refs 11 and 12. Briefly, pho-
2 Positive ply angles are measured clockwise from the longitudinal loading axis.
3 All original measurements were recorded in U.S. customary units.
toelastic coating materials are optically homogeneous when unstressed and optically
heterogeneous when stressed. That is, the index of refraction of the coating is a function
of the strain induced by the external load. When polarized light enters the coating, the light
vector splits into two polarized waves, which then proceed in the directions of the principal
strains at a particular point. Because the index of refraction is different in the two directions,
a relative retardation is formed, which forces the two waves to be out of phase when they
emerge from the coating. Therefore, all points having the same amount of principal strain
difference will possess an equal relative retardation. A color fringe corresponding to those
points can be viewed through a crossed reflection-type polariscope. Corrections on the order
of the fringes are required when the mechanical reinforcement from the coating and the
Poisson's ratio mismatch between the coating and the specimen are not negligible [11,12].
A 1-mm-thick coating material with a reflective backing (PS-1, Photolastic Division,
Measurement Group, Inc., Raleigh, NC) was machined to match the specimen width and
hole size. To prevent interference with the grips, the coating length of 51 mm was slightly
less than the unsupported length of the specimen. The coating was bonded to the specimen
with PC-1 adhesive after the introduction of fatigue damage.
In the present investigation, no corrections to the fringe order were attempted. Instead,
fringes measured in specimens with damage were compared to fringes in a virgin specimen
under the same stress (102 MPa over the unnotched area). At various states of damage,
fringe orders measured along sections of the specimen labeled A - A ' and C - C ' in Fig. 1
were plotted and compared.
Adiabatic Thermal Emission

Noncontact measurement of the adiabatic, reversible (isentropic) thermoelastic effect in
materials as a means of determining the dynamic state of strain or stress during cyclic loading
has recently been made feasible with the commercial availability of the SPATE 8000 (Stress
t
C,
B
R.
C
FIG. 1--Schematic of notched specimen coordinate notation and identification of sections
over which full-field data are plotted.
Pattern Analysis by measurement of Thermal Emission) instrument by Ometron, Inc.,

Herndon, VA. Through infrared optics, a computer-controlled camera measures (with a
resolution of 0.001~ the minute change in surface temperature occurring at the same
frequency as the applied sinusoidal load. The camera scans the specimen in a raster-like
manner, with each measurement point representing an average temperature change in a
circle approximately 0.5 mm in diameter. The only specimen preparation required is a thin
coating of flat black paint for a high, uniform emissivity and a diffuse surface finish. If
adiabatic and reversible deformations are maintained, the measured temperature change
can be related to the state of strain or stress in the material.
The theory linking temperature change with isentropic deformations of elastic matter was
established by Thomson in 1857 [13]. Blot later expressed the relationship for homogeneous
anisotropic materials as (summation implied over repeated indices)
To
T - To = --- otktC,~te,
C~
where
To and T = the temperatures of the material before and after deformation, respectively,
c~ = the volumetric specific heat at constant strain,
a~ = components of the thermal expansion tensor,
C,jkt = components of the isothermal elastic modulus tensor, and
e,, = components of the strain tensor [14].
Thus, a negative temperature change occurs during positive dilatation of an orthotropic

material with positive thermal expansion coefficients. A simplified relationship between the
plane-stress, isentropic stresses in a heterogeneous, orthotropic material, such as a graphite/
epoxy lamina, and the attendant temperature change measured in the laboratory is obtained
through consideration of the nonhomogeneous deformations and temperature changes of
the microconstituents (that is, the fibers and matrix). The result can be represented by an
equation of the form
T- To = K,0-~ + K20-2 (1)

where
To and T = the average, effective temperatures of the material before and after defor-
mation, respectively,
K~ and K2 = material constants involving the initial temperature, thermomechanical
properties, and volume fraction of each microconstituent, and
0-t and 0"2 the in-plane stress components derived from the theory of anisotropic plates
~
[15].
Because a typical graphite/epoxy lamina has a very small (or slightly negative) thermal
expansion coefficient parallel to the fibers (0".), a relatively small temperature change results
from uniaxial loads in the fiber direction. On the other hand, a very large temperature
decrease results from transverse tension in the lamina due to the large, positive transverse
thermal expansion coefficient (c~:2).
The accuracy of strain or stress measurements via infrared thermal emission measurements
has been verified previously for isotropic, homogeneous materials [16], but not for aniso-
tropic, laminated, composite materials. Therefore, the technique and related analytical tools
are still under development. The advantage of the technique is that it is an entirely non-
TABLE 1--Effective secant stiffness of virgin center-notched AS4/1808 laminates measured during the
first load cycle.
Tensile Stiffness, GPa Compressive Stiffness. GPa
Stacking
Sequence Quantity Min Max Mean Quantity Min Max Mean
(0/45/90/-45),4 9 30.4 32.3 31.0 9 27.1 29.9 28.4

(0/45/0/-45),4 12 41.3 44.8 42.8 12 36.1 39.1 37.6
contact, nondestructive evaluation tool. A limitation is the unknown amount of dissipative

versus adiabatic temperature change. Furthermore, it is difficult to isolate the individual
components of strain or stress from a single measured temperature change when the local
deformations are influenced by changing elastic properties and boundary conditions. A t
present, these factors restrict the thermal emission data to qualitative analyses of defor-
mations.
In the series of thermal emission measurements, all specimens of similar stacking sequence
were cycled at the same reduced stress limits, which were selected such that negligible
damage development occurred during the scanning process. The stress limits for the qua-
siisotropic and orthotropic specimens were 2.6 to 104.7 MPa and 2.6 to 130.2 MPa, re-
spectively. The entire gauge length of the specimen was scanned on both the front and back
sides. Full-field color contour maps of the temperature change were photographed on the
system monitor screen. In the included illustrations, the color scale is uncalibrated but is
directly proportional (including sign) to the difference T~, - T, where To and T are the
instantaneous surface temperatures at the minimum and maximum loads, respectively. Tem-
perature changes along sections B - B ' and C-C' (Fig. 1) were plotted and compared for
specimens with various states of damage.

Quasistatic Properties and Fatigue Life
Tensile and compressive quasistatic strengths and secant (zero to maximum cyclic load
excursions) stiffnesses for the notched specimens are given in Tables 1 and 2, where stress
has been computed over the gross (unnotched) cross-sectional area and strain has been
measured across the notch. Both laminates are stiffer and weaker in tension than in compres-
sion.
To achieve fatigue lives on the order of 5000 to 20 000 cycles, stress levels of -+203 MPa
and -+305 MPa were chosen for the quasi-isotropic and orthotropic laminates, respectively.
All specimens tested at these load levels failed in a compressive mode due to unstable
delamination growth in the ligament of material between the notch and straight edge.
TABLE 2--Strength of virgin center-notched AS4/1808 laminates.
Tensile Strength, MPa Compressive Strength, MPa

Stacking
Sequence Quantity Min Max Mean Quantity Min Max Mean
(0/45/90/-45~ 3 265 275 272 3 -304 -318 -312

(0/45/0/-45),4 3 354 392 373 3 -373 -394 -381
FIG. 2--Front view radiographs or fatigue damage in quasiisotropic specimens. (a) Specimen
3-12 at 15% of life. (b) Specimen 3-2 at50% of life. (c) Specimen 2-3 at 90% of life.
Fatigue Damage Initiation and Growth

Fatigue damage in the ( 0 / 4 5 / 9 0 / - 45),4 laminate initiates at the notch with matrix cracking
in all plies by the first or second load cycle, followed shortly by delaminations of the 0/45,
45/90, and 9 0 / - 4 5 interfaces in regions of dense matrix cracks (Fig. 2a). Delaminations
form earliest and grow fastest at interfaces closest to the surface. In situations where
delaminations have not yet formed between the surface 0 ~ ply and the underlying + 45 ~ ply
near the notch, short (up to 2 mm long), piecewise-linear ply fractures in the 0 ~ ply form
sequentially in a stepwise manner along the direction of an underlying +45 ~ matrix crack
(Fig. 3a). The 0 ~ ply fractures serve as catalysts for matrix cracking in the 0~ ply and for
%
I.OAD
(o) (b)
FIG. 3--Schematic of typical ply fracture patterns and associated matrix cracks. (a) 0~ ply
fracture seen in the quasi-isotropic and orthotropic laminates. (b) 45 ~ ply fracture seen in the
orthotropic laminate.
delamination of the associated 0/45 interface. This mode of damage is primarily a surface
effect, although it occasionally develops in interior plies near the end of life. The damage
growth rate is slowest between the tenth and ninetieth percentiles of life. Figure 2b illustrates
the characteristic X-shaped pattern of damaged material at an estimated 50% of fatigue life.
The damage condition in specimen 2-3 near the end of life (Fig. 2c) is atypical in that it
closely resembles the earlier damage state in specimen 3-2. More frequently, delaminations
will extend across the entire width of the specimen at the end of life. In specimen 2-3,
however, local fracture of the surface 0~ plies near the notch, as in Fig. 3a, was sufficient
to cause the precipitous loss of compressive stiffness that precedes specimen failure.
Fatigue damage initiation in the ( 0 / 4 5 / 0 / - 45)s4 laminate is dominated by 0 ~ matrix cracks
tangent to the notch (Fig. 4a). The cracks initiate symmetrically in all four quadrants, but
grow fastest in the upper-left and lower-right quadrants. This behavior is attributed to an
interaction of the 0 ~ matrix cracks and adjacent paths of ply fractures in the + 45 ~ plies
following those cracks only in the upper-left and lower-right quadrants (Fig. 3b). The -45,,
ply groups do not display this mode of damage. Staircase-type fractures of the 0 ~ plies, as
in Fig. 3a, also grow away from the notch along the + 45 ~ direction. As in the quasi-isotropic
laminate, this is primarily a surface effect, although large 0 ~ ply breaks have been noted to
pass through the adjacent 45~ ply and into the next 0 ~ ply. Matrix cracks in the off-axis plies
are concentrated along the length of the 0 ~ cracks during the early stages of damage accu-
mulation. At approximately half of fatigue life, delaminations at interfaces involving the
-452 ply groups have grown away from the notch and along the 0 ~ matrix cracks tangent
to the notch (Fig. 4b). Much of the new damage growth after this point in fatigue life is
associated with the tips of the large 0 ~ matrix cracks. At impending laminate failure, de-
laminations typically extend across the width of the specimen, resulting in a large stiffness
loss and a compressive mode of failure (Fig. 4c). Much of the delaminated area visible in
Fig. 4c is due to the 0/45 interface delaminations associated with several fracture paths in
the 0 ~ surface plies.
FIG. 4--Front view radiographs of Jatigae damage in orthotropic specimens. (a) Specimen
3-3 at 10% of life. (b) Specimen 3-7 at 50% of life, (c) Specimen 2-18 at 90% of life.
TABLE 3--Center-notched (0/45/90/- 45),,AS4/1808 laminate residual properties under fully-

reversed fatigue at 203 MPa.
Stiffnessa Strength a
Specimen Approximate No. of
I.D. % of Life Cycles E, Er S, Sc
3-12 15 1 305 0.93 0.96 1.11
2-1 15 904 0.94 0.92 0.'839
3-2 50 9 240 0,85 0.92 1.'i9
3-14 50 9 400 0.85 0.89 0.'896
2-3 90 11 024 0.67 0.73 0.'883
2-2 90 8 800 0.82 1.14 ... 0.'628
~ Normalized to values for virgin specimens.
Residual Stiffness and Strength

Residual strengths and stiffnesses for six quasi-isotropic specimens at various stages of
fatigue life are given in Table 3. X-ray radiographs of specimens used in the tensile residual
strength tests were seen previously in Fig. 2. A typical plot of tensile and compressive
stiffnesses during the fatigue lifetime, normalized to their respective values on the first load
cycle, is given in Fig. 5a. During the first 10% of life, the tensile and compressive stiffnesses
and the compressive residual strength decline rapidly, while the tensile residual strength
increases rapidly. During the middle 80% of fatigue life, the rate of stiffness change slows
markedly. Neither the tensile nor compressive residual strength changes appreciably during
this period. At impending laminate failure, tensile and compressive stiffnesses decline
sharply, as do both components of residual strength. Residual tensile strength at impending
failure is noted to be below the average tensile strength of virgin specimens. As evidenced
by the residual compressive strength data and postfailure fractography, fatigue failure of
the specimen occurs during the compressive portion of the load cycle. The relatively high
stiffness values indicated for specimen 2-2 reflect the nonsymmetrical distribution of damage
through the thickness. In this case, the side of the specimen on which the extensometer was
attached experienced much less strain than the opposite side because of an out-of-plane
deflection in the gauge length. Although in Table 3 the stiffness degradation in tension
generally exceeds that in compression, results from other fatigue tests at the same load level
indicate that this is not always true. The rates of change of stiffness and residual strength
were noted to correlate with the damage growth rate.
Residual properties for orthotropic specimens at three stages of fatigue life are given in
Table 4, and a typical normalized stiffness response during the fatigue lifetime is shown
graphically in Fig 5b. X-ray radiographs of residual tension test specimens were seen pre-
viously in Fig. 4. Stiffness response of this laminate differs from that of the quasiisotropic
laminate in that there is a more uniform degradation throughout fatigue life and a slightly
greater cumulative change at failure. Tensile stiffness change generally exceeds compressive
stiffness change throughout fatigue life. Up to approximately 50% of life, the residual tensile
strength is noted to increase more than in the quasi-isotropic laminate, while the residual
compressive strength initially increases, then decreases to a value nearly equal to virgin
specimen values. During the last 10% of fatigue life, when surface 0 ~ ply fractures and
delaminations are the primary modes of damage growth, both components of residual
strength decrease until the compressive strength reaches the applied cyclic load amplitude,
The tensile strength near the end of life is still greater than that of virgin specimens. With
0.8
0.6
C 3
tl.I
0.4
d
Q.E
LEOENfl, El TENSILE STIFFNESS
A COMPRESSIVE STIFFNESS
0.0 , , , _ . . . . . . . . . . , . . . . . , . . . . . , . . . .
0.0 3.0 6.0 9.0 12.0 lS.O
''0 1
CYCLES t T ~ 1
~ 0.8
7- o . s
W
0
W
N 0.4
0.2 LEOENDs Ill TENSILE STIFFNESS

m COMPRESSIVE STIFFNESS
0.0
' r ' - - - , - - - , - - , - - , - - , - - , - - - , - . -
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0

CYCLES ( THtRJSRNDS 1
FIG. 5--Typical normalized tensile and compressive stiffnesses during fatigue lifetime. (a)
Quasiisotropic specimen. (b) Orthotropic specimen.
T A B L E 4--Center-notched (0/45/0/- 45),, AS4/1808 laminate residual properties under fully-reversed

fatigue at 305 MPa.
Stiffness~ Strength ~
Specimen Approximate No. of
I.D. % of Life Cycles E, E, S, S,
3-3 10 250 0.94 0.92 1.16
3-5 10 350 0.94 0.94 ... 1.04
3-7 50 3 708 0.79 0.77 1.42 H - -
3-9 50 6 570 0.79 0.87 ... 1.02

2-18 90 10 460 0.67 0.76 1.10 H - -
4-7 90 5 760 0.66 0.73 ... 0.884
Normalized to values for virgin specimens.
BAKtS ET AL. ON DAMAGE INITIATION AND GROWTH IN NOTCHED LAMINATES 75
increasing amounts of fatigue damage, the typical fracture path seen in residual tensile
strength tests is offset from the transverse centerline through the hole to locations closer to
the tips of the large 0~ matrix cracks tangent to the hole.
Strain Redistribution
Full field measurements of isochromatic fringes and thermal emission in all of the spec-
imens listed in Tables 3 and 4 were recorded, in many cases on both front and back surfaces.
For brevity, only three images will be included here: a virgin specimen, a middle-life spec-
imen, and a late-life specimen. In all cases except the virgin specimen, the photoelastic data
were obtained on the same side of the specimen as the thermographic data. Hence, the
sensitivity of the two techniques to damage can be evaluated directly.
lsochromatic Fringe Measurements--Figure 6 illustrates the isochromatic fringe patterns

in three quasiisotropic specimens with sequential states of fatigue damage. The middle- and
late-life radiographs were seen previously in Fig. 2. Due to the increased compliance in the
region of intense matrix cracks and delamination adjacent to the notch, higher fringe orders
appear on the photoelastic coating directly over the damage zone. The existence of increased
fringe orders in the region surrounding the damage zone indicates that a larger proportion
of the load is being carried by the undamaged material. The slight asymmetry in Fig. 6c is
the result of a short fracture path in the back-surface 0 ~ ply directly under the coating, on
the left side of the illustration. As damage grows away from the notch in the transverse
direction, regions of high fringe order move toward the straight edges of the specimen. This
phenomenon is most prominent at the tips of the 0 ~ matrix cracks bounding the near-surface
0/45 delamination. Figures 7a and 8a quantify this effect along sections A - A ' and C-C' of
the specimen, respectively. Along section A - A ' , the increase in remote strain near the straight
edges is most evident in the late-life specimen. Fringe orders along section A - A ' are also
influenced by the growing areas of lightly loaded material directly above and below the
notch. Along section C-C', fringe orders are altered by the transverse growth of the damage
zone and the associated strain concentration.
FIG. 6--1sochromatic fringe patterns in the photoelastic coating of quasi-isotropie specimens.

(a) Specimen 3-10 with no damage (front). (b) Specimen 3-2 at50% of life (front). (c) Specimen
2-3 at 90% of life (rear).
---A A'---
4 ' I i I ' I ' J
----o-.~VIRGIN 50'4 LIFE " ~ v - - 9 0 ' / . LIFE

Y
I
A A_"x
~2
Z
0
-4 -2 0 2
XIRo
---A A'---~
/' - ' I I ' I '
50~176
LIFE ~ 90% LIFE
Y
3 I
= A~K X
=o
0 , I i
-4 -2 0 2
X/R o
FIG. 7--Birefringence distributions along section A - A ' . (a) Quasiisotropic specimens. (b)
Orthotropic specimens. Ro = hole radius.
In the orthotropic laminate, the fringe patterns become asymmetric about the load axis
after the appearance of damage (Fig. 9). This response is attributed to the asymmetric
distribution of damage seen previously in the X-ray radiographs of Fig. 3. High fringe orders
are induced across the 0~ matrix cracks extending tangentially from the notch in the upper-
left and lower-right quadrants due to high shear deformation across the crack, as can be
seen in a plot of fringe order along section A - A ' (Fig. 7b). A more uniform distribution of
fringes evolves in the load-bearing ligaments of material isolated by the 0 ~ cracks. The
regions of uniform fringe distribution directly above and below the notch have a larger area
C~-,
I ' I ' t I '~ i I ' I
~IRGIN .~-
C'l I,~
ul
o2
w
I i I I I i I , I
.4 -2 0 2 4
Y/R o
"-C C~
I m I ' I ' I I
VIRGIN ~ 50% LIFE "--'~90"/. LIFE
Y
c'l
ne
c
2
u.
0 I L I i I i I , I
-4 -2 0 2 4
Y/Ro
FIG. 8--Bire[ringence distributions along section C-C'. (a) Quasiisotropic specimens. (b)
Orthotropic specimens. Ro = hole radius.
than the corresponding regions in the quasiisotropic laminate. In the midlife specimen, a
short line of 0 ~ ply fracture to the left of the notch causes a strong, local concentration of
strain in the coating (Fig. 9b). At the late state of damage, fringes along section C - C ' are
relatively uniform due to strain redistribution and the absence of 0 ~ ply fracture on the front
surface (Fig. 8b). A delamination between the surface 0 ~ ply and the underlying 45 ~ ply
caused by 0~ ply fracture in the lower-right quadrant near the end of life results in greatly
reduced fringe order in the unloaded area (Fig. 9c).
FIG. 9--1sochromatic fringe patterns in the photoelastic coating of orthotropic specimens.

(a) Specimen 2-11 with no damage (front). (b) Specimen 3-7at50% of life (front). (c) Specimen
2-18 at 90% of life (front).
Adiabatic Thermal Emission Measurements--Thermographs of (0/45/90/- 45),4 specimens

with sequential states of damage development are shown in Fig. 10. The outstanding feature
of the series is the growth of a region of near-zero thermal emission around the notch.
Considering the thermoelastic relation of Eq 1, the decreased temperature change can be
caused by load relaxation in the surface ply via 0 ~ ply fracture and delamination, as in Fig.
3a. In the corresponding radiographs of Fig. 2, it is apparent that the low thermal emission
region encompasses the delaminated area under the surface ply. As surface ply delaminations
grow in the transverse direction, the "hot spots" originally located adjacent to the notch
shift immediately ahead of the delaminations (Fig. lla), reflecting the redistribution of
strain around the damage zones. The highest magnitude of temperature change in each
thermograph, normalized to the remote value, varied from 2.8 in the undamaged specimen
FIG. lO--Adiabatic thermal emission patterns in quasiisotropic specimens. (a) Specimen 3-

10 with no damage (rear). (b) Specimen 3-2 at 50% of life (front). (c) Specimen 2-3 at 90%
of life (rear).
B B'-"
800 I I '
VIRGIN ~ - - 5 0 % LIFE ------ 9 0 % LIFE
LU 6o0 Y
I
Z B~B'-X
Z
"l-
u 200 |
W
W
o. 0
!/',
V
"2""-4~ -2 0 2 4
X/R,
"- C C'~
600 . . . . . I . . . . .
VIRGIN ~ ~ 5 0 % LIFE ------ 9 0 1 , LIFE
A
Q
c.T
<
W
360
Z
ii~ \" J \ sz
~''.. ~ ii
W
~
~J
- ~ 0
V
-600
-6 -~ -2 0 2 4 6
Y/R,
FIG. 11--Adiabatic thermal emission profiles. (a) Section B-B' in quasiisotropic specimens.
(b) Section C-C' in orthotropic specimens. Ro = hole radius.
to 2.1 and 2.5 in the middle- and late-life specimens, respectively. In an inverse manner,
the trend in maximum temperature change agrees with the trend of increasing then decreasing
residual tensile strength during fatigue life. Furthermore, the specimens shown in Figs. 2
and 10 failed in tension along the transverse section of the laminate that includes the notch
and the hot spots. Hence, the region of high temperature change coincides with the region
that is involved in the tensile fracture event, and the magnitude of the temperature change
in those regions correlates well with the fracture load. In conjunction with radiographs,
thermographs provide an understanding of the effects of damage on the stress distribution

that was heretofore unavailable with radiographs alone. For example, 0 ~ ply fractures near
the notch on the back side of specimen 2-3 are visible to the discerning eye in Fig. 2c, but,
given the overall similarity of Figs. 2b and 2c, it is not obvious that there would be a significant
difference in residual strength. However, one may suspect such a strength change by noticing
the differences in the corresponding thermographs of Figs. 10b and 10c.
Due to the high Poisson's ratio of the ( 0 / 4 5 / 0 / - 45),4 laminate, the far-field temperature
change is very close to zero. Thermographs of virgin, middle-life, and late-life orthotropic
specimens indicate the severity of strain redistribution in the gauge length (Fig. 12). In this
laminate, the hot spots initially adjacent to the notch generally shift in the load direction,
close behind the tips of the 0 ~ matrix cracks tangent to the notch. Figure 11b illustrates the
movement of the hot spot away from the notch along profile C-C'. The effect is most
noticeable in the lower-right quadrant of Fig. 12b and in the upper-left quadrant of Fig.
12c. It is in these two quadrants that much of the middle- and late-life fatigue damage growth
occurs. As in the quasiisotropic laminate, the fracture path seen in residual tensile strength
tests frequently passes through, or at least near, the hot spots. The maximum, unscaled
temperature changes seen in the virgin, middle-life, and late-life specimens of Fig. 12 were
500, 430, and 475, respectively, which correlate in an inverse manner to the corresponding
values of residual tensile strength.

Damage
In both laminates, there exist matrix cracks parallel to the fibers in all plies, delamination
between plies of differing orientation, and macroscopic concentrations of fiber fracture
extending through the thickness of a lamina (local ply fracture). Damage in both cases
initiates along the hole boundary in the form of matrix cracking, followed by delamination.
Short, 0~ply fractures served as sites for delamination initiation away from the hole boundary.
The different patterns of matrix cracking and delamination in the two laminates can be
attributed to the directional nature of damage growth caused by the interaction of damage
in adjacent plies. For example, the lack of 90~ plies and the abundance of 10~ plies in the
FIG. 12--Adiabatic thermal emission patterns in orthotropic specimens. (a) Specimen 2-11
with no damage (rear). (b) Specimen 3-7 at 50% of life (front). (c) Specimen 2-18 at 90% of
life (front).
orthotropic laminate resulted in a routing of damage longitudinally along paths tangent to

the notch, rather than along the transverse centerline, as in the quasiisotropic laminate.
Residual Stiffness and Strength

Tensile and compressive stiffness degradation rates in the orthotropic material were more
uniform over the fatigue lifetime than in the quasiisotropic material, in which the rates were
highest at the beginning and end of life. Tensile stiffness degraded faster than compressive
stiffness in the orthotropic laminate. Residual compressive strength of the quasiisotropic
laminate degraded quickly early in life, and again at the end of life, causing a compressive
mode of failure under the reversed cyclic loads. The orthotropic laminate also failed in
compression, but the compressive residual strength did not degrade significantly until near
the end of life. Until late in the fatigue lifetime of both laminates, residual tensile strengths
increased to values above those for virgin specimens. At impending failure, residual tensile
strength degraded in both laminates, but not sufficiently to induce a tensile failure mode.
Photoelastic Coating
The isochromatic fringe patterns measured with the photoelastic coating technique at
various points in the fatigue lifetime provide an understanding of strain redistribution in the
presence of damage. In the quasiisotropic laminate, damage zones near the notch were
indicated by extremely high fringe orders. Zones of high fringe order in the undamaged
material shifted toward the straight edge of the specimen as delaminations grew away from
the notch, reflecting the damage-induced redistribution of load. In the orthotropic laminate,
the long 0 ~ matrix cracks tangent to the hole resulted in a line of concentrated fringes along
the cracks and a more uniform strain field in undamaged ligaments of material adjacent to
the notch. This result is attributable to the lack of shear load transfer capability across the
0~ matrix cracks tangent to the hole. In situations in which fractures in the surface ply grew
transversely to the notch (accompanied by delamination of the surface ply), high fringe
orders resulted over the fracture path, and low fringe orders resulted over the unloaded
lamina. In both laminates, fringe orders increased in the remaining load-bearing material
after the formation of fatigue damage near the notch. Fringe orders decreased in the regions
above and below the notch, reflecting the relaxation of strain there. Overall, more strain
redistribution was seen in the orthotropic laminate than in the quasiisotropic laminate due
to the more complete reduction of the notch effect by the damage. The greater increase in
residual tensile strength of the orthotropic laminate supports this observation.
Thermal Emission
Full-field strain distribution data near the notch measured with the adiabatic thermal
emission technique provide valuable insight into the nature of load redistribution in the
presence of damage, especially when used in conjunction with X-ray radiography. The
damage mechanisms that cause increased strains and, hence, increased thermal emission in
some region of the laminate are likely to cause that same region to be involved in damage
growth, and, eventually, the tensile fracture process. Measurements indicated that regions
of high thermal emission shifted to remain in front of growing damage zones. In the qua-
siisotropic laminate, the critical zones of high thermal emission shifted away from the notch
along a transverse direction, just ahead of the 0/45-interface delaminations under the surface
ply. In the orthotropic laminate, the critical zones shifted along the 0 ~ matrix cracks tangent
to the hole. The increase and decrease of temperature change measured in the critical zones
corresponded well with the increase and decrease of residual tensile strength of the speci-
mens. Regions of high thermal emission were frequently part of the tensile fracture surface.
In both laminates, there was no correlation between the amplitude of thermal emission and
the residual compressive strength, which can be attributed to a dependence of residual
compressive strength on the size and distribution of delaminations and their effect on the
stability of the notched laminates, rather than on the notch effect itself. The thermographs,
however, did indicate delamination growth and therefore can be used to indicate a potential
loss of compressive strength. In contrast with the photoelastic data, the thermographic
measurements indicated little deformation near fracture paths in the surface plies. This is
due to the sensitivity of thermal emission to actual surface material deformation, rather
than to a coating deformation.
General
Significant relations between fatigue damage and the concomitant changes in residual
stiffness, residual strength, and strain distribution in two notched laminates have been
presented. In particular, the associations existing between the residual tensile strength and
the distribution and magnitude of strains measured with the photoelastic coating and adi-
abatic thermography techniques were quite encouraging and should be investigated further.
Acknowledgment
The authors wish to acknowledge the support of the U.S. Air Force Office of Scientific
Research through contract No. 85.0087.
References
[1] Highsmith, A. L. and Reifsnider, K. L., in Composite Materials: Fatigue and Fracture, ASTM
STP 907, H. T. Hahn, Ed., American Society for Testing and Materials, Philadelphia, 1986, pp.
233-251.
[2] Highsmith, A. L. and Reifsnider, K. L., in Fracture of Fibrous Composites, AMD Vol. 74, C. T.
Herakovich, Ed., American Society of Mechanical Engineers, New York, 1985, pp. 71-87.
[3] Daniel, I. M., Experimental Mechanics, Vol. 25, No. 12, Dec. 1985, pp. 413-420.
[4] Kress, G. R. and Stinchcomb, W. W., in Recent Advances in Composites in the Unites States and
Japan, ASTM STP 864, J. R. Vinson and M. Taya, Eds., American Society for Testing and
Materials, Philadelphia, 1985, pp. 173-196.
[5] Sandhu, R. S., Gallo, R. L., and Sendeckyj, G. P., in Composite Materials: Testing and Design
(Sixth Conference), ASTM STP 787, I. M. Daniel, Ed., American Society for Testing and Ma-
terials, Philadelphia, 1982, pp. 163-182.
[6] Irvine, T. B. and Ginty, C. A., Journal of Composite Materials, Vol. 20, March 1986, pp. 166-
184.
[7] Bakis, C. E., Simonds, R. A., and Stinchcomb, W. W., "A Test Method to Measure the Response
of Composite Materials Under Reversed Cyclic Loads," in Test Methods and Design Allowables
for Fiber Composites (Second Symposium), ASTM STP 1003, C. C. Chamis and K. L. Reifsnider,
Eds., American Society for Testing and Materials, Philadelphia, in press.
[8] Bakis, C. E. and Stinchcomb, W. W., in Composite Materials: Fatigue and Fracture, ASTM STP
907, H. T. Hahn, Ed., American Society for Testing and Materials, Philadelphia, 1986, pp. 314-
334.
[9] Kawata, K., Takeda, N., and Hashimoto, S., Experimental Mechanics, Vol. 24, No. 12, Dec. 1984,
pp. 316-327.
[10] Daniel, I. M., Rowlands, R. E., and Whiteside, J. B., in Analysis of the Test Methods for High
Modulus Fibers and Composites, ASTM STP 521, American Society for Testing and Materials,
[11] Zandman, E, Render, S., and DaUey, J. W., Photoelastic Coatings, SESA Monograph 3, Society
for Experimental Stress Analysis, Westport, CT, 1977.
[12] Whitney, J. M., Daniel, I. M., and Pipes, R. B., Experimental Mechanics of Fiber Reinforced
Composite Materials, SESA Monograph 4, 1st ed., Society for Experimental Stress Analysis,
Brookfield Center, CT, 1982.
[13] Thomson, Sir William, "On the Thermo-elastic and Thermo-magnetic Properties of Matter,"
Quarterly Journal of Mathematics, Vol. 1, 1857, pp. 57-77.
[14] Biot, M. A., Journal of Applied Physics, Vol. 27, No. 3, 1956, pp. 240-253.
[15] Bakis, C. E., "On the Residual Strength of Notched Carbon Epoxy Laminates Under Reversed
Cyclic Loads," Ph.D. dissertation, Department of Engineering Science and Mechanics, Virginia
Polytechnic Institute and State University, Blacksburg, VA, 1988.
[16] Stanley, P. and Chan, W. K., Journal of Strain Analysis, Vol. 20, No. 3, 1985, pp. 129-137.
Models and Analysis
WoonBong Hwang t and Kyung S. Han I
Fatigue of Composite Materials Damage

Model and Life Prediction
REFERENCE: Hwang, W. and Hart, K. S., "Fatigue of Composite Materials--Damage Model

and Life Prediction," Composite Materials: Fatigue and Fracture, Second Volume, ASTM STP
1012, Paul A. Lagace, Ed., American Society for Testing and Materials, Philadelphia, 1989,
pp. 87-102.
ABSTRACT: Fatigue life prediction on composite materials is studied analytically using deg-
radation and damage models, resultant strains, and fatigue modulus. Definition of fatigue
modulus, new damage models using fatigue modulus and resultant strain, and prediction of
fatigue life of composite materials using degradation and damage models are discussed. This
approach can predict accurately the multi-stress level fatigue life as well as single-stress level
fatigue life of composite materials. Fatigue life is predicted by the following procedures: (1)
establish the fatigue modulus degradation model, (2) find fatigue life equation as a function
of fatigue modulus, (3) calculate the fatigue life using strain failure criterion. Degradation
models for composite damage are generalized; the three-parameter degradation model is found
most suitable to predicting fatigue life of composites. Also the predicted two-stress level fatigue
life using the proposed cumulative damage models is reasonably close to the experimental
data.
KEY WORDS: composite materials, fatigue (materials), failure, stress, strain, fatigue mod-
ulus, fatigue life, damage
Fatigue life prediction of composite materials has been the subject of many investigations
during the past two decades. Recently, the residual strength degradation [1-6] and modulus
degradation [7-12] approaches have been frequently used to predict fatigue life of composite
materials.
Broutman and Sahu [1] proposed a new cumulative damage theory using a linear strength
degradation equation. Hahn and Kim [2] introduced a nonlinear residual strength degra-
dation equation which assumes that the slope of the residual strength is inversely proportional
to some power of the residual strength itself. This nonlinear degradation model was inves-
tigated further by Yang et al. [3,4] for the purpose of reliability study in fatigue life of
composite materials. Charewicz and Daniel [5] proposed a damage model based on the
assumption that the residual strength degradation rate is a function of life fraction, n/N,
but not a function of residual strength.
Multi-stress level fatigue life and fatigue life distributions could be predicted well by the
above-mentioned residual strength degradation models. However, these models fail to pre-
dict single-stress level fatigue life because the applied stress-dependent parameter is difficult
to obtain by analytical method.
An advanced residual strength degradation model which can predict single- and multi-
stress level fatigue life of composite laminates is proposed by Reifsnider and Stinchcomb
Assistant professor and associate professor, respectively, Department of Mechanical Engineering,

Pohang Institute of Science and Technology, Pohang, Korea.
87
Downloaded/printed
[6]. They start with the nonlinear Broutman-Sahu type, which can be interpreted that the
strength degradation rate is a power function of fatigue cycles, n. After setting up a basic
equation, they modified it by defining critical element and introducing local failure function
which are deduced from the microstructural analysis in the composite laminates. To predict
fatigue life of composite laminates, S-N relation of unidirectional composites is still needed.
Broutman and Sahu [7] observed the continuous reduction of elastic modulus of composites
during fatigue test. The secant modulus degradation during fatigue test was investigated by
Hahn and Kim [8] and by O'Brien and Reifsnider [9]. Wang and Chim [10] proposed a
fatigue damage model defined by elastic modulus. They assumed that a fatigue damage
degradation rate is a power function of fatigue cycles, n, and proportional to a parameter
which is a function of damage. A fatigue life prediction method on composite laminates is
suggested by Poursartip et al. [11,12] using the elastic modulus degradation during fatigue.
One of the main differences between the residual strength degradation and the modulus
degradation approaches is the failure criterion. Usually, in the residual strength degradation
method, the only failure condition is that failure occurs when the residual strength degrades
to the applied stress. To predict fatigue life by modulus degradation methods, another failure
condition is needed in addition to the above condition. The secant modulus failure criterion,
which states "failure occurs when fatigue secant modulus degrades to the static secant
modulus," as proposed in Refs 8 and 9, is shown schematically in Fig. 1. A strain failure
STATIC
FINAL
LOADING
///-~ Es.(O/
I.IJ O'a
I--
EsN(N)
EsN(N)=EsN(O)
0 AIE(N) Eu
STRAIN
FIG, 1--Secant modulus failure criterion [8,9]. (The lower tip of the stress-strain loops of
final loading is translated to the origin. )
HWANG AND HAN ON FATIGUE OF COMPOSITE MATERIALS 89
or.
STAT
F(N)= F(O)%
(/) F(O)
IM
rr ~a
I--
FINAL
LOADING
r/ f i
0 Eu=•f
~. AE(N)
STRAIN
FIG. 2--Resuhant strain failure criterion [13,14].
criterion, which is "failure occurs when the fatigue resultant strain reaches the static ultimate
strain," as suggested in Refs 13 and 14, is shown schematically in Fig. 2. Another strain
failure condition, which could be interpreted as "failure occurs when the fatigue strain under
the final cycle equals the static ultimate strain," is found in Refs II and 12 and is shown in
Fig. 3.
In this study the previously proposed fatigue life prediction method [13-15] has been
extended by a more generalized fatigue modulus degradation model. Single-stress level
fatigue life is predicted as follows:
1. Assume that the fatigue modulus degradation rate at any fatigue cycle is followed by
a power function of number of fatigue cycles and fatigue modulus itself.
2. Find fatigue life equation as a function of fatigue modulus.
3. Calculate the fatigue life using strain failure criterion.
Multi-stress level fatigue life is predicted by cumulative damage approaches. First, the
fatigue damage is defined by fatigue modulus and resultant strain. Second, the damage is
formulated as a function of number of cycles, n, fatigue life, N, and applied stress level, q.
Then, following the usual procedure of cumulative damage theory, mutli-stress level fatigue
o-,,~-
STATIC
Or)
W ~a
E~
I-
C~ FINAL
LOADING
N)
E(N) = E(O) (7.

ira
V I
O Eu=AE(N)
STRAIN
FIG. 3--Elastic modulus failure criterion [ 11,12]. (The lower tip of the stress-strain loops
of final loading is translated to the origin. )
life is predicted. (Discussion on the structure of damage function is presented in Refs 14

through/6.)
E-glass fiber-reinforced epoxy resin materials (unidirectional composites type 1002) man-
ufactured by the 3M Company were used in this study. An MTS machine was used for static
and fatigue testing. Load was monitored by a load cell, and load and strain were recorded
by an x-y recorder for static test. Tensile strength measurements were conducted at a
displacement rate of 2.13 mm/min. The fatigue tests were performed in load control using
a sinusoidal waveform. The frequency was 1 to 3 Hz, which is believed to give a negligible
temperature rise during tests. The ratios of peak stress to ultimate strength (normalized
stress, q) used were 0.9, 0.85, 0.8, 0.75, and 0.7, and a constant minimum stress level of
0.05 was maintained.
Four series of tests were conducted to examine the cumulative damage caused by two-
stress level fatigue tests. To study low-high stress levels, the specimens were first subjected
to fatigue at the lower stress level to a certain number of cycles, nt. Those that survived
were then fatigued at the higher stress level up to failure. In the high-low test, specimens
were cycled at higher stress levels to n~ cycles and then fatigued at lower stress levels up to
failure.
Fatigue Modulus Concept

Due to degradation of composite materials under cyclic loading, the stress-strain curve
changes as the cycling continues (Fig. 4). Elastic modulus is obtained from a line n - r e .
Fatigue modulus is represented by a line between applied stress and resultant strain at a
specific loading cycle, n. Fatigue modulus is a slope of a line On' in Fig. 4. Therefore,
F ( n , q ) - e(n) - ~ ~(n) (1)
where
F ( n , q ) = fatigue modulus at nth loading cycle,
e(n) = resultant strain at nth loading cycle,
(~o = applied stress, and
q = ratio of applied stress, (~., to ultimate strength, (r..
Fatigue modulus, F, of a material is a function of loading cycle, n, and applied stress, q.

Initial and final conditions give the following relations:
F(O,q) = Fo = Eo (2a)
F ( S , q ) -- Fr (2b)
Fatigue modulus at 0th cycle, Fo, is assumed to be the same as elastic modulus, E,,, and
0'1'2' 3' n'
o I a a 4 n ,',+1 E(n)
STRAIN
FIG. 4--Fatigue modulus concept.
fatigue modulus at fracture is defined as Fj at the number of cycles to failure, N . Fo should

be constant regardless of applied stress, while FI depends on applied stress.
To make further derivation simple, it is reasonable to assume that applied stress has a
linear relation with resultant strain at any arbitrary loading cycle if the specimen undergoes
constant maximum loading. This assumption follows:
~. = F(n,) e(n,) (3)
where F(n,), fatigue modulus at nth cycle, is assumed to be a constant. Using this assumption,
it is concluded that for a single-stress level fatigue test, fatigue modulus is not a function of
applied stress but rather a function of loading cycle only, that is,
F = F(n) (4)
It is expected that fatigue modulus and conventional secant modulus differ a lot for short-
fiber composites. For unidirectional composites, differences will be small. However, fatigue
modulus concept and some assumptions (particularly the assumption that applied stress has
a linear relation with resultant strain at an arbitrary loading cycle) have to be verified by
experiments.
Fatigue Modulus Degradation Model

Assume that fatigue modulus degradation rate, d F / d n , is followed by a power function
of number of fatigue cycles, n, and fatigue modulus itself, F, that is,
dF C N c- t
dn A BFB_ t (5)
where A , B, and C are material constants. This model is more generalized than the one
proposed in Refs 13 and 14. All the possible cases of this generalized model are summarized
in Table 1, with fatigue life prediction equations (FLPE).
Single-Stress Level Fatigue Prediction

Integration of Eq 5 from nt to n, cycles gives
FS(n2) - P(n,) = - A(nf - n~) (6)
Substitution of n, = n and nt = 0 into the above equation follows:
FS(n) - Fn(O) = - A n c (7)
At failure, where n = N, Eq 7 becomes
F~ - Fg = - A N c (8)
Therefore, the number of cycles to failure is expressed as follows:
N =
{ [ (n/q}
M 1 - \Fo] ]J (9)
where M = F g / A .
As mentioned earlier, a failure criterion is needed to predict fatigue life using the above
equation. We chose the strain failure criterion suggested in Refs 13 and 14. This criterion
states that final failure of material occurs when the fatigue resultant strain reaches the static
ultimate strain. Based on the elastic stress-strain relation under static loading, the following
relation is obtained:
~. = E#u = FoG. (10)

where
or, = ultimate strength, and
e, = ultimate strain.
Fatigue modulus gives the following relation:
~o = Fr, oero (11)

where
tr, = applied stress,

Fr., = fatigue modulus at failure under the stress level, ~ra, and
ei., = resultant strain at failure under the stress level, or,.
Strain failure criterion (e, = er.a) gives the following relation between fatigue modulus and
stresses:
Ff,
a __ ~a
- q (12)
F0 ~.
Substitution of the above relation into Eq 9 provides
N = [M(1 - qS)ll'C (13)
Using the above equation, fatigue life of materials can be predicted under strain failure
criterion as long as applied stress level, q, and material constants, M, B, and C, are known.
Following the same manners in this paragraph, other fatigue life prediction equations (FLPE)
can be formulated from the other fatigue modulus degradation models. Those results are
presented in Table 1.
The conventional S - N curve (straight line in semilog scale) and Basquin's relation (straight
line in log-log scale) could be written as follows:
Conventional S - N curve
q = k logN + d (14)
where
q = applied stress level,
N = number of cycles to failure,
k = constant, slope of S - N curve in q - l o g N plane, and
d = constant, q-intercept of S - N curve in q - l o g N plane.
TABLE 1--Degradation models and fatigue life prediction equations (FLPE).

dF Cn c - l N = [M(1 - qn)],,c
1 A - -
dn B F B-I
(B,C ~ 0,1) ( M = F~/A)
dF N = [M(1 - q)]a,,c
2 -- = -ACnc
dn
(B = 1, C # O , 1 ) (M = Vo/A)
dF A N = M(1 - qn)
3
dn B F n-~
( B # 0 , 1 , C = 1) (M = F~/A)
4
dF
dn
- ACnC-~F N= [--~1 l n q l vc
(C # 0,1)
dF A N = exp [M(1 - q~)l
5
dn n B F s-I
(B ~ 0,1) (M = F f / A )
dF F N = q ~,'a
6 - A
dn n
7 dF us CnC- I N = (qB -- 1)l,C
dn = " o BFn_ 1
B a s q u i n ' s relation
tr, = % ( 2 N y (15)
where
~ra = applied stress,
2N = reversals to failure (1 cycle = 2 reversals),
~r = fatigue strength coefficient, and
b = fatigue strength exponent (Basquin's exponent).
The static failure occurs at one quarter cycle which could be considered as n = 0 or 1.
This point is represented as (1,0) or (1,1) in q - N plane. It is interesting that, regardless of
materials, proposed equations pass through either the point (1,0) ( F L P E 1, 2, 3, 4, and 7)
or (1,1) ( F L P E 5 and 6) in q - N plane while conventional S - N curve and Basquin's relation
do not pass one fixed point at static failure because of curve fitting difficulties [17]. This
aspect causes an inconsistency when the damage curve and residual strength degradation
diagrams are drawn with conventional life equations. For instance, a value other than static
strength is used [5,17] or life equation does not coincide with the residual strength degra-
dation curve [6] at the starting point.
Cumulative Damage Models

Model I
The fatigue modulus, which varies with the number of cycles during fatigue, could be
used to represent the damage. The fatigue damage, D, which satisfies the initial ( D = 0
when n = O) and final (D = 1 when n = N) conditions, could be expressed as follows:
Fo - F ( n )
D - (16)
F o - Fr
Equations 7 and 8 can be rewritten as follows:
1
F(n)Fo = 1- ~ nCJ]I,B (17)
Fo --~ U c] (18)
Substitution of the above two equations into Eq 16 gives the following final results:
111
O=[1]llB
-- 1 lib
(19)
1 - 1 - --~N c
Model II
Fatigue damage could be formulated as follows by the ratio of the resultant strain at nth
cycle versus the failure strain:
D = ~(n) (20)
~r
The resultant strain is defined in Fig. 4. By the fatigue modulus concept, the applied stress
and resultant strain could be written
era = F(n)~(n) (21)
From Eq 10 and the strain failure criterion, the failure strain could be expressed
~. = Foef (22)
Substituting Eq 21 and Eq 22 into Eq 20 gives
F0 (23)
D = q F(n)
The above equation also could be written as follows using Eq 17:
D = q (24)
Model III
Fatigue damage parameter is defined as follows:
D = lE(n) - lE0
(25)
lEf -- lEO
where
Or a
iE O = B (26)
Fo
Substitution of Eq 21, Eq 22, and Eq 26 into Eq 25 provides
q (27)
D-l_ q
From Eq 15, the above equation becomes
D=I_ q q [(1_1~ n c -1 j (28)
Model IV
Fatigue damage can be defined also as a power function of fatigue modulus, which is
expressed as follows:
F~ - FB(n)
D - Fon _ F~ (29)
Substitution of Eq 7 and Eq 8 into the above relation provides
Using the fatigue modulus degradation models presented in Table 1 and the above pro-
cedure, the four basic cumulative damage models defined by physical variables are for-
mulated as a function of the applied stress, q, the number of fatigue cycles, n, and fatigue
life, N. The results are presented in Table 2. Model III-3, n/N, is the same as Miner's
model; (n/N) c is the same as modified Miner's model; and model IV-5, lnn/lnN, is the
same as Hashin and Rotem's model [17-18] with the reference parameter equal to 1. (The
use of natural or common logarithm makes no difference.)
Measurement of Fatigue Modulus
To evaluate fatigue modulus, F(n), the successful measurement of resultant strain, r

is essential. Several studies [19-21] have reported on the measurement of resultant strain
(at the maximum applied stress) and creep or permanent strain (at the zero or minimum
applied stress) of composite materials during fatigue.
In Ref 19, both resultant and permanent strains of short glass and carbon fiber-reinforced
TABLE 2 - - C u m u l a t i v e d a m a g e models.
Damage Model Type
Degradation
Type I II III IV
-Mncj q (N) c
q "-------~_ 1-- -1
1- [ l - ~NCJ
1 ]~.B [1 - ~ n c J q[ / ~ }
1 - lnc
1 - [1-1n] l's
q--------~
q r( ___L__I _ 1 n
'
1 ]1,~
1 -- 1 -- ~ N
1 - exp( - A n c) q e x p ( A n c) q [ e x p ( A n c) - 1]
1 - e x p ( - A N c) 1 - q
I nn
I( ]
1 - 1 ~ln n q
1 - 1 - lln N
1 - n -a qn A q [n 4 - 1]
1 - N -A 1 -q
1 - [1 +nO] ~'B q q [( 1 ~,.B
1 - [1 + No] ''B [1 + n'l ''~ l - q L\I--~nU - 1/
thermoplastics were measured with a fatigue-rated extensometer having a 2.5-cm gauge

length.
In Ref 20, an lnstron type of machine (Shimazu Autograph IS-10 T) was used for the
low-cycle fatigue testing of short carbon fiber-reinforced nylon 6. The points of maximum
and minimum deformation were measured by reading the return points of the ball screw
rotation of the machine. Then both the resultant and permanent strains were calculated
from these results.
In Ref 21, residual elongations of unidirectional composites were measured from the
output of the linear variable differential transformer (LVDT) of the MTS machine. This
deformation was then converted to strain (permanent) within the gauge length by using a
calibration curve established separately.
We tried to measure the resultant and permanent strains with a fatigue-rated extensometer
having a 25.4-mm gauge length. But the dynamic loading effect and surface damage de-
veloped in the early stage of fatigue cause unnecessary movement of the extensometer and
prevent strain measurement. Instead, the resultant and permanent displacements were mea-
sured from the output of the LVDT of the MTS machine. The continual increase of both
resultant and permanent displacements with number of cycles was observed, suggesting
possible relation between the resultant strain and physical damage. However, the observation
results are not presented in this paper because of the difficulty in measuring strain exactly.

Tensile Strength and Single-Stress Level Fatigue Test
All test results are summarized in Table 3. The ultimate strength and elastic modulus
were found to be 770 MPa and 33.1 GPa, respectively. The static stress-strain behavior
shows almost linear nature until failure. The minimum tensile strength was slightly over
90% of its mean, and no specimen failed in its first fatigue loading at the stress level 0.9.
Two extreme scattering cases were observed in the single-stress level fatigue test. At stress
level 0.85 the standard deviation is larger than mean fatigue life. And the minimum fatigue
life at 0.75 is smaller than that at stress level 0.8. However, these two experimental results
are not too surprising in fatigue of composite materials.
Single-stress level fatigue life predictions and experimental data are compared in Fig. 5
and Table 4. The results show that FLPE 1, 2, and 5 predict fatigue life very well for this
particular system. The constants of each equation are estimated by the following least squares
methods:
Linear least squares method: Conventional S - N curve, Basquin's relation, FLPE 2, 4,

and 6
Nonlinear least squares method: FLPE 1, 5, and 7
Nonlinear weighted least squares method: FLPE 3 (weight = 1/An)
The results are presented in Table 5.
Two-Stress Level Fatigue test

An extreme scattering is observed in high-low (0.85-0.7) test. Five of the ten specimens
failed before 500 fatigue cycles at the first stress level, 0.85. However, once a specimen
survived 500 fatigue cycles at that applied stress level, it has certain amounts of fatigue life
at the second stress level, 0.7.
Two-stress level fatigue life was predicted by proposed cumulative damage models. For
single-stress level fatigue life, the experimental data were used. The poor single-stress level
fatigue life prediction by FLPE 6 and 7 prevents those cumulative damage models based
TABLE 3--Test summary.
Applied Arithmetic Standard Number of

Stress Mean Minimum Maximum Deviation Specimens
Static test (MPa) 770 695 860 48.9 21

Single-Stress Level Fatigue Test
0.9 513 47 1 243 346 10
0.85 3 694 359 15 276 4 322 12
0.8 10 922 4 044 17 751 4 194 10
0.75 27 577 3 324 70 584 21 382 10
0.7 47 739 17 439 113 539 24 642 10
Two-Stress Level Fatigue Test
0.7-0.85 (nl = 10 000) 2 031 105 10 179 2 983 10
0.85-0.7 (n~ = 500) 33 049 28 200 37 458 3 012 10
0.75-0.85 (nz = 5000) 3 240 157 11 401 2 986 16
0.85-0.75 (nl -- 500) 21 450 3 514 48 190 14 742 11
9 EXPERIMENTAL DATA
- - S- N CURVE
0.95
---- BASQUIN'S RELATION
..... PROPOSED FLPE 1
0.9
0.85
0.8
0.75
0.7
0.. I I I I I
2.5 3 3.5 4 4.5 5
Log N
FATIGUE LIFE
FIG. 5--Single-stress level fatigue test data and predictions.
TABLE 4---Single-stress level fatigue life comparison with experimental data.
Applied stress 0.9 0.85 0.8 0.75 0.7

Experimental data 513 3 694 10 922 27 577 47 739
Conventional S - N curve 841 2 546 7 710 23 340 70 664
Basquin's relation 920 2 492 7 200 22 222 74 150
FLPE 1 519 3 504 11 712 26 692 47 855
FLPE 2 586 3 138 10 319 25 978 55 232
FLPE 3 670 2 008 5 928 18 221 59 867
FLPE 4 608 3 058 9 968 25 694 57 257
FLPE 5 535 3 348 11 772 27 332 47 350
FLPE 6 47 377 3 446 36 352 450 900
FLPE 7 46 374 3 420 36 000 445 672
TABLE 5--Estimation of constants.
M A B C
FLPE 1 7.6845 4.4023 0.167 60

FLPE 2 46.675 0.241 70
FLPE 3 - 131.40 - 17.169
FLPE 4 0.018 87 0.268 28
FLPE 5 11.588 7.4169
FLPE 6 0.027 40
FLPE 7 -56.561 1.550 9
Conventional S-N curve k: -0.103 95 d: 1.2058

Basquin's relation r 1065.7 b: -0.057 14
o n d e g r a d a t i o n types 6 a n d 7 f r o m predicting two-stress level fatigue life. T h e p r e d i c t i o n

results are p r e s e n t e d in T a b l e 6. ( T h r o u g h o u t Tables 6a to 6d the e x p e r i m e n t a l d a t a a n d
p r e d i c t i o n s are failure cycles at the s e c o n d stress level after fatigued at the first stress level.)
T h e c o m p a r i s o n s h o w e d t h a t d a m a g e m o d e l type I predicts fatigue life well in l o w - h i g h
tests, while d a m a g e m o d e l type III predicts well in h i g h - l o w tests. O v e r a l l , it c o u l d be
c o n c l u d e d that the d a m a g e m o d e l III-I is best to use.
TABLE 6a--Two-stress level fatigue life prediction (low-high test).

ql = 70% r q2 = 85% tr.
n~ = 10 000
Experimental data: 2031
Damage Model Type

Degradation
Type I II III IV
1 3240 3694 3401 2921

2 2921 3692 3269
3 2173 3465 2926 2921
4 2834 3688 3229
5 3205 3690 3365 2573
TABLE 6b--Two-stress level fatigue life prediction (high-low test).

ql -- 85% tr,, q2 = 70% tr.
nl = 500
Experimental data: 33 049
Damage Model Type

Degradation
Type I II IIl IV
1 36 938 17 967 33 312 41 278

2 41 278 19 214 36 208
3 46 031 30 168 43 114 41 278
4 42 091 19 818 37 039
5 34 529 17 559 33 512 44 274
TABLE 6c--Two-stress level fatigue life prediction (low-high test).

ql = 75% tr,, q2 = 85% tr,
nl = 5000
Experimental data: 3240
Damage Model Type

Degradation
Type I II III IV
1 3327 3678 3326 3024

2 3024 3637 3188
3 2446 3227 2817 3024
4 2966 3618 3146
5 3152 3648 3283 2757
TABLE 6d----Two-stress level fatigue life prediction (high-low test).

q~ = 85% tr,, q2 = 75% tr,
nl = 500
Experimental data: 21 450
Damage Model Type

Degradation
Type I II III IV
1 22 269 14 632 21 206 23 844

2 23 844 16 701 22 630
3 26 237 22 293 25 428 23 844
4 24 188 17 268 22 992
5 22 971 14 685 21 553 25 288
Conclusions
Based on the fatigue modulus concept, a new approach has been suggested to predict
fatigue life of composite materials. Seven fatigue life prediction equations have been for-
mulated by fatigue modulus degradation models. Four basic damage models have been
defined by fatigue modulus and resultant strains. Twenty-three cumulative damage models
have been derived as a function of number of fatigue cycles, fatigue life, and applied stress
by the basic damage models and fatigue modulus degradation equations.
In this study, two possible ways (by finding unknown physical variables which could explain
fatigue phenomena, or by generalizing the existing strength and modulus degradation model)
have been proposed for future work to predict fatigue life of composite materials.
The following conclusions can be made:
1. Fatigue modulus can be used for composite materials analysis.

2. Damage and fatigue life of composite materials are related to fatigue modulus.
3. The relation between physical damage and fatigue modulus or resultant strain should
be investigated.
References
[1] Broutman, I. J. and Sahu, S., "A New Theory to Predict Cumulative Fatigue Damage in Fiberglass
Reinforced Plastics," in Composite Materials: Testing and Design (Second Conference), ASTM
STP 497, 1972, pp. 170-188.
[2] Hahn, H. T. and Kim, R. Y., "Proof Testing of Composite Materials," Journal of Composite
Materials, Vol. 9, 1975, pp. 297-311.
[3] Yang, J. N. and Liu, M. D., "'Residual Strength Degradation Model and Theory of Periodic Proof
Tests for Graphite/Epoxy Laminates," Journal of Composite Materials, Vol. 11, 1977, pp. 176-
203.
[4] Yang, J. N. and Jones, D. L., "Load Sequence Effects on the Fatigue of Unnotched Composite
Materials," in Fatigue of Fibrous Composite Materials, ASTM STP 723, 1981, pp. 213-232.
[5] Cbarewicz, A. and Daniel, I. M., "Damage Mechanisms and Accumulation in Graphite/Epoxy
Laminates," in Composite Materials: Fatigue and Fracture, ASTM STP 907, H. T. Hahn, Ed.,
1986, pp. 274-297.
[6] Reifsnider, K. L. and Stinchcomb, W. W., "A Critical-Element Model of the Residual Strength
and Life of Fatigue-Loaded Composite Coupons," in Composite Materials: Fatigue and Fracture,
ASTM STP 907, H. T. Hahn, Ed., 1986, pp. 298-313.
[7] Broutman, L. J. and Sahu, S., "Progressive Damage of a Glass Reinforced Plastic During Fatigue,"
Presented at the 24th Annual Reinforced Plastics/Composites Industry Conference, Society of
Plastics Industry, 1969, Sec. ll-D.
[8] Hahn, H. T. and Kim, R. Y., "Fatigue Behavior of Composite Laminates," Journal of Composite
Materials, Vol. 10, 1976, pp. 156-180.
[9] O'Brien, T. K. and Reifsnider, K. L., "Fatigue Damage Evaluation Through Stiffness Measure-
ments in Boron-Epoxy Laminates," Journal of Composite Materials, Vol. 15, 1981, pp. 55-70.
[10] Wang, S. S. and Chim, E. S.-M., "Fatigue Damage and Degradation in Random Short-Fiber SMC
Composite," Journal of Composite Materials, Vol. 17, 1983, pp. 114-134.
[11] Poursatip, A., Ashby, M. E A., and Beaumont, P. W. R., "The Fatigue Damage Mechanics of
a Carbon Fibre Composite Laminates: I--Development of the Model," Composites Science and
Technology, Vol. 25, 1986, pp. 193-218.
[12] Poursatip, A. and Beaumont, P. W. R., "The Fatigue Damage Mechanics of a Carbon Fibre
Composite Laminate: II--Life Prediction," Composites Science and Technology, Vol. 25, t986,
pp. 283-299.
[13] Hwang, W. and Han, K. S., "Fatigue of Composites--Fatigue Modulus Concept and Life Pre-
diction," Journal of Composite Materials, Vol. 20, 1986, pp. 154-165.
[14] Hwang, W., "Analysis of Fatigue of Composite Materials," M.S. thesis, State University of New
York at Buffalo, August 1984.
[15] Hwang, W. and Han, K. S., "Cumulative Damage Models and Multi-Stress Fatigue Life Predic-
tion," Journal o[ Composite Materials, Vol. 20, 1986, pp. 125-153.
[16] Leve, H. L., "Cumulative Damage Theories," in Metal Fatigue: Theory and Design, A. E Ma-
dayag, Ed., pp. 170-203.
[17] Hashin, Z. and Rotem, A., "A Cumulative Damage Theory of Fatigue Failure," Materials Science
and Engineering, Vol. 34, 1978, pp. 147-160.
[18] Hashin, Z., "Cumulative Damage Theory for Composite Materials: Residual Life and Residual
Strength Methods," Composites Science and Technology, Vol. 23, 1985, pp. 1-19.
[19] Mandell, J. E, Huang, D. D., and McGarry, E J., "Fatigue of Glass and Carbon Fiber Reinforced
Engineering Thermoplastics," Polymer Composites, Vol. 2, 1981, pp. 137-144.
[20] Jinen, E., "Accumulated Strain in Low Cycle Fatigue of Short Carbon-Fiber Reinforced Nylon
6," Journal of Materials Science, Vol. 21, 1986, pp. 435-443.
[21] Lorenzo, L. and Hahn, T. H., "Fatigue Failure Mechanisms in Unidirectional Composites," in
Composite Materials: Fatigue and Fracture, ASTM STP 907, H. T. Hahn, Ed., 1986, pp. 210-232.
Piet W. M. P e t e r s I
The Influence of Fiber, Matrix, and Interface

on Transverse Cracking in Carbon Fiber-
Reinforced Plastic Cross-Ply Laminates
REFERENCE: Peters, E W. M., "The Influence of Fiber, Matrix, and Interface on Transverse
Cracking in Carbon Fiber-Reinforced Plastic Cross-Ply Laminates," Composite Materials:
ABSTRACT: The transverse tensile strength of a unidirectional laminate is generally accepted

as a material property. The transverse strength of a single ply in angle-ply laminates is influ-
enced by the neighboring layers (effect of constraint) and is for this reason not a material
property. In the present investigation of four different carbon fiber-reinforced plastics (CFRP)
this influence of constraint and the effect of fiber, matrix, and interface on transverse cracking
are investigated. An intermediate ductile resin system proved to increase the transverse strength
significantly (if the fiber matrix bond is good) in comparison with a brittle resin system. A
highly ductile matrix (polyetheretherketone--PEEK), however, did not further or additionally
improve the transverse strength significantly. The intermediate ductile resin system proved to
be sensitive to the formation of voids under the cure conditions applied. The inhomogeneous
distribution of voids disturbed the expected constraining effect.
KEY WORDS: composite materials, transverse cracking, Weibull distribution, carbon fiber-
reinforced plastic (CFRP), interface, ductility
Crack formation in transverse plies of angle-ply laminates of fiber-reinforced plastics

is influenced by many parameters. Most important are properties of the fiber, properties
of the matrix, fiber volume fraction and fiber distribution, quality of the interface, defect
distribution, and constraining influence of neighboring layers. Most important fiber prop-
erties in relation to transverse cracking are the stiffness and thermal expansion coefficient
perpendicular to the fiber direction. Table 1 summarizes the properties of the most common
fibers [1 ].
The stiffness perpendicular to the fiber, being larger than the stiffness of the matrix, gives
rise to a strain magnification in the matrix if the laminate is loaded perpendicular to the
fiber direction. Based on an idealized square array of the fibers, Kies [2] developed a model
to determine the strain magnification factor as a function of fiber and matrix properties and
fiber volume fraction. The strain magnification factor (SMF) based on this model is given
by
SMF= 1/(1- (1-~-~) ~/4r~-') (1)
Scientist, DFVLR, Institut fiir Werkstoff-Forschung, 5000 K61n 90, West Germany.
103
Downloaded/printed
Copyright9 byby
TABLE 1--The elastic constants and thermal expansion coefficients ~;i and e~• of some
common fibers [1].
Fiber •l, GPa E~, GPa G,li, GPa a,, • 10-6/~ a• x 10-6/~
E-Glass 73 73 30 5 5
Aramid 133 5.4 12 - 4 59
C-Fiber (HT) 240 15 50 - 0.5 10
where
Vt --- the fiber volume fraction and

E,~,EI = the moduli of the matrix and of the fiber perpendicular to the fiber direction,
respectively.
The SMF as a function of the fiber volume fraction is presented in Fig. 1 for the three
most common types of fibers in a matrix with a tensile modulus of E,, = 4.0 GPa. If the
SMF is decisive for transverse cracking, the highest transverse strength can be expected for
aramid fiber-reinforced plastics (AFRP) and the lowest for glass fiber-reinforced plastics
(GFRP). G F R P did show the smallest transverse strength [3], however, carbon fiber-rein-
forced plastics (CFRP) showed the largest. Probably the lower fiber-matrix bond strength
of A F R P is the reason for this. Effectively, the transverse strength of single plies in angle-
ply A F R P laminates can be lower than for G F R P because the transverse straining capability
is largely consumed by the residual curing strain as a result of the high transverse thermal
expansion coefficient of the aramid fiber [3]. These results show that, before a realistic
prediction of the transverse strength of single plies in an angle-ply laminate can be made,
the influence of the parameters mentioned above should be investigated. This has hardly
been done for available commercial fiber-reinforced plastic systems. One of the reasons is
that it is almost impossible to vary one specific parameter without also changing the others.
18
u. 12
~
/
10
8.
4.
9169 ^ss
I~ 2'
e - 9 - , - , - , . , - i - , -
.0
FIBRE VOLUME
FIG. 1--The strain magnification factor SMF as a function of the fiber volume fraction for
the most common types of fibers according to the Kies model [2] (E., = 4.0 GPa).
PETERS ON CROSS-PLY CRACKING IN CARBON FIBER-REINFORCED PLASTICS 105
A recent investigation of the constraining effect of neighboring layers found that the
strength of a 90 ~ ply in an angle-ply laminate is higher than the transverse strength of a
unidirectional laminate [4,5]. The increase is small when the 90 ~ ply is thick and when the
stiffness of the neighboring layers is small. The increase is strongest for thin 90 ~ plies if they
are embedded between the stiff 0~ layers. This constraining effect not only results in an
increase in strength of the 90 ~ ply, but also in a decrease of the scatter of the strength
(increase of shape parameter) [6]. The strength distribution of 90 ~ plies can be found from
only a few specimens [3,6] by considering the specimen to consist of a chain of small
specimens (elements) all of which can fail. Thus, for each specimen as many data as cracks
occur are evaluated.
This technique is also applied in the present investigation, which deals with transverse
cracking in carbon fiber-reinforced plastics. Two-parameter Weibull distributions for the
respective 90 ~ plies are determined. From these strength distributions, the influence of fiber,
matrix, voids, and constraint on transverse cracking are determined qualitively.
Materials and Experiments

The materials, produced from prepregs of different suppliers, were cured under the con-
ditions presented in Table 2.
The laminates with epoxy matrix were produced by DFVLR-Braunschweig making use
of an autoclave. To obtain a fiat laminate surface, a 5-mm-thick aluminum plate was put
on top of the bleeder material (or in case of the no-bleed systems on top of the release
fabric) inside the vacuum bag. The CFRP laminates with the thermoplastic matrix were
produced by Messerschmitt-B61kow-Blohm, Miinchen. From the produced plates, which
were 16 mm wide and 200 mm long (in case of the IM6/APC2 laminates 250 mm long),
specimens were cut with a diamond wheel. The specimens were dried and stored in a
desiccator. The tests were performed in a displacement-controlled (2 mm/min) Instron
testing machine with mechanical grips. In the gripping area, the specimens with epoxy matrix
were reinforced with 40-mm-long aluminum alloy tabs, leaving a free length of 115 mm.
Specimens from the IM6/APC2 material had no tabs; emery paper was used to prevent
slipping in the grips. Cracking of the 90 ~ ply in the carbon-epoxy laminates was detected
with a piezoelectric transducer in connection with a computer, which stores the signal above
a certain threshold.
Cracking in the thermoplastic CFRP laminate cannot be detected with the piezoelectric
transducer because crack formation and growth are stable, which do not lead to a sudden
load drop. For this reason, crack formation was studied at the polished edge of some
TABLE 2--The autoclave cure cycle for the different CFRP material systems.
Cure Cycle Postcure
Cooling Pressure,
Prepreg Temp, ~ Time, hr Rate, ~ kPa Temp, ~ Time, hr
T300/914C 175 1 700 190 4

IM6/6376
no bleed 175 2.5 600
T800/6376
no bleed 175 2.5 600
IM6/APC2
thermoplastic 400 40 1000
specimens. These specimens were stepwise loaded and unloaded up to increasing levels of
strain, and the cracks in the gauge length l = 93 mm were counted.
Analysis
The method to determine the strength distribution of 90 ~ plies from a single specimen
will be discussed only briefly here. (For more complete information, see Refs 6 and 7.) The
occurrence of multiple cracking in the 90 ~ ply before the specimen finally fails makes it
reasonable to consider the 90 ~ ply as consisting of a chain of elements all of which can fail.
Crucial in this technique is the determination of the number of elements in a specimen. This
is based on the stress distribution near cracks in the 90 ~ ply. This stress distribution is
calculated with the aid of a shear lag analysis the element of which is given in Fig, 2. The
shear stress, which introduces again the stress in the broken 90 ~ ply, is active in the shear
transfer layer, a resin-rich area at the 0o/90 ~ ply interface.
The stress in the 90 ~ ply (ply 2) at a distance x from the crack was found to be
~2~ = ~2.= (1 - exp - ~/x) (2)
where
----" al g~ a2
and the elastic constants G, El, and E2 are the shear modulus of the shear transfer layer,
the modulus of the 0 ~ ply, and the modulus of the 90 ~ ply, respectively. The dimensions b,
a~, and a: are the thickness of the shear transfer layer and of the 0 ~ and 90 ~ plies, respectively.
The shear lag parameter G/b is the main difficulty in applying the shear lag model. Reifsnider
et al. [8] suggest b -- 0.0127 mm and G = 1.378 GPa for a graphite/epoxy. Highsmith et
al. [9] applied a moir4 interferometry method to determine the shear lag parameter G/b
experimentally. Scatter of the experimental data and their relatively insensitive dependence
on G/b make it difficult to determine the exact value of G/b. In the present investigation,
the shear transfer layer is suggested to be a resin-rich area of twice the fiber diameter
(b = 0.015 ram). Thus G is the shear modulus of the matrix Gin.
ul u2 ul
T T ill
]" Ill
,I l I=
Y~ - ~ ~ - -
ili il',
al =;: a2 -~ aI
FIG. 2--Element of the applied shear lag analysis.

The maximum shear stress, which occurs in the shear transfer layer at the crack tip
(x,y) = (0, --- a2/2) is
c ( e2.2
Tmax = " l"l-2glal] (4)
according to the applied shear lag analysis.

The undisturbed stress cr_,.~consists of a thermal and a mechanical component. The thermal
component is calculated with the aid of
ZiT (a2 - oq) a, El E2

or2'= E2 a,_/2 + E1 a1 (5)
where ct2 and ct~ are the thermal expansion coefficients of the 90~ ply and 0~ ply, respectively,
and AT is the temperature difference between curing and testing temperature. For the
materials with unknown thermal expansion coefficients, the term AT(a: - tx~) was deter-
mined from the curvature of unbalanced 02/902 laminates making use of the theory of a
bimetallic strip [10].
Close to an existing crack there will be no more crack formation because the stress here
is too low. This leads to the more or less regular crack spacing, which is often observed in
laminates with transverse plies. Thus the element length should be smaller than the crack
spacing. In that case, the condition that of the N existing elements n < N breaks only once,
which is necessary for a statistical evaluation of the occurring cracks, would be satisfied.
Another condition is that the stress in elements with broken neighbors can be assumed
undisturbed. This, however, results in a too large element which fractures more than once.
For this reason the element length is somewhat arbitrarily chosen as double the length in
which 90% of the stress in the 90 ~ ply is introduced again. This length follows from Eq 2
after substituting tr2~ = 0.9 ch.e. In most cases the element length is sufficiently short because
the number of cracks n is smaller than the number of elements N. In glass/epoxy laminates
the crack density is so high that the element length had to be reduced to half its size [3].
The next step in determining the strength distribution from the data of one specimen is
to measure the fracture strain of every occurring crack. This is done with the piezoelectric
transducer, which measures the load drop caused by sudden crack formation. If the crack
formation is a slow propagating process, the piezoelectric transducer fails to detect cracking;
in these cases cracks are counted at certain values of applied strain at a polished specimen
edge. Results from these experiments are fracture strain data of every occurring crack or
of crack j ( j < N) listed in increasing order of strength. The probability of failure F for the
occurrence of crack j now is calculated with
J
F - N + 1 (6)
where N is the total number of elements in the specimen. Now the assumption is made that
the strength distribution can be described by a two-parameter Weibull distribution:
F = 1 - exp - (~S.,o/~f.to)~ (7)

where
~-u0 = the characteristic fracture strain and

a = the shape parameter.
F r o m E q 7 it is clear that the two-parameter Weibull distribution can be presented as a

straight line if In ( - In(1 - F)) is plotted as a function of In (~f.to). The experimental data
are plotted this way, and a best two-parameter Weibull function is d e t e r m i n e d making use
of linear regression technique and least square analysis.
To compare the strength distribution of the 90 ~ ply elements of a different length, the
strength distribution is transformed to a distribution for elements with a length of l = 1
ram. This is done with the equation
~s v:/~ ,,, = ( v , / ~ ) TM (8)

which describes the influence of volume (V~, I,'2) on the characteristic strength ~ . This
equation can also be used to predict the influence of the ply thickness on the strength
distribution, assuming that the shape p a r a m e t e r which describes the distribution of defects
in the thickness direction a, is equal to the shape parameter which describes the defect
distribution in length direction a (a, = a).
Results
T300/914C
The results on this material were published before [6,7]; thus, only m a j o r findings are
reported here. D a t a and test results are presented in Table 3. The thermal strain was
calculated with ~ = 0.8 x 10-6/~ and c~2 = 28.9 • 10-6/~ for the 0 ~ and 90 ~ plies, re-
spectively [6]. Making use of E q 2 and E q 3, the element length was calculated based on
the elastic constants El = 128 GPa, E2 = 9.5 GPa, Gm = 1.48 G P a , and on a thickness of
the shear transfer layer of b = 0.015 mm.
The tests on four specimens per laminate showed relatively low scatter, so the results
TABLE 3--The thickness of the different plies, thermal strain in the 90~ ply ~?, the element length 1,,
the Weibull shape parameter a, and the characteristic fracture strain ~ (element length = I ram),
determined from (n) specimens of CFRP materials.
a~, mm a2, mm e2', % l,, mm a ~r, %
T300/914C
02/90J02 0.263 0.525 0.448 0.69 10.47 1.666 (4)
02/906/02 0.254 0.762 0.433 0.83 7.55 1.525 (4)
0J90J02 0.250 1.549 0.394 1.12 6.41 1.367 (4)
IM6/6376
0/904/0 0.134 0.535 0.492 0.715 11.41 1.629 (3)
0/90o/0 0.131 0.787 0.467 0.845 13.45 1.636 (2)
019012/0 0.133 1.598 0.407 1.124 7.75 1.507 (4)
T800/6376
0/904/0 0.140 0.559 0.509 0.707 11.71 1.814 (2)
0/906/0 0.139 0.788 0.489 0.823 15.81 1.867 (2)
0/9012/0 0.139 1.667 0.425 1.112 11~ 1.842 (6)
IM6/APC2
0/90~J0 0.130 1.565 0.641 1.213 5.72 2.327 (6) b
0/90tJ0 0.130 1.565 0.641 1.213 6.24 2.487 (6) c
~ Estimated value of a.
b Crack length c -> 0.3 mm.
cc = 1.57mm.
2.0
0.99 ~ ' . 1.5
0.90 cFr'p ,'~,'" 1.0
" T300/914C .5
,,,
o~
0.7o L =1 mm .0
'::3
J 0.50
~,, 0,30 /..'7 -! .0

Z "2
'~ o.20 --~ .s
LU
d
uu -2,0
,, 0.~0 90~; 984 _1
I
o --2.5
..J
.~j 0.05 at ~ , . --3.0
ar
o.ol
?"
7.Ss
e
=1o.47
--3.5
-'"
--4.s
J p
" " -5.0
0.005 /1 t" "-5.5
t 9
e" I -6.0
0.6 0:0 1:o ,:s 2
ELEMENT FRACTURE STRAIN,
FIG. 3--The probability of failure as a function of the strain for 90~ ply elements of unit
length of the investigated T300/914C system. The dashed lines represent a prediction of the
thickness effect with at = 6.34 and a, = 4.59, respectively, which forces the Weibull strength
distribution to intersect the experimental distribution at the characteristic fracture strain ~
(indicated with a plus sign).
clearly fall into three groups (for the three laminates). The mean WeibuU failure strain
distributions for the 90 ~ plies are indicated in Fig. 3. Thinner 90~ plies show a larger fracture
strain (larger than can be predicted with Weibull theory, for example, with a, - a = 6.3)
and a larger Weibull shape parameter. These are the two aforementioned effects of con-
straint.
T800/6376
The thermal strain in the system T800/6376 was determined with the aid of an unbalanced
laminate. The element length was determined with E, = 156.4 GPa, E~ = 8.62 GPa,
G m = 1.38 GPa (calculated from Em = 3.6 GPa [11]), and the thickness of the shear transfer
layer b = 0.015 mm. The results for the different 90 ~ plies are indicated in Table 3. The
test results of three to four specimens per ply thickness are demonstrated in Fig. 4. It was
found that there was a large scatter in test results. For example, Fig. 4c shows cracking in
two specimens at low strain and cracking in four specimens at high strain. In the latter
specimens only one or two cracks developed before the specimens failed or explosive de-
lamination between 0 ~ and 90~ plies, caused by a transverse crack, occurred. Premature
failure was also found in some specimens with thinner 90~ plies, but the results are not
indicated in Figs. 4a and 4b. For sake of comparison the results of the best specimens (see
Table 3) of the different 90 ~ plies are deduced to an element length of l = 1 ram. The
strength of the 90~2 ply is based on the few number of cracks in the best four specimens and
LN(FRACTURE STRAIN)
I .2 .3 .4 .5 .6
0.80
LL 0.60
J
0.40 .fJ
,<
0.20
LU
0.10 ,8!
::i?:%# t I
--2.0 ~
o.os
0.02 9
(D ' --4.5
o 0.01
13-
1.1 1 .3 1 .5 1 .7
ELEMENT FRACTURE STRAIN, ~.
LN(FRACTURE STRAIN) LN( FRACTURE STRAIN)

.I .2 .3 .4 .5 .6 - .4 - . 2 . 0 . 2 .4
0.60
, . . , ..... ' ........ ' .... i , , . . J . . . h . . . . . . . . . . . .
.0
/ _ _ ..- - ~ z - - _ -
LL 0 . 4 0 -.5
T808/6/376 ~ ~ /
o 9o~ 0 ~ -I .0
J Le=0"82 mm l~r # Le=l .11 mm ~ f
7 o.20 - 1 .5
u.
t.u 0 . 1 0
•215
, +o:23.,~
~
.,~.~, ,I
j ? ,
/r -2.0
2
//
E
LU ,,--,,o:,, 92 # ' - t ,'~ § 18~t/6 -2.5
0.05
~,~,, o:, o. 3~/,~ - 3 . 0 ~z
i
d
-3.5
_~ 0.02 - 4 . 0
0.01
9- 4 . 5
, , I I I -5.0
1 .2 1.4 1.6 1.8 0.6 8.8 1.0 1.2 1.4
ELEMENT FRACTURE STRAIN, % ELEMENT FRACTURE STRAIN, ~.
b c
FIG. 4--The Weibull fracture strain distribution for the 90~ plies of different ply thickness
of the carbon-epoxy T800/6376 (three to four specimens per laminate).
on an assumed shape parameter of a = 11. Figure 5 shows results similar to those for T300/
914C specimens (that is, at decreasing ply thickness the strength and shape parameters
increase).
IM6/6376
The thermal strain in the system IM6/6376 was determined with the aid of an unbalanced
laminate. The element length was determined with Eq 2 and Eq 3 with the elastic constants
PETERS ON CROSS-PLY CRACKING IN C A R B O N FIBER-REINFORCED PLASTICS 11 1
Et = 159.5 GPa, E., = 9.24 GPa, G,, = 1.38 GPa (calculated from E,, = 3.6 GPa [11]),
and the thickness of the shear transfer layer b = 0.015 ram. The results for the different
90 ~ plies are indicated in Table 3.
A complete review of the results is given elsewhere [3], so only the deduced fracture
strain distribution of elements with a length of I = 1 mm are presented in Fig. 6. The results
are similar to those for the system T800/6376, except that the distributions are shifted to
lower values of fracture strain.
IM6 / A PC2
Transverse cracking in this carbon fiber-reinforced thermoplastic could not be detected
with a piezoelectric transducer because crack formation and propagation are stable processes.
For this reason the cracks were counted at the polished edge of six specimens. The number
of cracks larger than 0.3 mm and across the complete 90 ~ ply thickness were counted at
increasing levels of strain. The thermal strain was measured from the curvature of an
unbalanced 0z/902 laminate, and the element length was calculated with the aid of the
measured elastic constants E1 = 165.2 GPa, E2 = 9.35 GPa, and Gm = 1.23 GPa (calcu-
lated from E m = 3.19 GPa [12]). The results indicated in Table 3 show a large thermal strain
eJ = 0.641%, which is due to the high consolidation temperature. The distributions for the
two different 90~ ply crack lengths are indicated in Fig. 7. The large scatter in test results
causes the coefficient of correlation R of the linear regression curve to drop at R = 0.80,
whereas for the carbon-epoxy systems this coefficient normally measures R > 0.95 [6].
LN(FRACTURE STRAIN)
- .2 .0 .2 .4 .6
2.0
0.99 1.5
0.90
T800/6376 1.0
L=lmrn .5
" 0.70 I.
"' ,.o,4~ I .0
x 1 .842
= 050 /// I 5
u_ 0 , 3 0 -1 . 0
i -
0.20 -~ .5 't
-2.8 -~
= e.le
U.o -2.5
0.05 -3.0 ~
-3.5
~ 0.02 -4.0
O
O~ ct=l I 81 4 5
. "q : ":, : ,', :

, ,', : '
. , , , "
. - 6 0
.
0'.8 1 .0 I .2 1 .4 1 .6 1 .8 2
ELEMENT FRACTURE S T R A I N , ~,
FIG. 5 - - T h e mean Weibull fracture strain distribution o f 90~ply elements o[ different thick-
ness (element length 1 = 1 ram) o f carbon-epoxy T800/6376.
LN(FRACTURE STRAIN)
-.4 - .2 .0 .2 .4 .6
- 2.0
0.99 1.5
0.90
IM-6/6376 1.0
U. L=lmm .5
uJ 0 . 7 0 1 . 507.
rv .0
d 0.50
-,5
U-
0.30 -I .0
.-I .5
iii
.J
ILl 9
ii
O.lO
0 9
~ 0.05 9012/ 904// 90s 9
../
-3.5
II)
0.02 -4.0
0
ns
n 0.01 -4.5
-5.0
0.005
-5.5
" -6.0
0.6 0.8 1.0 1.2 1.4 1.6 1.8 2
ELEMENT FRACTURE STRAIN~ ~.
FIG. 6 - - The mean Weibull fracture strain distribution of 90~ply elements of different thick-
ness (element length I = 1 mm) of carbon-epoxy IM6/6376.
Comparison o f the Different Materials

It is difficult to compare the transverse strength of all investigated CFRP laminates because
of the different influences of constraint and defect distribution on the strength and the
limited number of experiments on the IM6/APC2 material. For this reason only the results
of the 0/90t2/0 laminates are compared. For the carbon-epoxy systems the mean 90 ~ ply
fracture strain (F = 0.5) of 115-mm-long specimens is calculated from the measured Weibull
fracture strain distributions. This procedure leads to a relatively low fracture strain value
for the thermoplastic material (~H = 1.131%), which could be too low because of the large
scatter in test results, which causes the strength distribution to be unreliable. For this reason
the transverse fracture strain was also measured on five unidirectional specimens with a
length of 115 mm. The mean fracture strain measured 1.301%. The mean 90 ~ ply fracture
strain is presented in Fig. 8 as a function of the fracture strain of the used matrix material.
The systems Fibredux 914C, Fibredux 6376, and polyetheretherketone (crystallinity 30%)
have a fracture strain of 1.4% [13], 3.1% [11], and 10% [12], respectively. The fracture
strain of the respective 90 ~ plies shows an increase if the fracture strain of the matrix increases.
The increase is large at a low matrix fracture strain. At larger matrix fracture strains,
however, the additional increase in 90 ~ ply fracture strain is only small. Further, the resin
Fibredux 6376 with T800 fibers shows a larger 90 ~ ply fracture strain than the same resin
with IM6 fibers. Also indicated in Fig. 7 are the thermal strains in the respective 90~2 plies.
PETERS ON CROSS-PLY CRACKING IN CARBON FIBER-REINFORCED PLASTICS 1 13
LN(FRACTURE STRAIN)
.2 .3 .4 .5 .G .7
i . . . . i . . . . I . . i . . . . i . . . . i . . . .
0.60 ' ' .0
u_ IM6/APC2 ~ ~ -.5
w 0.40
-, 0/9012 /0 ~ ~..l /
lo =I . 2 1 m m ~ .
o. o --,.5
z
LU
~,,~ ~ + - 2 . 0 u.
Z 0.10
J + -2.5
LU d
I
ou- 0 . 0 5 ~ ~ . + - - 3 . 0 _~
)-- ~ S J
--3.5
J 0.02
;"< /7 / ~6"~' -'~.0
m 7 9 4-+
0.01 -4.5
O_
I I I I -5.0
1 .2 1 ,4 1 .6 1 .8 2.0 2.2
ELEMENT FRACTURE STRAIN, ~.
FIG. 7 - - T h e Weibull fracture strain distribution for cracks with a length of c > 0,3 mm (*)
and for cracks across the complete 90~ ply ( + ) in the 1M6/A PC2 0/90~2/0 laminate.
1.4 1.4
1.3' 9 IMS -1.3
"\" 1 . 2 1.2
0I T800
n~1.1 I
1.1
1.0 ~ 1.0
% II .9
m ,9'
I
Z
.8
ou_ ,0 ~b IMS
z 07 .7 Ot~
k--
IM6 .6
o3
I,.- J
m .5 .5 E
.4
I I "1"
.3 I---
II I ~1 I
~#/'I 131 .2
ii
031 tr I .I
,1
I I
.0 .O
I 2 3 4 10
MATRIX FRACTURE STRAIN, %
FIG. 8 - - T h e mean fracture strain and thermal strain o f a 90t2 ply in a 0/901z/0 laminate
with different matrix systems as a function o f the fracture strain o f the matrix.
A major drawback of the thermoplastic material is the high thermal strain caused by the
consolidation at higher temperatures in comparison with the epoxy matrix systems.
Discussion
The clearest effect of constraint is demonstrated by the results of the system T300/914C.
At a decreasing 90 ~ ply thickness, the fracture strain as well as the shape parameter increase.
Characteristic for the constraining effect is that the increase in fracture strain is stronger
than can be predicted based on the assumption that the shape parameter which describes
the thickness effect is equal to the shape parameter which describes the length effect (a, = a).
Although the fibers T800 and IM6 in the resin system Fibredux 6376 show a tendency toward
effects of constraint, it was found that in this resin system the inhomogeneous defect dis-
tribution plays an important role. Premature cracking, causing the strength distribution to
be shifted to lower values of fracture strain, was found in only a few of the specimens of
the produced plates. Thus the scatter in test results mainly occurs from specimen to specimen
and not in the test results of every single specimen. If the defects are responsible for
premature cracking, it must be concluded that the defect distribution is 0~ oriented. This is
confirmed by C-scan pictures (Fig. 9) made of some remaining material of the 0/9012/0
laminates. The two T800/6376 specimens clearly show voids, which are concentrated in the
0~ direction. The C-scan graph of the IM6/6376 material shows the same effect, although
the defects are smaller. Further, the thickness of the produced plates was found to vary
little but similarly across different cross sections perpendicular to the 0 ~ surface layer, and
FIG. 9--C-scan graphs of remaining pieces of 0/90t2/0 laminates. (a) Two specimens suc-
cessively cut out of the T800/6376 plate. (b) A part of the laminate cut out of lMr/6376.
PETERS ON CROSS-PLY CRACKING IN CARBON FIBER-REINFORCED PLASTICS 1 15
the voids occurred near the thickest areas. This indicates that for the variation of thickness
mainly the 0 ~ surface plies are responsible. The locally denser arrangement of fibers (thick
rovings) causes a very high pressure at the respective locations under the aluminum plate,
but just next to these locations there is a low pressure. Thus, it is likely that the voids
concentrate in these low-pressure areas. The specimens that showed premature cracking in
many cases contained voids visible at the specimens' polished edges. Because of the super-
imposed effect of this inhomogeneous defect distribution, the constraining effect in the
system Fibredux 6376 (Figs. 5 and 6) is not as clear as for the system Fibredux 914C. The
investigated specimens of the latter resin system did not show any voids at the polished
edge. Thus, under similar curing conditions, Fibredux 6376 tends to void formation, but
Fibredux 914C does not.
Comparing the failure strain distribution of 90~ plies of T800/6376 and IM6/6376, it
becomes clear that the resin system with T800 fibers shows a higher fracture strain distribution
(Fig. 8). This indicates that the strength of the interface is larger for the T800 fiber. Two
more parameters, which also contribute to the difference in 90~ ply strength, have to be
mentioned: the fiber volume (which is slightly lower for T800/6376) and the shape and
surface condition of the fiber. The T800 fiber is kidney-shaped and rough, whereas the IM6
fiber is circular and smooth.
The influence of matrix fracture strain on cross-plied cracking is demonstrated clearly by
the different materials (Fig. 8). It is emphasized that the increase of matrix fracture strain
is realized for systems with a lower stiffness (thus with a larger quasi-linear-elastic strain
range) and a larger plastic deformation. Brittle resins like Fibredux 914C are sensitive to
defects. The fracture of the 115-ram-long specimen thus is governed by the biggest defect
present in the material. Although the systems T800/6376 and IM6/6376 proved to show
more defects, the fracture strain of the transverse ply increased because of the increased
ductility of the resin. If, however, the strength of I-ram-long elements is compared (Figs.
3, 5, and 6) in the range where the defects are small (% > ~), the difference in 90 ~ ply
strength is smaller. In this range, fracture is influenced more by the strength of the interface
or the ultimate strength of the matrix. The use of a very ductile plastic (polyetheretherketone)
in comparison with an intermediate ductile plastic does not lead to a considerable additional
increase of the strength of a 90~2 ply (see Fig. 8). This is because the polyetheretherketone
only shows the high fracture strain (ductility) under one-dimensional loading conditions.
The matrix in the 90~ plies is stressed multiaxially so that the large ductility (based on plastic
shear deformation) cannot take place. The variability in strength of the IM6/APC2 material
is strong (a = 6.24), which indicates that the strength of the material can be improved if
the defects are reduced in size and number.
Based on the test results, the following simplified model can explain transverse cracking.
The transverse fracture strain of a unidirectional laminate can be accepted as a material
property; the transverse fracture strain of a single ply in an angle-ply laminate, however,
cannot. This fracture strain depends on many parameters. The real material property is
the ideal transverse fracture strain in the absence of defects indicated in Fig. 10 as a vertical
line (a = oo). This fracture strain is caused by cracking of either the matrix or the interface,
whichever is weaker. In actual laminates, however, there is a certain defect distribution
(influenced by the quality of prepregging and laminate production) which causes the strength
distribution to be shifted to lower values and under a certain angle a. The lower part of the
line is dominated by the big defects; the higher part of the line can approach (dependent
on the defect distribution) the ideal fracture strain. Figure 10 shows how some of the
parameters studied influence the transverse fracture strain. The ideal transverse fracture
strain can be improved by strengthening the matrix or the interface, whichever is weaker.
A positive effect on transverse cracking is exerted by neighboring layers (constraining effect).
ii
(.L ~. ~
~' t J
-- U
In (Fracture strain)
FIG. lO--Schematic presentation of the ideal Weibull distribution (a = ~o) and the distri-
bution as influenced by the different parameters investigated.
This causes the strength distribution to be shifted to larger fracture strains and an increase
in shape parameter a. (The latter effect was not found in glass/epoxy laminates [3]). Finally,
the ductility of a matrix system shifts the distribution to higher fracture strains by reducing
the severity of the defects.
Conclusion
The transverse fracture strain of a ply in an angle-ply laminate is an independent material
property only in the absence of defects. In this case failure of the weaker of either the matrix
or the interface causes 90 ~ ply failure. In a practical laminate (with defects), the transverse
strength is influenced by the constraining effect due to neighboring layers, the defect dis-
tribution, and the ductility of the matrix. The results of the four investigated carbon fiber-
reinforced plastics indicate that
1. The use of an intermediate ductile system results in a clear increase in transverse

fracture strain in comparison with a brittle matrix. A very ductile matrix system only brings
a minor additional improvement.
2. All CFRP materials show the influence of constraint (that is, the increasing fracture
strain and shape parameter at decreasing 90 ~ ply thickness).
3. The intermediate ductile resin system tends to nonuniformly distribute void formation,
which results in cracking of the 90 ~ ply at low strains. This premature cracking disturbs the
constraining effect.
References
[1] Kleinholz, R., Aramidfasern, Kohlenstoffasern und Textilglasfasern nach Mass," Vol. 20, Offen-
tliche Jahrestagung der AVK Freudenstadt, Vol. 20, 1985, pp. 40-1-40-5.
[2] Kies, J. A., "Maximum Strains in the Resin of Fiberglass Composites," NRL Report 5752, AD-
274560, U.S. Naval Research Laboratory, Washington, DC, 1962.
[3] Peters, P. W. M. and Meusemann, H., "On Cross-ply Cracking in Graphite, Glass and Aramid
Fibre Reinforced Plastics," International Conference on Composite Materials VI, European Con-
ference on Composite Materials II, London, 1987, pp. 3.508-3.525.
[4] Flaggs, D. L. and Kural, M. H., "Transverse Lamina Strength in Graphite-Epoxy Laminates,"
Journal of Composite Materials, Vol. 16, March 1982, pp. 103-116.
[5] Parvizi, A., Garrett, K. W., and Bailey, J. E., "'Constrained Cracking in Glass Fibre-Reinforced
Epoxy Cross-ply Laminates," Journal of Materials Science, Vol. 13, 1978, pp. 195-201.
[6] Peters, P. W. M., "'The Strength Distribution of 90~ Plies in 0/90/0 Graphite-Epoxy Laminates,"
Journal of Composite Materials, Vol. 18, 1984, pp. 545-557.
[7] Peters, P. W. M., "'Constrained 90-Deg Ply Cracking in 0/90/0 and +-45/90/-7-45CFRP Laminates,"
in Composite Materials: Fatigue and Fracture, ASTM STP 907, H. T. Hahn, Ed., American
[8] Reifsnider, K. L., Hennecke, E. G., and Stinchcomb, W. W., "'Defect-Property Relationship in
Composite Materials, ASML-TR-76-81, Part IV, June 1979.
[9] Highsmith, A. L., Stinchcomb, W. W., and Reifsnider, K. L., "Stiffness Reduction Resulting from
Transverse Cracking in Fibre-Reinforced Composite Laminates," VPI-E-81.33, Department of
Engineering, Science and Mechanics, Virginia Polytechnic Institute and State University, Blacks-
burg, VA, November 1981.
[10] Jones, E R. and Mulheron, M., "The Effect of Moisture on the Expansion Behavior and Thermal
Strains in GRP," Composites, Vol. 14, No. 3, July 1983, pp. 281-287.
[1l] Ciba Geigy, 6376-Data Sheet FFA 140e, 1984.
[12] Seferis, J. C., "Polyetheretherketone (PEEK): A Matrix Semicrystalline Polymer for High Per-
formance Composites," in Proceedings, National Academy of Athens, January 1985.
[13] Ciba Geigy, Fibredux 914 C Data Sheet FTA 49c, 1979, Duxford, Cambridge, England.
E. Gail Guynn, ~ Walter L. Bradley, ~ and Wolf Elber 2
Micromechanics of Compression Failures in

Open Hole Composite Laminates
REFERENCE: Guynn, E. G., Bradley, W. L., and Elber, W., "Micromechanics of Compres-
sion Failures in Open Hole Composite Laminates," Composite Materials: Fatigue and Fracture,
Second Volume, ASTM STP 1012, Paul A. Lagace, Ed., American Society for Testing and
ABSTRACT: Compression failures in laminates with holes are often preceded by the devel-
opment of a damage zone which grows with increasing compressive load. This damage zone
is due to local fiber buckling and/or shear crippling. The local softening in the damage zone
has been modeled using a Dugdale-type approach. This paper compares this analytical model
to the actual development of the damage zone size as a function of load. Comparison of the
model predictions to the experimental observations indicates that the Dugdale model does not
accurately predict the load-damage zone size relationship in open hole composite specimens
loaded in compression. Hole size and resin ductility were the two primary variables used in
this investigation. Three hole diameters were used in specimens 2.54 cm (1 in.) wide containing
5.08 cm (2 in.) gauge length. Matrix ductility and stiffness were also varied by using matrices
of polyetheretherketone (PEEK) and Hexcel F155. Very brittle resins were not studied because
the damage zone development that precedes unstable damage extension is too small to record.
Several coupon tests were interrupted prior to catastrophic damage growth to allow for a more
careful examination of the fiber shear crippling and/or buckling. These observations were
accomplished using nondestructive examination (C-scan and X-ray) and scanning electron
microscopy. A preliminary sectioning study was conducted through the thickness of the damage
zone of an AS4/PEEK specimen; each section was observed in the scanning electron micro-
scope.
KEY WORDS: composite materials, compression, shear crippling, failure modes, Dugdale
model, graphite/epoxy, graphite/PEEK
Nomenclature
ct Constraint factor: ct = 1 for plane stress; c~ = 3 for plane strain
bk Dimensions for partially loaded crack, in centimetres (inches), k = 1,2
c Half crack length, in centimetres (inches)
d Half crack length plus compressive plastic zone width, in centimetres (inches)
F~ Boundary-correction factor for a circular hole, due to r e m o t e uniform stress
Fg Boundary-correction factor for a circular hole in an infinite plate, partially loaded
crack
F~ Boundary-correction factor for two symmetric cracks emanating from a circular hole
in a finite-width plate subjected to a uniform stress
Fg Boundary-correction factor for two symmetric cracks emanating from a circular hole
in a finite-width plate subjected to partial loading
Research assistant and professor, respectively, Department of Mechanical Engineering, Texas A&M
University, College Station, TX 77843.
2 Director, U.S. Army Aerostructures Directorate, Hampton, VA 23665.
118
Downloaded/printed
Copyright9 by
GUYNN ET AL. ON MICROMECHANICS OF COMPRESSION FAILURES 119
p Length of compressive crush or damage zone, in centimetres (inches)

r Hole radius, in centimetres (inches)
S Remotely applied stress, Pa (psi)
cro Compressive stress supported in crush zone, Pa (psi)
w Specimen half width, in centimetres (inches)
xk Section dimension for Y - Z plane, in centimetres (inches), k = 1 to 3
zk Section dimension for X - Y plane, in centimetres (inches), k = 1 to 3
Introduction
Compression failures in laminates with holes are often preceded by the development of
a damage zone which grows with increasing compressive load. The compressive failure mode
observed depends strongly on the lateral support provided by the matrix to the fiber during
loading. In the absence of strong lateral support, all fibers would fail by buckling. However,
as matrix stiffness increases, buckling is suppressed and the fibers then fail in shear. The
available data in the literature strongly suggest that shear crippling involving fiber kinking
is the most common compression failure mode in graphite/epoxy composites [1]. These
failures begin with the kinking of a few fibers; these kinked fibers disrupt the stability of
adjacent fibers so that these also kink. The damage propagates until the shear crippled zone
completely transverses the specimen. Subsequently, and immediately, a more catastrophic
failure occurs producing brooming and delamination.
In the compression tests of this investigation, a damage zone, similar to a fiat, tension
fatigue crack in metals, initiates at the edges of the hole and propagates across the width
of the specimen, leading to final failure. The length or size of this zone increases with
increasing compressive load. The damage zone is virtually symmetric (ignoring load and
specimen assymetry) about the hole, and it is initiated by local fiber buckling and/or shear
crippling in the edges of the hole. Figure 1 (from preliminary testing with Wolf Elber at
NASA Langley Research Center) shows the initiation and propagation of this damage zone
across the specimen's width. However, this type of damage zone has been observed only in
ductile material systems. It was observed by Guynn and Elber during preliminary testing
that stable growth of the shear crippling zone in brittle systems is very short and cannot be
detected optically prior to catastrophic failure. However, the cracklike appearance of the
compression failures in ductile composite laminates is very similar to the Dugdale/Barenblatt
plastic zone model [2]. The Damage Zone Model (DZM) [3,4], which is very similar to the
Dugdale/Barenblatt model, has recently been developed in parallel studies to predict ac-
curately static fracture of notched composite laminates loaded in tension. Consequently,
because of these similarities, the Dugdale model has been studied to evaluate its ability to
predict the response of ductile, open hole composite laminates loaded in compression. It
should be noted that this research and analysis follow Newman's Dugdale type analysis for
metals [5].
Analytical Model
The Dugdale model, which works well for metals, has been tested for applicability to
ductile open hole composite laminates loaded in compression. This model was selected
because these failures resemble a fiat, tension fatigue crack in metals except that the stress
directions are reversed. (An example of this composite damage zone was shown in Fig. 1.)
The damage zone (hereafter called the crush zone) at each edge of the hole has been
compared to the plastic zone in metals. We have assumed that although the fibers are broken,
FIG. 1--1nitiation and propagation of damage zone prior to compressive failure in an AS4~
PEEK specimen. Hole diameter is 0.16 cm (1/16 in.). Damage zone lengths are (a) 0.05 cm
(0.02 in.), (b) 0.08 cm (0.03 in.), (c) 0.13 cm (0.05 in.) and 0.30 cm (0.12 in.), and (d) 0.36
cm (0.14 in.) and 0.36 cm (0.14 in.).
the debris in the crush zone continues to carry load. Thus, one could assume that as the
crush zone grows across the specimen width, a constant crushing pressure ao would exist
over the length of the crush zone, p. The crush pressure, ~o, is completely analogous to the
constant tensile flow stress assumed to be present in the Dugdale plastic zone of a specimen
containing a crack and loaded in tension.
Newman [5] applied the Dugdale approach to homogeneous, isotropic materials containing
a crack in a finite plate and also to cracks emanating from a circular hole in a finite plate.
His result, which is derived in detail in Ref 5, for cracks emanating from a circular hole in
a finite width specimen (see Fig. 2) is as follows
S F'h F ~ - a t r , , [ 1 - -~r2sin-' ( c / d ) ] F g F~" = 0 (i)
For a given S, w, r, c, et, and Cro, Eq 1 is solved for d using an iterative technique. However,
p = d - c, and thus the plastic zone size p is determined.
b2 ,
I
$
FIG. 2--Geometryfor Newman'sanalysis.
This work has followed Newman's analysis for cracks emanating from a circular hole in
a finite-width specimen, but the notched composite laminates were modeled with c = r.
However, instead of using an iterative solution, curves of S versus p have been generated
with (to as a variable parameter for the various hole sizes. The rearranged solution for Eq
1 is
acr~ 1 - -~r2sin-~ (c/d)]F~ F~

S = F~ F~ (2/
where
p=d-c.
The specimen configuration for this solution is shown in Fig. 3; the generated curves will
be presented later with experimental data in the section entitled "Results".
Compression failures in notched composite laminates are preceded by the development
of a damage zone which grows with increasing compressive load. This damage zone is
adjacent ot the center hole and is initiated by local fiber buckling and/or shear crippling
into the edges of the hole. As the compressive load increases, the damage zone progresses
across the width of the specimen to final failure. Starnes and Williams [6] compared two
graphite/epoxy systems and concluded that suppressing delamination (by either a delami-
nation-resistant resin or by transverse stitching) will raise the laminate failure strain to a
level at which this shear crippling failure mechanism becomes critical. Although Starnes and
Williams observed and reported the final compressive failure modes for these systems, the
details of the damage progression and development prior to catastrophic failure remain
unclear. The purposes of this research program were: (1) to determine the remotely applied
a ~ aO
t t t-f t- t
$
FIG. 3--Specimen configuration for the Dugdale model applied to open hole composite
laminates loaded in compression.
compressive stress, S, versus damage zone size, p, for these laminates; (2) to compare the
experimental data with the Dugdale type model; and (3) to determine the extent and nature
of the subsurface damage in the fiber buckling zone by a preliminary sectioning study through
the thickness of a damaged specimen.
Materials
Two composite material systems, AS4/PEEK and T2C145/F155, were investigated in this
experimental study. AS4/PEEK consists of high-strain carbon fibers (AS4) embedded in a
thermoplastic matrix known as polyetheretherketone (PEEK). PEEK is an aromatic poly-
ether resin which is primarily useful because of its exceptional toughness as well as chemical
resistance. PEEK became commercially available in 1981 [7] and is sold under the trade
name Victrex PEEK by Imperial Chemical Industries, Limited (ICI), United Kingdom. A
48-ply composite laminate of AS4/PEEK with a stacking sequence of [+-45/0J +45/02/-+45/
0/90]~ was investigated. This material was supplied by NASA Langley Research Center,
Hampton, VA, who obtained the laminate from ICI.
The T2C145/F155 prepreg, manufactured by Hexcel Corporation, San Francisco, CA,
consisted of graphite fibers embedded in a medium cross-linked density, medium toughness
epoxy with 6% elastomer (rubber particles) addition. The laminate of T2C145/F155 for this
study was manufactured at Texas A&M University. It contained 32 plies with a stacking
sequence of [(0/-+45/90)s]4. The 0 ~ fibers were placed in the surface ply to provide oppor-
tunity to directly observe the shear crippling process.
Methods
Compression tests were conducted in a specially designed high axial alignment Material
Test System (MTS) machine in the Materials and Structures Laboratory of Texas A & M
University (Fig. 4). The specimens tested were loaded in compression to failure in the
servocontrolled hydraulic test stand under displacement control at a relatively slow rate of
0.0025 cm/min (0.001 in./min) to provide more stable growth of the shear crippling zone.
The surface damage development adjacent to the center hole preceding failure was monitored
using a Wild M8 Zoom Stereomicroscope (shown in Fig. 4) equipped with a video system
for real-time recording. Prior to catastrophic failure, several tests were interrupted to provide
specimens with significant damage for sectioning studies and nondestructive examination
(NDE) of the shear crippling zone.
Compression specimens were 2.54 cm (1.0 in.) wide with a 5.08 cm (2.0 in.) gauge length,
each containing a center hole (see Fig. 5a). The center holes were induced into the specimens
using an ultrasonic drill press. The ultrasonic drill press uses a diamond-impregnated core
bit which is water-cooled through the core. The holes were actually bored using an ultrasonic
up/down motion, rather than the traditional rotational drilling methods. Hole diameters
were 0.16 cm 0/16 in.), 0.32 cm (V8 in.), or 0.64 cm (1/4 in.).
Sectioning was performed on an A S 4 / P E E K specimen that was loaded in compression
until a shear crippling zone formed. This specimen was 0.61 cm (0.24 in.) thick with a center
hole 0.16 cm (Vl6 in.) in diameter. Figure 5 shows a detailed schematic of this specimen
configuration. In order to section through the damage zone, the center 0.61 cm (0.25 in.)
of length, containing the center hole and adjacent damage (Fig. 5b), was cut from the
FIG. 4--Compression test facility with stereomicroscope and video attachments.
4~
$ m
0
z
Ill o
0
0
//GRIP/.
~ AREA// 0
6o
m
~c
..... il i
m
x~
r"
12.7 cm
(5.0 in~
;>'~''P>L I
/~tA//1 (b)
l Lt.(1.0 i~
(a)
FIG. 5--Specimen configuration. (a) Compression testspecimen. (b) Specimenfor sectioning
investigation,
specimen using a Struers Accutum Precision Saw (diamond blade). This center section was
then cut along its vertical centerline (the Y axis) into two halves, designated L and R. For
handling purposes, each of these halves was mounted in plastic, cold-mount specimen stubs.
The left half (L) was sectioned across the width of the specimen from the edge of the hole
toward the edge of the specimen, examining the Y-Z plane at each section, while the right
half (R) was sectioned through the thickness of the laminate, examining the X - Y plane at
each section. Table 1 details the sectioning specifications for each specimen. It should be
noted that, for simplicity, the Z axis has been omitted from the sketches shown on Table
1. For specimen L, the amount of material removed by sectioning has been designated x~
and is referenced across the width of the specimen from the edge of the hole, as shown in
Table 1. For specimen R, the material removed by the sectioning techniques has been
designated z, and is measured from the surface of the laminate, as shown in Table 1. The
Struers Precision Saw was used for the major sectioning cuts, while the thinner cuts and
final polishing (preparation for scanning electron microscopy--SEM) were accomplished
using a microprocessor-controlled grinding and polishing unit also manufactured by Struers
(Abramin Automated Polishing Unit). Each sectioned surface of specimens L and R was
examined in a JEOL 25 Scanning Electron Microscope.
NDE of the shear crippling zone (prior to catastrophic failure) was accomplished using
ultrasonic C-scans and dye-penetrant enhanced x-radiography. To act as an enhancing agent,
a zinc iodide solution was applied to the surface damage to penetrate the connected interior
damage. The zinc iodide solution consists of 60 g of zinc iodide, 10 ml of water, 10 ml of
isopropyl alcohol, and 10 ml of Kodak Photo-Flo 600 to act as a wetting agent.
Results
Macroscopic Compression Behavior
The average open hole compression strengths measured for the two materials and three
hole sizes tested are listed in Table 2. It should be noted that only one data point was
available for the T2C145/F155 specimens containing 0.64-cm (1/4-in.) holes. Figure 6 shows
a typical load-displacement curve for these open hole compression-loaded specimens. This
particular curve is for an AS4/PEEK specimen containing a 0.16-cm (IA6-in.) hole, and these
TABLE 1--Sectioning specifications.

Specimen L Specimen R
Section Section Section Section
Number Depth, xa Number Depth, z~
1 0.00 mm (0.000 in.) 1 0.00 mm (0.000 in.)
2 0.10 mm (0.004 in.) 2 0.25 mm (0.010 in.)
3 0.76 mm (0.030 in.) 3 0.52 mm (0.020 in.)
tv tv
----~ X
9 D-X
TABLE 2--Open hole compression strengths for the three hole diameters.
Compression Strengths, MPa (ksi), for Hole Diameter of

0.64 cm (1/4in.) 0.32 cm (1/8in.) (0.16 cm (1/16in.)
AS4/PEEK 355.7 (51.6) 422.6 (61.3) 495.0 (71.8)

T2C145/F155 297.1 (43.1) ~ 311.6 (45.2) 387.4 (56.2)
" Indicates only one data point available.
data are representative of the specimen chosen for the preliminary sectioning study. Spe-
cifically, the nonlinearity of this load-displacement curve should be noted. Figure 7 shows
an enlarged view of the more nonlinear and unloading portion of the curve. This part of
the curve is associated with visible damage initiation (labeled "pop-in" on Fig. 7) and unloads
due to stable damage zone development. This specimen was unloaded prior to catastrophic
failure for additional studies of the shear crippling zone.
Comparison o f Analytical and Experimental Results

Using Eq 2 and as a parameter, arbitrarily assigning values to (to for crush zone stress,
theoretical curves of S versus P where generated, as shown in Fig. 8. Experimental data (S
0 0
I=
I
I
-22240 -5000 I
9176
-44480 -10000
io
-66720 -15000
-e8960 -20000
J ' ' ' ~ ,
-.06 -.05 -.04 -.03 -.02 -.01 0
Displacement (in.)
I I I I
-. 1 5 Z 4 -. 1 0 1 6 -.0508 0
nl sp lacement (cm)
FIG. 6--Typical load-displacement data for open hole, compression-loaded composites,
loaded to damage zone formation. Data are from an AS4/ PEEK specimen containing a O.16-
cm ( I/16-in.) diameter hole.
P20-2
66720 -15000
f"
i,f
,/
-16000
27800
m -I7000
z
"0 -11120
IO
0
-J -18000
-50040
-19000
-88960 -20000 ' ' -~J

-.06 -.055 -.05 -.045
Dt sp 1 ~cement. (in.)
I I I .,..--J
-. 1524 -. 1397 - . 127 -". 1 1 4 3
Dtspllcement (cm)
FIG. 7--Enlarged view of nonlinear portion of load-displacement data presented in Fig. 6.
versus p) from two tests (one from each material system) have been plotted along with the
analytical curves in Fig. 8. Figure 9 shows the progressive growth of the crush zone from
initiation to failure in each of these tests. Both samples contain a 0.32-cm (Vs-in.)-diameter
hole. It should be noted that the observed S versus p relationship is much flatter than the
predicted relationships, independent of the chosen crush zone stress, tro, used in the analysis.
Sectioning Studies
Because of the magnitude of work involved in a detailed sectioning investigation, one
AS4/PEEK specimen was sectioned to evaluate the worth of this technique. This specimen
was loaded in compression until a shear crippling zone formed. The necessary load level
was approximately 99% of the average open hole compression strength for that particular
material and hole diameter (see Table 2 for details). Table 1 summarizes the section depth
or amount of material removed from each specimen, for each section examined in the SEM.
The initial SEM examinations were of the actual damage zone in the edge or circumference
of the hole, specimen L (Y-Z plane), and on the laminate surface, specimen R (X-Y plane),
prior to any polishing. These micrographs are shown in Figs. 10 and 11. Figure 10a shows
the shear crippled zone through the thickness of the laminate, looking into the edge of the
hole of specimen L. Careful observation shows that the majority of damage through the
laminate thickness is buckled 0 ~ fibers into the hole with only very local ply delamination
present, as necessary to accommodate the shear crippling and fiber microbuckling. Figure
10b shows the broken 0 ~ fibers in the edge of the hole. Furthermore, this micrograph shows
100
So o - lO0 Icst (689 ~ s ) )
toO-
~o o . 90 kst (6ZO MPa)
~0 o - 80 ksl (SSI MPI)
i
,,~o m 70 kst '(483 MPa)
Oa
'~0o " 60 k s i (414 141)(I)
9 u; ~O 0 m 50 kst (345 141=il)
300-
:~0-
I00-
O- 0
0.0 0,11 o~ o~ o.+' o~
Cmmh zon,, hmgth, In.
t "" i
0.0 O.4 0:6 of. ,:o ,~
Cnaeh zone lengfh, m
(a)
\% - I00 ksl (689 MPa)~

~% - 90 kSl (b20 MPa)
~0 " BO ks$ (',52 MPI)
l tl - ~-o0 - 70 ksl (483 MPa)

\ % - 60 ksl (414 MPa)
~o o - ';0 ksl (J45 MPI)
Ji"
0 __ .
"
---r-
Chlmh zone kmgth, In.
I!_1 !
o'~ O.4
! 1
O,5
d, o~ 0.4 6 0.8 1.0 1.2

r ~ l*~ISlth, om
(b)
F I G . 8--Applied compressive stress versus crush zone size for 0.32-cm ( I/8-in.) hole. (a)
A S 4 / P E E K specimen. (b) T2C145/FI55 specimen.
that the fibers generally are failed in bending under compression loading (only a few pure
compressive shear failures in the fibers are present), but they remain partially adhered to
the matrix. However, occasional longitudinal splitting was also noted. Figure 11 shows the
surface damage seen on specimen R (X-Y plane). Figure l l a shows the 45~ surface ply of
the damaged laminate. For orientation, a portion of the hole is shown in the upper left-
corner of this micrograph. Although little damage is seen in Fig. 1 la, closer examination
(Fig. llb) shows that the matrix is severely deformed by the compression loading and some
45 ~ fibers are broken.
FIG. 9--1nitiation and propagation of damage zone prior to catastrophic failure.
FIG. lO--Shear crippled zone in the edge of the hole, specimen L (Y-Z plane). (a) Mi-
crograph details buckled 0~ fibers through the thickness of the laminate. (b) Micrograph of
multiply broken or crushed 0~ fibers.
FIG. 1l--Specimen surface surrounding hole, specimen R (X-Y plane). (a) Micrograph of
45~ surface ply. (b) Micrograph details multiply broken 45~fibers.
Sectioning Across the Laminate Width--Figures 12 and 13 show the interior damage behind
the edge of the hole ( Y - Z plane) of specimen L in sections 2 and 3, respectively. Section
2 examines the damage 0.10 mm (0.004 in.) from the edge of the hole, while section 3 is
0.76 mm (0.030 in.) from the hole edge. Figures 12a and 13a show a reduction of 0 ~ fiber
breakage as specimen L is sectioned across the laminate width. Figures 12b and 13b detail
the 0 ~ fiber damage in the outer eight plies of the laminate of sections 2 and 3, respectively.
Figure 12b shows some damage (the white, triangle-shaped zone) has disappeared by section
3 (shown in Fig. 13b). The size of this damage is reduced in each ply, and the amount of
damage through the laminate thickness also is reduced. The 0 ~ plies (indicated by the arrow
on Fig. 12b) closest to the laminate surface appear at this section only to be buckled into
the hole. However, section 3 shows the now visible shear crippling zone in these outer 0 ~
plies. These micrographs clearly illustrate the reduction of fiber damage as the sections get
farther from the hole edge.
Sectioning Through the Laminate Thickness--Through-the-thickness damage in sections

1, 2, and 3 is examined in Figs. 11, 14, and 15, respectively. Figures 14a and 14b show
section 2 of specimen R ( X - Y plane), 0.25 mm (0.0100 in.) from the laminate surface, at
the beginning of the first group of 0~ plies. Figure 14a shows relatively little damage near
the hole edge in the 0 ~ plies; however, careful examination into the depth of the hole (Fig.
FIG. 12--Section 2 [xk = 0.10 mm (0.004 in.)] of specimen L. (a) Micrograph details
reduction in buckled tT fibers through the thickness of the laminate. (b) Micrograph shows
deformed +-45~ plies (white, triangle-shaped region) in addition to buckled 0~ plies.
FIG. 13--Section 3 [Xk = 0.76 mm (0.030 in.)] of specimen L. (a) Micrograph shows
relatively few buckled 0~ plies through the laminate thickness. (b) Micrograph shows shear
crippled near surface 0~ plies.
14b) indicates fiber damage. Section 3 is polished to this damage, 0.52 mm (0.021 in.), as
shown in Fig. 15a. The shear crippling zone initiates at the right-hand leg of the discontinuity
enclosed in the highlight box of Fig. 15a, and it extends at an angle of approximately - 3 0 ~
across the width of the specimen 0.70 mm (0.028 in.). The small arrow on Fig. 15a indicates
the shear crippling zone. Figure 15b shows the entire thickness of the 0~ ply buckling into
the hole. Additionally, the fibers are embedded in the extremely deformed matrix. Figure
16, a higher magnification micrograph of the damage shown in Fig. 14b, shows that the
ductile matrix is severely deformed with the fibers completely embedded in the matrix.
FIG. 14--Section 2 [Zk = 0.25 mm (0.010 in.)] of specimen R. (a) Micrograph of O~ ply
showing no damage. (b) Micrograph details damage present further through the thickness.
FIG. 15--Section 3 [zk = 0.52 mm (0.021 in.)] of specimen R. (a) Micrograph shows 0~
fiber microbuckling O.7 mm (0.028 in. ) across the specimen width. (b) Micrograph details ply
microbuckling.
Nondestructive Examination ( N D E)
Figure 17a shows the visible surface damage in a T2C145/F155 laminate containing a 0.16-
cm (Vlr-in.)-diameter hole. This laminate was loaded to 84% of the average open hole
compression strength for this particular hole size and material (see Table 2 for details). This
photograph indicates the presence of a large shear crippling zone. However, the C-scan
(Fig. 17b) and X-ray (Fig. 17c) both indicate that no resolvable delamination is present
within the laminate and also that the visible surface damage is significantly larger than the
interior damage. As noted earlier, very local delamination must accompany shear crippling
to accommodate the large interlaminar shear strains that occur locally. However, these
microscopic, local delaminations apparently do not propagate (to become macroscopic de-
FIG. 16--Micrograph shows buckled 0~ plies embedded in the matrix.
FIG. 17--Nondestructive examination o f a crush zone in a T2C145/F155 specimen con-

taining a O.16-cm (z/l~-in.) hole. (a)Photograph. (b) Ultrasonic C-scan. (c)Enhanced x-ray.
laminations) until final compressive failure when large-scale brooming and/or delamination
occurs. Thus, NDE shows that the only damage is buckling of surface 0 ~ fibers.
Discussion
From the results presented in Fig. 8, it is obvious that the Dugdale model does not
accurately predict the load-damage size relationship of notched composite specimens loaded
in compression. Possible reasons that this model does not exactly fit the data are as follows:
1. The constitutive relationship used in the Dugdale model is not an accurate description
of material response in the crush zone.
2. The crush zone size is measured on the surface of the specimen, but, based on the
NDE observations, this size is not an accurate indication of the actual crush zone size through
the laminate.
Although a Dugdale type model was chosen for open hole composite laminates loaded
in compression because of apparent damage similarities, the preliminary sectioning study
has revealed many differences in the damage and its propagation. Tension cracks (plane
stress or plane strain conditions) in metals primarily form symmetrically through the thickness
of the material and then propagate uniformly. Consequently, the visible surface crack is
representative of the actual damage through the thickness of the material. Because of the
inhomogeneity of composite laminates, the compressive crush zone initiates in the weaker
part (includes both the weaker edge of the hole and the weaker surface of the specimen)
of the laminate. Once the damage zone has formed, that particular portion of the specimen
is further weakened, leading to a reduction in stability (and lowered stiffness) for that part
of the laminate. The measured crush zone size is not an accurate measurement of the damage
through the thickness of the laminate. Unfortunately, only visible surface damage may be
measured during the test. However, it should be recalled from Fig. 8 that the observed
shape of the S-p relationship is much flatter than the predicted relationships, independent
of the chosen crush zone stress, tro. Thus, the model's inaccuracies cannot be explained by
inaccuracies in the crush zone size determination from real-time measurements at the surface.
The assumed constitutive relationship must be incorrect.
The Dugdale type analysis used by Newman [5] assumes an elastic-perfectly plastic con-
stitutive relationship for the material, as shown in Fig. 18a. A more realistic constitutive
relationship is shown for comparison in Fig. 18b. This relationship assumes that fiber buckling
reduces the load-bearing capacity of the material in the crush zone, giving load redistribution
to material adjacent to this zone. Once the crush zone has grown across the width of the
specimen, then the load on the crush zone again increases because there is no longer adjacent
material of greater stiffness to which load can be redistributed. This larger load in the crush
zone will usually lead to initiation of multiple delaminations which propagate immediately,
giving brooming and catastrophic failure.
It is clear that the constitutive relationship given in Fig. 18a would allow a monotonic
increase in load with crush zone size, the crush zone being assumed to support a stress of
~o, where tro> ~ro(X) throughout the width of the specimen, as seen in Fig. 19a. However,
the constitutive relationship shown in Fig. 18b would lead to a stress distribution similar to
that shown in Fig. 19b, which does not necessarily allow the load (proportional to the
integrated area under the P versus x - r curve) to monotonically increase with increasing
crush zone size.
The results of this preliminary sectioning investigation show that sectioning provides a
aO
(a)
(b)
FIG. 18--Constitutive relationships. (a) Dugdale type elasticmperfectly plastic model. (b)
Suggested model for crush zone in compression-loaded composite laminates.
a~
o0
oF-- 0-~I
@
i
i
w
distance from edge of hole (x-r)
(a)
stance from edge of hole ~(x-r)
(b)
FIG. 19--Stress distribution across remaining ligament width. (a) Dugdale type analysis.
(b) Crush zone in compression-loaded composite laminates.
viable method for determining the extent of the damage zone through the laminate thickness.
Sectioning (Figs. 10 through 13) has shown that the majority of damage in compression-
loaded, open hole composite laminates is shear crippled 0 ~ fibers. No ply delamination was
observed, although delamination did occur subsequently when catastrophic failure occurred.
It is expected that the rotation and kinking of the damaged 0~ fibers eventually provide the
trigger and driving mechanism for delamination, which is very obvious in final failure.
Furthermore, detailed sectioning of a specimen containing a larger crush zone and thus
further damage extent may show the delamination initiation. Additionally, Figs. 10b, 15,
and 16 show that the fiber/matrix interface remains intact for this AS4/PEEK laminate.
Because of the high ductility of the PEEK resin, severe deformation of the matrix occurred
during the compressive loading. The good fiber/matrix adhesion allowed the load to be
efficiently transferred to the fibers and thus led to a shear failure of the fibers and devel-
opment of the damage zone. Additional sectioning and SEM observations are expected to
lead to a complete three-dimensional schematic of this damage zone. Furthermore, two
future sectioning investigations are expected to produce an interesting comparison of the
microdamage and failure mechanisms in these crush zones. One sectioning study will observe
sections through the thickness of a laminate containing a poor fiber/matrix interface, while
the other study will observe sections through the thickness of a laminate with a stacking
sequence providing 0 ~ fibers in the surface ply.
Conclusions
The following conclusions may be drawn from this work:
1. Growth of a stable crush zone resulting from shear crippling of 0 ~ fibers precedes
compressive failure in tough matrix composite laminates with holes.
2. The assumption of an elastic-perfectly plastic constitutive relationship for material in
the crush zone does not accurately predict the load-crush zone size relationship. Such a
relationship also indicates a greater amount of stable crush zone growth than observed in
practice.
3. Additional measurements of toad supported by the crush zone are needed to allow a
more realistic constitutive model to be determined, and thus, a more realistic prediction of
the load-crush zone size relationship.
Acknowledgments
The authors gratefully acknowledge the material, financial, and mentor-type support
provided by N A S A Langley Research Center under grant number NAG-I-659, monitored
by John Whitcomb.
References
[1] Hahn, H. T. and Williams, J. G., "Compression Failure Mechanismsin Unidirectional Composites,"
NASA Technical Memorandum 85834, NASA Langley Research Center, Hampton, VA, Aug.
1984.
[2] Barenblatt, G. I., "The Mathematical Theory of Equilibrium Cracks in Brittle Fracture," Advances
in Applied Mechanics, Vol. 7, 1960, pp. 55-129.
[3] Aronsson, C. G. and Backlund, J., "Tensile Fracture of Laminates with Cracks," Journal of
Composite Materials, Vol. 20, May 1986, pp. 287-307.
[4] Aronsson, C. G. and Backlund, J., "Tensile Fracture of Laminates with Holes," Journal of Com-
posite Materials, Vol. 20, May 1986, pp. 259-286.
[5] Newman, J. C., Jr., "A Nonlinear Fracture Mechanics Approach to the Growth of Small Cracks,"
in Behavior of Short Cracks on Airframe Components, AGARD Conference Proceedings, Paper
CP328, Sept. 1982, pp. 6-1 through 6-26.
[6] Starnes, J. H. and Williams, J. G., "Failure Characteristics of Graphite-Epoxy Structural Com-
ponents Loaded in Compression," Mechanics of Composite Materials--Recent Advances, Z. Hashin
and C. T. Herkovich, Eds., Pergamon Press, Elmsford, NY, 1982, pp. 283-306.
[7] Hartness, J. T., "Polyether Matrix Composites," SAMPE Quarterly, Jan. 1983, pp. 33-36.
Joseph E. Grady, 1 Christos C. Chamis, 1 and Robert A. Aiello ~
Dynamic Delamination Buckling in

Composite Laminates Under Impact
Loading: Computational Simulation
REFERENCE: Grady, J. E., Chamis, C. C., and Aiello, R. A., "Dynamic Delamination
Buckling in Composite Laminates Under Impact Loading: Computational Simulation," Com-
posite Materials: Fatigue and Fracture, Second Volume, ASTM STP 1012, Paul A. Lagace,
Ed., American Society for Testing and Materials, Philadelphia, 1989, pp. 137-149.
ABSTRACT: A unique dynamic delamination buckling and delamination propagation analysis

capability has been developed and incorporated into a finite-element computer program. This
capability consists of (1) a modification of the direct time integration solution sequence which
provides a new analysis algorithm that can be used to predict delamination buckling in a
laminate subjected to dynamic loading and (2) a new method of modeling the composite
laminate using plate bending elements and multipoint constraints. This computer program is
used to predict both impact-induced buckling in composite laminates with initial delaminations
and the strain energy release rate due to extension of the delamination. It is shown that
delaminations near the outer surface of a laminate are susceptible to local buckling and
buckling-induced delamination propagation when the laminate is subjected to transverse impact
loading. The capability now exists to predict the time at which the onset of dynamic delami-
nation buckling occurs, the dynamic buckling mode shape, and the dynamic delamination
strain energy release rate.
KEY WORDS: composite materials, delamination, buckling, impact, fracture, finite elements
Composite laminates are subject to delamination, which causes a loss of both stiffness
and strength. Delamination is generally induced by static, dynamic, or fatigue loading.
Delaminated sublaminates are particularly susceptible to dynamic local buckling when sub-
jected to impact loading, The prediction of impact-induced dynamic delamination buckling,
the topic of this paper, is necessary for evaluating the durability of many composite structures.
The delamination buckling phenomenon has been observed experimentally under static
and fatigue loading conditions [1-4], and several analytical methods have been proposed
to model this damage mechanism. One-dimensional beam models and fracture mechanics
approaches have been used to gauge the stability of delaminations in compressively loaded
laminates [5,6]. Finite-element approaches [7,8] are often used for these analyses.
In the course of earlier research, experimental observations of dynamic delamination
buckling in transversely impacted laminates were reported based on high-speed photography
of the delamination buckling sequence [9]. This work motivated the present development
of a finite-element analysis technique to predict the occurrence of impact-induced dynamic
buckling in laminates with delaminations. It has been shown that the mechanism causing
extension of a delamination depends on the location of that delamination within the impacted
Aerospace engineer, senior aerospace engineer, and aerospace engineer, respectively, National
Aeronautics and Space Administration, Lewis Research Center, Cleveland, OH 44135.
Copyright by ASTM Int'l (all rights reserved); Mon Jan

137 16 20:12:47 EST 2012
Downloaded/printed
Copyright9 by by
laminate [9,10]. A delamination along the midplane of a symmetrical laminate will initially
extend in a predominantly mode II fashion. Delaminations off the midplane (closer to the
outer surface) of a laminate, however, are susceptible to buckling caused by compressive
stress in the delaminated region.
This local buckling may then induce a mode I-dominated extension of the delamination.
In the latter case, the ability to predict the onset of dynamic delamination buckling is essential
to determine if extension of the delamination will occur when the laminate is subjected to
a given impact load. The objectives of this paper are (1) to outline the computational
procedure for dynamic delamination buckling and delamination propagation, and (2) to
present typical results of this procedure for a delaminated composite laminate under impact
loading.
Dynamic Buckling Analysis

To perform the dynamic delamination buckling analysis, the direct time integration so-
lution sequence in the finite-element program is altered so that a linear buckling analysis is
performed at each time step. The buckling analysis requires solutions of the eigenvalue
problem:
[[K] + X[Kol] {0} = 0 (1)

where
[K] = the structural stiffness matrix, and
[K~] = the stress stiffness matrix.
The formulation of these matrices for the N A S T R A N Quad-4 plate element used here is
given in Ref 11.
Each scalar eigenvalue satisfying Eq 1 physically represents the nondimensional ratio
trA
h-
P.
where
tr = the time-dependent compressive longitudinal stress in the delaminated sublaminate,
A = the cross-sectional area of the sublaminate, and
P, = the critical compressive load that will cause buckling of the sublaminate.
When an eigenvalue reaches the critical value of unity 0rA = P ) , buckling in that mode
occurs. The eigenvectors {0} associated with each eigenvalue are the corresponding dynamic
buckling mode shapes. Figure 1 shows the altered finite-element solution scheme in detail.
The altered finite-element solution procedure was implemented in Version 65A of MSC/
NASTRAN using NASTRAN DMAP alters. The complete modification is shown in the
NASTRAN DMAP alter sequence in the Appendix. A functional description of the altered
solution sequence is given in Table 1.
Finite-Element Modeling Procedure
Figure 2 is a schematic diagram of the beamlike unidirectional [0],,, graphite-epoxy lam-
inate with an initial 5.08-cm (2.0-in)Z-long delamination through the width and located
z Original measurements were in inches.

GRADY ET AL. ON DYNAMIC DELAMINATION BUCKLING UNDER IMPACT LOADING 139
I flccd input Date 1
Acaamb,,
1 StIffnocc. I
I MaCll Mctr cad
I.,t,t.on ..,,l~ Dynamic Duckling Algorithm
l
~orm Applied-I
Old VICtOr J
I Integrate Equations t _ ~ 1
Of Motion:
t=t+At
Recover Current(6)]
Dllpllcamlntl
l,,*o7~ Strnac I I
J h*'t'ffn''" "''r'" l - - ' l ''rt't'~ ['<0]~
_-I
Io.t.u, .,I. o.-....l
l
PlOt Deformed
Struclurcl Geometry
J
FIG. l--Dynamic buckling analysis flow chart.
TABLE 1--NASTRAN DMAP alter sequence (refer to Appendix for listing).
Address Module Function
5 LABEL Begin buckling analysis

9 MATMOD Strip displacement vector {~} from transient solution
matrix at each time step
10 EMG Generate element stress stiffening matrices k,,
11 EMA Assemble global stress stiffening matrix K.
12-22 Eliminate boundary conditions; reduce K.
24 ADD Multiply K. by ( - 1.0)
25 DPD Read eigenvalue extraction data from input
27 READ Eigenvalue extraction: solve [K + hK.] {~} = 0 for
eigenvalues and eigenvectors
29 OFP Format and print eigenvalues
30 SDR1 Recover dependent degrees of freedom for eigenvectors
31 SDR2 Prepare eigenvectors for output
32 OFP Format and print eigenvectors
37 REPT Return to address 5
140 SECOND
COMPOSITEMATERIALS
Force
890
lime t
200 )Js
z
"f
111,
local delamlnatlon
w/
|.2,4
-,1.-
I- 50.8 cm
-t
FIG. 2--1mpact specimen geometry and loading.
halfway between the beam midplane and outer surface. The finite-element model of this
specimen consisted of a uniform mesh of 800 four-node isoparametric plate-bending elements
(QUAD-4 elements in MSC/NASTRAN) arranged as shown schematically in Fig. 3. Four
elements were used in the thickness direction, so each finite element represents 5 plies of
the 20-ply unidirectional laminate. Although a relatively fine, uniform finite-element mesh
was used for simplicity in this example, a somewhat coarser mesh is allowable in regions
where high stress gradients are not anticipated.
To make the most efficient use of computer time, larger elements may be used in regions
remote from the delamination without sacrificing any accuracy.
The allowable displacements of adjoining nodal points in the thickness direction were
42 Element~--
1.27 cm
,I
~ ~0"064 ~ 1
l TypicalElement
ments
FIG. 3--Finite-elementmodel of impact specimen.
CopyrightbyASTMInt'l(allrightsreserved);MonJan1620:12:47EST2012
Downloaded/printedby
(PDVSALosTeques)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
GRADY ET AL. ON DYNAMIC DELAMtNATION BUCKLING UNDER IMPACT LOADING 141
constrained, using multipoint constraint equation sets, so as to satisfy simple beam bending
assumptions away from the delaminated region and in each of the sublaminates in the
delaminated region. These assumptions are
1. Plane sections remain plane.

2. No strain in the transverse (Z) direction.
The first constraint set restricts neighboring nodal points in the thickness direction to
deform along a straight line over the cross section (shear deformation is neglected) while
the beam midplane remains unstrained. The second constraint set prevents neighboring plies
from penetrating each other. These constraints are relaxed near the delaminated region so
that the displacements of the nodal points defining the delamination are allowed to deform
as the solution dictates, independent of the deformation of the neighboring plies. This allows
the delaminated region to separate from the main laminate when a local compression occurs
in that area, as shown in Fig. 4. The constraints used here are shown schematically in
Fig. 5.
Typical unidirectional graphite-epoxy material constants were used in the finite-element
analysis; E~ = 134.4 GPa (19.5 x 106 psi), E2 = 10.3 GPa (1.5 x 10~ psi), G~2 = 5.0 G P a
(0.725 x 106 psi), v~2 = v~3 = vz~ = 0.3, and p = 1580 kg/m 3 (1.49 x I0 -~ lb-sVin.4). The
one-direction (longitudinal) is assumed to be aligned with the fibers, while the two-direction
(transverse) is perpendicular to the fibers and in the plane of the laminate. The three-
direction is perpendicular to the plane of the laminate. A representative triangular loading,
shown in Fig. 2, is applied at the midpoint of the laminate to simulate transverse impact.
Both ends of the laminate are assumed to be rigidly supported.

Figure 6 shows the midplane displacement and bending stress response of a [0],0~ laminate
without an initial delamination. It is apparent that the boundaries (clamped ends) have no
effect on the structural response before 250 ~.s for the specimen geometry studied. Fur-
thermore, if a delamination existed in the region of high compressive flexural stress 5 to 8
cm (2 to 3 in.) from the impact point, local buckling of the sublaminate region may occur
in the 150- to 250-1~s interval.
FIG. 4--1mpact specimen and buckled sublaminate.
z,w .... Original

Deformed
y,v
uI
~x,u 4 5.., -~
U4 ~
Constraints imposed a t each l o n g i t u d i n a l station '~':
1. U n t f o m r o t a t i o n a t each cross section

g| 9 g " t - 1,2,3,4
2. Plane sections remain plane

u t - ut§ 1 - tO t - 1.2,3
3. No transverse s t r a i n (c z - O)
wt Z w t 9 1.2,3,4
FIG. 5--Exaggera~d deformation of fini~-e&ment model.
A 5.08-cm (2.0-in.) long, off-midplane delamination is now introduced into the finite-
element model (as shown in Fig. 4) by removing the constraints described between two
adjacent plies in the region from x = 5.08 to x = 10.16 cm (x = 2 to x = 4 in.) from the
impact point. The calculated dynamic displacement of the partially delaminated specimen
under the same loading is shown in Fig. 7. Only the displacement of the right half of the
laminate is shown in order to more clearly highlight the opening profile of the delamination.
Results of the simultaneous dynamic buckling analysis are shown in Fig. 8. The lowest
eigenvalue reaches the critical value of 1.0 approximately 190 txs after impact, for the impact
loading condition shown in Fig. 2. These results indicate that a first-mode buckling is likely
to occur approximately 190 p.s after the impact event begins for the given loading conditions.
The buckling may then induce extension of the delamination in one direction or the other.
Progressive extension of the initial delamination is modeled by removing the constraints
between neighboring plies at the appropriate nodal points. This approach is used here to
perform finite-element analyses of impacted laminates with successively increasing delam-
ination crack lengths. A simple technique is then used to calculate the variation of the
dynamic strain energy release rate with time during the interval on which the impact load
acts. Using the applied force, F(t), and the displacement, y(t), at the point of impact, the
amount of energy stored in the structure at any time t is
W(t) = ~'(') F(l~) d~

Jo
G)
FOtCB :D
o >
0
..<
J 8lPO
m
t
.r-
0
Z
CD
300 Z
.251
BENDING '3TRES L~S/'I so us
DISPLACEMENT ED
so ps m
E r-
........... 100
0 .... 150
150
.125 ~ - - - 200
- - - 200 r
~-- 250 O.
r ilfii til~
G) z
- -
f
' ~ ~ "Vh..
~
,'MI!
~ U
I
~t I
J'\
. ,..:1",
E c
Q 0 0
r
r-
(3
L_
m
O,
\,,..// c
O9 z
m -.125 \,.JI -150 m
o
0
.-I
-.25 | r-
. . . . . 300 0
25 50 25 5O o
z
Position (cm) Position (cm)
FIG. 6--Midplane displacement profiles and distribution of maximum bending stress in impact specimen due to a central impact.
G3
t~q
v~
4~
,\
o
u
tO
s
~ A
T |u
O
to .4~
IN
hi' tr
9-
w
c
O
| I 1 1 I
\ r,-
t~
I I" I"
(w:~) | u o w o o e l O s l O
. . . . Original
First B u c k ~ ' Deformed
t0
onset of Iocsl
nmlnntlon buckling
IOO I~0 2o0
l l U [ (UJCROS[CONDS)
FIG. 8--Calculation of dynamic buckling instability point.
This calculation is performed for several successive increments in crack length, each incre-
ment being equal to one element length in the finite-element model.
Defining
OW W (ao + Aa) - W (ao)

Gr = ~ ..... = AA
where
a = crack length, and
a = crack area
as the strain energy release rate, the variation of strain energy release rate with time at each
tip of the 5.08 cm (2.0 in.) long initial delamination is shown in Fig. 9. This indicates that,
in the absence of flexural wave reflections from the boundaries, the delamination will initially
extend toward the point of impact. Variation of the energy release rate as the initial crack
length is extended toward the point of impact is shown in Fig. 10. The apparent trend for
the magnitude of the energy release rate to decrease as the crack length is increased indicates
that the available strain energy for delamination extension will decrease monotonically,
leading to stable delamination growth and even arrest.
Relevance to Experimental Observations

The dynamic delamination buckling analysis procedure presented here was used to predict
the observed delamination buckling of the cantilevered specimen shown schematically in
Fig. 11, which was tested as described in Ref 9. The impact force history and material
properties used in the analysis are from Ref 9. Figure 11 shows the calculated variation of
1
I. II
20
%
15
\
i
rr
t--
t~
t_
10
5
iiii
tll
/0
'Tow[ (UtC:RosrCONDS)
FIG. 9--Strain energy release rates for extension of original delamination in opposite di-
rections.
the lowest eigenvalue with time and the associated buckling mode shape. The eigenvalue
behavior indicates a first-mode delamination buckling approximately 150 p.s from the be-
ginning of impact. This is in reasonable agreement with the photographic data in Ref 9,
which shows delamination buckling and initial crack extension occurring two to three frames
(125 to 187.5 txs) from the beginning of impact. Further experimental investigation is cur-
rently being conducted to provide additional verification and will be reported in the near
future.
Possible Extensions
The methods described here can be extended in a straightforward manner to analyze the
problem of dynamic buckling and extension of an arbitrarily shaped embedded delamination
in a laminated plate. This problem has been analyzed previously under static loading con-
ditions [7,12]. The analysis procedure for the plate will be similar to that used for the
beamlike laminate geometry used here, although the physical meaning of the buckling
eigenvalue will be slightly modified due to the more complex stress field. The procedure
used here to model the composite laminate allows progressive extension of the delamination
to be simulated by removing the appropriate constraints between displacements at neigh-
boring nodal points. This slightly increases the total number of degrees of freedom to be
solved for in the finite-element analysis, while keeping the number of elements and the
element mesh constant as the crack extends. Further application of this crack-extension
modeling procedure to analyze the growth of both transply cracks and edge cracks in com-
posite laminates is possible. Embodying these extensions into the computational procedure
will make it possible to assess damage tolerance and structural fracture in composites sub-
jected to impact loads.
Conclusions
A dynamic delamination buckling analysis procedure for composite laminates with de-
laminations has been developed and incorporated into a finite-element computer program.
This algorithm enables the modified finite element program to predict the time at which
delamination buckling occurs, and the dynamic buckling mode shape. Preliminary experi-
mental verification has supported the validity of the analysis.
The procedure is based on a unique use of multipoint constraints which permits modeling
of the laminate using layers of plate-bending elements. This procedure greatly simplifies the
simulation of progressive crack extension. It was used to calculate the dynamic delamination
strain energy release rate for a composite laminate under impact loading.
L
20
% a2
15 ...... = a0
10
#\-.,
9i// \
n
/,,./ -,,,
5 ;// \
,.../ ,,,
I..
c ! ,'__/f. ! !
IJJ
50 ~ IDO 150 20~
'l'lu E (u4~:ROSECONDS)
FIG. tO--Strain energy release rates for progressive extension of original detamination
toward the point o f impact.
first bucklln9 mode st 150 pler

3.3
9 2 r~/sec
10 [90/0~5s Graphit~l ENxu
o n s e t of IOcll
J ~ / delsmlnntlon buckling
~." _,
I
IO II~X) 1~ ZOO
Till: (I~J{:R~SI:COND$)
FIG. 11--Lowest buckling eigenvalue for cantilevered laminate with an initial delamination.
Results obtained show that (1) delaminations near the outer surface of a laminate that is
subjected to transverse impact loading are susceptible to dynamic buckling and buckling-
induced propagation, and (2) in the absence of boundary effects, transverse impact at a
point near a delamination will cause the delamination to propagate toward the point of
impact with little or no extension in the opposite direction.
Further, application of this procedure to the problem of buckling and propagation of an
embedded delamination in a laminated plate is possible9 In addition, similar techniques can
be used to model different flaw types such as transply cracks and edge cracks. With these
modifications it would be possible to computationally assess the damage tolerance of com-
posite structures with initial or induced defects when subjected to impact loadings.
References
[1] Gillespie, J. W. and Pipes, R. B., "Compressive Strength of Composite Laminates with lnterlam-
inar Defects," Composite Structures, Vol. 2, No. l, 1984, pp. 49-69.
[2] Rosenfeld, M. S. and Gause, L. W., "'Composite Fatigue Behavior of Graphite/Epoxy in the
Presence of Stress Raisers," in Fatigue of Fibrous Composite Materials. ASTM STP 723, American
[3} Konishi, D. Y. and Johnston, W. R., "Fatigue Effects on Delaminations and Strength Degradation
in Graphite/Epoxy Laminates," in Composite Materials: Testing and Design (Fifth Conference),
ASTM STP 674, S. W. Tsai, Ed., American Society for Testing and Materials, Philadelphia, 1979,
pp. 597-619.
[4] Chai, H., Knauss, W. G., and Babcock, C. D., "'Observation of Damage Growth in Compressively
Loaded Laminates," Experimental Mechanics, Vol. 23, Sept. 1983, pp. 329-337.
[5] Chai, H., Babcock, C. D., and Knauss, W. G., "One Dimensional Modeling of Failure in Lam-
inated Plates by Delamination Buckling," InternationalJournal Sofids and Structures, Vol. 17, No.
11, 1981, pp. 1069-1083.
[6] Whitcomb, J. D., "'StrainEnergy Release Rate Analysis of Cyclic Delamination Growth in Com-
pressively Loaded Laminates," in Effects of Defects in Composite Materials, ASTM STP 836,
[7] Whitcomb, J. D., "Finite Element Analysis of Instability-Related Delamination Growth," Journal
of Composite Materials, Vol. 15, Sept. 1981, pp. 403-426.
[8] Kapania, R. K. and Wolfe, D. R., "Delamination Buckling and Growth in Axially Loaded Beam-
Plates," AIAA paper 87-0878-CP, presented at the 28th AIAA/ASME/ASCE/AHS Structure,
Structural Dynamics, and Materials Conference, 6-8 April 1987, Monterey, CA.
[9] Grady, J. E. and Sun, C. T., "Dynamic Delamination Crack Propagation in a Graphite/Epoxy
Laminate," in Composite Materials: Fatigue and Fracture, ASTM STP 907, H. T. Hahn, Ed.,
[10] Simitses, G. J. and Sallam, S., "'Delamination Buckling and Growth of Flat Composite Structural
Elements," AFOSR Technical Report, Sept. 1984.
[11] The NASTRAN Theoretical Manual (Level 15.5), MacNeal, R. H., Ed., MacNeal-Schwendler
Corp., Los Angeles, 1972.
[12] Whitcomb, J. D. and Shivakumar, K. N., "Strain Energy Release Rate Analysis of a Laminate
with a Postbuckled Delamination,'" NASA TM-89091, Feb. 1987.
APPENDIX
Dynamic Buckling D M A P Alter Sequence for MSC/NASTRAN
ALTER 526
PARAML CASECC//DTI/I/71/V.N,TSET $
pARAML CASECCIIDTI/I/6//V,N,DEFSET $
PARAML UGV//TRAILER/I/V.N.NOCUGV $
PARAM /INOPIV,N,COLNUM=I $
5 LABEL RAALDOP $
PARAM //ADD/COLNUM/COLNUM/3 $
PARAM //LE/V,N,GETOUT/NOCUGV/COLNUM
COND QUITRAA,GETOUT
MATHOD UGV,,,,,/COLUGV,/I/COLNUM
I0 EtIG EST,CSTM,MPT,DIT,,COLUGV,ETT,EDT/KDELM,KDDICT . . . . /
I / O / O / / / / V ,N,TS E T/V ,N,DE FS E T/////////K 6 R OT
EMA GPECT,KDDICT,KDELM,BGPDT,SIL,CSTM/KDGG,/-I
EQUIV KDGG.KDNN/MPCF2 $
CQND LBL2D,MPCP2
MCE2 USET,GM,KDGG,,,/KDNN,,,
15 LABEL LBL2D $
EQUIV KDNN,KDFF/SINGLE
COND LBL3D,SII|GLE
SCEI USET,KDNN,,,/KDFF,KDFS,,,, $
LABEL LBL3D $
20 EQUIV KDFF,KDAA/OMIT
COND LBL5D,OMIT
SMP2 USET,GD,KDFF/KDAA $
LABEL LBL5D
ADD KDAA,/KDAAM/C,N,(-L.O,Q.O)/C,N,(O.O,Q.O] $
25 DPD DYNAIIICS,GPL,SIL,USET,,/GPLD,SILD,USETD. . . . . . . EED,EQDYN/V,N,
LUSET/V,N,LUSETD/V,N,NOTFL/V,N,NODLT/V,N,NOPSDL/V,N,NOFRL/ V,
N,NONLFT/V,N,NOTRL/S,N,NOEED/C,N,123/V,N,NDUE $
CDND ERROR3,NOEED
READ KAA,KDAAH,,,EED,USET,CASECC/LAMA,PHIA,MI,OEIGS/C,N,BUCKLING/
S,N,NEIGV/C,N,2 $
COND ERROR6,NEIGV $
$$OFP OEIGS, LAMA,,, ,//S,N,CARDNO $
OFP LAMA,,,, , / / S , N, CARDNO
30 SDR1 USET,,PHIA,,,GO,GM,,KFS,,/PHIO,,BQG/C,N,11C,N,BKLI
SDR2 CASECC,CSTM,MPT,DIT,EQEXIN,SIL,,,BGPDT,LAMA,BQG,PHIG,EST,/
OBQOI,0PHIG,OBESI,OBEFI,PPHIG/C,N,BKLI $
OFP DPHIG,OBQGI,OBEFI,OBESI,,//S,N,CARDND $
COND PS,JUI.IPPLOT $
PLOT PLTPAR,GPSETS,ELSETS,CASECC,BGPDT,EQEXIN,SIL,,PPHIG ,GPECT,OBESI/
PLOTX3/V,N,NSIL/V,N,LUSET/V,N,JUMPPLOT/V,N,PLTFLG/S,N,PFILE
35 PRTMSG PLOTX3,//
LABEL P3
REPT RAALOOP,I000
"JUMP QUITRAA $
LABEL ERROR3
40 PRTPARM //C,N,-3/C,N,BUCKLING
JUMP QUITRAA $
LABEL ERROR~
PRTPARM / / C , N , - 6 / C , N , BUCKLING
LABEL QUITRAA $
ENDALTER
R o n a l d D. Kriz I
Edge Stresses in Woven Laminates at

Low Temperatures 2
REFERENCE: Kriz, R. D., "Edge Stresses in Woven Laminates at Low Temperatures,"

Composite Materials: Fatigueand Fracture, Second Volume, ASTM STP 1012, Paul A. Lagace,
ABSTRACT: Woven glass-epoxy laminates are used as nonmetallic components at low tem-
peratures in magnetic fusion energy structures. Previous damage studies on G-10CR and G-
llCR cryogenic grade woven laminates revealed that most of the damage occurred in the
laminated interior. An existing, generalized plane strain, finite element model was modified
to predict stress states at the laminate interior and free edges. Finite element results dem-
onstrated that the weave geometry reduces edge stresses at low temperatures. Delamination
edge stresses in woven laminates are more sensitive to small changes in temperature than those
in nonwoven laminates.
KEY WORDS: woven composites, edge stresses, low temperatures, delaminations, glass/
epoxy, finite elements
Woven glass-epoxy laminates are used at low temperatures for a variety of applications
in magnetic fusion energy (MFE) structures. These woven laminates are used mainly as
electrical, thermal, and permeability barriers, which provide minimal structural support.
However, when cooled to liquid helium temperatures (4 K), large internal stresses arise
from the differential thermal contractions of dissimilar components within the laminate.
When these thermally induced stresses are combined with stresses induced by small me-
chanical loads (magnetic loads), microcracks may initiate at loads well below ultimate. The
formation of large numbers of microcracks could limit the usefulness of these materials as
permeability barriers for containing cryogenic liquids.
Previous studies [1-3] measured the mechanical properties of two popular glass-epoxy
woven materials, G-10CR and G-11CR. Later studies [4,5] related individual damage events
to the mechanical performance of G-10CR and G-11CR at small strains and low tempera-
tures. All tests were performed on thin laminates in which internal and edge damage could
be monitored periodically with increasing load. Results from these investigations showed
that bimodulus behavior resulted at strains well below the ultimate strain and that damage
originated in the interior but not at the edge. For nonwoven laminates, damage in the form
of ply cracks and delaminations originates at the edge because large stresses exist there;
this edge damage is representative of damage in the interior. For woven laminates, edge
damage is not representative of internal damage.
The edge stress problem was first defined by Pipes and Pagano [6] for nonwoven thin
laminates loaded in tension normal to the y-z plane; it is the generalized plane strain problem.
Materials research engineer, Fracture and Deformation Division, Institute for Materials Science
and Engineering, National Bureau of Standards, Boulder, CO 80303.
2This study was sponsored by the Department of Energy, Office of Fusion Energy.
150
Downloaded/printed
Copyright9 by by
ASTM lntcrnational www.astm.org
KRIZ ON EDGE STRESSES IN WOVEN LAMINATES AT LOW TEMPERATURES 151
For woven laminates, Kriz [4,5] modified the generalized plane strain problem to include
in-plane loads and fiber axis rotation, 0, in the x - z plane far from the edge at y = 0. In-
plane and out-of-plane loads were calculated by simple two-dimensional plate theory. For
woven laminates, the formulation for the edge stress problem in the y - z plane is identical
to the generalized plane strain problem defined in Ref 5, except for a 90 ~ rotation about
the z axis. The finite element formulation used in this study is given in Appendix A of Ref
5. The only exception to the previous study is that there are no applied loads in the y - z
plane. The same elastic-thermal properties from Ref 5 are used in this finite element study;
they are listed in Table 1.
The finite element grids for the woven and nonwoven edge stress problem are shown in
Fig. 1. For the [90/0], nonwoven grid, fibers are oriented at either 0~ or 90~ to the load axis
(x direction). For the [90/0]s woven grid, fiber bundles are oriented at either 0~ or 90~ to
the load axis, and the fiber bundles oriented at 90 ~ can have an additional warp angle, 0,
in the y - z plane.
Both grids shown in Fig. 1 represent only one-quarter of the total symmetric grid geometry.
There are four possible symmetric geometries. For the nonwoven laminate we can have the
sequence [90/0], shown in Fig. 1, or the sequence [0/90],, which can be obtained by inter-
changing the top surface and midplane. Similarly, we classify woven laminate geometries
at the edge as [90/0]~ and [0/90]~. Figure 2 shows the total symmetry for these two weave
geometries.
It should be noted that the nonwoven laminate [0/90], is not the same as the woven
laminate [0/90]~ where 0 = 0~ because 0 = 0~ is not physically possible but can be ap-
proached in a limiting sense. For the purpose of comparing the woven and nonwoven
laminates in the same figure we refer to the nonwoven laminate as a woven lamin_ate with
a warp angle, 0 = 0~ This comparison requires the same base to height ( B : H ) ratio. If we
use the same grid the B : H ratios are unavoidably different for different warp angles. Hence
comparisons are only made with the nonwoven 0 -- 0~ case. To avoid this problem, different
grids would need to be generated for laminates with different 0. This would require higher
density grids for laminates with larger 0 if the B : H ratio were constant. The grids shown
in Figs. 1 and 2 represent the limit of the present computer memory. Hence it was possible
to generate only one grid for different 0. When comparing results of laminates with different
TABLE 1--Elastic-thermal properties [5].
Isotropic Constituents Transversely Isotropic

Composite"
Bisphenol E-Glass
Property Epoxy Fiber Glass-Epoxy (Vr = 0.50)
Young's modulus, GPa E = 7.16 E = 82.22 E~ = 44.69

E, = 17.88
Shear modulus, GPa G = 2.77 G = 32.55 GI, = 6.801
G, = 6.888
Poisson's ratio v = 0.293 v = 0.263 u, = 0.278
v, = 0.298
Coefficient of
thermal expansion, OLI= 5.963
10-rK 1 m = 23.39
" Vt = volume fraction of glass fibers. Subscript 1 indicates fiber direction; subscript t indicates
transverse plane.
~ z
Top surface
[9om] s
X>< "
O ~'~-'/ ~-J ~ ~l/" ~ / ~ / ' ~ / ~ . ~ ~ / ~ / ~ W
~Z Midplane ~Y
ozol~
i
FIG. l--Finite element grids of woven and nonwoven (0 = O) laminates [0/90]",.
0 this approximation may be significant. Comparisons between woven and nonwoven mid-
plane stresses are additionally confused by unavoidable regions of epoxy for woven laminate
within intervals 0.0625 < y / B < 0.4375 and 0.5625 < y / B < 0.9375 for all 0. Only in the
region 0.9375 < y / B < 1.0 are woven and nonwoven laminates similar. Hence comparison
between woven and nonwoven edge stresses is confined to this region.
Both grids have a higher density of elements near the edge, where we predict a large
gradient of stress as y approaches B. Each element is a constant strain triangle with three
degrees of freedom per node. The nonwoven grid has 2044 elements with 1115 nodes; the
woven grid has 2394 elements with 1286 nodes. The global stiffness matrix of the woven
grid is larger, with 3805 degrees of freedom and a half bandwidth of 346.
In this study, we modeled edge stresses for woven and nonwoven thin laminates and
determined the influence of weave geometry and thermal effects on edge stresses. With
these results, we explain why edge damage differs for woven and nonwoven laminates.
Edge Stress Analysis

Figure 3 defines simple geometries for woven and nonwoven laminates loaded in tension
in the x direction. The symmetric stacking sequence [0/90], for the nonwoven laminate is
denoted by a subscript (s), and the ordering of the plies is 0 ~ near the top surface and 90 ~
near the midplane. Similar stacking sequences exist near the free edge of woven laminates,
where we use a superscript (0) to denote the weave curvature in the y-z plane. Stresses in
the y-z plane develop from the mismatch in Poisson's contraction between dissimilar ply or
fiber bundle orientations. For a [0/90], nonwoven laminate, the edge stress, ~r:, develops
because the transverse contraction in the 0~ ply is greater than the transverse contraction
in the 90 ~ ply. Hence, along the midplane nearest the edge, ~z is in tension because the 0 ~
and 90 ~ plies are bonded together. To satisfy equilibrium, <r: must change from tension to
compression as we move along the midplane (z = 0) away from the edge (y = B). For the
[90/0], laminate, the opposite is true: (r_, is compressive along the midplane nearest the
edge. The contribution of weave geometry and thermal loads to the edge stress problem is
not easily understood by simple intuitive models. For a better understanding of the influence
of weave geometry and thermal loads on edge stresses, we extended the existing edge stress
model of Pipes and Pagano [61 for some simple weave geometries.
Both grids model out-of-plane displacements but only for unit depth. Hence, no out-of-
plane curvatures are modeled in this study. We assume that this approximation is the most
serious for the larger warp angles. Obviously, a three-dimensional finite element model
would be more accurate, but we considered only the influence of weave curvature in the
y-z plane on the edge stresses in this study.
Mechanical and thermal loads were applied separately to woven grids with 15~ and 45 ~
of curvature. Typically, the edge stresses are normalized by the mechanical load (out-of-
plane strain) (see Ref 6). In this study we also normalized the edge stresses by mechanical
[9o/oles
. Z . .
[oloo1~
-:.
~z,~:~-'.:=..:'. - - ~ ' . , < ~ = ~ ~.~
o, . - . . . . . : . . . . . . . . . -..-~- ...... . . . . . . .,- ~ y
&'y
FIG, 2--Complete finite element grids of [90 / 0 ]~and [0/90] ~,woven laminates with complete
symmetry showing both edges.
EDGE
NONWOVEI'J (O/90)O
s
(0190)s 1)" "~TL
' [~
' O~c
'~ y
EDGE
(90/0)0
J ^ ~// WOVEN
(x
LOAD
FIG. 3--Geometries defined for woven and nonwoven symmetric laminates loaded with
uniform axial strain, where 0 is the warp angle.
(ex) and thermal (AT) loads separately and then superposed results with ex = 0.01 and
AT = - 100 K. For simplicity we show separate results before superposition for only one
laminate. For this laminate the mechanical load ~, = 0.01 and the thermal load AT = - 100
K were chosen because comparable edge stresses were predicted with these two loads when
0 = 0~

In Fig. 4, we compare the (r: stress distribution of the [0/90], nonwoven grid in Fig. 1
with the results of previous studies [7,8]. In Fig. 4 the values of (r: are normalized by the
applied out-of-plane strain, e,. They compare well with results of Wang and Crossman [7],
who used a similar finite element model with constant strain triangles, and Pagano [8], who
used an analytic approach based on Reissner's variational principle. Our grid geometry was
justified by this comparison. The approximation of the following elastic properties were
used to calculate (r,:
E~ = 137.9 GPa (20 Msi)
E, = 14.5 GPa (2.1 Msi)

(1)
Gt, = G,, = 5.86 GPa (0.85 Msi)
v/, = v, = 0.21
where the unidirectional elastic properties are denoted with l and t subscripts: l indicates
the fiber direction and t indicates the transverse direction. Two subscripts indicate a plane:
G , and v,, are shear modulus and Poisson's ratio, respectively, in the plane transverse to the
fiber direction. Using this notation the following relationships are written for hexagonal
(transversely isotropic) symmetry, where there are only five independent elastic constants.
Et v,/ = E , Pit
(2)
v, = E , / 2 G , - 1
Better approximations of elastic properties for glass-epoxy are listed in Table 1, where
u, ~ v,, and G, ~ G,,. The influence of these approximations on edge stresses is given in
Ref 9. Such approximations of elastic properties are often made for a variety of elasticity
solutions in composite materials, and they can alter the model predictions significantly. In
this study the differences of properties in the It and tt planes are insignificantly small.
Thermal and mechanical stress distributions along the midplane for woven ([0/90]s 45) and
nonwoven ([0/90],) laminates are shown separately in Fig. 5a. For superposition of thermal
and mechanical loads we are required to choose nonarbitrary values for e, and AT. For
MFE applications these materials typically operate at 4 K (liquid helium) or AT = -400~
From Fig. 5a we see that at such low temperatures thermal stresses would be dominant.
For purposes of comparing mechanical and thermal effects it would be more physically
meaningful to choose a strain and temperature that would yield equivalent edge stresses.
Hence, we choose a moderate strain, e, = 0.01 (1%) and a corresponding temperature
200
z
~_ B=8 H 25
160 1- B0~- ~ _ ~
90o H 20
120 f C>y .-
~; Wang and Crossman [7] 15 -~r
N = 6, Pagano [8] j x
80 Present Study / 10 bN
b"
40 5
o o
~ y7
-40 i , i i i i, i i i ,, -5
0 0.2 0.4 0.6 0.8 1.0
y/B
FIG. 4--Comparison of finite element results of this stud), with results of previous studies
for ~ stress along the midplane (z = 0).
0.4 B 15
Interface Interface~; Interface
Z ~x= 0.01
0.2
~ ~ v e ~ v,
o ~. ,8t,,r;_,8
o ~ ~" : ,'r =-,L_o8K3
n
-2 -0.2 I~
0 ~ - 0 3,,
t:>
-40 -5
~4 Z B =1.33 H -0.4
-10
H~ e~e:[sJ~8]e
-80
-6 -0.6
AT
-15
0 0.2 0.4
y/B
0.6 0.8 l.O
8 8.2 0.4
y/B
0.6 8.8 1.0
FIG. 5--(a) Comparison of midplane stress distribution for [0/90]~~5 woven and [0/90]~
nonwoven laminates with mechanical and thermal loads shown separately. (b) Comparison of
midplane stress distribution for [0/901,~-~ woven and [0/90]~ nonwoven laminates with and
without thermal loads superposed.
change, AT = - 100 K, because equivalent edge stress is predicted for the nonwoven lam-
inate in Fig. 5a. In the remainder of this discussion we use these values for all laminates
with and without the superposition of equivalent thermal stresses. Only in Fig. 5a are the
interface locations shown where large stresses are predicted because of the discontinuity in
epoxy and composite elastic properties across these interfaces. These large interfacial stresses
are artifacts of the finite element method and can be eliminated by modeling the elastic
properties in a continuous manner (see Ref 10). Because of these interface discontinuities
we limit our discussion to stresses away from these epoxy-composite interfaces.
Stress distributions with and without superposed thermal loads are compared in Fig. 5b
along the midplane for woven ([0/90],45) and nonwoven ([0/90],) laminates. The most sig-
nificant result for a woven laminate is that cr. is compressive near the edge without the
thermal load. With the thermal load, the compressive stress, gz, increases. For the nonwoven
laminate, the opposite is true: ~r: is tensile near the edge and increases in tension with the
thermal load. The largest midplane tensile stress exists in the epoxy region where delami-
nations could initiate, but compressive stresses near the free edge would limit the growth
of such a delamination. Hence, the weave geometry can suppress delaminations at the
midplane near the free edge.
For a woven laminate the presence of a free edge also shows the symmetry of the th stress
distribution. If no free edge existed at y / B = 1.0, the midplane cr, distribution would be
symmetric about y / B = 0.5. With the presence of a free edge at y / B = 1, the symmetry
is skewed where the largest or. stress occurs at the interior near y / B = 0. The largest tensile
~rz stress occurs in the interior (0.15 < y / B < 0.8).
Stress distributions along the midplane for [90/0],45 woven and ([90/0],) nonwoven lam-
inates are compared in Fig. 6. For the nonwoven laminate, ,r~ is compressive near the free
edge, as expected. The ~r: stress near the edge in the woven laminate is in tension and
decreases toward compression with the addition of the thermal load AT = - 100 K. Again,
we observe a beneficial effect of weave curvature on the edge stresses at low temperatures.
The symmetry of the crz distribution for [90/0], 45 is skewed, decreasing crz near the free edge
Z r = 0.01 / 15
~L.
I"I
B "I o
B = 1.33 H
]
~/-a~-SK
H I_ ~ . ~ ~ gO edge" [90101: 10
T ~1~-2"~1~ 4>y"
e=450
40 AT= -100 K 5
~ o 05
b 0
-40
-10
-80
-1 g
0 0.2 0.4 0.6 0.8 1.0
y/B
FIG. 6--Comparison of midplane stress distributions for [90/0Is ~ woven and [90/0]~ non-
woven laminates with and without thermal loads superposed.
O'z, ksi
-15 -I0 -5 0 5 l0 15
1.0. I I I I
0o ~ EDGE TOP SURFACE "~Ex = 0.01 Z
-- VIEW ~B=l.33H ~
LOAD
4
DIRECTION
z/H 0.5
90* ~ EDGE
4
LOAD
DIRECTION
~.T I NE
C
- I 0 0 -80 -60 --40 --20 0 20 40 60 80 1O0
Crz, MPa
FIG. 7--Comparison of edge stress distributions through the thickness for [0/90], ~5 woven
and [0/90]s nonwoven laminates with and without thermal loads superposed.
O-z, ksi
--20 --15 --10 -5 0 5 I0 15 20
n I I I I I
TOPSURFACE ~ Ex = 0.01 Z
~ LOAD
EDGE
DIRECTION
1/\\ J
VIEW
0~ ~ ~ Y
1/
Z/H
0o
EDGE
VIEW
LOAD
'q~T'RE"(~T
ION
/~T =-lOOK
o:oo i/
AT =-lOOK ~T=OK
MIDPLANE
I I I I I I I I I I
--140--120--100'--80 --60 --40 --20 0 20 40 60 80 100 120 140
o"z , MPa
FIG. 8--Comparison of edge stress distributions through the thickness for [90/O]J~ woven
and [90/0], nonwoven laminates with and without thermal loads superposed.
80
10
[0/90]S 5 4"x=O.O 1
40 .... [9010]: B= 4,97 H

&T = -100 K
cO
0. v
o
=OK'~X / Ii ,' aT=OK / "J ~

-5
-40 &T = -100 K
-10
-80 , ~ ~ ~ ,
o 0.2 0:4 o:e o'.8 1.o
y/B
FIG. Q--Comparison of midplane stress distributions for [0/90]~ I~ and [9010]~~ woven lam-
inates with and without thermal loads superposed.
and increasing trz in the interior near y / B = 0. The largest compressive stresses occurred
at 0.4 < y / B < 0.6 for the [90/0If laminate. The opposite is true for the [0/90]/5 laminate.
Edge stress distributions through the thickness (0 -< z / H <- 1.0) for woven ([0/90], '5,
[90/0],45) and nonwoven ([0.90], [90/0],) laminates are compared in Figs. 7 and 8. Again,
we observe a beneficial effect of weave curvature on the midplane edge stresses at low
temperatures for the [90/0]~ and [0/90]~ laminates. Away from the midplane the opposite
was true for these woven laminates where there are large tensile delamination stresses, crz,
near z / H = 0.5, which increase with the thermal load. Hence, we may expect delamination
initiation at the 0/90 interface but not at the midplane.
All the above results assumed no weave curvature in the x-z plane. Although this restriction
is not a serious limitation for small weave angles, it is most significant at the free edge for
large angles. With curvature at the edge in the x-z plane, we expect the m. stress in the [90/
0 I f to decrease at z / H = 0.5 and increase at z / H = 0 (the midplane). At the edge of the
[0/90]f laminate, we expect tr~ to decrease at z / H = 0.5 and z / H = 0. To study the exact
influence of weave curvature in the x-z plane on edge stresses requires a three-dimensional
finite element model. In this study, we described only the effect of weave curvature in the
y-z plane on edge stresses.
o-z , ksi
-15 -10 -5 0 5 10 15
1.0 I I I I I
TOPSURFACE
r~
[o/9o]~
t\
.... [9o/o]~,~
'~
B = 4.97H
E x = 0.01
Z/H 0.5
1
/// f/
AT=0 K-"~ ; AT=OK

F"-I
[ I
I I
AT=-100Kq I ~,'XT - -lOOK
I I
I I
I I
MIDPLANE
I i I
II I I 1
-100 - 8 0 -60 -40 -20 0 20 40 60 80 I0(
Oz., MPa
FIG. lO--Comparison of edge stress distributions through the thickness for [0/90]~t5 and
[90/0]~~ woven laminates with and without thermal loads superposed.
For the purpose of discussion, we selected extreme geometries to compare general trends
between woven and nonwoven geometries. Actual woven G-10CR and G-11CR laminates
have weave angles between 10~ and 15~ Both [0/90], 15 and [90/0], ~5geometries exist at the
edge of a real laminate. In Figs. 9 and 10, we compare edge stress distributions for these
geometries. Again, we see the beneficial effect of weave geometry at low temperatures for
the [90/0], 15, but large tension stresses occur at the midplane of the [0/90], ~5laminate. When
the warp angle, 0, increases to 45 ~ these midplane stresses become compressive. The op-
posite trend occurs at the 0/90 interface of a [0/90]`O laminate when 0 increases from 15 ~ to
45 ~ Hence, the region of possible delamination changes from the [0/90]~ midplane to the
[90/0],~ interface as 0 increases.
Although the weave geometry is, in general, beneficial in reducing edge stresses at low
temperatures, we expect delaminations and transverse crack formation in the fill-fiber bun-
dles to occur at or near the laminate free edge for higher thermal-mechanical loads. For
some weave geometries it is possible that damage may initiate in the interior. The hetero-
geneous nature of the weave geometry and stress distributions at and near the free edge
explains why the damage at the edge is not representative of damage at the interior.
Conclusions
In general, the weave geometry is beneficial in reducing edge stresses at low temperatures.
For a [0/90],~ laminate, the edge stress normal to the midplane changes to compression with
increasing warp angle. For [0/90],o woven laminates, edge stresses at the midplane are tensile
and decrease to compression with decreasing temperature. For both woven laminate types,
[0/90]~ and [90/0]~, the free edge skews the symmetry of the midplane stress distribution by
decreasing the magnitude of the symmetric edge stress. This trend in skewed symmetry also
occurs at low temperatures. Unlike nonwoven laminates, edge stresses in woven laminates
are dominated by thermal loads. The larger warp angles demonstrated increased sensitivity
of edge stresses to small changes in temperature. If delaminations occur at the higher thermal-
mechanical loads, initiation will most likely occur in the interior. If delaminations occur at
the edge, initiation will most likely occur at the midplane of the [0/90]~ laminates and change
to the 90/0 interface of the [90/0]`O laminates with larger warp angles. All conclusions are
based on predictions where weave curvature is confined to the y-z plane. Improved results
are possible if the weave curvature is included in the x-z plane.
References
[1] Kasen, M. B., MacDonald, G. R., Beekman, D. H., Jr., and Schramm, R. E., "Mechanical,
Electrical, and Thermal Characterization of G-10CR and G-11CR Glass/Epoxy Laminates Be-
tween Room Temperature and 4 K,'" in Advances in Cryogenic Engineering, Vol. 26, Plenum
Press, New York, 1980, pp. 235-244.
[2] Kasen, M. B., "Mechanical and Thermal Properties of Filamentary-Reinforced, Structural Com-
posites at Cryogenic Temperatures, 1: Glass-Reinforced Composites," Cryogenics. Vol. 15, 1975,
pp. 327-349.
[3] Kasen, M. B., "'Cryogenic Properties of Filamentary-Reinforced Composites: An Update,"
Cryogenics, Vol. 21, 1981, pp. 323-340.
[4] Kriz, R. D., "Influence of Damage on Mechanical Properties of Woven Composites at Low
Temperatures," Journal of Composites Technology & Research, Vol. 7, No. 2, 1985, pp. 55-58.
[5] Kriz, R. D. and Muster, W. J., "Mechanical-Damage Effects in Woven Laminates at Low Tem-
peratures," in Materials Studies for Magnetic Fusion Energy Applications at Low Temperatures--
VIII, NBSIR 85-3025, National Bureau of Standards, Boulder, CO, 1985, pp. 49-86.
[6] Pipes, R. B. and Pagano, N. J., "'Interlaminar Stresses in Composite Laminates Under Uniform
Axial Extension," Journal of Composite Materials, Vol. 4, 1970, pp. 538-548.
[7] Wang, A. S. D. and Crossman, F. W., "Some New Results on Edge Effect in Symmetric Composite
Laminates," Journal of Composite Materials, Vol. 11, 1977, pp. 92-106.
[8] Pagano, N. J., "Stress Fields in Composite Laminates," International Journal of Solids and Struc-
tures, Vol. 14, 1978, pp. 385-400.
[9] Kriz, R. D., "'Effect of Material Properties on Interlaminar Stresses in Angle-Ply Composite
Laminates," Report No. VPI-77-16, Virginia Polytechnic Institute and State University, Blacks-
burg, VA, 1977.
[10] Talug, A., "Analysis of Stress Fields in Composite Laminates with Interior Cracks," Ph.D. thesis,
Department of Engineering Science and Mechanics, College of Engineering, Virginia Polytechnic
Institute and State University, Blacksburg, VA, August 1978.
A l a n J. Russell t and Ken N. Street t
Predicting Interlaminar Fatigue Crack

Growth Rates in Compressively
Loaded Laminates
REFERENCE: Russell, A. J. and Street, K. N., "Predicting lnterlaminar Fatigue Crack

Growth Rates in Compressively Loaded Laminates," Composite Materials: Fatigue and Frac-
ture, Second Volume, ASTM STP 1012, Paul A. Lagace, Ed., American Society for Testing
and Materials, Philadelphia, 1989, pp. 162-i78.
ABSTRACT: The fatigue crack growth rates of delaminations under mixed mode (I and II)
loading conditions were investigated. Double cracked-lap-shear specimens were used to obtain
da/dN versus ~G data for AS4/3501-6 graphite/epoxy. This, together with previously reported
results for pure mode I and mode II made it possible to establish a mixed mode fatigue
criterion. The validity of these findings was then tested by using them to predict the growth
rates of through-the-width delaminations in laminates subject to constant amplitude com-
pressive Ioadings. Symmetric honeycomb sandwich specimens with delaminations located at
a depth of four, six, eight, or ten plies beneath the surface were loaded in compression at 2.5
Hz, and the fatigue crack growth rates were determined. The experimental results were gen-
erally in good agreement with the predicted behavior although fiber bridging and load-induced
crack closure resulted in predicted rates being too high in some cases.
KEY WORDS: delamination, fatigue, growth rate, mixed mode, compression, buckling,
graphite/epoxy, interlaminar
Nomenclature
A,AI, AH, A,, Constants in fatigue crack growth rate law
I1 Half-length of delamination
Aa Delamination growth
b Specimen width
C Compliance
d Half-wavelength of buckled region
Bit Flexural stiffness of buckled plies
EL, Es, E5 Axial moduli of laminate, buckled region, and subbuckled region
f Frequency
f.f. Mode I and mode [I fractions of total strain energy release rate
Gi, Gli, G Mode I, mode II, and total strain energy release rates
GM, Gp Total strain energy release rates due to M and P, respectively
AG Strain energy release rate range
Gic, Gilt, G~ Mode I, mode II, and mixed mode fracture energies
hL,hB, hs Thicknesses of regions L, B, and S
KI,KIt Mode I and mode II stress intensity factors
t Research scientist and composites group leader, respectively, Defence Research Establishment
Pacific, FMO Victoria, British Columbia, Canada V0S IB0.
162
Downloaded/printed
Copyright9 by by
RUSSELL AND STREET ON INTERLAMINAR FATIGUE CRACK GROWTH RATES 163
M Moment
N Number of applied load cycles
nB Number of plies in region B
n,nl, nll, nm Exponent in fatigue growth law
PL, PB, Ps Axial loads in regions L, B, and S
R Fatigue load ratio
W, W,x W,xx Shape and first and second derivatives w.r.t, x of buckled region
Wo Shape of buckled region prior to loading
OLM,(XP Mode I fractions of Gu and Gp, respectively
F Shear modulus
8 Out-of-plane displacement amplitude including 8o
8o Value of 8 prior to loading
EL,EB,~S Axial strains of regions L, B, and S
0 Angle of rotation of end of buckled plies
Introduction
The applicability of linear elastic fracture mechanics and the strain energy release rate,
G, to interlaminar fracture of organic matrix composite materials has been well established.
However, to be of general use it has been necessary to characterize delamination behavior
under mode II loading (forward shear) and in some instances mode III (tearing), in addition
to mode I (tensile opening). Moreover, for brittle epoxy matrices the mode II fracture
energy, G~,, is significantly greater than that for mode I [1]. This not only means that the
individual components of the strain energy release rate, G~, GH, and Gm must be determined
but that the fracture properties under mixed mode loading conditions must also be known.
Fortunately, it appears that for a number of composite materials there is a simple relationship
[1,2] that predicts the fracture energy, G,, under any combination of modes I and II
Gt GII GI + G,
Gt'-'-~,+ allc -- Gc - 1 (1)
The first equality in Eq 1 is also valid for fatigue at subcritical loads as is shown in the
Appendix. A further complicating factor is the presence of fiber bridging which occurs under
mode I-dominated conditions and results in an increase in fracture resistance during quasi-
static crack growth. Fiber bridging effects can be difficult to predict because they are very
sensitive to lay-up, fiber volume fraction, the amount of crack opening, moisture, and
temperature [3].
The effects of loading mode and fiber bridging also influence delamination growth under
fatigue conditions. Previously the authors investigated interlaminar fatigue crack growth
under both pure mode I [4] and pure mode II [5] Ioadings. It was found that in both shear
and tension the crack growth rate, da/dN, could be related to the strain energy release rate
range, AG, by a relationship of the form
da/dN = A(AG/G~)" (2)
However, for a given value of da/dN, greater values of AG were required in mode II than
in mode I. Moreover, fiber bridging, which occurred in the mode I tests only, resulted in
a rapid decline in da/dN during fatigue under constant AG conditions. Thus the fatigue
crack growth rate depended not only on AG but also on the previous growth history of the
delamination.
This paper investigates the problem of mixed mode interlaminar fatigue crack growth in
AS/3501-6 graphite/epoxy laminates. First, fatigue testing of cracked-lap-shear (CLS) spec-
imens were carried out to establish a mixed mode interlaminar fatigue criterion and to
determine the extent of fiber bridging under mixed mode loading. Then the experimental
data were used to predict delamination growth rates in the well analyzed problem [6-10]
of out-of-plane buckling in compressively loaded laminates.
Mixed-Mode Fatigue
Double cracked-lap-shear (DCLS) specimens were chosen for the mixed mode fatigue
tests because of their linear response and the independence of Gx and G . on crack length.
Unidirectional laminates of Hercules AS4/3501-6 graphite/epoxy, 24 plies thick and con-
taining Teflon-coated glass fabric inserts between plies 6 and 7 and plies 17 and 18 were
laid up by hand and cured according to the manufacturer's recommended cure cycle. Fol-
lowing an 8-h post cure at 177~ the specimens were cut from the laminates and tabbed as
shown in Fig. 1. Testing involved precracking the specimens 1 to 2 mm from the end of the
inserts, loading to fracture at a rate of 0.5 mm/min and then fatiguing under constant load
amplitude at 2 Hz and an R ratio of 0.05. To restrict the delamination extension following
fracture and to ensure that both cracks arrested with approximately the same length, the
specimen was clamped 10 mm beyond the precracks. This clamp was removed prior to
fatigue loading. Crack growth was monitored visually under low-power magnification and
was facilitated by applying a thin white coating to the polished edges of the specimen. An
extensometer was used to measure the elastic modulus, and the strain energy release rate
was calculated from the following equation:
G = ~-~ (E-h)1
lj
(E-h) (3)
where P is the load, b the specimen width, E the longitudinal tensile modulus, and h the
thickness. The subscripts 1 and 2 refer to the thin and thick ends of the specimen, respectively.
Equation 3 can be derived in the same way as that for the single CLS specimen [11] because
as far as the total strain energy release rate is concerned the DCLS is simply two single CLS
specimens back to back.
In most cases, constant fatigue crack growth rates were observed for a given load although
occasionally some fiber bridging occurred accompanied by a gradual decrease in da/dN.
Thus it would appear that the mode I fraction in these DCLS specimens represents the
transition limit for the occurrence of fiber bridging under fatigue conditions. The constant
crack growth rates obtained, as well as the initial or maximum values (na = 0) of da/dN
when fiber bridging occurred, are plotted against AG/Gc in Fig. 2. Also shown for comparison
are the previously obtained data for mode I and mode II. The mixed mode data fall between
the pure mode data and have a slope greater than that for mode II but less than that for
[O]lz [0]~4
75ram ~
250 mm
FIG, 1 - - D o u b l e cracked-lap-shear specimen. Width = 15 mm.
f , ,/ ~,1
AS4/3501-6 2O~ DRY / / /
1 0 .2 9 D C L S DATA ( R = O 0 5 ) /I /
- - - a/o,:o.6a / / / ' /
f =2Hz, R=O / / ~./
~'/ _/ ~ ' /
0
*/
7 7,, ,,,'/~ /
o 10-3 ,o'/ / 9 ~/
E
Z
"0
"U
,.47/ 4._'/
I,-
<
//
9t - 10 -4
I-
o
n.
o
n.
o
tu
~ 9 1 0 -s
I-
<
//" /
U.
1 0 - 6 ~
0.1 0.2 0.4 0.6 0.8 1.0
AG/G c
FIG. 2--ComparL~on of double cracked-lap-shear fatigue data with pure mode I [4] and
pure mode H [5] results.
mode [. This suggests that a rule-of-mixtures type of relationship might provide an empirical
fit to the mixed mode fatigue data, namely
da/dN = A~(AG/G.) ~ (4)

where
A., = fiAi + f]iA]]
f, = G,/CG, + C,,)
.f, = 1 - f ,
nrn = f i n , + fIlnI,
Linear finite-element analysis (FEA) of the DCLS specimen (see section entitled Analysis)
gave a value for f[ of 0.37 and the resulting curve predicted by Eq 4 is plotted in Fig. 2. It
can be seen that this agrees well with the experimental data and hence Eq 4 was used to
predict the fatigue crack growth rates in the compressively loaded laminates.
Analysis
The mode I and mode II components of the strain energy release rate for delamination
growth due to out-of-plane buckling were determined as a function of delamination depth,
delamination length, and laminate stiffness. The method of analysis used was a refinement
of Whitcomb's approach [8], which made use of buckling theory to linearize the problem,
superposition analysis to break down the loads into workable components, and linear F E A
to calculate GI and G~[.
Whitcomb compared his results with nonlinear F E A and showed that the major source
of error was in accurately determining the postbuckling load, PB, and the out-of-plane
deformation, 8. Recently, Rothschild et al. [9] showed that the assumption of clamped-end
boundary conditions can lead to an underestimation of 8, particularly in the case of deep
delaminations. By allowing rotation of the ends of the buckled region, 8 could be made to
agree with the value obtained from nonlinear F E A . A similar procedure was adopted in
the present work except that nonlinear F E A was avoided by determining the end rotation,
0, from a linear F E A .
With reference to Fig. 3 and following the arguments of Ref 8, the shape of the buckled
region is assumed to be
8
w(x ) = -~ [cosOrx / d) - cos( ~ra/ cl) ] (5)
where d is defined so that
2d sin('na/ d)
(6)
~r8 tan 0
because this satisfies the boundary conditions at x = -+a, namely w = 0 and w,x = -Ttan0.
The relationship between bending moment and curvature
M(x) = - P s ( w + 0.58 cosOra/d)) = D,b(w,~, - w.... ) (7)
results in a postbuckling load
~2Dllb (8 - 8o)
P~ = d---S - ~ (8)
i ~Ps! :
2a
2d
FIG. 3--Out-of-plane buckling geometry with rotation 0 at ends of buckled plies.
M G=O
/
FIG. 4--Reduction of linearized loads into three components by superposition.
where 8o is the out-of-plane displacement prior to loading and Dll is the flexural stiffness
of the buckled plies.
8 is fixed by the compatibility requirement that the shortenings of the buckled and sub-
buckled regions are equal
~2(2a - d ) . (82 _ 8~ + 2aeB = 2a~s (9)

8d 2
Then making use of the relationships between laminate loads and strains
PL = ~LELhLb = P8 + Ps = ~aEBhBb + e, Eshsb (lo)

Equations 6 through 10 can be solved iteratively to obtain values for P~ and 5. Whitcomb's
results for clamped ends can be retrieved by setting tan 0 = 0 and d = a.
Next, superposition was used to reduce the problem to three independent loadings as
shown in Fig. 4. Of these only the moment M which is equal to 0.5 8 PB cos Ova~d) and
the load P = P , - PBELhL/(EBhB) result in strain energy release rates, GM and Ge, re-
spectively. GM was determined from the compliance, C, which is simply half that of a double-
cantilever-beam (DCB) specimen [12] plus a term which allows for rotation at the crack tip
4L 3 1.2L
C = Ebh3 + ~ + L20uo (11)
where 0Mo is the rotation per unit applied moment. It then follows that
p2 dC 6M 2 0.6M 2 M0M
GM - 2b dL - -------3
Eb2h + Fb2L2--~ + b ~ (12)
In using this equation L is made equal to a/2. The error in this assumption is not large
because the first term predominates and is independent of L. The strain energy release rate,
Ge, was calculated using Eq 3 because the geometry is identical to one half of a DCLS
specimen.
Linear F E A were carried out for both loading conditions using meshes similar to the one
shown in Fig. 5 with four-noded, two-dimensional isoparametric solid elements (ANSYS
(-20,1.4) (20,1.4)
( - 2 0 , - 1.96) (20,- 1.96)

FIG. 5--Finite element mesh used for modeling ten-ply-deep delaminations. The coordinates
are in millimetres, and the crack tip lies at the origin.
STIF42) in plane strain. The crack closure method of Rybicki and Kanninen [13] with a
crack growth increment of 0.03 mm was used to determine GI and G.. The F E A results are
compared with the strength of materials values in Table 1. For both loading conditions, the
total strain energy release rates obtained by the two methods are in good agreement. Also,
the mode I fraction was relatively insensitive to delamination depth. Thus it was possible
to use the strength-of-materials expressions for GM and Ge combined with the F E A values
for aM and c,e, to calculate the components of G~ and G . as shown below
G? = auGM G[ = eteGe (13)
and
G~ = (1 - otu)GM G~ = (1 - ote)Gp (14)
Superposition of the strain energy release rates [14] then gives
G~ = [(G~) ~ - (G/)~ 2 (15)

and
G,, = [(G~) ~ + (G~)~ 2 (16)
The negative sign in Eq 15 is required because M results in crack opening whereas the load
P results in crack closure as is evident from Fig. 4 and previously discussed by Whitcomb
[81.
Thus, the analysis procedure consisted of an initial estimation of PB and M which allowed
0 = 0M - 0p to be calculated from the data in Table 1, which in turn enabled a better
TABLE 1--Comparison of strain energy release rates from strength of materials (SM) and finite
element ( FE) analyses.
Depth, ns (plies) GsM, J/m z GFE, J / m ' Gsu/GFe GI/G 0, radians
M/b, 20N
4 114.9 116.1 0.990 0.603 0.150/nB2
6 35.8 36.3 0.986 0.612 0.151/n~2
8 16.6 16.9 0.982 0.619 0.152/nB2
10 8.39 8.56 0.980 0.628 0.153/nB2
P/b, IOMN/m
4 22.9 23.6 0.969 0.397 0.0107
6 37.4 38.1 0.981 0.392 0.0109
8 56.7 57.2 0.991 0.386 0.0112
10 82.0 81.8 1.003 0.374 0.0116
estimate of P~ and M. This process was then repeated until constant values of PB and M
were achieved, usually after three or four iterations, at which point G~ and G, were cal-
culated. Finally, AG/Gc was determined from Eq A4, and da/dN was evaluated using Eq
4 and the values of A and n shown in Fig. 2. Figure 6 shows examples of predicted fatigue
crack growth rates for delaminations located at different depths. Although the shapes of
the curves are similar, increasing the depth raises the critical delamination length for fatigue
crack growth but also shifts the curves to higher values of da/dN. Also shown in Fig. 6 is
the gradual reduction in the mode I fraction of the strain energy release rate as the delam-
inations grow.
i i i i i i
GI/G
0.5 10 P L u
0.4
0.3
0.2
10-2 0.1
0
8 PLY/
qp
o
1~
Z
1:1
'o
I.-
~(
\
dc
"1"
O
gg
O
t,1
/,
10. s
\ EL= 1200/~m/m
EL=Es--Es=130 GPa
~.~ D,, = 360 Nm
h L : 3.3rnm,~o: 0
10. 1 I I ! I
2O 40 60 80 100 120 140 160
OELAMINATION LENGTH, 2a (ram)
FIG. 6--Examples of predicted fatigue crack growth rates for delaminations at four, six,
eight, and ten plies beneath the surface. Laminate strain, eL = 1200 txm/m.
Delamination Growth Under Compressive Loading

Experimental
Honeycomb sandwich specimens containing through-the-width delaminations in both face
sheets (Fig. 7) were used for the compression tests. Recent work [15] has demonstrated the
validity of this type of specimen for characterizing compression failure with and without
defects. Delaminations were located at a depth of four, six, eight, or ten plies below the
surface. Laminates were manufactured in the same way as for the mixed mode fatigue tests
and then cut into 22-ram-wide strips prior to bonding to the aluminum honeycomb and end
fittings with 3M AF-126 film adhesive. Loading-pin holes, 12.5 mm in diameter, were drilled
in both ends of the specimens which were then mounted in the load frame as shown in Fig.
8. A small compressive load was applied, and the alignment bolts were tightened. Alignment
was checked by ensuring that the load/strain response and the out-of-plane buckling were
the same on both sides of the specimen.
Fatigue tests were run under a constant load amplitude at a frequency of 2.5 Hz and an
R ratio of 0.05. A n extensometer was used to monitor the total out-of-plane deformation
of both delaminations, and this output was used to periodically stop the test for in situ
measurements of delamination growth which were made in the same manner as for the
mixed mode fatigue tests. The load amplitude and initial delamination length were selected
so as to avoid fast interlaminar fracture as the delaminations grew.
1 2 . 5 mm
i~ ~-i
;TEEL
NUM T A B S
Y GRAPHITE/EPOXY
E
E
o IINATION
rlNUM H O N E Y C O M B
26 ADHESIVE
ING PIN HOLE
FIG. 7--Honeycomb sandwich specimen used for compression tests.
FIG. 8--Compression specimen mounted in load frame showing location of extensometers

for measuring strain and out-of-plane displacement.

Figure 9 shows a comparison of the measured and predicted out-of-plane displacements
as a function of laminate strain ~L for a 30-ram-long delamination four plies deep. This
demonstrates clearly the need to allow for rotation of the ends of the buckled region. The
effect of an initial imperfection of 0.025 mm is also shown in Fig. 9. While this improves
0.8
0.7
'E 0.6
Z
0.5
0
,.I
0.4
3
< 0.3 O.025mm#/>()~

(50= ~ ~--~--~o= 0
L
o
~
0
0.2 ///
'- ""-0 = 0
2a=3Omm
0.1 EXPERIMENTAL / / / EL = 1156 #m/m
CURVE ~.. / [ AS4/3501-6, [0]74
DEPTH =4 PLIES
0 500 1000 1500 2000
LAMINATE STRAIN, (#m/m)
FIG. 9--Comparison of predicted (solid curves) and measured out-of-plane displacements
as a function of laminate strain for a 30-ram-long, four-ply-deep delamination.
i ! ! i
I
2a=3Omm
AS4/3501-S,[OIz4
9 MEASUREOVALUES
DEPTH=G PLIES
I0 i I I
70 8 90 100 110 12Q
DELAMINATION LENGTH, 2a (ram)
FIG. lO--Comparison of predicted (solid curves) and measured out-of-plane displacements
as a function of delamination length for an eight-ply-deep delamination.
! i i
AS4/3501-6, 20~ DRY

-1 f = 2 . 5 Hz, R = O . 0 5
10
DEPTH = 4 P L I E S
+ x L~L= 2 3 8 7 # m / m
e| eL=20?9#m/m
4.x
16 2 +~
A
@
U
x 4.
E
Z x 2387~m/m
| | x
x +
'U 1 6 3 |
6) x +
UJ 4.
I- x 4.
| x + +
E x x x
-r,
I-
3=
O | 2079#m/m
E
U
< 1() 4
E
O
|
|
|
|
|
I I I I I
. 5 20 40 60 80 100 120 141
D E L A M I N A T I O N L E N G T H , 2 a (ram)
FIG. 11--Comparison of predicted and measured values of d a / d N in specimens with de-

laminations located four plies beneath the surface. See Fig. 6 for details of the predicted curves.
the fit at low strains it tends to overestimate B at higher strains. The failure of any of the
theoretical curves to predict the buckling load correctly may result from a slight curvature
of the specimens due to excess resin bleed at the ends of the laminate rather than to a
failure of the model. In any case, at higher strains and longer delamination lengths where
most of the fatigue data were generated, good agreement between predicted and measured
displacements was found for ~o = 0, provided that end rotation was allowed. A n example
of this is shown in Fig. 10 for an eight-ply-deep delamination.
Plots of fatigue crack growth rate versus delamination length are shown in Figs. 11 through
14. In spite of the efforts that were made to carefully align the specimens, it was generally
found that crack growth was more rapid on one side of the specimen than on the other,
often by as much as a factor of 2. This not only reflects the steepness of the da/dN versus
I I
A S 4 / 3 5 0 1 - 6 , 20 C, DRY
f : 2 . 5 Hz, R : 0 . 0 5
/ ~ DEPTH : 6 PLIES
1 ~ +x ~L = 1 6 6 1 # m / m
61 #m/m
4.4. 4-4.4.4-4.
4. 4-4- 4.4. 4-4-~ 4 . 4.4-
4- 4- o ~ o ~ 4. 4- 4- 4- 4-
10 "3 _t "~ | | 9 | ?@
|
!:
1() l i
40 60 80 100
DELAMiNATION LENGTH, 2 " ( m m )
FIG. 12--Comparison of predicted and measured values of da/dN in specimens with de-
laminations located six plies beneath the surface. See Fig. 6 for details of the predicted curves.
&G plots but also the fact that small amounts of bending can greatly influence the fraction
of opening mode at the crack tip. However, it can be argued that when a symmetric response
was obtained, the specimen was indeed seeing pure compression. This would not have been
true for a specimen with a delamination on one side only, where the extent of bending would
have been unknown or at best difficult to determine. It is also likely that in those tests
where there was a significant difference between the two sides, the true crack growth rate
lies somewhere between the two measured values and hence the test provided an upper and
lower bound on da/dN.
Figures 11 through 14 also show the crack growth rates predicted using the same material
properties as shown in Fig. 6. Generally, there was reasonably good agreement between
the predicted and measured values of da/dN although there were some consistent discrep-
ancies. If the initial delamination length was sufficiently short that G J G was greater than
0.4, then fiber bridging occurred, and, although the delamination began growing at the
predicted rate, it quickly fell behind. Figures 12 and 14 show examples of this behavior.
Because the bridged fibers persisted for some time before failing they exerted a significant
effect on da/dN long after the delamination had grown to the point where no new fiber
bridging occurred.
Another type of discrepancy was observed for long delaminations. In this case the crack
growth rate declined faster than predicted especially at the lower fatigue loads. Examples
of this are shown in Fig. 11. In the four-ply-deep specimens, G~ falls to zero at 2a = 60 mm
at which point GH is set equal to the total strain energy release rate. However, as the
delamination continues to grow, the closure at the crack tip continues to increase, and it is
likely that the resulting friction reduces da/dN. This explanation is consistent with the results
of Wu [16], who showed that a negative KI resulted in an increase in K, for the longitudinal
splitting of unidirectional fiber-glass.
Finally, the analysis appears to consistently underestimate da/dN at small fractions of
16 2
4.
x
4.
Q
E
+ 4. 4- 4-
Z
'0
x
I +x ~. '~
10-3
UJ
I-
Z
1--
o
E
o
r
E AS4/3501-6,20~ DRY
u
f : 2 . 5 H z , R=O.05
•L:1155 #m/m
DEPTH = 8 PLIES
lO .4 I I
60 80 100 120
DELAMINATION LENGTH, 2a (ram)
FIG. 13--Comparison of predicted and measured values of da/dN for specimens with
delaminations located eight plies beneath the surface. See Fig. 6 for details of the predicted
curve.
10 -1 ! ! I I 1
E 162
z
+ 4- + + +
4-
I-
4,
4" X X
4" X X X
on ,
u §
E
x
AS4/3501-6,20~ DRY
ld 3
f = 2.SHz, R=0.05
(~L=1212~m/m
DEPTH = 10 PLIES
I I I I i
70 80 90 100 110 120 130
DELAMINATION LENGTH, 2 a (rnm)
FIG. 14--Comparison of predicted and measured values of da/dN for specimens with
delaminations located ten plies beneath the surface. See Fig. 6 for details of the predicted curve.
mode I. This could arise from either a failure of the fatigue criterion to correctly predict
da/dN over this range or else from an error in the G~ and/or GI, values. The latter could
be checked by comparing the present results with those from a nonlinear finite element
analysis. However, to verify the fatigue criterion it will be necessary to devise a specimen
capable of generating interlaminar fatigue crack growth rate data at between 5% and 10%
mode I.
Conclusions
The following conclusions concerning the mixed mode interlaminar fatigue properties of
AS/3501-6 graphite/epoxy and their applicability to delamination growth under compressive
loading can be made as a result of this investigation:
1. A mixed mode fatigue criterion of the form

da
d"'N = Am(AG/G~)"~
where Am and n~ are related to the pure mode I and mode II values by a rule-of-mixtures
gave a good fit to the experimental data obtained at a mode I fraction of 0.37.
2. A strength-of-materials-based analysis for determining GI and GII for delamination
growth due to out-of-plane buckling has been developed. The use of finite-element analysis
has been limited to the determination of the ratio of GI/G and the rotation of the ends of
the buckled plies, both of which are relatively insensitive to delamination depth.
3. The analytical model and mixed mode fatigue criterion successfully predicted the de-
lamination growth rates due to out-of-plane buckling in the absence of fiber bridging. It
was possible to predict when fiber bridging would occur, however further work will be
required to quantify the resulting reduction in da/dN.
4. For long delaminations, closing forces at the crack tip have a retarding effect on the
delamination growth rate particularly at lower strains.
Acknowledgments
This work is part of the program in support of the Department of National Defence
technology requirements for damage tolerance analyses of advanced materials in high per-
formance aircraft. The authors are grateful to E. Jensen for his expertise in carrying out
much of the experimental part of this investigation.
APPENDIX
Consider loading along path O A in Fig. A1 such that G I / G . = m where m is constant.

Failure occurs at P' where GI = G~', GI~ = G , ' and
GI'/GIc + GII'/G.c = 1 (A1)
The mixed mode fracture energy for this value of rn is, therefore,
Gc = Gj' + G . ' (A2)
GI
G'C~p A
G'O G.. G.' G..c ' G.,
FIG. Al--lnteraction between constant mixed mode loading path and mixed mode failure
en velope.
Substituting GI' = mG~i' into Eq A1 and Eq A2 and eliminating G . ' gives
m 1 m+l
-- + - (A3)
GI~ Gu~ G~
Finally, for any point P along OP', Eq A3 can be multiplied by the corresponding value of
G , to give
G__
L + G . _ Gt + G , _ AG (A4)
Gic Gilt Gc Gc
because for fatigue at R = 0, AG = G~ + Gn.
References
[1] Russell, A. J. and Street, K. N., in Delamination and Debonding of Materials, ASTM STP 876,
W. S. Johnson, Ed., American Society for Testing and Materials, Philadelphia, 1985, pp. 349-
370.
[2] Johnson, W. S. and Mangalgivi, P. D., in Toughened Composites, ASTM STP 937, N. J. Johnston,
[3] Russell, A. J., "'Micromechanisms of Interlaminar Fracture and Fatigue," Polymer Composites,
Vol. 8, 1987, pp. 342-351.
[4] Russell, A. J. and Street, K. N., in Composite Materials: Testing and Design (Eighth Conference),
ASTM STP 972, American Society for Testing and Materials, Philadelphia, 1988, pp. 259-277.
[5] Russell, A. J. and Street, K. N., in Toughened Composites, ASTM STP 937, N. J. Johnston, Ed.,
[6] Chai, H. Babcock, C. D., and Knauss, W., lnternationatJournalofSotids Structure, Vo[. 17, 1981,
pp. 1069-1083.
[7] Ashizawa, M., "Fast lnterlaminar Fracture of a Compressively Loaded Composite Containing a
Defect," presented at the Fifth DOD/NASA Conference on Fibrous Composites in Structural
Design, New Orleans, Jan. 1981, Douglas Paper No. 6994.
[8] Whitcomb, J. D., in Effect of Defects in Composite Materials, ASTM STP836, American Society
for Testing and Materials, Philadelphia, 1984, pp. 175-193.
[9] Rothschild, R. J., Gillespie, J. W., and Carlsson, L. A., in Composite Materials: Testing and Design
(Eighth Conference), ASTM STP 972, American Society for Testing and Materials, Philadelphia,
1988, pp. 161-179.
[10] Flanagan, G., in Composite Materials: Testing and Design (Eighth Conference), ASTM STP 972,
[11] Brussat, T. R., Chiu, S. T., and Mostovoy, S., "Fracture Mechanics for Structural Adhesive
Bonds," AFML-TR-77-163, Wright Patterson Air Force Base, Dayton, OH, 1977.
[12] Mostovoy, S., Crosley, P. B., and Ripling, E. J., Journal of Applied Polymer Science, Vol. 10,
1966, pp. 1351-1371.
[13] Rybicki, E. E and Kanninen, M. E, Engineering Fracture Mechanics, Vol. 9, No. 4, 1977, pp.
931-938.
[14] Tada, H., Paris, P. C., and Irwin, G. R., The Stress Analysis of Cracks Handbook, Del Research
Corp., Hellerton, PA, 1973, p. 1.13.
[15] Lagace, P. A. and Vizzini, A. J., in Composite Materials: Testing and Design (Eighth Conference),
ASTM STP 972, American Society for Testing and Materials, Philadelphia, 1988, pp. 143-160.
[16] Wu, E. M., in Composite Materials, Vol. 5 (Fracture and Fatigue), L. J. Broutman, Ed., Academic
Press, New York, 1974, pp. 191-247.
Delamination
Alice L. Glessner, 1 Michael T. Takemori, 2 Michael A. Vallance, 3
and Steve K. Gifford 2
Mode I Interlaminar Fracture Toughness of

Unidirectional Carbon Fiber Composites
Using a Novel Wedge-Driven
Delamination Design
REFERENCE: Glessner, A. L., Takemori, M. T., Vallance, M. A., and Gifford, S. K.,
"Mode ! Interlaminar Fracture Toughness of Unidirectional Carbon Fiber Composites Using
a Novel Wedge-Driven Delamination Design," Composite Materials: Fatigue and Fracture,
ABSTRACT: The mode I interlaminar fracture behavior of unidirectional carbon fiber com-
posites with PMR-15 polyimide and PEEK (polyetheretherketone) matrix resins has been
studied at various temperatures and testing rates. These materials were evaluated for strain
energy release rate. Gx, using a novel wedge-driven delamination (WDD) design. In this WDD
configuration, Gt is directly proportional to the measured load necessary to advance the sample
onto the wedge. The crack front remains essentially stationary in the laboratory reference
frame during steady crack growth, facilitating the monitoring of the fracture process. Fur-
thermore, the crack velocity can be set to any desired value by controlling the machine
crosshead speed. Steady crack growth was observed in the polyimide laminates, which exhibited
an average fracture toughness of 370 J/m 2 at 25~ The measured GI values were slightly
dependent on test rate and showed an increase in toughness at increasing temperatures in the
range of 25 to 200~ The toughening mechanism of bundle pullout seemed to be more
dominant at higher temperatures, possibly accounting for the higher toughness values. The
PEEK laminates exhibited the stick/slip crack growth behavior typical of ductile matrix com-
posites. Slow growth regions were identified by whitened areas on the fracture surface, while
the fast fracture areas remained unwhitened. The strain energy release rates of the composite
corresponding to both the onset of fast fracture and the arrest of the crack were measured as
a function of temperature and rate. While the arrest GI values were fairly independent of
temperature and rate, the G~ values for instability onset increased with increasing temperature.
KEY WORDS: crack extension force, crack path tortuosity, crack tip process zone. double-
cantilever beam, fiber bridging, fiber bundle pullout, interlaminar fracture toughness, stick/
slip crack growth, strain energy release rate
The interlaminar fracture behavior of continuous fiber composites based on both ther-
mosetting and thermoplastic resin matrices has been extensively investigated [1-21]. In these
studies, mode I interlaminar fracture tests have been p e r f o r m e d mainly on double-cantilever-
Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis,

MN 55455.
2 Polymer Physics and Engineering Branch, Corporate Research and Development, General Electric
Co., Schenectady, NY 12301.
3 Plastics R & D Labs, Ciba Geigy, Ardsley, NY 10502.
181
Downloaded/printed
Copyright9 by
beam (DCB) samples. Recently, Vallance et al. [22] have developed a novel adaptation of
the wedge-driven delamination (WDD) technique which allows simple velocity-controlled
mode I interlaminar fracture testing. This new W D D approach permits a direct measurement
of the interlaminar fracture toughness G~c, constant (or programmed) crack velocity testing,
a stationary crack front with respect to the laboratory reference frame, and a relatively
simple extension to modes 1I and III testing.
In the work presented here, the W D D technique has been used to study the interlaminar
fracture behavior of unidirectional carbon fiber composites based on a thermosetting matrix
resin and on a thermoplastic matrix resin. Observations were made on the differences in
behavior for the two resin-based laminates at various temperatures and crack velocities.
The transfer of matrix resin toughness to the laminates is discussed with special consideration
given to fractographic observations and the nature of the crack propagation behavior.
Mode I Wedge-Driven Delamination Test Configuration

The mode I wedge-driven delamination (WDD) testing configuration is shown in Fig. 1.
In the W D D testing design [22], the sample is forced upon a stationary low-friction wedge,
which is composed of an even number of bearings, alternately rotating clockwise and coun-
terclockwise. The clockwise rotating bearings are coaxial about one axis of rotation, while
the other half of the bearings are coaxial about a second axis of rotation, offset from the
first. Because of the offset rotation axes, each set of bearings contacts one half of the
symmetrical split laminate. The counter rotating offset bearings provide a low-friction wedge,
which is a key to this testing design. The uncracked end of the W D D test sample is clamped
tightly between two steel plates with considerable care taken to align the sample with the
plane of the crack bisecting the distance between the two roller axes.
A finite deformation model of cantilever deflection was arrived at using Euler-Bernoulli
slender-beam analysis [22]. The nonlinear homogeneous differential equation governing
deformation of the beam was derived from an expression for system potential energy using
variational calculus. The solution, obtained by change of variables, was used to model the
deflection of one half of a split laminate by a frictionless roller, as shown in Fig. 2.
L sinl3+ ( ~ ) f o 'c~
(1)
R-'-~= + cosl~ f l sin0 [sin(13 - 0)]-t,2 dO
PR2 = sinl3
E1
sin0 [sin(13 - 0)] -"2 dO + cosl3 )}' (2)
where terms are defined in Fig. 2. S, = S - H/2 and R, = R + H/2. An energy conser-
vation equation has been written for the case of finite, nonlinear, elastic deformation with
crack growth under the influence of one force
C~ OP /dA
(3)
Here G is the instantaneous crack extension force, 2A is the crack area (both surfaces), and
P(x,5) is the unloading deflection curve following loading to 5. Combining Eqs 1 through
GLESSNER ET AL. ON MODE I INTERLAMINAR FRACTURE TOUGHNESS 183
FIG. 1--Mode I wedge-driven delamination (WD D ) testing apparatus. Wedging device with
offset counter-rotating bearings (left) and side view of wedge and split laminate (right).
9 c -I- L .I
FIG. 2--Definition of terms for modeling the deflection of a split laminate by a frictionless
roller.
3 in several steps leads to
Gt = P / B (4)
where B is the laminate width. Previous authors have apparently overlooked this simple
equality. The subscript I signifies mode I opening. A final expression of some utility is the
asymptotic expression for vanishing 13
lim PRz~ = 9(1 + S~/R~)2- (5)

~,o E1 (L/Re) 4
Although the derivation has not been shown here, the assumptions should be mentioned.
The derivation of Eq 3 assumed isothermal deformation without inertia. The derivation of
Eqs 1 and 2 assumed negligible shortening as a result of axial compression and frictionless
rotation of the rollers on their axes. No translation or rotation was allowed at the crack
front (cantilever condition).
The great advantage of the W D D configuration lies in the simple computation of the
interlaminar crack extension force, Gt, given by Eq 4. G~ is calculated simply by dividing
the applied compressive load by the sample width. Thus, for samples of constant width, the
time dependence of GI is easily obtainable directly from the load record alone. The mode
I interlaminar fracture toughness G~cis then determined from the value of Gx that corresponds
to the initiation of crack growth. Thus, the WDD technique avoids the more tedious tech-
niques associated with the conventional DCB test specimens [1,23,24].
Another significant advantage of our stationary wedge, moving-sample W D D technique
is a virtually stationary crack front in the laboratory reference frame. For steady crack
growth conditions, the average crack velocity along the sample length is the same as the
machine crosshead velocity, while the crack front position, relative to the wedge rollers, on
average, remains constant. Therefore, the crack front remains motionless relative to exter-
nally fixed recording equipment. Furthermore, the crack velocity can be set to any desired
value by controlling the machine crosshead speed. For discontinuous crack growth condi-
tions, such as stick/slip crack growth, the crack front position moves relative to the wedge
rollers. During crack arrest, the crack front moves closer to the wedge rollers (at the
crosshead speed), and, during fast fracture, the separation increases rapidly. The average
distance, however, remains relatively constant. Any desired crack velocity profile can be
programmed easily. Stationary geometry also means that the crack propagation at the end
of a test duplicates the propagation at the beginning with regard to the ratio of shear stress
to flexural stress and the unloading rate d G / d L . This last quantity relates to the extent of
propagation resulting from a crack front instability. Because the geometry is repeating, the
load, on the average, stays constant, similar to the tapered DCB experiment. Here the crack
front is straight, however, not curved. In the DCB experiment, when the displacements
become very large during extended crack growth, finite-deformation elasticity solutions are
required [12]. They are rarely needed in the W D D geometry, although the formulation of
Eqs 1 through 4 is generalized for finite deformations.
Materials and Methods

Two different unidirectional carbon-fiber-reinforced laminated composite systems were
tested for mode I interlaminar fracture toughness. One system was based on a thermosetting
polyimide developed by researchers at the NASA Lewis Center [251, PMR-15 (polymer
from monomer reactants), with a calculated average prepolymer molecular weight of 1500
g/g-mole.
Irnidized structure of PMR-15 prepolymer.
These carbon-fiber-reinforced materials have been advocated for continuous use in envi-
ronments up to 316~ [26]. The composites were prepared in the form of 30 by 30 cm 20-
ply unidirectional laminates by vacuum-bag autoclaving with an average thickness of
0.32 cm.
Several sections transverse to the fiber axis were metallographically polished for obser-
vation under reflected light in a Zeiss optical microscope. The overall void content of the
PMR-15 plaques was determined using a Zeiss IBAS Interactive Image Analysis System to
be less than 0.3%.
The second system was based on Imperial Chemical Industries-America's Victrex P E E K
(polyetheretherketone), a semicrystalline polymer typically used in composite applications.
PEEK has a melting point of 334~ and a glass transition temperature of 143~ The repeat
unit of PEEK is
Unidirectional carbon tape is impregnated with one formulation of this thermoplastic by

ICI to form their APC-2 (Aromatic Polymer Composite 2) prepreg. Twenty plies of APC-
2 prepreg were molded in-house into laminates with a total thickness of 0.28 cm and length
of 23 cm. The compression molding was performed according to ICI specifications in a
matched-die, positive-pressure tool on a Tetrahedron MTP-14 24-ton press with dual open-
ings (one for quenching), and programmable force, heating, and cooling. The plaques
demonstrated a mean void content of 0.9 vol%. The enthalpy of melting for PEEK resin
in the composite panels was determined on a Perkin Elmer Differential Scanning Calorimeter
(DSC VII) and used to calculate percent crystallinity of the resin, where the enthalpy of
fusion was taken as 130 J/g [27]. The resulting crystalline weight fraction was found to be
0.26, where the analysis was performed as in Ref 28.
C-scan images were obtained with a Panametrics H Y S C A N ultrasonic scan system in

order to verify crack front straightness and lack of damage within the original panels. A
single 10-MHz spherically focused transducer (Panametrics V311-70753, 1.27-cm aperture,
2.54-cm focal length) was mounted on the programmable, two-axis scanner so that wave
propagation through water was perpendicular to the plane of the laminate. The gated peak-
height detector of the Panametrics 5052 U A ultrasonic analyzer was set to monitor the
reflection off the metallic tank bottom (pulse-echo geometry), and a gray-scale image of
peak attenuation was recorded on an electrical discharge Raytheon Line Scan Recorder with
slew rate synchronized to that of the two-axis scanner. No flaws, with the exception of crack-
starter inserts, were detected in any of these laminates.
Both sets of laminates were cut into 1.9-cm-wide specimens, with the fiber direction along
the length of the specimens, with a high speed saw using a diamond-impregnated disk blade.
A n interlaminar starter crack was introduced at the midplane of one end of each sample
during the lamination process by inserting a folded piece of Kapton polyimide (for the P E E K
samples) or Teflon PTFE (for the PMR-15 samples) film between the tenth and eleventh
plies. The folded end of the inserted film faced the interior of the sample and was slit with
a razor blade prior to testing. Because of the finite thickness of the inserted film, a resin-
rich area formed directly ahead of the film fold during the lamination process. Thus, it was
necessary to grow the crack a short distance past this region (using the W D D setup) before
starting the test,
The W D D testing was performed on an Instron Model 1331 servohydraulic test stand,
which was controlled and monitored by a Hewlett-Packard (HP) 9920 modular computer.
Programmed stroke control for constant or sequentially stepped displacement rates was
accomplished with an HP 6944A Series 200 Multiprogrammer containing the necessary
voltage digital to analog ( D / A ) , timer/pacer, and memory cards. Stroke, force, and tem-
perature data were sampled at the same rate that the displacement command was stepped,
which varied from 1.2 s/point for the slowest displacement rates to 1.2 • 10 -4 s/point for
the fastest rates. A Scan Control card and a 500-kHz analog-to-digital ( A / D ) card were
used to implement the acquisition of data from the three feedback channels. The loads were
measured with an Interface Super-Mini 222-N (50-1b) load cell and buffered on a 64K memory
card for subsequent transfer to the computer.
All testing was performed within an ATS Series 3170 environmental chamber, which was
temperature controlled to within 0.5~ Liquid nitrogen was used as the active coolant.
Temperature readings were made with a platinum resistance thermometer placed directly
on the sample. The load cell, which was outside of the chamber, was thermally isolated
from that part of the fixturing that was inside the chamber by a liquid-cooled, load coupling.
The crack front position was recorded with a Panasonic WV-1800 video camera and a
Panasonic 6300 VHS video cassette recorder. Canon optics (TV Zoom Lens, V6x16, 16-
100 mm, 1:19) were used. Crack length measurements were then obtained from the recorded
videotape with the aid of a F O R . A Model VPA-1000 Video Position Analyzer and a F O R . A
Model TG-160 Title Generator, both digitally interfaced to the computer using parallel-bit,
transistor-to-transistor-logic (TTL) signals. The time, as generated by the computer's real-
time clock, was imposed on the image during recording using the TG-160. During analysis,
the operator locates the crack tip with the VPA-1000, then transmits the position to the
computer over the interface. The time, as recorded on the image, is scrolled in via the
computer's keyboard. Figure 3 is a schematic of the interconnections. The flexural parameter
EI was measured for half of a fractured W D D coupon (see Fig. 2) at 25~ in three-point
bending. This value was found to be on average 0.616 Nm -~for the PMR-15 panels and 0.578
Nm z for the P E E K samples.
, load cell
oven
CCTV c
T serv~ actuator
I .Iv,deo- ! [rnu,t,- I
-]cassette I ]programmer]
L ,Irecorder [ [ I
I I vnoeo
I .
I L.-I bile
. . ~ 7I I c
mt/r~176176
pos,hon ~1 I generator I I I
|analyzer 11[ J [ "Y
FIG. 3--Schematic of WDD testing configuration.
Mode 1 W D D Test Results

PMR-15 / Carbon Fiber Laminates
W D D samples cut from the PMR-15/CF laminates were tested at displacement crosshead
rates ranging from 0.0254 to 25.4 mm/s and temperatures between 25 and 200~ Each
sample was tested at a constant temperature. In some cases, the crosshead displacement
rate was held constant over four or more consecutive 12.7-mm sections, while in others, the
testing rate was increased by one order of magnitude for each successive 12.7-mm section.
Typical mode I W D D data are shown in Fig. 4 for an actuator displacement rate of 0.0254
mm/s and a temperature of 200~ For W D D samples of constant width, Eq 2 shows that
the interlaminar fracture driving force, G~, is directly proportional to the instantaneous load
values, hence the time dependence of G~ is easily obtainable from the load trace, as shown
in Fig, 4a. In this figure, G[ is observed to increase steadily for about 200 s, then level off,
while the corresponding crack growth data show little crack growth during the first 200 s,
with subsequent steady crack growth. Crack length in Fig. 4b corresponds to dimension L
in Fig. 2.
Because of the moving sample/stationary wedge nature of the W D D experimental ar-
600I"l'"l'"'rl''''rrl''llrl'A ~ ,,, t,,~, i,, ,, i, ,,, i,, ,~

-- B 9 Calculated
o Msasured
G 500 ooooOoOoOOOOOOOooooOoooof
ooo~176 9o
400IoOOOo~
o * e*ae~
LLIkllllllllrllldlllllllaO0 500
I00 200 300 5
I I I I
TIMEisec~ rIMEIsec)
FIG. 4--(a) Typical G, data for PMR-15/CF composites using the mode I WDD testing
geometry. (b) Calculated and measured crack length for the same sample as in Fig. 4a.
rangement, steady crack growth is associated with a constant crack tip position in the
laboratory reference frame. Variations in the value of G[, a material variability, as indicated
by variations in the time behavior of the load trace, are associated with deviations of the
crack velocity from the machine crosshead velocity. This can be seen from the dependence
of load on crack length L in Eq 5.
Defining C as the length of the uncracked portion of the beam
dL dC d~
d--7 + ~ + ~= L + (? + ~ = 0 (6)
as can be seen in Fig. 2. In the present cases the actuator speed ~ is a constant, go.
Because the data for the initial 200 s of Fig. 4b correspond to initial crack loading with
no crack growth (C = 0), the crack length dependence L is L = - go- Excellent agreement
can be seen in Fig. 4b.
Beyond the initial 200 s, steady crack growth conditions are attained ((? = - g o ) and the
crack tip position remains nearly stationary in the lab frame (L = 0). This is clearly seen
in Fig. 4b, where the crack tip position fluctuates about an average position of 59 mm at
times greater than 200 s.
Using Eq 5 and values of P / E I relevant to the split laminate under test, the crack tip
positions were calculated and compared to the measured values in Fig. 4b. These curves
show good agreement. Conversely, the measured crack tip positions could be used in con-
junction with Eqs 4 and 5 to calculate the crack driving force Gv The form of Eq 5 suggests
that small errors in the measurement of L will result in large errors in calculated values of
G~ if L rather than P is measured and used in the calculations.
The results of some of the mode I W D D tests on the PMR-15/CF unidirectional laminates
are shown in Figs. 5 and 6. These curves have been drawn through 40 G~ values without
smoothing. Values of G[ were obtained by averaging over 80 successive data measurements
and correspond to a total of 1.27 mm of crosshead displacement. Both figures show data
taken at five different temperatures between 25 and 200~ The data in Fig. 5 were taken
at a constant crosshead displacement rate of 2.54 mm/s over the full 50 mm of data presented,
while Figs. 6a and 6b show data taken at four different rates, increasing from 0.0254 to 25.4
mm/s in order-of-magnitude increments. In Fig. 5, steady crack growth behavior was ob-
served. The fluctuations in G~ values were probably caused by local variations in crack front
processes, such as fiber bridging and crack branching. The increase in G~ observed during
GLESSNER ET AL. ON M O D E I I N T E R L A M I N A R FRACTURE TOUGHNESS 189
r
r I ~ I ~ [ '
0.6 -- 0.254 -
cm/sec
0.5 200 ~
Gi ~ f~ ...
(kJim2) 0.4 ~ V ' ~ " / ' - / ~"'---'-'-~-\ .,,/1-00
r~ . . . . . ~ " , . ~ . j t X
02 "
1.27 2.54 3.81 ~08

STROKE (cm)
FIG. 5--Temperature dependence of Gr for PMR-15 / CF laminates.
the first 1 cm reflects the initial loading period, when no crack growth is occurring. Once
steady growth commences, GI is relatively constant.
Because of the relative constancy of G,, the interlaminar fracture toughness Glc was taken
as the average value. This value of Gk reflected the strain energy release rate during mode
I constant velocity steady state crack propagation.
Typical WDD data for two different laminates are given in Figs. 6a and 6b. The data
show evidence of a slight rate dependence at higher temperatures and a much stronger
temperature dependence, with Glc increasing at higher rates and higher temperatures. The
estimated rate dependence (approximately 10%) is smaller than the sample-to-sample scatter
(approximately 20%), which makes the existence of a rate dependence somewhat uncertain.
However, when testing is performed at successive crack velocities on a single specimen, the
scatter in the data (reflecting local crack growth fluctuations) is relatively small, and the
rate dependence appears real. This can be seen for the five samples shown in Figs. 6a and
6b.
I I I I I I I 0,00254 0.0254 0.254 2 54

em/sec
Q; =ire n_ _I_
!
0,6
06
GI Gi 200 ~
(kOlm 2) (kJImZ) 0 5
0.5
- - -- 100 ..
o . . . , ' ..... ...%.. , ' ' " -..' "...... 9 ..... / .... ..-'. ... ..'"-">'"+"
~.. ..................... '-........ Yo.........."
_l_ =I= _I_
0,3 J~- 0,00254 ' 0,0254 0.254 ' 254
cm/sec
0.2 I I
1,27
I I
254
I I
3.81
I
508
__ a_ . . . . 1 _ ~ ........ l_ , I .......... L .......
12 7 2,54 3.81 5.08
Stroke (cm) Stroke (cm)
FIG. 6--(a) Temperature and rate dependence of G~for PMR-15/CF laminates: Plaque I.
(b) Temperature and rate dependence of GI for PMR-15/ CF laminates: Plaque 2.
The temperature dependence is quite evident, however, with G;~ increasing by 50% from
390 to 570 J/m 2, when going from 25 to 200~ at 25.4 mm/s. The corresponding values at
0.0254 mm/s are 340 and 500 J/m 2, respectively. The large sample-to-sample scatter prevents
us from making a definitive statement about the absolute temperature dependency, however.
PEEK~Carbon Fiber Laminates

The mode I W D D interlaminar fracture toughness of P E E K / C F unidirectional laminates
was measured at crosshead rates ranging from 0.0254 to 25.4 mm/s and temperatures between
- 2 5 and 120~ Each sample was tested at constant temperature. As with the PMR-15/CF
composites, the crosshead rate was held constant throughout the test in some samples, while
in other samples, the testing rate was increased by one order of magnitude for each successive
25.4-mm section.
A strikingly different mode I interlaminar crack growth behavior pattern was observed
for the P E E K / C F unidirectional composites. Instead of steady crack propagation as seen
in the PMR-15/CF composites, repeated stable/unstable/arrest crack growth behavior was
observed, as shown in Fig. 7. This stick/slip pattern is characterized by periods of crack
arrest and very rapid crack jumps. Crack arrest is enabled by the inherently stable test
geometry of the WDD samples.
Mode I W D D results shown in Fig. 7 are for a sample tested at 0.508 mm/s and 25~
The initial increase in G[ represents the sample loading period. Once steady crack growth
commenced, G[ was approximately constant. The sudden transition to unstable, fast fracture
appears as a sudden drop in G~. The amount of steady crack growth before instability was
not constant, and the crack often progressed to fast fracture with no prior steady growth.
Crack lengths were calculated using measured G, values and Eq 5 for a laminate tested
at 25~ and 0.254 mm/s. Both the measured and calculated values of crack tip positions are
plotted in Fig. 8. Excellent agreement is observed between the two curves. At high load
values (short crack lengths), disparity between measured and calculated crack lengths must
certainly result from the trigonometric linearization implicit in Eq 5. More accurate pre-
dictions would result from the use of Eqs 1 and 2.
Unlike the case for PMR-15/CF laminates, where the interlaminar fracture toughness G,,
was taken as the average value of G~ during steady crack growth, two different values of
G, were chosen to characterize the interlaminar fracture toughness of the P E E K / C F lam-
inates. The two values chosen were the value of G[ at the onset of the crack instability,
designated Gu (for unstable), and the value of G, at crack arrest, GA (for arrest).
A summary of the mode I WDD P E E K / C F interlaminar fracture data is shown in Fig.
9. A very strong temperature dependence is observed for the Gu values, with Gv rising
rapidly above 70~ that is, as the glass transition temperature of PEEK (143~ is ap-
whiter zones on frocture surfoce

GI ( j / M 2 } / 9 ~ ~
.ooo
3000L
I000
" '
e~ , I , , I , , , , , , ,
-o 2.54 506 762 10.16
STROKE (CM)
25 ~ 0.0508 CM/S
FIG. 7-- Typical G, data for PEEK/CF composites using the mode 1 WDD testinggeometry.
' I ' I ' I ' I ' I ' 1 ' I ' I ' I '
5
9 Calculated
[] Measured
o
4 060
-i-
o;:0oo:;
I-
{..9
z SlOe -O~nlo $o
w
.J
rr __ O02 o
{O
2 ' I , I , I , 1 , I , _1 , I , I , I ,
10 20 30 40 50 60 70 80 90 100
TIME (sec)
F I G . 8--Calculated and measured crack length for a P E E K / CF sample tested at 25~ and
O.0254 cm/s.
proached. There is nearly a 120 ~ increase in Gu from about 2.2 k J / m 2 at - 2 5 ~ to 4.9

kJ/m 2 at 120~ Temperature, on the other hand, appears to have little effect on the
arrest values GA (within the range of temperatures tested).
Rate has little effect on either Gu or GA over the rates and temperatures tested. The
variation in G~, or GA (at constant T) falls within the experimental uncertainty. Apparently,
however, rate and temperature affect the length of stable crack growth occurring before
' ' ' I ' ' ' I ' ' I I '
9 O 0.00254 c m / s
9 o 0.0254
9 D 0.254
5 /I, ~ 2.54
l
9 O 25.4
G i (k j / m 2 ) 3
2
0-
1
O o ARREST
0 I , ~ I , , , I , , , I , , , I ,
- 40 0 40 80 120
TEMPERATURE (~
FIG. 9--Summary of mode 1 WDD PEEK/CF interlaminar fracture data (solid symbols
are G~ values; open symbols are GA values).
instability. High rates and low temperatures favor little or no stable growth between slip
instabilities, while slow rates and higher temperatures favor longer distances of stable growth.
While several cases of minimal stable crack growth were seen at low rates and high tem-
peratures, no cases of long stable growth processes were observed at fast rates and low
temperatures.
Fractography
P M R - 1 5 / Carbon Fiber Laminates
Because of the steady mode I interlaminar crack growth behavior of PMR-15/CF uni-
directional composites, the overall fracture surface appearance was macroscopically quite
uniform. On a finer scale, however, two distinguishing features were evident. First, fiber
bundle pullouts could be observed at scattered locations, some extending for long distances
(several centimetres) along the length of the fracture surface. The fiber pullouts ranged from
individual fibers to bundles several millimetres wide. During the fracture process, these
fibers bridged the two fracture surfaces, thus increasing the resistance to crack propagation.
The multiple cracking which accompanies fiber bundle pullouts results in an increase in
fracture surface area and hence an increase in the apparent interlaminar fracture energy of
the composite.
The second observation was made on metallographically polished cross sections of the
FIG. lO--Optical micrograph of a metallographically polished cross section of a PMR-15

laminate showing the tortuous crack path and ply structure with crack direction into the plane
of the rnicrograph (arrow indicates fiber bundle pullout).
FIG. 11--Temperature dependence o f PMR-15 / CF laminate cross sections showing an in-

crease in fiber bundle pullout at higher temperatures.
sample (plane normal to the crack growth direction), as shown in Fig. 10. The crack path
front is not straight. A close examination of the interlaminar ply structure shows that the
individual plies themselves are, in fact, not perfectly straight either (a result of the lamination
procedure for aligned plies). The resultant crack front tortuosity reflects the imperfect ply
structure, as well as other local fluctuations in the crack advance process, such as multiple
cracking, fiber bundle pullout, and local defects. (An example of a fiber bundle pullout has
been indicated in Fig. 10.)
Fiber bundle pullouts were observed to be more prevalent at higher temperatures in the
PMR-15/CF laminates (Fig. 11). These results may contribute to the increased interlaminar
fracture toughness seen at higher temperatures.
On a microscopic scale, the fiber-matrix behavior during interlaminar fracture can be
examined from scanning electron micrographs of the fracture surface. Some dilational de-
formation of the resin can also be observed, as shown in the micrograph of Fig. 12. This
deformation is much less pronounced than that seen in the more ductile thermoplastic resins,
as discussed below for the P E E K / C F laminate.
P E E K ~ C a r b o n Fiber Laminates
The appearance of the fracture surface for the P E E K / C F laminates was considerably
different from that of the PMR-15/CF laminates, as might be expected due to the stick/slip
crack growth behavior observed in the PEEK laminates and the much higher crack growth
resistance. The most dramatic difference is the alternating light and dark regions on the
FIG. 1 2 - - S E M micrograph o f a ~,pical PMR-15 fracture surface at IO0~ and 2.54 • 10 -3

cm/s.
fracture surface, which correspond to stable and unstable growth behavior, respectively. In
Fig. 7 the lines above the Gt trace indicate the lighter portions of the fracture surface along
the 100 mm of crack advance. The light fracture surface regions correlate very well with
the stable crack growth regions of constant GI, where the crack velocity is nearly equal to
the crosshead velocity. The dark regions thus represent the fracture surfaces created during
the unstable crack jump periods.
Scanning electron micrographs of the fracture surface indicate a very sharp transition
between the regions of stable and unstable growth, as shown in Fig. 13. This transition is
equally distinct at the point of instability (slow to fast, Fig. 13a) and the point of arrest (fast
to slow, Fig. 13b). Surprisingly, the two transitions would be nearly indistinguishable if not
for the viewer's prior knowledge of the crack growth direction. The only distinguishing
feature is a slight shift in the skew of the drawn ligaments that belies the crack growth
direction [3]. The electron micrographs in both the stable and unstable fracture surface
regions exhibit considerable resin retention on the fiber surfaces. Ductile flow of the resin
is quite evident in the presence of tufts on the fracture surface in both regions. Perhaps the
distinguishing feature between the two regions is the extent of the deformation process that
occurs. The steady crack growth region is characterized by a high intensity of drawn resin
extending nearly a fiber diameter away from the surface. This causes considerable light
scattering off the surface and a resultant whitened appearance. The unstable fast fracture
region, on the other hand, reveals more moderate deformation. The reduced light scattering
gives the fracture surface a more darkened appearance.
Similar to the PMR-15/CF laminates, fiber bundle pullout and crack path tortuosity were
observed in all samples.
Discussion: lnterlaminar Crack Growth Behavior
Interlaminar Fracture Toughness

The G,c results that we have obtained for the PMR-15/CF laminates (370 J/m 3 at 25~
compare quite favorably with those observed for other (nonrubber-toughened) thermosetting
resins (for example, various epoxy/CF and epoxy/GF laminates) with reported values rang-
ing from 80 to 356 J/m 2 [1,4, 7,8,11,12,16]. Our values also show good agreement with those
obtained (400 J/m 2) using the conventional DCB testing geometry on similar PMR-15/CF
laminates/
The PEEK/CF laminate results for Gu (2.5 kJ/m 2 at 25~ agree very well with those
obtained by Leach et al. [19], but are somewhat higher than those obtained for P E E K / C F
laminates by Gillespie et al. [7,16], who report average values of 1.75 kJ/m 2. Differences
in fabrication techniques, as well as the strong fiber volume fraction dependence reported
for PEEK laminates [7,16], may be responsible for the difference. These mode I interlaminar
fracture toughness values compare favorably with other thermoplastic laminates, such as a
polyamide/CF laminate, with a reported value of 2.73 kJ/m 2 [8].
A correlation between neat matrix resin toughness and composite interlaminar fracture
toughness has been reported by Hunston [4], who observed that the more brittle thermo-
setting resin-based laminates were tougher than their corresponding neat matrix resins,
whereas the tougher thermoplastics (and toughened thermosets) were tougher than their
corresponding laminates. Because the neat PEEK resin has a reported Glc value of greater
than 4.8 kJ/m 2 at 23~ [29], our P E E K / C F results agree qualitatively with the Hunston
correlation.
4 D. Ward, General Electric Aircraft Engine Business Group, private communication, 10/86.
FIG. 13--(a) SEM micrograph of a PEEK/CF composite at the initiation of fast fracture
(crack growth from right to left) at 70~ and 2.54 • 10 ~cm/s (actuator speed).
Crack Tip Process Zone

Several models [7-9] have been developed to account for the fracture surface appearance,
especially the presence of the tufts. In most models, it is generally believed that the con-
straints imposed by the rigid fiber layers lead to cavitation in the matrix or the fiber-matrix
interfacial region ahead of the crack front within a zone called the process zone.
The process zone is distinct from the commonly described plastic zone in that damage in
the process zone is composed of discrete regions of cavitation or voiding rather than a more
continuous zone of plastic flow. As the crack advances and the two fracture surfaces are
pulled apart, the resin regions between adjacent voids stretch, leaving tufts of drawn material
on the resultant fracture surfaces.
The P E E K / C F laminates exhibit high interlaminar fracture strength caused by localized
but intense dilation in the process zone. These values are nonetheless reduced from the neat
resin values because of the severe constraints placed on deformation by the noncontributory
fiber reinforcement. Lee [30] has presented an analogy with adhesive bond strength to
explain this behavior. Adhesive bond strength is observed to attain optimum values when
the adhesive bond layer thickness is roughly equivalent to the plastic zone size that would
occur in the neat adhesive material. Slightly reduced strength is observed for thicker adhesive
layers, while considerable reduction in adhesive strength is seen for thinner bond layer
thicknesses as a result of the severe restriction of the plastic deformation. Treating the
FIG. 13--(b) SEM micrograph of a PEEK/CF composite at the arrest of fast fracture (crack
growth from left to right) at 70~C at 0.254 cm/s (actuator speed).
interlaminar region as an adhesive bond layer, Lee surmises that composites based on ductile
thermoplastics, such as PEEK, with large neat resin plastic zone sizes, should be expected
to have reduced interlaminar fracture toughness values, while those based on less ductile
or brittle thermosets, such as PMR-15, with considerably smaller plastic zone sizes, are not
expected to be adversely affected. While Lee's analogy may ignore the severe anisotropic
constraints imposed by the fibers, the principal concepts are applicable.
Interracial Adhesion
Another important factor in determining the nature of the process zone damage is the
strength (or weakness) of the matrix-fiber interface. Optimum interlaminar toughness may
sometimes require less than perfect adhesion, as researchers [31,32] have observed that
controlled intermittent (partial) bonding can lead to enhanced fracture toughness over full
bonding along the fiber length. Weak bond regions can increase the amount of energy
absorption through processes such as crack front bifurcation and secondary crack propa-
gation. The strong bond regions are necessary for integral (undamaged) laminate stiffness
and strength.
Adhesion in PMR-15/CF laminates was sufficient to enable good interlaminar fracture
toughness values when compared to other thermosetting resins.
Fiber Bridging, Multiple Cracking

On a larger scale, mechanisms that affect the path of the growing crack or the density of
secondary crack formation in the process zone play an important role in determining the
measured interlaminar energy, particularly where matrix resin toughness is intrinsically low.
The nesting or intermingling of fibers between plies, for example, acts as an obstacle, causing
the crack to deviate from the plane, and necessitating local mode III crack propagation.
These processes are crack area and crack energy intensive, thus increasing the apparent
fracture strength.
Fibers or fiber bundles may also bridge the fracture plane, increasing the energy necessary
to propagate the crack. This bridging of fibers or bundles can be caused by secondary crack
formation on adjacent weak planes (with the bridging fibers in between). For the crack to
cross from one plane to another, the fiber bundle must bridge the crack. Examples of fiber
bundle pullouts can be seen in Fig. 12. Both P E E K / C F and PMR-15/CF laminates showed
fiber bundle pullout and bridging.
Stick~Slip Crack Growth

The onset of fast fracture during stick/slip propagation in the P E E K / C F laminates oc-
curred rather abruptly, as viewed in the load traces (Fig. 7) and on the fracture surfaces
(Fig. 13). As seen in Fig. 7, the transitions from stable to unstable growth are randomly
spaced. The nucleation of such transitions is related undoubtedly to some feature of the
microstructure, which is itself spaced randomly. While certain researchers have postulated
sequences of events leading to the transition [7,16], none to our knowledge have accounted
for its stochastic nature. Critical imperfections may cause the fast fracture to initiate, thus
accounting for the larger scatter of the Gu values (approximately 50%) compared to those
seen for the GA values (approximately 30%). This may also explain the absence of steady
growth sometimes observed at low rates and high temperatures, where steady crack growth
behavior is expected. The arrest values are not expected to be so dependent on imperfections.
Visually, however, it is difficult to detect any evidence for such a defect-dominated process
on the fracture surface (see, for example, Fig. 13). Thus, the mechanisms responsible for
the onset of fast fracture are not clearly understood.
Thermoplastic versus Thermoset Resin

Thermoplastic resin composites usually exhibit stick/slip behavior, as observed in PEEK/
CF and polyamide/CF [8] laminates. Unidirectional composites with untoughened ther-
mosetting resins, however, usually exhibit steady crack growth in interlaminar fracture
testing. This is particularly true for epoxy laminates [7], as well as for the PMR-15/CF
laminates reported herein.
At controlled crack velocities, composites from ductile resins exhibit a large crack tip
process zone, which facilitates stable crack growth at high G~ values. The "'slip" behavior
occurs at some critical moment, perhaps associated with the presence of some defect. The
defect, manifested as a decrease in crack growth resistance, causes the release of strain
energy into kinetic energy (essentially inertia-dominated behavior). At high crack velocities,
the process zone is reduced in size (viscoelastic behavior of thermoplastics, akin to a ductile/
brittle transition), and the crack grows against a much diminished resistance. It is only the
stabilizing effect of test geometry in the W D D (or DCB) experiment that brings the crack
to a halt, reactivating the large process zone. Of course, some randomly spaced local high
crack resistance regions may enhance crack arrest. Composites based on highly networked
resins, like epoxies, which are very defect sensitive, even at the slower laboratory testing
rates, show very little tendency toward the formation of ductile process zones. Rather than
the essentially two-state behavior demonstrated by their tougher analogs, these materials
exhibit steady crack growth.
Conclusions
A novel adaptation of the wedge-driven delamination (WDD) technique [22] has been
implemented to study the interlaminar fracture behavior of unidirectional carbon fiber
composites based on a thermosetting resin, PMR-15, and a thermoplastic resin, PEEK. This
WDD technique allowed a direct measurement of the interlaminar fracture toughness, which
bypassed some of the tedious calculations or measurements required when using the double
cantilever beam approach. Furthermore, this novel W D D testing methodology provided
direct crack velocity control and a stationary crack tip with respect to the laboratory reference
frame.
The PMR-15/CF laminates showed higher interlaminar fracture toughness values than
those reported for various untoughened epoxy laminates. Higher values were observed at
higher temperatures, and a slight indication of higher toughness at higher rates could be
perceived. The PMR-15/CF laminate, typical of other thermoset-based composites, derived
much of its interlaminar fracture strength from its composite nature. Although increased
resin toughness at higher temperatures may play a role, multiple cracking, fiber pullout,
fiber bridging, and crack path tortuosity contributed heavily to composite toughness. At
high temperatures these obstacles to crack growth appeared to be enhanced, thus leading
to higher toughness values. Fiber nesting and intermingling of plies also contributed, thus
making the interlaminar fracture strength quite dependent on the laminate fabrication pro-
cedures.
The P E E K / C F laminates demonstrated very high interlaminar fracture toughness values,
but with a stick/slip type of crack propagation. The (higher) values at instability showed a
significant temperature dependence, whereas the (lower) arrest values showed negligible
dependence. Neither showed any noticeable rate dependence within the range tested. For
the P E E K / C F laminates, a typical thermoplastic composite, the severe constraints imposed
by the fiber layers greatly reduced the extent of the process zone. The strong temperature
dependence of the crack growth resistance during stable propagation reflected the strong
temperature dependence of the neat resin dilational flow process. The mechanisms for the
onset of crack instability are stochastic and still unclear.
References
[1] Keary, E E., Ilcewicz, L. B., Shaar, C., and Trostle, J., Journal of Composite Materials, Vol. 19,
March 1985, pp. 154-177.
[2] Bascom, W. D., Bitner, J. L., Moulton, R. J., and Siebert, A. R., Composites, Vol. 11, Jan. 1980,
pp. 9-18.
[3] Bascom, W. D., Boll, D. J., Fuller, B., and Phillips, P. J., Journal of Materials Science, Vol. 20,
1985, pp. 3184-3190.
[4] Hunston, D. L., Composites Technology Review, Vol. 4, 1984, pp. 176-180.
[5] Hartness, J. T., SAMPE Journal, Vol. 20, 26 Sept./Oct. 1984, pp. 26-31.
61 Donaldson, S. L., Composites, Vol. 16, April 1985, pp. 103-112.
Gillespie, Jr., J. W., Carlsson, L. A., and Smiley, A. J., Composites Science and Technology, Vol.
28, 1987, pp. 1-15.
[8] Su, K. B., "Mechanisms of lnterlaminar Fracture in a TP Matrix Comp Laminate," Fifth Inter-
national Conference on Composite Materials, San Diego, CA, 1985, p. 995.
[9] Bradley W. L. and Cohen, R. N., in Delamination and Debonding of Materials, ASTM STP 876,
410.
[10] Lee, S. M., Journal of Composite Materials, Vol. 20, March 1986, pp. 185-196.
[11] Wilkins, D. J., Eisenmann, J. R., Camin, R. A., Margolis, W. S., and Benson, R. A., in Damage
in Composite Materials, ASTM STP 775, K. L. Reifsnider, Ed., American Society for Testing and
[12] Devitt, D. E, Schapery, R. A., and Bradley, W. L., Journal of Composite Materials, Vol. 14, Oct.
1980, pp. 270-285.
[13] Han, K. S. and Koutsky, J., Journal of Composite Materials, Vol. 15, July 1981, pp. 371-388.
[14] Whitney, J. M., Browning, C. E., and Hoogsteden, W., Journal of Reinforced Plastics and Com-
posites, Vol. 1, Oct. 1982, pp. 297-313.
[15] O'Brien, T. K., Johnston, N. J., Morris, D. H., and Simonds, R. A., SAMPE Journal, Vol. 18,
July/August, 1982, pp. 8-15.
[16] Gillespie, J. W., Jr., Carlsson, L. A., Pipes, R. B., Rothschilds, R., et al., "Delamination Growth
in Composite Materials," NASA Contractor Report, NAG-I-475, 1985.
[17] Scott, J. M. and Phillips, D. C., Journal of Materials Science, Vol. 10, 1975, pp. 551-562.
[18] Richards-Frandsen, R. and Naerheim, Y., Journal of Composite Materials, Vol. 17. March 1983,
pp. 105-113.
[19] Leach, D. C., Curtis, D. C., and Tamblin, D. R. in Toughened Composites, ASTM STP 937,
American Society for Testing and Materials, Philadelphia, 1987.
[20] Kasen, M. B., Journal of Composites Technology and Research, Vol. 8, No. 3, Fall 1986, pp. 103-
106.
[21] Johannesson, T., Sjoblom, P.. and Selden, R., Journal of Materials Science, Vol. 19, 1984, pp.
1171-1177.
[22] Vallance, M. A., Gifford, S. K.. Glessner, A. L., and Takemori, M. T., to be published.
[23] Landes, J. D. and Begley, J. A., in Fracture Toughness, ASTM STP 514, American Society for
[24] Brown, W. F., Jr. and Srawley, J. E., in Fracture Toughness Testing Methods, ASTM STP 381,
[25] Serafini, T. T., Delvigs, P., and Lightsey, G. R., Journal of Applied Polymer Science, Vol. 16,
1972, pp. 905-915.
[26] Harper, J., Whittenberger, J. D., and Hurwitz, E I., Polymer Composites, Vol. 5, July 1984, pp.
179-185.
[27] Dawson, P. C. and Blundell, D. J., Polymer, Vol. 21, May 1980, pp. 577-578.
[28] Vallance, M. A. and Tomkinson-Walles, G. D., "Linear Viscoelastic Response of High Perform-
ance Thermoplastic Composites," Proceedings, Thirty-first International SAMPE Symposium and
Exhibition, Las Vegas, 1986, pp. 410-419, SAMPE, Covina, CA.
[29] Jones, D. P., Leach, D. C., and Moore, D. R., Polymer, Vol. 26, Aug. 1985, pp. 1385-1393.
[30] Lee, S. M., Journal of Composite Materials, Vol. 20, March 1986, pp. 185-196.
[31] Mai, Y. M. and Castino, E, Journal of Materials Science, Vol. 19, 1984, pp. 1638-1655.
[32] Atkins, A. G., Journal of Materials Science, Vol. 10, 1975, pp. 819-832.
Carlos R. Corleto I and Walter L. Bradley I
Mode II Delamination Fracture

Toughness of Unidirectional
Graphite/Epoxy Composites
REFERENCE: Corleto, C. R. and Bradley, W. L., "Mode I1 Delamination Fracture Tough-

ness of Unidirectional Graphite/Epoxy Composites," Composite Materials: Fatigue and Frac-
ture, Second Volume, ASTM STP 1012, Paul A. Lagace, Ed., American Society for Testing
ABSTRACT: The mode II delamination fracture toughness of a ductile and a brittle unidi-
rectional graphite/epoxy composite has been studied using the end-notched flexure (ENF)
test and the end-loaded split laminate (ELS) test. The stress field in the vicinity of the crack
tip of a split laminate beam under mode I and mode II (ELS test) conditions has been
determined by means of a finite-element analysis. Also, the micromechanics of mode lI fracture
have been studied during in situ and postmortem observations of the fracture process. Both
the ENF and ELS tests give similar values for Gn,. However, because elastic material behavior
is assumed in the analysis, the G,c results for a ductile composite are somewhat uncertain
because permanent deformation is observed. The ELS test provides a pure shear stress state
in the vicinity of the crack tip. The formation of hackles in composites made with brittle resins
provides a more tortuous path for the crack leading to an increased resistance to delamination
under mode I1 conditions compared to mode [. However, extensive resin deformation and
yielding play a more significant role in the fracture resistance for mode II loading of composites
made using a rubber-toughened epoxy.
KEY WORDS: graphite/epoxy, delamination, fracture toughness, failure analysis, mode II,
finite-element analysis
O n e of the most important factors hindering the use of fiber-reinforced composite materials
in structural applications is their inherently p o o r damage tolerance for delamination. The
resistance of composites to delamination can be well characterized by the delamination
fracture toughness, measured as energy dissipated per unit area of crack growth. W h e r e
behavior is essentially linearly elastic, the fracture toughness can be measured as an energy
release rate, G, of the material. Delamination in composites can occur due to tensile stresses
(mode I), in-plane shear stresses (mode II), and out-of-plane tearing stresses (mode III).
As a result, a complete understanding of these delamination processes is needed to properly
design composite structures and develop materials with improved fracture toughness char-
acteristics. Several areas must be addressed to thoroughly understand delamination. First,
a critical assessment of current methods used to measure resistance of composites to delam-
ination is needed to verify the validity of the analysis for the various test configurations.
Furthermore, the geometry independence of the measured energy release rates must be
established by comparing measured values for different test geometries. Finite-element
Graduate research assistant and professor, respectively, Department of Mechanical Engineering,

Texas A&M University, College Station, TX 77843.
201
Downloaded/printed
Copyright9 by
analyses may be used to determine the state of stress at the crack tip where beam theory
does not give details about the stress field. Second, a study of the physical mechanisms
involved during the fracture process must be undertaken by means of in situ and postmortem
fracture observations. Investigations of this nature should help in the understanding of the
relationship between the macroscopic resistance to crack propagation and the microme-
chanisms of delamination fracture.
Recent investigations indicate that increasing mode II loading, particularly for brittle
systems, results in a significantly greater resistance to crack propagation as measured by the
total energy release rate required to propagate the crack [1]. This indicates the need to
thoroughly understand mode II delamination in order to properly characterize this mode
of failure.
Several test configurations have been proposed for measuring the resistance to delami-
nation under mode II loading including the end-notched flexure (ENF) test [2] and the end-
loaded split laminate (ELS) test [3]. The end-notched flexure test configuration can be seen
in Fig. la. This test consists of a three-point bend loading configuration for a split laminate
specimen. The ELS test configuration, shown in Fig. lb, also uses a split laminate beam
specimen loaded at the cracked end and fixed at the opposite end. Both the ENF and ELS
test methods evaluate G.c from equations derived using linear beam theory and assume
linear elastic material behavior [2,3]. A current review of the literature indicates that both
test configurations are being used currently to measure G.c [2-4].
During a recent finite-element analysis [5], the ENF test was found to be a pure mode
II fracture test within the constraints of small deflection theory. The study also revealed
that the interlaminar normal stress is identically zero along the beam center line, and the
interlaminar shear stress exhibits the expected singularity. In the case of the ELS test, a
study similar to the one reported in Ref 5 is still needed.
W2 1:)/2
(a)
P=2pa
(b)
FIG. l - - M o d e 11 test configurations: (a) ENF test configuration and associated parameters;
(b) ELS test configuration and associated parameters.
CORLETO AND BRADLEY ON MODE II DELAMINATION FRACTURE TOUGHNESS 203
As pointed out earlier, under increasing mode II loading, particularly for more brittle
systems, a significantly greater resistance to crack propagation has been observed. Post-
mortem observation studies in scanning electron microscopy (SEM) indicate that this increase
in fracture toughness is the result of the development of hackles on the fractured surfaces
[1]. However, understanding the hackle formation process as well as damage zone devel-
opment ahead of the crack tip requires further in situ and postmortem observations of
delamination in the scanning electron microscope.
The objectives of this investigation have been (1) to measure the mode II delamination
fracture toughness of a ductile and a brittle graphite/epoxy composite using the ENF test
and the ELS test and to compare the results, (2) to determine the stress field in the vicinity
of the crack tip of the ELS specimen under mode II and mode I loading conditions by means
of a finite element analysis, and (3) to study the micromechanics of fracture using real-time
and postmortem observations of the fracture process and correlate them to the macroscop-
ically obtained energy release rates for composites made with brittle and ductile resins.
Analysis
End-Notched Flexure ( ENF) Test
The ENF test consists of loading a split laminate beam specimen using a three-point bend
fixture. Figure la shows the ENF test configuration and associated test parameters. Cal-
culation of the strain energy release rate G,, assuming linear beam theory and linear elastic
material behavior, gives the following result [6]
9p2a2C
G. - (1)
2b(2L 3+ 3a 3)
where
a = the initial crack length,
b = the width,
C = the measured compliance,
L = half-span length, and
P = the applied load at the center pin.
The strain energy release G~I can also be obtained by means of an experimental method for
determining the relationship between compliance and crack length. In this case, GI, is given
by [6]
3p2ma2bo
G,l = 2b---'--7~ (2)
where m is obtained experimentally by measuring the compliance as a function of crack

length on one of the specimens whose width is bo and fitting the data to the relationship
(determined from linear beam theory)
C = ma 3 + constant (3)
End-Loaded Split Laminate ( ELS) Test

The energy release rate for pure mode II in this case is determined by asymmetrically
loading a split-laminate beam specimen. Figure ib shows the ELS test configuration and
associated test parameters. Linear beam theory and linear elastic material behavior are
assumed by the analysis. A complete explanation of this analysis can be found in Ref 3.
From this analysis, Gx~ can be expressed by
9P2a 2
GII - 4bZEllh3 (4)
P is the load applied at the cracked end of the beam and h is half the specimen thickness
(see Fig. lb). The axial modulus E . also can be shown to be
P
EH - 2 b ~ h 3 (3a 3 + L 3) (5)
where
= the total beam tip deflection.
Area Method
In all the data reduction methods previously described, the critical energy release rate
was calculated using the measured critical load Pc at the onset of crack growth and the crack
length 'a' prior to crack extension. Therefore, G,c is a measure of the energy required to
initiate crack propagation. However, as the crack propagates, Gttc does not necessarily
remain constant, in which case a propagation and arrest value of GH, can arise. In the area
method, an alternate approach for evaluating fracture toughness, GI~ is interpreted as the
energy required to create a new cracked surface area. Gtt can then be given by
U
Gn = bAa (6)
where U is the area between the load-deflection curve for loading and unloading of a very
small change in crack length, Aa (see Fig. 2). An important advantage of this method is
that only elastic material behavior is required to predict G~. Therefore, for geometrically
6
FIG. 2--Alternate interpretation o f G.
I t
69.85 88.9
I I
70.485 76.2 77.47
FIG. 3--Finite-element mesh.
nonlinear and/or nonlinear elastic material responses, this method gives an average energy
release rate for the observed crack extension Aa. For unstable crack growth, it gives an
average value for GHc which typically falls between the initiation and arrest values measured
for linear behavior. The load displacement record should return to the origin to guarantee
that no significant far field deformation or damage is included in U in Eq 6. The inclusion
of energy dissipation in far field damage in the fracture energy term U in Eq 6 would give
an erroneously high estimate of Gnc. For materials in which GH is independent of crack
growth rate and crack growth distance (that is systems with minimal fiber bridging and/or
plastic wake), the average and initiation G, should be identical. Furthermore, where slow
stable crack growth occurs, the average G~iccalculated from the area method should equal
the Guc for initiation calculated from linear beam theory. Another advantage of this method
is that G~ is obtained without knowledge of the material's elastic properties and it is always
given by Eq 6, regardless of the test configuration used.
Split laminate beam specimens approximately 152.4 mm long and 25.4 mm wide for the
ENF test and 292.1 mm long and 25.4 mm wide for the ELS test were cut from 24-ply,
precracked, laminated panels of Hercules AS4/3502 (thickness of 3.05 mm) and Hexcel
T6C145/F185 (thickness of 4.83 mm) graphite/epoxy composites. The AS4 graphite fibers
have an average tensile strength of 3585 MPa, an average tensile modulus of 234 GPa, and
a density of 1.80 g/cm 3 [7]. The T6C145 graphite fibers have an average tensile strength of
3550 MPa, an average tensile modulus of 234 GPa, and a density of 1.77 g/cm 3 [8]. The
3502 epoxy resin is a brittle resin with a tensile elongation of only 1.5% and a fracture
toughness of approximately 70 J/m z [9]. The F185 resin, which is much more ductile, has
a strain to failure of 8% and a fracture toughness of approximately 6000 to 8000 J/m: [9].
The starting crack was introduced on both panels by means of a Teflon insert 0.127 mm
thick. Tabs made of 12 x 12 x 1.4 mm aluminum angle and 4 x 9 • 0.4 mm rectangular
tubing were bonded to the cracked end of the ELS test specimens with a mixture of EA901
structural adhesive and B1 curing agent. A UN-10-A gray brittle coating manufactured by
Measurements Group Inc., Raleigh, North Carolina, was also applied to the lateral sides
of the specimen to facilitate monitoring crack locations.
End-Notched Flexure ( ENF) Test

Five specimens for each graphite/epoxy system were tested using a three-point bend fixture
with a total length of 101.6 mm attached to a closed loop servohydraulic test machine. Prior
to testing, the specimens were precracked to provide a sharp crack tip from which to initiate
the mode II fracture. Load point displacements were measured by the ram displacement,
and loads were monitored by a 2225-N load cell. Real-time analog plots of the load-deflection
curve were made on an x-y recorder. The tests were conducted by positioning the specimens
in the three-point bend fixture such that the location of the crack tip was approximately
midway between the center and one of the outer loading pins. Under displacement control
conditions and a ram rate of 2.54 mm/min, the specimens were loaded until delamination
crack growth occurred. Then, the specimens were unloaded.
Because machine stroke displacement was used to measure specimen deflection, a machine
compliance experiment was performed to make the necessary deflection corrections to the
data obtained during the mode II delamination tests.
G~tc calculations were made using Eq 1 and measured compliances after appropriate
machine compliance corrections had been made. It should be noted that corrections asso-
ciated with shear compliance were not taken into consideration because no reliable values
of the shear modulus G~2 were available. However, these are usually small, and this omission
was not thought to have a significant effect on the measured G,c values. The measured
compliances were obtained by curve fitting the best straight line to the data during the
loading part of the test. Equations 2 and 3 were also used to calculate Gl~c. To use these
two equations, a compliance calibration was performed with one of the tested specimens.
Finally, Guc values calculated by the area method were determined by means of Eq 6. The
energy absorbed in the creation of a new cracked surface was approximated numerically by
integrating the area under the load-deflection curves obtained.
End-Loaded Split Laminate ( ELS) Test

Two specimens for each graphite/epoxy system chosen for this study were tested in a
mode I / m o d e II fixture [3] attached to a closed loop servohydraulic machine. A mode I
precrack was introduced to the specimens to sharpen the initial crack created by the Teflon
insert and to provide an initial crack-length-to-beam-length (a/L) ratio of approximately
0.55, at which stable crack growth is expected, z
Mode II delamination testing was done under displacement control conditions at a rate
of 10.2 mm/min and 6.4 mm/min. The slower displacement rate was used when testing the
AS4/3502 specimens to obtain some stability in crack growth. Loads were recorded with a
445-N load cell. Displacements were monitored using the ram displacement of the test
machine. The tests were conducted by loading the specimens until the crack was allowed
to grow approximately 10 to 40 mm. At that point the test was stopped, and the new crack
location was marked and recorded. This procedure was repeated several times with the same
2 Schapery, R. A., Mechanics and Materials Center, Texas A&M University, College Station, TX
77843, private communication.
specimen until the crack tip was approximately 15 to 25 mm from the uncracked end.
The critical energy release rate G,< was calculated by means of Eq 4 and Eq 6. The axial
modulus Eu was calculated from the load-deflection test data and Eq 5. The energy absorbed
in the creation of a new cracked surface was approximated numerically by integrating the
area under each loop in the load-deflection curves.
Finite-Element Analysis
The stress field in the vicinity of the crack tip of a split laminate beam specimen under
mode II and mode I was determined using a finite-element analysis. This was a linear analysis
made with a finite-element algorithm developed by Henriksen. 3 The algorithm is based on
a nonlinear code updated with Lagrangian formulation using six and/or eight-noded iso-
parametric elements with two degrees of freedom per node.
To overcome the difficulty imposed by the singularity at the crack tip, a substantial
refinement of the mesh in the vicinity of the crack tip and the use of triangular six-noded
elements around the crack tip with midside nodes displaced to the quarter point were
implemented. These crack tip elements contain a singularity of the form 1VFrr, providing a
stress field that agrees with the theoretical stress singularity of linear fracture mechanics.
These "special" nodes also contain rigid body motion and constant strain modes, thus
satisfying all the necessary conditons for monotonic convergence [10].
The generated mesh consists of 791 nodes and 294 elements (Fig. 3). Mode I loading was
simulated by applying a symmetric load at the cracked end of the mesh. Mode II loading
was introduced by asymmetrically loading the cracked end (ELS test).
The material properties used in this finite-element analysis were from typical properties
of unidirectional AS4/3502 graphite/epoxy composite. Linear beam theory as welt as linear
elastic material behavior were assumed in the analysis. Stress contour plots were obtained
from the output data, and the stress distribution was plotted as a function of distance ahead
of the crack tip.
665
532 ')
399
266
133 ~ , A54/,3502
0
.51 1.02 1.53 2.04 2.55
DEFLECTO
I N(mm)
FIG. 4--Typical load-deflection curve ofAS4/3502 tested using the ENF test configuration.
3 Henriksen, M., Engineering Department, Colorado School of Mines, Golden, CO 80301, private
communication.
Scanning Electron Microscopy (SEM)

Split laminate specimens of Hercules AS4/3502, approximately 30.5 mm long, 5.1 mm
wide, and 1.02 mm thick and of Hexcel T6C145/F185 graphite/epoxy composites, 30.5 mm
long, 5.1 mm wide, and 1.27 mm thick, were cut from eight-ply and six-ply laminated panels.
Specimen dimensions were determined by the space allowed by the tensile stage of the
scanning electron microscope and stable crack growth criteria [6]. All specimens were pol-
ished with 0.3 alumina powder and cleaned ultrasonically in a scanning electron microscope
Freon 113 solution which is harmless to the resin.
Mode II real-time observations of the delamination fracture process were determined
using the scanning electron microscope. Mode II delamination was achieved by means of a
specially designed three-point bend fixture installed to the tensile stage of the scanning
electron microscope. Postmortem studies were performed to confirm that the real-time
observations were representative of the bulk delamination behavior.
All surfaces were coated with a 150-A-thick gold/palladium film to minimize charging
effects associated with the nonconductive nature of the epoxy resins of the composites. The
experimental results were recorded on both video tape and on standard tri-x film.

End-Notched Flexure (ENF) Test Results
Table 1 shows the ENF test results for both graphite/epoxy systems investigated. Method
I corresponds to G,c values obtained using beam theory equations (Eq 1) and measured
compliances. Method II results are based on the compliance calibration (Eq 2 and Eq 3.)
Method III refers to data reduction based on the area method (Eq 6) and the energy
dissipated under the load-deflection curve. Delamination crack growth occurred at typical
load values of 556 to 613 N for AS4/3502 graphite/epoxy and was unstable for this material.
Figure 4 shows a typical load-deflection curve of AS4/3502 graphite/epoxy showing the
TABLE 1--End-notched flexure (ENF) test results.
Gu,, J / m-'
Material Specimen I II III
AS4/3502 ! 574 583 616

2 567 563 593
3 633 621 619
4 585 582 630
Average 590 587 615

T6C145/F185 a a
1 . . . . . . 2240
2 . . . . . . 2695
3 . . . . . . 2485
4 . . . . . . 1995
Average . . . . . . 2354
a No calculation of GHc is possible due to significant nonlinearity in load-deflection curves. If one

uses linear portion of curves to measure compliance and Pmaxfor crack extension (ignoring nonlinear
behavior near P~,,,), GHclower bound value of 1225 may be calculated.
1780
1424'I I
1068
712
.76 1.52 2.29 3.05 3.81
DEFLECTION (m)
FIG. 5--Typical ~ a ~ d e f l e ~ o n curve ~ ~ / H ~ tes~d z~mg ~ e E N F test configu-
ra~on.
sudden drop of load as unstable delamination occurred. This indicates that the G~rc values
measured using the area method will include not only the energy required to create a crack
surface but also the energy dissipated in dynamic crack growth processes. Because the work
required to create a cracked surface is rate dependent and usually smaller at higher rates,
the G., calculated for unstable crack growth using the area method can be either higher
(due to dynamic effects) or lower (due to rates effects on toughness) than the initiation
value calculated using beam theory. The small nonlinearity just prior to crack extension is
probably associated with the development of a crack tip damage zone.
The average critical energy release rate G.c for AS4/3502 using methods I, II, and III of
data reduction were 590, 587, and 615 J/m 2, respectively. These values are all within 5%
of each other, which can be considered within the experimental error associated with the
different methods of data reduction. Note that methods I and II, both based on linear beam
theory, give almost identical results.
Figure 5 shows a typical load-deflection curve obtained for mode II delamination testing
of T6C145/F185 graphite/epoxy. Crack growth was stable as seen in the upper part of the
curve. However, the point at which crack growth occurred is not well defined. Therefore,
the maximum load reached during each test was used to estimate the onset of crack growth.
These values were between 1390 and 1632 N. The curve also indicates an inelastic behavior
of the material (that is, the unload curve did not return to the origin). Therefore, all the
assumptions of the standard analysis for such tests have been violated. It should be noted
that the load-deflection curves at different crack lengths used for the compliance calibration
also showed nonlinear behavior. The Gitc values predicted by methods I. II, and I l i are all
ambiquous. However, the energy release rate obtained by the area method gives a better
approximation of the critical energy release rate Gin. because at least the geometrically
nonlinear behavior of the specimen is taken into consideration by this method. The calculated
average G,,c of 2400 J/m 2 should be considered an upper bound estimate of the mode II
critical energy release rate of T6C145/F185 because the inelastic behavior is not accounted
for in this analysis. A J-integral approach needs to be developed to obtain a more meaningful
measure of the mode II delamination fracture toughness of this system.
End-Loaded Split Laminate ( ELS) Test Results

The ELS test results are listed in Table 2. Method I corresponds to Gnc calculated using
Eq 4 and with E n calculated from Eq 5 and test data. Method II corresponds to the area
method (Eq 6). Average Eu values of 126 GPa for AS4/3502 and 110 GPa for T6C145/
F185 were calculated. Figure 6 shows a typical load-deflection curve obtained after testing
one of the AS4/3502 specimens. As seen in this figure, unstable crack growth is observed
for the first crack extension despite the a/L ratio of at least 0.55 introduced to all ELS
specimens during the mode I precrack done prior to mode II testing, which according to
the stability considerations of this loading and geometry should have been stable (see note
at bottom of page 206). This is probably the result of the additional energy storage in the
fixtures. Nonlinearity, especially at displacements greater than 51 mm, may also be associated
with the limitations of linear beam theory used in the stability analysis. An average G~c of
595 J/m 2 (method I) and 543 J/m 2 (method II) were found for AS4/3502. These results are
within 10% of each other. However, considerable scatter was observed in the Glz~data from
each specimen (Table 2). This scatter is believed to be the result of the high degree of error
involved in measuring crack locations which is done by monitoring the crack as it propagates.
In contrast, this scatter was very small in the E N F results. A typical load-deflection curve
for T6C145/F185 is seen in Fig. 7. Stable crack growth is observed for all load/unload loops.
This is evidenced by the smooth decrease in loading rate followed by a decrease in load
until the crack was arrested. It should be noted that it is not known when crack growth
TABLE 2--End-loaded split laminate ( ELS) test results.

G,c, J/m 2
Material Specimen I II
AS4/3502 1 630 595

613 560
613 543
648 543
2 595 525
595 525
508 508
Average 600 543

T6CI45/F185 a
1645 2153
2030 2030
2275 2415
2275 2363
2188 1873
2275 2118
1960 2643
2223 2240
2415 2083
2590 2170
2713 2765
2800 2328
Average 2282 2265
a Nonlinearity is too great to satisfy linear beam theory assumptions implicit in the approach.
66.5
_53.2
39.9
26.6
13.3 AS4/3502
0
12.7 25.4 38.1 50.8 63.5 76.2 88.9
DEFLECTION (m)
FIG. 6--Typical load-deflection curve ofAS4/3502 tested using the ELS test configuration.
started during this process, which was a common uncertainty for this material for both types
of mode II tests. A t least some of the observed nonlinear behavior is due to inelastic behavior
because the curves fail to return to the origin as the specimen is unloaded. Nonlinear behavior
associated with the limitations of linear beam theory is also observed for deflections greater
than 51 mm.
The average mode II energy release rate values for T6C145/F185 were found to be 2275
J/m 2 (method I) and 2260 J/m 2 (method II). However, a systematic or monotonic increase
in G,c for method I can be observed throughout the test. This is the result of the nonlinearity
in the load-deflection curve seen in Fig. 7. Although this nonlinearity is less severe than
the nonlinearity observed for this material tested using the ENF test, it is sufficient to
invalidate the data analysis done using linear beam theory (method I). The scatter of the
results for method II (area method) is more random and is associated with the experimental
difficulty in obtaining a precise measure of the change in crack length. However, these
222.5
%
178.0
~.~ 133.5
89.0
44.5
0
0 12.7 25.4 38.1 50.8 63.5 76.2 88.9
DEFLECTION (ram)
FIG. 7--Typical load-deflection curve of T6CI45/ F185 tested using the ELS test configu-
ration.
values are an overestimation of G.c because the curves did not return to the origin as the
specimens were unloaded, resulting in the incorporation of some far field energy dissipation
in the calculated work per unit area of crack extension.
A comparison of the toughness GII~calculated using both test configurations indicates that
for AS4/3502 (a composite utilizing a brittle resin), similar results were obtained for the
ENF and ELS tests. The T6C145/F185 delamination fracture toughness tests using the ENF
and ELS loading configuration do not give valid results because the assumptions associated
with each test method were violated. Therefore, in order to properly characterize the fracture
toughness of this material and tougher graphite/epoxy composites in general, an analysis
that allows for material and geometric nonlinearity such as a J-integral is needed. It is worth
noting that similar results were obtained for the ENF and ELS specimens when data were
analyzed using the area method on this tougher composite system. However, both results
should be considered upper bounds of the mode II critical energy release rate of the material,
as previously noted,
One difficulty in reducing the measured load-displacement curves to obtain G.c should
be mentioned. A small degree of nonlinearity was noted during loading for the composite
that utilized the brittle resin; a greater amount was noted in the composite made with a
ductile resin. If care is not taken to measure the slope consistently from specimen to spec-
imen, then apparent variations in Gilt will be, in fact, the result of careless data reduction.
Where our results showed nonlinearity (mainly at the very beginning of the test and near
maximum load), these nonlinearities were averaged with the linear portions of the curves
in the compliance determination, and measured compliances procedures. A small, systematic
decrease in compliance with increasing load has been noted for both ductile and brittle
resin/fiber systems. This is thought to be due to friction effects which develop by contact
between the two legs of the partially split laminate, but further work is necessary to confirm
this point. Both Eq 1 and Eq 4 indicate such an error will cause the calculated G,,. to be
artificially high if the net effect of friction is to raise the value of P at which crack extension
occurs.
Finite-ElementAnalysis
Figure 8 shows the finite-element results of a split laminate beam specimen tested under
pure mode I conditions (double cantilever beam test). Figure 8a is a stress contour of the
normal (Syy) stress. A perfectly symmetric stress condition is shown through the beam
thickness indicating the expected mode I loading. The stress is seen to rapidly increase as
the crack tip is approached. Figure 8b shows the Syy stress distribution ahead of the crack
tip at the midplane of the specimen. The Syy stress, a principal normal stress along this
plane, rapidly drops off with distance from the crack tip until it is compressive at a distance
of 0.76 mm ahead of the crack tip. Finally, it gradually approaches a zero stress level at
approximately 8 mm from the crack tip, which is maintained all the way to the end of the
beam. Similar behavior was observed when the stress field ahead of the crack tip was
calculated by means of a high order plate theory [11]. An identical loading case, but assuming
isotropic material behavior, also indicated that compressive stresses developed in the spec-
imen ahead of the crack tip. This suggests that these compressive stresses are due to geometry
and loading rather than material anisotropy. However, this compressive stress state is not
expected to significantly affect delamination because it develops far enough away from the
crack tip. The dashed line in Fig. 8b indicates the magnitude of the shear stress Sxy, which
is seen to be zero everywhere along the delamination plane for the mode I loading. Therefore,
this confirms that a pure mode I loading macroscopically does indeed give a pure mode I
delamination stress microscopically around the crack tip.
t "" ..... "" " "
/ 6 . 9 HPa a. \
-6.9} \
i/
)
0 Pg, .-"
9
,'"
s
. I I I
2.5 mm Ahead c r a c k - ~ Behind crack 2.5 mm
(a)
0.07
0.06 -
0 0.05 -
0. MODE I
(.9
v 0.04 -
>-
X
(/I 0.03-
o 0.02 -
>.
v) 0.01
Sxy Syy/
0.00 J.
-.0! ! ! , ,
-10 -8 ~7 -6 -5 -4 -3 -2 -1 0
Disfance Ahead of Crack Tip (ram)
(b)
FIG. 8--Finite-element mode I results: (a) Syy (normal) stress contour plot; (b) Mode 1
stress field ahead of crack tip in the plane of delamination.
-6.9 MPa .;
--.-.2D.2_] ...... " 0 Pa
................ 0 Pa
-6.9 MPa :-...
2 . 5 mm Ahead c r a c k
4- Behind c r a c k
2.5 nm~
(a)
0.07
0.06
o 0.05
rt MODE II
0
v 0.04-
>.
x
O3 0.03 -
L..
o 0.02 - Sxy /
V) 0.01 -
0.00 -
$yy
-.01
-10 . 'o . '0 --'7
. '6. "5 "4 "3 '2 "1 o
Distance Ahead of Crack Tip (ram)
(b)
FIG. 9--Finite-element mode 11 results." (a) Sxy (shear) stress contour plot; (b) Mode 11
stress field ahead of crack tip in the plane of delamination.
Figure 9 shows the results of a split laminate beam specimen subjected to an asymmetric
loading which is postulated to give a pure mode II (shear) state of stress. Similar to the
normal stress under mode I conditions, the stress field is perfectly symmetric across the
laminate, as seen in Fig. 9a. However, the shape of the stress field appears more narrow
and elongated than for mode I loading. Figure 9b shows the stress field in the plane of the
crack as a function of distance ahead of the crack tip. In this case, the shear stress can be
seen to decrease monotonically to a constant value which extends all the way to the end of
the beam. The dashed line indicates the normal stress distribution along the delamination
plane to be zero (that is, macroscopic mode II loading does indeed appear to give a micro-
scopic mode II stress along the delamination plane). As can be seen, the shear stress ahead
of the crack tip for mode II loading decays much more slowly than the normal stress for
mode I loading. Furthermore, the extent of the stress field above and below the delamination
plane is much smaller for mode II than for mode I, as seen in the stress contour plots. These
results are consistent with the SEM observations of damage zone size and shape to be
reported in the next section.
SEM Observations o f Mode H Delamination

Figure 10 presents real-time SEM observations of delamination for loading of AS4/3502.
Figure 10a shows a portion of the damage zone ahead of the crack tip which developed as
the specimen was being loaded (left side of micrograph). This damage zone, characterized
FIG. [O--Mode 1l in situ delumtnution of AS413502.
by extensive microcracking, was found to be long and narrow like a Dugdale zone, being
at least 180 Ixm long and extending about 5 Ixm above and 5 Ixm below the plane of
delamination. This elongated damage zone is at least twice as long as the damage zone re-
ported ahead of the crack tip for mode I delamination of the similar material AS4/3501-6
[12]. This is probably because for mode II, the resin-rich region between plies behaves
like a soft material between rigid platens with all the strain being localized in this region,
and also because of the slower rate of decay of the stress field under mode 1I compared to
mode I revealed by the finite-element analysis. The hackle formation process often associated
with mode II delamination of composites made with relatively brittle resins can also be seen
in Fig. 10. First, microcracks start to form at approximately 45 ~ to the plane of the plies
(Fig. 10a), perpendicular to the principal normal stress direction. Then, macroscopic crack
extension occurs as the sigmoidal microcracks begin to coalesce, forming the hackles (Figs.
10a and 10b). During this process, some resin deformation must have accompanied the
FIG. 11--Mode H in situ delarnination of AS4/3501-6 [1 ].
FIG. 12--Mode H in situ delamination of T6C145/F185.
coalescence of microcracks into hackles as the gap between adjacent hackles results from
rotation at the base as microcrack coalescence takes place. Note how the final hackle
orientation is steeper than the 45 ~ angle of orientation of the initial microcracks from which
the hackles form. No clear indication of crack growth direction is given by the hackle
orientation on the fractured surfaces. Arrows at the upper left corner of all micrographs
indicate crack growth direction. Figure 11 further illustrates the difficulty in trying to infer
crack growth direction from hackle direction alone [13].
In situ SEM observations of the T6C145/F185 composite are shown in Fig. 12. The damage
zone developed ahead of the crack tip as the specimen was being loaded. This damage zone
was approximately 1000 Ixm long and extended at least 25 gm above and 25 I~m below the
plane of delamination near the macroscopic crack tip, progressively decreasing to about 5
i~m at the end of the damage zone. The damage zone is characterized by extensive micro-
cracking. While the microcracking for mode I delamination is clearly in the bulk [14,15], it
appears that the microcracking seen for mode II loading of the very ductile T6C145/F185
is in the gold/palladium film applied to avoid charging. This microcracking in the coating
begins at approximately 3% strain [16]. Thus, the "damage zone" here is a measure of the
size of the nonlinear deformation zone ahead of the crack tip that has at least 3% strain.
This deformation/damage zone is approximately 15 times larger by area than the damage
zone for AS4/3502, indicating that much more energy is dissipated in the delamination
process for this tougher composite material (T6C145/F185). Note also that the microcracks
do not coalesce to form hackles as macroscopic crack extension occurs, because they are
primary coating cracks, not resin cracks. Instead, the resin extensively deforms by shear
deformation until a complete separation of the surfaces occurs due to local shear deformation
(Fig. 12b). Note the large strain undergone by the resin in areas well above and below the
resin-rich area where the crack formed. The extensive resin deformation and large damage
zone observed explains the nonlinear behavior observed in the load-deflection curves for
this material (Figs. 5a and 7). Furthermore, the much greater resistance to delamination of
T6C145/F185 compared to AS4/3502 measured for mode II loading (2260 J / m 2 versus 590
J/m 2) is consistent with these observations. The long, narrow shape of the damage zone for
both systems is consistent with what one would expect based on the finite-element analysis.
Postmortem fractography of the fractured surface under mode II delamination of AS4/
3502 can be seen in Fig. 13. The most significant artifact observed in these micrographs are
the hackles formed on the fractured surface. Note the featherlike appearance of the hackles
which have developed at a very steep angle (larger than the 45 ~ angle at which the microcracks
FIG. 13--Mode H post-mortern fractography of AS4/3502.

FIG. 14--Mode 11 post-mortem #actography of T6CI45/ FI85.
were observed to develop during the in situ observations) (see Fig. 13a). Therefore, the
hackles undergo some rotation before full separation of the fractured surfaces occurs. This
is in good agreement with the final hackle angle orientation from the in situ observations
(Fig. 10). Figures 13b and 13c show the detailed river pattern markings that develop on the
surface. No clear indication of crack growth direction is seen from the river pattern markings
on the hackles or the direction in which the hackles point (Fig. 10a). In fact, the direction
the hackles and the river pattern markings point will be determined by whether microcrack
coalescence occurs on the upper or lower boundary of the resin-rich region between plies,
and this is a random process.
Postmortem observations of the fractured surface under mode 1I delamination of T6C145/
F185 composite can be seen in Fig. 14. Note the extensive deformation undergone by the
resin, indicating that the resin failed due to yielding instead of the failure leading to hackle
formation of the more brittle graphite/epoxy systems such as AS4/3502. Comparing Fig.
14 with the in situ SEM observations for this material (Fig. 12) confirms our earlier hypothesis
that the failure in the system is ductile and the observed microcracking was only in the
coating.
As it has been shown already, the critical energy release rate for T6C145/F185 is ap-
proximately four times larger than for AS4/3502. Therefore, it can be inferred that although
the formation of hackles provides a more tortuous path for the crack leading to an increased
fractured toughness for mode II loading compared to mode I results for brittle composites,
extensive resin deformation can play an even more significant role in the fractured toughness
resistance of tougher graphite/epoxy systems. In fact, the reported fracture toughness of
T6C145/F185 under mode I conditions where little to no hackle formation is observed is
similar to the fracture toughness of the material under mode II conditions [91. Fracture was
by shear deformation in both cases.
Conclusions
1. Both the E N F and ELS tests give similar values for the critical mode II energy release
rate G~xc. However, because elastic material behavior is assumed in the analysis, the G~c
results for the ductile composite are somewhat uncertain because a small permanent de-
formation was observed.
2. Different sizes and shapes for the damage zone are observed for mode I to mode II,
and these differences are consistent with the difference in the size and shape of the stress
field predicted by finite-element analysis.
3. The deformation damage zone developed ahead of the crack tip at the onset of crack
growth under mode II delamination conditions is much larger by area for the composite
made using a more ductile epoxy (F185) than for the composite made with a brittle epoxy
(3502).
4. The formation of hackles provides a more tortuous path for the crack, leading to an
increased resistance to delamination under mode II conditions compared to mode I for
brittle composites. However, the extensive resin deformation and yielding play a more
significant role in the fracture toughness resistance of tougher composites giving similar
fracture toughness values under mode I and mode II conditions.
Acknowledgment
This work was made possible by the generous support of the Air Force Office of Scientific
Research under the direction of George Haritos.
References
[1] Tse, M. K., Hibbs, M. E, and Bradley, W. L., "'lnterlaminar Fracture Toughness and Real Time
Fracture Mechanisms of Some Toughened Graphite/Epoxy Composites," Toughened Composites,
ASTM/STP 937, N. J. Johnston, Ed., American Society for Testing and Materials, Philadelphia,
1987, pp. 115-130.
[2] Murri, G. B., "Gut, Measurement of Toughened Matrix Composites Using the End-Notched Flexure
(ENF) Test." ASTM Committee D30.02 Task Group on Interlaminar Fracture Toughness, Dallas,
TX, 1984.
[3] Vanderkley, P. S., "Mode I-Mode II Delamination Fracture Toughness of a Unidirectional Graph-
ite/Epoxy Composite." Master of Science thesis, Texas A&M University, College Station, TX,
Dec. 1981.
[4] Russell, A. J. and Street, K. N., "The Effect of Matrix Toughness on Delamination: Static and
Fatigue Fracture under Mode II Shear Loading of Graphite Fiber Composites," Toughened Com-
posites, ASTM/STP 937, N. J. Johnston, Ed., American Society for Testing and Materials~ Phil-
adelphia, 1987, pp. 275-293.
[5] Gillespie, J. W., Carlsson, L. A., and Pipes, R. B., "'Finite Element Analysis of the End-Notched
Flexure Specimen for Measuring Mode II Fracture Toughness," Composite Science and Technology,
Vol. 27, 1986, pp. 177-197.
[6] Carlsson, L. A., Gillespie, J. W., and Pipes, R. B., "'On the Analysis and Design of the End-
Notched Flexure (ENF) Test Specimen for Mode II Testing," Journal of Composite Materials,
Vol. 20, November 1986, pp. 594-604.
[7] Specification Sheet of AS4/3502 Graphite/Epoxy Composite, Hercules, Inc., P.O. Box 98, Magna,
UT 84044.
[8] Bulletin CFM4B, Celanese Corp., 26 Main Street, Chatham, NJ 07928.
[9] Jordan, W. M., "'The Effect of Resin Toughness on Fracture Behavior of Graphite/Epoxy Com-
posites," Ph.D. dissertation, Texas A&M University, College Station, TX, 1985.
[10] Barsoum, R. S., International Journal for Numerical Methods in Engineering, Vol. 10, 1976, pp.
25-37.
[11] Whitney, J. M., Composites Science and Technology, Vot. 23, 1985, pp. 201-219.
[12] Bradley, W. L. and Corleto, C. R., "The Significance of Hackles in the Failure Analysis of
Graphite/Epoxy Composite Materials," Mechanical Engineering Division of the Texas Engineering
Experiment Station, Texas A&M University, College Station, TX, 1986.
[13] Hibbs, M. E and Bradley, W. L., "Correlations Between Micromechanical Failure Processes and
the Delamination Fracture Toughness of Graphite/Epoxy Systems," in Fractography of Modern
Engineering Materials: Composites and Metals, ASTM STP 948, J. E. Masters and Joseph J. Au,
Eds., American Society for Testing and Materials, Philadelphia, 1987, pp. 68-97.
[14] Nam, S. W., Bradley, W. L., and Chakachery, E. "'A Study of Coating Cracking During In-Situ
Fracture Studies in SEM," unpublished research, Department of Mechanical Engineering, Texas
A&M University, College Station, TX, 1986.
[15] Bradley, W. L. and Cohen, R., "'Matrix Deformation and Fracture in Graphite Reinforced
Epoxies," in Delamination and Debonding of Materials, ASTM STP 876, W. S. Johnson, Ed.,
[16] Jordan, W. M. and Bradley, W. L., -Micromechanics of Fracture in Toughened Graphite-Epoxy
Laminates," in Toughened Composites, ASTM STP 937, N. J. Johnston, Ed., American Society
for Testing and Materials, Philadelpia, 1987, pp. 95-114.
T. K e v i n 0 ' Brten,
" 1 G r e t c h e n B. M u r r i , 1 a n d Satish A . S a l p e k a r 2
Interlaminar Shear Fracture Toughness and

Fatigue Thresholds for Composite Materials
REFERENCE: O'Brien, T. K., Murri, G. B., and Salpekar, S. A., "lnterlaminar Shear
Fracture Toughness and Fatigue Thresholds for Composite Materials," Composite Materials:
Fatigue and Fracture, Second Volume, ASTM STP 1012, Paul A. Lagaee, Ed., American
ABSTRACT: Static and cyclic end-notched flexure (ENF) tests were conducted on three
materials to determine their interlaminar shear fracture toughness and fatigue thresholds for
delamination in terms of limiting values of the mode II strain energy release rate, G,, for
delamination growth. Data were generated for three different materials: a T300/BP907 graph-
ite/epoxy, an $2/SP250 glass/epoxy, and an AS4/PEEK (polyetheretherketone) graphite/ther-
moplastic. The influence of precracking and data reduction schemes on the mode II toughness
and fatigue behavior is discussed. Finite-element analysis indicated that the beam theory
calculation for G,, with the transverse shear contribution included was reasonably accurate
over the entire range of crack lengths. However, compliance measurements for the three
materials tested and the variation in compliance with crack length differed from the beam
theory predictions. For materials that exhibited linear load-deflection behavior, GHc values
determined from compliance calibration measurements provided the most conservative and
accurate estimate of the interlaminar shear fracture toughness. Cyclic loading significantly
reduced the critical G, for delamination. A threshold value of the maximum cyclic G . below
which no delamination occurred after one million cycles was identified for each material to
quantify the degradation in interlaminar shear fracture toughness in fatigue. In addition,
residual static toughness tests were conducted on glass/epoxy specimens that had undergone
one million cycles without delamination. These residual static tests, and the initial static tests
on the tough AS4/PEEK graphite/thermoplastic, exhibited nonlinear load-deflection behavior.
For these cases, the load at deviation from nonlinearity was used to determine the interlaminar
shear fracture toughness. A linear mixed-mode delamination criterion was used to characterize
the static toughness of several composite materials; however, a total G threshold criterion
appears to be sufficient for characterizing the fatigue delamination durability of composite
materials with a wide range of static toughnesses.
KEY WORDS: delamination, fatigue, interlaminar fracture, mode II, strain energy release
rate, ENF test, PEEK, glass/epoxy, graphite/epoxy
Nomenclature
A, P a r a m e t e r s d e t e r m i n e d f r o m fit of c o m p l i a n c e calibration d a t a (i = 0,1,3)
a D e l a m i n a t i o n length
b B e a m width
C Flexural c o m p l i a n c e of E N F s p e c i m e n
Co Flexural compliance of u n c r a c k e d E N F s p e c i m e n (a = 0)
' Senior research scientist and research scientist, respectively, Aerostructures Directorate, U.S. Army
Aviation Research and Technology Activity (AVSCOM), NASA Langley Research Center, Hampton,
VA.
2 Research scientist, Analytical Services and Materials, Inc., Hampton, VA
222
Downloaded/printed
Copyright9 by
O'BRIEN ET AL. ON SHEAR FRACTURE TOUGHNESS AND FATIGUE THRESHOLDS 223
CSH Flexural compliance calculated from beam theory including the contribution of
transverse shear
C~ Flexural compliance of uncracked E N F specimen (a = 0) including the contri-
bution of transverse shear
Ell Axial modulus of lamina in fiber direction
E[~ Axial modulus of lamina calculated from compliance measurement in three-point
bend test
E:I~ Axial modulus of lamina measured from tension test of E N F specimen
Modulus of a unidirectional lamina transverse to the fiber direction
612 In-plane shear modulus of a unidirectional lamina
GI3 Transverse shear modulus of a unidirectional lamina
G Total strain energy release rate for delamination growth
Gc Critical value of strain energy release rate for delamination onset
G,h Threshold maximum cyclic G for delamination onset in fatigue
GI Strain energy release rate for delamination growth due to interlaminar tension,
mode I
Glc Interlaminar tension fracture toughness
G. Strain energy release rate for delamination growth due to interlaminar shear,
mode II
G~ r Mode II strain energy release rate as calculated by beam theory
GStn Mode II strain energy release rate calculated by beam theory with transverse
shear contribution
G~ Mode II strain energy release rate from finite-element analysis
Gllc Interlaminar shear fracture toughness
Critical mode II strain energy release rate at delamination onset calculated from
compliance calibration measurements
G,CtCS,, Critical mode II strain energy release rate for delamination onset calculated from
compliance calibration including transverse shear
GiSCc Critical mode II strain energy release rate for subcritical delamination growth
Gi~csn Critical mode II strain energy release (including transverse shear) for subcritical
delamination growth
Gllth Threshold maximum cyclic Gn for delamination in fatigue
h Beam half-thickness
L Beam half-span
P Out-of-plane load
t,. Critical load at delamination onset
P#L Load at onset of nonlinear behavior
R Ratio of minimum to maximum cyclic load
r2 Coefficient of determination (perfect data fit, r 2 = 1)
x,y,z Cartesian coordinates
8 Center point out-of-plane displacement
~c Critical center point displacement at delamination onset
Ix Coefficient of sliding friction for the delaminated E N F specimen
1)12 Poisson's ratio of a unidirectional lamina
Introduction
Delamination failures commonly occur in highly loaded composite structures. One of the
predominant loads experienced by many composite structures is interlaminar shear. Several
tests have been used to calculate interlaminar shear strength of composites, but all have
severe limitations [1]. Perhaps the most popular of these tests is the short beam shear (SBS)
test, which consists of a small unidirectional coupon loaded in three-point bending. Attempts
have been made to generate interlaminar shear S-N curves for composite laminates using
this SBS test [2]. However, short beam shear test specimens often fail in a mode different
than interlaminar shear [3]. Furthermore, the interlaminar shear S-N data generated by the
SBS test may not represent the generic material behavior and, hence, may not be applicable
to composite structures of differing layups and thicknesses [4].
To assess the delamination durability of composites under cyclic loads, tests for inter-
laminar fracture toughness have been conducted to determine fatigue thresholds for delam-
ination in terms of limiting values of the strain energy release rate, G, associated with
delamination growth [5-8]. Because these fatigue thresholds are calculated in terms of G,
they represent generic material behavior independent of the composite layup or geometry.
The end-notched fexure (ENF) test was recently developed and evaluated for measuring
the interlaminar shear fracture toughness, GHc, of composite materials [9-20]. This ENF test
consists of a 24-ply unidirectional beam loaded in three-point bending (Fig. 1). The specimen
contains an insert at the midplane of one end to simulate an initial delamination. The crack
tip is extended beyond the front of the insert before loading to obtain a sharp crack tip.
The load measured at the onset of delamination from the precrack is substituted into an
analysis for GH to calculate the interlaminar shear fracture toughness, Gilt. In this study,
several techniques previously proposed to calculate Git were evaluated and compared.
In a previous study [12], the static ENF test was shown to be useful as a means of screening
various materials for improved interlaminar shear fracture toughness. In this study, both
static and cyclic flexural loading was applied to the beam to generate interlaminar shear
fracture toughnesses, G,c, and fatigue thresholds, G~,h, for glass/epoxy, graphite/epoxy,
and graphite/thermoplastic materials. Cyclic loads were applied by means of a roller support
fixture (Fig. 2). This fixture allows the specimen to rest on two pins or rollers, which are
mounted on ball bearings, while the load is applied to the center of the specimen by another
roller (Fig. 3). lnterlaminar shear fatigue thresholds were compared to G~ thresholds for
mixed-mode delamination generated from cyclic edge delamination tests. A delamination
fatigue failure criterion is proposed based on the delamination fatigue thresholds measured
from a variety of interlaminar fracture toughness tests.
Materials
Unidirectional panels with midplane inserts were manufactured and cut into ENF test
specimens for three materials. Test specimens of $2/SP250 glass/epoxy were manufactured
-r
P P_P_
z
FIG. 1--Diagram of ENF specimen.
FIG. 2--Roller fixture for three-point bend test.
from prepreg supplied by the 3M Co. by Bell Helicopter Textron under NASA contract
NAS1-18199. Test specimens of T300/BP907 graphite/epoxy were manufactured at N A S A
Langley from prepreg supplied by American Cyanamid. Test specimens of AS4/PEEK were
cut from panels manufactured by Imperial Chemical Industries. All test coupons were
approximately 152.4 mm (6 in.) long by 25.4 mm (1 in.) wide and 24 plies thick. The average
ply thicknesses for the T300/BP907, AS4/PEEK, and $2/SP250 were 0.16, 0.132, and 0.2413
mm (0.0063, 0.0052, and 0.0095 in.), respectively. Table 1 lists material properties, average
ENF specimen thicknesses, 2h, and fiber volume fraction, Vj, measured for the three ma-
terials. The axial modulus in the fiber direction, EN, was measured from tension tests on
the 24-ply ENF specimens. The transverse modulus, En, shear modulus, G,2, and Poisson's
ratio, v~2, were measured from 90~ -+45~ and 0 ~ tension tests. The fiber volume fractions
of the materials were calculated from the fiber areal weight of the prepreg divided by the
product of the fiber density and the average ply thickness measured for the ENF test
specimens. The crystalline percentage of the AS4/PEEK composites was 24% as determined
by wide angle X-ray scattering [21].
FIG. 3--ENF specimens loaded in the test fixture.
Test Procedures
Precracking
Previous studies have indicated that G~tcvalues measured in the E N F test by propagating
a crack from the insert (that is, measured without a precrack) will decrease with insert
thickness [12]. Other studies have indicated that the lowest values of G,c were measured
when a sharp precrack was grown from the insert before the bending load was applied [10,
14-20]. In this study, three precracking techniques were used. The first technique was to
clamp the specimen across the width slightly ahead of the insert and then wedge the crack
surfaces open until a sharp crack formed and grew to the clamp. This created a tension
(mode I) precrack. The second technique was to move the specimen in the three-point bend
fixture so that the initial crack length was nearly equal to the half-span length (that is, the
end of the insert was close to the center roller). The specimen was then loaded in three-
point bending to propagate the crack from the end of the insert to a position under the
center roller. This produced a shear (mode II) precrack. Then the specimen was positioned
in the test fixture to the desired initial crack length and was tested. The third technique was
identical to the second except that the precrack was grown at a relatively high cyclic load,
requiring relatively few load cycles, and then was repositioned and tested under a static
load. The advantage of the last technique was that the straightness of the delamination front
could be confirmed after the test by examining the fracture surfaces.
TABLE 1--Material properties.
Material E . , Msi E22, Msi Gi2, Msi vx2 2h, in. h"E,/GI: Vr, %
$2/SP250 6.31 2.50 0.60 0.25 0.2270 0.1355 49.5
T300/BP907 17.11 t.20 0.83 0.29 0.1505 0.1167 56.3
AS4/PEEK 21.23 1.50 0.67 0.37 0.1243 0.1224 61.7
As will be shown later, the static shear precrack was used to generate the majority of the
test data. By using this approach, the Gnc and cyclic G,h values obtained were representative
of the interlaminar shear fracture toughness, and fatigue threshold for delamination growth,
due to interlaminar shear stresses at the tip of a delamination that was created by large
interlaminar shear stresses. These large shear stresses could have developed during a high
load that the structure experienced during its lifetime. Previous work oncomposite fatigue
suggests that it is the high loads in the spectrum that are the most damaging in terms of
creating delaminations and subsequent reductions in residual strength and life. Therefore,
the common practice used for metals of coaxing a precrack by applying blocks of low cyclic
loads over many cycles was not adopted. The tension precrack was not used, even though
it may yield slightly more conservative values of Gllc, because it was assumed that if a pure
shear stress state existed at the delamination front, then the stress state in the material that
created the delamination must have also been pure shear.
Crack Length Determination

To locate the crack, the sides of the graphite composites were painted white with water-
soluble typewriter correction fluid. The glass composites were translucent and, therefore,
did not require an enhancement of the edge to locate the crack tip. Figure 3 shows the
graphite and glass composites as they appeared when loaded in the three-point bend ap-
paratus. The initial crack length was measured from the centerline of the right-hand roller
to the end of the crack before testing. The average of the crack length measurements from
both specimen edges was used in the data reduction to minimize error associated with an
uneven precrack. This crack length measurement could sometimes be verified after the test
by splitting the laminate into two pieces and observing the difference between the precrack
fracture surface and the fracture surface caused by the three-point bending. For example,
Fig. 4 shows the change in fracture surface appearance for two AS4/PEEK composites that
either had a static mode II precrack and was then cycled in mode II, or had a cyclic mode
II precrack and was then loaded statically in mode II. The static and cyclic shear fracture
surfaces have markedly different appearances, which allows an accurate determination of
initial crack length from the end of the precrack to the imprint of the support pin on the
outer surface.
FIG. 4--Difference in static and fatigue fracture surfaces for AS4~ PEEK.
Static Tests
Precracked E N F specimens were positioned in the three-point bending fixture to the
desired initial crack length. The fixture had 6.35-mm (0.25-in.) diameter steel rods, supported
by annular ball bearings encased in aluminum channels at each end, that applied the loads
across the specimen width. In addition, the fixture had a degree of freedom out of the plane
of the test specimen due to additional pinned joints between the cross heads and the rollers.
This extra degree of freedom assured uniform loading across the width of the laminate for
all three rollers9 This combination of loading pins and annular ball bearings created a
frictionless roller system ensuring simple support conditions for both static and cyclic loading.
However, because the E N F specimen was precracked on one side only, the deformation
was asymmetric under the applied loading. Hence, small side forces could develop causing
the specimen to shift on the rollers during the test. To prevent this from occurring, a small
restraining bar was attached to the fixture at the uncracked end of the specimen (Figs. 2
and 3). This restraining bar was free to move with the specimen as it deflected 9
All tests were conducted with a span length of 769 mm (3 in.), that is, L = 389 mm
(1.5 in.) in Fig. 1. Loads were applied using a servohydraulic test stand in stroke control at
a rate of 2.54 mm (0.1 in.)/min, until the delamination grew. The delamination grew to a
point immediately under the center load point for most crack lengths tested. Some specimens
were repositioned to a new crack length to yield additional toughness values from the same
specimen 9 Center point displacements were measured with a direct current differential
transducer (DCDT) whose rod was supported by a spring as shown in Figs9 2 and 3. The
load-displacement behavior of the specimen and applied load versus machine stroke were
plotted on an x-y-y' recorder. Typical load-displacement results for the three materials
tested are shown in Fig. 5. The load--displacement plots were linear for the glass/epoxy
laminates, slightly nonlinear for the graphite/epoxy laminates, and significantly nonlinear
for the graphite/thermoplastic laminates 9
Fatigue Tests
Precracked ENF specimens were positioned in the three-point bend apparatus to the
desired initial crack length. Specimens were loaded statically in stroke control to the mean
load and then cycled sinusoidally at a frequency of 5 Hz at a maximum constant load
amplitude corresponding to an R ratio of 0.1. Specimens were cycled until the onset of
stable delamination growth was detected by a combination of visual observation and a drop
off in the cyclic load at a constant cyclic stroke2 The number of cycles to delamination onset
4OO
600F
P, Ib400F /
200V~//I I J
300
P, Ib 200
100
/ /
I I I 1
/
I
0 .03 .06 .09 0 .05 9 10 0 .03 .06 .0q
6, in. 6, in. 6. in.
FIG. 5--Typical ENF load-displacement plots.
3 The drop off in cyclic load typically occurred first, with visual observation confirming delamination
growth after no more than 5% reduction in the maximum cyclic load.
was recorded, and the specimen was reloaded to the mean load to record the compliance
at the new crack length.
Residual Static Tests

Several $2/SP250 glass/epoxy laminates were tested statically after undergoing 106 or
more cycles at low cyclic loads below the threshold for delamination growth. The procedure
used for these tests was identical to the procedure used for the initial static tests. The load-
displacement plots were nonlinear for these residual static tests, similar to the initial static
tests on AS4/PEEK.
C o m p l i a n c e Calibration
Two precracked ENF specimens of each material were placed in the three-point bend
apparatus repeatedly, to simulate crack lengths ranging from a = 0 to a = L, by shifting
the position of the specimen in the three-point bend apparatus and thereby changing the
distance between the right end roller and the delamination front. At each unique delami-
nation length position, the specimen was loaded high enough to obtain a load-deflection
plot but without extending the delamination. The slopes of the load-deflection plots were
measured to obtain a record of specimen compliance as a function of crack length.
Analysis
In this section, several techniques proposed for calculating GH in the ENF specimen will
be outlined and compared. In subsequent sections, the ENF test data will be reduced using
several of these methods and will be compared. Finally, based on the observations from
this study, particular data reduction techniques will be recommended.
B e a m Theory
Figure 1 shows the ENF specimen configuration where L is the half-span length, 2h is
the laminate thickness, b is the laminate width, P is the applied load, and a is the initial
delamination length as defined by the distance between the load support and the delami-
nation front. A closed-form equation for the strain energy release rate associated with
delamination growth due to interlaminar shear was derived in Ref 9 for this test using linear
beam theory. This analysis yielded
9 p2 a2C
G~r - (1)
2b(2L 3 + 3a 3)
where C is the flexural compliance defined as the ratio of the center point deflection, 8, to
the applied load, P, and derived from the linear beam theory as
2L 3 + 3a ~ 8
C = 8Ettbh 3 - p (2)
Substituting Eq 2 into Eq I yields the following equation for G11in terms of the axial modulus,
9P2a 2
G~ r - 16Ellb2h 3 (3)
Transverse Shear and Friction Contribution

The beam theory equations may be modified to include the influence of transverse shear
deformation [14] where
Csn 2L3 + 3a3 [ 2(1.2L + 0.9a)h2Elt]

- 8E.bh3 1 + (2L 3 + 3--~G-~,~ J (4)
9P2a2C [ ]
GS( - 2b(ZL3 + 3a3) 1 + O.2(E./G,3) (h/a) 2 (5)
For transversely isotropic materials, the transverse shear modulus, G13, is assumed to be
equal to the inplane shear modulus, G~2. For shear compliant materials where Etl/Gx3 is
high, or for thick beams, the contribution of the shear deformation terms may be significant.
Also, Eq 5 indicates that for any material and span length, the contribution of transverse
shear will be the greatest for the shorter crack lengths.
The contribution of friction to crack growth retardation has been estimated previously
[14]. Friction decreases the energy available for crack propagation such that
3p2Ixa
GH(Ix) = GS~ 4E.b2h 2 (6)
where IX is the coefficient of sliding friction for the fracture surface. For reasonable values
of Ix, the reduction in G,c after including the friction contribution was found to be only 2
to 5% for graphite composites with typical test coupon geometries [14]. No attempt was
made in this study to quantify Ix or its influence on Gn.
Compliance Calibration Method

An alternate method for determining Guc is the use of a compliance calibration curve.
An experimental curve of normalized compliance, C/Co (where C is the compliance as
measured by beam theory [Eq 2] and Co is the compliance for the beam with no crack)
versus normalized crack length cubed, (a/L) a, can be constructed using the technique de-
scribed earlier. A linear regression fit of the data yields
C/Co = Ao + A3(a/L) 3 (7)
where A3 is the slope of the line fit to the data and A0 is the y-intercept. The linear beam
theory would yield A0 = 1 and A3 = 1.5. G,c by this approach is obtained by differentiation
of C in Eq 7 with respect to a, and multiplication by Pfl/2b, yielding
Gf c -
3A3P~,a2Co (8)
2bL 3
A similar compliance calibration technique may be employed by rearranging Eq 4 for
beam compliance with transverse shear such that
Csn- 8Enbh-~
2La 3 [ 1 + 1.2"y + 0.9~,(a/L) + 1.5(a/L) 3] (9)
where 3' = (h/L)2(Ell/G13), and the compliance of an uncracked beam is
2L 3
CO, - 8Eubh_ ~ (1 + 1.23') (10)
Dividing Eq 9 by Eq 10 yields
Csn/C~ = Ao + A,(a/L) + A3(a/L) 3 (11)
The beam theory with transverse shear would yield A0 = 1, A~ = 0.93'/(1 + 1.23'), and A3
= 1.5/(1 + 1.23'). These coefficients may also be determined independently from a least
squares fit of the compliance versus crack length data. Differentiating Eq 11 with respect
to a, and multiplying by Pc2/2b yields
0
G,CCSn _ P;Cs. (A, + 3A~(a/L) 2) (12)
2bL
Influence of Nonlinear Behavior

In Refs 16, 18, 19, and 20, deviations from the linear load-displacement curve for tough-
ened matrix composites (such as the AS4/PEEK graphite/thermoplastic) in the ENF fracture
test were attributed to the onset of subcritical crack growth and inelastic shear response of
the material in the crack tip region. Scanning electron microscope photographs indicate that
interlaminar shear fracture consists of the formation of matrix cracks at 45 ~ to the original
crack followed by a coalescence of these matrix cracks for extension of the delamination.
In brittle matrix composites, such as the T300/5208 shown in Fig. 6, there is very little
matrix yielding, and the formation and coalescence of these cracks occur simultaneously.
Hence, the load-displacement record is linear. However, in tough matrix composites, such
as the AS4/PEEK shown in Fig. 7, significant matrix yielding occurs during the formation
and coalescence of these matrix cracks. Furthermore, these two events do not necessarily
occur simultaneously, and the load-displacement record is nonlinear.
One conservative approach is to estimate the mode II interlaminar fracture toughness
using the load, PNL, at which nonlinear response is first observed. In this way, Guc may be
thought of as a strain energy release rate parameter for subcritical crack growth. In Refs
16, 18, 19, and 20, G~SC~was evaluated by substituting the initial linear compliance, C, and
the load at onset of nonlinear behavior, P,~L, in the strain energy release rate calculation of
Eq 1 yielding
9p2La2C
G sc - (13)
2b(2L 3 + 3a 3)
Similarly, this subcritical fracture toughness may be estimated by using PNL in Eq 5 to include
the contribution of transverse shear, or in Eqs 8 and 12 to incorporate the compliance
calibration information.
Finite-Element Analysis
Finite-element analyses of the ENF specimen have been performed to evaluate the strain
energy release rate for delamination growth [13,15,18,20]. In Ref 13, a two-dimensional
FIG. 6--Fracture surfaces of brittle (5208) material.
plane strain model using 1000 four-noded isoparametric elements with 2400 degrees of
freedom was used to model the ENF specimen. The virtual crack closure technique was
used to calculate Gn [22]. The element size in the vicinity of the crack was 0.02 by 0.02 mm
(0.000 79 in. by 0.000 79 in.). A nodal coupling technique implementing multipoint con-
straints was used to prevent overlapping of crack surfaces. In addition to the frictionless
case, assumed coefficients of friction were used to estimate friction forces at nodes located
on the crack surfaces. The ratio of G~E/GIr was plotted as a function of normalized crack
length, a/L, for two graphite/epoxy ENF specimens with two different span lengths subjected
to the same applied load. The results indicated that the deviation in Gn between the finite-
element analysis and beam theory was significant at the shorter crack lengths (a/L <=0.3).
However, if G~E/GSFis plotted, the agreement is fairly good over the entire range of crack
lengths (Fig. 8).
In Refs 15, 18, and 20, a two-dimensional plane stress analysis using four-noded isopar-
ametric elements with Gauss numerical integration of order two was used with the virtual
crack extension technique to calculate G, for the ENF specimen. Elements at the crack tip
had dimensions of 0.0127 by 0.0127 mm (0.0005 by 0.0005 in.). The frictionless contact
problem was incorporated in the analysis by connecting duplicate nodes across the crack
interface with nonlinear truss elements with zero tensile stiffness and infinite compression
stiffness. The finite-element results indicated that although the beam theory expressions for
compliance were accurate, the beam theory expressions for GII in Eqs 1, 3, and 5 may be
FIG. 7--Fracture surfaces of tough (PEEK) material.
conservative by 10 to 48%, with the greatest deviation from beam theory occurring at
a/L >=0.4. Further finite-element results were generated for a wide range of crack lengths
in Ref 20. As indicated in Fig. 8, these results also showed that the finite-element values
of G , diverged from the beam theory results for delamination lengths greater than 0.4 L.
Recently, a two-dimensional plane strain finite-element analysis was performed to verify
if either one of the previous analyses, which yielded contradictory results, was valid [23].
The E N F specimen was modeled using eight-noded, isoparametric elements. The laminate
modeled consisted of 24 plies, each having a ply thickness of 0.142 mm (0.0056 in.). The
span length was 76.2 mm (3 in.), that is, L = 38.1 mm (1.5 in.). Material properties typical
1.6 'T ,1 2h
I.~-~ T J/Trethewey et af
/<,,,,,, (EII/GI3= 25.7.
FE 1.4
Gn Mall and Kochhar / ~ L = 2 in )
Eli/G13
_ = 20.5, / ./
GFI 1.2 t -- 2 0 i n , . t T / ) /r-Present~ analysis
.uu// \(Ell/G13= 25.7, L = 1.5 in.)
1
0.5 I
all
FIG. 8--Comparison of G. from various analyses.
of a graphite/thermoplastic composite were used:
E, = 16.7 Msi
E22 = 1.40 Msi
GI2 = ai3 = 0.65 Msi
vl2 = 0.3
v13 = 0.3
Figure 9 shows a plot of a typical mesh. Three mesh refinements were performed by sub-
dividing the square elements at the delamination front into four smaller square elements of
equal size. The element size at the delamination front was 0.1905, 0.0953, and 0.0476 mm
(0.0075, 0.003 75, and 0.001 875 in.) for the coarse, medium, and fine meshes, respectively.
The element size was gradually increased away from the delamination front in both direc-
tions. The coarse, medium, and fine meshes had 312, 410, and 516 elements, respectively.
Initially, the delaminated surfaces were allowed to deform freely. Results indicated that
the delaminated surfaces would cross into one another, which is physically impossible.
Therefore, the nodes along the delamination front were constrained to move the same
distance in the vertical direction using multipoint constraints. Various a / L ratios between
0.2 and 0.9 were modeled in the analysis. This was achieved by shifting the supports and
the central load point by the same amount along the mesh, which is analogous to shifting
the beam in the three-point test fixture to test different crack lengths, as was done in the
compliance calibration testing. The crack tip always remained between the central load and
the same end support. Compliance values corresponding to the measured center-point dis-
placement for a unit central load were calculated for the three meshes. Compliance values
converged as the element size decreased from the coarse to the fine mesh. The G , values
were calculated using the virtual crack extension method. These G , values also converged
as the element size decreased from the coarse to the fine mesh. Furthermore, Gn values
calculated using the global change in compliance agreed with the G , values calculated using
the local virtual crack closure technique.
Figure 10 shows GH/P 2 as a function of a / L for the ENF specimen. Good agreement was
observed between the G re values and the GSxn values calculated using Eq 5. A maximum
deviation of 6.7% was observed for a / L = 0.4 (Fig. 8). A plane stress analysis was also
performed at this crack length using the coarse mesh. The difference in the plane strain and
plane stress values for G , was only 0.79%. Hence, Eq 5 yields G s" values that are reasonably
IIIII lllllll
IIIII Imlllll
IIIII ~ 1 1 1
I I I I I ILII I I
II1~ IEIIII
I~1 Imlltl
- - a 9
FIG. 9--Nni~-e~ment m~h ~ r ~e ENF ~er
3 • I0 -4
"t ~P ,J-2h /
z I.-~L'I T
G__ng
p2' (in'-Ibs}-I i GSH /
I I I I
0 .2 .4 .6 ,8 1
all
FIG. lO--Normalized mode l l strain energy release rate as a [unction o f delamination length
in the ENF test.
accurate, although slightly conservative compared to the finite-element results. The small
reduction (between 2 and 5%) in G , attributed to friction on the delaminated surfaces [14]
may eliminate some of this difference. For example, by assuming that
3 p 21xa
GS[~ -~ G~ E (14)
4Ellb2h z
then GS~n calculated by Eq 5 may represent an accurate value of the mode II strain energy
release rate. Hence, GS~n values represent reasonably accurate calculations of the strain
energy release rate over a large range of a / L values in the E N F test.
Results
Compliance calibration curves were generated for all three materials tested. Specimen
compliance was defined as the ratio of the center point deflection to the out-of-plane load.
Center point deflection was measured by the D C D T mounted under the specimen (Figs. 2
and 3). Compliance was measured from the slope of the load-deflection plot for each crack
length tested. The compliance calculated using the DCDT-measured deflection was less than
the compliance calculated using the stroke of the hydraulic ram. Hence, because the com-
pliance used to calculate G . was always calculated using the center point deflection measured
directly from the DCDT, no correction for machine compliance was necessary. Compliance
was measured for delamination lengths from a = 0 to a = 38.1 mm (1.5 in.) in increments
of 2.54 mm (0.1 in.).
The average of the two crack lengths measured from both edges was used to reduce the
compliance calibration data. Furthermore, because calculation of the slope is very sensitive
to errors in crack length measurement, the specimens used for compliance calibration tests
were split apart, and the edge measurements were examined for crack front deviation through
the width. A distinct crack front was visible on the fracture surfaces at the transition from
the shear precrack to the tensile fracture surface created by splitting the tested beam into
two pieces. Both of the glass/epoxy laminates had relatively straight delamination fronts.
The T300/BP907 laminates had a deviation in delamination lengths between the two edges
of 5.08 mm (0.20 in.) and 4.064 mm (0.16 in.). One of the A S 4 / P E E K laminates had a
deviation in delamination lengths between the two edges of 3.81 mm (0.15 in.), and the
other laminate had a straight delamination front.
Two specimens of each material were used to generate compliance calibration data. Data
3.0
[13AS4/PEEK Beamtheory-./
T300/BP007 /
C__ Z.0 ~ '
CO a 3
C _
L0 T0 _A0 §
[ I I 1 I
.2 .4 .6 .8 1.0
FIG. 11--Linear regression analysis of compliance calibration data.
were fit to Eq 7 using a linear least squares regression analysis. Figure 11 shows the data
and the linear fit to Eq 7 for one specimen of each of the three materials. Table 2 lists the
values of the slope, A3, the y-intercept, Ao, and the goodness of fit, r 2, calculated for each
test of the three materials. If the beam theory is accurate, the slope should be 1.5, and the
y-intercept should be 1. Of the three materials tested, only the A S 4 / P E E K material had a
slope close to 1.5. The T300/BP907 and the $2/SP250 had slopes that were significantly
less than 1.5.
These same data were fit to Eq 11 to determine the coefficients A0, A~, and A3 in the
cubic polynomial representation for the compliance of the E N F specimen with transverse
shear incorporated. Table 3 lists the values of these parameters as determined by the least
squares regression analysis. Also listed in Table 3 are values for A0, A,, and m3, calculated
from beam theory with transverse shear using properties from Table 1. As noted earlier in
the linear regression analysis, the A3 coefficient of the (a/L) 3 term is less than anticipated,
based on beam theory with transverse shear for the two epoxy materials. The additional A~
coefficient of the linear (a/L) term, which was not present in the linear regression analysis,
also appears in Eq 12 for GII. Table 3 shows that these values were also different from those
calculated using beam theory.
TABLE 2--Compliance calibration data, C/C0 = A0 + A3(a/L)~.
Roller Fixture Knife-Edge Fixture Beam Theory

Material Spec. No. A3 Ao r2 A3 A0 r2 A3 A0
$2/SP250 255- 6 1.151 1.003 0.999 1.226 1.000 0.989 1.50 1.0
255-17 1.230 0.983 0.990 1.408 1.000 0.998
AVG. 1.191 1.317
T300/BP907 0-2-11 1.313 1.012 0.999 1.316 0.991 0.996 1.50 1.0
0-2-9 1.296 1.036 0.992 1.318 0.994 0.997
AVG. 1.305 1.317
AS4/PEEK 3-20 1.446 0.991 0.991 1.801 1.007 0.997 1.50 1.0
3-19 1.482 1.002 0.995 1.540 1.004 0.997
AVG. 1.464 1.671
TABLE 3--Compliance calibration data, CsH/C~ = Ao + At(a/L) + A3(a/L)3.
Roller Fixture Beam Theory and Shear
Material Spec. No. A3 A~ A0 A3 A~ Ao
$2/SP250 255- 6 1.140 0.0113 1.000 1.399 0.839 1.000

255-17 1.338 -0.1199 1.014
AVG. 1.239 - 0.0543 1.007
T300/BP907 0-2-11 1.259 0.0606 0.996 1.412 0.847 1.000
0-2-9 1.101 0.2160 0.980
AVG. 1.180 0.1383 0.988
AS4/PEEK 3-20 1.527 -0.0905 1.014 1,408 0,845 1.000
3-19 1.511 -0.0324 1.011
AVG. 1.519 -0.0615 1.013
To determine the contribution of the roller test fixture compliance to C, calibration tests
were performed on the same specimens using a rigid knife-edge support fixture described
in Ref 12. Table 2 lists the values of A0, A3, and r 2 determined from these tests. For the
epoxy matrix materials, the slope, A3, was closer to the beam theory value of 1.5 for tests
conducted on the more rigid knife-edge fixtures. The difference in slope calculated for the
two fixtures was small for the graphite/epoxy (T300/BP907), however, the difference was
significant for the glass/epoxy ($2/SP250). For the graphite/thermoplastic (AS4/PEEK),
which showed the best agreement with the beam theory using the roller fixture, the slope
calculated using the knife-edge supports exceeded the beam theory value.
As a further check of the beam theory, the axial moduli of the compliance calibration
specimens were measured in a uniaxial tension test using a 25.4-mm (1-in.) long extensometer
as described in Ref 12. These Et~x measurements were compared to E( x values calculated
from compliance measurements using Eq 2 where
1 + 1.5(a/L) 3
Efx= 4b(h/L)3 C (15)
The E( x values were calculated from compliance measurements for each crack length,
measured in 2.54-mm (0.1-in.) increments, from 0 to 38.1 mm (1.5 in.). As shown in Fig.
12, Etex values increased slightly at the longer crack lengths. An average value of E~x over
20
AX F Knife edgefixture
15 [- Ell
~gg
Ell, Msi 10 "-- Roller fixture
0 L_t i t I
0 .5 1.0 l..~
a. in.
FIG. 12--Axial stiffness comparison for T300/BP907.

TABLE 4--Comparison of measured axial moduli with values from flexural compliance.
E ((, Msi
E ax, Msi
Knife-Edge
Material Spec. No. Roller Fixture Fixture Tension Test
$2/SP250 255- 6 5.42 5.95 5.83

255-17 5.56 5.92 6.79
AVG. 5.49 5.94 6.31
T300/BP907 0-2-11 12.44 13.38 17.25
0-2-9 12.20 13.59 16.97
AVG. 12.32 13.48 17.11
AS4/PEEK 3-20 15.76 15.64 22.76
3-19 15.45 16.06 19.70
AVG. 15.61 15.85 21.23
the range of crack lengths was calculated for each specimen. Table 4 compares the various
measurements of Elt. The E~( values were considerably less than the E ax values, with the
more compliant roller fixture yielding the lowest values. Table 5 lists the difference in the
E~(and E~(values for the three materials. The graphite composites had over 20% difference
in measured and calculated axial stiffness. If the transverse shear correction is included in
the flexural compliance (Eq 4), then the estimate of E~x becomes
1 + 1.5(a/L) 3
(16)
EfX = 4b(h/L)3C - (h/L)2[1.2 + 0.9(a/L)]/GI3
The second term in the denominator reflects the contribution of transverse shear. Hence,
including transverse shear increases E~x, but only by 4% to 5%. Therefore, the E N F spec-
imens are not behaving as the beam theory would predict, even after the influences of
transverse shear and test fixture compliance are included.
Examination of the specimen edges in an optical microscope indicated that the outermost
three layers were thicker than the interior plies. This may result from resin bleeding from
the interior to the top and bottom surfaces during cure. This resulting inhomogeneity yields
surface plies that are thicker, more resin rich, and hence, less stiff than the interior plies.
When the laminate is loaded in bending, the specimen appears to be much less stiff than
when it is loaded in tension because the outermost plies carry the majority of the load in
bending. Furthermore, this effect will be slightly different in the cracked and uncracked
portion of the E N F specimens, resulting in a change in compliance with crack length that
differs from the beam theory predictions. When this occurs, as evidenced by A3 coefficients
TABLE 5--Difference in measured and calculated axial modulus, ( E ~ - E~X)/Efix.
Material Roller Fixture Knife-Edge Fixture
$2/SP250 0.130 0.059

T300/BP907 0.280 0.212
AS4/PEEK 0.265 0.253
T A B L E 6--Static E N F data f o r T300/ BP907.
Fracture Toughness, in.-lb/in. "~
Spec. No. a, in. C, in./lb Pc, lb Co, in./lb G,~r G,Sn~ Gtcc Gtccsn Precrack
1-2-3 0.80 1.67 x 10 -4 430 1.39 x 10 .4 10.73 11.13 9.56 9.50 shear
1-2-8 0.70 1.70 • 10 -4 420 1.50 x 10 -4 8.50 8.91 7.52 7.75 shear
1-2-9 0.80 1.88 x 10 4 355 1.57 • 10 .4 8.23 8.54 7.33 7.29 tension
0-2-8 1.00 2.04 x 10 .4 375 1.47 x 10 .4 13.24 13.55 12.00 11.45 none
0-2-6 1.10 1.71 • 10 .4 390 1.13 x 10 -4 13.18 13.44 12.04 11.39 none
I-2-5 0.80 1.70 x 10 .4 450 1.42 x 10 -4 11.97 12.40 10.66 10.59 none
that are less than predicted by the beam theory (Tables 2 and 3), the compliance calibration
method should be used to reduce the data.
Static Tests
Tables 6 through 8 list the load at delamination onset, Pc, delamination length, a, com-
pliance, C, and critical values of G. using the various data reduction schemes for the T300/
BP907, AS4/PEEK, and $2/SP250 laminates, respectively. These data are summarized in
Fig. 13. Critical values of G u w e r e calculated using the measured compliance, crack length,
and load at onset of delamination, Pc, in the beam theory Eq 1. The contribution of transverse
shear was calculated using E . , G~2, and h from Table 1 in Eq 5. For all three materials,
GiS~ values were slightly greater than G~cr values, reflecting the small contribution of trans-
verse shear. Initial compliance values for the uncracked beam, C,~, were calculated from Eq
7 for each specimen tested using the measured compliance and crack length along with the
coefficients A3 and An, calculated using the roller fixtures in Table 2. Then, G[( and
G~Ci~TM values were calculated from Eqs 8 and 12 using appropriate coefficients from Tables
2 and 3. For all three materials, G cc values were lower and, hence, more conservative than
the beam theory values calculated with or without transverse shear. The largest differences
occurred for the materials whose slope, A3, deviated the greatest from the beam theory
slope of 1.5 in Eq 7 (Table 2).
Table 6 lists results for T300/BP907 graphite/epoxy. As noted earlier, the G~cc values
yielded the lowest (that is, the most conservative) estimate of toughness. The G~Clcsn values
TABLE 7--Static E N F data for A S 4 / P E E K .
Load, lb Fracture Toughness, in.-lb/in, z
Spec. No. ,. in. C. iu./Ib p, p,,~ G,'~f G,~,"~ O,If C,I~s" O,;~ C,;f"
3-6 0.65 2.94 • 10-4 355 330 9.30 9.84 9.09 8.85 7.86 7.65
3-10 0.60 2.25 • 10 -4 383 348 7.23 7.72 7.06 6.78 5.83 5.59
3-4 1.00" 3.32 x 10 -4 335 295 17.20 17.62 16.89 17.16 13.10 13.31
3-4 0.85 3.13 x 10 -4 356 310 15.01 15.52 14.71 14.79 11.15 11.21
3-12 1.100 2.95 x 10 -4 319 260 15.22 15.52 14.97 15.27 9.94 10.14
3-12 1.05 2.63 • 10 -4 355 315 16.09 16.44 15.81 16.10 12.45 12.68
3-12 1.00 2.54 • 10 -4 340 305 13.55 13.88 13.31 13.53 10.71 10.88
3-9 0.80 2.33 x 10 `4 615 255 30.63 31.80 30.00 29.99 5.16 5.16
3-7 0.60 2.42 • 10 -4 503 300 13.41 14.32 13.10 12.57 4.16 4.47
AVG. 15.29 15.85 14.99 15.00 8.98 9.01
~Cyelic shear precrack.
240 SECOND C O M P O S I T E MATERIALS
TABLE 8---Static E N F data for $21SP250.
Fracture Toughness, in.-lb/in, z
Spec. No. a, in. C, in./lb Pc, lb Co, in./lb G~ r GS~C GFc G,Ctcsn
255-20 0.85 1.60 x 10 -4 306 1.31 • 10 -4 5.66 5.87 4.76 4.95

255-20 0.90 1.42 • 10 4 363 1.13 • 10 -4 7.64 7.89 6.38 6.66
255-20 0.95 1.46 • 10 -4 357 1.12 • 10 -4 8.09 8.33 6.81 7.08
255-18 0.60 1.18 X 10 4 680 1.10 • -4 11.99 12.89 9.69 10.80
255-17 0.60 1.25 • 10 -4 505 1.16 X 10 -4 6.98 7.51 5.64 6.31
255-6 0.60 1.19 X 10 -4 530 1.11 X 10 -4 7.31 7.86 5.91 6.62
255-19 0.60 1.21 • 10 -4 510 1.12 • 10 -4 6.89 7.40 5.56 6.23
255-10 0.60 1.20 x 10 -4 522 1.12 x 10 -4 7.16 7.70 5.78 6.47
255-3 0.60 1.20 • 10 4 496 1.12 • 10 -4 6.47 6.95 5.22 5.84
AVG. 7.58 8.05 6.19 6.77
were slightly different, indicating the small contribution due to transverse shear. The three
tests run without precraeks, where the delamination was initiated at the insert, yielded the
highest apparent toughness. Because these delaminations started in the resin-rich pocket at
the tip of the insert, they were not considered to be representative of naturally occurring
delaminations. The tension precracked specimen yielded slightly lower toughness than the
two shear precracked specimens. H o w e v e r , because these specimens had experienced large
tensile deformations at the crack tip, they were also not considered to be representative of
a naturally occurring delamination in a region of high shear stresses. T h e r e f o r e , all remaining
tests for the three materials were run on specimens that were precracked in shear, either
statically or cyclically, to obtain a realistic interlaminar shear fracture toughness of the
delaminated composite with shear deformation at the crack tip.
Table 7 lists results for A S 4 / P E E K graphite/thermoplastic. As shown in Fig. 13, there
was a significant degree of scatter in the data from these nine tests. This scatter could be
reduced if the high and low data points were discarded leaving only seven data points. As
noted earlier, the GIcc values yielded lower (that is, more conservative) estimates of tough-
ness than the b e a m theory values, with or without transverse shear (Table 7). H o w e v e r , the
A S 4 / P E E K specimens exhibited significant nonlinear load-displacement behavior before
unstable delamination growth. Therefore, both the beam theory and the compliance cali-
bration methods do not reflect the change in energy with crack growth that occurs after the
Open bars - calculated

o Beam theory
g- r~ Beamtheory with
3O transverse shear
P ,,
Solid bars - compliance
?5
calibration
P~NL
GI-[c' 20 <> Beam theory
6 7- 9'~ Beam theory with
in. -Ibs 15 transverse shear
in.2 lO Open symbols - Pc
5 Solid symbols - PNL
o
1300/BP907 $2/5P250 AS4/PEEK
FIG. 13--1nterlaminar fracture toughness from the E N F test.
TABLE 9 Residual static E N F data for $2/SP250.
Load, lb Fracture Toughness, in.-lb/in. 2
Spec. No. a, in. C, in./lb Pc Put GIsir GIS~ GIcc GIccsn GlSCc Gxscsn
255-8 0.90 1.44 • 10 4 340 270 6.79 7.02 5.67 5.93 3.58 3.74
255-16 0.90 1.28 x 10 -4 372 255 7.22 7.42 6.04 6.31 2.84 2.96
255-4 0.65 1.37 • 10 -4 366 300 4.61 4.90 3.74 4.11 2.51 2.76
255-9 0.70 1.33 • 10 -4 365 300 5.02 5.30 4.10 4.44 2.77 3.00
AVG. 5.91 6.16 4.89 5.20 2.92 3.12
compliance has increased at loads above PNL. H o w e v e r , if subcritical crack growth is assumed
to occur at the onset of nonlinearity, then E q 14 may yield a m o r e accurate representation
of the interlaminar shear fracture toughness than E q 1. T h e r e f o r e , PNL was used in the
compliance calibration equations (8 and 12) to calculate the critical value of G . for onset
of subcritical delamination growth. Calculated values of G~c and GIscs" for the A S 4 / P E E K
specimens tested are listed in Table 7. These values, calculated using the load at onset of
nonlinear behavior, PNL, are significantly lower than toughness values calculated using the
critical load at unstable delamination growth, Pc. Interestingly, specimen 9 had the highest
CC
Pc and, hence, the highest value of G~ic, but had the lowest PNL, thus yielding a relatively
low value of GS~c. The two specimens that were precraeked under cyclic shear loading, using
a relatively high maximum load and a low n u m b e r of cycles, had toughness values similar
to the rest of the specimens that were precracked under static shear loads.
Table 8 lists results for the $2/SP250 glass/epoxy. As noted earlier, the G~Ic values yielded
the lowest (that is, the most conservative) estimate of toughness. In addition to the static
tests, residual toughness tests were conducted on four specimens that had been subjected
to relatively low m a x i m u m cyclic loads for at least one million fatigue cycles without de-
lamination growth. These residual toughness values (Table 9) were consistently lower than
the static toughness values measured (Table 8). F u r t h e r m o r e , these tests exhibited significant
nonlinearity, similar to the behavior of the A S 4 / P E E K static tests. T h e r e f o r e , Eqs 8 and
TABLE 10--Fatigue E N F data f o r T3001BP907.
GH.... in.-Ib/in. 2
Spec. No. a, in. C, in./lb P~,, lb G~ r G sn G cc G ccsn Cycles, N
0-4-9 1.08 2.00 • 10 -4 250 6.23 6.36 5.68 5.38 280

0-4-8 0.75 1.71 • 10 -4 260 3.65 3.80 3.24 3.27 490
0-4-4 0.95 1.71 • 10 -4 200 2.98 3.06 2.69 2.58 10 140
0-4-2 1.18 2.20 • 10 -4 150 2.66 2.70 2.44 2.30 1 745
0-4-3 0.90 2.00 • 10 -4 170 2.36 2.43 2.12 2.05 7 900
0-4-7 0.85 1.70 x 10 -4 170 1.86 1.92 1.67 1.63 18 830
0-2-5 0.97 2.10 • 10 -4 130 1.58 1.62 1.43 1.37 23 400
0-4-11 0.65 2.00 • 10 -4 160 1.29 1.36 1.13 1.20 28 100
0-2-6 0.80 1.80 x 10 -4 140 1.23 1.27 1.09 1.09 37 240
0-2-11 1.00 2.00 z 10 -4 110 1.12 1.14 1.01 0.97 331 850
0-2-8 1.00 2.12 • 10 -4 100 0.98 1.00 0.89 0.85 562 200
0-2-9 0.90 2.25 • 10 -4 90 0.74 0.77 0.67 0.65 >1 000 000
TABLE ll--Fatigue E N F data for $2/SP250.
Gn .... in.-lb/in, z
Spec. No. a, in. C, in./Ib P .... lb G~ r GS~ G cc G ccsn Cycles, N
255-15 1.00 1.50 x 10 -4 150 1.56 1.60 1.32 1.36 13 500

255-15 0.95 1.52 x 10 -4 150 1.49 1.54 1.25 1.30 16 800
255-14 0.90 1.42 x 10 -4 170 1.67 1.72 1.40 1.46 19 000
255-11 0.60 1.18 • 10 -4 244 1.53 1.65 1.50 1.39 300 000
255-8 0.90 1.41 x 10 -4 160 1.47 1.52 1.23 1.29 >1 000 000
255-2 0.90 1.50 • 10 4 160 1.57 1.61 1.31 1.37 >1 000 000
255-16 0.90 1.43 x 10 -4 150 1.31 1.36 1.10 1.15 >1 000 000
255-4 0.65 1.29 x 10 4 150 0.73 0.78 0.59 0.65 >1 088 000
255-7 0.80 1.33 • 10 -~ 140 0.91 0.94 0.75 0.79 >1 000000
255-9 0.70 1.48 • 10 4 145 0.88 0.93 0.72 0.78 >1 550 000
12 using PNL were used to calculate GlSlc a n d GtStCcsH for these s p e c i m e n s ( T a b l e 9). A s n o t e d
earlier for t h e A S 4 / P E E K material, the G sc were lower t h a n the G zcc values.
Fatigue Tests
Tables 10 t h r o u g h 12 list the cycles to d e l a m i n a t i o n onset, N, m a x i m u m cyclic load, Pm~x,

d e l a m i n a t i o n length, a, c o m p l i a n c e , C, a n d m a x i m u m cyclic G~ values using the v a r i o u s
data r e d u c t i o n s c h e m e s for the T 3 0 0 / B P 9 0 7 , $2/SP250, a n d A S 4 / P E E K l a m i n a t e s , respec-
tively. Figures 14 t h r o u g h 16 show the n u m b e r s of cycles to d e l a m i n a t i o n o n s e t as a f u n c t i o n
of m a x i m u m cyclic GfI c level for the t h r e e m a t e r i a l s tested. All t h r e e m a t e r i a l s exhibit
significant r e d u c t i o n s in critical G~ values for d e l a m i n a t i o n o n s e t with fatigue cycles, with
an a p p a r e n t t h r e s h o l d value for d e l a m i n a t i o n o n s e t in fatigue as i n d i c a t e d by the p l a t e a u s
in Figs. 14 t h r o u g h 16. H e n c e , cyclic loading significantly reduces t h e critical GH for d e l a m -
ination onset. T h e s e t h r e s h o l d values of GII may be c o m p a r e d to GHc using the E N F test to
quantify t h e d e g r a d a t i o n in i n t e r l a m i n a r s h e a r fracture t o u g h n e s s due to fatigue.
Discussion
Ideally, G~cc a n d GxccsH values for each s p e c i m e n s h o u l d be d e t e r m i n e d using values of
the A3 a n d AI coefficients m e a s u r e d for each s p e c i m e n . H o w e v e r , this w o u l d b e very t i m e
TABLE 12--Fatigue E N F data for A S 4 / P E E K .
Gtt.... in.-lb/in, z
Spec. No. a, in. C, in.lib P .... lb G~ r G sn G~Ic G~Icxn Cycles, N
3-22 0.60 2.35 • 10 -4 332 5.68 6.07 5.55 5.32 580

3-15 0.60 2.34 • 10 -4 257 3.39 3.62 3.31 3.17 2 250
3-17 0.60 2.29 • 10 4 182 1.66 1.77 1.62 1.56 30 000
3-11 0.60 2.33 x 10 -4 133 0.90 0.97 0.88 0.85 144 000
3-13 0.60 2.25 • 10 -4 125 0.77 0.82 0.75 0.72 >1 000 000
O'BRIEN ET AL. ON S H E A R F R A C T U R E T O U G H N E S S A N D FATIGUE T H R E S H O L D S 243
10F N 1 f=SHz
6
cc in - Ib
GI]max' in2 4
o I I I
o 2 4 6 8 lOx 105
Cycles, N
FIG. 14--GCCm~as a function o f cycles to delamination onset for T3001BPg07.
consuming. Therefore, all the data were reduced using the average value of these coefficients
calculated from the two compliance calibration tests for each material listed in Tables 2 and
3, and Co was determined from Eq 7 using the measured values of C and a for the particular
test. An alternative would be to measure Co directly for each specimen, while still using the
average coefficients from just a few compliance calibration tests.
Static Tests
Because the compliance calibration values are the most conservative for materials that
exhibit linear load--displacement behavior, and because they represent the actual change in
compliance with delamination growth for the specimens tested, the interlaminar shear frac-
CC
ture toughness is best represented by G.c. However, for materials that exhibit nonlinear
load-displacement behavior, as was observed in the AS4/PEEK static tests or in the residual
static tests on the $2/SP250 following high cycle fatigue, the interlaminar shear fracture
toughness is characterized most conservatively by G sc values. Because the growth of a mode
II delamination actually corresponds to the coalescence of small tensile cracks oriented at
45 ~ to the delamination plane in the resin layer between the plies, then mode II crack growth
may be more stable if the resin toughness is high, or if the material at the delamination
front has been cyclically deformed. Hence, the critical G,, for onset of subcritical (that is,
10
f=SHz
(N= 1) R=0.1
Static
6
GCC in - Ib
I1 max' in 2 4
2
o o
I i I I I
0
0 2 4 6 8 lO • 105
Cycles, N
FIG. 15--G~,,~ as a function of cycles to delamination onset for $2/SP250.
301
r f = 5 Hz
R=O.I
25
20
cc in - Ib
GII max' . 2 15
lO
oL- 0~ 21 4 o 6 5 8 I0 x
Cycles, N
FIG. 16--GCC,~as a function of cycles to delamination onset for AS41PEEK.
stable) delamination growth may provide a better, and more conservative, measure of the
interlaminar shear fracture toughness in these cases.
Fatigue Tests
The ENF tests conducted in this study showed a significant, in some cases an order of
magnitude, reduction in the mode II delamination durability of the three materials studied
compared to their static interlaminar shear fracture toughnesses (Figs. 14-16). This reduction
in delamination resistance during cyclic loading was also observed in studies conducted using
mixed-mode edge delamination tension tests [5-8]. Figure 17 shows the reduction in critical
total G for delamination onset for two edge delamination tension (EDT) layups of T300/
BP907 [7]. The two layups have intermediate and low percentages of mode II, with the
remainder mode I. Also shown in Fig. 17 are the critical G cc values for delamination onset
(Fig. 14) for static loading and for fatigue (with the same frequency and R ratio as the E D T
tests). The static Gc values are different for the three tests, with the layups having the highest
percentage of G. showing the largest apparent toughness. However, the G,, values are nearly
identical for the two edge delamination layups and the ENF specimens. Hence, the static
10 F,~ Static Fatigue

8 kT IN=l) R = 0.1 f = 5HZ tayup *'~II Test
9 ~ 024 100% ENF
6~-~ ~' v +30/• 30/gol~)s 43~ EDI
in - Ib
Gmax'
in
4
F Lo..v.n_~__ {•
x 9 A
10, EDT
, 5>"
Cycles, N
FIG. 17--Comparison of static Gc and fatigue G,, values for delarnination onset for T300/
BP907 E D T and ENF tests.
f=SHz
Gmax,
io2 ~ (3~2/-352/oz/902)s
i0~
l l l
0 2 4 6 8 10x 105
Cycles. N
FIG. IS--Maximum cyclic O as a function o f cycles to delamination onset for A S 4 / P E E K
E D T and ENF tests.
toughness of this material will vary with the ratio of mode I and II at the delamination front,
but the fatigue threshold depends only on the total G, independent of the mode ratio.
Figure 18 shows the reduction in critical total G for delamination onset for a (352/-352/
02/902), edge delamination tension (EDT) layup of AS4/PEEK [8]. The total G for this
layup consists predominantly of tension, G~, with only a small shear component, G,. Also
shown in Fig. 18 are the critical G. values from the ENF test for delamination onset in
fatigue (Fig. 16) under the same frequency and R ratio. Although these two tests are different
in that one consists of pure interlaminar shear and the other is predominantly interlaminar
tension, the G,h values are nearly identical. Hence, as was noted earlier for the T300/BP907
material, the fatigue threshold for the AS4/PEEK appears to depend only on the total G,
independent of the mode ratio.
Previously a linear delamination failure criterion was proposed for delamination failure
under static loads [24]. This criterion had the form
GI GII
G,'~ + ~ = 1 (17)
Figure 19 from Ref 24 shows interlaminar fracture toughness data plotted from the literature
for materials with matrices ranging from very brittle to very tough. Pure mode I data (G~c
values) were generated using double cantilever beam (DCB) specimens and are plotted on
12
~'~, EDT t
-K~, DCB
8
in
L \.
4 FJ"'-4W..~ ""')~-Hx205 ~
"\..
0 2 4 6 8 I0
in - Ib
Gll' in2
FIG. 19--Mixed mode fracture toughness [24].
the ordinate. Pure mode II data (G.c values) were generated using ENF specimens and are
shown on the abscissa. Mixed mode data (Go values) were generated using edge delamination
tension (EDT) and crack lap shear (CLS) specimens and are plotted at the appropriate
coordinates according to the G~ and G. component for each test. For all the materials, the
data fit the linear criterion given by Eq 17. However, for the brittle materials, like the epoxy
matrix materials in this study, Gxc "~ G.c, whereas for the toughened matrix materials, such
as AS4/PEEK, Gtc was nearly equal to Guc. Therefore, for the toughened matrix materials,
noting that G --- G~ + G. and that G~c = G.c, Eq 17 reduces to
G = Gc (18)
A linear failure criterion similar to Eq 17 may be assumed for delamination onset under
cyclic loading as
GI G, = 1 (19)
Gtt'--'~h+ Gmh
However, as indicated in Figs. 17 and 18, the threshold value for delamination onset in
fatigue appears to be independent of mode ratio, even for the brittle epoxy matrix materials
where the static toughness is dependent on mode ratio.
Figure 20 shows the static and fatigue delamination strain energy release rate data for
T300/BP907, Linear failure criteria are plotted for the static interlaminar fracture toughness
and fatigue threshold between the predominantly mode I (EDT) and mode II (ENF) tough-
ness values. The fatigue envelope is lower than the static envelope, with the greatest apparent
reduction occurring for the pure mode II tests as indicated in Fig. 17. The static envelope
is skewed because Glc ~ G.c, whereas the fatigue envelope approaches a 45~ line reflecting
the near equality of the threshold values of Gj and G.. Therefore, for the brittle epoxy
matrix composites, the fatigue delamination criterion of Eq 19 simply reduces to
Gm,x = G,h (20)
Figure 21 shows the static and fatigue delamination strain energy release rate data for
AS4/PEEK. Linear failure criteria are plotted for the static interlaminar fracture toughness
and fatigue threshold between the predominantly mode I (EDT) and mode II (ENF) tough-
ness values. The fatigue envelope is lower than the static envelope, with the greatest apparent
Static Fatigue Layup ~}~,G]] Test

(15=1) R=O.] f=PHZ
61 9 O 024 100% ENF
9 z~ ( +35/O/gO)s I0~ EDT
Grin- Ib 42
in2 .~ GI + GII=I
o
I I I I I I
0 2 4 6 8 I0
GII, in-._Ib
in 2
FIG. 20--Mixed mode fracture toughness and fatigue threshold criteria for T300/ BP907.
O'BRIEN ET AL. ON SHEAR FRACTURE T O U G H N E S S AND FATIGUE T H R E S H O L D S 247
Static Fatigue [ayup %GII Test

(N=I) R=O.1 f=-SHZ
9 [] 024 lOO% ENF
Gr~ in
2 - GI Gll
I i I I I I I I
0 2 4 6 8 I0 12 14
in - Ib
GII' in2
FIG. 21--Mixed mode fracture toughness and fatigue threshold criteria for AS4/ PEEK.
reduction occurring in pure mode II tests. The static envelope is skewed because G~c "~
Gfsc, whereas the fatigue envelope approaches a 45 ~ line reflecting the near equality of the
threshold values of G~ and G,. Therefore, the fatigue delamination criterion of Eq 20 may
apply for the tough AS4/PEEK composite as well as for the epoxy matrix composites.
However, several inconsistencies in AS4/PEEK data have been noted in the literature [8].
This may be illustrated by comparing the data in Figs. 19 and 21. The static toughness data
for PEEK composites shown in Fig. 19 indicate that a total G criterion should apply for this
material (that is, G~c ~ G.c). However, the data in Fig. 21 indicate that G~c < G,c. The G.c
value in Fig. 19 was calculated using ENF specimens that exhibited unstable propagation
from a tensile precrack. This value is similar to the G sc values reported in this study using
ENF specimens with shear precracks. However, the G.c value extrapolated from the pre-
dominantly mode I E D T layup is lower than the G~c value in Fig. 19 that was measured
from DCB tests. This difference is minimized if the influence of residual thermal stresses
to G in the EDT test are included in the data reduction [8,25]; however, this would also
increase the apparent G,h for this predominantly mode I case over the Gmhmeasured in this
study using the ENF test, which is difficult to rationalize physically. Because of the many
variables that may influence toughness for the semicrystalline PEEK thermoplastic matrix
composites, a detailed study should be conducted using a variety of tests on panels of this
material as it is currently produced to fully characterize the toughness of AS4/PEEK.
Residual Static Tests

The degradation in residual G.c values for $2/SP250 laminates after 106 cycles (Tables 8
and 9) indicates that matrix damage was created at the delamination front even though no
coalescence occurred resulting in delamination growth. These data would suggest that de-
lamination growth might occur at very long fatigue lives, perhaps on the order of 107 to 109
cycles. In some composite structures, such as helicopter rotor blades and hubs, these long
lives are very common. Hence, for very long-term durability, the G,h values measured at
10~ cycles may be unconservative. Testing for delamination onset after 107 or more cycles
is needed to confirm this perception. This long-term testing may be very time consuming
and costly. However, the flatness of the Gin,, versus cycles curves between 105 and 106 cycles
(Figs. 14 through 18) suggests that G,h values will probably decrease very little beyond 106
cycles.
Conclusions
Based on the analysis and reduction of test data for the materials tested in this study, the
following conclusions were reached:
1. Finite-element analysis of the end-notched flexure (ENF) specimen indicates that the
beam theory calculation for GH with the transverse shear contribution included is reasonably
accurate over the entire range of crack lengths. These GS~ values are slightly conservative
compared to the finite-element results, but the difference is minimal when the contribution
of friction is included with the finite-element results. Hence, GtSin values represent reasonably
accurate calculations of the strain energy release rate associated with interlaminar shear in
the ENF test.
2. ENF specimen compliance measurements and the variation in compliance with delam-
ination length for the materials tested differed from the beam theory. This difference was
attributed to the variation in ply thickness and, hence, the variation in fiber volume fraction
through the thickness of the ENF specimens. For the materials tested in this study, the axial
stiffness estimated from flexural tests was consistantly lower than values measured on the
same ENF specimens in axial tension tests.
3. For the materials that exhibited linear load-deflection behavior in the ENF three-point
bending test, GHc values determined from compliance calibration measurements provided
the most conservative and accurate estimate of interlaminar shear fracture toughness.
4. Cyclic loading significantly reduced the critical G~I for delamination onset. The max-
imum cyclic GII level below which no delamination was observed after 106 cycles (that is,
the threshold cyclic GH for delamination in fatigue) was determined for the three materials
and was compared to the static G~ic using the ENF test to quantify the degradation in
interlaminar shear fracture toughness due to fatigue.
5. Toughened matrix materials, and brittle matrix materials that underwent low-load/
high-cycle fatigue loading, exhibited nonlinear load-displacement behavior in the ENF test.
For these cases, GII calculated using the load at deviation from linearity may provide a more
accurate and conservative estimate of the interlaminar shear fracture toughness.
6. Although a linear mixed-mode failure criterion is needed to characterize the static
interlaminar fracture toughness of some composite materials, a total G threshold criterion
appears to be sufficient for characterizing the fatigue delamination durability of composite
laminates with a variety of static toughnesses.
Acknowledgments
The authors wish to acknowledge the contribution of Tracy Bridges of NASA Langley
who designed and fabricated the three-point bend fixture and performed the static and
fatigue tests in this study. The authors also wish to acknowledge the contribution and
suggestions of Dr. I. S. Raju of Analytical Services and Materials, Inc., for the finite-
element analysis that was performed.
References
[1] Abdallah, M. G., "Review of the State of the Art of Advanced Composite Interlaminar Shear
Test Methods," Proceedings, 1986 SEM Spring Conference on Experimental Mechanics, New
Orleans, LA, June 1986, Society for Experimental Mechanics, Bridgeport, CN.
[2] Adams, D. O. and Kearney, H. L., "Full-Scale Fatigue Testing of Advanced Fiber Composite
Components," Journal of the American Helicopter Society, Vol. 31, No. 2, April 1986, p. 66.
[3] Whitney, J. M. and Browning, C. E., "On Short-Beam Shear Tests for Composite Materials,"
Experimental Mechanics, Vol. 25, No. 3, Sept. 1985, p. 294.
[4] O'Brien, T. K., "Interlaminar Fracture of Composites," Journal of the Aeronautical Society of
India, Vol. 37, No. 1, Part III, Feb. 1985, p. 61.
[5] O'Brien, T. K., "Mixed-Mode Strain-Energy-Release Rate Effects on Edge Delamination of
Composites," in Effects of Defects in Composite Materials, ASTM STP 836, American Society for
Testing and Materials, Philadelphia, 1984.
[6] O'Brien, T. K., "Generic Aspects of Delamination in Fatigue of Composite Materials," Journal
of the American Helicopter Society, Vol. 32, No. 1, Jan. 1987, p. 13.
[7] Adams, D. E, Zimmerman, R. S., and Odem, E. M., "Determining Frequency and Load Ratio
Effects on the Edge Delamination Test in Graphite Epoxy Composites," in Toughened Composites,
ASTM STP 937, American Society for Testing and Materials, Philadelphia, 1987.
[8] O'Brien, T. K., "Fatigue Delamination Behavior of PEEK Thermoplastic Composite Laminates,"
in Proceedings, American Society for Composites First Technical Conference, Dayton, OH, 1986,
p. 404; Journal of Reinforced Plastics, Vol. 7, July 1988, p. 341-359.
[9] Russell, A. J., "On the Measurement of Mode II Interlaminar Fracture Energies," Defence
Research Establishment Pacific (DREP), Victoria, British Columbia, Canada, Materials Report
82-0, Dec. 1982.
[10] Russell, A. J. and Street, K. N., "Moisture and Temperature Effects on the Mixed-Mode Delam-
ination Fracture of Unidirectional Graphite/Epoxy," in Delarnination and Debonding of Materials,
ASTM STP 876, W. S. Johnson, Ed., American Society for Testing and Materials, Philadelphia,
1985, p. 349.
[11] Russell, A. J. and Street, K. N., "Factors Affecting the InterlaminarFracture Energy of Graphite/
Epoxy Laminates," in Progress in Science and Engineering of Composites, Proceedings of the
Fourth InternationalConference on Composite Materials (ICCM-IV), Tokyo, 1982, Elsevier, New
York, p. 279.
[12] Murri, G. B. and O'Brien, T. K., "Interlaminar GHcEvaluation of Toughened Resin Composites
Using the End-notched Flexure Test," AIAA-85-0647, Proceedings, Twenty-sixth AIAA/ASME/
ASCE/AHS Conference on Structures, Structural Dynamics, and Materials, Orlando, FL, April
1985, AIAA, New York, p. 197.
[13] Mall, S. and Kochhar, N. K., "Finite Element Analysis of End Notched Flexure Specimens,"
Journal of Composites Technology and Research, Vol. 8, No. 2, Summer 1986, p. 54.
[14] Carlsson, L. A., Gillespie, J. W., and Pipes, R. B., "On the Analysis and Design of the End
Notched Flexure (ENF) Specimen for Mode II Testing," Journal of Composite Materials, Vol. 20,
Nov. 1986, p. 594.
[15] Gillespie, J. W., Carlsson, L. A., and Pipes, R. B., "'Finite Element Analysis of the End Notched
Flexure Specimen for Measuring Mode II Fracture Toughness," Composites Science and Tech-
nology, Vol. 27, 1986, p. 1.
[16] Carlsson, L. A., Gillespie, J. W., and Trethewey, B. R., "Mode II Interlaminar Fracture of
Graphite/Epoxy and Graphite/PEEK," Journal of Reinforced Plastics and Composites, Vol. 5, July
1986, p. 170.
[17] Carlsson, L. A., Gillespie, J. W., and Whitney, J. M., "Interlaminar Fracture Mechanics Analysis
of the End Notched Flexure Specimen," in Proccedings, American Society for Composites First
Technical Conference, Dayton, OH, 1986, p. 421.
[18] Gillespie, J. W., Carlsson, L. A., Pipes, R. B., Rothschilds, R., et al., "Delamination Growth in
Composite Materials," NASA Contractor Report 178066, 1985.
[19] Smiley, A. J. and Pipes, R. B., "Rate Sensitivity of Interlaminar Fracture Toughness in Composite
Materials," in Proceedings, American Society for Composites First Technical Conference, Dayton,
OH, 1986, p. 434.
[20] Trethewey, B. R., Carlsson, L. A., Gillespie, J. W., and Pipes, R. B., "Mode II Interlaminar
Fracture During Static and Fatigue Loading," Center for Composite Materials Report CCM 86-
26, University of Delaware, Newark, DE, Sept. 1986.
[21] Wakelyn, N. T., "Resolution of Wide Angle X-Ray Scattering from a Thermoplastic Composite,"
Journal of Polymer Science: Part A: Polymer Chemistry, Vol. 24, 1986, p. 2101.
[22] Rybicki, E. E and Kanninen, M. F., "A Finite Element Calculation of Stress Intensity Factors
by a Modified Crack Closure Integral," Engineering Fracture Mechanics, Vol. 9, No. 4, 1977, pp.
931-938.
[23] Salpekar, S. A., Raju, I. S., and O'Brien, T. K., "'Strain Energy Release Rate Analysis of the
End-Notched Flexure Specimen using the Finite Element Method,"Journal of Composites Tech-
nology and Research, Vol. 10, No. 4, Winter 1988.
[24] Johnson, W. S. and Mangalgiri, P. D., "Influence of the Resin on Mixed-Mode Interlaminar
Fracture," in Toughened Composites, ASTM STP 937, American Society for Testing and Materials,
Philadelphia, 1987.
[25] O'Brien, T. K., Johnston, N. J., Raju, I. S., Morris, D. H., and Simonds, R. A., "Comparisons
of Various Configurations of the Edge Delamination Test for Interlaminar Fracture Toughness,"
in Toughened Composites. ASTM STP 937, American Society for Testing and Materials, Phila-
delphia, 1987.
Yves J. Prel, 1 Peter Davies, 2 Malk L. Benzeggagh, 3 and
Francois-Xavier de Charentenay 4
Mode I and Mode II

Delamination of Thermosetting and
Thermoplastic Composites
REFERENCE: Prel, Y. J., Davies, P., Benzeggagh, M. L., and de Charentenay, E-X.: "Mode
I and Mode 11 Delamination of Thermosetting and Thermoplastic Composites," Composite
Materials: Fatigue and Fracture, Second Volume, ASTM STP 1012, Paul A. Lagace, Ed.,
ABSTRACT: A study of the delamination behavior of three classes of composite material is

described: glass/epoxy, graphite/epoxy, and graphite/PEEK (polyetheretherketone). Mode
I loading was applied to double cantilever beam (DCB) specimens of two thicknesses, while
three different types of mode II specimen were tested: end-notched flexure (ENF), end-notched
cantilever beam (ENCB), and cantilever beam enclosed notch (CBEN). The latter was de-
veloped specifically for this study and is described in detail. A mixed mode specimen is also
described. Results from both static and fatigue tests are discussed in terms of fracture mechanics
values.
KEY WORDS: composite materials, delamination, mode I, mode II, mixed mode, glass/
epoxy, graphite/epoxy, graphite/PEEK, unidirectional
In recent years, a concerted effort directed toward greater understanding of composite

delamination has resulted in the development of improved composite materials. However,
as tougher matrices are proposed, the techniques employed to characterize them must also
be developed; fracture mechanics has proved an effective tool, and currently a number of
tests using this approach are used widely. While no standards exist yet, considerable ex-
perience has been gained, particularly with the double cantilever beam (DCB) specimen
for mode I testing. Since early applications of the cleavage test to glass/epoxy composites
[1], many papers have described different aspects of this test. Thus both experimental
procedures [2-5] and data reduction methods [6-9] have been presented in some detail.
Data from a range of mode II (shear) tests have also been reported [10-13], but these tests
are used less widely at present.
An investigation into the response of composite materials to delaminating loads requires
a consistent data base, including confirmation of specimen geometry independence and,
where possible, the use of several different test methods. The work presented here is in
three parts, aimed at establishing such a data base using existing and specially developed
fracture mechanics tests and then leading to a discussion of the behavior of different types
Engineer, Centre National d'Etudes Spatiales, Evry, France.

2 Researcher, Ecole Polytechnique F6d6rale de Lausanne, Switzerland.
3 Lecturer, Division Polym~res et Composites, Universit6 de Technologie de Compi~gne, BP 649,
60206 Compi~gne, France.
4 Research director, Peugeot S.A., Velizy, France.
251
Downloaded/printed
Copyright9 byby
of composite material. The materials studied are glass/epoxy, graphite/epoxy, and graphite/
PEEK.
In the first section, results are given for a series of mode [ tests on DCB specimens of
two thicknesses under static loading. A comparison of the performance of two materials
under fatigue loading is also given. In the second part, results from two existing shear tests,
employing the end-notched flexure (ENF) [10] and end-notched cantilever beam (ENCB)
[12] specimens under static loading, are compared with results obtained using a test
developed at Compi~gne [14]. The latter, using the cantilever beam enclosed notch (CBEN)
specimen, is described in some detail, and both analytical and experimental results are
presented. A comparison under fatigue loading is given again. In the third section, the
development of this test methodology is followed with the description of a mixed mode I-
II specimen.
Materials
The unidirectional composite materials studied in this work are presented in Table 1
together with their modulus values used in subsequent calculations. All were obtained in
the form of thin (4.5 to 5.2 mm) and thick (20 mm) panels molded using the prepreg suppliers'
recommended temperature and pressure cycles. The graphite/epoxy was postcured for 4 h
at 190~
Thin films were placed at the midthickness of panels during molding to act as starter
cracks. The film material was PTFE for the epoxies and aluminum for the PEEK-based
composites, of similar thickness (50 I~m).
Specimen Preparation
The panels were cut with a diamond-impregnated disk to the dimensions shown in Fig.
1. They were stored in a controlled environment chamber at 23~ until testing.
Mode 1 Procedure
Thin DCB specimens were loaded through bonded hinges, whereas holes were drilled
directly in thick specimens for load introduction (Fig. 1). Loading rate was 2 mm/min. At
least four specimens of each material and thickness, with different starter crack lengths,
were tested to allow compliance calibrations to be made for each material and specimen
thickness. The calibration employed was that proposed by Berry and described in Ref 8:
a n
D
c h (1)
TABLE 1--Materials and properties for epoxy composites, moduli given for Vf = 60%.
vf , %
Material Type Fiber Thick Thin Matrix E~, GPa G~, GPa Source
Glass/epoxy E-glass 60 60 DGEBA 46 5 Brochier SA (MI0)
Graphite/epoxy T300 55 60 914 141 6.2 Ciba-Geigy
Graphite/PEEK AS4 61 61 polyether- 134 5. l ICI plc (APC2)
etherketone
PREL ET AL. ON MODE I AND MODE II DELAMINATION 253
5
a) ~7~o
i.ao_l
b, o
i i
i_ ao=,
I . o~ L I
el 1 "~
"V l a.t
r_ L _,
L
FIG. 1--Specimen dimensions (ram) and loading arrangements. (a) Thin DCB, (b) Thick
DCB, (c) ENF, (d) CBEN, (e) ENCB, (f) Mixed mode.
The plot of log compliance versus log initial crack length enabled the empirical parameters
n and h to be found. An R curve was then plotted for each specimen, representing the value
of Gj as a function of the effective crack length a'. The latter was calculated from the
compliance at each point on the load-displacement curve and then
nP~
GI - 2Ba' (2)
Values of G~c at initiation were defined by the first acoustic emission and corresponded
closely to the first nonlinearity on the load-displacement curve. Values of GIp were deter-
mined as the mean value of G~ during propagation, corresponding to the plateau of the R
curve. A detailed description of this method of mode I data reduction is to be found in Ref
8, together with a number of examples of R curves for different materials.
Mode I fatigue tests were performed on thick DCB specimens under load control, with
an R ratio (Pm,,/Pm,,) of 0.1 and a sinusoidal load frequency of 1 Hz. Load and displacement
were continuously recorded to enable crack length to be determined from the static com-
pliance calibration. These crack length values were checked by applying a thin layer of
white ink to the specimen side and following propagation with a traveling microscope. The
data recording system has been described previously [14].
Mode II Procedure
Three different shear delamination tests were used, employing the CBEN, ENCB, and
E N F specimens loaded in bending. Their dimensions and configurations are shown in Fig.
1. The ENF specimen was introduced by Barrett and Foschi for testing of wooden beams
[13] and has proved well-suited to testing thin specimens of relatively brittle composites
[10]. The ENCB specimen is a thick version of the end-loaded split (ELS) specimen used
by Vanderkley [12]; Bathias and Laksimi used similar specimens in mode II fatigue tests
[15]. Most specimens were precracked a few millimetres from the starter film in mode I,
but some specimens without precracks were also tested.
The test setup shown in Fig. 2 was employed to load the CBEN and ENCB specimens
at 5 mm/min; E N F specimens were loaded in a three-point bend fixture at 0.5 mm/min.
Data reduction methods are described fully later.
Mode II fatigue tests were performed on CBEN and ENCB specimens using load control,
with an R ratio of - 1 and frequency of 5 Hz.
Mixed Mode Procedure

The mixed mode specimen is prepared by removing a section of the CBEN specimen to
free the upper half of the beam as shown in Fig. 1. The specimen was mounted in the setup
shown in Fig. 2. To study the contributions of mode I and mode II, either a or L may be
varied. In this work, a was varied from 20 to 70 mm with L = 130 mm, and then L was
varied from 90 to 130 for a = 50 mm. Tests were performed under displacement control of
2 mm/min.
I
Do GRIPPING SYSTEM
il ( ,,LoAo.,L.
B: SPECIMEN [.- ACOUSTIC ISOLATION
I ~J C: TRANSDUCER F: HYDRAULIC PISTON
G= COUNTERWEIGHT
H: MOBILE HEAD
a0:INITIAL CRACK LENG1H
MIXED MODE AND MODE II SET UP

FIG. 2--Experimental setup used to test CBEN, ENCB, and mixed-mode specimens.
TABLE 2--Mode I fracture energies. (Standard deviation in brackets.)
Material Thickness, Initiation, Propagation,

Type mm Gtc, J/m-' Gre, J/m 2 Comments
E-glass/DGEBA 4.5 264 (34) 515 (89) Fiber bridging equilibrium

20 228 (23) ... Bridging-no plateau
reached
T300/914 5 185 (6) 176 (6) No bridging
20 185 (22) 151 (22) No bridging
Stable Unstable
Onset A rrest
AS4/PEEK 5.2 1460 (90) 2400 (140) 2470 (220) 970 (240)
20 1500 to 3 5 0 0 Unstable 2378 (550) 563 (81)
Results
Mode I Static
The results obtained from mode I tests on DCB specimens of two thicknesses for each
type of material arc shown in Table 2.
Fiber bridging in the glass/epoxy specimens raised propagation values, whereas no bridging
occurred in the graphite/epoxy. The formation of fiber bridges has been shown to be related
to specimen stiffness [16], matrix toughness [17], and the thickness of the interply layer
[18]. For the thin glass/epoxy specimens tested here, a constant propagation value of 515
J/m 2 was reached, whereas for the thick specimens a continuous increase up to values in
excess of 1000 J/m 2 was recorded [14]. When bridges form, the resulting increase in delam-
ination resistance conceals any matrix-toughening effects, DCB tests on unidirectional spec-
imens will then be of limited use in describing the behavior of other layups.
There was little difference between initiation values for thick and thin specimens of either
of the epoxy composites. For thick graphite/PEEK specimens, however, the values char-
acterizing initiation from the starter film tended toward thin specimen values only when
long starter cracks were used as shown in Fig. 3. This may be explained in terms of the
matrix-rich zone in front of the starter film; higher strain rates corresponding to short cracks
result in a smaller damage zone, so that it is the matrix alone, rather than the adjacent
plies, which is loaded. Therefore, for short starter crack lengths, the values obtained tend
toward the pure matrix toughness values. At the lower strain rates corresponding to longer
cracks, the plies above and below the matrix-rich region are also loaded. The initiation
values obtained thus become increasingly representative of the composite rather than the
pure matrix, and a previous fractographic study has shown that, for long starter crack lengths,
the crack initiation takes place at the boundary between the ply and the matrix-rich region
for both thick and thin specimens [19].
Both stable and unstable propagation occur in graphite/PEEK [20]. Therefore, three
values were used to characterize propagation, all determined using the compliance calibration
obtained from specimens with different starter crack lengths. Values for stable propagation
in thin specimens (thick specimens showed nearly all unstable propagation) are considerably
higher than the initiation values, due at least in part to the formation of bridges of fiber
bundles. Unstable propagation was characterized by values at the onset of instability, which
are similar for both specimen thicknesses, and arrest values. There was considerable scatter
4
[] Thick
Gic
9 Thin
kJ/m2
[] []
[]
A 9 m
o 3~)
6'o s'o t2o 15o
Starter crack length , mm
FIG. 3--Mode 1 G~c values for different initial starter crack lengths, graphite~PEEK.
in the values characterizing the onset of unstable propagation in graphite/PEEK, but their
similarity to the stable propagation value suggests that local events, such as fiber bundle
breakage, rather than a global critical strain energy release rate, determine the onset of
instability.
Mode I Fatigue
Results are presented in the form of AG versus da/dn plots for glass/epoxy and graphite/
PEEK in Fig. 4. The former can support AG values well in excess of its G~c value as fiber
bridges increase G~p at least as rapidly as the applied G~ increases with crack propagation.
For graphite/PEEK, however, the applied AG increases until a value within the spread of
values obtained for the onset of instability under static loading and then unstable propagation
occurs.
The results were fitted to a power law of the form
d a / d N = C(AG) m (3)
For the glass/epoxy an exponent of 1.6 is obtained. A value of 3.7 has been published
for a glass-cloth-reinforced epoxy under similar loading conditions [15], but the higher resin
volume fraction and more clearly defined interply layer of that material would be expected
to facilitate delamination growth. For graphite/PEEK, a considerably higher m value of
10.5 was found.
Mode II Static Tests

Mode II tests were performed on the CBEN specimen under static and cyclic loading.
Additional tests were also carried out on ENCB and E N F specimens to provide a comparison.
The critical strain energy release rate, Guc, was calculated by the Irwin-Kies compliance
method:
p2 dC
GHc - 2w da (4)
The compliance calibration, C = f(a), was obtained using several methods as shown in
Table 3.
da ram/cycle
10 -I
MODE I
R-0.1, f=lHz
9 GLASS/EPOXY
* G R A P H I T E / PEEK
10 -2
10-3
I~ m = 1 0 . 5
10-4
I I I I I ; ; ; ~ o I I I I ~ ,
10 2 10 3 10 4
G I (J/m 2)
FIG. 4--Mode I fatigue crack propagation results, glass~epoxy and graphite~PEEK.
TABLE 3--Methods used to obtain compfiance calibrations for the three specimen types, and
reference suffixes.
Specimen Type
Method CBEN ENCB ENF
Beam theory BT
Finite element FE FE
Static tests ST
Fatigue tests FAT FAT
The C B E N Specimen
Two methods were used to determine the compliance calibration:
1. The finite-element method: Specimens of different crack lengths were modeled using
two-dimensional quadratic elements in plane stress. The material behavior was considered
to be linear elastic and transversely isotropic. Boundary conditions at the built-in ends are
accounted for by contact elements, as described in Ref 21. Contact elements are used also
on the crack faces to avoid the interpenetration of the two half-beams (Fig. 5). Contacts at
the supports, loading points, and crack faces are assumed to be friction-free. Previous work
has indicated that this leads to an error of less than 5% for the CBEN specimen geometries
used in this work [22].
The compliance calibration obtained may be fitted to a curve of the form CpE = et + [3a3
(FE = finite element).
2. Fatigue crack propagation tests: The load-displacement plots from static tests show a
slightly nonlinear behavior up to unstable fracture, which makes exact compliance deter-
mination difficult. In addition, the compliance change with crack length is relatively small
FIG. 5--Finite-element mesh used to determine compliance of CBEN specimen.
Compliance
mm l N x I0 -3
2 1 . . . . . . .,
C=a +Ba 3
L =120mm
1.5
FAT
i i t L t i
50 70 90 110
crack length , mm
FIG. 6--Compliance calibration for glass / epoxy CB EN specimens by two different methods.
compared with that measured in the mode I tests, and because each point on the C = f(a)
plot comes from a different specimen considerable scatter is obtained. Therefore, a more
reliable approach is to use fatigue crack propagation tests on single specimens.
These tests, for R = - 1 a n d f = 5 Hz under load control, are carried out with continuous
recording of load and displacement, which enables the compliance change with crack length
to be determined. Again, the relation obtained may be fitted to a curve of the form CFA1 =
a + 13a3.
3. Comparison of the two methods: In Fig. 6, the compliance calibrations obtained by
both methods for glass/epoxy are plotted. The values of c~ and 13 determined using each
method are presented in Table 4 for the three materials. These coefficients reflect the
considerable differences in stiffness between the graphite- and glass-reinforced materials.
While reasonable agreement is obtained between theory and experiment, the compliance
values calculated by the analytical expression are very sensitive to changes in dimensions,
particularly thickness, which may account for differences. The modeling of the clamping
conditions is also prone to error because the distance over which the beam is fixed must be
selected in the finite-element model.
Once the compliance calibration has been carried out, GIIC may be determined.
TABLE 4---Values o f the parameters et (ram~N) and f5 ( N -~ m m -~') for the compliance calibrations,
C B E N specimen, based on 20 • 20 m m 2 section, L = 120 ram.
ct(• I0 4) 13(x 10-~~
Material FE FAT FE FAT
Glass/epoxy 13.6 11.6 5.43 4.44

Graphite/epoxy 5.11 1.98
Graphite/PEEK 5.11 4.73 1.97 1.93
The critical strain energy release rate is calculated using the following expression:
p2 dC 3P 2
Gnc = 2 w da 2w 13a2 (5)
The theoretical coefficients, a and 13, which were determined for a 20 mm • 20 mm cross
section, are corrected for differences in geometries of individual specimens by a factor of
the form (wh3/12), where w = width and h = thickness. P is the maximum load, and two
cases are examined. In the first case, tests were performed on specimens containing the film
starter crack but not precracked in mode I. In this case, P corresponds to the onset of
unstable propagation without subcritical propagation. In the second case, precracked spec-
imens were tested and the load-displacement plot showed more nonlinear behavior before
the maximum load. This is due to a slow propagation phase visible by eye in the glass/
epoxy. Examples of the load-displacement plots recorded in the two cases are shown in
Fig. 7.
LOAD
kN
6
L =120ram
a =77ram
0
Dlsplacernent. mm
FIG. 7--Load-displacement plots, glass~epoxy C B E N specimens. (a) N o precrack and (b)

precracked.
PREL ET AL. ON MODE I AND MODE II DELAMtNATION 261
TABLE 5--Mean values of O,c (J/m2), CBEN specimen. (Standard deviation in brackets. )
No Precrack Precracked
Material FE FAT FE FAT
Glass/epoxy 3510 2870 2510 2050
(457) (374) (112) (92)
Graphite / epoxy 795 598
(50) (55)
Graphite / PEEK 4250 3960 2695 2507
(500) (464) (193) (177)
The Guc values for the three materials, calculated using both compliance expressions, are
presented in Table 5.
Tests on E N C B and E N F Specimens

For the ENCB specimen, G~c is also calculated using the compliance method. Three
methods were used to obtain the compliance calibration, and, for glass/epoxy, the results
may also be accurately fitted by curves of the form C = et + 13a3. Values of the parameters
(x and 13 for glass/epoxy, presented in Table 6, may be compared with values obtained for
CBEN specimens of the same free length in Table 4. The ratio of the coefficient is about
4, the CBEN compliances evolving relatively little with crack length.
For the ENF specimen, G~c is calculated using the following analytical expression [23],
which includes a shear correction:
El ( h] ~) a 2
Giic - 2Etw2h [ 1 + 0.2 ~313 (6)
where
h = specimen thickness and
w = width.
For ENCB specimens, Gnc is determined using the maximum load because no propagation
was observed before this point. Load-displacement plots for ENCB tests on graphite/PEEK
(shown in Fig. 8) illustrate the increasingly stable propagation for longer initial precracks;
plots from ENF tests evolve in a similar manner [24]. However, for graphite/PEEK and
for precracked glass/epoxy ENF specimens, some propagation may be noted before the
maximum load. Therefore, two values are calculated: a value at the onset of nonlinearity
using the initial crack length and a value at maximum load using a crack length calculated
TABLE 6--Values of the parameters e((mm/N) and f3 (N -1 mm -2) for compliance calibrations on
glass~epoxy, for ENCB specimen based on 20 mnm x 20 mm section, L = 120 ram.
Parameters FE FAT ST
a 1.37 x 10_3 1.27 1.36
13 1.9 x 10-9 1.95 1.71
LOAD
kN
5 84
2 4
Displacement , mm
FIG. 8--Load-displacement plots, graphite~PEEK, ENCB specimens.
from the compliance at that point. These values are lower and upper bounds because it is
not clear exactly when crack propagation begins. Tables 7 and 8 summarize the results for
the three materials. Values calculated by the finite-element method for glass/epoxy CBEN
and ENCB specimens agree quite closely.
Mode 1I Fatigue Tests

Under fatigue loading, resistance to both initiation and to propagation of a delamination
may be measured. Results for initiation resistance of glass/epoxy have been published
previously [25]. Here, propagation has been studied, and the ENCB and CBEN specimens
have been compared. The loading conditions (for R = - 1, f = 5 Hz, L = 120 mm in both
cases) may be represented by either AG or G,,. AG and Gm are defined as follows:
AG = (AP)2(
d C ) 2 w~ a = 3(AP)Z13aZ2w
TABLE 7--Mean values of G,, (J/m 2) obtained on ENCB specimens. (Standard deviations
in brackets.)
No Precrack Precracked
Material ST FE ST FE
Glass/epoxy 3000 3340 2110 2345

(213) (238) (135) (150)
Graphite/PEEK 1780 2610
(178) (260)
TABLE 8---Mean values o f Guc (J/m 2) from tests on ENF specimens, a (Standard deviations
in brackets. )
Material No Precrack, BT Precracked, BT

Glass/epoxy 3640 (377) 1715 (211)-2275 (218)
Graphite/epoxy 1035 (44) 518 (27)
Graphite/PEEK 2140 (46)-4413 (143) 1109 (230)-3678 (280)
"Where two values are given, these correspond to nonlinearity and maximum load points.
where
Z~P = 2Pro and P,, = the maximum load applied.
3 p,,2[~a2
G,. - 2-~ (7)
that is, A G = 4Gin.

Figure 9 shows crack propagation plots, primarily to illustrate the results for the two types
of test on glass/epoxy. However, a few points are also plotted for graphite/PEEK, but more
tests are required to confirm these values. Glass/epoxy crack lengths can be measured
directly by traveling microscope. For graphite/PEEK, however, visual detection of the crack
front is very approximate in mode II. Therefore, a compliance calibration was obtained by
d___~a ram/cycle d.~a_aram~cycle

dN dN
10-I 10-1
o~.
.oD ,,
R=-I , f=5Hz ~ "" oo~
ua
10 -2 10 -2
o~o
,~',o m=2.8
10-3 ~0 ~ 10 -3
10-4
"~""
u
9o
(a)
10 -4
"if"" (b)
9 GLASS/EPOXY 10 -5 9 GLASS / E ~ X Y
10-~
9 8 CBEN s p e c i m e n s 3 ENCB specimens
9 GRAPHITE / PEEK
102 103 104 10Z 10] 104
aGii (Jim l ) t~GII (Jim 2)
F I G . 9 - - M o d e l l fatigue crack propagation results from (a) C B E N and (b) ENCB speci-
mens.
placing strain gauges along the specimen length to indicate the passage of the crack. From
the strain gauge responses, the crack position at a number of points could be identified. By
plotting the compliances corresponding to these points, intermediate crack lengths could be
calculated.
The results for glass/epoxy shown in Fig. 9 may be fitted to equations of the form
da da
d-N = C(AGII)m or -~ = C(Giim) m (8)
and the coefficients C and m are also presented on the figure. It may be noted that the
coefficient m is quite similar for tests on the glass/epoxy using the different specimens. Few
published data are available for comparison; Bathias and Laksimi found a value of 7.6 for
a glass-fabric-reinforced epoxy, but for R = 0.01 [15].
Mixed Mode Tests

To determine the contributions of mode I and mode II to the delamination initiation for
the mixed mode specimen shown in Fig. 1, an analysis was performed using the finite-
element method in conjunction with linear elastic fracture mechanics. This analysis, which
has been described previously [21], enables the following fracture parameters to be deter-
mined:
9 Total strain energy release rate, G

9 The J integral
9 Integrals J, and Ji,
9 Stress intensity factors, K, and K,,
The strain energy release rate may be calculated in two ways: by the energy method using
virtual crack propagation and by the calculation of the integrals J, and J,, for a contour
around the crack tip as proposed by Bui [26].
From Ji and J,, it is possible to calculate K, and K,, using the relation
K, 2
J' E,* (9)
where
i = I, II, and
E,* = the effective modulus.
Two types of test may be distinguished:
1. For a constant free length, L, the ratio J,,/J~ is independent of the crack length over
the range examined here. These results are summarized in Table 9.
2. For a constant crack length, a, the ratio J,,/J, depends on the free length, the mode
II contribution increasing as free length is reduced.
The results for the two cases are presented in more detail in Ref27. Here only the constant
L case will be considered. The existence of this constant J,,/Ji ratio for a constant free length
allows the Irwin-Kies compliance method to be applied to determine the total strain energy
release rate. Values of G for the constant L case, determined both by the finite-element
TABLE 9---Fracture parameters calculated by virtual crack advance and J integral methods, for
mixed-mode specimen, constant free length = 130 mm.
Crack Length. G, J, Jl, J., K~, K.,

mm J/m z J/m -~ J/m 2 J/m 2 MPa/m ~ MPa/m 1'2 J,/Jf
20 657.3 655.24 356.89 298.35 2.174 2.751 0.836

30 597.84 593.18 324.47 268.71 2.073 2.611 0.828
40 784.9 793.48 415.2 378.28 2.345 3.098 0.911
50 817.37 796.31 430.78 365.53 2.389 3.045 0.848
55 750.01 763.45 396.4 367.05 2.291 3.052 0.926
70 951.7 906.55 470.56 435.99 2.497 3.326 0.9265
Mean (SD) 760 (120) 751 (111) J~i/~ = 0.879(0.042)
method GrE (using the experimental critical load) and from the experimental compliance
values GExp, are presented in Table 10.
The values of J are all greater than the value of Gic of 228 J/m 2 obtained in the static
mode I test, whereas the JH values are all lower than the corresponding G,c of around 3500
J/m 2. For the range of mixed mode loadings considered here, the failure of this glass/epoxy
is dominated by the mode I component [27].
Discussion
A considerable amount of data is presented here, some of which confirm previous work
and some which are original and require further discussion. First, we will consider the mode
I and mode II tests individually, and then, we will examine the similarities and differences
in the behavior of the three materials under the two types of loading. Finally, the mixed
mode test and the direction of further work will be discussed.
There are many published results for mode I delamination resistance, and the values
presented here agree quite closely with those given elsewhere. For example, Ashizawa found
a value of 200 J/m 2 for propagation in 6-mm-thick specimens of the same graphite/epoxy
system [28], while published values for graphite/PEEK (APC-2) range from 1330 J/m 2 at
initiation [29] to 2400 J/m 2 for stable propagation [30].
No previous reports have been found of studies in which such thick DCB specimens of
these materials have been tested. It is reassuring to note in Table 2 that the principal
difference for the two epoxy-based composites is in the effect of fiber bridging on propagation
TABLE IO---G,o,o~(J /m 2) obtained by finite element and experimental compliance calibration.
GFE GEXP
Crack Length, mm Pc, N (Finite Element) (Experimental)
20 550 665.3 831.7

30 590 620.2 741.6
40 780 829.7 919.5
50 930 818.2 772
55 990 722.4 571.4
Mean (SD) 733 (92) 767 (115)
values due to the increased stiffness of the thicker specimens. Initiation values for the two
thicknesses are very similar in both cases. This suggests that the differences that arise from
curing of panels of different thicknesses, such as varying cooling rates through the thickness
and different residual thermal stresses, are secondary considerations for the mode I delam-
ination resistance of these materials.
When thick panels of graphite/PEEK are formed, an additional parameter may be the
development of higher levels of crystallinity, but previous work has indicated that the degree
of crystallinity at the center of the thick specimens was only slightly higher than that measured
on the surfaces and on thin specimens (28% compared to 25%) [19]. While differences in
morphology may accompany this change, recent work suggests that it is unlikely to affect
delamination resistance [31]. Therefore, differences in behavior between thick and thin
specimens are due largely to higher strain rates in thicker specimens. This is particularly
apparent at initiation (as seen in Fig. 3), but it also results in less stable propagation in
thicker specimens. In developing a standard for DCB tests, it may be desirable to specify
at least two loading rates for toughened composites so that strain rate sensitivity can be
detected.
Fiber bridging poses a major problem in mode I delamination characterization, and single
values are clearly inadequate to fully represent material behavior. The use of both initiation
and propagation values or R curves is preferable, but the evaluation of the contribution of
tougher matrix materials to delamination resistance is not straightforward when bridging
occurs. First, it should always be specified when presenting data whether or not bridging
has been observed. The plateau value of G~p obtained from tests on thin specimens embodies
both matrix-toughening mechanisms, such as microcracking and matrix plasticity, and the
contribution of a zone of bridging fibers. Comparisons of propagation values, or values
obtained by the area method [9], for untoughened and toughened composites, therefore,
can be extremely unfavorable to the untoughened materials. Further study of the fiber-
bridging phenomenon is required to assess the influence of fabrication variables and fiber
characteristics.
The threshold value of ziG and the rate exponent, m, which characterize the mode I
fatigue test describe dynamic initiation and propagation behavior. Threshold values were
not determined, but Fig. 4 shows the higher values of AG required to achieve the
same crack propagation rate in the graphite/PEEK as in the glass/epoxy, over the range
studied.
The mode II tests described in published work have been performed mostly on specimens
with open defects (that is, starter cracks growing into the specimen from a free edge). This
is the case for the ENF and ENCB specimens, for example. However, cracks very often
propagate from enclosed defects within components, and a realistic loading situation might
be envisaged in which an enclosed defect in a short cantilever beam is loaded in bending.
In preliminary tests on specimens without artificial (film) defects, failure inevitably occurred
in shear at the built-in end of the beam [25]. Therefore, the CBEN specimen was developed,
with an enclosed starter film, to represent a particularly severe loading situation. Two aspects
of the results obtained are discussed here: the influence of the type of defect and the influence
of specimen type and thickness.
The results in Table 5 for Gnc for the three materials with and without precracks suggest
that the graphite/PEEK is more sensitive to the presence of a starter film than the epoxy-
based composites. Precracking results in a proportionately greater drop in G,c for graphite/
PEEK, and, again, this may be due to the ductile matrix-rich region in front of the starter
film. While it might appear an attractive solution to precrack mode II specimens, and less
scatter is recorded in results from precracked specimens, two points should be noted. First,
resin-rich regions and thin films are realistic defects in composites and, thus, their behavior
under load is of interest. Second, precracking under mode I or fatigue loading will result
in a damaged area, the presence of which may alter subsequent mode II resistance.
In comparing results from the different tests, it is simplest to start with the two involving
thick specimens, the CBEN and ENCB. A major difference between these two loading
arrangements is revealed by the values of 13in Tables 4 and 6. There is relatively low variation
in the compliance of the CBEN specimens with crack length compared with the ENCB
specimen. It should be noted, however, that, even for the ENCB specimen, there is a
considerable difference between experimental and theoretical values for graphite/PEEK.
This may be due partly to uncertainty in the determination of the initial slope of the nonlinear
load-displacement curves. When the finite-element compliance calibrations are used with
experimental critical loads, good agreement is obtained between Guc values from the two
specimens for glass/epoxy and graphite/PEEK. Overall, there does not appear to be a
significant difference between G,c values obtained on the two specimens, and the two sets
of fatigue tests also give very similar results.
The interpretation of E N F tests on tough composites is complicated by nonlinear behavior,
which has been noted previously [24]. For the precracked graphite/epoxy E N F specimen,
where no nonlinearity is observed, Table 8 shows a GHc value similar to that obtained on
the CBEN specimen. For the precracked glass/epoxy E N F specimen a small nonlinearity
gives a range of values close to those from CBEN and ENCB specimens. However, for the
graphite/PEEK ENF tests, a wide range of values may be obtained, depending on when
crack initiation is assumed to occur. This test clearly requires further study, and recent
papers have suggested a nonlinear fracture mechanics approach and a modification to the
beam theory data reduction method for tough composites [24,32].
If we now consider the behavior of the different materials under mode I and mode II
loading, a number of observations may be made. The ratio of mode II to mode I fracture
energies at initiation varies considerably, from around 10 for the glass/epoxy to 3 or 4 for
the graphite/epoxy to 1.5 for the graphite/PEEK. The reasons for such differences may be
explained in terms of the different mechanisms acting, and fracture surfaces of the glass/
epoxy [27], the graphite/epoxy [33], and the graphite/PEEK [34] have been published
elsewhere.
Clearly, once mode I and mode II have been studied, the aim is to establish a failure
criterion allowing more realistic loading conditions to be treated. The mixed mode specimen
was developed with this aim in view, and the results for the glass/epoxy underline the
complexity of this project. A number of authors have examined mixed mode fracture, and
a general consensus is that as mode II contribution increases, so the overall Gc to failure
increases [35-37]. The results shown here for glass/epoxy follow this trend. As matrix
toughness is increased, so the difference between mode I and mode II fracture energies
decreases; Berglund and Johannesson [38] and Russell and Street [29] have published results
for graphite/PEEK, indicating similar mode I and mode II values. The results of Berglund
and Johannesson were obtained on specimens promoting unstable intralaminar failure, but
the present work confirms this conclusion. Johnson and Mangalgiri [39] have related the
increased G~c values associated with tougher matrices to an increased ability to dilatate;
shear deformation does not require volume dilatation, so relatively small increases are
recorded for corresponding Giic values.
The extension of this work is to vary the free length of the mixed mode specimen to
obtain different combinations of mode I and mode II loading, and, hence, to plot the full
mixed mode failure envelope for each of the three materials. The final phase is the study
of the influence of temperature and loading rate on this envelope.
Conclusions
The conclusions from this study may be summarized as follows:
9 In mode I, increasing thickness from 5 to 20 mm had no effect on initiation fracture

energies of epoxy-based composites. For graphite/PEEK, initiation energies in thick
specimens with short crack lengths were higher than those in thin specimens due to a
strain rate effect. Propagation energies in glass/epoxy, which included a fiber-bridging
contribution, were dependent on specimen stiffness.
9 The comparison of two mode II tests on thick specimens, CBEN and ENCB, showed
very similar behavior under both static and fatigue loading.
9 Finally, a mixed mode test has been developed allowing mode I and mode II contri-
butions to be studied on specimens of identical thickness to those tested in mode I and
mode II.
Acknowledgments
Financial support for this work was generously provided by Renault, l'Agence Fran~aise
pour la Ma~trise de l'Energie, and la R6gion de Picardie. The authors also gratefully ac-
knowledge the gift of materials from Ciba Geigy, Aerospatiale, and ICI plc. Finally, the
assistance of members of the Division ModUles Num6riques en M6canique at Compi~gne,
and in particular J. M. Roelandt, is greatly appreciated.
References
[1] McKenna, G. B., Mandell, J. E, and McGarry, E J., Society of the Plastics Industry, 29th Annual
Technical Conference, 1974, section 13C.
[2] de Charentenay, E X., Bethmont, M., Benzeggagh, M., and Chretien, J. F., in Mechanical
Behaviour of Materials (ICM3), Vol. 3, Pergamon Press, 1979, p. 241.
[3] de Charentenay, E X. and Benzeggagh, M., in Proceedings, Third International Conference on
Composite Materials, ICCM3, A. R. Bunsell, Ed., Vol. 1, 1980, p. 186.
[4] Benzeggagh, M. L., Prel, Y., and de Charentenay, F. X., in Proceedings ICCM5, The Metallurgical
Society, 1985, pp. 127-139.
[5] Wilkins, D. J., Eisenmann, J. R., Camin, R. A., Margolis, W. S., and Benson, R. A., in Damage
in Composite Materials, ASTM STP 775, K. L. Reifsnider, Ed., American Society for Testing and
[6] Devitt, D. E, Schapery, R. A., and Bradley, W. L., Journal of Composite Materials, Vol. 14,
Oct. 1980, pp. 270-285.
[ 7] Keary, P. E., Ilcewicz, L. B., Shaar, C., and Trostle, J., Journal of Composite Materials, Vol. 19,
March 1985, pp. 154-177.
[8] de Charentenay, E X., Harry, J. M., Prel, Y. J., and Benzeggagh, M. L., in Effects of Defects
in Composite Materials, ASTM STP 836, American Society for Testing and Materials, Philadelphia,
1984, p. 84.
[91 Whitney, J. M., Browning, C. E., and Hoogsteden, W., Journal of Reinforced Plastics and Com-
posites, Vol. 1, 1982, pp. 297-313.
[10] Russell, A. J. and Street, K. N., in Progress in Science and Engineering of Composites (ICCM4),
T. Hayashi, K. Kawata, and S. Umekawa, Eds., Japan Society of Composite Materials, Tokyo,
1982, p. 279.
[11] Giare, G. S., Engineering Fracture Mechanics, Vol. 20, No. 1, 1984, p. 11.
[12] Vanderkley, P. S., "Mode I-Mode II Delamination Fracture Toughness of a Unidirectional Graph-
ite-epoxy Composite," Masters thesis, Texas A&M University, College Station, TX, Dec. 1981.
[13] Barrett, J. D. and Foschi, R. O., Engineering Fracture Mechanics, Vol. 9, 1977, p. 371.
[14] Benzeggagh, M. L., Prel, Y. J., and de Charentenay, E X., in Developments in the Science and
Technology of Composite Materials, Proceedings of First European Conference on Composite
Materials, (ECCM1), Bordeaux, France, September 1985, EACM, Bordeaux, pp. 291-314.
[15] Bathias, C. and Laksimi, A., in Delamination and Debonding of Materials, ASTM STP 876,
237.
[16] Phillips, D. C. and Wells, G. M., Journal of Materials Science Letters, Vol. 1, 1982, pp. 321-
324.
[17] Johnson, W. S. and Mangalgiri, P. D., "Investigation of Fiber Bridging in DCB Specimens,"
NASA Technical memo 87716, 1986.
[18] Russell, A. J., "Micromechanisms of Interlaminar Fracture and Fatigue," Polymer Composites,
October, 1987, pp. 342-351.
[19] Davies, P., Benzeggagh, M. L., and de Charentenay, F. X., "Delamination of Continuous Fibre
Reinforced Thermoplastic Composites," Journ~es Nationales sur les Composites (JNC5), Editions
Pluralis, Paris, Sept. 1985, pp. 17-32 (in French).
[20] Carlile, D. R. and Leach, D. C., in Proceedings, Fifteenth National SAMPE Technical Conference,
Cincinnati, October 1983, SAMPE, pp. 82-93.
[21] Prei, Y., Benzeggagh, M. L., and Roelandt, J. M., "A Finite Element and Experimental Study
of the Mode I and Mode II Delamination Tests," Journ~es Nationales sur les Composites (JNC5),
Editions Pluralis, Paris, September 1985, pp. 49-62 (in French).
[22] Prel, Y., "A Study of Delamination under Mode II Static and Fatigue Loading in a Glass/Epoxy
Unidirectional Composite," Doctoral thesis, Universit6 de Technologic de Compi~gne, France,
June 1987 (in French).
[23] Carlsson, L. A., Gillespie, J. W., and Pipes, R. B., "On the Analysis and Design of the End
Notched Flexure (ENF) Specimen for Mode II Testing," Journal of Composite Materials, Vol. 28,
November 1986, p. 594.
[24] Carlsson, L. A., Gillespie, J. W., and Trethewey, B. R., Journal of Reinforced Plastics and
Composites, Vol. 5, July 1986, pp. 170-187.
[25] Prel, Y., Benzeggagh, M. L., and de Charentenay, E X., "Effects of Defects on Mode II Inter-
laminar Shear Behaviour," Journ~es Nationales sur les Composites (JNC5), Editions Pluralis, Paris,
September 1985, pp. 191-208 (in French).
[26] Bui, H. D., "Associated Path Independent J-integrals for Separating Mixed Modes," Journal of
Mechanics and Physics of Solids, Vol. 31, No. 6, 1983, pp. 439-448.
[27] Benzeggagh, M. L., Davies, P., Gong, X. J., Roelandt, J. M., et al., "A Mixed Mode Specimen
for Interlaminar Fracture Testing," Composite Science and Technology, accepted for publication.
[28] Ashizawa, M., "Improving Damage Tolerance of Laminated Composites through the Use of New
Tough Resins," presented at the Sixth Conference on Fibrous Composites in Structural Design,
New Orleans, January 1983, McDonnell Douglas papers.
Fatigue Fracture Under Mode II Shear Loading of Graphite Fiber Composites," Toughened Com-
posites, ASTM STP 937, N. J. Johnson, Ed., Philadelphia, 1987, p. 275.
[30] Leach, D. C. and Moore, D. R., Composite Science and Technology, Vol. 23, 1985, pp. 131-161.
[31] Curtis, R. T., Davies, P., Partridge, I. K., and Sainty, J. P., in Proceedings, ICCM6-ECCM2,
London, July 1987, Vol. 4, Elsevier Applied Science, London, p. 401.
[32] Gillespie, J. W., Carlsson, L. A., and Pipes, R. B., Composite Science and Technology, Vol. 27,
1986, pp. 177-197.
[33] Davies, P., Benzeggagh, M. L., and de Charentenay, E X., in Proceedings, 32nd SAMPE Sym-
posium, April 1987, Anaheim, CA, SAMPE, pp. 134-146.
[34] Davies, P. and de Charentenay, E X., Proceedings, ICCM6-ECCM2, London, July 1987, Vol. 3,
Elsevier Applied Science, London, pp. 284-294.
[35] Jurf, R. A. and Pipes, R. B., Journal of Composite Materials, Vol. 16, 1982, pp. 386-394.
[36] Russell, A. J. and Street, K. N., in Delamination and Debonding of Materials, ASTM STP 876,
370.
[37] Bradley, W. L. and Jordan, W. M., "The Relationship Between Resin Ductility and Composite
Delamination Fracture Toughness," in Proceedings, International Symposium on Composite Ma-
terials and Structures, Technomic Publishing Co., Beijing, China, June 1986, pp. 445-450.
[38] Berglund, L. and Johannesson, T., in Proceedings, First European Conference on Composite
Materials (ECCM1), Bordeaux, France, September 1985, AEMC, pp. 259-264.
[39] Johnson, W. S. and Mangalgiri, P. D., "Influence of the Resin on Interlaminar Mixed-Mode
Fracture," NASA Technical Memo. 87571, NASA, Washington, DC, July 1985.
Wen S. Chan ~ and Albert S. D. Wang 2
Free-Edge Delamination Characteristics in

S2/CE9000 Glass/Epoxy Laminates Under
Static and Fatigue Loads
REFERENCE: Chan, W. S. and Wang, A. S. D., "Free-Edge Delamination Characteristics

in S2/CE9000 Glass/Epoxy Laminates Under Static and Fatigue Loads," Composite Materials:
ABSTRACT: The static and fatigue characteristics of delamination growth in the $2/CE9000
glass/epoxy laminates under cyclic loading and various environmental conditions were inves-
tigated. From the resulting data, a growth model with a = 3 • 10 ~and 13 = 5 was established.
In the many cases reviewed in this paper, the values of a and 13remain unchanged until there
is a change in temperature, a change in moisture content, or a change in the R ratio.
KEY WORDS: composites, delamination, fatigue, damage tolerance, growth model, glass/
epoxy
It is well known from the literature [1] that both the initiation and growth of free-edge
delamination in laminates are influenced by the lamination geometry, that is, the ply fiber
orientation, ply thickness, ply stacking sequence, and so forth. Hence, by fabricating a group
of laminates in which a given geometrical factor is varied, one can determine the effect of
this factor on the delamination growth characteristics. This fact makes it possible to establish
a rational model with which to separate the geometrical factors from the material factors.
In this regard, Wang and Crossman [2] applied the energy release rate method of classical
fracture mechanics to model free-edge delamination in laminates subjected to quasistatic
loading. The delamination growth criterion is of the form
G ( ~ , a) = Gr (1)
In this equation, the quantity G ( ~ , , a) is the strain energy release rate that exists at the
crack front of a delamination of size a; the delamination is assumed to grow stably under
the applied laminate stress ~,. It is noted that G(~,, a) is computed from a finite-element
analysis that includes not only the variables a and ~x, but also the elasticity properties of
the basic material ply, the ply fiber orientation, ply thickness, ply stacking sequence, and
the location of the considered delamination.
Associate professor, Department of Mechanical Engineering, University of Texas at Arlington,

Arlington, TX 76019.
2 Professor, Department of Mechanical Engineering and Mechanics, Drexel University, Philadelphia,
PA 19104.
270
Downloaded/printed
Copyright9 byby
CHAN AND WANG ON FREE-EDGE DELAMINATION CHARACTERISTICS 271
TABLE 1--List of specimens in static tests.
Specimen
Laminate Number Test Condition
(m35/0/90), A19 Static to failure

A20 Hot (180~ 82~
A21
A31 Static to failure
A32 Wet specimen
A33
A44 Hot (180~ 82~
A45
A47 Cold ( - 67~ - 55~
A48
(35/0/-35/90), B1 Static to failure

B2 Room temp/dry
B3
B4
(0/-+35/90), C1 Static to failure
C2 Room temp/dry
C3
Static Test
The static specimens and the conditions under which they were tested are listed in Table
1. An environmental chamber was used for specimens tested at hot (180~ 82~ wet, hot/
wet, and cold ( - 67~ - 5 5 ~ conditions.
The specimens were loaded at a rate of 0.05 in./min in a Tinius Olsen servohydraulic test
machine and monitored by a 4-in. extensometer on the loading axis. A n X-Y plotter was
attached to record the load-strain curve during loading. A dye penetrant was applied to the
edge of the specimen under load to highlight the delamination and facilitate observation.
Stress at the onset of delamination was determined, and photographs were taken to document
the delamination.
Fatigue Test
The fatigue specimens, test parameters, and test procedure are presented in Table 2. A
tensile load with a sinusoidal waveform was applied in all tests; the oscillation frequency
was kept at about 5 Hz. Damage inspections were made at several scheduled intervals, for
example, after 1, 10, 100, 1000, 10 000 cycles, and so forth. At each inspection interval, an
optically opaque and crack-penetrating chemical (DIB) was applied to the specimen's free
edges and a plane view X-ray taken. The X-ray photograph showed whether or not free-
edge delamination had occurred; if it had, the size of the delamination at three specific
locations on the photograph was measured and the average used as the representative value.
In this manner, a plot of the delamination size as a function of fatigue cycle number N was
obtained. An experimental plot of this type manifests the phenomenological behavior of
the free-edge delamination growth in the specimen being tested.
TABLE 2--List o f specimens in fatigue tests on $2/CE9000 laminates.
Case Specimen Max

Laminate Number Number Stress, ksi R Test Procedure ~
(• 1 A7 20.5 0.1 Fatigue to 106 cycles at 20.5 ksi (141

A8 MPa)
A9
A13 20.5 0.1 First two cycles at 30.2 ksi (208 MPa),
A14 then fatigue to 106 cycles at 20.5
A15 ksi (141 MPa)
2 A1 26.37 0.1 Fatigue to failure
A2
A3
3 Same as 29.3 0.1 Fatigue to failure at 29.3 ksi (202
Case 1 MPa)
(35/0/- 35/90)~ 4 B5 27 0.1 Fatigue to failure
B6
B7
5 B8 30 0.1 Fatigue to failure
B9
B10
6 Bll 30 0.1 First two cycles at 33 ksi (228 MPa),
B12 then fatigue to failure
B13
(0/• 7 C4 31.12 0.1 Fatigue to failure
C5
C6
8 C7 34.58 0.1 Fatigue to failure
C8
C9
9 C10 34.58 0.1 First two cycles at 38 ksi (262 MPa),
C11 then fatigue to failure
C12
(-+35/0/90)~ 10 A25 21.48 0.1 Temperature, 180~ (82~ fatigue
A26 to failure
A27
11 A22 23.87 0.1 Temperature, 180~ (82~ fatigue
A23 to failure
A24
(• 12 A28 21.48 0.1 Temperature, 180~ (82~ First two
A29 cycles at 26.26 ksi (181 MPa), then
A30 fatigue to failure
13 A34 27.25 0.1 Wet specimen. Fatigue to failure
A35
A36
14 A37 30.28 0.1 Wet specimen. Fatigue to failure
A38
A39
15 A40 27.25 0.1 Wet specimen. First two cycles at
A41 33.31 ksi (230 MPa), then fatigue
A42 to failure
16 AI0 29.3 0.5 Fatigue to 1.5 x 106 cycles at 20.5
All ksi (141 MPa), then 29.3 ksi to
AI2 failure
17 A4 29.3 0.5 Fatigue to 106 cycles at 26.37 ksi (208
A5 MPa), then 29.3 ksi (202 MPa) to
A6 failure
18 A16 29.3 0.5 First two cycles at 30.2 ksi (208 MPa),
A17 then fatigue to 750 000 cycles at
A18 20.5 ksi (141 MPa), then 29.3 ksi
(202 MPa) to failure
NOTE: 1 ksi = 6.895 MPa.

o Specimens were tested at room temperature/dry conditions unless otherwise stated.
CHAN AND WANG ON FREE-EDGE DELAMINATtON CHARACTERISTICS 273
In each type of test, three replicate specimens were used. The scatter of the data indicates
the variability encountered in the experiment. Unless otherwise specified, all tests were
performed under ambient room temperature ( - 7 5 ~ and humidity ( - 6 5 % relative) con-
ditions.
Analysis of Test Results

The unidirectional mechanical properties of $2/CE9000 glass/epoxy tape under various
environmental conditions are listed in Table 3. The nominal thickness of the tape was 0.0085
in. (0.216 mm). The interlaminar fracture toughness was calculated on the basis of the
measured far field strain and the ply thickness.
This paper reports on a study whose objectives were to investigate the static and fatigue
characteristics of delamination growth and to establish a database for the development of
a delamination growth model under cyclic loading and various environmental conditions.
The specific goal of the study was to determine, for a selected material system, the parameters
et and 13 in the following growth model:
dN - a (2)
where
"d = a / b = the ratio of the delamination length to half of the coupon width, and
G = the total strain energy release rate corresponding to the maximum stress amplitude.
In addition, the dependence of a and 13 on fatigue amplitude ratio and environmental

factors was explored.
The material studied was $2/CE9000-9 unidirectional glass/epoxy tape, a Ferro Corp.
product. The matrix CE9000-9 is a 350~ curing resin.
Experiment
Laminate specimens were tested to quantify their delamination characteristics under var-
ious loads and environmental conditions. Two major tasks were included in the study. The
first was to record the growth of free-edge delamination as a function of static load and of
TABLE 3--Unidirectional material properties o f $2/CE9000 glass~epoxy.
Room Temp, Hot (180~ Room Temp, Hot (180~ Cold (-67~
Dry Dry Wet Wet Dry
El, Msi 8.09 &09 6.64 6.95 6.84

E2, Msi 2.28 1.76 1.97 1.84 2.87
U12 0.288 0.280 0.29 0.30 0.270
Gt2, Msi 1.06 0.92 0.841 0.56 1.30
NOTE: 1 Msi = 6.895 GPa; ~ = (~ - 32) x s/9.

fatigue load and cyclic time. The second was to correlate the growth behavior of the observed
and predicted delamination.
Specimen Description and Preparation

The following family of laminates suggested by NASA publication 1092 [3] for edge
delamination tension tests were used: (-35/0/90), (35/0/-35/90), and (0/---35/90),. These
layups were chosen to minimize the applied strain required to measure delamination before
final failure. Each specimen was 10 in. (254 mm) long and approximately 1.5 in. (38.1 mm)
wide without necking along the edges. Glass laminate end tabs were used on all specimens.
For wet conditions, the specimens were pretreated before testing. This consisted of placing
the specimens in a moisture chamber with 95% relative humidity until they gained 1%
moisture in weight. The specimens were then sealed in a bag and placed in a dry chamber
for protection until time for testing. Before running a test, the specimens were removed
from the sealed bag and checked by ultrasonic inspection for any imperfections that might
have occurred during preparation.
O'Brien's equation [4], shown below, was used to compute interlaminar fracture toughness
for ply delamination:
1
Gc = ~ [EL^M -- E*]ec:t (3)
$2 / CE9000 GLASS / EPOXY ( +- 35/0/90)5
[ I Delamination strain ~ Ultimate strain I Scatter range

(avg value)
2.8
2.754 I
2.4
2.0
A
v
1.6
r"
.m
CO
Uq
1.2
P
'
0
RT/Dry Hot/Dry Cold/Dry RTANet Hot/Wet
FIG. 1--Static onset of delamination and ultimate strain of (+-35/0/90), $2/CE9000 glass/
epoxy laminates under various environmental conditions.
Laminate ( +_35/0/90)5
1.0- [100 I[]e3~l", ~ 9
[] Hot (180~ / dry
[] Cold (-67~ / dry
9 Room temp / wet
0.8- O Hot (1800F)/ wet
/~ Room temp / dry
A
E 0,6"
O
0.4 84
O
A
0 []
l ~
0.2-
0
0 1'0
Load (kips)
FIG. 2--Static-edge delamination growth versus load.
$2 / CE9000 GLASS / EPOXY ( +- 35/0/90)S

A
e-
_o 0.5
0.48
~ 0.4 0.41
C
~ 0.3 0.32 0.33
-I
~. 0.2
~ o.1
C
"= 0
RT/Dry RT/Wet Hot/Dry Hot/Wet Cold/Dry
C
1 in-lb/in2 = 0.175 kJ/m2

FIG. 3--Effect of environment on interlaminar fracture toughness of ( +_35/ 0 / 90)~ $2/CE9000.
"M
$2 / CE9000 GLASS / EPOXY

60
m
Ultimate strain 0
i Delamination strain I Scatter range 0
(avg value) Z
0
0
( • 35/0/90)s 0
(0/+ 35/90)s "0
o
69
--I
rn
2.8 i I r162...... (35/0#35190)5 ~
2.754
m
2.4 2D
i
r-
I
2.0
*~ 1.6
C
i
*- 1.2
! !
0.8
0.4
0 VA F '' F?
RT/Dry Hot/Wet RT/Dry Hot/Wet RT/Dry Hot/Wet
FIG. 4--Effect of stacking sequence on delamination and ultimate strain of laminates under room-temperature and
hot/wet conditions.
where
ec = strain at the onset of delamination,
t = laminate thickness,
ELAM = laminate modulus,
E* = delaminated laminate modulus, and
Gc = interlaminar fracture toughness.
Static Tests
Figure 1 shows the test results for the onset of delamination and ultimate strain of the
(-+35/0/90), specimens under various environmental conditions. Room temperature data
from Ref 5 are included in the figure for comparison with the other environmental conditions
of this study. The figure shows that specimens tested at the - 6 7 ~ (-55~ con-
dition have the lowest onset-of-delamination strain. The determination of the onset-of-
100
go
Laminate ( +-35/0/90)s Test Data
A A7,13 R = 0.1
80 9 A8,14
9 A9,15 Omax = 20.5ksi
70 Analytical Data
C 13..5
o 60 cB = 3 x 10-s
predicted
r=
m
-o 50
Log10 N
FIG. 5--Delamination growth in laminates tested in Case 1.
delamination size was based either on direct test measurements or on extrapolation of the
data in Fig. 2 to 5% of the specimen width. Figure 2 shows that as the load increases, the
delamination growth in this laminate is slower under room-temperature/dry conditions than
under other environmental conditions.
All the specimens tested show the delamination wandering at the 90/90 and 0/90 interfaces.
Under the same load level, specimens tested at - 6 7 ~ condition exhibited more ex-
tensive matrix cracking in the 90 ~ ply and +-35~ ply than specimens tested at the other
conditions.
The interlaminar fracture toughness was calculated by Eqs 3 and 4. The laminate modulus
ELAMwas calculated by classical lamination theory, and the delaminated laminate modulus
E* was computed as
3E(~_35/% + Eg0
E* - (4)
4
1 O0
go Laminate ( • 35/0/90)s Test Data
A1 R :, 0.1
80 I"1 A2
O A3 Omax = 26.37 ksi
70 Analytical Data
6=5
C a = 3 x 10-S
.o 60 predicted
C
~
"0 50
~ 40
30
O
20
I0 J 0
a
In
D
0 I 2 3 4 5 fi
Log10 N
FIG. 6--Delamination growth in laminatestestedin Case2.
1 O0 O~D
gO Laminate ( +_35/0/90)S Test Data
/% A13,7 R=0.1
Q A14,8
omax = 29.3 ksi
80 O A15,9
70 Analytical Data
B=S
e- ~ = 3xl0-S
o
60 predicted
C
a
m 50
~ 40
L
i
30
20
I0
I I I I , , , I , , I
0 I 2 3 4 5 fi
LOglO N
The results of interlaminar fracture toughness are presented in Fig. 3. The results clearly
indicate that the (-+35/0/90), laminate has a much lower fracture toughness under - 6 7 ~
dry (-55~ conditions than under the other test conditions. This suggests that cold/
dry conditions are the most critical for damage tolerance design.
The effect of stacking sequence on interlaminar fracture of laminates at room tempera-
ture and hot/wet conditions was also investigated. Both onset-of-delamination and ultimate
strain were measured for laminates with three different stacking sequences: (---35/0/90),,
(35/0/-35/90)s, and (0/---35/90),. Figure 4 shows the results at both room-temperature/dry
and hot/wet conditions.
Fatigue Tests
The case numbers given in Table 2 identify a group of specimens and the fatigue tests to
which they were subjected. The results of these tests are discussed below in sequential order
by case number. Note that the load ratio R ( = trmJtrma~) was chosen to be 0.1 for all cases
except 16, 17, and 18, where it was chosen to be 0.5.
Case 1: RT/Dry, R = 0.1, Laminate (--35/0/90),, Specimens A7, 8, 9, 13, 14, 15

The first two cycles of specimens A13, 14, and 15 were at 30.2 ksi (208 MPa), then all
six specimens were subjected to a maximum amplitude fatigue stress of 20.5 ksi (141 MPa),
or 70% of the static onset-of-delamination load, which is about 29.3 ksi (202 MPa). The
results are shown in Fig. 5 as delamination size (percent of half width, or 0.75 in.) versus
log N. Delamination begins at about 2000 cycles, and, after 106 cycles, there is only about
10% delamination. The delamination growth curve shown in Fig. 5 was determined ana-
lytically from Eq 2. The value of G used in the equation was calculated from a finite-element
analysis [2] for a strain level corresponding to the maximum applied load. The value of
Gc = 0.4 in.-lb/in. 2 (0.07 k J / m 2) was obtained from the static test results shown in Fig. 2.
The ct = 3 x 10 -s and 13 = 5 were obtained from curve-fitting of the test results, et and 13
1O0 600
gO Laminate (35/0/-35/90)S Test Data
B5 R = 0.1
80 r'l B6
C) B7 Omax = 2? ksi 0
Analytical Data
70
8=5
c
O el = 3 x 1 0 - 5
6O Gr = 0.52in-lb/in 2
C
, predicted
~ 5o
0
30
20 6
10
Ob
_ _ o . . . . . ,_
_ 1
0 2 3 4 5 5
Loglo N
!00
90
Laminate (35/0/-35/90)S Test Data
/ % B8 R=0.1
80 [] 89
O B10 Omax = 30 ksi
70 Analytical Data
6=5
0
a = 3x10-5
r
60 Gc = 0.52in-lb/in 2
e-
predicted
'I= 50
"6
~u 40
Q.
30
[]
,o
I , I I , ,
~e- v
0 1 2 3 4 5
Loglo N
were empirically determined from the test results. The test data were first plotted in the
form of log (da/dN) versus log (G/Gc) (see Eq 2). Because a linear relationship was ob-
served, a linear curve in the logarithmic scale was then fitted. The details of this technique
can be found in Ref 6.
Case 2: RT/Dry, R = 0.1, Laminate (_+35/0/90),, Specimens A1, 2, 3

The specimens were subjected to a (rma~of 26.37 ksi (182 MPa), or 90% of the static onset-
of-delamination load. The results are shown in Fig. 6. The analytical prediction, shown by
the solid curve, was obtained as described in Case 1.
Case 3: RT/Dry, R -- 0.1, Laminate (• Specimens A7, 8, 9, 13, 14, 15

The specimens tested in Case 1 were reloaded under a (rmax of 29.3 ksi (202 MPa) to
observe the delamination growth in a crack with an initial size equal to about 10% of the
width of specimen. The results are shown in Fig. 7. The analytical prediction, shown by the
solid curve, was obtained as described in Case 1.
Case 4: RT/Dry, R = 0.1, Laminate (35/0/ - 35/90),, Specimens B5, 6, 7

The specimens were subjected to a ~ma, of 27 ksi (186 MPa), or 90% of the static onset-
of-delamination load. The results are shown in Fig. 8. The analytical prediction, shown by
the solid curve, is again based on Eq 2 with the values for c~ and 13 the same as in Case 1.
The Gc value, however, for B type specimens is 0.52 in.-lb/in. 2 (0.091 kJ/m~). This number
was calculated from Eq 3.
Case 5: RT/Dry, R = 0.1, Laminate (35/0/-35/90)s, Specimens B8, 9, 10

The specimens were subjected to a ~m,x of 30 ksi (207 MPa), which is approximately equal
to the static onset-of-delamination load. The results are shown in Fig. 9. The analytical
! O0
Laminate (35/0/-35/90)S Test Data

90
/~ Bll R = 0.1
[] B12 Ornax = 33 ksi
80 C) B13 (1st 2 cyc)
Omax = 30 ksi
70 Analytical Data
B-S
e.
o
a = 3xlO-S
r
60 Gc m 0.52 in-lb/in 2
e.
predicted
m
50
4O
Q.
30
I0
, , , _ 1 i . 1
0 I 2 3 4 5 6
LogloN
FIG. lO--Delamination growth in laminates tested in Case 6.
CHAN AND WANG ON FREE-EDGE DELAMiNATION CHARACTERISTICS 283
I O0
90 Laminate (01 +_35/90)S Test Data
C4 R=0.1
80 13 CS
0 C6 Omax = 31.12ksi
70 Analytical Data
B=5
C Q = 3xl0-S
r
60 Gc = O.68in-lb/in 2
C predicted
_m
50
40
O.
30
8
!0
&
I ---- I I
0 I 2 3 4 5 5
Loglo N
prediction, shown by the solid curve, differs from that in Case 4 only in the magnitude of
the load amplitude.
Case 6: RT/Dry, R = 0.1, Laminate (35/0/-35/90),, Specimens B l l , 12, 13

The specimens were preloaded to 33 ksi (228 MPa) quasistatically, inducing a delamination
of about 4%. They were then reloaded cyclically to a trm,, of 30 ksi. The experimental data
and predicted delamination growth curve are shown in Fig. 10. The predicted growth curve
in this case was computed in the same way as in Case 5, except the delamination growth
starts from 4% rather than from 0%.
Case 7: RT/Dry, R = 0.1, Laminate (0/• Specimens C4, 5, 6

The specimens were subjected to a ~rm~xof 31.12 ksi (215 MPa), or 90% of the static onset-
of-delamination load. The results are shown in Fig. 11, along with the predicted growth
I00 DO
90 Laminate (0/+- 35/90)S Test Data
C7 R = 0.1
BO [] C8
Omax = 34.58 ksi
O C9
70 Analytical Data
13=5
a = 3x10-5
0
60 Gc = 0.68in-lb/in a
C predicted
50
Q.
40
,/
30
20
10
~k
0 I 2 3 4 5 6
LOglO N
curve. The prediction is based on Gc = 0.68 in.-lb/in. 2 (0.12 kJ/m2), where a and 13 remain
as given in Case 1.
Case 8: RT/Dry, R = 0.1, Laminate (0/+35190), Specimens C7, 8, 9

The specimens were subjected to a ~rma,of 34.58 ksi (238 MPa), which is the static onset-
curve. Here again, a Gc of 0.68 in.-lb/in. 2 (0.12 kJ/m 2) is used, and a and 13 remain as given
in Case 1.
Case 9: RT/Dry, R = 0.1, Laminate (0/_35/90),, Specimens C10, 11, 12

The specimens were preloaded to 38 ksi (262 MPa), inducing a delamination of about
4%. They were then reloaded cyclically to a ~rm,, of 34.58 ksi (238 MPa) until final failure.
The experimental data and predicted delamination growth curve are shown in Fig. 13. This
case is similar to Case 8, except that a predelamination was introduced before fatigue loading.
Case 10: 180~ R = 0.1, Laminate (--35/0/90),, Specimens A25, 26, 27

The specimens were subjected to a ~rm~,of 21.48 ksi (148 MPa), or 70% of the static onset-
of-delamination load, under hot (180~ dry conditions. The results are shown in Fig. 14.
Because the material's properties were changed under high temperature, the behavior of
the delamination growth also changed drastically. A prediction of the growth behavior is
shown by the solid curve. The prediction curve is based on a Gc of 0.37 in.-lb/in. 2 (0.065
kJ/m2), which was determined in the static tests; a value for 13 of 4.5 was selected from the
best fit of experimental data. The reduction of 13 from 5 to 4.5 indicates that the material
softened because of the increase in temperature; it also implies that an increase in energy
dissipation occurred at the crack front during each load cycle.
!00 OD~
90 Laminate (0/-+ 3 5 / 9 0 ) s Test Data
C10 R = 0.1
80 [] Cll Ornax = 3 8 k s i
O C12 (1st 2 cyc)
Omax B 3 4 . 5 8 ksi
70
Analytical Data
c 13=5
o
60 a I. 3xl0-S
e-
Gc = 0.68in-lb/in 2
mto predicted
~O 50
eL 4O
3O
2O
tO
I , | I I I , I
0 1 2 3 4 5 5
Log10 N
100
gO
Laminate ( -+ 35/0/90)s Test Data
A25 R = 0.1
80 [] A26 Omax = 21.48 ksi
O A27 Temp = 180~
70
Analytical Data
c
13=4.5
o 60 a = 3x10-5
c
Gc = 0.37in-lb/in ~
predicted
.~ 5o
~ 40
20
I0
I~r '~ q~ ' ' . I

0 I 2 3 4 5 6
LOgl0N
Case U: 180*F/Dry, R = 0.1, Laminate (• Specimens A22, 23, 24

The specimens were subjected to a trm~xof 31.12 ksi (215 MPa), or 90% of the static onset-
growth data are shown in Fig. 15, along with the prediction curve. Of course, the prediction
is again based on a Gc of 0.37 in.-lb/in. 2 (0.065 kJ/m z) and a 13 of 4.5.
Case 12: 180~176 R = 0.1, Laminate (• Specimens A28, 29, 30

The specimens were preloaded to 26.26 ksi (181 MPa), which was just sufficient to initiate
delamination (~2%). They were then fatigue loaded under a crm~xof 21.48 ksi (148 MPa),
as in Case 10. Figure 16 shows the delamination growth data, along with the prediction
curve calculated with a Gc of 0.37 in.-lb/in. 2 and a 13 of 4.5.
CHAN A N D WANG ON FREE-EDGE DELAMINATION C H A R A C T E R I S T I C S 287
Case 13: RT/Wet, R = 0.1, Laminate (+-35/0/90),, Specimens A34, 35, 36

The specimens were previously conditioned in a wet environment (see section entitled
Specimen Description and Preparation), then loaded under a ~rm~xof 27.25 ksi (188 MPa)
in a moisture protective environment. This load level is about 80% of the static onset-of-
delamination load. The delamination growth of these specimens is shown in Fig. 17, along
with a predicted growth curve based on a Gc of 0.45 in.-lb/in. 2 (0.079 kJ/m z) and a 13 of 3.
Here again, the reduction of 13 from 5 in dry specimens to 3 in a wet environment indicates
the change in material property due to moisture absorption.
Case 14: RT/Wet, R = 0.1, Laminate (-+35/0/90),, Specimens A37, 38, 39

As in Case 13, the specimens were conditioned in a wet environment. They were then
fatigue loaded under a (rm~xof 30.28 ksi (209 MPa), which is about 90% of the static onset-
of-delamination load. The experimental data and predicted delamination growth are shown
in Fig. 18. The values for Gc and 13 used here are the same as in Case 13.
I O0 ggl
go
Laminate ( + 35/0/90)s Test Data
80 A22 R = 0,1
1-1 A23 Omax = 23.87 ksi
O A24 Temp = 180~
70
Analytical Data
6O 6=4.5
r a = 3xl0-S
C Gc = 0.37 in-lb/in 2
r predicted
-~ 5 O
~ 4o
L
30
20
10
Ow ,j
0 I 2 5 4 5 6
Loglo N
I00
gO Laminate ( + 35/0/90)s Test Data
/~ A28 R=0.1
[] A29 Ornax = 2 6 . 2 6 k s i
80 O A30 (1 st 2 cyc)
O'max = 2 1 . 4 8 k s i
Temp = 180~
70
Analytical Data
r
6=4.5
.9 60 a = 3x10-5
Gc = 0.37in-lb/in 2
predicted
.~ 50
40
eL
30
20 j
!0
0 1 2 3 4 5
Logl0N
Case 15: RT/Wet, R = 0.1, Laminate (• Specimens A40, 41, 42

The specimens were conditioned in the same wet environment as those in Case 14. They
were first loaded to 33.31 ksi (230 MPa) to start free-edge delamination, then fatigue loaded
under a O'ma x Of 27.25 ksi (188 MPa), as in Case 13. The experimental data and predicted
delamination growth are shown in Fig. 19.
Case 16: R T / D r y , R = 0.5, Laminate (• Specimens A I 0 , 11, 12

The specimens were fatigue loaded under a trm,, of 20.5 ksi (141 MPa) to about 1.5 x
106 cycles; at this point, onset of delamination occurred. The specimens were then fatigue
loaded under a trm, of 29.3 ksi (202 MPa) until failure. The experimental data are shown
in Fig. 20. Based on the data for a tr~a, of 29.3 ksi, a prediction was made using a 13 of 9.
The increase of 13 to 9 for an R of 0.5 from a 13 of 5 for an R of 0.1 indicates the effect of
fatigue amplitude ratio. This effect should theoretically be reflected in the quantity G in Eq
3 rather than in the value of 13, but this question cannot be resolved here with limited data.
Case 17: RT/Dry, R = 0.5, Laminate ( _ 35/0/90),, Specimens A4, 5, 6

The specimens were fatigue loaded under a O'ma x of 26.37 ksi (182 MPa) to 106 cycles,
resulting in a delamination of about 15%. The specimens were then fatigue loaded under
a trm~xof 29.3 ksi until failure. The experimental data and the predicted delamination growth
are shown in Fig. 21. The prediction, as in the previous case, is based on a 13 of 9.
Case 18: RT/Dry, R = 0.5, Laminate (-+35/0/90),, Specimens A16, 17, 18

The specimens were preloaded to 30.2 ksi (208 MPa), inducing a delamination of about
5%. The laminates were then fatigue loaded under a (r~ax of 20.5 ksi (141 MPa) to 750 000
cycles. A t the end of the cycling, the delamination grew to about 10%. The laminates were
I00
90
80 Laminate ( 235/0/90)S Test Data
A34 R = 0,1
!"1 A35 Omax = 27.25ksi
70 O A36 Wet specimens
C Analytical Data
O 6O
B=3
= 3 x 10-S
nl
n 50 Gc = 0.45in-lb/in 2
predicted
$ 40
r
30 O
2o
I0
0 I 2 3 4 5 6
Loglo N
i O0
Laminate (_+ 35/0/90)s Test Data

go
A37 R = 0.1
[] A38 amax = 30.28ksi
O A39 Wet specimens
80
Analytical Data
70
13=3
a = 3xl0-S
C
Gc = 0.45in-lb/in 2
o 60 predicted
C
"~
"U 5O a
CI.
30
20
!0
0 I 2 3 4 5 6
Loglo N
lOOt- .
90
Laminate ( ! 35/0/90)S Test Data
80 /% A40 R = 0.1
[] A41 (;max = 33.31ksi
0 A42 (1st2 cyc)
70 Omax = 27.25 ksi
Wet specimens
60 Analytical Data
B=3
Q = 3x10-5
.~ 5o Gc = 0.45in-lb/in2
predicted
40,
/-
G)
L I
3~t
20 i 0n
I0
0
0 I 2 3 4 5 6
Loglo N
100 CO6
go
Laminate ( +_35/0/90)S Test Data
~al~ A10 R = 0.5

BO 9 [] All amax = 20.5ksi
0 A12 (1.5xl06cyc)
Omax = 29.3 ksi
70
Analytical Data
t-
O 8=9
9= 60 a = 3x10"5
m
e-
Gc = 0.406in-lb/in 2
predicted
9~ 50
aD
Q.
30
0
20
0 A
I0
0
0
0 a
I 2 3 4 5 6
Log10 N
F I G . 20--Delamination growth in laminates tested in Case 16.
i O0 ,
90
Laminate ( + 35/0/90)5 Test Data
,ll~ A4 R = 0.5
80 9 I-I A5 (~max : 26.37 ksi
9 O A6 (lxl06cyc)
Omax = 29.3 ksi
70
Analytical Data
C
O 13:9
'= 60
C
m a : 3 x 10-$
Gc : 0.406in-lb/in a
predicted
9~ so
30
20
10
, , , , I , ,
A,
0 I 2 3 4 5 6
Log10 N
100 D
L a m i n a t e ( _+35/0/90)S Test Data

90
~b,/~ A16 R=0.5
9 [] A17 Oma x = 30.2 ksi
80 eO Ala ( 1st 2 cyc)
Omax = 20.5 ksi
(750,000 cyc)
Omax == 29.3 ksi
70
Analytical Data
e-
o B~.9
60
._~ = 3 x 10-5
E Gc = 0.406in-lb/in 2
- predicted
-~ s o
40
r
3O
20 8
10
A A I
1"
0 ! 2 3 4 5 6
Log lo N
then fatigue loaded under a trmax of 29.3 ksi until failure. The test data are shown in Fig.
22, along with prediction curves based on a 13 of 9.
Summary
Of the environmental conditions under which the S2/CE9000 glass/epoxy laminates were
tested, the cold/dry ( - 67~ - 5 5 ~ condition resulted in the lowest value for interlaminar
fracture toughness. This suggests that cold/dry conditions are the most critical for damage
tolerance design.
A growth model for the $2/CE9000 glass/epoxy laminates has been established. Under
room temperature/dry conditions, the power 13 appears to have a value of about 5 for the
glass/epoxy system used in these tests. For graphite/epoxy systems [7], the value of 13 is 8.
The glass/epoxy system is less stiff than the graphite/epoxy system, so a value of 5 appears
to be in the right direction.
In the many cases reviewed here, the values of a and 13 remain unchanged until there is
a temperature change, a change in moisture content, or a change in the R ratio. A change
in temperature, moisture content, or both affect the properties of the material, but a change
in the R ratio does not. Clearly, the delamination growth law, as used in this paper, needs
further refinement in order to include properly the factor R.
Acknowledgment
This work was done when the first author was associated with Bell Helicopter Textron,
Inc. The work was supported by the Army Aviation Technology Directorate under Contract
No. DAAK-84-C-0002 to Bell Helicopter Textron. Barry Spigel was technical monitor. The
authors would like to thank C. Rogers of Bell Helicopter Textron, Inc., project engineer
of this contract, for his interest in this work. The first author of this paper extends his
appreciation for the fruitful discussions with Barry Spigel and J. Martin of Bell Helicopter
Textron.
References
[1] Wang, A. S. D., "Fracture Mechanics of Sublaminate Cracks in Composite Materials," Composite
Technology Review, Vol. 6, No. 2, 1984, pp. 45-62.
[2] Wang, A. S. D. and Crossman, E W., "Initiation and Growth of Transverse Cracks and Edge
Delamination in Composite Laminates, Part 1: An Energy Method," Journal of Composite Ma-
terials, Vol. 14, No. 1, 1980, pp. 71-87.
[3] Standard Testfor Toughened Composites, NASA Reference Publication 1092, April 1982.
[4] O'Brien, T. K., "Characterization of Delamination Onset and Growth in a Composite Laminate,"
Damage in Composite Materials, ASTM STP 775, K. Reifsnider, Ed., American Society for Testing
[5] Chan, W. S., Rogers, C., Cronkhite, J. D., and Martin, J., "Delamination Control of Composite
Rotor Hubs," Journal of the American Helicopter Society, Vol. 31, No. 3, July 1986, pp. 60-69.
[6] Bucinell, R. B., "A Stochastic Delamination Growth Model for Fatigue Loaded Laminated Com-
posite Materials," Ph.D. dissertation, Drexel University, Philadelphia, September 1987.
[7] Wang, A. S. D., Chou, P. C., Lei, C. S., and Bucinell, R. B., Cumulative Damage Model for
Advanced Composite Materials, AFML-TR-84-4004, 1984.
Shankar Mall, t Ki-Tae Yun, 2 and N a n d K. K o c h h a r 3
Characterization of Matrix Toughness Effect

on Cyclic Delamination Growth in Graphite
Fiber Composites
REFERENCE: Mall, S., Yun, K.-T., and Kochhar, N. K., "Characterization of Matrix Tough-
ness Effect on Cyclic Delamination Growth in Graphite Fiber Composites," Composite Ma-
terials: Fatigueand Fracture, Second Volume, ASTM STP 1012, Paul A. Lagace, Ed., American
ABSTRACT: Three composite systems, having different interlaminar fracture toughnesses,

were tested to characterize the matrix toughness effect on delamination growth behavior in
composites under fatigue loading. With each composite system, three specimen types were
tested: (1) double-cantilever-beam, (2) cracked-lap-shear, and (3) end-notched flexure spec-
imens for mode I, mixed mode I-II, and mode II Ioadings, respectively. The measured de-
lamination growth rate data were correlated with the corresponding strain energy release rates,
GI, Gll-tl), and Gw The cyclic delamination growth resistance of composites, expressed in
terms of static interlaminar fracture toughness, that is, AG/Gc, decreases with increasing
matrix toughness. And this decrease depends on the loading mode. Further, this decrease in
cyclic delamination growth resistance of composites, expressed in terms of normalized inter-
laminar fracture toughness (AG/Gc), increases when fatigue loading mode is changed from
mode I through mixed mode I-II to pure mode II.
KEY WORDS: delamination, delamination growth rate, strain energy release rates, fracture
mechanics, fatigue, composite materials, matrix toughness
Nomenclature
a Length of debond, mm
da Debond growth rate, mm/cycle
dN
b Width of specimen, mm
C Curve-fit parameter for power-law equation
C Compliance, mm/N
Ell Young's modulus of laminate in longitudinal direction, GPa
GI Mode I strain energy release rate, J/m 2
Gtc Critical mode I strain energy release rate, J/m 2
G(H[) Mixed mode I-II strain energy release rate, J/m 2
(~([- II)C Critical mixed mode I-II strain energy release rate, J/m 2
Gn Mode II strain energy release rate, J/m 2
Department of Aeronautics and Astronautics, Air Force Institute of Technology, Wright-Patterson

Air Force Base, OH 45433.
2 Department of Aerospace Engineering and Engineering Mechanics, University of Texas, Austin,
TX 78712.
3 Engineering Mechanics Research Corp., Troy, MI 48099.
296
Downloaded/printed
Copyright9 by by
MALL ET AL. ON MATRIX TOUGHNESS EFFECT ON CYCLIC DELAMINATION 297
aiic Critical mode II strain energy release rate, J/m ~'

Gr Total strain energy release rate (GI + Gtt), J/m 2
h Half depth of specimen, mm
L Half span of specimen, mm
n Curve-fit parameter for power-law equation
P Load, N
Introduction
A considerable amount of current research activity is devoted to the study of failure
mechanisms in fiber-reinforced, laminated composite materials. Delamination is the most
commonly observed failure mode in these composites. It is now widely recognized that the
ability to anticipate and quantify detamination failure using fracture mechanics analysis plays
an important role in the understanding of composite behavior; in the screening, selection,
and development of improved composite materials; and in establishing the damage tolerance
and durability design criteria for composite structures. Most of the work reported in this
area has been concerned with delamination growth under static loading [1]. Less attention
has been given to delamination growth under cyclic loading.
There is an obvious need to investigate the role of cyclic delamination in material de-
velopment, screening, selection, and design. The present study focused on the influence of
the matrix toughness on the delamination growth under cyclic loading by measuring cyclic
delamination growth rate in three composites with different interlaminar fracture toughness.
Delamination in composite laminates is generally constrained to grow between layers because
of the presence of continuous fibers above and below each interface. This physical restraint,
in the presence of external loads, can result in a variety of loading modes at the delamination
tip. These modes may range from pure mode I (opening or peel) through various combi-
nations of mode I and mode II (sliding or shear) to pure mode II loading. Mode III (tearing)
or any combination of mode III may also be present. Thus, the complete characterization
of delamination growth behavior requires its investigation for all these loading modes. In
the present study, cyclic delamination growth behavior was investigated for mode I, mixed
mode I-II, and mode II loadings.
C?
2547 7 0~
L3mm
254 mm .I 0
FIG, 1--Double-cantilever-beam specimen.
203 mm
~ LAP
7 _/ /o
STRAP
254 mm
FIG. 2--Cracked-lap-shear specimen.
Background
For damage tolerance analysis of composite structures, a complete characterization of
delamination growth under fatigue loading is required. Wilkins et al. [21 were the first to
characterize the delamination growth in composite with a brittle epoxy (T300/5208) in mode
I and mixed mode I-II loadings using double-cantilever-beam (DCB) and cracked-lap-shear
(CLS) specimens. Their study showed that correlations between measured delamination
growth rates and the corresponding strain energy release rates resulted in the power-law
relationship of the form
da
d-"N = c(G)" (1)
which is similar to cyclic crack growth law used to characterize fatigue crack in metals. Later
studies [3,4] endorsed the findings of Wilkins [2]. Ramkumar and Whitcomb [3] investigated
the mode I and mixed mode cyclic delamination growth in T300/5208 graphite/epoxy (a
brittle, fiber-reinforced composite system). On the other hand, Bathias and Laksimi [4]
studied the cyclic delamination growth in a fabric-reinforced composite (glass fabric--"Bro-
chief'--in 1452 resin) under mode I and mode II loadings.
Recently, Russell and Street [5] investigated the effect of matrix toughness on the cyclic
delamination growth behavior under mode II loading by testing four different graphite fiber
composite systems having widely different mode I interlaminar fracture toughness. This
study showed that the benefits of improved matrix toughness on the composite properties
/ ~ STARTER CRACK
12 ! -,= " .
FIG. 3--End-notched .flexure specimen.
50 I I I I
AS4/APC-2
45 oO0 o ~ 1 7 6
40
/~176 oo
mm o
o
~P
3o oo
I I I
25~ 400 800 12OO 16OO 2000
N, cycles
FIG. 4-- Typical variation of delamination length with fatigue cycles in ENF specimen (AS4~
APC-2).
were reduced or eliminated entirely under mode II fatigue conditions. However, no such
study is available for the mode I and mixed mode I-II fatigue loadings. The objective of
the present study was, therefore, to investigate the effect of matrix toughness on the de-
lamination growth behavior under mode I and mixed mode I-II fatigue loadings. Also, mode
II fatigue loading was included in the present study for the sake of completeness. Three
graphite fiber composite systems with different mode I interlaminar fracture toughness were
tested. For each composite system, three types of specimens, DCB, CLS, and end-notched
flexure (ENF), were employed. The results of the present study, as well as previous studies
[2-5], are then discussed in terms of damage tolerance and long-term durability design
considerations in composite structures, and in composite material development, screening,
and selection processes.
Specimen Configurationsand Preparation

The following three composite systems were investigated (in order of increasing tough-
ness):
1. A brittle composite system--T300/3100 ( g r a p h i t e / b i s m a l e i m i d e - - A m e r i c a n Cy-

anamid).
TABLE 1--Composite material properties.~
Modulus~, GPa Poisson's Ratio b
Composite EH E22 Gi2 v 12 v23
T300/3100 131.0 10.8 5.8 0.30 0.50

IM6/R6376 148.2 10.3 5.0 0.35 0.50
AS4/APC-2 127.6 10.5 5.3 0.30 0.50
a E3 3 = E=; vt3 = v~2;G13 = a l 2 .

b The subscripts 1, 2, and 3 correspond to the longitudinal, transverse, and thickness directions,
respectively, of an unidirectional ply.
IO00 I I I I I
T300/31OO /
800 - 0.4
600 - 0.3
GT ,
dim2 ooo oz G I / G r r
200~ - 0.1
O'
-0 Joo 200
I
300 400
I I
500
"l 0
60o
STRESS, MPa
FIG. 5--Variation of strain energy release rates with applied load for CLS specimen with
25.4-mm crack length (T300/3100).
2. An intermediate tough composite system--IM6/R6376 (graphite/toughened epoxy--

Ciba-Geigy).
3. A very tough composite system--AS4/APC-2 (graphite/thermoplastic polythereth-
erketone (PEEK--ICI).
Unidirectional laminates of 24 plies (nominal thickness of 3 mm) were prepared from

prepeg according to its manufacturer's recommended laminating procedures. A 0.025-mm-
thick Teflon film was inserted along one edge of the panel prior to processing to provide
the midplane delamination starter crack. DCB, CLS, and ENF specimens were machined
from these laminates (Figs. 1-3). The metallic hinges were attached to both sides of the
DCB specimen at the end adjacent to the starter notch to allow for unconstrained rotation
at the end during load introduction. The CLS specimens were obtained by machining away
the half side of the panel up to the midplane delamination to form the extended portion of
specimen, that is, strap. The lengths of strap and lap adherends were 254 and 203 ram,
respectively (a total of 76 mm was for grip support on both ends).
Experiments
The purpose of this experimental program, which included static and fatigue tests of DCB,
CLS, and ENF specimens, was to measure the interlaminar fracture energy under static
TABLE 2--Flexural modulus.

Composite E,,, GPa
T300/3100 125.5
IM6/R6376 134.5
AS4/APC-2 119.3
0.008 I I I I
o EXPERIMENTAL AS4/APC-2
O.OOE
FEM- ~ -
C~ 0.004 -
mm/N LINEAR BEAM

" THEORY
0'002f
% I
02
I
Q4
I
Q6
I
08 1.0
o/L
FIG. 6--Relations between compliance, C, and normalized crack length, a/L, of ENF
specimen (AS4/APC-2).
loading and the delamination growth rate under cyclic loading. Similar experiments are
described in detail for DCB and CLS specimens in Refs 6 and 7.
E N F Specimen
Prior to static testing the delamination was advanced from the end of the insert to create
a natural crack front by fatigue loading in mode II condition. Then this specimen was loaded
in three-point bending using a custom-made fixture. The static test was run in a displacement-
controlled mode. The center point displacement was measured by a direct current differential
transducer. The delamination growth occurred in the unstable manner which resulted in
sudden drop in the measured load versus displacement record. The critical strain energy
4OO
oo_As4,
Pc-2
FEM~ / / /
J/rn2
G~, 2oo //...._LN
,TEA#oRB
AM
yE
Ioc //
O0 ~'~ I I I I
0.2 0.4 0.6 0.8 1.0
cl/L
FIG. 7--Variation of strain energy release rate with normalized crack length, a/L, for a
constant displacement of 1.80 mm (AS4/ A PC-2).
302 S E C O N D COMPOSITE M A T E R I A L S
TABLE 3--Interlaminar fracture toughness of three composites."
Composite Glc, J/m 2 G(HI)c, J/m2 G.c, J/m 2
T300/3100 170 452 548

IM6/R6376 473 599 650
AS4/APC-2 1205 1397 1502
Average value of at least three specimens.
release rate, G.c, was then computed with this measured critical load from the finite-element
analysis as discussed in the following section.
The fatigue tests of ENF specimens were conducted under constant amplitude cyclic
displacement at a cyclic frequency of 2 Hz with the ratio of minimum to maximum load (or
displacement) in a fatigue cycle equal to 0.1. Fatigue cycles, applied loads, and displacements
were monitored continuously throughout each test. Because delamination length was almost
impossible to monitor during the fatigue test, it had to be determined indirectly by a rela-
tionship between measured compliance and delamination length established under static
loading from the same specimen. After computing the delamination length for each periodic
compliance measurement, the delamination growth rates (da/dN) were obtained by curve
fitting and differentiating the crack length, a, versus fatigue cycle, N, data. A typical relation
between the crack length and fatigue cycles is shown in Fig. 4. It can be seen that the
delamination growth rate initially increased and then decreased. This behavior was expected
because strain energy release rate, G,, varied in a similar manner as discussed in the following
section. Several tests were conducted at different constant amplitude cyclic displacement
using several ENF specimens.
Finite Element Analysis

Finite element method (FEM) was employed to compute strain energy release rates for
CLS and ENF specimens. Strain energy release rate, GI, for the DCB specimen was
2000
1600 9 AS4/APC-2
9 IM6/R6376
9 T300/3100
1200
GI ,
J/m e
800
400
O I t
0 400 800 1200 1600 2000
Grr, J / m z
FIG. 8--Interaction between Gt and G..
i0-~ I
T30013100
i0-~
do
dN I0-'
mm/cycle o DCB
a CLS
10-5 a ENF
lO-e ... I I
50 I00 500 I000
AG, Jim z
FIG. 9--Relations between strain energy release rate and delamination growth rate ( T300/
3100).
computed from the linear beam theory [7]. (It was shown in Ref 7 that G[ obtained from
FEM and linear beam theory are in agreement and hence there was no need for the FEM
analysis of the DCB specimen.)
C L S Specimen
The CLS specimens were analyzed with a finite-element program, called G A M N A S [8],
to compute strain energy release rates. This two-dimensional analysis accounted for the
geometric nonlinearity associated with the large rotation in the unsymmetric CLS specimen.
The importance of this nonlinear analysis is discussed in Ref 6. This FEM analysis followed
the same procedure described in Ref 6 in which similar composite bonded CLS specimens
were used. A typical finite-element model consisted of about 1000 isoparametric four-node
10-2 I I
IM6/R6376
10-3
o DCB
do
w
10-4
dN a CLS
mrn/cycle o ENF
I0"
I I
~~ I00 5oo ~ooo
AG, J / m 2
FIG. lO--Relations between strain energy release rate and delamination growth rate (IM6/
R6376).
IO-Z i I
AS4/APC-2
io-3
dQ
- - I
io-4
dN o DCB
mm/cycle CLS
10-5 o ENF
I I I
J0"50 IO0 500 ~000
~G, d/m 2
FIG. 11--Relations between strain energy release rate and delamination growth rate (AS4~
APC-2).
elements and had about 2000 degrees of freedom. The element size near the crack tip was
0.025 • 0.025 mm. Plane-strain conditions were assumed in the F E M analyses. The material
properties of all composite systems are listed in Table 1. The strain energy release rates,
Gr, Gt, and GH were computed using the virtual crack closure technique [9]. Figure 5 shows
the typical variation of computed strain energy release rates GT, and G I / G . as a function
of nominal applied stress. Further, these strain energy release rates were constant for a
significant delamination region, that is, up to 125 mm. The delamination growth rate data
were measured over this region.
ENF Specimen
A detailed finite-element analysis of the E N F specimen was conducted by Mall and
Kochhar [10]. A similar procedure was employed to analyze all tested ENF specimens in
the present study. One difficulty encountered in testing composite specimens, loaded in
bending, is the evaluation of flexural modulus, EH, which may depend on span-to-depth
ratio [11]. In the present study, a combined numerical-experimental approach was used to
evaluate Elt. The measured compliance of an E N F specimen with no delamination (that is,
a = 0) was matched with its counterpart from F E M analysis where E~ was varied. This
provided the true estimate of E u for the tested specimen. These E , values are presented
in Table 2.
TABLE 4--The constants in Eq 5 for three composite systems.
Mode I Mixed-Mode I-II Mode II

Composite c n c n c n
T300/3100 3.68E-19 7.03 1.87E-16 5.53 3.29E-18 5.80

IM6/R6376 1.64E-19 6.40 7.12E-17 5.46 6.12E-15 4.62
AS4/APC-2 1.87E-17 4.80 1.90E-13 3.77 9.11E-13 3.66
MALL ET AL. ON MATRIX T O U G H N E S S EFFECT ON CYCLIC DELAMINATION 305
TABLE 5---Cyclic delamination growth rate exponent, n.
Mixed Mode
Composite Gtc, J/m 2 Mode I I-II Mode II Reference
T300/5208 103 8.02 and 10.08 6.07 3

AS 1/3501-6 110 9.4 5.79 5,14
AS4/2220-3 160 5.71 5
T300/3100 170 7.03 5.53 5.80 Present study
IM6/R6376 473 6.40 5.46 4.62 Present study
C6000/F155 495 4.52 5
AS4/APC-2 1200 4.8 3.77 3.66 Present study
AS4/APC-2 1330 3.0 3.88 5,14
Figure 6 shows a typical comparison of the measured compliance with FEM and the
theoretical compliance obtained from the simple linear beam theory [12], which is given as
2 L 3 + 3a 3
C - (2)
8E.bh 3
This clearly shows that the FEM and linear beam results are in good agreement with the
experimental values. Figure 7 shows a typical comparison of G . obtained from FEM analysis
I I I i I I i 1 I /
lO-i-
1
(ci) ASI/3501-6
(b) T500/5208
(c) T300/3100
(d) IM6/R6376
lO-Z-- (e) AS4/APC-2 J - !
da I~
(f) AS4/APC-2
,JA
dN ,//,','/I
mm/cycle io_4
10-e
(f) e
Cd/) II I I
i0 -~ _
(b) (a) "-1
i n i i I I 1 I I I
o.ol 0.02 004 0.06 0.08 03 0.2 0.4 0.6 0.8 1.0
AG I/G m
FIG. 12--Comparison o f delamination growth rate o f different composites for mode 1.'Solid
line = present; dashed line = previous studies [3,14].
and from a linear beam theory as given in the following [12]
9P2a 2
G,, - (3)
16b'-E.h 3
where P is the applied load, EH is the flexural modulus, and a, b, and h are shown in
Fig. 3.
Resulls and Discussions

Static Tests
The results of static tests from all three types of specimens are presented in Table 3 and
also shown in Fig. 8 to show the interaction of the mixed-mode loading and matrix toughness
on the interlaminar fracture toughness of the composites. Recently, Johnson and Mangalgiri
[13] investigated toughness behavior of seven matrix and adhesives systems to assess the
influence of the resin toughness on interlaminar fracture under mode I and mixed mode
(various combinations of mode I and mode II). This study showed that, in general, the
higher the G[c value, the closer G~c is to GHc. The brittle materials are much more sensitive
to the G~ components than are the tougher materials. The tougher materials are almost
equally sensitive to Gt and G,. In other words, fracture of brittle resins is controlled by G~
I i I i I i I t I
I0"1-- (o) T300/5208

(b) T300/3100
(C) IM6/R6376
(d) AS4/APC-2
10-2 _
do p
- / / -
dN
mm/cycle
,o-0- /
I0 -~ - (dl (o) -
I I I I I I I I I
0.01 0.02 0.04 0.06 0.08 0.1 0.2 0.4 0.6 0.8 1.0
'AGII-]I)IGlz-II)c
FIG. 13--Comparison of delamination growth rate of different composites for mixed mode
l-H: Solid line = present study; dashed line = previous study [3].
MALL ET AL. ON MATRIX T O U G H N E S S EFFECT ON CYCLIC DELAMINATION 307
I I I I I I i I I
i0-1 - (o) T500/3100

(b) ASI/3501-6
(c) IM6/R6576
(d) AS4/2220-3
(e) C6000/FI55
10-2 -
(f) AS4/APC-2 t"/
dN
do 7
10-3
(g) AS4/APC-2
.•////zT//
/,'/,:///
/
mm/cycle io-4L /,/,',7 / /

/"'//," /
10-5
/
I0 e (ol
I I i I I I I I I
0.0:9 0.04 0.06 0.08 Ol 0.2. 0.4 0.6 0.8' LO
AG~/GTrc
FIG. 14--Comparison of delamination growth rate of different composites for mode II:
Solid line = present study; dashed line = previous study [5].
components, and the fracture of tough resins is controlled by the total strain energy release
rate, Gr. The results of the present study are in agreement with this previous study [13].
For the brittle composite system, T300/3100, G.~c is more than twice Gw, and for the
intermediate tough and for very tough composites (that is, IM6/R6376 and AS4/APC-2)
Gtt c is approximately 0.75 and 0.8 of Gw, respectively. Further, mixed mode fracture data
from the CLS specimen fall almost on the straight line joining Gz( and G.c data. Thus, a
mixed mode delamination failure criterion under static loading can be expressed in the
following form as suggested in Ref 13
(~1 + ~GII = 1 (4)
Cyclic Delamination Behavior

The measured delamination growth rate data were correlated with the corresponding
strain energy release rate range for all three composite systems, as shown in Figs. 9 through
11. They obeyed a relationship of the form
da
--~ = c(aG) ~ (5)
where AG ( = Gmax -- Gram) is strain energy release rate range. The solid line in Figs. 9
through 11 showing this relation was fitted to the data by using a least square regression
analysis. The values of c and n are presented in Table 4. Table 5 presents the values of
cyclic delamination growth rate exponent, n, for various composites from the present study
as well as from previous studies [3,5,14]. These composites are arranged in this table in the
increasing order of their interlaminar fracture toughness, G~c. This comparison clearly
shows that the exponent n in the power-law relationship, Eq 5, in general, decreases with
increasing matrix toughness for all three loading modes. Further, this exponent, n, decreases,
in general, when the fatigue loading mode is changed from pure mode I through mixed
mode I-II to pure mode II loading in each composite system with one exception of T300/
3100.
The principal effect of increasing the matrix toughness was, therefore, a reduction in the
slope n which physically means the magnitude of the delamination growth rate for a given
AG. The values of n obtained in the present study as well as in previous studies range from
3 to 10. The values of n for the graphite fiber composites are quite high compared to typical
values of n for fatigue crack growth in aluminum and steel alloys that range from 1.5 to 2.5.
The high value of n means that a small change in applied load would cause a large change
in delamination growth rate in composites. Thus, the cyclic delamination growth in com-
posites is more sensitive to errors in design loads than are typical cracks in metallic structures.
And this sensitivity is greater in brittle composite than in the ductile composite. Further, it
may be difficult to design composite material structures, especially those with a brittle matrix,
for finite life against delamination failure. Minor design alterations or small analysis errors
could cause a much shorter life than the design value. A viable alternative would involve
an infinite-life approach. For this purpose, the no-delamination-growth threshold, Gth, may
be an important material property for composites.
No attempt to evaluate the threshold value of delamination growth, G,h was done in the
present study. However, a reasonable estimate about the role of matrix toughness on this
no-delamination-growth threshold as well as on delamination growth resistance can be made
from the measured delamination growth rate data. For this purpose, the measured delam-
ination growth rates were plotted as the function of normalized strain energy release rate
ranges, AGJG~c, AG,_n~/G,_Hc), AGn/G.c in Figs. 12 through 14. Similar relations obtained
from previous studies also are shown in these figures. These figures clearly show that the
normalized delamination growth resistance and threshold value under fatigue loading de-
DECREASI
N]1G
SHEARMODE
dQ
dN
ITOU/
NCREASI
~ NG
GHNESS
zxGIG c
FIG. 15--Schematic representation of effect of toughness and loading mode on cyclic de-
lamination growth.
crease with increasing toughness. This decrease is dependent on loading mode. Pure mode
II resulted in comparatively more decrease than mixed mode I-II and mode I. Figures 12
through 14 show that tougher matrix composites will experience more fatigue degradation
than the brittle matrix composites, and this degradation increases when the fatigue loading
mode is changed from pure mode I through mixed mode I-II to pure mode II. This is shown
schematically in Fig. 15. These results show an interesting feature that the increase in
interlaminar fracture toughness of composites by improving matrix toughness does not
translate directly in improved delamination resistance under fatigue loading. These findings
should be considered in screening, selecting, and developing improved composite materials,
and in establishing the damage tolerance and durability design criteria for composite struc-
tures.
Concluding Remarks
An investigation of laminated composites was undertaken to characterize delamination
growth mechanisms under mode I, mixed mode I-II, and mode II fatigue loadings. Three
composite systems, in order of increasing interlaminar fracture toughness, were tested:
graphite/bismaleimide (T300/3100), graphite/epoxy (IM6/R6376), and graphite/thermo-
plastic polyetheretherketone (AS4/APC-2). With each composite system, three specimen
types were tested: (1) a double-cantilever-beam specimen for mode I loading, (2) a cracked-
lap-shear specimen for mixed mode I-II loading, and (3) an end-notched flexure specimen
for mode II loading. The results of this study and previous studies [3,5,14] led to the following
conclusion:
During fatigue loading the normalized delamination growth resistance of composites

expressed in terms of static interlaminar fracture toughness, that is, AG/Gc, decreases
with increasing matrix toughness. And this decrease depends on the loading mode.
Further, this decrease in normalized delamination growth resistance (AG/Gc) increases
when fatigue loading mode is changed from Mode I through Mixed Mode I-II to pure
Mode II.
References
[1] O'Brien, T. K., "Interlaminar Fracture of Composites," NASA TM 85768, June 1984.
[2] Wilkins, D. J., Eisenmann, J. R., Camin, R. A., Margolis, W. S., and Benson, R. A., "Char-
acterizing Delamination Growth in Graphite-Epoxy," in Damage in Composite Materials, ASTM
STP 775, K. Reifsnider, Ed., American Society for Testing and Materials, Philadelphia, 1982, pp.
168-183.
[3] Ramkumar, R. L. and Whitcomb, J. D., "Characterization of Mode I and Mixed-Mode Delam-
ination Growth in T300/5208 Graphite Epoxy," in Delamination and Debonding of Materials,
ASTM STP 876, W. S. Johnson, Ed., American Society for Testing; and Materials, Philadelphia,
1985, pp. 315-335.
[4] Bathias, C. and Laksimi, A., "Delamination Threshold and Loading Effect in Fiber Glass Epoxy
Composite," in Delamination and Debonding of Materials, ASTM STP 876, W. S., Johnson, Ed.,
Fatigue Fracture Under Mode II Shear Loading of Graphite Fiber Composites," in Toughened
Composites, ASTM STP 937, N. J. Johnston, Ed., American Society for Testing and Materials,
[6] Mall, S., Johnson, W. S., and Everett, R. A., Jr., "Cyclic Debonding of Adhesively Bonded
Composites," in Adhesive Joints: Their Formation, Characteristics, and Testing, K. L. Mittal, Ed,
Plenum Press, New York, 1984, pp. 639-658.
[7] Mall, S. and Johnson, W. S., "Characterization of Mode I and Mixed-Mode Failure of Adhesive
Bonds Between Composite Adherends," in Composite Materials: Testing and Design (Seventh
Conference), ASTM STP 893, J. M. Whitney, Ed, American Society for Testing and Materials,
[8] Whitcomb, J. D. and Dattaguru, B., "User's Manual for GAMNAS--Geometric and Material
Nonlinear Analysis of Structures," NASA TM 85734, Jan. 1984.
[9] Rybicki, E. F. and Kanninen, M. E, "A Finite Element Calculation of Stress Intensity Factors
by a Modified Crack Closure Integral," Engineering Fracture Mechanics, Vol. 9, No. 4, 1977, pp.
931-938.
[10] Mall, S. and Kochhar, N. K., "Finite-ElementAnalysis of End-Notch Flexure Specimens," Journal
of Composites Technology and Research, Vol. 8, No. 2, Summer 1986, pp. 54-57.
[11] Zweben, C., Smith, W. S., and Wardle, M. W., "Test Methods for Fiber Tensile Strength, Com-
posite Flexural Modulus and Properties of Fabric-Reinforced Laminates," in Composite Materials:
Testing and Design (Fifth Conference), ASTM STP 674, S. Tsai, Ed., American Society for Testing
[12] Russell, A. J., "Factors Affecting the Interlaminar Fracture Energy of Graphite/Epoxy Lami-
nates," in Progress in Science and Engineering of Composites, T. Hayashi, K. Kawata, and S.
Umekawa, Eds., Proceedings of ICCM-IV, Tokyo, 1982, pp. 279-286.
[13] Johnson, W. S. and Mangalgiri, P. D., "Influence of the Resin on Interlaminar Mixed-Mode
Fracture," in Toughened Composites, ASTM STP 937, N. J. Johnston, Ed., American Society for
[14] Russell, A. J. and Street, K. N., "A Constant AG Test for Measuring Mode I InterlaminarFatigue
Crack Growth Rates," presented at Eighth ASTM Symposium on Testing and Design of Composite
Materials, Charleston, SC, 1986.
Structural Aspects
Ralph F. Foral I and D o n a l d R. Gilbreath 2
Delamination Failure Modes in

Filament-Wound Composite Tubes
REFERENCE: Foral, R. E and Gilbreath, D. R., "Delamination Failure Modes in Filament-

Wound Composite Tubes," Composite Materials: Fatigueand Fracture, Second Volume, ASTM
1989, pp. 313-325.
ABSTRACT: Filament-wound composite tubes of helical hoop construction were tested in

axial tension at room temperature and at elevated temperatures. Stress analysis of such tubes
shows the presence of tensile interlaminar stresses, tending to produce delamination. The test
results show that delamination failures do occur, and that material performance depends on
laminate stacking sequence as well as test temperature. In specimens with no inside hoop
layers, the tensioned inside helical fibers pulled away from the laminate in a localized delam-
ination failure mode, called transverse fiber pullout. The predicted interlaminar stresses at
which this occurred are considerably lower than the expected through-the-thickness strength.
The interlaminar stresses producing these failures must be accounted for in design of curved
laminates, because they can produce premature failure at a load level substantially below that
predicted by in-plane analysis.
KEY WORDS: delamination, failure, filament-wound composites, stacking sequence effects,
transverse fiber pullout, elevated temperature testing
Composite materials, made of continuous fibers and a resin matrix, are being used in
structural applications in a variety of component shapes. When applications require a flat
composite laminate, the analytical techniques for predicting its strength are generally well
established. When the laminate is curved, however, strength prediction techniques are less
well understood. In curved laminates, in-plane tensile stresses can produce out-of-plane
stresses which tend to delaminate the material. This effect has been pointed out by Chang
and Springer [1], who analyzed composite bends under combined moment and shear. The
resulting interlaminar stresses can produce premature failure at a load level substantially
below that predicted by in-plane analysis and must be accounted for in design.
This paper reports a combined experimental and analytical investigation of delamination
failure modes in curved composite laminates. The specimens were thin-wailed composite
tubes loaded to failure in axial tension at room temperature and at elevated temperatures.
Laminate stresses were calculated using a three-dimensional generalized plane strain elas-
ticity solution. Representative test data, demonstrating features of the delamination failure
modes, are presented along with calculated stresses. Results demonstrate the influence of
laminate stacking sequence, as well as elevated temperature, on laminate performance. In
laminates with no inside hoop layers, the tensioned inside helical fibers tear away from the
Professor, Department of Engineering Mechanics, University of Nebraska-Lincoln, Lincoln, NE

68588-0347.
2 Research and development engineer, Brunswick Corporation, Defense Division, Lincoln, NE 68504;
formerly, graduate student, Department of Engineering Mechanics, University of Nebraska-Lincoln.
313
Downloaded/printed
Copyright9 byby
laminate in a localized transverse fiber pullout failure mode. This occurs at predicted inter-
laminar stresses much lower than usually listed for through-the-thickness tensile strength [2].
Test Specimens and Procedures

The specimens were laminated thin-walled tubes of filament-wound helical hoop con-
struction. In the laminate, each helical layer was of two-ply, balanced construction with
adjacent plies of + ~b, then - ~b, fiber angle orientation to the cylinder axis. Fiber orientation
in the hoop plies was very nearly 90 ~ to the cylinder axis. Details of specimen construction
are given in the individual test descriptions below. All tubes were 70.6 mm inside diameter
and 350 mm long. The ends were reinforced with additional hoop windings and machined
to interface the end grips. The end grips, shown in Fig. 1, introduce axial load to the tube
wall through six knurled and tapered segments. These are wedged against the inside surface
of the test specimen by a tapered central spindle. External radial support is provided by a
backup collar and insert. The end of the spindle is threaded to allow connection through
spherical seats to the testing machine. Performance of the end grips has been completely
satisfactory, transferring concentric axial loads up to 340 kN with no indication of slippage
or grip-induced failure.
Testing was performed in a 1000-kN Southwark Emery universal testing machine. Axial
strains were measured with special strain transducers [3]; loads and strains were recorded
on an IBM PC by a Measurements Group 2100 System strain gauge conditioner/amplifier
and a Metrabyte analog-to-digital ( A / D ) converter. Elevated temperature tests were con-
ducted with an Applied Test Systems 1800 W electrical resistance clam-shell furnace heating
the specimen. Sheathed thermocouple probes were used to measure specimen temperature,
using techniques described in Ref 4.
Dimensions of the tubular specimens were chosen so that the central test section would
take on a uniform axial strain state, free from bending-extension coupling and end effects
due to the grips. Previous work cited in Ref 3 indicates that this can be accomplished with
a radius-to-thickness ratio greater than 10 and a total tube length greater than the desired
gauge length plus four times the radius. Our finite-element analyses showed that the present
specimens were of proper proportions, with a central uniform axial strain section at least
130 mm long. It is important to note that laminate symmetry, essential for most flat laminate
testing, is not a requirement for uniform strain in a tubular specimen. In the tubes, bending-
extension coupling, including that induced by lack of laminate symmetry, occurs near the
end grips and rapidly diminishes in magnitude within the grip-influenced region. The central
test section, sufficiently remote from the ends, takes on uniform axial strain.
-'~~~. TubulaSpecl
r men
Axla~ "~
FIG. 1--Schematic--end grip configuration.
FORAL AND GILBREATH ON DELAMINATION IN FILAMENT-WOUND COMPOSITES 315
h r
R-x
FIG. 2--Schematic--tubular specimen under axial tension.
Analysis
Consider a thin-walled composite tube of laminated helical hoop construction subjected
to a centric axial load P (Fig. 2). Insight of how the interlaminar stresses arise in such a
tube can be provided by a simple fiber analysis. In the laminate, the helical plies, with
unidirectional fibers aligned at an acute angle ---+ to the cylinder axis, carry most of the
axial load. The hoop plies, with unidirectional fibers making an angle of nearly 90 ~ to the
cylinder axis, mostly serve to react tendencies of the helical fibers to straighten, and thus
to move inward. The helical fibers follow a helical path with radius of curvature R/sin 2 +,
where R is the tube radius. For a thickness t, of helical fibers with unit width to remain in
equilibrium, an outward-directed resultant radial stress must act; the magnitude of this stress
is t, 9 sin" + / R times the stress in the fiber. Neglecting the load-carrying ability of the resin,
the helical fiber stress is P / ( 2 9 ~r 9 R " T , 9 cos 2 +), where T, is the total helical fiber
thickness; the corresponding radial stress at a thickness t, is P 9 t, 9 tan 2 +/(2 9 ~r 9 R-' 9 T,).
This is an interlaminar normal stress tending to produce delamination or to buckle stabilizing
hoop layers. As shown in the above simplified analysis, the magnitude of the interlaminar
stress depends not only on the axial load and location through the laminate, but also on the
helical wind angle and the tube radius.
The constitutive behavior of each ply of the laminate referred to the z, h, r coordinate
system (Fig. 2) is given by Refs 5,6
c,, o o
"Y~, (1)
L;:j Cl6 C26 C~6 c ~ j k ~ , z h - ctzhAT
relating the normal and shear stresses (tr's and r's) to the normal and shear strains (~'s and
3,'s). The C,~ terms define the elements of the stiffness matrix in terms of the unidirectional
ply properties and the off-axis angle + [6]; the a terms are the linear coefficients of thermal
expansion, assumed constant over the temperature change AT.
An existing generalized plane strain elasticity solution [7] was used to predict the stresses
and strains in the tubes at a point away from the ends. In the analysis, each layer is assumed
to be homogeneous and orthotropic with respect to the cylinder axes z, h, r. Under this
assumption, two adjacent helical plies, with fibers oriented at + + and - + to the cylinder
axis, are treated as a single homogeneous, orthotropic material. The shear stresses T:h in
each ply are assumed to be equal and opposite, producing negligible shear strain ~/a, for the
two-ply layer. These assumptions closely describe overall behavior of the two-ply helical
layer, especially if the helical plies are thin.
In the assumed orthotropic layers, Eq 1 simplifies to
{o,} [cz Czrl,-,- Al (2)

r Cz, Ch, Cr,J[~., a, ATJ
where the elements of the stiffness matrix are given in terms of the moduli of elasticity E
and Poisson's ratios v as [6]
C.,z = E=(1 - VhrVrh)/A Czh = (Vh: + V,zVhr)Ez:/A

C~h = Whh(J -- .rz~.tzr)/A Czr = ( v , + v , : v ~ ) E z z / A (3)
C. = E.(I - v..hvhz)/A Chr = (v,h + v..hvr:)Ehh/A
A = 1 - PzhVhz -- VhrPrh -- VrzPzr -- 2PhzPrhPzr
At a generic point within a layer, the average radial and tangential stresses must satisfy the
equilibrium equation
dffr ~ r -- ffh
d---r + - - r - 0 (4)
and the corresponding strains are given by
du u
~' = ~rr ~h = -r (5)
in terms of the radial displacement u. With the assumed orthotropic behavior, tangential
displacements are zero through the laminate.
In Ref 7, Eqs 2 through 5 are combined and solved for the radial displacement u. Solutions
for each layer (k = 1,N) are combined by imposing conditions of continuity between layers
and applying the boundary conditions. At the interface of the outer surface of the kth layer
and the inner surface of the (k + 1)th layer, the radial displacement and radial stress are
continuous,
uo ~ = u, T M (m),, k = (a,)) +' (6)
while the inner and outer surfaces of the tube are stress free,
(03: = (o,)F = 0 (7)
Finally, a state of generalized plane strain is assumed to exist:
~ / = ~?*' (8)
The solution so obtained has been programmed for digital computer solution. For the
laminates studied, predicted in-plane stresses and strains were found to correlate closely
with simpler analyses [8] which do not, however, predict the radial stresses important to
this analysis. To facilitate failure analysis of the helical plies, the ~zh shear stresses were
calculated by using the predicted values of normal strains e,, eh, and er in Eq 1.
Unidirectional ply properties were calculated using the unified three-dimensional micro-
mechanics equations of Chamis [9]. For the hoop layers, these are the elastic constants used
in Eq 3 because the material axes for these layers coincide with the cylinder axes. To calculate
the elastic constants for the helical layers, with balanced -+qb plies, the inverted form [5] of
Eq 1 was used. It is of interest that, although the resultant shear stress ~h on the balanced
helical layer is assumed to be zero, its effect on thickness change for each ply is additive.
As a result, predicted values of through-the-thickness Poisson's ratio can be negative for
helical layers with highly anisotropic plies, as was pointed out by Herakovich [10]. This can
have an important effect on interlaminar stresses in the cylindrical laminates considered
here.
Test Results and Discussion

All failures occurred in the central test section, with little apparent effect of the end grips.
In all tests except high temperature ones, audible crackling began at intermediate load
levels. Based on calculated stresses, this was interpreted to be cracking of the resin system
transverse to the fibers.
Stacking Sequence and Elevated Temperature Effects

Two sets of specimens were tested to measure the effect of hoop layer placement and
elevated temperature on performance. The laminates were constructed using Celanese Corp.
Celion 6000 carbon fibers with an epoxy novolac resin system. Reported neat resin tensile
properties of the resin system are 26.3 MPa strength, 4.04 GPa modulus of elasticity, and
1.75% elongation.
Details of specimen construction are listed in Table 1. The important difference between
the two laminates is that one, laminate A, included hoop plies on the inside of the laminate,
while the other, laminate B, did not. All specimens were conditioned at 82~ for at least
24 h, then stored in a plastic bag with desiccant until testing. The tubes were tested to failure
in axial tension at room temperature, and at selected steady-state elevated temperatures.
A constant load rate of 2.5 kN/s was used in all tests. The elevated temperature tests
involved heating the central test section to a specified temperature and maintaining that
temperature constant while applying axial tensile load to failure. Techniques for the elevated
temperature tests were essentially those reported previously for tube compression tests [4].
Figure 3 plots axial load at failure versus average wall temperature for the two sets of
TABLE 1--Tubular specimen details--Celion 6000~epoxy novolac laminates.
Laminate A, With Inside Hoops B. No Inside Hoops
Stacking sequence
(inside to outside) [90/-+20/90/-+20/90] r [-+20/90/---20/90] r
Ply count [2/2/1/2/2]r [2/2/2/2]T
Fiber thickness (mm)
Helical 0.46 0.46
Hoop 0.57 0.49
Laminate thickness (mm) 1.72 1.59
Average fiber volume content (%) 60 60
" 300
z
With inside hoops
a
<~
O 200
-.I
ee--
U.I \
Iv' No inside hoops ~ --~...
:D ". 9 \
--. \
100 \ \
U. \0 9
\ \
0 I I i I I I
0 100 200 300
TEMPERATURE ( d e g . C)
FIG. 3--Failure load versus wall temperature for Celion 6000~epoxy novolac specimens
(Table 1).
specimens. The solid circles represent test points, which have been connected by the broken
lines, indicating performance trends. The difference in performance of the two laminates is
striking. At room temperature, the failure load for the laminate without inside hoops is only
60% of that for the other laminate. As expected, strength of both laminates decreased with
increasing temperatures, showing the effect of resin softening and loss of strength on per-
formance. At the highest temperature, near the heat distortion temperature of the resin
system, the performance of both laminates drops to very low levels. In the specimens with
inside hoop layers, post-test inspection showed increased delamination throughout the test
section as the test temperature increased, with the resin system increasingly unable to resist
delamination and/or collapse of the stabilizing hoop layers. The force of the delamination
was sufficient to severely deform the stainless steel sheathed thermocouple probe, which
had been fitted inside the specimens to measure inside surface temperature. In the specimens
without inside hoop layers, the inside helical fibers pulled away from the laminate in a
localized failure mode, called transverse fiber pullout. The transverse stresses at which this
failure occurred are substantially lower than expected transverse strength, as is discussed
below.
To compare performance of the laminates on an analytical basis, stresses were calculated
for the same baseline conditions, an axial tensile load of 100 kN and a temperature decrease
of 100~ Predicted radial stress distributions under these conditions are plotted in Fig. 4.
Tensile radial stresses, tending to produce delamination, are predicted for both laminates,
reaching their maximum values at the outer surface of the helical layers. The peak value is
25% higher in the laminate without inside hoops, and it occurs in the unsupported inside
helical layer. The temperature decrease, simulating cool down from cure, produces tensile
radial stresses through most of the laminates. An increase in temperature, therefore, as in
F O R A L A N D G t L B R E A T H O N D E L A M I N A T I O N IN F I L A M E N T - W O U N D COMPOSITES 319
§
90 +20 90 +20 90 +20 90 +20 90
I I I I I I I
0.8
eL
=E
e 0.4
I
Ill
n-
I-
-100 C ~
r
.,I
< 0
3<
-0.4
B
"-- I I I I I I I I I I I I
1 1.02 1.04 1.02 1.04
RADIUS RATIO RIR I

FIG. 4--Predicted radial stress distributions at baseline conditions--Celion 6000~epoxy no-
volac laminates A and B (Table l ).
§
90 +20 90 +20 90 +20 90 +20 90
I ~ I I I I I I
2.
a
'5
v
I
w
a,, 1
I-
_l
<
o
<
n-
-1
-2 I I I I I i I i I I I I
1.02 1.04 1.02 1.04
RADIUS RATIO R/R i

FIG, 5--Predicted radial stress distributions at room temperature failure--Celion 6000~epoxy
novolac laminates A and B (Table 1 ).
the elevated temperature tests, decreases these stresses algebraically, working to prevent
delamination. In-plane stress distributions are not plotted because they are quite regular,
and their values coincide quite closely with predictions of conventional in-plane analysis [8].
The helical layers take on high tensile stress in the fiber direction, carrying most of the axial
load. The hoop layers, stabilizing the helical layers, carry compressive stresses in the cir-
cumferential (fiber) direction.
Figure 5 plots the predicted radial stress distributions in both laminates at room temper-
ature failure load levels. Effects of cure were calculated using a 235~ temperature decrease
to simulate cool down from the highest temperature in the cure cycle. The residual stresses
due to cure are plotted along with the stresses due to axial load, and the resultant of these.
The peak radial stress values are much lower than expected room temperature transverse
allowables. No through-the-thickness strengths for filament-wound laminates of the Celion
6000/epoxy novolac material are available.
During room temperature testing, behavior of the specimens with inside hoop layers was
characterized by a steady increase in load to a sudden, catastrophic failure. The mode of
failure was typical of in-plane, fiber-dominated failures, with all helical fibers severed along
a localized axial plane. Only a few hoop fibers held the two pieces of the specimen together.
Post-test inspection showed little evidence of delamination, The predicted helical fiber strain
at failure is 0.01, 83% of ultimate fiber elongation. The in-plane transverse stresses due to
cure and load are additive tension, and transverse cracking is predicted for both the helical
and hoop layers. If cure stresses are neglected, the Tsai-Hill in-plane theory [11] still predicts
failure, with a negative 7% margin of safety.
In contrast, room temperature failure in specimens without inside hoop layers occurred
at load levels much below those predicted with in-plane techniques. These specimens dis-
played a steady increase in load to near the maximum load, at which point there were load
hesitations, until a sudden load drop-off occurred. Visual inspection of the specimen showed
no apparent external damage; removal of the end grips, however, revealed extensive damage
inside the tube, as shown in Fig. 6. Without the stabilizing effect of the inside hoop layers,
the inside helical fibers apparently pulled away from the remainder of the laminate in a
transverse fiber pullout failure mode. The predicted radial stress at the inner helical-hoop
interface is 1.62 MPa (Fig. 5); concomitant fiber direction stress in the inside helical ply is
833 MPa, 47% of the allowable. As before, transverse cracking is predicted in both the
helical and hoop layers when cure stresses are included. Neglecting these, the in-plane Tsai-
Hill failure theory predicts no failure, with a positive 51% margin of safety for the laminate.
Transverse Fiber Pullout Tests

Three sets of Kevlar 49/epoxy specimens were tested to measure the effect of wind angle
on transverse fiber pullout. Wind angles selected were 10~ 20 ~ and 30 ~ Details of specimen
construction are listed in Table 2. Neat resin tensile properties of the epoxy resin system
are 58.6 MPa strength, 2.92 GPa modulus of elasticity, and 3.1% elongation. The specimens
were fabricated in 1974, but not used then because the goals of that test program changed.
They were stored together in the laboratory environment, with no special conditioning, until
used in the current program.
All specimens were tested to failure in axial tension at room temperature at a constant
1.2 kN/s load rate. Figure 7 plots axial strain versus load for three typical specimens; the
readings of the three strain transducers, equally spaced around the circumference, are closely
grouped at all load levels, demonstrating the concentricity of loading. Behavior was nearly
linear until failure. In Fig. 8, the failure loads are plotted versus wind angle. The circles
represent test points, which have been connected by a broken line, indicating trends in
performance. The open circles in the figure represent specimens wound with 80% of the
FIG. 6--View of inside surface after failure--Celion 6000~epoxy novolac laminate B ( Table
1).
hoop fiber thicknesses listed in Table 2. Figure 9 plots predicted radial stress distributions
at failure--due to cure, axial load, and the resultant of these. Effects of cure were calculated
using a 115~ temperature decrease to simulate cool down from the highest temperature in
the cure cycle. Peak radial stresses are much lower than expected transverse strength. No
through-the-thickness strengths for filament-wound laminates of the Kevlar 49/epoxy ma-
terial are available.
In test, the 10~ specimens displayed a steady increase in load to near the maximum load,
TABLE 2--Tubular specimen details--Kevlar 49~epoxy laminates.
Stacking sequence (inside to outside) [-+d~/90/++/90]r

Ply count [2/2/2/2]r
Helical wind angle (deg), ~b 10 2O 30
Fiber thickness (mm)
Helical 0.66 0.66 0.66
Hoop 1.25 1.07 0.85
Laminate thickness (mm) 3.23 2.87 2.57
Average fiber volume content (%) 59 60 59
1.2 -~" 3 0 ~ ~ 2 0 "--k

1.11 \ ~A ".ar oJJ.
0.9
0.8
i 0.7
0.6
~ 0.5
0.4
0.3
0.2
~ 0 20 4-0 60 80 1 O0 120 140 160 180 200

AXIAL LOAD (kN)
FIG. 7--Axial strain versus axial load for Kevlar 49~epoxy specimens ( Table 2 ).
30s
Z
v
r
,< 200
O \
,_1
UJ
nr' .,,,.
,_1
,<
u_ 100
I I I
VO 10 20 30
HELICAL WIND ANGLE ( d e g )

FIG. 8--Failure load versus helical wind angle for Kevlar 49~epoxy specimens (Table 2 ).
§
+10 90 +10 90 +20 90 +20 90 +30 90 +30 90
I-- I, t ~ I I-- I I -- I L~ I I
2.4t - I
a.
1.8 i
1.2,-
-1
IlC 0.6
1.04 1.08 1 1.04 1.08 1 1.04 1.08
RADIUS RATIO R/R i
FIG. 9--Predicted radial stress distributions at failure for Kevlar 49/epoxy specimens
( Table 2 ).
at which point load hesitations occurred, then sudden catastrophic failure. Except for the
load hesitations, the mode of failure was typical of in-plane fiber dominated failure. Post-
test inspection showed that most helical fibers were severed; in addition, they were torn
away from the laminate to the inside, as shown in Fig. 10. The predicted helical fiber strain
at failure is 0.015, 85% of the ultimate elongation. Transverse matrix cracking was predicted
for both the helical and hoop layers. Neglecting cure stresses, the in-plane Tsai-Hill theory
predicts failure, with a negative 7% margin of safety.
The 30~ specimens behaved much differently than the 10~ specimens. After a steady
increase in load, failure occurred first as a load drop off, then as continuous crackling with
almost constant load as crosshead displacement continued. The only apparent damage was
a localized pull away of the inner helical layers from the inside of the laminate, as shown
in Fig. 11. At failure, the calculated maximum interlaminar stress in the inner helical layer
(Fig. 9) is 2.38 MPa; concomitant longitudinal stress in the helical layer is 469 MPa, 34%
of the allowable. In-plane Tsai-Hill criteria, neglecting cure stresses, predict no failure, with
a positive 41% margin of safety.
Behavior of the 200 specimens was not as consistent as the others 9 Both specimens plotted
in Fig. 8 failed by transverse fiber pullout, similar to the 30 ~ specimens. The predicted radial
stress at the inner helical hoop interface (Fig. 9) correlates closely with the radial stress at
failure of the Celion 6000 (20 ~ laminate (Fig. 5). One of the specimens showed little
capability to carry more load; the other continued to carry increasing load, with much
crackling, to catastrophic failure at 227 kN. It appears that local conditions on the inside
surface of the specimens can influence strength in this highly localized mode of failure.
Transverse fiber pullout is potentially an important mode of failure in any curved laminate.
This failure mode can occur whenever curved inner layer fibers in a curved laminate carry
tensile stress9 The mechanism of failure is more localized than for conventional delamination
FIG. lO--View of inside surface after failure--Kevlar 49~epoxy 10~ specimen (Table 2).
failure, and different allowables must be used to predict its occurrence. Tests to determine
through-the-thickness strength allowables [2] are designed to provide a uniform interlaminar
stress distribution. The resulting allowable stresses (43 MPa for AS4/3501-6 graphite/epoxy
in [2]) describe failures in which the entire laminate separates as a unit, usually along the
thin resin layer between plies. To use uniform through-the-thickness tensile strength to
predict the highly localized transverse fiber pullout is probably unconservative.
FIG. l l--View of inside surface after failure--Kevlar 49~epoxy 30 ~ specimen (Table 2).
Conclusions and Recommendations

Delamination is a potentially important failure mode in filament-wound tubes of helical
hoop construction tested in axial tension. The inherent delamination tendencies can be
controlled by proper placement of the stabilizing hoop layers. Tubes that failed with little
indicated delamination at room temperature failed by delamination at elevated temperature.
Further study is required to optimize the required thickness and placement of the hoop
layers, and to determine through-the-thickness allowables for filament-wound laminates,
both at room temperature and at elevated temperature.
Tubes without hoop layers on the inside surface failed in a localized delamination failure
mode, called transverse fiber pullout. This failure mode, characterized by low transverse
allowables, can occur whenever curved inner layer fibers in a curved laminate carry tensile
stress. Additional testing is needed to establish design allowables for this failure mode, both
for laid-up and filament-wound laminates. Local conditions at the surface, like helical wind
pattern and associated crossovers, excess resin, and so forth, may be important in establishing
these allowables.
Once suitable allowables are established, an interactive failure theory must be developed.
The simple criterion of tensile radial stress alone was not justified in our tests. A n improved
analysis, accounting for the local helical layer construction, is needed to better predict
interlaminar stresses and provide improved understanding of the failure modes.
Acknowledgments
The test specimens were fabricated by Brunswick Corporation, Defense Division, Lincoln,
NE, as part of their Internal Research and Development Program. Brunswick and the
Engineering Research Center at the University of Nebraska-Lincoln provided support for
the work. Siu-Ping Ko of University of Nebraska-Lincoln assisted in computer analysis of
the laminates.
References
[l] Chang, E K. and Springer, G. S., Journal of Composite Materials, Vol. 20, Jan. 1986, pp. 30-
45.
[2] Lagace, P. A. and Weems, D. B., "A Through-the-Thickness Strength Specimen for Composites,"
TELEC Report 86-21, Massachusetts Institute of Technology, Cambridge, Nov. 1986.
[3] Foral, R. E and Humphrey, W. D., Journal of Composites Technology and Research, Vol. 7, No.
1, 1985, pp. 19-25.
[4] Foral, R. E, Baldwin, D. D., and McGee, J. M., in Proceedings, 1986 Spring Conference on
Experimental Mechanics, Society for Experimental Mechanics, New Orleans, 8-13 June 1986, pp.
399-405.
[5] Hyer, M. W., "Response of Thick Laminated Cylinders to External Hydrostatic Pressure," Tech-
nical Report No. 87-4, Department of Mechanical Engineering, University of Maryland, College
Park, May 1987.
[6] Vinson, J. R. and Sierakowski, R. L., The Behavior of Structures Composed of Composite Ma-
terials, Martinus Nijhoff, Dordrecht, 1986, pp. 43-49.
[7] Newhouse, N. L., "Mechanical and Transient Hygrothermal Loading of Multilayer Orthotropic
Cylinders," Ph.D. dissertation, University of Nebraska-Lincoln, Jan. 1984, pp. 41-74.
[8] Whitney, J. M. and Halpin, J. C., Journal of Composite Materials, Vol. 2, No. 3, July 1968, pp.
360-367.
[9] Chamis, C. C., SAMPE Quarterly, Society for the Advancement of Material and Process Engi-
neering, April 1984, pp. 14-23.
[10] Herakovich, C. T., Journal of Composite Materials, Vol. 18, Sept. 1984, pp. 447-455.
[11] Jones, R. M., Mechanics of Composite Materials, Scripta Book Co., Washington, DC, 1975, pp.
76-80.
Kevin J. Saeger 1 and Paul A. Lagace L
Fracture of Pressurized Composite

Cylinders with a High Strain-to-Failure
Matrix System
REFERENCE: Saeger, K. J. and Lagace, R A., "Fracture of Pressurized Composite Cylinders

with a High Strain-to-Failure Matrix System," Composite Materials: Fatigue and Fracture,
ABSTRACT: An experimental and analytical investigation was conducted to examine the

fracture behavior of pressurized graphite/epoxy cylinders and coupons that utilize a high strain-
to-failure matrix system. The material system is a five-harness satin-weave cloth made of
Hercules AS4 fiber impregnated with American Cyanamid's CYCOM 907 epoxy. Tests were
performed on standard tensile coupons to determine the elastic constants of the material and
the notched and unnotched fracture characteristics of a quasi-isotropic (0,45)s laminate of this
material system. Two notch types, holes and slits, were examined, and existing failure criteria
were explored as means to extrapolate these coupon data to the prediction of the failure
pressures of axially slit cylinders, which were also tested. The fracture characteristics of this
composite system were compared to a baseline system with Hercules 3501-6 epoxy. In the
laminates using the "tough" matrix system, delamination was not a significant damage mode.
Previous work had shown localized delamination in the 3501-6 laminates. However, both the
coupon tests and the cylinder tests indicate that this high strain-to-failure matrix system is
more notch sensitive than the baseline epoxy system with the relatively brittle 3501-6 matrix.
KEY WORDS: composite materials, notched strength, pressurized cylinders, fracture, "tough"
matrix systems
With the growing use of composite materials in primary aircraft structures, the issue of
postimpact strength retention is increasingly important. C o m m o n occurrences such as the
dropping of a wrench or contact with maintenance vehicles can cause virtually indetectable
damage, which can significantly affect the load-carrying capability of the structure. Natural
occurrences such as hail storms, bird strikes, or impact by loose stones on the runway can
also cause this type of damage. O n e of the primary damage modes in these types of incidents
is delamination. In an effort to suppress postimpact delamination, high strain-to-failure
( " t o u g h " ) matrix systems are being considered. These matrix systems have been shown in
various studies to reduce significantly the amount of delamination associated with an impact
event [I]. H o w e v e r , postimpact strength is not the only concern of the designer. Cutouts
are a necessary part of many structures, often occurring in the form of windows, doors,
access ports, and other needs. These cutouts cause stress concentrations in the structure,
which the material must be capable of carrying.
The notch sensitivity of flat coupon specimens made of composite materials has been
LResearch assistant and associate professor, respectively, Technology Laboratory for Advanced Com-
posites, Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, Cam-
bridge, Mass. 02139.
326
Downloaded/printed
Copyright9 by
SAEGER AND LAGACE ON FRACTURE OF PRESSURIZED COMPOSITE CYLINDERS 327
widely studied (see Ref 2) with the great majority of the work conducted on brittle (that
is, low strain-to-failure) matrix systems. These studies have shown that composites are indeed
notch sensitive, and various failure criteria have been proposed to explain/correlate this
phenomenon. Of the literature available on the effect of high strain-to-failure matrix systems,
the work done by Poe [3] is of particular interest. Poe, who studied the fracture of composites
using the 5208 matrix and the BP-907 matrix, found that when large crack-tip damage was
present, the fracture stress was elevated. The BP-907 matrix tended to suppress this damage,
and, consequently, the fracture stresses were reduced as compared to laminates using the
more brittle 5208 matrix.
To make use of this abundance of coupon data for the prediction of the notch sensitivity
of more complicated structures, it is necessary to develop a methodology to extrapolate the
data properly. Work has been done concerning the notch sensitivity of coupons and pres-
surized cylinders using a five-harness satin-weave fabric of AS1 fiber impregnated with
Hercules 3501-6 epoxy [4-8]. Work done by Graves and Lagace [5] on cylinders with axial
slits showed that a relatively simple correction term could be applied to extrapolate coupon
data to the fracture of cylinders. A survey of previous work on pressurized cylinders may
be found in Ref 8. This survey shows that, for the material systems considered, simple
correction factors could be applied, which correlate flat plate and cylinder data. The purpose
of this investigation is to determine if the material "toughness" has any effect on these
correction terms.
Objectives
The principal objective of this investigation is to determine the similarities and differences
of the failure strengths and the failure mechanisms between cylinders constructed using a
composite system with a matrix of a baseline "brittle" epoxy (Hercules 3501-6) and identical
cylinders constructed using a high strain-to-failure epoxy (CYCOM 907 from American
Cyanamid) for the matrix. Specifically, this investigation examines whether the equations
found for the prediction of failure pressures in composite cylinders using a brittle epoxy
matrix are applicable for cylinders using a high strain-to-failure matrix system. The failure
modes of these cylinders utilizing these two matrix systems are also to be compared.
Approach
The manufacture and testing of composite cylinders are tedious as well as expensive
processes and, thus, few data are available. On the other hand, a great deal of literature
has been written concerning the fracture behavior of fiat composite coupons. If the fracture
mechanisms occurring in these coupon specimens are similar to the fracture mechanisms
occurring in pressurized cylinders, then a relatively simple correlation should exist to predict
the fracture of the structural configuration of a pressurized cylinder using tensile coupon
data. Any such correlation between flat plate fracture and cylinder fracture should preferably
be failure criterion independent. That is, the ideal "geometrical correction factor" would
work equally well with whatever equation is used to predict/correlate coupon failure. The
obvious approach to obtain such an ideal correction factor is to find a simple relation between
the stress field induced in a flat specimen and that induced in a more intricate specimen.
For such a correction factor, only the important details of the stress solutions need be
considered.
Therefore, standard tensile coupon specimens (as shown in Fig. 1) were first tested to
evaluate their fracture characteristics. A total of 70 coupons were manufactured and tested;
25 were unnotched and 40 were notched. The unnotched coupons were used to obtain the
TOP VIEW SIDE VIEW
7-
75 mm "~- GLASS/EPOXY
. , TAB
I
200mm
..,,-GRAPHITE/EPOXY
POXY
FM-123 FILM ADHESIVE
GLASS/EPOXY
r~ 50 mm~-
FIG. 1--Configuration of coupon specimen.
material constants for the system used. A summary of the test program used to determine
the basic material properties is shown in Table 1.
A (0,45)s layup was chosen for the cylinder tests. The fabric used in this investigation has
equal transverse and longitudinal elastic constants, When this material is then used in a
(0,45)s layup, the resulting laminate is quasi-isotropic. The fact that these are fabric laminates
is represented by using parentheses and commas (,) rather than brackets and slashes [/] in
the laminate notation. The angle represents the orientation of the warp fibers, with the fill
fibers at 90~ to this. Thus, a 45 ~ ply has warp fibers at + 45 ~ and fill fibers at - 4 5 ~
Forty notched coupons were first tested to determine the fracture characteristics of the
material system in this configuration. Four notch sizes and two notch types were used in
this investigation, The nominal notch sizes are 3.18 mm, 6.35 mm, 9.53 mm, and 12.7 ram.
The notch types are holes and slits. Both holes and slits were tested to determine if this
laminate is sensitive to notch shape or only to notch size as is the case of the A370-SH/
3501-6 material [5}. Five coupons of each notch type and each notch size were tested. A
TABLE 1--Testing program for basic material properties.
Number of
Layup Specimens Property Measured
(0)4 10 E~, vt2, longitudinal strength

(90)4 5 E:2, v,~, transverse strength
(-+45)~ 5 G~2
CopyrightbyASTMInt'l(allrightsreserved);MonJan1620:12:47EST2012
Downloaded/printedby
(PDVSALosTeques)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
TABLE 2--Test matrix for (0,45)s Coupons
Notch Type
Notch Size, mm Hole Slit
Unnotched 5~
3.18 5 5
6.35 5 5
9.53 5 5
12.7 5 5
~ indicates the number of specimens tested.
summary of the test matrix for the (0,45)s coupons, notched and unnotched, is provided in
Table 2.
A total of 12 cylinders were tested in this investigation. A typical cylinder specimen is
shown in Fig. 2. The only notch type used in the cylinder tests was a slit aligned along the
longitudinal axis. Slit lengths of 51 mm, 76 mm, 102 mm, and 152 mm were used. A n
overview of the test matrix for cylinders is given in Table 3.
Manufacture of Specimens
The material system used was AS4 fiber cloth, in a five-harness satin-weave configuration,
impregnated with CYCOM 907 epoxy. The fabric had an American Cyanamid designation
of CYCOM 907 GF6K5H-60". The 0~ direction is taken as the warp direction of the fabric,
and the 90 ~ direction is taken as the fill direction. For the manufacture of coupons, plates
of the material measuring 350 mm by 305 mm were laid up in the desired orientation and
cured in an autoclave. The cure cycle, as suggested by the manufacturer, begins with 1 hour
at 116~ with a full vacuum. At the end of this 1-hour hold, the pressure is raised to 0.69
MPa. When the pressure inside the autoclave reaches a value of 0.1 MPa, the vacuum is
vented and a temperature rise to 177~ begins. This temperature is held for 1 h and then
the plate is cooled at an average rate of 2.8~ No postcure is recommended or used.
Five coupons were cut from each panel with a water-cooled diamond wheel cutter. Glass/
epoxy loading tabs were bonded onto each end of the specimen with American Cyanamid
P = INTERNAL
PRESSURE
FIG. 2--Configuration of cylinder specimen.
TABLE 3~Test matrix for O, finders.
Number of
Slit Length, mm Cylinders Tested
51 3
76 3
102 3
152 3
FM-123-2 film adhesive. This resulted in the specimen configuration shown in Fig. 1. In
specimens requiring such, slits were manufactured using a jeweler's saw (0.2 mm nominal
thickness) held in a pin vise. The slit was made by hand using a metal rule as a guide. All
of the slits were oriented perpendicular to the loading direction. The holes were made using
diamond-crusted drills and reamers with water cooling. All notches were at the specimen
center. Far-field longitudinal and transverse gages were used on the coupon specimens to
measure elastic constants.
The cylinders were laid up by hand on an aluminum mandrel. Graphite/epoxy plies were
measured and cut with razor knives. In order for the manufactured cylinders to be wrinkle-
free after cure, no slack can be left in the plies during layup. Because of this, the plies were
cut so as to overlap by approximately 10 mm at splices. During layup, these overlaps were
heated with a heat gun. This gave enough tack in the overlapped regions to hold the plies
taut during layup. A method of seam overlap was chosen to obtain as large a seam-free test
section as was possible. The slits later placed in the cylinders were placed directly opposite
these seams in all cases. Once layup of the plies was complete, cure was again accomplished
in an autoclave using the same cure cycle as for the plates. The axial slits in the cured
cylinders were manufactured in a manner entirely analogous to the procedure used for
coupons. The only difference was that care had to be taken to ensure that the slit was
parallel to the axial direction. The final specimen is 152 mm in diameter and 610 mm in
length as indicated in Fig. 2.
The cylinder tests are accomplished via pressurization, thus requiring a sealed specimen.
Slit sealing was accomplished through the use of a thin (0.8 mm thick) piece of aluminum
patch, held in place with masking tape. Because all tests used internal pressurization, the
tape served only to hold the patch in the correct position until pressurization was begun. A
thin patch was able to be used because the slit width was extremely small; therefore, very
little of the patch was unsupported. The use of a patch of this size and the manner in which
it was held in place ensured that very little load would be transferred through the patch.
The patch was then covered with a piece of vacuum bag, which was sealed to the inside of
the cylinder with vacuum tape.
The ends of the cylinders were sealed with aluminum endcaps first developed and used
by Rogers [9]. These caps are 25.4 mm thick with an outside diameter of 330 mm. A circular
groove 12.7 mm deep and 3.18 mm wide is cut in the cap at a radius of 152 mm so that
there is allowance for equal spacing on the inner and outer surfaces of the cylinder. A two-
part potting epoxy (Epon 815 resin and V40 hardener supplied by the Miller-Stephenson
Company) was used to bond the clean endcaps to the cylinder. Strain gauges were applied
on a line running colinear with the slit. These gauges were placed at various distances from
the slit tip, and the readings from these gauges were later used to verify analysis procedures
used to calculate the stress field.
Test Procedure
All coupon tests were conducted using an MTS 810 testing machine equipped with hy-
draulic grips. These monotonic-to-failure tests were done under stroke control at a rate of
12.7 mm per 700 s, which gives a strain rate of approximately 5400 microstrain per minute
over the 200-mm test section. Data were recorded at 0.5-s intervals using an automated data
acquisition system.
The cylinders were tested in a blast chamber. The pressurization was accomplished with
bottled nitrogen via a 6.36-ram-diameter copper tube attached to one of the endcaps with
standard pipe fittings. This tube was then run outside the blast chamber to the bottled
nitrogen. The other endcap was attached with similar tubing and fittings to a Dynisco PT
119G-7.5C pressure transducer. The transducer signal and strain gage signals were fed to
the data acquisition system. The output from the transducer was also fed to an X-Y plotter
so that the pressure could be monitored during the test. Pressurization was performed
manually, via a pressure regulator, attempting to keep a constant pressurization rate of 0.3
MPa per minute, as shown by an X-Y plotter, until failure occurred. The data were sampled
every 0.5 s.
Correlations and Results

Coupons
Much work has been done concerning the fracture behavior of flat specimens in uniaxial
tension. Two criteria shown to yield excellent correlations are the Mar-Lin equation and
the Whitney-Nuismer average (or point) stress criterion [10]. The Mar-Lin equation is based
on the theoretical singularity associated with a crack terminating at the interface between
fiber and matrix. The strength of this singularity has been shown by Fenner [11] to be a
function of the Poisson's ratios and shear moduli of the constituents, fiber and matrix, and
of the angle between the interface and the crack. For a generic fiber-matrix combination,
the Mar-Lin equation has the form:
~t = H~(2r) -m (1)
where
~r = the average failure stress based on virgin area,

2r = the notch length,
H~ = a composite fracture parameter which is determined experimentally, and
m = the value of the singularity at the fiber-matrix interface for a given material system
(for the current material system m = 0.28).
It has been shown [12] that this equation is equally applicable to various notch geometries
if the projection of the notch length on a line perpendicular to the loading direction is used
as the equivalent notch length.
The Whitney-Nuismer average stress criterion states that failure will occur in the laminate
when the average stress over some distance from the edge of the notch, ao, reaches the
unnotched strength, cro, of the given laminate. Unlike the Mar-Lin equation, this criterion
takes on different forms for different notch shapes. For a crack in an infinite flat plate, this
criterion has the form:
ao 11/2
In this equation, Cro is the unnotched fracture strength of the laminate and is determined
experimentally from virgin specimens. The parameter ao is determined experimentally from
notched specimens. Once again, crI is the fracture stress of the notched specimen, and 2r is
the notch length. The value of tro found in this investigation is 508 MPa, and the value for
ao, found by pooling the hold and slit data, is 5.65 mm. For the Mar-Lin equation, the value
of Hc was found to be 576 MPa(mm) ~28.
Both the Mar-Lin equation and the Whitney-Nuismer average stress criterion yielded
excellent correlations for the strength of notched coupons. The correlations of these criteria
with experimental data of the failure stresses of coupons with various slit lengths are shown
in Fig. 3 and is shown in Fig. 4 as a function of hole diameter. In all cases, failure was
associated with the manufactured notches. These two criteria give nearly identical results
for the range of notch lengths considered, 3.18 mm to 12.7 mm, for both notch types
considered. The horizontal line segment of the Mar-Lin correlation indicates that the fracture
strength cannot exceed the measured unnotched fracture strength.
Cylinders
For the case of the cylinders, the internal pressure generates a biaxial stress field with a
2 to 1 ratio in the hoop (2) to the longitudinal (1) direction as referenced to Fig. 2. However,
it has been shown [12], for axially loaded coupons, that a slit aligned along the load direction
does not affect the laminate strength. Thus, the longitudinal stress is ignored for these
cylinders with slits oriented along the longitudinal (1) direction and only the hoop stress is
considered.
When the hoop stress is used in the Mar-Lin equation to extend the prediction of failure
to pressurized cylinders, a correction term must be used to take into account the effects of
the specimen geometry. This correction factor, which was derived for the asymptotic case
by Folias [13], relates the stress state at the crack tip in a pressurized isotropic cylinder to
~ 500
~V"400
~
300
I.tJ
200 MAR-LIN
D
F- oeo= WHITNEY-NUISMER ASC
100
n-
It.
I I I I I I I I
2 4 6 8 10 12 14 I
SLIT LENGTH, mm
FIG. 3--Experimental data and correlations for tensile coupons with slits.
SAEGER AND LAGACE ON F R A C T U R E OF P R E S S U R I Z E D C O M P O S I T E C Y L I N D E R S 333
o 500
Q.
400
03
LtJ
rY
I-- 5 0 0
03
LLI
r 200 MAR-LIN
Z)
I'-- 9 9 9 9 WHITNEY- NUISMER ASC
C~
<Z I 0 0
fie
LL
I I I I I I I I
2 4 6 8 10 12 14 16
HOLE DIAMETER, mm
FIG. 4--Experimental data and correlations for tensile coupons with holes.
the stress state at the crack tip of a flat plate and indicates a stress intensification at a notch
due to the localized bending in the pressurized cylinder. This correction factor is a function
of the cylinder geometry and of the material properties. Thus, the failure stress of a cylinder
is found by dividing the fiat plate solution by this correction factor, K, yielding:
%,,~176 = (%,,p,,,o)/K (3)
The correction factor, K, is given by the equation:
K = (1 + 0.317 k2) '': (4)
In this equation, the shell parameter h is given by:
M = ~Eh r 4 (5)
where
E = the modulus of the material,
h = the shell thickness,
R = the shell radius,
D = the bending stiffness of the material, and
r = the half-crack length.
This correction is valid only for isotropic materials. A similar curvature correction for
specially orthotropic materials has been given by Krenk [14] and yields virtually the same
results in this case.
Unlike the Mar-Lin equation, which required only the stress solution at the crack tip, the
Whitney-Nuismer average stress criterion requires the knowledge of the hoop stress as a
function of the distance from the notch tip, and, thus, an asymptotic correction factor cannot
be used. A knowledge of the stress state would allow the determination of the value of the
integral of the hoop stress over the distance ao (as determined from coupon tests). The hoop
stress and, more importantly, the integral of the hoop stress were obtained in this investi-
gation using the finite difference method [15]. The formulation of the problem was such
that the structural effects were included. With the integral of the hoop stress thus obtained,
it was a simple step to determine the average stress over the ligament from the slit tip to
the distance ao (as determined from coupon tests). When this average stress reached the
unnotched fracture strength, as determined from coupon tests, failure was predicted to
Occur.
Although both the Mar-Lin equation and the Whitney-Nuismer average stress criterion
give very good results when used to correlate coupon data, when they are used to predict
the failure pressures for pressurized cylinders with slits, the Whitney-Nuismer average stress
criterion correlates the experimental data better than does the Mar-Lin equation. As can
be seen in Fig. 5, the Mar-Lin equation predicts failure pressures slightly higher than those
measured during actual tests. Over the range of slit lengths considered, 51 mm to 162 mm,
the failure pressures predicted using the Whitney-Nuismer average stress criterion corre-
spond excellently with the experimental data.
It should be noted that the inability of the extrapolated and corrected Mar-Lin equation
to correlate the data may be the consequence of an inability to apply the asymptotic correction
factor so as to properly capture the behavior of the material at the slit tip. This emphasizes
the need to characterize the stress field properly, irrespective of the criterion used for the
extrapolation of the fracture data. The average stress criterion is inherently more "forgiving"
of local inaccuracies in calculating the stress field because the stress is averaged over a finite
area. This illustrates an advantage of an average stress approach in developing such a
methodology.
Comparison of Material Systems

An important objective of this investigation was to determine the effects of a high strain-
to-failure matrix system on the notched strength of composite cylinders. To gain an under-
standing of the possible effects the matrix may have, coupons were first investigated. First,
the unnotched properties of the test laminate are considered. A summary of the unnotched
properties for both laminates using the 3501-6 and using the CYCOM 907 laminate are given
in Table 4. The unnotched strength found by Graves and Lagace for a (0,45)~ laminate made
with a similar fabric impregnated with the 3501-6 matrix is 626 MPa. The unnotched strength
MAR -LIN
o
Q- 1.2
9 .... c
W~ 1.0
m 0.8
1,1
n," 0.6
0..
w 0.4
n.- OIQ
-J 0.2
~. I I I I I I I I
0.0
0 40 80 120 160
SLIT LENGTH, mm
FIG. 5--Experimental data and correlations for pressurized cylinders with axial slits.
TABLE 4--Material comparison.
Matrix System
Property (Symbol), Units 3501-6 CYCOM 907
Thickness t, mm 1.4 1.67
Unnotched strength ~ro, MPa 626 508
Unnotched fracture load No, kN/m 876 848
Longitudinal modulus EL, GPa 54.1 46.6
Longitudinal stiffness EL " t, MN/m 75.7 77.8
Mar-Lin fracture parameter Hc, MPa(mm) ~z8 763 576
Mar-Lin fracture parameter
corrected for thickness Hc " t, MPa(mm) ~-z8 1068 959
found in this investigation for the (0,45), laminate incorporating the C Y C O M 907 matrix is
508 MPa. This represents a reduction of 19% in the ultimate strength. However, these two
stresses have not been calculated from specimens of equal thickness. The CYCOM 907
laminates have an average cured thickness of 1.67 mm, while the average thickness for the
3501-6 laminates is 1.4 mm. This is an increase in thickness of 16%. Because identical weaves
were used, this increase in thickness should be caused by an increase in the matrix volume
(decrease in fiber volume) of the cured laminate. If it is assumed that the added thickness
is caused by an increase in the matrix content of the laminate and it is further assumed that
the matrix carries very little of the longitudinal load, then the load at failure of the two
unnotched laminates should be similar (that is, within the difference between AS1 and AS4
fibers). If the unnotched failure stresses ~o are multiplied by their respective thicknesses to
obtain the load per unit width at failure, No, a value of No for the 3501-6 laminates of 876
KN/m is obtained, and the N,, for the CYCOM 907 laminates is found to be 848 KN/m.
These values for No are within 3.3% of each other. Furthermore, the average longitudinal
modulus measured for the (0,45)~ CYCOM 907 coupons is determined experimentally to be
46.6 GPa. The longitudinal modulus value found by Graves and Lagace for the (0,45), 3501-
6 laminates is 54.1 GPa. Again, the 3501-6 laminate appears to be 14% stiffer than the
CYCOM 907 laminate. If these values are again corrected for thickness, the 3501-6 laminate
has a value for Es of 75.7 M N / m and the CYCOM 907 laminate has a value for Es of
77.8 MN/m. These values are within 3% of one another. As the AS1 and AS4 fibers have
virtually the same longitudinal modulus, this difference is simply due to normal experimental
error.
Using the Mar-Lin equation to fit the data for both holes and slits, a value for Hc of 574
MPa(mm) ~ is obtained. The value for Hc found by Graves and Lagace for the 3501-6
laminates was 763 MPa(mm) ~28. This shows that the Hc for the 3501-6 laminate is 33% higher
than for the CYCOM 907 laminate. As was done previously, these values of Hc should be
normalized by the thickness. The normalized Mar-Lin fracture parameter for the 3501-6
laminate is 1068 MPa(mm) ~-~8. For the CYCOM 907 laminate, this normalized parameter
takes a value of 958 MPa(mm) ~28. This shows an 11.5% decrease in the fracture load for
the high strain-to-failure matrix system. These limited data indicate that, as Poe [3] sug-
gested, a high strain-to-failure matrix system will give a lower bound for the fracture strength
of a given fiber system. This was explained by Poe by the fact that a matrix, which is capable
of large plastic strains, will restrain typical energy release modes found in laminates with
relatively brittle matrices.
Typical laminate damage modes include delamination, matrix splitting, fiber pullout, and
matrix plasticization. With a high strain-to-failure matrix system, the first three of the damage
modes presumably could be suppressed. The effect of these damage modes at the crack tip
arguably could be accounted for by a softening of the laminate in this damage zone. This
is somewhat analogous to the plastic zone created in ductile materials in nonlinear elastic
fracture mechanics. In any case, this damage zone should act to relieve the stress concen-
tration at the notch tip and should allow for higher fracture stresses. If these damage modes
are suppressed, thereby suppressing the size of the damage zone, the stress concentrations
would remain unalleviated. The failure of coupons using the 3501-6 matrix showed localized
delamination associated with the fracture process [16]. The coupons tested in this investi-
gation showed little to no delamination. This suggests that the CYCOM 907 matrix suppresses
secondary damage.
Because different correlations and experimental techniques were used by Graves and
Lagace [5] and in this investigation, a direct comparison between the failure pressures of
cylinders using the 3501-6 matrix and those using the CYCOM 907 matrix is difficult to
make. However, as can be seen from Fig. 5, the Mar-Lin equation gives a somewhat
optimistic value for the failure pressures experimentally obtained in this investigation. On
this basis, it may be said that the failure pressures of cylinders using the CYCOM 907 matrix
system (with a normalization for the thickness) is less than 90% of the normalized failure
pressures obtained for cylinders incorporating the Hercules 3501-6 matrix system.
Summary
As assessment of the effects of a high strain-to-failure matrix system on the fracture
strength of cylinders with longitudinal slits was made. Coupled with tests on basic coupon
specimens, these tests provide insight into these effects. Based on the results of this inves-
tigation, a few important conclusions can be drawn. The Mar-Lin equation and Whitney-
Nuismer average stress criterion are applicable for the prediction of the notched fracture
strengths of coupons using a high strain-to-failure matrix system. Also, it was found that
the material elastic constants and unnotched fracture strength values obtained using a high
strain-to-failure matrix system, American Cyanimid CYCOM 907, were identical to these
same parameters obtained from a baseline epoxy, Hercules 3501-6, when these two systems
were compared on a constant thickness basis. This high strain-to-failure matrix system does,
however, show a decrease in the notched fracture strength on the order of 10% as compared
to the baseline system employing 3501-6 epoxy. This is attributed to the suppression of
energy-releasing damage modes in the CYCOM 907 laminates, which do occur in laminates
with a brittle epoxy.
Although a comparison with previous cylinder tests is tentative, it does appear that the
cylinders using this high strain-to-failure matrix are more notch sensitive than are cylinders
incorporating the baseline epoxy system. More importantly, this study shows that the meth-
odology of extrapolating coupon data to cylinder fracture can be accomplished for high
strain-to-failure systems in a similar manner as for the baseline brittle epoxy system.
References
[1] Williams, J. G. and Rhodes, M. D., "The Effect of Resin on the Impact Damage Tolerance of
Graphite-Epoxy Laminates," National Aeronautics and Space Administration, NASA TM-83213,
October 1981.
[2] Awerbach, J. and Madhukar, M. S., "Notched Strength of Composite Laminates: Predictions and
Experiments--A Review," Journal of Reinforced Plastics, Vol. 4, January 1985.
[3] Poe, C. C., Jr., "Fracture Toughness of Fibrous Composite Materials," National Aeronautics and
Space Administration, NASA TP-2370, 1984.
[4] Chang, S. G. and Mar, J. W., "'The Catastrophic Failure of Pressurized Graphite/Epoxy Cylinders
Flawed with Slits and Holes", TELAC Report 85-26, Massachusetts Institute of Technology,
Cambridge, MA, March 1985.
[5] Graves, M. J. and Lagace, P. A., "'Damage Tolerance of Composite Cylinders," Composite Struc-
tures, Vol. 4, No. 1, 1985, pp. 75-91.
[6] Kageyama, M., "The Effect of Delaminations on the Failure of Pressurized Graphite/Epoxy
Cylinders," TELAC Report 86-18, Massachusetts Institute of Technology, Cambridge, MA, May
1986.
[7] Chang, S. G. and Mar, J. W., "The Catastrophic Failure of Pressurized Graphite/Epoxy Cylinders
Initiated by Slits at Various Angles," Journal of Aircraft, Vol. 22, No. 6, June 1985.
[8] Lagace, P. A. and Saeger, K. J., "'Damage Tolerance Characteristics of Pressurized Graphite/
Epoxy Cylinders," presented at the Sixth International Symposium on Offshore Mechanics and
Arctic Engineering. OMAE Paper No. 87-607, March 1987.
[9[ Rogers, J. C., "An Investigation of the Damage Tolerance Characteristics of Graphite/Epoxy
Pressure Vessels," TELAC Report 81-12, Massachusetts Institute of Technology, Cambridge, MA,
September 1981.
[10[ Lagace, P. A., "Notch Sensitivity and Stacking Sequence of Laminated Composites," in Composite
Materials: Testing and Design (Seventh Conference), ASTM STP893, J. M. Whitney, Ed., Amer-
ican Society for Testing and Materials, Philadelphia, 1986, pp. 161-176.
[11] Fenner, D. N., "Stress Singularities in Composite Materials with an Arbitrarily Orientated Crack
Meeting an Interface," International Journal of Fracture, Vol. 12, No. 5, October 1976, pp. 705-
721.
[12] Brewer, J. C., "Tensile Fracture of Graphite/Epoxy with Angled Slits," TELAC Report 82-16,
Massachusetts Institute of Technology, Cambridge, MA, December 1982.
[13] Folias, E. S., "Asymptotic Approximations to Crack Problems in Shells," Mechanics of Fracture,
Vol. 3, Leyden, Noordhoff International, 1977, pp. 117-160.
[14] Krenk, S., "Influence of Shear on an Axial Crack in a Cylindrical Shell," International Journal
of Fracture, Vol. 14, No. 2, April 1978, pp. 123-143.
[15] Saeger, K. J., "Damage Tolerance of Composite Cylinders with a High Strain-to-Failure Matrix
System," TELAC Report 86-11, Massachusetts Institute of Technology, Cambridge, MA, May
1986.
[16] Graves, M. J., "The Catastrophic Failure of Pressurized Graphite/Epoxy Cylinders," TELAC
Report 82-10, Massachusetts Institute of Technology, Cambridge, MA, September 1982.
Christos C. Chamis t and Carol A . Ginty ~
Fiber Composite Structural Durability

and Damage Tolerance: Simplified
Predictive Methods
REFERENCE: Chamis, C. C., and Ginty, C. A., "Fiber Composite Structural Durability and
Damage Tolerance: Simplified Predictive Methods," Composite Materials: Fatigue and Frac-
ture. Second Volume, ASTM STP 1012, Paul A. Lagace, Ed., American Society for Testing
ABSTRACT: Simplified predictive methods and models ("theory") to evaluate fiber/polymer-

matrix composite material for determining structural durability and damage tolerance are
described. This "'theory" includes equations for: (1) fatigue and fracture of composites without
and with defects; (2) impact resistance and residual strength after impact; (3) thermal fatigue:
and (4) combined stress fatigue. Several examples are included to illustrate applications of the
"'theory" and to identify significant parameters and sensitivities. Comparisons with limited
experimental data are made.
KEY WORDS: fatigue, fracture, impact, residual strength, thermal, hygral (moisture), com-
bined load, stress concentration, thermal cycles, thermal cracking
Nomenclature
B Thermal fatigue degradation coefficient
E Elastic modulus as defined by subscripts
G Shear modulus
g Gravity acceleration
5 Ply-stress influence coefficient as defined by subscripts
Kr Stress concentration factor
M Moisture
N Number of cycles
R Ratio of cyclic stress, min/max
S Strength as defined by subscripts
T Temperature
To Reference temperature
Tc;o Dry glass transition temperature
Tow Wet glass transition temperature
x,y,z Global (structural axes) coordinates
1,2,3 Ply-material axes coordinates
A Change
0 Ply angle orientation
v Poisson's ratio as defined by subscripts
Senior aerospace scientist and aerospace engineer, respectively, National Aeronautics and Space
Administration, Lewis Research Center, Cleveland, OH 44135.
338
Downloaded/printed
Copyright9 byby
CHAMIS AND GINTY ON STRUCTURAL DURABILITY AND DAMAGE TOLERANCE 339
p Density as defined by subscripts

tr Stress as defined by subscripts
Subscripts
C Compression
C Composite property
L,T Longitudinal, transverse
0 Reference property
C Ply property
S Shear
S Symmetric when following a laminate designation
T Tension
x,y,z Respective coordinate direction properties
1,2,3 Respective ply-material axis properties
Introduction
A major and continuing concern in the fiber-composites community has been the accurate
prediction (even a good approximation) of the structural durability and damage tolerance
of fiber composite structures in service environments. Hygrothermomechanical service en-
vironments are of major concern, that is, temperature, moisture, mechanical loads (static,
cyclic, and impact), and various combinations of these environments. In response to this
concern, a recent research effort at the NASA Lewis Research Center was directed toward
the development of the methodology required to predict the life and/or durability and
damage tolerance of composite structural components in aerospace propulsion environments.
This paper describes the research effort to develop simplified and approximate predictive
methods and models for determining the structural durability and damage tolerance of fiber
composite structural components subjected to hygrothermomechanical aerospace environ-
ments.
These simplified predictive methods and models ("theory") have evolved over the years
by investigating a broad range of composite behavior. These models are of generic, iso-
parametric form for all contributing variables (moisture, thermal, stress, and fatigue cycles).
The models are applicable to structural components with and without defects. The defects
can be those resulting from fabrication, induced damage (impact), or inadvertent damage.
The models are based on composite micromechanics and ply-stress influence coefficients,
thereby rendering them generic and applicable to all types of fiber/polymer-matrix com-
posites including intra- and interply hybrids.
The mathematical form of the models and the significance of the terms in the equations
are described in detail. Use of the models for various applications is illustrated by select
examples. These examples include the following (in part):
1. Fatigue and fracture of smooth and notched laminates.

2. Impact resistance.
3. Residual strength after impact.
4. Residual stresses.
5. Fatigue and fracture after thermal cycling.
6. Crack development after thermal cycling.
7. Thermal and mechanical load cycles to microcrack formation.
Predicted results are compared with available experimental data and discussed with respect
to their significance and application to design.
Fundamental Considerations
Several fundamental aspects underlie the development of the simplified predictive methods
included in this paper. These aspects and respective justifications are as follows:
1. Holes, slits, and impact damage (defects) induce similar strength degradation in fiber-
composite laminates where the characteristic dimensions of these defects are negligible
compared to the planform dimensions of the laminate. If this is not the case, the effects of
the defects must be evaluated using appropriate structural analyses. Experimental data for
holes and slits are shown in Fig. 1 [I], Fig. 2 [2], and in Table I [2]. It is worth noting in
Table I that the failure modes are almost identical for the smooth (unnotched) and defected
(notched) laminates.
2. Fatigue degrades all ply strengths at approximately the same rate. Experimental data
for longitudinal compression, transverse compression, and interlaminar shear [+-45]+ fatigue
are shown in Fig. 3 [3]. Additional relevant data are included in Ref 4.
3. All types of fatigue degrade laminate strength linearly on a semilog plot including: (a)
mechanical (tension, compression, shear, and bending); (b) thermal (elevated and cryogenic
temperature); (c) hygral (moisture); and (d) combinations (mechanical, thermal, hygral,
and reverse-tension compression). Experimental data for compression fatigue are shown in
Fig. 4 [11.
4. Laminates generally exhibit linear behavior to initial damage under uniaxial or com-
bined loading including hygrothermal effects.
120 - -
~, ,,'-- HOLE I I
"0
30 I I J
0 .125 .375 .625
DEFECT SIZE, I•.
FIG. 1--Laminate static fracture data, ([0/+--45/90]2s, T3OO/ E ) (1 ksi = 6.894 O Pa; 1 in.
= 2.54 cm) [11.
250
0 SOLIDSPECIMEN
0 SPECIMENWITH 0.2S-iN. F 1BER LOAD
LENGTH THROUGHSLIT DIRECTION-~ DIRECTION ,---FIBER
200 Z~ SPECIMENWITH O.2S-]N. " ." DIRECTION
DIAMETER THROUGHHOLE
-0
9 150
,7,
2-1N.-WIDE SPECIMEN
50
WITH 0.25-IN. DIA/IETER
CENTERED THROUGHHOLE
I I I I
0 • ~30 • :t60 ~75 •
LN~INATE ORIENTATION, DEG
FIG. 2--Angle-plied laminate longitudinal strengths (1 ksi = 6.894GPa; 1 in. = 2.54 cm)
[21.
5. All ply stresses (mechanical, thermal, and hygral) are predictable by using linear
laminate theory.
6. Stress concentration factors for circular holes are available. They can be obtained from
the literature [5] or can he predicted by using finite-element analysis.
Simulation of Defects and Stress Concentrations

The simulation of defects (holes, slits, and impact damage) for approximate analysis is
depicted schematically in Fig. 5 and is consistent with the justifications previously mentioned.
It can be seen from Fig. 5 that impact damage and slanted slits are also simulated with
circular holes. The difference between the horizontal slit and the slanted slit is the stress
state in the "removed" vicinity of the slit. This state is uniaxial for the horizontal slit but
is combined for the slanted slit.
The stress concentration factor at any point around the perimeter of the circular hole,
for combined stresses, is obtained by superposing the respective stress concentration factors
for the individual stress states. This is depicted schematically in Fig. 6 where equations for
stress concentration factors are given under the respective schematic [5]. The equations are
explicit and relatively simple. (The notation is readily deduced from the schematics and is
also summarized in the section entitled "Nomenclature.") These equations reduce to the
well-known stress concentration factors for isotropic material as follows:
1. Kr (') = 3 (0 = 90 ~ for trc~ or 0 = 0~ for tro,y).

2. K r cy~ = - 1 (0 = 0 ~ for tr~x or 0 = 90 ~ for ~rc.,).
TABLE I. - FRACTURE MODESOF [• s GRAPHITE-FIBER/EPOXY

LAMINATES (DETERMINED BY SEM ANALYSIS)
[Longitudinal tension, LT; transverse tension, TT;
intraply shear, S.]
Ply orien- Fracture modes
tation,
[• s Sol i d Notched specimen Notch specimen
deg specimen with s l i t with hole
0 LT LT,S LT,S
3 LT,S LT,S LT,S
5 LT,S LT,S LT,S
lO LT,S LT,S LT,S
IS LT,S LT,S LT,S
30 LT,S LT,S LT,S
45 S,LT S,LT S,LT
60 TT,S TT,S TT,S
75 TT TT TT
90 TT TT TT
FATIGUE
0 LONG[1-UD[NAL, 0~
I'-I TRANSVERSE,900
/% INTERLANINARSHEAR, r ~
OPEN SYMBOLSDENOTEROOMTEMPERATURE,WET
SOLID SYMBOLSDENOTEROOPITEMPERATURE,DRY
~ ~ ~N"~ 0 0
.G - - RO0. TEMPE,;"., "%,'%,~0 ~
~. "%~'%',~-"~ /--ROOMTEMPER-
"~.." ~ ATURE, DRY
So ttTGo-
,oJ J ,\ -...
I I I I",
0 2 q 6 8 10
LOG. FATIGUECYCLES
FIG. 3--Compressive life and durability AS graphite-fiber~epoxy [3].
1.0
0 ~ []
UNNOTCHED -J
T
I I I I I I I
1 2 3 4 5 6 7
L06, FATIGUE CYCLES
FIG. 4--Notch laminates are not sensitive to compression fatigue (for [0/+-45/0/9012s, T300/
934) (1 ksi = 6.894 GPa) Ill.
3. K r "'~ = - 4 (0 = 45 ~ and 225 ~ for trc~,.), K r ~x'~ = 4 (0 = 135 ~ and 315 ~ for cr,~),
Kr I''~ = 0 (0 = 0, 90, 180, 270 and 360 ~ for ~r,~y).
In order to use the stress concentration equations in Fig. 6, the following factors must be
known: (1) the laminate elastic constants about the structural axes, and (2) the tangential
modulus E,.0e at the point where the stress concentration factor is being calculated. These
are easily determined by using laminate theory and the well-known transformation equations
[6,7]. For general laminates under combined stress, the position 0 at which try00is m a x i m u m
is generally u n k n o w n . O n e way to determine the position 0 is to calculate the stress con-
centration factors K r at regular small angle intervals (about 5 ~ and then determine the
maximum ~rL.oofrom
a~oo = max(K/xkrcx, + K r I'~ Gr,.yv + Kr""kr,,,) (1)
(0 -< 0 -< 180 at 0 = 5 ~ intervals)
where cr.... ~c,.,., and ~c,y are the far-field stresses and are generally known from structural
analysis.
The ply-stress concentrations at that 0 are then calculated by using laminate theory or
ply-stress influence coefficients [7,8]. To use the ply-stress influence coefficients, it is usually
convenient to transform the local stress, ~c00, to an equivalent structural axes-stress state
obtained from the following equations:
~c,, = ~,oo sin: 0 ]

fiery = O'coo COS 2 0
~rc,~ = - 89 tr~oo sin 20 )
(2)
llltt 11111 11111 tltll
%1/
0
lllll 11111 lllll lllll

HOLE SLIT INPAET DANAGE SLANTED SLIT
+ 0 f-
EOUI VALENT
FIG. 5--Simulation of defects in laminates for approximate analysis.
'
( ~--X X ~ X
K~X) = Ocee : Ecoo K~Y) =~ = Eceo K~XV) :~ = ecee

(]CXX ECXX OCYY ECyY OCXY 2Ecxx
x I-.o cos2 o x I-.o s,.2o x [-%i,, 9 .o~ ,D,,I

+ (1 + RDI ) SIN 2 O] + (1 + RDI) COS2 O] X SIN 20
.HE.E "o: [E~xx/E~] v2 AND "DI L \E~ G~x~J
FIG. 6--Superposition of respective stress concentration factors. Defect simulated by circular

hole with diameter equal to defect size. Linear behavior to first ply or interply damage. Note:
creedis the hoop stress at 0 [5].
The ply stresses can now be determined from the following ply-stress influence coefficient
matrix equations:
~,.~2t = /~lX/X :Jr/r ,gT/S I or,,,. (3)

L dS/X 5S/Y ~JS/S_] [~r.,,J
This equation is easily used when the ply-stress influence coefficients are available [8] as
summarized in the Appendix. If not readily available, they can be determined from laminate
theory, in general.
In summary, the step-by-step procedure for determining the ply-stress concentrations to
be used for durability and damage tolerance assessment of composites by using simplified
methods is as follows:
1. Replace the defect with an equivalent hole in an infinite medium.

2. Determine the elastic properties of the laminate.
3. Generate tables for the stress concentration factors at regular angle intervals (about
5~ by using equations in Fig. 6 and the tangential modulus equation which can be found in
the Appendix.
4. Determine the far-field stress state due to the loading conditions. Note that this stress
state must be transformed for the slanted slit case as noted in Fig. 5.
5. Determine the maximum value for c~co0from Eq 1.
6. Determine the laminate-stress concentrations due to tL,, from Eq 2.
7. Determine the corresponding ply-stress concentrations from Eq 3.
Fatigue and Fracture for Mechanical Load Only

Durability and damage tolerance are generally assessed by determining the stress that the
composite can sustain (1) under repeated loading (fatigue), (2) after damage, and (3) under
combinations thereof. The simplified methods (equations) for determining the fatigue stress
of composites without and with defects with constant stress amplitude are summarized in
Fig. 7 for mechanical loads only. These simple methods predict the fatigue and fracture in
the first ply in its weakest (first-to-fracture) mode, which may be thought of as either defect
initiation or defect growth initiation. The hygrothermal effects are accounted for in the
equations as well as the stress concentration effects. To use these equations to determine
the ply-fatigue uniaxial stress S,,, at initial defect growth, all the other variables in the
equations must be known. For example, the service environment temperature T and the
number of cycles N are known from the design requirements. The glass transition temper-
ature in the presence of moisture T~w can be estimated from the equation [8] in the Appen-
dix. The cyclic stress degradation coefficient B is usually determined from experimental
data [3].
Comparisons of the predicted results from the equation for composites without defects
are shown in Fig. 3 for graphite and epoxy and in Fig. 8 [9] for glass and polyester. In both
these figures the agreement is considered good because it appears to provide a lower bound.
Comparisons for composites with defects under reverse fatigue are shown in Fig. 9. The
two predicted curves in this figure show the difference between the static fracture stresses
for tension and compression. As can be seen, the experimental data are approximately
between the two predicted curves.
I O
S~cYc = [ T G w - T ] I12
Sto LTGD - ToJ
- B LOG N
I
(A)
Io
s~o [TG0- To] 0

- B LOG N
I
(B)
1
(C)
FIG. 7--Fatigue and fracture for mechanical load only. Criteria for first ply failure are (!)
transply cracking and (2) delamination. (a) Laminate without defects. (b) Laminate with
defects. (c) Cyclic load, constant stress amplitude.
CYCLIC STRESS
RATIO
(MIN/MAX),
70 c-BASELINEDATA, R
/ R = 0 0 -0.10
GO / I-I -. 25
,~ / A -. 50
5o . . ~ ~ . 0 0 0 - i .oo
40 '-"~--~.. - " ~ zx zz zx zx
I / ""
o I I I I "I
5 q 5 6 7
LOG FATIGUE CYCLES
F I G . B--Comparison o f effect o f mean stress on fatigue endurance (tests conducted at 70~
and 50% RH) [l ksi = 6.894 MPa; I~ = 5/9 (~ - 32~
C H A M I S A N D G I N T Y ON S T R U C T U R A L D U R A B I L I T Y AND D A M A G E T O L E R A N C E 347
ENSI0" TENSI0"
SOt-- IR:-'
O-- " ...........
30 - - COMPRESSION..-/ ~ ' - - . ~
o
d 20- 2S
. . . . . SCXXT = 110 KSl
ccYc = I - 0.05 LOG N
10- Sco Scxxc = 80 KS[
o I I I I I I I
2 5 q 5 6 7 8
LOG FATIGUE s
FIG. 9--Comparisons of predicted fatigue strength with data (I0/~-45/0/90L, s T3OO/E lam-
inate) (l ksi = 6.894 MPa).
Impact Resistance
The simplified methods (equations) for predicting the composite damage due to impact
are outlined in Fig. 10. The notation used in the equations is self-evident. The quantified
damage is that due to internal delamination which is generally not visually detectable. The
energy balance is determined by assuming the following: (1) the impacting projectile does
not impart any velocity to the target; (2) the projectile does not absorb any energy; and (3)
(1) CRITERIA
INITIAL DAMAGEWITH NO PENETRATION
INITIAL DAMAGEWITH INTERLAMINAR DELA~IINATION
T-PROJECTI LE
(2) EQUIVALENCE \
RIGID PROJECIILE SPHERE (D)
DAMAGEDAREABOUNDARIESARE
0.5 x PROJECTILE-PROJECTEDAREA
p ,D,v
ENERGY BALANCE
1 ~ D3 p V 2 = 1 ~ D2 t
c
IMPACTING VELOCITY TO INITIAL DAMAGE

9 . ~-TARGET
v [o i iC i s,,]
HEIGHT-DROP EQUIVALENT
FIG. lO--lmpact resistance for local laminate characteristic.
that delamination is "generally" initiated simultaneously in several interply locations through

the thickness. This third assumption is based primarily on observations of dissected, impacted
laminates which generally show several interply delaminations but no surface damage [1].
The important composite parameters in these equations are (1) the laminate thickness tc
and (2) the composite interlaminar shear strength Sty3. The impacting velocity or height-
drop to induce delamination will increase linearly with composite thickness and/or with
composite interlaminar shear strength. For example, if it is assumed that (1) the typical
interlaminar shear strength for graphite-fiber/epoxy composites is 12 ksi, (2) the laminate
is 2.54 mm (0.1 in.) thick, and (3) a 25.4-mm (1-in.) diameter steel ball is the projectile,
then the impacting velocity to initiate delamination is
V =
[ 0.75 ( 386.4 ins2 0.3-11~
i. \1.0 in.// 12 000 i-~2J
V = 1077 in./s ~- 90 ft/s
which is reasonable and consistent with suggested literature values [10].

In another example, the height-drop equivalent for delamination initiation of a 25-1b tool
box is estimated by using the same laminate. By employing the last equation in Fig. 10,
assuming that the impactor is 25.4 mm (1 in.) in diameter, and the equivalent density is
about 50 lb/in), the height-drop equivalent H is
H = [1.5 (0.1 in./1 in.) (12 000 lb/in.Z)/50 lb/in. 3]
H= 36 in. = 3.0ft
which is also reasonable based upon literature values [11].

The two previous numerical examples demonstrate the applications of equations in Fig.
10 for estimating impact loads to be used, in turn, for the sizing of laminates during prelim-
inary design phases. It is understood that these laminate designs must be validated with
more sophisticated analyses (for example, finite element) and verified with strategically
selected experiments.
Thermal Fatigue and Thermal Cycles to Initial Cracking

The simplified method, criteria, concept, and equation to predict the thermal cycles to
initial transply cracking (referred to as microcracking in the literature [I2]) are summarized
in Fig. 11. It is important to note in this figure that the cyclic temperature amplitude AT is
measured from the cure temperature which is different than the glass transition temperature
(dry or wet). The predictive model (equation) for the number of thermal cycles Nr is shown
at the bottom of Fig. 11. To use this equation, the various temperatures Tc~w, Tc~o, T, and
To, the cyclic temperature (elevated, room, cryogenic) amplitude AT, the room temperature
ply-transverse strength 5f22o , the ply-transverse stress tr~.2,,y,.,and the thermal fatigue deg-
radation coefficient B must be known.
The step-by-step procedure for using this equation is as follows:
1. Obtain T~o from the material supplier.

2. Determine T~w from the equation in the Appendix for the mission-specified moisture.
3. Obtain T from the mission-specified conditions. This is usually the maximum and
minimum cyclic temperature.
//~-CURE TEMPERATURE
/
ELEVATED
TEMPERATURE
"
AT r TIME
L-ROOM
TEMPERATURE . ~ CRYOGENIC
TEMPERATURE
THERMAL CYCLES
LoG -' I('~ - T - ~

NT B LVGD - To/ S ,,o
0s = MAXIMUM TRANSVERSEPLY STRESS AT ~T TO YIELD
MAXIMUM (Op,y~,cyc/Sp,~,yo) RATIO
FIG. 11--Thermalfatigue and cycles to initial cracking (first ply initial transverse cracking).
4. Room temperature in general is the reference temperature, T,,.

5. Obtain S,z,_,, from the material supplier or by using micromechanics equations [13].
6. Calculate the cyclic ply-transverse stress ~22~,., by using laminate theory at the corre-
sponding A T.
7. Select an appropriate value for the thermal fatigue degradation coefficient B. Values
for this coefficient are not currently available in the literature to the authors' knowledge.
However, some guidelines for selecting estimates are described in Ref 3.
To demonstrate, this procedure was used to predict the number of thermal cycles to
transply crack initiation for two laminate configurations with three different cyclic temper-
atures (AT = - 1 0 0 , -280, and -600~ The results are summarized in Table II. As
expected, the laminate configuration and the cyclic temperature influence the number of
thermal cycles to transply crack initiation. One very interesting and surprising result is the
high sensitivity of the thermal cycles to relatively small changes in the magnitude of the
thermal ply-transverse stress at cryogenic temperatures. As can be seen in the table, a 2%
increase in cryogenic stress magnitude causes a 44% decrease in thermal cycles.
Factors that contribute to the transverse ply stress will also contribute significantly to the
number of thermal cycles to initial transply cracking. The most prominent of these factors
and most difficult to quantify accurately include: (1) temperature profile through the laminate
thickness; (2) thermal cyclic degradation coefficients B; and (3) in situ ply-transverse
strength.
The sensitivity of the temperature profile through the laminate thickness on the number
of thermal cycles to initial transply cracking is shown in Table III. As can be seen, it is very
significant. It also illustrates, in part, the difficulty that would be encountered in pinpointing
this number by measurement. The results in this table also show that the smallest number
of cycles occurs for the ply with the greatest temperature difference. These results are
consistent with the findings reported in Refs 12 and 14.
The sensitivity of the number of thermal cycles to (1) thermal cyclic degradation coefficient
B and (2) in situ ply-transverse strength St.o are shown in Fig. 12a for elevated temperature
TABLE I I . - SUMMARYOF THERMAL CYCLES PREDICTION FOR

LAMINATES [02/902] s AND [03/90] s AT THREE CYCLIC
TEMPERATURES (T300 GRAPHITE-FIBER/EPOXY MATRIX)
[Cure temperature assumed at 350 ~ glass t r a n s i t i o n
temperature, 420 ~ dry conditions, M = 0 percent;
assumed value for B = 0 . I 0 . ]
Condition or property Cyclic temperature,

AT, ~
-280 -100 -600
Use temperature, T, ~ 70 250 -250

Thermal cyclic ply-transverse 4.7 1.8 (a)
stress, a~cvc, ksi
Ply-transversE strength, 8 6 II
S~o, ksi, (at 70 ~
Thermal cycles (NT) to I0 902 52 399 (b)
i n i t i a l transply cracking
aFor laminate [02/902] s, 9.3 ksi; for laminate

L [03/90]2, 9.5 Rsi.
UFor laminate [02/902] s, 162; for laminate [03/90] s,
91.
CONVERSION FACTORS: 1 ~ = 5/9 (~ - 32 ~
I ksi = 6.894 GPa.
TABLE I l l . - NUMBEROF THERMAL CYCLES TO INITIAL TRANSPLY

CRACKING IS VERY SENSITIVE TO TEMPERATURE PROFILE
THROUGH-THE-LAMINATE THICKNESS ([02/902] s) T300
GRAPHITE-FIBER/EPOXY (LAMINATE)
[Cure temperature, 350 ~ glass t r a n s i t i o n temperature, 420 ~
thermal cyclic degradation c o e f f i c i e n t , O.l; room temperature
dry ply-transverse t e n s i l e strength, 8 k s i . ]
Ply o r i e n - Temperature, AT, Use Ply residual Thermal

tation, oF temper- stress, cycles,
deg ature, oQ2Z , NT
Elevated Cryogenic ~ psl
-100 . . . . . . 250 1771 52 399

-125 225 2250 51 785
-600 -250 9347 29
-575 -225 8924 282
90 150 . . . . . . 200 2695 35 748

175 175 3174 23 259
-550 -200 8544 484
-525 -175 8121 819
CONVERSION FACTORS: I ~ : 5/9 (~ - 32 ~

1 ksi : 6.894 GPa.
18 - -
0 PREDICTEDPLY STRENGTH
16
~ I-I 80 PERCENTPREDICTEDVALUE
Z~ 65 PERCENTPREDICTEDVALUE
0 50 PERCENTPREDICTEDVAUE
lq "~I C) 1.1 x (PREDICTEDVALUE)
0 1.25 x (PREDICTEDVALUE)
12
10
6 "~,
q
2
d
(A)
12 - - I
10 --
_q
o .1 .2 .3 ., .5 .~ .~ .8 .9 1.o
(B)
FIG. 12--Thermal cycles to failure, sensitivity to degradation coefficient B and to in situ

ply strength [O/90j]s T3OO graphite fiber/epoxy laminate [I~ = 5/9 (~ - 32~ (a) Elevated
temperature, T = 175~ (b) Cryogenic temperature, T = -250~
cycling and in Fig. 12b for cryogenic temperature cycling. The following characteristics are
observed from these two figures:
1. The thermal cycles Nr are very sensitive to B for values around 0.1 and also to SL.o.
2. The sensitivity is more manifest in cryogenic temperature cycling than in elevated
temperature cycling.
3. The sensitivity diminishes progressively as B increases beyond 0.3 for both temperatures
except for the 50% S~.ocurve at cryogenic temperatures.
The important conclusion from this discussion is that the predictive model in Fig. 11
provides a simple and effective means to estimate the thermal cycles to initial transply
cracking and its sensitivity to various participating parameters.
Combined Hygrothermomechanical Cyclic Loading

The predictive equation for combined cyclic loading is shown in Fig. 13. The combined
loading includes (1) steady state, (2) mechanical load cycling, (3) thermal cycling, and (4)
hygral (moisture) cycling. The notation used in the figure is appropriately defined. Note
that the equation, as shown, applies only to one uniaxial ply stress and for constant amplitude
of each of the cyclic loads.
The equation is general, however, and can be applied to a variety of conditions, for
example, (1) variable cyclic amplitude can be handled by applying the equation to each
amplitude and then summing the corresponding terms and (2) combined stress states can
be handled by using available combined stress failure criteria [7]. In either case, care should
be exercised in using the respective cyclic stress degradation coefficients for the various ply
stresses. The recommended (default) value from Ref 3 is 0.1 for this coefficient for all types
of stresses when no other values are available.
The predictive equation in Fig. 13 was used to estimate the number of cycles for various
t t I t
STEADY-STATE CYCLIC STRESS, L CYCLIC TEMPER- CYCLIC ROISTURE, f't
STRESS, L ATURE, T
\ /L
s.j&=
Sf,0 FHT ;
(Sj~cYC~ =
\S--~O / L FHT- BL LOG N L : (S/,cYc'~ =
S~o/T F.T - ~ Lo* "T
=
: [To - T]i'
S/'O 'JR FHT - ~ LOG NR :
FIG. 13--Combined hygrothermomechanical cyclic loadings (KT = stress concentration

factor).
TABLE IV. - CYCLES TO INITIATE TRANSPLYCRACKS DEPENDON LAMINATE CONFIGURATION AND TYPE OF LOADING
CONDITIONS (AS GRAPHITE-FIBER/EPOXY AT 0.62 FVR)
[Fatigue degradation coefficients are assumed to be BM = DT = BL = 0. I . ]
Ca) Ply-transverse stress influence coefficients
Loading condition Laminate configuration
[• s [902/• s [• s
Stress along 0~ P!Y direction, OcxX 0.145 0.170 -0.022
Use temperture, Td -16.681 -12.280 -7.698
Use moisture, M -1298.265 -1170.719 -571.146
Transverse ply-strength tension/compression, 8/20 8/20 8/20
Sd22T/SQ22C, ksi
(b) Cycles to f a i l u r e
Loading condition Laminate configuration Comments
[• s [9021• s [:30/03] s
Applied stress, NL 36 SO0 4220 >lOOxlO 6 acx x = 30 ksi
Applied stress and residual stress, NL I I >lOOxlO 6 acx x = 30 ksi + oR
Elevated temperature cycling b, NT 76 620 234 060 987 000 T = 250 ~
Cryogenic temperature cycling b, NT 180 275 260 >lOOxtO 6 T = 300 OF
Moisture cycling c, NM >lOOxlO 6 >lOOxlO 6 >lOOxlO 6 M = 1 percent
aCure t e m p e r a t u r e , 350 *F.

b C y c l i c t e m p e r a t u r e s a r e from room t e m p e r a t u r e to T.
C C y c l i c m o i s t u r e i s from d r y to M.
CONVERSION FACTORS: 1 ~ = 5 / 9 ( * F - 32 OF);

1 ksi = 6 . 8 9 4 GPa,
loading conditions for three different laminates:
1. [ - 45/O/90],--quasiisotropic.
2. [902/-+-lO],--pressure vessel.
3. [-+30/03],--engine blade.
The results obtained are summarized in Table IV. It can be seen that: (1) mechanical load
with residual stress is the most critical; (2) moisture cycling, only, is insignificant for 1%
moisture by weight and can be neglected; and (3) blade laminate configurations are not
likely to exhibit transply cracks under any cyclic loading conditions within practical ranges.
The significant point is that the combined cyclic stress degradation effects can be evaluated
for a variety of conditions by using the predictive models described in Fig. 13.
Summary of Results
The results of an investigation of simplified predictive methods and models ("theory")
for composite structural durability and damage tolerance are summarized as follows:
1. The theory consists of simplified closed form equations used to assess composite ma-
terial property, structural durability, and/or damage tolerance.
2. The theory accounts for fatigue and fracture for the most commonly occurring inad-
vertent defects in composites.
3. The theory is limited to predicting cycles to initial damage or initial defect extension.
4. The theory is based on integrated composite mechanics, accounts for hygrothermal
effects, and is applicable to combined hygrothermomechanical cyclic loading.
5. The theory is suitable for preliminary design.
6. The theory is expedient for parametric studies (composite systems), laminate config-
urations, temperatures and/or moistures, and combined loading conditions.
7. The theory predicts results that appear to be reasonable for assumed representative
conditions and for the limited data available. However, it should be used judiciously because
additional verification is needed.
More specific findings are summarized as follows:
1. Fatigue cycles to initial cracking depend on (a) composite material, (b) laminate con-
figuration, and (c) type of cycle loading.
2. Thermal cyclic transply crack initiation and location depend on (a) temperature gradient
through the thickness, (b) environmental effects on all properties, especially on transverse
strength, and (c) fatigue degradation rate of transverse strength.
3. Composite impact resistance depends on (a) laminate thickness and (b) interlaminar
strength. Subsequent residual strength depends on the respective stress concentration factor.
4. Cyclic cryogenic temperatures significantly degrade the laminate fatigue resistance
more so than elevated temperatures.
5. Moisture cycling (up to 1% by weight) has generally negligible degradation effects on
composite fatigue resistance.
6. Thermal fatigue cycles are very sensitive to low values (less than about 0.2) of the
fatigue degradation coefficient.
APPENDIX
Summary of Useful Equations
The equations summarized in this appendix are required to perform the calculations of
the simplified predictive methods. (See "Nomenclature" for the notations.)
Ply-stress influence coefficients are as follows:
(1) For trcx~# 0 (trc~y = trc~r = 0),
.~JLIX = crt'l'''2t= E''---'!1(cos 2 0 - vc.ysin 2 0)

I~cxx Ecxx
5TIX- ~rr22 - Et,22

Ec= [(v,,n - vc.r) cos 2 0 + (1 - vc..vr,..) sin z 0]
O'cxx
,r -- ~,'lz __ -- G
- o2 (1 + v~.r) sin 20)
~ E~
(2) For r = tr=,, = 0),
Et, ll
~JLIY- ~m - ~ (sin z 0 - V~y cos 2 0)
g T/Y - trr22 -
Er22
E~yy [ ( i - v~yvt,,z) cos z 0 + (v,,n - V~y) sin 2 0]
O'cyy
5JS/ Y = crt'2-''A = ~ ( 1 + 1)cxy) sin 20

~cyy
(3) For Cr.y ~ 0 (cr~x = cr~y = 0),
9 L / S = crt'l---2 = .Et,, (1 - v,.21)sin 20

(~cxy 2Gcxy
5 T / S - crt'22 - -- -Et~22(1 - vt.zl) sin 20

ff cxy Gcxy
~S/S . ~. . . Gt'12 cos 20

I~cxy ac,y
The tangential modulus is
E~oo E~----~ ~ + 4 y sin: 20
The glass transition temperature for wet composite is
TGw = (0.005 M - 0.1 M + 1)Too
References
[1] Porter, T. R., "Compression and Compression Fatigue Testing of Composite Laminates," NASA
CR-168023, National Aeronautics and Space Administration, Washington, DC, 1982.
[2] Irvine, T. B. and Ginty, C. A., Journal of Composite Materials, Vol. 20, No. 3, Mar. 1986, pp.
166-184. (NASA TM-83701.)
[3] Chamis, C. C. and Sinclair, J. H., in Composite Materials: Testing and Design (Sixth Conference),
ASTM STP 787, I. M. Daniel, Ed., American Society for Testing and Materials, Philadelphia,
1982, pp. 498-512.
[4] Grimes, G. C. in Test Methods and Design Allowables for Fibrous Composites, ASTM STP 734,
C. C. Chamis, Ed., American Society for Testing and Materials, Philadelphia, 1981, pp. 281-340.
[5] Lekhnitskii, S. G., Anisotropic Places, Gordon and Breach, New York, 1968.
[6] Agarwal, B. D. and Broutman, L. J., Analysis and Performance of Fiber Composites, Wiley, New
York, 1980.
[7] Murthy, P. L. N. and Chamis, C. C., "Integrated Composite Analyzer (ICAN) Users and Pro-
grammers Manual," NASA TP-2515, National Aeronautics and Space Administration, Washing-
ton, DC, 1986.
[8] Chamis, C. C., "Prediction of Fiber Composite Mechanical Behavior Made Simple," NASA TM-
81404, National Aeronautics and Space Administration, Washington, DC, 1980.
[9] Lark, R. E and Chamis, C. C., "Hygrothermomechanical Evaluation of Transverse Filament Type
Epoxy/Polyester Fiberglass Composites," NASA TM-83044, National Aeronautics and Space
Administration, Washington, DC, 1983.
[10] Liu, D., Malvern, L. E., and Sun, C. T. in Proceedings, 1986 SEM Spring Conference on Ex-
perimental Mechanics, Society for Experimental Mechanics, Bethel, CT, 1986, pp. 863-868.
[11] Sollars, T. A. in Proceedings, 28th AIAA/ASME/ASCE/AHS Structures, Structural Dynamics
and Materials Conference, Part 1, American Institute of Aeronautics and Astronautics, New York,
1987, pp. 362-375.
[12] Wilson, D., Wells, J. K., Hay, J. N., Lind, D., and Owens, G. A., SAMPE Journal, Vol. 23,
No. 3, May-June 1987, pp. 35-41.
[13] Chamis, C. C., "Simplified Composite Micromechanics Equations for Strength, Fracture Tough-
ness, and Environmental Effects," NASA TM-83696, National Aeronautics and Space Adminis-
tration, Washington, DC, 1984.
[14] Adams, D. S., Bowles, D. E., and Herakovich, C. T., Journal of Reinforced Plastics and Com-
posites, Voi. 5, No. 3, July 1986, pp. 152-169.
Eugene Dan-Jumbo, 1 Alan R. Leewood,2 and C. T. Sun 3
Impact Damage Characteristics of

Bismaleimides and Thermoplastic
Composite Laminates
REFERENCE: Dan-Jumbo, E., Leewood, A. R., and Sun, C. T., "Impact Damage Char-
acteristics of Bismaleimides and Thermoplastic Composite Laminates," Composite Materials:
ABSTRACT: Impact damage in laminates of four different composite systems was investi-
gated. These composite systems included two thermoplastic-based composites (PEEK and
Torlon) and two epoxy-based composites (toughened and brittle BMIs). At low impact ve-
locities, PEEK (polyetheretherketone) and Torlon composites showed significantly less matrix
cracking and delamination and greater strength retention capability than the two epoxy-based
composites. However, beyond a certain threshold velocity, the thermoplastic composites ex-
perienced a sudden drop in flexural strength, more so than the epoxy-based composites.
Between the two epoxy-based systems, the toughened BMI composite exhibited much better
impact resistance properties.
KEY WORDS: thermoplastic composite, brittle BMI, toughened BMI, impact damage, de-
lamination, matrix cracks, contact law, residual strength
Due to the brittleness of epoxy resins, epoxy-based fiber composites are susceptible to
impact damage. Thermoplastics, having greater toughness, are considered to have the po-
tential for alleviating this problem [1].
Although damage inflicted by low-velocity impact appears quite complicated, the major
failure modes include only matrix cracking, delamination, and fiber breakage [2]. As pointed
out by Sun and Rechak [3], the delamination mode of failure is induced by matrix cracks
which occur prior to other failure modes. Thus, suppression of matrix cracking will suppress
delamination. It is conceivable that the use of tougher matrices will yield composites that
are more resistant to impact damage.
Impact damage and dynamic response characteristics were studied experimentally for four
different composite systems. Two were epoxy-based composite systems with brittle and
toughened BMI matrices, respectively. The other two composites were thermoplastic-based
systems--graphite/polyetheretherketone (PEEK) and graphite/Torlon composites. All lam-
inates were made using a ~r/4 quasiisotropic construction with various thicknesses. Impact-
induced failure modes were examined using sectioning and microphotographic techniques.
The contact rigidity, dynamic response, dynamic failure strain, and flexural residual strength
were also investigated experimentally.
Composite Structural Certification and Analysis, General Dynamics, Fort Worth, TX 76101.
2 Vice president, Advanced Composite Engineering, West Lafayette, IN.
3 Professor, School of Aeronautics and Astronautics, Purdue University, West Lafayette, IN 47907.
356
Downloaded/printed
DAN-JUMBO ET AL. ON IMPACT DAMAGE CHARACTERISTICS 357
TABLE 3--Contact coefficients.
k otc,
Laminate No. (lbf/in. ~'5) (N/cm 15) (in. • 10 -3) (cm X 10 -3)
1 1.125 54 • 107 1.2368 • 107 1.54 3.91

2 1.098 63 x 107 1.2073 x 107 0.77 1.96
3 1.248 55 x 107 1.3720 x 107 1.26 3.20
4 1.207 55 x 107 1.3270 x 107 0.996 2.53
5 0.859 4 • 107 0.9439 x 107 0.755 1.92
6 1.128 46 x 107 1.2400 x 107 1.047 1.047
Experimental Procedures
Specimen
T h e c o m p o s i t e m a t e r i a l s used in c o n s t r u c t i n g the l a m i n a t e s for this work were A m e r i c a n
C y a n a m i d ' s t o u g h e n e d B M I ( T 3 0 0 / x 3100), U.S. Polymeric's Brittle B M I ( T 3 0 0 / V 3 7 8 A ) ,
I C I ' s semicrystalline P E E K , a n d A m o c o ' s a m o r p h o u s Torlon. Two different fibers, I M 6
a n d AS4, were c o m p a r e d for the P E E K c o m p o s i t e l a m i n a t e . All of the l a m i n a t e s were
f a b r i c a t e d in 304.8 by 304.8 m m (12 by 12 in.) panels. T h e eight l a m i n a t e c o n s t r u c t i o n s are
listed in T a b l e 1. T h e s e l a m i n a t e s are all quasiisotropic. T h e P E E K a n d B M I c o m p o s i t e
TABLE 2--Specimen dimensions and impact velocities.
Specimen No. Width, in. Thickness, in. Span, in. Impact velocity, m/s
Laminate ! (Toughened BMI--40 ply)

1 0.96 0.213 4 20.7
2 0.96 0.213 4 15.5
3 0.96 0.215 4 14.9
Laminate 2 (Toughened BMI--24 ply)
1 0.96 0.125 4 21.2
2 0.96 0.125 4 15.2
3 0.96 0.125 4 14.1
Laminate 3 (Brittle BMI----40ply)
1 0.96 0.216 4 21.0
2 0.96 0.220 4 27.2
3 0.96 0.213 4 15.0
4 0.96 0.210 4 15.2
5 0.96 0.214 4 15.2
Laminate 4 (Brittle BMI--24 ply)
1 0.96 0.126 4 21.0
2 0.96 0.128 4 15.0
3 0.96 0.129 4 15.0
4 0.96 0.131 4 14.9
Laminate 5 (PEEK--24 ply)
1 0.96 0.131 4 20.1
2 0.96 0.133 4 29.3
3 0.96 0.140 4 21.4
Laminate 6 (Torlon C--24 ply)
1 0.96 0.209 4 21.2
2 0.96 0.207 4 30.4
3 0.96 0.206 4 21.4
laminates consist of unidirectional plies approximately 0.134 62 mm (0.0053 in.) thick, while
the Torlon composite is a woven fabric.
Two types of specimens were used for the impact test: beam-like specimens, 25.4 by 177.8
mm (1 by 7 in.), which were also used in residual strength measurement; and 152.4 by 152.4
mm (6 by 6 in.) panels used for determining the geometrical effect.
Impact Testing
Impact tests were performed using a pressurized air gun to shoot a hard steel ball (22.225
mm [7/8 in.] diameter) into the specimen. The bali's velocity was measured by using two
light-emitting diode (LED) photovoltaic cell pairs. The LEDs were fixed 254 mm (10 in.)
apart, and the time interval between the interruption of the light beam was measured. The
impactor velocity was obtained using an experimentally obtained calibration curve. The
specimen was positioned with the help of a pointer so that the impactor would hit the marked
center point.
Several beamlike specimens from laminates 1 through 6 were first impacted at various
velocities for impact damage examination (see Table 2). Note that for these impact tests,
the span between the end-clamps of the specimen was 15.142 mm (4 in.). A steel ball 22.225
mm (7/8 in.) in diameter was used as impactor.
Laminates 7 and 8 were used for comparing different fibers (AS4 versus IM6) in PEEK
composites. To enhance bending effects, the span for laminates 7 and 8 was chosen to be
152.4 mm (6 in.).
Sectioning of Damaged Area

Specimens with impact-induced damage were inspected with an optical microscope at
various cross sections. This technique highlighted the degree of matrix cracking, delami-
nation, and fiber breakage. Matrix cracks and delaminations were also visualized using
enhanced X-ray photography. After the specimens were impacted, they were treated with
a dye which penetrated inside the laminate to fill the voids caused by microcracks and
delaminations. A limitation of this technique is that the penetrant must have an avenue to
the damaged areas.
Residual Strength Test

Additional beamlike specimens of laminates 1, 3, 5, and 6 were impacted and then
subjected to a three-point bending test. The span in this test was 76.2 mm (3 in.). The
bending strength was recorded.
ExperimentalResults--Impact Damage
In the following discussion of experimental results, the laminates are grouped according
to their overall thickness to avoid the geometrical effect due to different moments of inertia
of the specimen cross sections. Thus, the 24-ply Torlon laminate (laminate 6) is grouped
with the 40-ply BMI (laminates 1 and 3). X-ray radiographs are not presented because they
reveal the same impact damage information as the microphotographs from sectioning.
Laminates 2, 4, and 5
All these laminates have a [-45/0/+45/90]3, construction with a thickness of approxi-
mately 3.302 mm (0.13 in.). From the X-ray photographs as well as microphotographs of
FIG. 1--Microscopic photograph of longitudinal cross section (laminate 2, specimen 2,

toughened BMI, 24-ply, 15-m/s impact).
sections of the impacted specimens, it is clear that laminate 5 (PEEK) suffered the least
matrix cracking and delamination, followed by laminate 2 (toughened BMI). Laminate 4
(brittle BMI) seemed to be most susceptible to impact damage.
Figures 1 and 2 show the longitudinal sections of laminates 2 and 4, respectively, after
impact at 15 m/s. The brittle BMI composite laminate sustained appreciable matrix cracking
and delamination (see Fig. 2). At the same impact velocity, the toughened BMI composite
(Fig. 1) exhibited significantly less delamination.
At 21 m/s, both brittle and toughened BMI laminates suffered extensive matrix cracking
and delamination (see Figs. 3 and 4). A close examination of these microphotographs shows
that although delamination in laminate 2 (toughened BMI) was less severe than that in
laminate 4 (brittle BMI), fiber breakage in the 0~ plies was more extensive in laminate 2.
Similar results were found by Sun and Rechak [3] in a graphite/epoxy laminate using adhesive
layers to toughen the interfaces. They noted that while adhesive layers were quite effective
in suppressing delamination, they might cause more fiber damage at high impact velocities.

brittle BMI, 24-ply, 15-m/s impact).
FIG. 3--Microscopic photograph of longitudinal cross section (laminate 4. specimen 1,

brittle BML 24-ply, 21-m/s impact).
At 21 m/s impact velocity, the A S 4 / P E E K laminate (laminate 5) displayed no damage.

When impact velocity was increased to about 29 m/s, extensive matrix cracking and delam-
ination appeared as shown in Fig. 5. Permanent indentation on the top surface left by the
impact is also visible. Delamination is more severe near the midplane of the laminate,
possibly due to more severe transverse shear stress. Excessive transverse shear stress could
produce matrix cracks, which in turn could induce delamination. Comparing laminates 4
and 5, impacted at 15 m/s and 29 m/s, respectively, it is evident that laminate 5 suffered
less matrix cracking and delamination than laminate 4.
Laminates 1, 3, and 6
These laminates are approximately 5.3 mm (0.21 in.) thick. Note that laminate 6 consists
of woven AS4 graphite fabric and Torlon matrix. In general, it was found that these thicker
laminates exhibited considerably less impact damage than the corresponding 24-plied lam-
inates at the same impact velocities. This behavior can be attributed to the expected reduction
in bending and transverse shear stresses in thicker laminates.

FIG. 5--Microscopic photograph of" longitudinal cross section (laminate 5, specimen 2,

PEEK, 24-ply, 29-m/s impact).
A comparison of toughened (laminate 1) and brittle (laminate 3) laminates at an impact

velocity of about 21 m/s is illustrated in Figs. 6 and 7. As with the corresponding thinner
laminates, the toughened BMI composite has higher resistance to matrix cracking and
delamination. At this velocity, the Torlon composite laminate (laminate 6) showed no visible
damage.
While the Torlon laminate exhibited greater damage resistance at lower impact velocities,
it suffered massive fiber breakage at higher velocities (30 m/s). As shown in Fig. 8, a major
bending crack propagated through many laminae. Matrix cracks and delamination were also
present, although they were highly localized. This behavior is likely due to the high toughness
FIG. 6--Microscopic photograph of longitudinal cross section (hlrninate 1..wecimen 1.


brittle BMI, 40-ply, 21-m/s impact).
of the Torlon matrix as well as the fabric construction of the laminates. The wavy nature
of the interface in this laminate might also impede delamination crack growth.
IM6 and AS4 Fiber Comparison

The effect of changing fiber from AS4 to IM6 was investigated using the PEEK laminate,
laminates 7 and 8. Impact velocities ranged from 22 m/s to 30 m/s. The span between
FIG, 8--Microscopic photograph of longitudinal cross section (laminate 6, spec#nen 2,

AS4/ Torlon, 24-ply, 30-m/s impact).
supports was 152.4 mm (6 in.). Impact specimens were X-ray photographed and sectioned
to observe impact damage. It is evident from the results that neither fiber system shows a
distinct advantage in impact damage tolerance.
Plate Effect
All of the previous comparisons were made using beamlike specimens. To investigate size
effects on impact damage, 152.4 by 152.4 mm (6 by 6 in.) plate panels were impacted. The
plate specimen was clamped along two opposite edges leaving the other two edges free. The
distance between the clamped supports was 101.6 mm (4 in.). Laminates 1, 3, 5, and 6 were
used in this investigation.
Except for the degree of damage, the plate specimens did not differ from beam specimens
in failure modes or impact tolerance properties. The greater bending stiffness in the plate
specimen was able to reduce the bending and transverse shear stresses and, consequently,
the impact damage. It is also evident that PEEK and Torlon laminates have superior impact
damage resistance, as was found in the beam specimens.
Residual Strength
The postimpact load-carrying capability of a composite laminate is of prime concern to
the design engineer. After a tool-drop type accident where no damage is visible from the
surface, the structure is still expected to carry the full spectrum of loading.
For this study, virgin and residual flexural strengths were measured using the three-point
bending test. This test was chosen over the popular compression test to avoid additional
structural instability effects due to the residual deflection left in the impacted specimen.
Four matrix material systems (laminates 1, 3, 5, and 6) were evaluated. The residual load
versus impact velocity is plotted in Fig. 9. Note that laminates 1, 3, and 6 have almost the
same initial strength because they have the same thickness. The PEEK laminate (laminate
5) is thinner, and its initial strength is lower.
In all cases the residual strength decreased as the impact velocity increased. It is apparent
from Fig. 9 that the PEEK- and Torlon-based composites (laminates 5 and 6) exhibit a
"threshold velocity" before which little reduction in residual strength occurs. At velocities
greater than the threshold value, these composites quickly lose their load-carrying capacity.
This behavior of a "'threshold velocity" was also observed during microphotographic damage
assessment. Due to the high toughness of PEEK and Torlom matrix cracking and delami-
nation.appeared in the beamlike specimens only at impact velocities of about 25 m/s or
more in laminates 5 and 6. This explains why there is little reduction in strength in these
laminates prior to this velocity. Beyond this point, matrix damage occurred and substantial
fiber breakage followed. This explains the sudden drop of strength in laminates 5 and 6
when impacted at velocities higher than 25 m/s.
Both toughened and brittle BMI composite laminates exhibited a more gradual decline
in residual strength. Delamination in these composites was induced by both shear and
bending cracks at low-impact velocities. The extent of delamination progresses as the impact
velocity increases, thus the gradual reduction of residual strength. It is likely that this contrast
in behavior is tied to the relative brittleness of the epoxy-based matrix material. It is also
because of this brittleness that matrix cracks easily branch out into delamination to dissipate
impact energy rather than penetrating and breaking the fiber in adjacent layers.
Recently, Sun and Rechak [4] performed residual strength tests on graphite/epoxy lam-
inates with and without adhesive layers. Adhesive layers were found to be effective in
suppressing delamination. At high-impact velocities, the laminate containing adhesive layers
1250
n LAMINATE NO. 6
Q- LAMINATE NO. 5
I000 L
3 L~ LAMINATE NO. 3
O LAMINATE NO. 1
0 0
m~ 75(
r
>
500
o
250 t ,
O r
[]
o
o.oo Io.oo 20.00 30.00 40.00 50.00
IMPACTOR VELOCITY (M I SEC)
FIG. 9--Residual load relations for a three-point bend test with a 3-in. span.
was found to have suffered a greater drop in both tensile strength and flexural strength due
to greater fiber breakage.
From the present results and those obtained in Refs 3 and 4, we may conclude that a
tough matrix may provide excellent impact resistance properties at low-impact velocities.
However, beyond a certain threshold velocity, the use of tough matrix materials may result
in more laminate tensile and flexural strength reduction than the use of brittle matrix
materials.
Dynamic Response Behavior

Impact damage is closely related to the dynamic stress field produced by dynamic contact
between the impactor and the laminate. Contact force depends on contact rigidity as well
as structural flexibility. Thus, contact behavior and dynamic impact response of the laminate
should also be included in the comparison of impact resistance properties of composite
laminates.
Contact B e h a v i o r
Contact laws between the 7/8-in.-diameter steel ball and the composites were experi-
mentally established using the procedure proposed by Yang and Sun [5]. For loading, the
contact law is a s s u m e d to be
F = koL1'5 (1)
where
F = c o n t a c t force,
k -- c o n t a c t rigidity coefficient, a n d
-- i n d e n t a t i o n .
F o r u n l o a d i n g , we have
[0~ -- 0/.o] 25
(2)
F = F,. L'a,,, _ eloJ
where
F m = c o n t a c t force at u n l o a d i n g ,
am = i n d e n t a t i o n at u n l o a d i n g , a n d
so = p e r m a n e n t i n d e n t a t i o n after c o m p l e t e unloading.
In E q 1 a n d E q 2, the two coefficients k a n d ao are to be d e t e r m i n e d e x p e r i m e n t a l l y .
4800
1
K = 1.12554E+07
4000
3200
/
2400
lag
1500
G
0 .800 1.500 2.400 3.200 4.000 4.800 5,600
ALPHA (0.00ICM)
FIG. lO--Curve-fit of the experimental static indentation data during loading ( - - = curve
fit) for laminate 1.
TABLE 3--Contact coefficients.
k o~c,
Laminate No. (lbf/in. ~) (N/cm tS) (in. X 10 -3) (cm x 10 3)
1 1.125 54 x 107 1.2368 x 107 1.54 3.91

2 1.098 63 x 107 1.2073 x 107 0.77 1.96
3 1.248 55 x 107 1.3720 x 107 1.26 3.20
4 1.207 55 • 107 1.3270 x 107 0.996 2.53
5 0.859 4 x 107 0.9439 x 107 0.755 1.92
6 1.128 46 x 107 1.2400 x 107 1.047 1.047
T h e 25.4-mm (1-in.) wide s p e c i m e n was c l a m p e d at b o t h ends leaving a 5 0 . 8 - m m (2-in.)

gauge length for the i n d e n t a t i o n test. T h r e e loadings at t h r e e different locations were c a r r i e d
out. T h e loading curve for l a m i n a t e 1 p l o t t e d in t e r m s of F versus a is s h o w n in Fig. 10.
T h e n these d a t a were used to o b t a i n the c o n t a c t rigidity coefficient k in E q 1 using a least
s q u a r e s fit. T h e p o w e r law fits the data very well. T h e values of k thus o b t a i n e d for l a m i n a t e s
1 t h r o u g h 6 are listed in T a b l e 3.
4000
i
$
s
4000
/
3200
/
/
I
2400
,00o -
80o ~y
.000 ,.600 ,.,06 .,00 ,.oo0 ,.000 5.600

ALPHA (O.O01CMI
FIG. 11--Curve-fit of the experimental static indentation data during unloading ( - - - -
curve fit) for laminate 1.
Comparing the values ofk for all the composite systems, it is seen that the PEEK composite
(laminate 5) has significantly lower contact rigidity. This indicates that for a given contact
force, the resulting indentation in the PEEK composite would be larger, yielding a larger
contact area, and, therefore, a lower contact pressure. A larger contact area with lower
pressure will reduce the transverse shear stress concentration and thus minimize local matrix
cracking. The effect of contact area on impact damage has been studied by Sun and Rechak
[3]. For practical purposes, the toughened and brittle BMI composites (laminates 1 through
4) can be regarded as having the same contact rigidity.
Unloading tests were performed with three unloading levels. For each laminate specimen,
all unloadings were done at the same location.
Because the value of ct depends on the unloading level, theoretically, many unloading
levels have to be considered. In Ref 5, a relation between loading and unloading coefficients
was found, that is,
k
- = ac, (3)
S
where
F~
s = (4)
is the unloading coefficient for the unloading from F,. Thus, using Eq 3, an unloading curve
750 1
v = 3 M/S
5OO
250
r\.
^
V
-250
-500
0 500 970
TIME (Micro-Sec)
FIG. 12--Dynamic strain response during low-velocity impact of laminate l at gauge 2
(l-in. strip, toughened BMI, 40-plies).
is necessary to determine the value of a,. Using this relation and Eqs 1 through 3, we obtain
Ot...Zo )2/5
= 1 - a<, (5)
OLm \OLin I
from which % can be calculated for any given value of c~,,. Also, we set
% = 0 if a,,~% (6)
The experimentally generated loading and unloading curves together with Eq 3 were used
to determine ct,. The results are presented in Table 3. These values were used in Eq 2 to
predict the unloading path. Comparison of the experimental and theoretical unloading curves
for laminate 1 is shown in Fig. 11. Good agreement is evident.
The value of a<, is the indentation beyond which permanent deformation results. The
smaller the value, the larger the expected crater due to impact. Among the four composite
systems tested, it is noted that the PEEK composite has the lowest c~,.
Dynamic Strain Response

Dynamic strain histories at two locations produced by impact loading were recorded using
strain gauges. Gauge 1 was placed opposite the center of impact, and gauge 2 was located
25.4 mm (1 in.) away from gauge 1. For low-velocity impacts (3 m/s), a pendulum impact
set up was used. For high-velocity impacts, a compressed air gun was used.
Figures 12 through 14 show the strain history responses for an impact velocity of 3 m/s
750
1
V = 3 MIS
Pi A
500
V \
~-_,,~250
:E
i o
/
i; X'\
-250 V
-5000
500 970
TiME (Micro.Sec)
FIG. 13--Dynamic strain response during low-velocity impact of laminate 3 at gauge 2
(1-in. strip, brittle BMI, 40 plies).
DAN-JUMBO ET AL. ON iMPACT DAMAGE CHARACTERISTICS 369
750
1
V = 3 MIS
500 1
250
-250
xf F \/
V
-500 t
O 500 970
TIME IMicro.Sec)
(l-in. strip, Torlon, 24 plies).
at gauge location 2 in laminates 1, 3, and 6, respectively. Note that these three laminates
have almost identical thicknesses. From the amplitude and the oscillatory behaviors of these
response curves, one can conclude that the Torlon composite (laminate 6) has the lowest
bending stiffness. The responses of both BMI composite laminates are quite similar except
that the toughened BMI composite exhibits more rapid decay in amplitude. This can be
taken as an indication that toughened BMI composite may have a greater damping property.
Dynamic Failure Strain

During high-velocity impact, the strain gauge right under the impact point (gauge 1) broke
due to matrix cracking. The exact strain level at which the matrix cracked in the bottom
ply cannot be determined from these data. However, a range of failure strains can be
estimated. From Figs. 15 and 16 the matrix failure strains in both the toughened and brittle
BMI (laminate 3) were estimated to be approximately 0.75%. This indicates that toughened
BMI performs no better than brittle BMI as far as resistance to bending cracks is concerned.
On the other hand, Fig. 17 indicates that the Torlon-based composite did not crack until
2% strain.
Conclusions
An experimental investigation of impact tolerance properties of four composite material
systems was carried out. Following are the major findings:
1. Toughened BMI composite is superior to brittle BMI composite in impact damage
&- L
V = 15.4 M I S
12,000
10.000
,/
6,000
9,-r.
m -...R,nnn. ,,
z
2,000
/-,
o / ,- f"J ~
-2,000
V
-4,000.
0 500
TIME IMicro-Sec) 971
FIG. 15--Dynamic strain response during low-veloci O, impact of laminate 1 at gauge 1

(1-in. strip, toughened BMI, 40 plies).
25,000
rI V l = 15.1 MIS
,
20,000 , I
15,000
--..r
2-':
10,000
I
sooof,j L
0 500 970
TIME [Micro-Sec)
(l-in. strip, brittle BMI, 40 plies).
40~
-.
,o.,,,. j J
-- 1
-20,00~
500 970
TIME (Micro-Sac]
FIG. 17--Dynamic strain response during high-velocity impact of laminate 6 at gauge l
(l-in. strip, Torlon, 24 plies).
resistance. The former has greater capability of retaining flexural strength than the latter
for high-velocity impacts.
2. Both P E E K and Torlon composites remain free from matrix cracking and delamination
until a threshold impact velocity is reached, beyond which their residual flexural strengths
drop suddenly due to extensive fiber breakage. Thus, for high-velocity impacts, tougher
PEEK and Torlon composites may be no more impact damage resistant than the BMI
composites.
3. P E E K composite has a lower contact rigidity and is much more susceptible to per-
manent indentation than the other composite systems.
4. Partially due to its fabric construction, Torlon composite laminate has a much higher
dynamic failure strain than either toughened or brittle BMI laminate.
A ckno wledgments
This work was supported by an I R A D program with General Dynamics, Fort Worth
Division. The first author extends his appreciation to R. J. Stout of General Dynamics for
his encouragement in this research effort.
References
[11 Dorey, G., Bishop, S. M., and Custis, E T., "On the Impact Performance of Carbon Fiber Laminates
with Epoxy and PEEK Matrices," Composites Science and Technology, Vol. 23, 1985, pp. 221-237.
[2] Joshi, S. P. and Sun, C. T., "Impact Induced Fracture in a Laminated Composite," Journal of
Composite Materials, Vol. 19, Jan. 1985, pp. 51-66.
[3] Sun, C. T. and Rechak, S., "Effect of Adhesive Layers on Impact Damage in Composite Laminates,"
in Composite Materials: Testing and Design (Eighth Conference), ASTM STP 972, I. D. Whitcomb,
[4] Sun, C. T. and Rechak, S., "'Effect of Toughened Interfaces on Impact Residual Strength of a
Graphite/Epoxy Laminate," to be published.
[5] Yang, S. H. and Sun, C. T., "Indentation Law for Composite Laminates," in Composite Materials:
Testing and Design (Sixth Conference), ASTM STP 787, I. M. Daniel, Ed., American Society for
A. P. Christoforou, 1 Stephen R. Swanson, ~ and S. W. Beckwith 2
Lateral Impact of Composite Cylinders

REFERENCE: Christoforou, A. P., Swanson, S. R., and Beckwith, S. W., "Lateral Impact
of Composite Cylinders," Composite Materials: Fatigue and Fracture, Second Volume, ASTM
STP 1012, Paul A. Lagace, Ed., American Society for Testing and Materials. Philadelphia,
1989, pp. 373-386.
ABSTRACT: An analytical study was conducted to explore and understand the mechanics of
the lateral low-velocity impact of composite cylinders by a heavy mass. Two simple models
are presented, which when compared with the experimental data demonstrate that the fun-
damental features of the impact event are represented. Although the results are preliminary,
a failure model based on linear laminated shell theory and linear fracture mechanics gave a
reasonable representation of the observed impact damage. A nonlinear ring model gave good
insight into the physics of the impact event in the presence of delaminations and the resulting
nonlinear effects.
KEY WORDS: composite pressure vessels, impact damage, residual strength, delamination,
IM7 carbon fiber
In contrast to the number of studies dealing with impact in flat laminate geometries,
relatively little has been published on impact of composite cylinders. However, the potential
for impact damage exists in pressure vessels and other structures. To address the need to
understand impact of curved geometries, we have carried out a combined experimental and
analytical study using instrumented drop-weight impact of cylindrical tubes. Pressure testing
of these cylinders was then used to determine the extent of damage.
Studies of impact Ioadings of advanced composites have been carried out by a number
of investigators [1-9]. It has been shown in general that composites can be sensitive to
impacts. The resulting damage can consist of a number of modes, including matrix cracking,
delamination, and fiber breakage. The damage may or may not be visible on the surface.
The degree of impairment of subsequent function of the laminate depends on both the
nature of the damage and the subsequent loading. For example, compression after impact
has been shown to be sensitive to delamination and matrix damage due to the loss of structural
stiffness [2-4, 7]. Attempts have been made to model various aspects of the impact process.
The structural response of flat plates to impacts of various velocities has been measured
[10-12]. The results show that there is a range of conditions in which the mass of the
impactor is large with respect to the mass of the laminate, so that the structural response
can be modeled as a single degree of freedom spring-mass system [13]. At higher velocities,
a more involved treatment is necessary because other structural response modes are excited
[101.
Nonlinear response of laminates to impact, which appears to be typical, complicates the
analysis of the structural response. In tests of flat laminates, the resulting displacements are
often large enough to produce membrane action as well as bending. Delamination and
University of Utah, Salt Lake City, UT 84112.

2 Hercules Aerospace Co., Magna, UT 84044.
373
Downloaded/printed
Copyright9 byby
softening due to matrix cracking are further sources of nonlinearity. Attempts have been
made to include these effects into response models [13].
High local contact stresses are produced in impact. The force-deformation at the inden-
tation due to the contact has been shown to follow a law similar to that for isotropic materials
[14] obtained from a Hertz analysis. These stresses may be high enough to damage fibers
and/or matrix without additional bending effects. It is likely that cutting of the fibers at the
surface may depend on the shape of the impactor.
Relatively little has been reported on impact of curved geometries. Clearly the curvature
adds to the stiffness of the structure and thus affects the impact force. An experimental
study of impact of pressure vessels has been presented [15]. One of the interesting findings
of that study was that the relationship between strength loss due to impact damage and the
geometric parameters of the vessel is quite complex. For example, increasing the wall
thickness of the vessel actually caused an increase in relative strength loss at constant impact
energy over a certain range of geometries.
In previous work [16] we have carried out impact tests on cylinders made of IM7 carbon
fiber, followed by pressure tests to determine strength loss. Additionally, delaminations
resulting from the impact were measured with C-scan. The impacts were produced by a
relatively heavy drop weight impactor (mass of 5.03 kg) at impact energies up to 47.5 J (35
ft-lb). In addition to the impacts of empty cylinders, impacts were applied to cylinders
supported by either an RTV rubber core or an aluminum tube core. These latter tests were
performed to attempt to simulate actual service conditions of rocket motor cases. After
impact, the cylinders were modified by end fittings and tested in the University of Utah
biaxial test facility. This apparatus and specimen were designed to minimize the stress
concentrations at the ends of the specimen [17] so that accurate measurements of the strength
loss due to impact could be made. The biaxial loadings were intended to represent pressure
loading in the cylindrical section of a rocket motor case, which is often the critical loading
that drives the design. Impairment of the tension-load-carrying ability of the cylindrical
section of a rocket motor case by accidental damage due to impact is of significant interest.
The results shown in Fig. 1 indicate that loss in strength of up to 40% or more was achieved
over the range of impacts studied, with all of the three internal support conditions used.
Clearly it is necessary to be able to relate these impact tests to actual service conditions. It
is believed that more detailed understanding of the mechanics of the impact event is required
to be able to do this. Thus the present study was undertaken to add to this understanding.
In the following we present the results of models of the impact event. These models are
compared with the experimental data and demonstrate that the fundamental features of the
impact are represented. However, the importance of nonlinear effects is quite pronounced.
Thus, preliminary results of a simplified nonlinear model are also presented.
Experiments
The experimental results shown in Fig. 1 were obtained from impacts of cylindrical tubes
made of IM7/55A carbon/epoxy. The cylinders were filament wound with a [90/-2212T
layup (indicated from the inside to the outside of the cylinder wall) with a total wall thickness
of 1.52 mm (0.060 in.). The cylinders were 96.5 mm (3.80 in.) inside diameter by 420 mm
(16.5 in.) in length. The elastic properties of this composite material system are given in
Table 1.
The cylinders were supported in a wooden cradle, as indicated in the inset in Fig. 2, and
subjected to lateral impact by a 5.0-kg (11.1-1b) falling weight. The instrumented contacting
tup was hemispherical in shape, with a diameter of 15.9 mm (0.625 in.). More details of
the experiments are given in Ref 16.
CHRISTOFOROU ET AL. ON LATERAL IMPACT OF COMPOSITE CYLINDERS 375
1200 " 180 I ~ m I I t I

tO00 - 160 / ~ A l u m i n u m Support -
140
.~ 800 - 12o
oo
600 -
~80 i
400 . 60 ' ~RTV Support
40 -
200 -
ZO -
0 0 I m J I m i m
S lO 15 20 25 30 35
(ft-lbs)
I. I I, I I
0 I0 20 30 40
Impact Energy, (Joules)
FIG. l--Effect of the internal support conditions and impact energy on the hoop stress at
failurefor carbon/epoxycy~nders.
Static Linear Model
In previous work [16] it was shown that it is possible to calculate the impact response for
the empty cylinders using a simple spring-mass model, for the relatively heavy mass used
in the impact experiments. Note that the impact mass of 5.0 kg is approximately 14 times
that of the entire cylinder. The inputs to this model are the mass of the heavy impactor and
the stiffness of the cylinder acting as a nonlinear spring of negligible mass. The impact force
and deflection can be calculated successfully using the elementary equations of dynamics
for a single degree of freedom system.
The stiffness used in the model was measured by statically loading the cylinder in a test
machine, while measuring the force and deflection. The stiffness was also estimated by cross-
plotting the force versus time and deflection versus time measured during the impact, so as
to give force versus deflection directly from the impact measurements. This would correspond
to the cylinder stiffness only if the cylinder mass were negligible, as assumed in the single
degree of freedom model. A comparison of the measured static and dynamic force deflection
response for the cylinders is shown in Fig. 2. The results shown in Fig. 2 indicate that the
present problem of low-velocity impact of a cylinder by a heavy impactor can be treated as
an equivalent static problem of a concentrated force acting on the cylinder.
TABLE 1--Material properties for IM7/55A carbon/epoxy.

Fiber-direction Young's modulus, E, 169 GPa (24.6 Msi)
Transverse Young's modulus, E22 6.72 GPa (0.975 Msi)
In-plane shear modulus, Gt2 5.26 GPa (0.763 Msi)
Major Poisson's ratio, vt2 0.266
4 900 I I I I I I I I
800 - - S t a t i c Loading
- i x 20.8 doules (15.31 ft-lbs)~ Impact "
0 34.4 Joules (25.38 ft-lbs)~Experiment
700 --V 19.3 Joules (14.25 ft-lbs)(
3 1:3 47.5 Joules (35.00 ft-lbs)J
600
i 50o
-- ~70
2 .,~
,~400
300
1 - 200
100
0 0 ~ i 13 l i , i i
0 .I .? . .4 .5 .6 .7 .8 .9
(inJ
l I I I I
0 0.5 1.0 1.5 2.0
Deflection, (cm)
FIG. 2--Comparison of measured static and dynamic force deflection response for an empty
carbon~epoxy cylinder (in cradle).
The next task in the model development was to calculate the static force-deflection re-
sponse as well as to estimate the stresses in the cylinder under the equivalent static force.
To do this we used a solution to the problem of a linear orthotropic cylindrical shell subjected
to asymmetric concentrated load presented by Kliger, Forristall, and Vinson [18] and by
Vinson and Chou [19]. The solution, which includes the effects of transverse shear defor-
mation, is based on an expansion of the displacements, rotations, and localized loading in
terms of a double Fourier series which satisfies the end boundary conditions of simple
support. The radial pressure load is assumed to be applied over a small rectangular area.
A computer program to implement this approach was presented in Ref 18 and was used in
our analysis.
We attempted to compare the computed stiffness from the Fourier series program with
experiments in which the concentrated load was applied on each side of the cylinder in a
symmetric diametrically opposed manner. This was intended to more accurately represent
the loading in the impact experiments because the cradle (shown in the inset to Fig. 2)
tended to provide support over the entire bottom half of the cylinder. The symmetric loading
analysis can be considered to have zero displacement at the horizontal symmetry plane. The
above program was modified to handle this case of a cylinder loaded by two centrally located
and diametrically opposed concentrated loads. The load can be expanded in a double Fourier
series as
F(x,O) = ~, ~ Fro. cos (nO) sin (m'rrx/1) (1)

n=0 m=l
The coefficients Fro, of the Fourier series are solved for by assuming that the load can be
expressed as
1~ < • < 1_~, -t~l < 0 < ~1, and ~r + qh < 0 < rr + qh
otherwise (2)
where
Fo = F/2Rqh(12 - 11) (3)

The size of the contact zone, 2R~1 x (12 - 11), was estimated for our geometry to be
approximately 3 • 3 mm (0.125 x 0.125 in.). The geometry, nomenclature, and boundary
conditions are indicated in Fig. 3.
Force Deflection Response

In our initial comparisons of computation and experiment we did not fully appreciate the
importance of modeling the "simple support" boundary condition assumed in the analysis.
The importance of this boundary condition became apparent in the initial comparison of
the Fourier series analysis with experimental data obtained on a specimen with free ends.
The finite length of the specimen, 420 mm (16.5 in.), apparently creates nonzero deflections
and out of roundness at the ends which violate the simple support boundary condition of
the analysis. As a result the experimentally determined stiffness was lower than the analytical
prediction by 40%. An additional experiment was performed with short plugs inserted at
the ends in an attempt to impose the simple support condition on the specimen. The measured
stiffness was increased relative to the free end cylinder, but still was about 20% lower than
the analysis.
To further investigate the effect of the boundary conditions on lateral stiffness, two
t /_.... Supportconditions v - w= =
v=w-~,o-O
FIG. 3--Geometry of analyzed cylindrical shell.
3.5 - 800 , , , , , ,
3.0 - 700
2.5 -
600 ~OrU~iet Series
i
"F
500
Z.O-
v
2 400
1.5
o 300
1.0
zoo
O. 5 - 100
O _ O
0 O.l 0.2 0.3 0.4 0.5 0.6 0.7
(in)
1 I I I
0 0.5 1.0 1.5
Deflection, (cm)
FIG. 4--Force deflection response for a long aluminum cylinder loaded by two centrally
located and diametrically opposed concentrated forces.
aluminum tubes of different lengths were tested. First, a 42-cm (16.5-in.)-long aluminum
tube was tested and compared to the Fourier series analysis. As with the composite tube
with open ends, the experiment again was in poor agreement with the analysis. Next, a 183-
cm (72-in.)-long aluminum tube was tested. The results are shown in Fig. 4. The initial
slope of the experimental force deflection curve for this long cylinder matches the Fourier
series solution quite well, indicating again the importance of the end condition on lateral
stiffness for relatively short cylinders. During the test, significant yielding occurred in the
aluminum tube as evidenced by the permanent deformation seen in Fig. 4. This permanent
deformation was not observed in the composite tubes.
In conclusion, the Fourier series analysis correlates reasonably well with the linear portion
of the force deflection response of the cylinder, if the simple support conditions at the ends
assumed in the analysis are also used in the experiment. However, the agreement between
analysis and experiment appeared to be better for the isotropic aluminum cylinder than for
the orthotropic composite cylinder. It should be pointed out that the analysis is based on a
single orthotropic material. Thus the coupling terms in the D matrix and the nonzero B
matrix for the nonsymmetric laminate could not be represented in the analysis. As a final
point, the nonlinearity of the composite cylinder response is significant, as can be seen in
the measurements shown in Fig. 2.
Failure Analysis
The laminate strains were calculated for the cylinder under the equivalent static load by
using the Fourier series analysis. As in the usual laminated plate theory, the laminate strains
were rotated into the fiber direction for each lamina of the [90/+--2212r cylinder laminate.
CHRISTOFOROU ET AL. O N LATERAL IMPACT OF C O M P O S I T E CYLINDERS 379
TABLE 2--Failure properties for IM7/55A carbon~epoxy.
Maximum fiber strain allowable (tension) 1.8%

Maximum fiber strain allowable (compression) - 1.0%
Interlaminar shear strength, "re 41.3 MPa (5990 psi)
These strains were then used to estimate the extent of fiber damage from the impact. A
maximum fiber strain failure criterion was used in accordance with previous results on
laminate failure [17]. The fiber strain allowable for the IM7/55A carbon/epoxy system used
was taken as 1.8% in tension and - 1.0% in compression, based on coupon data. The failure
values for this composite system are listed in Table 2. We characterized the size of the
damage (broken fiber) zone simply by the extent in the axial direction for which strains at
the inner (90 ~ lamina were predicted to exceed the allowable values. Figure 5 shows the
predicted length of the fiber damage zone as a function of impact energy. Because the hoop
stress in the specimens was twice the axial stress (pressure vessel loading), a crack perpen-
dicular to the 90 ~ fiber should affect the strength the most. Further, the orientation of the
crack makes the failure mechanism insensitive to the axial stress [20,21].
The square root of energy has been used to plot the damage relationships rather than the
energy itself. From the basic assumption of an equivalent quasistatic loading used for the
present impact experiments and analysis, the applied force for a linear system is proportional
to the square root of the energy. It is believed that the resulting damage would thus depend
on the applied force rather than the energy directly.
A fracture mechanics approach was used to estimate the amount of strength degradation
due to the impact. In this approach we idealized the predicted broken fiber region as a crack
of length 2c. From linear fracture mechanics the failure stress of a wide sheet of isotropic
5 2.51 ! I ! I I
2.01
4 a
3 1.51
z . "~.oI
I
m
1 o.51
o. ol I
0 ] 2 3 4 $ 6
(ft-lbs) 'j
i J I I I , I I
0 1 2 3 4 5 6
~-. (Ooules)89
FIG. 5--Predicted fiber damage zone length as a function of impact energy.
TABLE 3--Failure properties for AS4/3501-6 carbon~epoxy.

Maximum fiber strain allowable (tension) 1.4%
Fracture toughness 714.4 GPa ~mcm (65 200 psi ~mln.)
material containing a crack is given by Ref 22
trt = K,c(~rc) v2 (4)

where
K~c = the fracture toughness of the material, and
c = the crack half-length.
Due to the lack of an experimental fracture toughness Ktc for IM7/55A, an estimate was
calculated by using the fracture toughness for AS4/3501-6 listed in Table 3. This value,
obtained from tests on a plate with a through crack, was modified to give an estimate of
the appropriate value for the cylinder tests according to the following formula:
K,c(IM7) = K,c(AS4) 9 -~/. ~(IM7)

trt ~,(AS4) (5)
where
Gqt = the stress in the fibers aligned with the loading direction for an unnotched specimen
loaded by the same far field failure stress ~I as for the notched specimen, and
~, = the respective strain allowable for the two fiber systems.
Equation 5 is based on the failure of a critical ply ahead of the crack, as suggested by Poe
[23].
The fracture toughness value of 1534 GPa V~cm (140 000 psi V~ln.) calculated by Eq 5
was then applied to the cylinder by again identifying the critical ply, calculating the far field
stress for that ply by the usual lamination theory and substituting in Eq 4.
The degraded strengths predicted by Eq 4 are plotted in Fig. 6 along with the experimental
data. As can be seen, the models predict a sharp drop of strength with increasing energy
(and thus increasing contact force). Because of the scatter in the experimental values of
strength retention, it is difficult to be more definite about the agreement between the model
and experiment.
Nonlinear Effects
Experimental evidence from our testing demonstrated that thin carbon/epoxy cylinders
under concentrated lateral loads deflect in a very nonlinear relationship with the applied
load. C-scan observations and sectioning of the specimens showed that large delaminations
occur under this type of loading, and furthermore the delaminations occur at the midsection
of the cylinder wall.
Due to the complexity of the shell theory, which will be complicated even further by the
presence of delamination, a simpler analysis was used to gain some insight into the physics
of the problem. To do this, the structural response of an isotropic ring with delamination
subjected to a concentrated load was studied using an energy method. The response of the
ring was calculated by a strength of materials approach.
1,25 I I 1 1 I I
1.o
,3 o o
~ o
5 o.75
~ 0.5
Predicted . _
~ 0 0
0
0
O ExperimentalData
~. 0.25
I I ~ I I I
O I 2 4 5 6
(ft-lbs) 89
t i J J I I I I
0 l Z 3 4 5 6 7
w~, (Joules)89
FIG. 6--Effect of impact energy on the residual strength ratio for empty carbon~epoxy
cylinders.
For the ring shown in Fig. 7, the bending strain energy of a ring with a delamination of
size 0~ is given by
~,,-' M~
U= 2 ~o ": - - R
M~dO + 2 ,,2 ~ RdO (6)
2.0 I i I I I
0 C-Scanof Cylinders After Impact
1.5 Ring Model ~ Ir .

Prediction
o" 1.0
0
8
0.5
2 4 6 8 IO 12
(U/Uo)89
FIG. 7--Effect of impact energy on the delamination size in hoop direction for carbon/
epoxy cylinders.
where Mo is the bending moment in the ring which is given by
Mo =
_~
sin0 +
6 cos ~- -
30~ + 7r
8) (7)
Equation 6 was derived by assuming that the load in the delaminated region was carried
by the two sublaminates and the magnitude of the force in each section was assumed to be
half of the force carried in the total ring. The assumption makes the bending stiffness of
the delaminated region of the ring four times smaller than that of the undelaminated ring.
This type of assumption has been used previously in the development of the end-notched
flexure (ENF) specimen for the characterization of mode II interlaminar fracture toughness,
as presented in Refs 24 and 25.
Integrating Eq 6 yields
U = "-~ + ~ 01 - ~ sinOl + 4 - 3 cos (8)
where c~ is defined as
Oi
6 cos ~- - 8
a - (9)
30, + 7r
Using Castigliano's theorem, the deflection of the ring under the load can be calculated
as
OU
= -- (10)
OF
which yields
5 = ~ + ~0t - ~sin01 + 4- 3cos et (11)
The compliance, C, of the ring is given by the relation
C = ~- (12)
F
The interlaminar fracture toughness is characterized by the strain energy release rate, G,
which is related to the change in compliance with crack extension, or in our case, delami-
nation extension
F 2 dC
G - (13)
2W d(RO 0
where W = width.
CHRtSTOFOROU ET AL. ON LATERAL IMPACT OF COMPOSITE CYLINDERS 383
Combining Eqs 11, 12, and 13 yields
G = 4E----~ - ~ cos 0, + 3a sin ~- + ~ a 2 (14)
By setting G equal to G~ (G~ = 9.63 N/cm [5.5 lb/in.] 3, Eq 14 becomes a criterion for
delamination crack propagation. For crack initiation, the interlaminar shear stress allowable
is used. Using simple beam theory, the critical load required to start delamination is given
by
F~ 4
= g t-'r~ (15)
where r~ is defined as the interlaminar shear stress allowable listed in Table 2.

Using Eqs 14 and 15 in conjunction with Eq 8, we were able to calculate the delamination
size in the rings as a function of the applied energy. A comparison of the calculations with
the C-scan data of the cylinders after impact is shown in Fig. 7. It can be seen that the
correlation is reasonable, indicating that the simple model used may represent the physics
of the problem if not the quantitative details.
Finally, by using Eqs 14 and 15 in conjunction with Eq 11, the nonlinear force deflection
response of the delaminated ring can be calculated. A comparison of the predicted nonlinear
force deflection response of the ring and the measured force deflection response of the
cylinders is shown in Fig. 8. One major conclusion derived from Fig. 8 is that delamination
0.7 I _
, l i i I l
1.75 "
K0 - initial stiffness
v
1.5 0.6 K0 = 622 N/cm . . . . Ring model
o
(355.4 1b/in)
o K0 = 4028 N/cm 0 Impact data
1.25 - 0.5 (2300 lb/in) (in cradle)
w-
w- K0 - 4623 N/cm ~ Pinched cylinder
0.4 (2640 l b / i n )
1.0
r
e 0.75 - 0.3
,,o,, -'O~
ssO
u. 0.5 o.z . .~o
0.25 - 0.1 ,16-6
o- o ~ l
0.1 0.2 0.3 0.4 0.5 0.6 0.7
(~n)
i I I I i I I l
0 0.25 0.5 0.75 1.0 1.25 1.5 1.75
Deflection. (cm)
FIG. 8--Force deflection response for carbon~epoxy cylinders under concentrated forces.
3 Tadaro, T. J., private communication with Hercules Aerospace Co., Magna, UT, February 1987.
is a first-order cause of the nonlinearity seen in the cylinder subjected to both impact and
static loading.
Discussion
The major thrust of the present work is to develop models useful for interpreting lateral
impact of composite cylinders. Toward this end it has been shown that, for the relatively
heavy impact mass used in our experiments, the impact event can be treated as equivalent
to a static, concentrated, lateral loading. While the response to this loading is actually quite
nonlinear, at least some general trends have been obtained by means of a linear analysis
based on Fourier series model. However, a simplified nonlinear response model based on
a delaminating ring, as well as other experimental evidence, suggests that delamination may
be a major factor in the response.
The assumption of an equivalent static loading is, of course, a major simplification of a
complicated problem. While this assumption will be less appropriate with lower impact
masses, and thus higher impact velocity for a given impact energy, it seems clear that for
our experiments this is an important idealization. However, it should still be pointed out
that relatively little is known about lateral loadings of laminates.
Although the impact force-deflection response of the laterally loaded cylinder is seen to
be quite nonlinear, it still seems useful to start with a linear analysis. The Fourier series
model available in the literature and used in our work seems particularly appropriate because
it represents the inherent features of the geometry and loading. These features include
orthotropic stiffness (idealized as a single orthotropic material), cylinder geometry, and
localized loading. The comparisons with experimentally measured static stiffness indicate
that if proper attention is paid to the boundary conditions, reasonably good agreement
between analysis and experiment can be obtained.
On the other hand, it is clear that the use of this linear analysis to predict damage resulting
from the impact is on much less firm ground. Clearly the nonlinearity is important. For
example, if delamination occurs, not only the stiffness but also the strain distribution through
the material thickness would be affected. In addition to delamination, failure of the laminate
will also affect the stress distribution through softening of the individual plies. Thus a failure
prediction based on a linear analysis cannot be expected to accurately describe the actual
laminate damage due to the impact. It is also probable that the fracture mechanics model
is not a precise description of the effect of the damage resulting from the impact. However,
the resulting strength retention curve seems quite reasonable and in general agreement with
the experimentally measured residual strengths. The predicted residual strengths are lower
than the experimental data due to the fact that the predicted fiber crack in the 90~ plies was
not always observed experimentally. However, it was observed for the case of the two low
data points shown in Fig. 6. For those specimens a 2.5-cm-long region of broken fibers was
visible on the inside surface of the cylinder. No explanation is available for this difference
in results of similar tests other than data scatter. Thus it may be that the overall physics of
the impact is in fact captured to some degree by the linear modeling.
Although much simplified by the lack of axial coordinate dependence, the nonlinear ring
model appears to give us additional insight into the impact event. Principally, the similarity
of stiffness loss with delamination in the ring model with that measured experimentally
suggests that delamination may indeed be a major contributor to the stiffness loss. Clearly
other factors are also involved, such as the geometric effects of the large deformations.
However it seems that delamination may be a predominant effect.
Finally, it should be noted that our impact experiments cited above included cylinders
with internal support as well as the empty cylinders modeled in the present work. One of
the major effects of the internal support is that it tends to increase the value of the con-
centrated radial stress at the point of impact relative to the bending stress. This concentrated
force may be high enough to produce damage by itself. We are presently carrying out a
companion study in an effort to establish loss of strength due to these Hertz-type loadings.
Conclusions
We have presented two simple models, which when compared with the experimental data
demonstrate that the fundamental features of the impact appear to be represented. The
results are preliminary, and more work needs to be done along the same lines in order to
establish more complete models.
The failure analysis model based on a Fourier series solution for an orthotropic cylinder
under concentrated load, combined with a linear fracture mechanics model for subsequent
strength loss, gave satisfactory agreement with the main trends of the experimental impact
data.
A nonlinear ring model was used to predict delamination sizes in the cylinder resulting
from impact. Agreement with measured data showed that delamination could be the principal
cause of nonlinearity observed in the force deflection response of the impacted cylinders.
References
[1] Labor, J. D., "Impact Damage Effects on the Strength of Advanced Composites," in Nondes-
tructive Evaluation and Flaw Criticality for Composite Materials, ASTM STP 696, R. Pipes, Ed.,
[2] Starnes, J. H. Jr., Rhodes, M. D., and Williams, J. G., "Effects of Impact Damage and Holes
on the Compressive Strength of a Graphite/Epoxy Laminate," in Nondestructive Evaluation and
Flaw Criticality for Composite Materials, ASTM STP 696, R. Pipes, Ed., American Society for
[3] Sharma, A. V., "Low Velocity Impact Tests on Fibrous Composite Sandwich Structures," in Test
Methods and Design Allowables for Fibrous Composites, ASTM STP 734, C. Chamis, Ed., Amer-
ican Society for Testing and Materials, Philadelphia, 1981, pp. 54-70.
[4] Williams, J. G. and Rhodes, M. D., "Effect of Resin on Impact Damage Tolerance of Graphite/
Epoxy Laminates," in Composite Materials: Testing and Design (Sixth Conference), ASTM STP
787, I. M. Daniel, Ed., American Society for Testing and Materials, Philadelphia, 1982, pp. 450-
480.
[5] Cabrino, G., "Residual Strength Prediction of Impacted CFRP Laminates," Journal of Composite
Materials, Vol. 18, 1984, pp. 508-518.
[6] McQuillen, E. J., Gause, L. W., and Llorens R. E., "'Low Velocity Transverse Normal Impact
of Graphite Epoxy Composite Laminates," Journal of Composite Materials, Vol. 10, 1976, pp.
79-91.
[7] Levin, K., "Effect of Low-Velocity Impact on Compression Strength of Quasi-lsotropic Laminate,
in Proceedings (First Technical Conference), The American Society for Composites, 7-9 Oct. 1986,
Dayton, OH, pp. 313-325.
[8] Poe, C. C., Jr. and Kennedy, J. M., "An Assessment of Buffer Strips for Improving Damage
Tolerance of Composite Laminates," Journal of Composite Materials Supplement, Vol. 14, 1980,
pp. 57-70.
[9] Wardle, M. W. and Tokarsky, E. W., "Drop Weight Impact Testing of Laminates Reinforced with
Kevlar Aramid Fibers, E-Glass, and Graphite," Composites Technology Review, Vol. 5, No. 1,
1983, pp. 4-10.
[10] Takeda, N., Sierakowski, R. L., and Malvern, L. E., "'Wave Propagation Experiments On Ball-
istically Impacted Composite Laminates," Journal of Composite Materials, Vol. 15,1981, pp. 157-
174.
[11] Sun, C. T. and Chattopadhyay, S., "'Dynamic Response of Anisotropic Laminated Plates Under
Initial Stress to Impact of Mass," ASME Journal of Applied Mechanics, Vol. 42, 1975, pp. 693-
698.
[12] Dobyns, A. L., "'Analysis of Simply-Supported Orthotropic Plates to Static and Dynamic Loads,"
AIAA Journal, Vol. 19, 1981, pp. 642-650.
[13] Shirakumar, K. N., Elber, W., and Illg, W., "Prediction of Impact Force and Duration Due to
Low-Velocity Impact on Circular Composite Laminates," ASME Journal of Applied Mechanics,
Vol. 52, 1985, pp. 674-680.
[14] Yang, S. H. and Sun, C. T., "'Indentation Law for Composite Laminates," Composite Materials:
[15] Lloyd, B. A. and Knight, G. K., "Impact Damage Sensitivity of Filament-Wound Composites
Pressure Vessels," presented at the JANNAF Meeting, New Orleans, LA, Aug. 1986.
[16] Christoforou, A. P., Swanson, S. R., Ventrello, S. C., and Beckwith, S. W., "Impact Damage of
Carbon/Epoxy Composite Cylinders," Proceedings, 32nd International SAMPE Symposium/Ex-
hibit, Anaheim, CA, 6-9 April 1987, Vol. 32, pp. 964-973.
[17] Swanson, S. R. and Christoforou, A. P.. "'Response of Quasi-Isotropic Carbon/Epoxy Laminates
to Biaxial Stress," Journal of Composite Materials, Vol. 20, 1986. pp. 457-47l.
[18] Kliger, H., Forristall, G., and Vinson, J., "Stresses in Circular Cylindrical Shells of Composite
Materials Subjected to Localized Loads," AFOSR Technical Report No. 73-0494, Jan. 1973.
[19] Vinson, J. R. and Chou, T. W., Composite Materials and Their Use in Structures. Applied Science
Publishers, London, 1975.
[20] Mar, J. W. and Lin, K. Y., "'Fracture of Boron/Aluminum Composites with Discontinuities,"
Journal of Composite Materials, Vol. 11, 1977, pp. 405-421.
[21] Graves, M. J. and Lagace, P. A., "'Damage Tolerance of Composite Cylinders," presented at the
Business Aircraft Meeting and Exposition, Wichita, KS, 12-15 April 1983.
[22] Broek, D., Elementary Engineering Fracture Mechanics, Martinus Nijhoff Publishers, The Hague,
1983.
[23] Poe, C. C., Jr., "A Unifying Strain Criterion for Fracture of Fibrous Composite Laminates,"
Engineering Fracture Mechanics, Vol. 17, No. 2, 1983, pp. 153-171.
[24] Russell, A. J. and Street, K. N., "Moisture and Temperature Effects on the Mixed-Mode Delam-
ination Fracture of Unidirectional Graphite/Epoxy," in Delamination and Debonding of Materials,
ASTM STP 876. W. S. Johnson. Ed., American Society for Testing and Materials, Philadelphia,
L985, pp. 349-370.
[25] Carlsson, L. A., Gillespie, J. W., Jr., and Pipes, R. B., "On the Analysis and Design of the End
Notched Flexure (ENF) Specimen for Mode II Testing," Journal of Composite Materials. Vol. 20,
1986, pp. 594-604.
L a w r e n c e E. Malvern, ~ C. T. Sun, t a n d Dahsin L i u 2
Delamination Damage in Central Impacts at

Subperforation Speeds on Laminated
Kevlar/Epoxy Plates
REFERENCE: Malvern, L. E., Sun, C. T., and Liu, D., "Delamination Damage in Central
Impacts at Subperforation Speeds on Laminated Kevlar/Epoxy Plates," Composite Materials:
ABSTRACT: An experimental research program has shown that in central impacts at normal
incidence by small hard impactors at subperforation speeds on continuous-filament fiber-
reinforced, laminated composite plates with brittle matrices, the major damage is delamination
at the interfaces between unidircctionally reinforced layers of different orientation. This is
accompanied by some matrix cracking between the fibers of the layers.
Composites with three kinds of fibers (glass, Kevlar, and graphite), each in a brittle epoxy
matrix, have been investigated. This paper reports new results on Kevlar and compares them
with some previously published results on glass and graphite. Postimpact visual examination
of the semitransparent glass and Kevlar laminates revealed the extent of the delamination.
Relationships between projected delamination area and the kinetic energy imparted to the
plate are reported. For those plates where the overlapping dclaminations could bc resolved
and thc total delamination area could be determined, the evidence supports the assumption
of a single linear relation between total delamination area and imparted kinetic energy for
each system observed at these subperforation speeds, but there is a limited amount of this
kind of data available. Bending strength and stiffness reductions of the damaged panels are
also reported.
KEY WORDS: composite materials, fiber composites, impact damage, glass/epoxy, Kcvlar/
epoxy, graphite/epoxy, delamination, flexural strength, flexural stiffness
Continuous-filament fiber-reinforced, laminated composites are widely used in applica-

tions in which their high stiffness and strength-to-weight ratios are advantageous, but brittle-
matrix composites are susceptible to impact damage. The general objective of the research
reported here is to gain understanding of the mechanisms of impact damage in such materials,
which will be useful to designers.
An experimental research program at the University of Florida has shown that in central
impacts at normal incidence by small hard impactors at subperforation speeds on continuous-
filament fiber-reinforced, laminated composite plates with brittle matrices, the m a j o r d a m a g e
is delamination at the interfaces between unidirectionally reinforced layers of different
orientation. This is accompanied by some matrix cracking between the fibers of the layers.
Composites with three kinds of fibers (glass, graphite, and Kevlar), each in a brittle epoxy
Department of Aerospace Engineering, Mechanics and Engineering Science, University of Florida,

Gainesville, FL 32611.
-' Department of Metallurgy, Mechanics and Materials Science, Michigan State University, East
Lansing, MI 48824.
387
Downloaded/printed
Copyright9 by
matrix, have been investigated. This paper reports the results on Kevlar and compares them
with some of the previous results on glass and graphite. Postimpact visual examination of
the semitransparent glass and Kevlar laminates with a bright light behind them revealed the
extent of delamination. Transient delamination crack speeds, determined in some glass/
epoxy plates with only two interfaces by photographing the delamination patterns at 25qxs
intervals were in the range of 200 to 500 m/s. The Kevlar/epoxy laminates were not suffi-
ciently transparent for transient measurements by this method. Postimpact delamination
areas in opaque graphite/epoxy laminates were determined by ultrasonic C-scan and X-ray
methods. The X-ray can resolve overlapping delaminations in plates with two to four in-
terfaces, which is comparable to the visual results with Kevlar.
Relationships between the projected delamination area and the kinetic energy imparted
to the plate by the impactor are reported for several different stacking sequences. The
relationship between the total delamination area (the sum of the overlapping delaminations)
and the imparted kinetic energy is examined for some of the plates for which the overlapping
delaminations could be resolved.
Strength and stiffness retention fractions, determined by three-point bend tests of damaged
specimens and undamaged control specimens, have been determined for a range of sub-
perforation speeds on plates of different stacking sequences for each of the three composite
systems. The percentage reduction in strength is greater than the stiffness reduction, as
determined by this method.
In each test series, several laminated plates of the same type were subjected to impact at
a sequence of increasing subperforation speeds, and the damage produced was correlated
with the impact energy imparted to the plates. The impacts were produced by a projectile
fired with the gas gun assembly shown schematically in Fig. 1. The system differs from that
used previously at the University of Florida [1-3] only in the method of recording the
Fiber OpticB u n d ~
Gauge
,,0h, ou,cs, /
MICR0 PRO;e::cSt:,: ~ _ .nlpactOr
GuBS / .arre, Ve.oci,,/

Sensor AB I1
Triggering
Valve
FIG. 1--Schematic of gas gun and velocity measuring system.
MALVERN ET AL. ON DELAMINATION DAMAGE IN CENTRAL IMPACTS 389
impactor velocities. In the tests reported here the impactor was a flat-ended steel cylinder
of length 25.4 mm, diameter 9.5 mm, and mass 14.6 g.
The new velocity sensor [4] uses two fiber-optic-transmitted light paths through a cylin-
drical holder near the impacted plate. Signals from the two impactor-interrupted light paths,
50.8 mm apart, are fed into a microprocessor, which reads out times from which it is possible
to reduce (1) the average inbound speed through the sensor, (2) the arrival time at the
plate, (3) the time when the rebounding projectile leaves the plane of the initial front face
of the target plate, and (4) the average rebound speed through the sensor. (Note that the
difference between (2) and (3) is not necessarily the actual contact time; independent direct
measurements [2,3] and calculations [5] both show that the impactor often loses contact
with the plate before the maximum plate deflection occurs and makes contact again during
the plate spring-back.) The microprocessor can also provide a trigger for transient recorders.
The rebound speed is used to determine the rebound kinetic energy (as much as 10% to
20% of the incident impactor kinetic energy at these subperforation speeds) in order to
determine the energy imparted to the plate for correlation with the amount of impact damage.
Figure 2 is a photograph of the clamping fixture used to hold the target plate. As is pointed
out later, the fixture does not achieve an ideal clamped boundary condition.
The target plates in most of the investigation were nominally 152 mm square. Most of
the glass-epoxy plates and a few early tests on Kevlar/epoxy plates were 15-ply plates with
a thickness around 3.4 ram. The test series to be reported here used plates formed from 15
plies of Fiberite hy-E 177144AA-2 Kevlar/epoxy prepreg tape with thickness around 3.8
mm. Three 00/90 ~ stacking sequences were used in the series:
3-layer (0~/905/0d plates with 2 interfaces,
5-layer (0J90J03/90d0 d plates with 4 interfaces, and
15-1ayer (0/90/...)~5 alternating-ply plates with 14 interfaces.
FIG. 2--Clamping fixture to hold the target plate.
These are similar to the layups previously studied for glass/epoxy plates [1-9]. In glass/
epoxy, the multiple-ply layers were found to offer somewhat greater impact energy ab-
sorption without perforation than the alternating-ply plates [6]. This motivated the inves-
tigation of multiple-ply laminates in all three systems. It turns out that the multiple p l i e s d o
not provide any advantage in graphite/epoxy or Kevlar/epoxy.
The graphite/epoxy laminates [10-12] were fabricated from Hercules AS4/3501-6 prepreg
tapes. Because when this prepreg is cured in a layup it is only about half as thick as the
Fiberite hy-E Kevlar/epoxy prepregs, the number of plies was doubled to 30 to give an
average specimen thickness of 3.8 mm, the same as in the Kevlar/epoxy test series and
comparable to the glass/epoxy. Three symmetric crossply laminates were used in the graph-
ite/epoxy test series:
3-layer (Ot./90~o/O,~) plates with 2 interfaces,

5-layer (0J90r plates with 4 interfaces, and
15-layer (02/902/02...) plates with 14 interfaces.
Takeda [2] has described in detail the procedure for preparing glass/epoxy plates from
prepreg tapes and preparing the package for autoclave curing in the Baron-Blakeslee BAC-
4 autoclave. Similar procedures were followed with graphite and Kevlar. Prepreg tapes are
cut into 305-mm (12-in.) squares at various orientations to the fiber direction; stacked on
an aluminum tool plate with Teflon-coated vent cloth above and below the tape stack,
topped with glass cloth breather layers, a Mylar film, and another aluminum plate; and
enclosed in a vacuum bag. Heating rates are controlled according to procedures recom-
mended by the prepreg tape manufacturer, with some slight modifications to improve the
product quality.
The Fiberite hy-E 177144AA-2 Kevlar/epoxy prepreg layups were cured with an autoclave
pressure of 310 KPa (45 psi) and a suitable temperature cycle, including holding for 1.5 h
at 127~ (260~ after raising the temperature at a rate of 1.11~ to 4.44~ (2~ to 8~ per
minute. A vacuum of 635 mm (25 in.) of mercury was applied to the vacuum port on the
vacuum bag throughout the cure. At the end of the 1.5-h holding period, the autoclave
heater was turned off. After the system cooled to 65.5~ (150~ the autoclave pressure
was released, but the plates were not removed from the autoclave until it had cooled
essentially to room temperature.
The Hercules AS4/3501-6 graphite/epoxy prepreg layups were given a more complicated
cure cycle. A higher autoclave pressure, 586 KPa (85 psi) and the vacuum of 635 mm (25
in.) of mercury were applied at the beginning. Temperature was raised at a rate of 1.67~
to 2.78~ (3~ to 5~ per minute to 116~ (240~ and held for about 1 h. Then the autoclave
pressure was increased to 690 KPa (100 psi), the vacuum was released, and temperature
was raised at the same rate to 177~ (350~ and held for about 2 h. Then the heat was
turned off. The autoclave pressure was released after the system had cooled to 93.3~
(200~ and further cooling to about room temperature was allowed before the plates were
removed.
After cure, specimens were cut from the plates with a diamond saw. Usually four specimens
were cut from each plate, three to be impacted at three different speeds (in the range of
20 m/s to 100 m/s) and one for a control specimen to be tested to failure for comparison
to determine the strength and stiffness retention factors for the damaged specimens. A few
larger 229-mm-square Kevlar/epoxy specimens were tested to verify that the central damage
pattern was not significantly affected by the boundary distance at these impact speeds.
Postimpact determination of the delamination pattern in glass and Kevlar could be done
visually with a bright light behind the plates. Figures 3 and 4 are photographs of 229-mm-
FIG. 3--Photograph showing two delamination areas in three-layer Kevlar/ epoxy plate.
square Kevlar/epoxy plates showing the delamination pattern in a three-layer plate with
two interfaces and a five-layer plate with four interfaces, respectively, with a 600-W light
behind the plates. For these cases it was possible to resolve the overlapping delaminations
and determine the total delamination as the sum of the individual areas. For the five-layer
plate views from both front and back were used, but only the back views are shown in Figs.
3 and 4.
In the two photographs, the fiber direction of the front and back layers is vertical. In the
two-interface plate of Fig. 3, the delamination in the first interface is small and peanut-
shaped and extends horizontally a little beyond the boundaries of the larger vertical delam-
ination of the second interface. The first-interface delamination has its greatest extent in
the direction of the fibers of the second layer, and the last-interface delamination has its
greatest extent in the direction of the fibers of the last layer. The boundaries of the individual
delaminations in Fig. 3 are quite easy to see in the original photograph (possibly not so easy
in the printed reproduction). The only uncertainty in the original might be the right and
FIG. 4--Photograph showing four delamination areas in five-layer Kevlar/epoxy plate.
left boundaries of the second delamination in the small region where the first delamination
overlaps them.
Determination of the boundaries of the four delaminations in Fig. 4 requires a little more
imagination. The first and third delaminations extend horizontally in the photograph, while
the second and fourth extend vertically. In the original photograph, it is fairly easy to resolve
the outer portions of the boundaries, but the central region is not so certain. The first one
is horizontal and peanut-shaped, and the second is believed to be vertical and peanut-shaped.
The third one is not symmetrical with respect to a vertical centerline, but extends farther
to the right than to the left, and it is squared off somewhat on the right side. The fourth
one extends upward farther than it does downward.
For plates with more than four interfaces, it was not possible to resolve the overlapping
delaminations in the photograph. The total delamination could be determined by an edge
replication examination of slices cut from the plate [13], which requires destruction of the
plate.
Transient deflection measurements were made on a few glass/epoxy and Kevlar/epoxy

plates with a noncontacting eddy-current probe, which senses the motion of an aluminum
foil element cemented on the back face of the specimen. Figure 5 shows examples of the
central deflection versus time for one 00/90 ~ alternating-ply glass/epoxy plate with 15 plies,
and one 00/90 ~ alternating-ply Kevlar/epoxy plate with 18 plies, impacted at such a low
speed that very little damage occurs. It is noteworthy that even at these low speeds, the
deflection is of the order of the plate thickness (about 3.4 ram). Evidently even at this low
speed the use of small-deflection plate theory to calculate the deflections may be question-
able. Comparisons were made between the observed central deflections in a glass/epoxy
plate and the deflections calculated with a shear-deformable small-deflection elastic plate
theory [5]. The experimental plates were nominally 152 mm square, with a clamped boundary
leaving an exposed area 140 mm square. Because the clamping is not perfect, it was found
that good agreement with the shape of the central deflection versus time could only be
obtained by assuming that the plate was slightly larger (170 mm square), but then the
calculated maximum deflection was greater than the observed deflection (2 mm instead of
1.4 ram). In the calculations a Hertzian impact law was used for the interaction of the
impactor (assumed rigid) and the elastically deforming plate. The calculation predicted times
during the first vibration cycle after impact at which the impactor lost and regained contact
with the plate. These times were similar to contact measurements made on other glass/
epoxy plates [2] and are associated with the irregularities in the shape of the first cycle of
the plate vibration. Subsequent cycles (not shown) appeared sinusoidal. The Kevlar/epoxy
response has a somewhat shorter period than the glass/epoxy. The time to the maximum
deflection of 2.36 mm in the first half cycle is 395 p,s for the Kevlar/epoxy; corresponding
figures for the glass/epoxy are 2.47 mm and 501 I~s, as is shown in Fig. 5.
.52 mm)
(0.0 #s, 0
500 ps /
GLASS-EPOXY [0~176 ....

/ Impact Velocity: 19.8 m/sec
~ )// Maximum Deflection: 2.47 mm
(501 /~s,2.47 mm)

/~s, 1.24 mm)
~p/(0.0, s, 0.0 mm) / ~ mm)
(395/Is, 2.36 mm)

FIG. 5--Transient central deflections in a glass~epoxy plate and in a Kevlar/epoxy plate.
] I i I/ I I V~ I
200 I i ////z~ -
/ /
160 ~ / // -
I / /
i~ 12~
I-
/ till/
o, I
~ eo
I-l=/ i i
'
bl
/1/./
l ,' r'l r'l3-Layer
mO 40 V I / , / " 0 - - - - - 0 5-Layer -
0. ~ h i - - - - - & 15-Layer
i
o_io i i ] I I I [
0 20 40 60 80
KINETIC E N E R G Y , (Joules)
FIG. 6--Projected delamination area versus imparted kinetic energy in three o,pes of
Kevlar/ epoxy laminates.
Results of Delamination Versus Imparted Kinetic Energy

Figure 6 shows plots of projected delamination area versus imparted kinetic energy for
the three 00/90 ~ layups of the test series of 15-ply Kevlar/epoxy plates. The projected
delamination area does not resolve the overlapping areas, so that, as expected, the projected
area is largest for the 3-layer plate and smallest for the 15-layer plate, with the 5-layer
intermediate. For each of the three layups the projected delamination is approximately
linear with the imparted kinetic energy. The straight lines in the figure were fitted to the
data for each layup by least squares. The threeqayer data extends to much smaller energies
than the others, because the delamination approached the boundaries at the higher impact
speeds when the delamination was confined to only two interfaces.
Figure 7 shows similar plots for 15 tests at various speeds on each of the three similar
layups of the 30-ply graphite/epoxy plates [11,12]. Delamination areas in the opaque graph-
ite/epoxy were determined by ultrasonic C-scan and by X-ray. The plotted points shown in
Fig. 7 are from C-scan data, while two fitted lines are shown for each layup, one based on
the C-scan data and one based on the X-ray data. Despite the data scatter and considerable
disagreement between the results from the two methods, the results indicate a linear rela-
tionship between imparted kinetic energy and projected delamination for each layup.
The impact data tables A1 to A4 in the Appendix show measured values for most of the
plotted points. In Tables A1 to A3 for graphite/epoxy, large differences may be seen in
many cases between the projected delamination as determined by the C-scan and as deter-
mined by the X-ray. The C-scan tends to overestimate the delamination because other types
of damage besides delamination (for example, matrix cracking) may obstruct the ultrasonic
signal. (See Ref 14 for a discussion of matrix cracking in impacted glass/epoxy plates.) The
X-ray often underestimates delamination because the iothalamate sodium that was injected
did not penetrate to the limits of the delamination. The plotted points for projected delam-
160 r- / / o 3 - Layer I D t
l o /I ~S-Layer !R.aw s.
[._ /~ 0 15 " Layer J o! c-scan
| // / ~ Least-Squares of c-scan
F-- ~_ /// - - - Least-Squares of x-ray
~ 120
=< 8 o ~- oo/I// y ..
I I/ 11
t- /,, J
,ok ~
0 I~/~J~ ~ I - - '~ l - - I I I
0 10 20 30
KINETIC E N E R G Y (Joules)
FIG. 7--Projected delamination area versus imparted kinetic energy in three types of graph-
ite~epoxy laminates.
ination were based on the C-scan data, while the total delamination had to be based on the
X-ray. This explains the anomalous result of some tests in which the projected delamination
area appears to be greater than the total delamination area.
If the energy per unit delaminated area has any real significance, then the results should
plot on a single line when the total delamination area is plotted instead of the projected
delamination area. Figure 8 shows such a plot for three 3-layer and three 5-layer plates of
the 15-ply Kevlar/epoxy test series, where the resolution of the overlapping areas could be
made and the total delamination area determined: these results are consistent with the
assumption of a single line whose slope indicates an energy absorption per unit delamination
area of about 114l J/m-' after a threshold value of 0.197 J, but these numbers are based on
very limited data.
Figure 9 shows similar results for total delamination in graphite/epoxy, based on X-ray
data [11,12] for 8 three-layer plates and 11 five-layer plates. The results support the as-
sumption of a single linear relationship, but it is based on limited data, especially for the
three-layer plates. The slope of the fitted line for the five-layer plates indicates an energy
absorption per unit area of delamination of 3440 J/m 2 in these graphite/epoxy plates. The
accuracy of this value may be suspect because of the problems noted above concerning the
X-ray method.
This value for delamination energy absorption in graphite/epoxy is comparable to some
early results on filament-wound glass/epoxy [15,16], which showed a single linear plot of
impact energy versus delamination area with a coefficient of 3150 J/m-' in plates with two
or four interfaces. A second series on plates fabricated with Scotchply Type 1003 prepreg
tapes [2,3] gave a coefficient of 7500 J/m', more than twice the previous value. An extensive
N A S A study [17] of graphite/epoxy laminates has also shown large differences in impact
energy absorption for different resins.
The numerical values quoted in this report for both the graphite/epoxy and the Kevlar/
epoxy should not be regarded as definitive because of the small number of tests and because
of the great variability among different matrices. It seems clear, however, that when the
5O
W~J 4 0
30
20
I,- 10
/ n n 3-Layer
o----o 5-Layer
I I I I I I I I I
0 2 4 6 8 10
KINETIC ENERGY, (Joules)
FIG. 8--Total delamination area versus imparted kinetic energy in two types of Kevlar/
epoxy laminates.
superior stiffness-to-weight ratio of graphite/epoxy is not required the glass/epoxy laminates

should be considered because of their good impact energy absorption and low cost.
Fracture surface energies based on these results would be half the values quoted because
two surfaces are formed at each delamination. The lowest value determined this way on
glass/epoxy is still an order of magnitude higher than some fracture surface energies de-
termined by static double-cantilever-beam tests on pure epoxy [18] or on glass/epoxy lam-
inates [19].
z 120 A
0 o 3-Layer
I--. -- &5-Layer
z~
i E .o
<~
..I ~ A
9-J "~ 4 0
,< -- O
I.-
0
I,-
~ o Ol- I I I I I
0 10 2o 3o
KINETIC ENERGY (Joules)
FIG. 9--Total delamination area versus imparted kinetic energy in two types of graphite/
epoxy laminates.
Structural Degradation of Impacted Plates

The results reported in the previous section support the assumption that in these laminated
plates with a brittle epoxy matrix the total delamination produced by these impacts by small
steel impactors at subperforation speeds is a linear function of the kinetic energy imparted
to the plate for each of the composite systems studied. An important problem remains,
however, namely the effect of the damage sustained in stopping the impactor on the use-
fulness of the damaged plate as a structural element.
Retained strength and stiffness after impact could be characterized in several different
ways. Tests could be made in tension, compression, or bending, for example, in either
monotonic or cyclic loading of test coupons cut from the damaged region. Different types
of tests and different coupon sizes might produce different results. The choice of test could
be guided by the particular function the damaged element was expected to perform.
In this investigation static three-point cylindrical bending tests were performed in which
the 152-mm-square plate was supported by two knife edges, each about 12.7 mm from a
plate edge and monotonically loaded by a central wedge with a rounded edge parallel to
the supports. The two edge supports are each attached to the foundation by a single pin
parallel to the direction perpendicular to the two edge supports. The edge supports are free
to rotate about this pin, so that no twisting moment is applied to the plate. This type of
test had the advantage that it was easy to perform, gave a clearly identifiable point of
maximum load, and proved to be surprisingly reproducible. The tests measured the residual
strength and stiffness of the whole structural element rather than the local strength and
stiffness of the damaged region. The results obtained were therefore particular to the size
of plate employed. Of course, the same damage area in a much larger plate would produce
a much smaller percentage decrease in strength and stiffness.
[n each test the two edges selected for support were chosen so that the plate was tested
in the stiffer orientation, that is with the fibers of the outer layers perpendicular to the
supports and the loading wedge. The three-point cylindrical bending tests were performed
on impact-damaged plates and on undamaged control specimens cut from the same plate.
Because the compression response is more sensitive to delamination than the tension re-
sponse, the plates were tested with the back side in compression, which gave the greater
degradation because the greater delamination was in the interface near the back side (op-
posite from the impacted side).
The load P versus deflection w curve was determined; the strength was determined from
the maximum load P~,x and the stiffness from the slope d P / d w of the linear portion of the
curve. To take account of the thickness variations between specimens not cut from the same
plate, the results were expressed as apparent elastic flexural modulus E r and apparent
strength (r.... calculated as though the response were linearly elastic up to failure [10-12],
by the following equations:
Apparent E t = ( d P / d w ) L 3 / 4 b h 3 Apparent crm,x = 3Pm~xL/2bh 2 (1)

where
L = the distance between supports (usually 127 mm),
b = the width of the plate (usually 152 mm), and
h = the plate thickness, which was measured on each specimen.
Figures 10 and 11 show examples of apparent strength and stiffness fractions retained by
impact-damaged graphite/epoxy laminates, plotted versus imparted kinetic energy. EOr and
~r%,~ denote stiffness and strength of undamaged control specimens cut from the same plate.
1.2 -
o 3-Layer
1.0 ~ A 5-Layer
o 15-Layer
oE0.8
O
o
"~o.6
10 0.4 \~ -.<o
0.2
I I I I I I I l
0 10 20 30 40

FIG. lO--Strength retention fractions in graphite~epoxy plates.
The 3-layer plate has the most degradation; the 15-layer plate has the least. The stiffness
reduction is much less than the strength reduction in all cases. Therefore, nondestructive
evaluation procedures based on stiffness measurements are likely to underestimate the
damage. Pre- and postimpact vibration tests on graphite/epoxy plates [11], for example,
show little variation in the natural frequencies of the plates, supported as cantilevers. Damp-
ing measurements were much more sensitive to the damage, but the damping results showed
a great deal of scatter, which was believed to be caused by uncontrolled damping in the
supports.
Figures 12 and 13 show similar examples of apparent strength and stiffness retention
fractions for the test series of Kevlar/epoxy laminates, versus imparted kinetic energy. Figure
14 shows strength and stiffness reduction in glass/epoxy 3-layer plates, plotted versus de-
lamination area. Again the strength reduction is greater than the stiffness reduction, but in
the 5-layer and 15-layer Kevlar/epoxy laminates the stiffness reduction is much greater than
it was in the graphite/epoxy plates. All the strength and stiffness data shown in Figs. 10
through 13 have been presented as retained fractions of the strength and stiffness of un-
damaged control specimens cut from the same plate. To give some idea of the relative
1"2 F a a
~J~loLO _
~ ----o
0.8
k-
<
m 0.6 o
~.
u)
0.4
o.2
I ~176
~, 5-Layer
o 1S-Layer
o I I I I I I I I
o 10 20 30 40

FIG. 11--Stiffness retention fractions in graphite~epoxy plates.
1.0 li~ I I I I I I
I ~~\ KEVLAR/EPOXY
.s \\\
Z
\\
i\
O .6
I--
Z
UJ
I--
UJ \
iv- O O3-Layer
"r .4 oo
~ ~\\ o-------o 5-Layer
I-- A--- -- .--.~15- Layer
Z
W
n,-
I,- .2 - \\
(n \
o \
\
0 P l i i i i l
0 8O
KINETIC ENERGY, (Joules)
FIG. 12--Strength retention fractions in Kevlar/epoxy plates.
._o &
1.0 ~ I I I I I I I
__~ KEVLAR/EPOXY
O .8
O \
< \
Z N
O
W
.6-
-
~
\: ~
".
N
-
re" .4- ~ -
if)
ill
Z
14.
B :\ A
~ .2--
O Q 3-Layer
u) o ~ - ~ o 5-Layer
-- ,~------~ 15-Layer
0 I I I I I I I
0 20 40 60 80
KINETIC ENERGY, ( J o u l e s )
FIG. 13--Stiffness retention fractions in Kevlarlepoxy plates.
400 SECOND C O M P O S I T E MATERIALS
GLASS - EPOXY [0~/90~/0~]

1.0
O
~ 0.8
u.
0.6
uJ
rr
ii1
Z O STIFFNESS REDUCTION FACTOR
L,L
~ 0.4 9 STRENGTH REDUCTION FACTOR
Z
"r
I-
Z
w 0.2
I,.-
I 1 I I !
2 4 6 8 10
DELAMINATED AREA (IN 2)

FIG. 14--Strength and stiffness reduction in three-layer glass~epoxy plates versus delami-
hated area.
performance of the three systems studied, typical flexural strengths and flexural moduli for
each of the three layups of the three material systems tested are listed in Tables 1 and 2.
The flexural modulus E ' r and flexural strength (r'~,,,x were computed for typical control
specimens by Eq 1, based on cylindrical three-point bend tests as previously described.
These results are for a specific kind of test, and some caution is in order in generalizing
to other kinds of loading. The apparent superiority of the three-layer plates is, of course,
a result of the fact that these are bending tests with outer layer fibers in the direction
perpendicular to the supports. The three-layer plates have two thirds of their fibers in the
outer layers, and it is not surprising that this configuration is stronger and stiffer in bending
than the others. As would be expected, the graphite/epoxy is the stiffest in all configurations
by a considerable margin. It is also the strongest in all of the configurations tested, but the
margin is not quite so great over the E-glass/epoxy as it was for the stiffness.
TABLE 1--Typical flexural strength (r~ of undamaged control specimens.

System Glass / Epoxy Kevlar/ Epoxy Graphite / Epoxy
3-layer 1120 MPa 425 MPa 1490 MPa

TABLE 2--Typical flexural modulus E~ of undamaged control specimens.
System Glass/Epoxy Kevlar/Epoxy Graphite/Epoxy
3-layer 43 GPa 51 GPa 130 GPa

In comparing the performance of the different materials, it should be noted that the results
quoted are for one batch of each specific material and that the laminate parameters differed
somewhat for the different materials. Materials and some typical laminate parameters are
summarized in Table 3,

This paper has concentrated on experimental research at the University of Florida and
has not attempted to review other work dealing with low-speed impact on composite lam-
inates. Many of the papers cited give references to other work. Reference 20 contains several
papers on impact, as does a recent N A S A Langley Workshop Proceedings volume [21]. See
also papers by Sun and his colleagues at Purdue University [22-25] including representations
of the force law between the impactor and target plate or beam and numerical calculations
of the propagating elastic waves.
The experimental research program at the University of Florida has shown that, for all
three systems studied, in central impacts by small hard impactors at subperforation speeds
on continuous-filament fiber-reinforced, laminated composite plates, the major damage is
detamination at the interfaces between unidirectionally reinforced layers of different ori-
entation, accompanied by some matrix cracking between the fibers of the layers.
For each of the systems, the total delamination data available are consistent with the
assumption that at subperforation speeds the total delamination area is a linear function of
the kinetic energy imparted to the plate by the impactor. The slopes of the fitted straight
lines indicated values of the energy per unit area required for delamination of about 1141
J/m-' for the Kevlar/epoxy, 3440 J/m 2 for the graphite/epoxy, and as much as 7500 J/m-'
for the E-glass/epoxy. Details of the particular materials and plate configurations studied
are given at the end of the preceding section and in the section entitled "Experimental
Procedure."
For each of the systems and each of the plate configurations studied, the flexural strength
and stiffness degradation of the impacted plates, as measured in cylindrical three-point bend
tests, is approximately linear with the kinetic energy imparted to the plate. Examples for
graphite/epoxy, Kevlar/epoxy, and glass/epoxy were given in the preceding section.
The energy-absorption advantage of the multiple-ply layers, previously observed in glass/
epoxy, was not found in the Kevlar/epoxy and graphite/epoxy systems. Therefore, further
TABLE 3--Summary of materials used and some laminate parameters.~
Number Thickness, Estimated Fiber

Material of Plies mm Mass, g Volume Fraction
Glass/epoxy MMM Scotchply 1003 15 3.3 135 0.5

Kevlar/epoxy Fiberite hy-E 177144AA-2 15 3.8 125 0.7
Graphite/epoxy Hercules AS4/3501-6 30 3.8 143 0.7
~ All specimens were 152 mm square.
research on impact d e l a m i n a t i o n , especially in g r a p h i t e / e p o x y , should p r o b a b l y be c o n c e n -

trated on plates w h e r e the layers have only one or two plies in a given direction. This m e a n s
that postimpact d e l a m i n a t i o n d e t e r m i n a t i o n will have to be d o n e by a destructive t e c h n i q u e .
It also m e a n s that the d e l a m i n a t i o n s in central impacts will usually be c o n f i n e d to a region
near the impact point and may be m o r e a m e n a b l e to analysis by simple d a m a g e a c c u m u l a t i o n
models without the necessity of resorting to d e l a m i n a t i o n fracture crack p r o p a g a t i o n anal-
yses.
Acknowledgments
The results r e p o r t e d here w e r e o b t a i n e d with s u p p o r t from the U . S . A r m y R e s e a r c h
Office, R e s e a r c h Triangle Park, NC, u n d e r C o n t r a c t No. D A A G 2 9 - 8 3 - K - 0 1 0 7 . Facilities
and p e r s o n n e l o f the University o f Florida C e n t e r o f Excellence P r o g r a m in N e w Materials
were used. The transient deflection m e a s u r e m e n t s s h o w n in Fig. 5 were m a d e by Dr. S.
Chaturvedi (now a m e m b e r o f the Civil E n g i n e e r i n g Faculty at O h i o State U n i v e r s i t y ) while
he was a p o s t d o c t o r a l associate at the University of Florida.
APPENDIX
Tables of Impact Data
TABLE Al--lmpact data, graphite~epoxy Hercules AS4/3501-6, 3 layer.
Projected Delamination
Velocity, m/s Area, cm: Total Area Retention Factors
Specimen Imparted Delaminated
ID V, V, Energy,J By C-Scan By X-Ray by X-Ray, cm2 Strength Stiffness
GR3, IA) 32.40 10.68 6.67 82.5 36.0 39.2 . . . . . .
GR3, IB) 44.56 12.00 13.12 150.0 150.0 . . . . . . . . .
GR3 2A) 40.64 10.87 10.92 105.1) 85.0 . . . . . . _.
GR3 2B) 32.94 10.75 6.91 44.0 26.0 26.0 . . . . . .
GR3 5A) 31.55 10.68 6.28 45.0 26.0 26.3 . . . . . .
GR3 5B) 46.06 K73 14.57 145.0 . . . . . . . . . . . .
GR3 8A) 24.71 12.11 3.30 26.0 . . . . . .
GR3 8B) 29.16 12.99 4.86 27.0 1310 iii . . . . . .
GR3 8C) 28.48 8.29 5.29 12.0 8.11
GR3 8D) 20.20 6.64 2.59 28.5 9.5 1"1".'0 015'7 1.03
GR3 13A) 24.52 6.02 4.02 25.2 10.0 ...
GR3 13C) 22.62 7.50 3.24 20.0 11.5 0128 0163
GR3 (13D) 22.58 7.52 3.23 18.2 7.0 7~0 0.45 0.97
GR3 (14A) 24.47 9.97 3.56 33.2 18.5 18.5 0.70 11.99
GR3 (14B) 28.32 13.26 4.46 34.6 5.0 5.5 0.49 11.81
GR3 (14C) 33,01 14.56 6.25 83.5 14.0 14.0 0.41 11.64
GR3 (14D) 35.52 5.73 8.76 94.6 . . . . . . 0.43 0.79
MALVERN ET AL. ON DELAMtNATION DAMAGE IN CENTRAL IMPACTS 403
T A B L E A2--1mpact data, graphite~epoxy Hercules AS4/3501-6, 5 layer.
Velocity, m/s Area, cm 2 Total Area Retention Factors
ID V, V, Energy, J By C-Scan By X-Ray by X-Ray, cm-' Strength Stiffness
G R 5 3A) 32.40 7.94 7.03 24.0 13.5 18.0 . . . . . .

GR5, 3B) 38.96 10.85 9.97 28.0 24.0 33.0 . . . . . .
GR5, 3C) 56.82 14.05 21.56 46.0 21.0 33.4 . . . . . .
GR5, 3D) 38.96 10.85 9.97 40.5 40.0 56.8 . . . . . .
GR5, 6A) 59.00 19.91 21.98 98.0 47.5 71.0 . . . . . .
GR5~ 6B) 25.25 20.96 1.41 13.2 6.0 6.0 . . . . . .
GR5 ! 6C) 46.10 12.03 14.1 45.0 39.0
GR5 i 6D) 63.66 14.51 27.37 93.0 57.0 67.5
GR5 q 9A) 31.93 5.83 7.02 19.3 15.0 0175 0190
GR5 i 9B) 46.10 12.03 14.11 49.2 39.0 54.0 0.43 0.28
GR5 9C) 56.82 14.05 21.6 84.0 55.0 70.2 0.59 0.91
GR5 i 9D) 66.15 14.24 29.7 120.5 96.5 115.5 0.43 0.77
GR5 (I1A) 32.71 8.50 7.10 23.6 14.0 9.9
GR5 ( l l B ) 40.16 9.24 10.88 41.6 19.0 24.5 0172 0192
GR5 ( l l C ) 52.16 14.52 17.88 69.0 35.0 52.0 0.55 0.81
GR5 ( l l D ) 56.82 14.05 21.6 67.0 41.0 59.0 . . . . . .
T A B L E A3--1mpact data, graphite~epoxy Hercules AS4/3501-6, 15 layer.
Velocity, m/s Area, em z Total Area Retention Factors
ID V, Vr Energy, J By C-Scan By X-Ray by X-Ray, cm 2 Strength Stiffncss
GRI5 (4A) 32.40 7.94 7.03 11.0 10.0

GR15 (4B) 38.96 10.85 9.97 4.6
GRI5 (4C) 70.85 20.63 32.7 28, 7 22-.'0 01;5 1107
GRI5 (4D) 71.35 14.75 34.7 33.8 20.0 0.59 0.95
GRI5 (7A) 31.89 7.94 6.80 4,4 2.0
GRI5 (7B) 45.32 11.22 13.74 8.2 6.0
GR15 (7C) 28.57 7.46 5.42 14.0 7.0 01~4 "
GR15 (7D) 66.93 11.28 31.0 25.3 17.0 0.53 0~75
GRI5 (10A) 60.55 11.08 25.2 19.0 9.0 0.62 0.97
GRI5 (10B) 52.05 12.13 18.25 16.4 10.0
GRI5 (10C) 40.41 9.51 10.99 6.0 4.5
OR15 (10D) 28.57 7.46 5.42 1.0 H .
GR15 (12A) 68.28 19.51 30.5 38.0 20,5

GRI5 (12B) 71.95 20.63 33.9 40.0 30.0 o)io o~5
GR15 (12C) 74.82 21.17 36.7 39.0 26.0
GR15 (123) 63.90 18.54 26.6 30.0 22.5
T A B L E A 4--1mpact data, Kevlar / epoxy ( Fiberite Hy-E 177144A-2).
Projected Delami-
nation
Velocity, m/s Energy, J Area. cm2 Total Area Retention Factors
Specimen Delaminated
ID V', V', Incident Imparted By C-Scan Visual Visual. cm2 Strength Stiffness
3-layer
KE3 (10A) 20.6 3.1 2.4 a 17.8 16.49 19.1 0.88 1.02
KE3 (10B) 20.6 3.1 2.4" 15.6 16.30 18.0 0.84 1.01
KE3 (10C) 20.6 3.1 2.4a 17.5 ... 32.6 0.75 1.02
KE3 (liD) 27.7 5.61 4.7 a 71.2 . . . . . . 0.41 0.76
KE3 (llC) 31.24 7.14 5.75 a 81.0 . . . . . . 0.38 0.73
5-layer
KE5 (14D) 25.1 4.6 3.55a 20.8 17.0 27.8 0.66 0.94
KE5 (14C) 31.64 14134 7.3 5.82 29.1 22.3 47.6 0.65 0.97
KE5 (15C) 31.24 13.99 7.1 5.71 27.0 24.8 47.4 0.66 1.04
KE5 (14A) 45.57 14.59 ... 13.63 89.5 88.8 ... 0.32 0.75
KE5 (15A) 45.75 14.56 ... 13.76 101.1 89.0 ... 0.39 0.78
KE5 (14B) 66.90 10.41 ... 31.95 189.5 144.5 ... 0.14 0.28
KE5 (15B) 67.91 ... 31.95 203.8 155.2 ... 0.14 0.32
15-1ayer
KE15 (12D) 31.73 10.61 6.5 15.1 7.6 ... 0.70 1.02
KEI5 (13A) 45.69 18.30 12.89 41 36.3 ... 0.52 1.06
KE15 (12A) 46.42 17.40 13.50 42.6 31.1 ... 0.56 1.06
KEI5 (12B) 68.00 14.06 32.40 150.6 69.7 ... 0.23 0.48
KE15 (13B) 69.40 18.77 32.70 144 89.0 ... 0.24 0.55
KE15 (12C) 91.38 62.4 210.5 190.5 ... 0.13 0.41
KE15 (13C) 93.06 li133 62.4 194.2 183 ... 0.14 0.37
~ Estimated imparted energy for six cases where Vr was not recorded.
References
[1] Ross, C. A. and Sierakowski, R. L., "Studies on the Impact Resistance of Composite Plates,"
Composites, Vol. 4, 1973, pp. 157-161.
[2] T a k e d a , N., "Experimental Studies of the Delamination Mechanisms in Impacted Fiber-Reinforced
Composite Plates," Ph.D. dissertation, University of Florida, Gainesville, FL, 1980.
[3] Takeda, N., Sierakowski, R. L., and Maivern, L. E., "Studies of Impacted Glass Fiber-Reinforced
Composite L a m i n a t e s , " SAMPE Quarterly, Vol. 12, 1981, pp. 9-16.
[4] Malvern, L. E., Sun, C. T., and Liu, D., " D a m a g e in Composite L a m i n a t e s from Central Impacts
at Subperforation Speeds," in Recent Trends in Aeroelasticity, Structures and Structural Dynamics,
Proceedings of Symposium in M e m o r y of Professor R. L. Bisplinghoff, P. Hajela, Ed., University
of Florida Press, Gainesville, FL, 1987, pp. 298-312.
[5] Petersen, B. R., "'Finite E l e m e n t Analysis of Composite Plate Impacted by a Projectile," P h . D .
dissertation, University of Florida, Gainesville, FL, 1985.
[6] Cristescu, N., Malvern, L. E., Ross. C. A . , and Sierakowski, R. L., "Failure Mechanisms in
Composite Plates Impacted by Blunt-Ended Penetrators," in Foreign Object Impact Damage to
Composites, ASTM STP568, American Society for Testing and Materials, Philadelphia, 1975, pp.
159-172.
[7] Sierakowski, R. L., Ross, C. A . , Malvern, L. E., and Cristescu, N., "'Studies on the Penetration
Mechanics of Composite Plates," Final Report, Grant No. DAAG29-76-C-0085, U.S. A r m y Re-
search Office, Research Triangle Park, NC, Dec. 1976.
[8] Sierakowski, R. L., Ross, C. A . , and Malvern, L. E., "Studies on the Fracture Mechanisms in
Partially Penetrated Filament Reinforced Laminated Plates," Final Report, G r a n t No. D A A G 2 9 -
79-G-0007, U.S. A r m y Research Office, NC, 1981.
[9] Sierakowski, R. L., Ross, C. A . , Malvern, L. E., and Doddington, H. W., "'Studies on the Fracture
Mechanisms in Partially Penetrated Filament Reinforced L a m i n a t a e d Plates," Technical Report
No. 1, Grant No.'DAAG29-79-G-00007, U.S. A r m y Research Office, NC, 1982.
[10] Malvern, L. E. and Sun, C. T., "Delamination Sensing and Modeling in Localized Impacts on
Filament-Reinforced Laminated Plates," Final Report, Contract No. DAAG29-83-K-0107, U.S.
Army Research Office, NC, 1986.
[11] Liu, D., Sun, C. T., and Malvern, L. E., "Structural Degradation of Impacted Graphite/Epoxy
Laminates," The Shock and Vibration Bulletin, Bulletin 56, Part 2, Aug. 1986, pp. 51-60.
[12] Liu, D., Malvern, L. E., and Sun, C. T., "Delamination in Central Impact in Graphite/Epoxy
Laminates," in Proceedings 1986 SEM Conference on Experimental Mechanics, Society for Ex-
perimental Mechanics, Bethel, CT, 1986, pp. 863-868.
[13] Liu, D., Lillycrop, L., Malvern, L. E., and Sun, C. T., "The Evaluation of Delamination--An
Edge Replication Study," Experimental Techniques, Vol. 11, May 1987, pp. 20-25.
[14] Liu, D. and Malvern, L. E., "'Matrix Cracking in Impacted Glass/Epoxy Plates," Journal of
Composite Materials, Vol. 21, July 1987, pp. 594-609.
[15] Sierakowski, R. L., Malvern, L. E., and Ross, C. A., "'Dynamic Failure Modes in Impacted
Composite Plates," in Failure Modes in Composites 111, T. T. Chiao, Ed., American Institute of
Mining, Metallurgical and Petroleum Engineers, NY, 1976, pp. 73-88.
[16] Malvern, L. E., Sierakowski, R. L., Ross, C. A., and Cristescu, N., "'Impact Failure Mechanisms
in Fiber-Reinforced Composite Plates," in High Velocity Deformation of Solids, IUTAM Sym-
posium, Tokyo, 1977, K. Kawata and J. Shiori, Eds., Springer-Verlag, Berlin and Heidelberg,
1979, pp. 120-131.
[17] Williams, J. D. and Rhodes, M. D., "The Effect of Resin on the Impact Damage Tolerance of
Graphite/Epoxy Laminates," in Composite Materials: Testing and Design (Sixth Conference),
ASTM STP 787, I. M. Daniel, Ed., American Society for Testing and Materials, Philadelphia,
1982, pp. 450-480.
[18] Bascom, W. D., Jones, R. L., and Timmons, C. O., "Mixed Mode Fracture of Structural Ad-
hesives," in Adhesion Science and Technology, Vol. 9-B, L. H. Lee, Ed., Plenum Press, New
York, 1975, p. 501.
[19] Yeung, P. and Broutman, L. J., "The Effect of Glass-Resin Interface Strength on the Impact
Strength of Fiber Reinforced Plastics," Polymer Engineering and Science, Vol. 18, 1978, pp. 62-
72.
[20] Foreign Object Impact Damage to Composites, ASTM STP 568. American Society for Testing and
Materials, Philadelphia, 1975.
[21] Tough Composite Materials, Recent Developments (1983 NASA Langley Workshop), Noyes Pub-
lications, Park Ridge, NJ, 1985.
[22] Yang, S. H. and Sun, C. T., "Indentation Law for Composite Laminates," in Composite Materials:
[23] Tan, T. M. and Sun, C. T., "'Use of Statistical Indentation Laws in the Impact Analysis of Laminated
Composite Plates," Journal of Applied Mechanics, Vol. 52, 1985, pp. 6-12.
[24] Sankar, B. V. and Sun, C. T., "'Low-Velocity Impact Response of Laminated Beams Subjected
to Initial Stress," AIAA Journal, Vol. 23, 1985, pp. 1962-1969.
[25] Joshi, S. P. and Sun, C. T., "Impact Induced Fracture in a Laminated Composite," Journal of
Composite Materials, Vol. 19, 1985, pp. 51-66.
Author Index
A H
Aiello, R. A., 137 Han, K. S., 87
Hwang, W., 87
B
Bakis, C. E., 5, 66
Beckwith, S. W., 373 K
Benzeggagh, M. L., 251 Kochhar, N. K., 296
Berg, M., 29 Kriz, R. D., 150
Bradley, W. L , 118, 201
C L
Chamis, C. C., 137,338 Lagace, P. A., 326
Chan, W. S., 270 Lee, J. W., 19
Chinatambi, N., 45 Leewood, A. R., 356
Christoforou, A. P., 373 Liu, D., 387
Corleto, C. R., 201
D M, O
Dan-Jumbo, E., 356 Mall, S., 296

Daniel, L. M., 19 Malvern, L. E., 387
Davies, P., 251 Murri, G. B., 222
DeCharentenay, F.-X., 251
E, F P
Peters, P. W. M., 103
Elber, W., 118
Poursartip, A., 45
Foral, R. F., 313 Prel, Y. Y., 251
G R, S
Gerharz, J. J., 29 Reifsnider, K. L., 66
Gifford, S. K., 181 Russell, A. J., 162
Gilbreath, D. R., 313
Ginty, C. A., 338 Saeger, K. J., 326
Glessner, A. L., 181 Salpekar, S. A., 222
G6kg61, O., 29 Simonds, R. A., 5
Grady, J. E., 137 Stinchcomb, W. W., 5, 66
Guynn, E. G., 118 Street, K. N., 162
407
Sun, C. T., 356,387 Wang, A. S. D., 270

Swanson, S. R., 373
T, V, W, Y
Takemori, M. T., 181 Yaniv, G., 19
Yih, H. R., 66
Vallance, M. A., 181 Yun, K. T., 296
Subject Index
A B
Absorbed moisture Beam theory, 229,230 (See also Linear beam
effect on epoxy resin matrix material, 29 theory)
Adhesion, interracial, 197 Behavior of materials
Adhesive layers graphite fiber composite laminates, 5-18
effectiveness in delamination suppression, Bend test
363 cylindrical, 397
wedge-driven delamination, 196, 197 of damaged specimens, 388
Adiabatic thermal emission, 68, 78-80(figs), residual load relations, 364
81 Bending stress, 141, 143(fig), 238
AFRP (see Aramid fiber-reinforced plastics) Birefringence distribution, 76-77(figs)
Airbus fin box program Bismaleimides (BMI), 356, 357(tables)
candidate material systems, 34 Block-loading sequence, notched
environment effect laminates, 47(fig), 54, 56(figs). 57
on life, 39(fig) BMI( See Bismaleimides)
on residual strength, 38(fig) Brittle bismaleimides, 356, 357(tables)
fatigue life results, 40 Brittle composite systems, 299
loading sequence, 31 Brittle epoxy, 387-389
quasirealistic loading, 33, 41(table), 42 Brittle matrix (epoxy) composites
residual strength, 40, 41(table), 42 tests to determine behavior, 6
Aircraft structures Brittle systems
design of composite materials for postim- bismaleimides, 256, 257(tables)
pact strength, 326 thermoplastic composites, 256,257(tables)
Airframe structures unidirectional graphite/epoxy composites,
design, 29-44 203,240,248
Aluminum cylinders, 377-378 Brittle thermoplastics, 356, 357(tables)
Anisotropic materials Buckling
temperature changes linked before and compression failures, 119, 121
after deformations, 69 compressive stress, 138, 162
Analytical methods out-of-plane, 164, 170-172(figs)
to model damage mechanisms, 137 Buckling--micrographs, 129-133(figs)
Aramid fiber-reinforced plastics (AFRP),
1104 C
ASTM Standards
D 2344-76, 34d Cantilever beam enclosed notch (CBEN)
Autoclave curve cycle specimens, 253, 254(fig), 256, 258(table),
for CFRP material systems, 105(table) 260(fig, table) 261(table)
Axial moduli, 238(tables) tests, 266-267
Axial strain, Kevlar expoxy specimens Carbon/epoxy cylinders
vs axial toad, 322(fig) impact energy, 357(fig), 381(figs)
409
Carbon/epoxy cylinders--Continued Compression failures

loading, 380 composite laminates, 118, 120(fig), 121,
material properties, 375(table) 137
Carbon fiber composites notched composite laminates, 121
fracture behavior, 182, 185, 190 open-hole composite laminates, 118
PEEK laminates--fracture surface, 193, Compression loading. 128
195 Compression stiffness. 14, 119
reinforced plastics (GFRP), 104 Compression strengths
Carbon/epoxy prepregs--specimens, 33 open-hole composite laminates, 126(table)
Carbon fiber-reinforced plastics, 1160 Compression tests
(CFRP), 103 open-hole composite laminates. 119, 123,
CBEN (See Cantilever beam enclosed notch) 124(fig)
Center-notched graphite/epoxy laminates, 66 PEEK materials, 8
CFRP (See Carbon fiber reinforced plastic) Compressive fatigue, 343(fig)
Cleavage test Compressive forces, 5, 30-31
glass/epoxy composites, 251 Compressive life and durability
Coating materials, photoelastic graphic fiber epoxy, 342(fig)
index of refraction Compressive strength compared to
stressed and unstressed, 68 residual strength, 43(fig)
Compliance calibration Compressive stress, 121-122
interlaminar shear fracture toughness, 230, Compressive stress vs crush-zone size
235,236-237(tables). 242-243,248 open-hole composite laminates, 128(fig)
mode I and mode II delamination, Computer program
258(table), 265(table) algorithm, 147
Composite cylinders to predict delamination buckling, 137
models for interpreting lateral impact, 384 Constant strain triangles, 154
Composite laminates Contact behavior and dynamic impact re-
behavior, 251-252,339-355 sponse
compression buckling, 137-149 vs impact resistance properties, 364, 372
cyclic delamination growth, 296 Contact law, 356, 365,366(table)
delamination buckling, 137-149 Contact stresses
delamination characteristics, 270 impact of pressure vessels, 374
for aircraft structures, 326 Coupon specimens--form
material properties, 299(table), 326 Airbus fin box program, 32(fig)
predictive methods for structural durabil- Crack closure, 162
ity and damage tolerance, 341,353 Crack density, 20, 21(fig), 22
structural durability and damge toler- Crack extension, 146, 209
ance-predictive methods, 339-355 Crack extension force. 181
Composite materials Crack formation, 107
compression failures, 122 Crack growth behavior
damage initiation, 66 graphite/epoxy composites, 209
fatigue life, 87 mode I--wedge driven delamination de-
Kevlar/epoxy plates, 387 sign, 190, 193, 198
matrix toughness, 5 mode II fractography, 216-219(figs)
open-hole laminates--material failures, 122 Crack growth rates
transverse cracking, 103 constant crack tip position, 188
Composite pressure vessels, 373 finite-deformation elasticity, 185
Composite structure certification, 29 interlaminar fatigue, 168-176(figs)
Composite structures interlaminar sheer fracture toughness, 240-
damage tolerant materials, 5 242
Compression, 162, 170(fig), 171(fig) notched laminates, 51-55
Compression behavior--macroscopic, 125, Crack length, 227
126(table) Crack path tortuosity, 193, 195
Compression buckling, 137 Crack propagation, 202,203
SUBJECT INDEX 411
Crack tip damage bend test, 388, 397

effect on fracture stress, 327 by low-velocity impact, 356
Crack tip process zone, 188, 196-197, 209 failure prediction, 384
Crack velocity, 184 fiber fractures, 21
Cracked-lap-shear (CLS) fatigue model geo- in composite laminates, 19, 23, 27
metrics, 46, 164-165(figs) in notched laminates, 66
Cracked-lap-shear specimens initiation and growth, 66, 81
dimensions, 298(fig) kinking of fibers, 135
strain energy release rates, 300(fig) life prediction, 87
Cracking formation in transverse plies, 103 matrix cracks, followed by delamination,
Cracking, premature, 114-115 80
Cracking, transverse, 320, 348 micrograph studies, 129-131(figs)
Cracks, 29-30 modes II delamination, 217
Critical energy release rate, 204, 209 models, 19, 87, 90-96(equations, tables),
Critical strain energy release, 6 119, 120(fig)
Cross-ply composite laminates modes, 326, 335
fatigue life, 19 notched laminates, 66
Crush-zone size. 128(fig) open-hole composite laminates, 123, 135
Cure cycle for CFRP material systems, 105 rotation and kinking of fiber, 135
Cyclic delamination behavior, 296, 304 theory, 87
(table), 307, 308(fig) tolerance, 255,271,297
Cyclic forces, 5, 15 tolerance analysis, 298
Cyclic loading tolerance of fiber composites to aerospace
delamination resistance, 244, 248 environments, 339
durability and damage tolerance, 352 tolerant materials, 5,339
free edge delamination, 283 under cyclic tensile loading, 19
graphite fiber composite, 6 Damage mechanism models, 137
notched laminates, 66 Damage tolerance, 201
Cyclic shear delamination, 227(fig) DCB (See Double-cantilever beam)
Cyclic shear precrack, 227(fig) Defects in composites, 353
Cyclic stress levels Deflection
normalized vs PEEK materials, 8, 9(fig, compliance calibration curves, 235
table), 10, 16(figs) glass/epoxy vs Kevlar/epoxy plate, 393
normalized vs specimens cycled to failure, of split laminate by frictionless roller,
9(table), 10(fig), 16(figs) 184(fig)
Cylinder tests Deformations
manufacture of specimens influenced by changing elastic properties,
experimental procedure, 329(fig) 70
material properties, 328(fig) model of cantilever deflection, 182
quasi-isotropic laminates, 328(table) under impact loading. 142
test matrix for cylinders, 330(table) Degradation of composite materials
Cylinders models, 94-97(equations, table)
carbon/epoxy compound material system residual strength, 87
elastic properties, 374, 375(table) stiffness, 401
impact energy, 375(fig) under cyclic loading, 91
pressure vessel impact stresses, 374 Delaminating loads
Cylindrical bending tests, 397 response of composite materials, 270
Delamination
adhesive layers, 363
D at crack intersections, composite lami-
nates, 19
Damage bismaleimides and brittle thermoplastics,
accumulation, 19 356
affect on stress distribution, 80 composite cylinders, 373-385
Delamination--Continued Delamination resistance

crack propagation, 383 interlaminar shear fracture toughness and
effect of impact energy, 380, 381(fig) fatigue, 244
extension, 382 mode II fracture toughness, 202
fatigue loading, 49(fig) of thermosetting and thermoplastic com-
following matrix cracks, 80 posites, modes I and II, 266, 309
following rotation and kinking, 135 Design load levels for composite airframe
glass/epoxy under static and fatigue con- structures, 40
ditions, 270, 276(fig), 335 Destructive evaluation techniques, 7
growth under fatigue conditions, 163,164, Displacement behavior, 243
168-176(figs) Double-cantilever-beam (DCB)
impact damage, 356 interlaminar fracture toughness, 181-182
Kevlar/epoxy plates, 387,394 loading arrangements, 252
lengths determined using dye-penetrant X- mode--Syy stress, 212
ray radiography, 47 model geometrics, 46
low-temperature edge stresses, 150, 160 specimen dimensions. 253(fig), 297(fig)
mode of failure, 356 specimens, 252, 253(fig), 297(fig)
notched laminates, 45-50, 51(fig) Double cracked lap-shear specimen. 164-
open-hole composite laminates, 132 165(figs)
orthotropic specimens, 76-80 Ductility, 103
pressure vessels, 374 Dugdale model, 118, 122(fig), 133-135(figs)
quasi-isotropic specimens, 77-80 Durability design criteria for composite
resistance ability of laminates, 5, 14-15, structures, 297. 299
17 Dynamic buckling analysis, 138, 139(fig)
shear fracture toughness, 224, 229, 240- Dynamic delamination buckling, 138
242,244-245 Dynamic displacement, 142
X-ray radiographs, 49(figs) Dynamic failure strain, 356. 369
Delamination behavior Dynamic loading effect, 97
cyclic growth in graphite fiber, 296 Dynamic response behavior
thermosetting and thermoplastic compos- composite cylinders, 356, 364, 367-
ites, 251 371(figs), 376(fig)
Delamination buckling
analysis, 138, 145-146 E
fatigue crack growth rates, 162-173 Edge delamination, 6.63
instability point, 146(fig) Edge delamination growth
under impact loading, 137 vs load, 275(fig)
under static and fatigue loading condi- Edge delamination layups, 244
tions, 137 Edge delamination tension tests, 274
Delamination damage Edge-notch flexure (ENF) model geomet-
low-impact speeds, 389 rics, 46
Delamination failure Edge stresses
in filament-wound composite tubes, 314, at low temperatures, 160
318(fig), 319(fig) thermal/mechanical loads, 150, 152-154,
matrix toughness effect, 297 156-159(figs)
Delamination fracture toughness, 209 Elastic modulus of composites, 88, 91
Delamination growth, 244, 246-247 Elastic properties
Delamination growth damage modes, 46 carbon/epoxy. 374, 375(table)
Delamination growth position/direction, 48 influence on deformations, 70
Delamination growth rates woven laminates at low temperatures,
glass/epoxy laminates, 273,278-294(figs) 154-155
graphite fiber composites, 303(fig), 305 Elastic thermal properties, 151(table)
holes in carbon fiber, 45 Elevated temperature, testing. 317
notched laminates, 51-56(figs), 64-65 ELS (See End-loaded split laminate)
under cyclic loading, 297,308 ENCB (See End-notched cantilever beam)
SUBJECT INDEX 413
End-loaded split laminate (ELS) test, 202 Failure, 87, 109-110(figs), 118
(fig), 203-204, 206, 210(table), 211 Failure analysis
(fig) cylinder laminate, 378, 379(table)
End-notched cantilever beam (ENCB) tests unidirectional graphite/epoxy, 201
specimens, 253(fig), 254(fig), 258(table), Failure mechanisms (See also Dynamic fail-
261(table), 262-263(figs), 267, ure strain)
263(figs) delamination, 313,317
End-notched flexure (ENF) test in fiber-reinforced laminated composites,
mode II delamination, 202(fig), 203, 297
204(fig), 206, 208(table), 209(fig) inside surface of tube, 320, 321(fig), 323,
shear fracture toughness, 224 324(figs)
static ENF data, 239-241(tables) load levels, 320, 322(figs)
test procedures, 226 radial stress distribution, 320
Energy conservation equation, 182 Failure process in structural materials, 1
Energy release rate--crack extension, 205 FALSTAFF (See FAtigue Loading STAn-
ENF (See End-notched flexure) dard For Fighter Aircraft)
ENSTAFF (See Environmental loading stan- Fatigue
dard) composite airframe structure, 29
Environmental conditions composite materials, 66. 222, 241-
glass/epoxy laminate testing, 295 242(tables), 243
m fatigue and residual strength testing, 33 crack propagation, PEEK, 257(fig)
in proof testing, 29 damage model and life prediction, 87
Environmental effects, 29 cross-ply composite laminates, 19
Environmental fatigue loading, 43 cycles, power function, 88
Environmental loading standard (EN- glass/epoxy laminates, 270
STAFF, ENvironmental FAL- graphite fiber composites, 5,296
STAFF), 43 loading, 252, 266-267, 345
Epoxy-based fiber composites modulus degradation, 89
impact damage, 356 smooth and notched composites, 257(fig)
Epoxy composites stress of composites, 345,353
fiber--susceptible to impact damage, 256 testing, 90, 266, 271
laminate material systems, description, Fatigue crack growth, 162, 169(fig), 172-
357(tables) 173(figs)
materials and properties, 252(table) Fatigue crack propagation, PEEK, 257(fig)
Epoxy matrix materials Fatigue cycles, 240
fatigue evaluation, 29, 33 Fatigue damage
tests to determine behavior, 6 degradation rate, 88
Epoxy resins, 356 growth, 71
Equations--energy release rates in cross-ply laminates. 19
compared with model geometrics, 46, 51- in notched laminates, 45, 67, 71(fig)
54, 57, 60, 62-64 influence on laminate integrity, 8
Euler-Bermouli slender-beam analysis, 182 initiation, 71, 72(fig)
Evaluation techniques models, 88, 90-96(equations, tables)
specimen deploy, 7 Fatigue damage degradation rate, 88
specimen sectioning, 7 Fatigue damage model, 88
ultrasonic pulse-echo C-scan, 7 Fatigue delamination, 246
X-ray radiography, 7 Fatigue design load level (FDLL), 40
Extensometer, 97 Fatigue ENF data, 241-242(tables)
Fatigue growth law exponent, 163
F Fatigue life
composite airframe structure, 29
Fabric prepregs composite materials, 87
for Airbus fin box and Tornado composite cross-ply composite laminate predictions,
structures, 34 19, 25-26(figs), 27
Fatigue life--Continued PEEK, 266

failure criteria, 19 thresholds for composite materials, 228,
multistress level, 89, 98(table), 100- 242-245
101(tables) Fatigue tests, PEEK, 266
notched laminates, 79 FDLL (See Fatigue design load level)
predictions, 89 Fiber breakage
single stress level, 89, 98(table), 99(fig, ta- impact damage in brittle systems, 356
ble) Fiber bridging
Fatigue life distributions, 87 glass/epoxy specimen, 255
Fatigue life prediction PEEK, 266
equations, 90-93, 94-96(equations, ta- Fiber buckling, 121
bles) Fiber bundle pullout
multistress level, 89-93, 99-101(tables) during fracture process, 192, 195
single stress level, 87, 89-93 temperature effects, 193(fig)
Fatigue load factor, 30(fig) Fiber composites
Fatigue loading conditions Kevlar/epoxy plates, 387
dynamic buckling, 137 predictive methods for structural stability
Fatigue loading and damage tolerance, 338
durability and damage tolerance, predic- Fiber damage (See Damage)
tive methods, 345 Fiber kinking, 135
energy release rate, 63 Fiber-matrix bond strength, 104
notched laminate delaminations, 48, 58 Fiber-reinforced composite materials
(fig), 60 notched laminates, 66
thermosetting and thermoplastic compos- plastic cross-ply laminates, 103-105
ites, 252, 266-267, Fiber strain failure, 379(table), 380(table)
X-ray radiographs, 49(figs) Finite element analysis
FAtigue Loading STAndard for Fighter Air- deformation, 142(fig)
craft (FALSTAFF) delamination buckling, 141
environmental conditions grids of woven laminates, 153(fig)
humidity, 33 ENF specimens, 231,232-234(figs), 248
temperature, 33 grids of woven/nonwoven laminates,
flight load sequences, 33, 36(table) 152(fig)
loading program, 31 interlaminar fatigue crack growth, 166-
Fatigue loads, 34(fig) 167, 168(table)
Fatigue modulus, 87, 89, 91(fig), 92, 93- mode II fracture toughness, 201,207(fig),
97(equations, tables) 212, 213(fig)
Fatigue prediction, 92-97 modeling procedures, 138, 139-140(table,
Fatigue proof testing program, 31 figs)
Fatigue response tests models, 138, 154
center notched T300//5208 graphite-epoxy woven glass-epoxy laminates, 150
composites, 7 Fiber pullout, 335
secant stiffness measurements, 10 Filament-wound composite tubes, 314,323-
tough matrix composites, 6, 10, 15-18 325
Fatigue stress Finite element analysis
of composites, 345,353 filament-wound composite tubes, 314
Fatigue tests glass/epoxy laminates, 280
composite materials, 8, 9(table), 42(fig, ta- graphite fiber composites, 302-303
ble), 90, 98(table), 99 Finite element method
crack growth rates, 170 compliance calibration, 258(fig, tables),
glass/epoxy laminates, 271 259(fig), 260(table), 265(table), 267
graphite fiber composites, 302 Flexural modulus
interlaminar fracture toughness, 192, 198 graphite fiber composites, 300(table)
list of specimens, 272(table), 279 Flexural residual strength
mixed mode loading, 162, 164 Kevlar/epoxy plates, 387
notched CFRP laminates, 46 thermoplastic composite laminates, 356
SUBJECT INDEX 415
Flexural stiffness GFRP (See Glass fiber-reinforced plastics)

Kevlar/epoxy plates, 387 Glass/epoxy
Flight-by-flight loading sequence, 31 delamination, 252, 387
Flight load sequences effects of environmental conditions,
Tornado program, 36(table) 274(fig), 276(fig), 277
Forced deflection response free-edge delamination, 270
aluminum cylinder, 377, 378(fig, 393 mechanical properties of woven materials,
carbon/epoxy cylinder, 383(fig) 150
Fractography shear fracture toughness, 222
carbon fiber laminates, 182 unidirectional material properties,
epoxy composites, 255 273(table)
mode II delamination fracture, 218- Glass fiber-reinforced plastics (GFRP), 104
219(figs) Graphite/epoxy
Fracture, 137 compression failures, 118
Fracture characteristics cross-ply laminates
composite cylinders, 327 fatigue evaluation, 29
Fracture mechanics fatigue life prediction, 19, 20(fig), 26(fig)
analysis, 137, 297 stress-life curve, 23(fig), 24-25(figs)
crack propagation, 202 delamination damage, 387
crack-tip damage, 327, 332-333 matrix toughness effect, 298
cylinders, 327 mode II delamination, 201
graphite fiber composites, 296 notched laminates, 67, 69
mixed-mode loading, 163 shear fracture toughness, 222
model geometrics, 46 thermosetting and thermoplastic compos-
predicting crack growth rates, 162 ites, 252
stress, 327, 332-333 Graphite fiber composites
thermoplastic composites, 251,267 fatigue response, 10, 14, 15-18
toughness, 297 matrix toughness, 296
Fracture modes Graphite fiber notched laminates, 45
graphite fiber/epoxy laminates, 342(table) Graphite polyetheretherketone composites
Fracture strain, 108(table), 111(fig), 115 (PEEK)
Fracture strength of cylinders, 330-335 compression failures, 118
Fracture stress crack propagation, 256-257(figs)
composite cylinders, 332-333(figs) fatigue, 6
static fracture stresses for tension and fatigue crack propagation 257(fig)
compression, 342(fig), 345 material behavior, 266-267
Fracture toughness thermoplastic-based composite, 356
composite cylinders, 382 Graphite epoxy composites, 6
cyclic delamination growth, 297 Grid geometry, 154
graphite fiber composites, 6 Growth rate, 162
mixed mode, 245(fig) Growth model, 270
testing--interlaminar, 185-188
unidirectional graphite/epoxy composites,
201
Free-edge delamination growth, 271,274 !t
Friction, 230
Fringe patterns Hackle orientation, 217, 219, 220
in orthotropic laminate, 76 High temperature
in quasi-isotropic laminate, 77 effects on epoxy resin matrix material, 29
High strain-to-failure (tough) matrix systems
G effect on delamination, 326
High-velocity impact
Gas gun, 388(fig) dynamic failure strain, 369
Geometrics--woven/nonwoven laminates, Hygrothermomechanical cyclic loading,
154(fig), 160 352(fig), 353
Interlaminar fatigue
crack growth rates, 162, 190
Impact damage Interlaminar fracture, 222
composite cylinders, 373 Interlaminar fracture tests
fiber composites, 387 double-cantilever-beam (DCB) samples,
in brittle bismaleimides and thermoplastic 182
composite systems, 356 wedge-driven delamination, 182, 184
Impact damage characteristics Interlaminar fracture toughness
bismaleimides and brittle thermoplastics: adhesion effects, 197
delamination, fiber breakage and ma- carbon fiber laminates, 182, 185, 189
trix cracks, 358 environment effect, 275(fig), 278
size effect using plate panels, 363 glass/epoxy laminates, 273
Impact damage resistance, 363 graphite fiber composites, 6, 302(table),
high-velocity impact damage, 369 305(table), 308
Impact data testing materials and methods, 185-186
discussion, 394-395 Interlaminar shear fracture toughness, 224,
tables (appendix), 402-404 231,243
Impact delamination Instron type machine, 97
graphite/epoxy, 402 Irwin-Kies compliance method, 264
Impact energy Isochromatic fringe patterns, 75-78(figs), 81
absorption, 395 Isothermal deformation, 184
carbon/epoxy cylinders, 373-381
effect on delamination, 380,381(fig) K
Impact forces, 6, 137
Impact-induced dynamic buckling, 137 Kevlar/epoxy plate
Impact loading and unloading deflection, 393
dynamic delamination buckling, 137 delamination damage, 387, 392(fig)
high-velocity, 368, 371(fig) multiple-ply laminates, 390, 392(fig)
low-velocity, 368-371(figs) Kinetic energy
Impact resistance Kevlar/epoxy laminates, 389, 394(fig),
predicting composite damage, 347(fig) 396(fig)
properties of composite laminates, 364(fig) Kevlar/epoxy stiffness and strength reten-
Impact specimens tion, 399(figs)
bending stress, 143(fig) graphite/epoxy laminates, 395(fig), 396(fig)
buckled sublaminate, 141(fig) graphite/epoxy stiffness and strength re-
displacement profiles, 144(fig) tention, 398(fig)
geometry and loading, 140(fig) Kevlar epoxy specimens, 321(table) 322-
Impact testing (See also High-velocity im- 324(figs)
pact, Low-velocity impact) Kinking of damaged fibers, 135
of composite cylinders, 374
PEEK, 359-362(figs), 363 L
Torion, 359-362(figs), 363
In-plane stress distribution, 316, 320 Lamina properties
In-plane Tsai Hill theory, 323 predictive model for damage, 19
Interfacial adhesion, 197 Laminate defects, 341,344(fig), 345
Infrared optics Laminate static fracture data, 340(fig)
Interface, 103 Life, 5
Interlaminar crack growth behavior Linear beam theory, 204, 207,229
crack tip process zone, 196-197 Linear elastic fracture mechanics, 163,204,
double cantilever beam (DCB) testing ge- 207
ometry, 195 Linear failure criterion
fracture toughness, 195 for delamination, 246
PEEK/CF laminate results, 195 Linear finite element analysis, 166-168
unidirectional composites, 192 Linear laminate theory, 26, 341
SUBJECT INDEX 417
Linear load displacement behavior, 243 Materials

Linear regression analysis of compliance cal- failures in open-hole composite laminates,
ibration data, 236(fig) 122
Load-bearing structural components, 66 Matrix cracking
Load-carrying capability bismaleimides and thermoplastic lami-
postimpact unloading tests, 363 nates, 356, 367
Load deflection curves, 209, 211(figs) in impacted glass/epoxy plates, 394
Load displacement data Matrix cracking in notched laminates
graphite/epoxy composites, 212 fatigue damage development, 45, 49-
interlaminar shear fracture toughness, 50(figs), 59(fig), 65
228(fig), 231,240, 248 fatigue damage growth, 71
open hole composites, 126-127(figs) fatigue damage initiation, 71(fig), 72(fig)
thermosetting and thermoplastic compos- Matrix cracks, 19-20, 29-30, 356
ites, 260(fig), 267 Matrix deformation during compressive
Load-induced crack closure, 162 loading, 135
Load levels for composite airframe struc- Matrix failures, 19
tures, 40 Matrix fracture strain, 115
Load, mechanical Matrix plasticization, 336
fatigue and fracture, 345,346(fig) Matrix splitting, 335
Load redistribution, 81 Matrix toughness, 5-18, 296, 298
Loading Mechanical and thermal loads
carbon epoxy cylinders, 380 stress distribution, 156
geometry of cylinder, 376, 377(fig) Mechanical loading (See Loading programs)
lateral, of cylinders, 384 Mechanical tests
Loading conditions method for reversed cyclic loading, 6
crack growth rates, 162 Methodology and materials
fatigue and matrix toughness, 6 interlaminar fracture toughness testing,
glass/epoxy laminates, 271 185-186, 187(fig)
graphite fiber composites, 297,301 Microcracks, 216, 217,266, 348
predictive methods for durability and dam- Micrograph studies, 129-131(figs)
age tolerance, 352(fig), 353 Midplane displacement, 141, 143-144(figs)
Loading impact, 137 Midplane stress distribution, 156(fig), 159(fig)
Loading programs Midplane stresses
evaluation of airframe structure, 31, 35 at low temperatures, 159. 160
fatigue and fracture, 346 woven and nonwoven glass-epoxy, 152
mechanical, cycling, 352 Mixed mode
Low load fatigue tests, 46 cyclic delamination growth rate, 305(table)
Low temperatures, 150, 159, 160 delamination growth rates, 306(fig)
Low velocity impact damage edge delamination tension tests, 244
bismaleimides and thermoplastic compos- failure criterion, 248
ite, 356, 364(fig), 367-371(figs) fracture toughness, 245(fig)
composite cylinders, 356 interlaminar fatigue, 162, 164-165(figs)
delamination, 388-390 tests, 254
in brittle composite systems, 356 delamination initiation, 264
Kevlon/epoxy plates, 389 linear elastic fracture, 264
strain energy release rate. 264
M stress intensity factors. 264
Mode I
Macroscopic compression behavior, 125 delamination, 266-267
Mar-Lin equation double cantilever beam specimens, 252
prediction of failure to pressurized cylin- double cantilever beam test, 212
ders, 331-334 fatigue tests, 253,256, 257(fig)
Material properties fracture energies, 255(table), 267
carbon/epoxy cylinders, 375(fig) graphite/PEEK crack length. 256(fig)
Mode I--Continued Photoelasticity

static, 255 coatings, 68, 69, 75(fig)
test methodology, 252 Plane strain problem, 151
Mode II Plane stress analysis, 234
delamination, 215(fig), 217(fig) Plasticization, 335
delamination growth rates, 307(fig) Plastics, fiber reinforced, 103-105
fatigue tests, 254 Ply stresses, 341
fractography, 218-219(figs) Polyetheretherketone (PEEK) laminates
graphite/epoxy composites, 266-267 brittle bismaleimides, 356-357
loading, 203 compressive strength test results, 13(table)
shear delamination tests, 253 Imperial chemical industries--America
shear fracture toughness, 222 system. 185
static, 255,256 interlaminar fracture toughness, 190
strain energy release rate, 235(fig) material description, 357(tables)
test configuration, 202(fig) matrix fracture strain, 115
test methodology, 252,254 matrix toughness, 300
Mode III residual strength test results, 13(table),
tearing, 297 14(fig)
Model geometrics, 46 shear fracture toughness, 222
Modulus degradation, 87 stiffness curves, low cyclic stresses vs high
Moisture, 31.45 cyclic stresses, 14
Monotonic loading conditions, 6 stiffness ratio vs normalized life plots, 10-
11, 12(figs)
N test methods, 6
thermoplastic-based systems, 356,
NASTRAN DMAP, 138, 139(table) 357(tables)
NDE (See Nondestructive examination) thermoplastic composite laminates, 356-
Neat resin tensile properties, 320 357(tables)
Newman's analysis for cracks. 121(fig) Torlon composites, 357(tables)
Nondestructive examination(NDE), 132-133 Victrex PEEK, 122
Nondestructive evaluation, 7, 66 Postimpact delamination
Nonlinear behavior, 231 Kevlar plate, 388, 390, 391(fig)
Nonwoven laminates--damage, 150 unloading tests, 363,367-368
Notched strength, 326 Precracking test procedures, 226, 227(fig)
Predictive methods and models
O
for structural durability and damage tol-
Open hole composite laminates, 118 erance of composites, 339-355
compression failures of materials, 122 Prepreg materials, 34
compression strengths, 126(table) Pressure load--radial, 376
load displacement data, 126-127(figs) Pressure vessels
Orthotropic behavior, 316 potential for impact damage, 373
Orthotropic specimens strength loss due to impact, 374
isochromatic fringe patterns, 78 Pressurization
photoelastic coatings, 74-77(figs) test procedure for cylinders, 328-331
residual and compressive strength, 81 Pressurized cylinders, 326
tensile and compressive strengths, 73(fig),
74(fig) Q
Out-of-plane displacements, 171, 172(figs)
Overload, 45 Quasi-isotropic construction of laminates, 356
Quasistatic specimens
photoelastic coatings, 75(fig)
isochromatic fringe measurements, 75,
PEEK (See Polyetheretherketone) 76-77(figs)
Photoelastic coating technique, 81 residual and compressive strength, 81
SUBJECT INDEX 419
tensile and compressive stiffness, 70, 71 Shear deformation. 76, 121

(fig), 74(fig) Shear fatigue thresholds, 224
thermography, 78(fig) Shear lag analysis, 106(fig), 107
Shear strength, 222
Shear stress, 107, 201
R Short beam shear (SBS) test, 224
Radial pressure load, 376 Simplistic loading
Radial stress distribution, 318, 319(figs), 320, Tornado program, 31
321(tables), 322-323(figs) SPATE (See Stress pattern analysis by mea-
Radiography, 67.71-72(figs) surement of thermal emission)
Rebound kinetic energy. 389 Specimen geometry
Refraction index of photoelastic coating ma- composite materials, 7(fig)
terials, 68 notched laminates, 47(fig)
Residual load resistance Stacking sequence
bend test, 364 effect on interlaminar fracture, 279
vs impact velocity, 363 effects, delamination failure modes, 313,
Residual properties, center-notched lami- 317(table), 321(fig)
nates, 73(table), 80 on laminated Kevlar/epoxy plates, 388
Residual static tests, 229, 240-241(tables), Static-edge delamination growth
243(figs), 245(fig), 247 vs load, 275(fig)
Residual strength Static force deflection, 376(fig)
bend test, 364 Static indentation data, 366-371(figs)
composite airframe structure, 29-30, 42 Static linear model
composite cylinders, 373 impact response for empty cylinders, 375
degradation model, 87 Static loading
impact resistance, 356 delamination
nonlinear equation, 87 buckling, 146
notched laminates under reversed cyclic notched laminates, 48, 63
loading, 66, 80 X-ray radiographs, 49(figs)
tests, 31 energy release rate, 63
Residual toughness tests, 241 shear fracture toughness, 244
Reversed cyclic loading. 6, 67 Static shear delamination, 227(fig)
Rocket motor cases Static shear precrack, 227(fig)
impact tests of cylinders to simulate actual Static tension, 8
service conditions, 374 Static testing, 90
Rotations, 376 Static tests
interlaminar fracture toughness, 302
(table), 306
interlaminar shear fracture toughness,
228-229, 239, 240-241(tables). 243,
SBS (See Short beam shear) 247
Scanning electron micrographs, 194(fig), 208 mode II, 256
Secant modulus failure criterion, 88(fig) specimens, 271(table)
Secant stiffness measurements, phases Stereomicroscope, 123(fig)
normalized life, 10 Stick/slip crack growth, 181, 199
normalized stiffness, 10 Stiffness
notched laminates, load cycle, 70(table) composite cylinders, 375-376
tension-compression-stiffness-versus-life, composite laminates, 7, ll(fig)
10(fig) degradation, 81
Sectioning studies, 125(table), 127 fatigue damage in notched laminates, 45,
Sectioning studies--specimen micrographs, 47, 57-59(figs). 67
130-131(figs) fiber-reinforced plastic cross-ply lami-
Servohydraulic test machine, 271 nates, 103
Shear crippling, 121, 127, 131-132(fig) Kevlar/epoxy plates, 388, 397
Stiffness--Continued Stress concentration factors, 341

notched laminates--tensile and compres- Stress criterion
sive strengths, 70(tables), 73(table), effect on fatigue endurance, 346(fig)
74(fig, table) Mar-t~in vs Whitney-Nuismer, 332-335
Stiffness degradation, 19, 81,401 material comparisons, 335(table)
Stiffness matrix, 316 Stress distribution
Strain, 87 ahead of crack tip
Strain distribution, 66, 75, 80-81 mode I conditions, 212,213(fig)
Strain energy release rate mode II conditions, 214(fig)
composite cylinders, 382 graphite/epoxy composites, 207
crack-lap-shear (CLS) specimens, 300(fig), woven laminates, 154-155, 156-159(figs)
303 Stress intensity factors, 162
cyclic delamination behavior, 304(fig), 307 Stress levels
delamination buckling, 146-147(figs) cyclic, 20(fig)
delamination fracture testing, 203 Stress-life properties (S-N curves), 19, 24-
graphite fiber composites, 6 27(figs)
interlaminar fatigue crack growth, 162, Stress pattern analysis by measurement of
163, 168(table), 169 thermal emission (SPATE), 68-69
interlaminar fracture toughness, 181 Stress prediction, 316, 320, 323
mixed mode tests, 264, 265(table) Stress/strain
mode I and mode II during cyclic loading, 66, 91
delamination of thermoplastic compos- failure criterion, 89(fig)
ites, 257, 260 secant modulus failure criterion, 88(fig)
graphite fiber composites, 296 SPATE 8000 (Stress pattern analysis by
mode II ENF test, 235(fig) measurement of thermal emission) in-
notched laminates, 45, 46 strument, 68-69
shear fracture toughness, 212,224, 229 Structural components
Strain failure criterion, 88, 89(fig), 93 under long-term cyclic loading, 66
Strain magnification, 103, 104(fig) Structural degradation of impacted plates,
Strain measurement techniques, 67-68 397
Strain response (See Dynamic strain) Structural durability of fiber composites in
Strain-to-failure matrix system, 336 aerospace environments, 339
Strength Struers accutum precision saw, 125
airframe structures, 31 Subperforation speeds (See Low-impact de-
degradation rate, 88 lamination damage)
graphite fiber composite laminates, 5
notched laminates, 70(tables), 73-81(figs,
tables) T
Strength and stiffness reduction
glass/epoxy plates vs delaminated area, Temperature
400(fig, table) effect on epoxy resin matrix material, 29
pressure vessels, 374 loading programs, 31
Strength and stiffness retention, 388, 397 Temperature and loads--sequences, 34(fig)
Strength distribution, 106, 115 Temperature dependence
Stress fracture toughness testing, 189(figs), 199
bending, 141 Temperature linked with deformations of
composite materials, 87 elastic matter, 68-69
crack density, 22 Temperature profiles
crack growth rate, 51-53, 62, 63 flight load sequences, 36(table)
cross-ply laminates, 20-27 Temperature tests
distribution, 80, 106-107 delamination failure modes. 313, 317
notched laminates, 67 (table), 318-319(figs), 320
temperature effects, 29 Tensile and compressive strengths, 70
Stress concentration equations, 343,344(fig) (tables), 73, 74(fig, table)
SUBJECT INDEX 421
Tensile delamination stresses, 159,201 vs thermoset resin in interlaminar fracture

Tensile forces, 5 testing, 198
Tensile fracture process, 81 Thermoplastics, 5
Tensile properties, neat resin of epoxy resin Thin-walled composite tubes
system, 320 interlaminar stresses, 315
Tensile static tests on notched CFRP lami- loading, 314
nates, 46 stresses and strains, 315
Tensile strength Tinius Olsen servohydraulic test machine. 271
use to predict transverse fiber pullout, 324 Torlon
Tension. 239 impact damage, 356, 357(tables), 358
Tension-compression cyclic loading Tough matrix vs brittle matrix materials
tests to identify damage and failure, 6 impact resistance, 364
Tension tests Toughened BMI matrixes, 356, 357(tables)
edge delamination, 274 PEEK, 358,360(figs)
Test methods for Mode I Transverse sheer stress, 367, 376
double cantilever beam test, 6 Tornado program
edge delamination test, 6 candidate material systems, 34, 35(fig)
Test methods environment effect
compression tests for open-hole composite on life, 39(fig)
laminates, 123 on residual strength, 38(fig)
response of composite materials, 87 fatigue design load levels, 41(table)
under reversed cyclic loads, 7, 67 quasirealistic loading, 33, 41(table)
Test procedures simplistic loading, 31, 41(table)
cylinders, 328-331 temperature profile and flight load se-
Tornado program, 37(fig) quences, 36(table)
Testing test setup, 35
interlaminar fracture toughness materials wing loading sequence. 31
and methods, 185-186 Tough matrix systems
shear fracture toughness, 226 cyclic delamination growth, 300
Tests pressurized composite cylinders, 326
crack growth rate/stress range, 51 resistant to lamination driven fracture. 17
Thermal cracks, 29-30 tests to identify behavior, 6
Thermal cycles Transply cracking. 348
predictions for laminates. 348,350(table) Transverse cracking
to failure. 351(figs) carbon fiber-reinforced plastics, 105, 111
to initial transply cracking, 350(table) fatigue life of composite materials. 19.
Thermal cyclic loading. 352 20(fig), 27
Thermal cycling. 29 Transverse fiber pullout, 314. 320. 321(ta-
Thermal emission, 68.78-79(figs), 81 ble), 322(figs), 324
Thermal expansion coefficients, 107 Transverse fracture strain, 115
Thermal fatigue, 348 Transverse matrix cracking. 320. 323
Thermal loading (See Loading programs) Transverse shear, 230, 238-239
Thermal loads--stress distribution, 156, Transverse shear stress, 367,376
157(figs) Transverse strength
Thermal ply transverse stress fiber-reinforced plastics, 103-104, 105. 112
at cryogenic temperatures, 349 Tsai-Hill theory
Thermal strain, 109, ll3(fig) in-plane stress prediction, 316, 320, 323
Thermally induced stress
combined with magnetic loads. 150, 156 U
Thermography, 78(fig). 80
Thermoplastic-based composites Unidirectional carbon tape, 185
PEEK, 356, 357(tables), 372 Unidirectional composite materials
Torlon, 356, 357(tables), 372 delamination, 252
Thermoplastic resin composites glass epoxy, 273(table)
Unidirectional composite materials--Con- PEEK/CF interlaminar fracture data, 190-

tinued 191(figs), 199
interlaminar fracture toughness, 273 testing design, 182, 183(fig)
material properties, glass/epoxy, 273(table) testing materials and methods, 185-189
mechanical properties, 273 Whitney-Nuismer average stress criterion
Unidirectional graphite/epoxy composite correlations for strength of notched
delamination on fracture toughness, 201- coupons, 332-334
221 Woven composites, 150
Weibull distribution
V fatigue evaluation, 40
fracture strain distribution, ll0-111(figs),
Velocity measuring system, 388(fig) ll2-113(figs), ll6(fig)
Voids, 114-115 transverse cracking, 103, 105, 108(table),
109(fig)
W
Wild M8 zoom stereomicroscope, 123(fig)
Wing loading sequence (FALSTAFF), 31
WDD (See Wedge driven delamination)
Wedge-driven delamination X
carbon fiber composites, 182 X-ray radiography, 67, 73, 81

STP1012 Eb.1415051 1 PDF

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

STP1012 Eb.1415051 1 PDF

Diunggah oleh

Hak Cipta:

Format Tersedia

Copyright by ASTM Int'l (all rights reserved); Mon Jan 16 20:12:47 EST 2012

Paul A. Lagace, editor

1916 Race Street

Copyright 9 by AMERICAN SOCIETY FOR TESTING AND MATERIALS 1989

Peer Review Policy

Printed in Ann Arbor. MI

FATIGUE AND DAMAGE GROWTH

Effects of Matrix Toughness on Fatigue Response of Graphite Fiber Composite

Fatigue Life Prediction of Cross-Ply Composite Laminates--J. w. LEE,

Consideration of Environmental Conditions for the Fatigue Evaluation o f

Fatigue Damage Development in Notched (0,/---45), Laminates--A. POURSARTIP

Damage Initiation and Growth in Notched Laminates Under Reversed Cyclic

MODELS AND ANALYSIS

Fatigue of Composite Materials--Damage Model and Life Predictionmw. HWANG

The Influence of Fiber, Matrix, and Interface on Transverse Cracking in Carbon

Micromechanics of Compression Failures in Open Hole Composite Laminates--

Dynamic Delamination Buckling in Composite Laminates Under Impact Loading:

Edge Stresses in Woven Laminates at Low TemperaturesmR. D. KRIZ 150

Predicting lnterlaminar Fatigue Crack Growth Rates in Compressively Loaded

Mode I Interlaminar Fracture Toughness of Unidirectional Carbon Fiber

Mode II Delamination Fracture Toughness of Unidirectional Graphite/Epoxy

Interlaminar Shear Fracture Toughness and Fatigue Thresholds for Composite

Mode ! and Mode II Delamination of Thermosetting and Thermoplastic

Free-Edge Delamination Characteristics in $2/CE9000 Glass/Epoxy Laminates

Characterization of Matrix Toughness Effect on Cyclic Delamination Growth in

Delamination Failure Modes in Filament-Wound Composite T u b e s - - R . F. FORAL

Fracture of Pressurized Composite Cylinders with a High Strain-to-Failure Matrix

Fiber Composite Structural Durability and Damage Tolerance: Simplified

Impact Damage Characteristics of Bismaleimides and Thermoplastic Composite

Lateral Impact of CompositeCylinders--A. P. C H R I S T O F O R O U , S. R. S W A N S O N , A N D

Oelamination Damage in Central Impacts at Subperforation Speeds on Laminated

This volume is a collection of peer-reviewed papers based on presentations made at the

Effects of Matrix Toughness on Fatigue

ABSTRACT: To evaluate the possible advantages of using a tough resin, polyetheretherketone

KEY WORDS: composite materials, thermoplastics, polyetheretherketone, fatigue, damage,

1 Laboratory engineer, research associate, and professor, respectively, Department of Engineering

Material and Specimens

FIG. 1--Specimen geometry.

FIG. 2--Specimen gripping arrangement.

Results of Static Tension and Compression Tests

Fatigue Test Results

TABLE 1--Fatigue test results.

Material Cyclic Stress, Percentage of Monotonic

PEEK 207 (30.0) 71 2 410

a Load cycling halted before fatigue failure.

TABLE 2--Residual compressive strength test results.

Material Normalized Normalized Normalized

PEEK 0.440 1 000 000 0.910 0.89

" Normalized to 290 MPa (42.0 ksi) unfatigued compressive strength.

Results of Residual Strength Tests

TABLE 3--Residual tensile strength test results.

Material Normalized Normalized Normalized

PEEK 0.445 1 000 000 0.920 1.14

strength) and residual compressive strength (normalized by monotonic compressive strength).

Fatigue Life Prediction of Cross-Ply

ABSTRACT: Fatigue damage development in cross-ply laminates consists of transverse matrix

Prediction of CDS Life

E2(k) = E2(0)[1 + 13~h] (2)

The axial modulus of the laminate at a transverse crack density h is

From the equations above it follows that

log ~r2 = a(log N) 2 + b(log N) + log F2re (5)

Residual Life After CDS

log tr~ = a'(log N ) 2 + b'(log N ) + log F~re (8)