Sie sind auf Seite 1von 16

http://jcm.sagepub.

com/
Materials
Journal of Composite
http://jcm.sagepub.com/content/39/16/1417
The online version of this article can be found at:

DOI: 10.1177/0021998305050432
2005 39: 1417 Journal of Composite Materials
C. Kyle Berkowitz and W. Steven Johnson
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure

Published by:
http://www.sagepublications.com
On behalf of:

American Society for Composites


can be found at: Journal of Composite Materials Additional services and information for

http://jcm.sagepub.com/cgi/alerts Email Alerts:

http://jcm.sagepub.com/subscriptions Subscriptions:
http://www.sagepub.com/journalsReprints.nav Reprints:

http://www.sagepub.com/journalsPermissions.nav Permissions:

http://jcm.sagepub.com/content/39/16/1417.refs.html Citations:

What is This?

- Jul 26, 2005 Version of Record >>


at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
Fracture and Fatigue Tests and Analysis
of Composite Sandwich Structure
C. KYLE BERKOWITZ*
G. W. Woodruff School of Mechanical Engineering
Georgia Institute of Technology
Atlanta, GA 30332, USA
W. STEVEN JOHNSON
School of Materials Science & Engineering and
G. W. Woodruff School of Mechanical Engineering
771 Ferst Dr., Georgia Institute of Technology
Atlanta, GA 30332-0245, USA
(Received March 26, 2004)
(Accepted October 13, 2004)
ABSTRACT: A composite sandwich system is investigated in this research. Quasi-
static fracture toughness and fatigue crack growth experimental and analytical
approaches are the focus. The particular system studied is comprised of a Nomex
(aramid fiber) honeycomb core with graphite/epoxy facesheets (skins).
A modified version of the double cantilever beam (DCB) specimen geometry is
used for experimentation. The critical strain energy release rate, G
c
, is used to
characterize the fracture toughness of the facesheetcore joint. Fatigue crack growth
testing is also performed. Novel analytical and experimental techniques are coupled
and utilized to address challenges presented by the material system, especially
difficult crack visualization. Crack length and growth can be estimated with an
empirical approach, employing a compliance calibration. Experiments can also be
simulated once several constants are estimated, aiding design. Many of these
techniques can be generalized to other adhesive DCB experimentation.
Results show that cold tests result in higher fracture toughnesses and slightly
slower fatigue crack growth rates than room temperature tests. The hot temperature
has less significant impact. Although only a limited amount of very slow growth data
(<10
6
mm/cycle) is measured, the material appears to behave with a fatigue thresh-
old of -7% of fracture toughness; and a Paris crack growth model is successfully fit
with an exponent of -3.2. Results also show the failure of this system is always in the
aramid paper core material.
KEY WORDS: sandwich structure, debond growth, fatigue crack growth, failure
modes, temperature effects, Nomex honeycomb.
*Author to whom correspondence should be addressed. Present address: 7220 Richardson Road, Georgia Tech
Research Institute-ATAS, Smyrna, GA 30080, USA.
Journal of COMPOSITE MATERIALS, Vol. 39, No. 16/2005 1417
0021-9983/05/16 141715 $10.00/0 DOI: 10.1177/0021998305050432
2005 Sage Publications
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
INTRODUCTION
S
ANDWICH STRUCTURES ARE normally composed of high strength thin skins
(facesheets) bonded to a significantly thicker, less dense, and weaker core material.
Sandwich structures are most useful in applications that put a premium on stiffness-
to-weight ratios, such as in many aerospace applications. The focus of this paper is
a sandwich structure with graphite/epoxy facesheets and a Nomex (aramid fiber)
honeycomb core. This and similar systems appear on many commercial jets, especially as
control surfaces. These composite structures tend to develop damage through long-term
use, and this creates an enormous maintenance concern since the structural properties
are strongly dependent on the bond integrity. These debonds can occur in the film
adhesive used between the facesheets and core, or damage can occur in the core itself. The
goal of this research program is to characterize the fracture and fatigue parameters of a
particular material system, so that they can be incorporated into broader life predictions
and a damage tolerant methodology. The properties were experimentally measured, as
functions of test temperature, and novel experimental and data analysis strategies were
emphasized.
BACKGROUND
Sandwich structures have been widely applied in both aerospace and marine
applications, as well as anywhere where weight and stiffness are the driving performance
parameters. The first production parts that employed the idea of the modern sandwich
structure were produced in the 1940s, but the basic idea existed much earlier [1]. While
sandwich structures have been widely researched, only in the past two decades has
this work addressed fracture and failure issues. Most of this work has focused on
structures with foam core materials, and it has addressed both experimental fracture
toughness determination and FEA to model failure criteria and predict parametric
dependencies [2,3]. Sandwich buckling and post-buckling analysis has received much
attention as well.
Avery and Sankar [4] and Ratcliffe and Cantwell [5,6] are among the only researchers in
the open literature who have performed experimental fracture work on Nomex-cored
sandwich structures, which is a very common material in commercial aviation. Fatigue
debond growth, as in the classic Paris formulations used in metals, has virtually been
ignored or not well reported. The maintenance of commercial aircraft, which are used
at a very high capacity for decades in many cases, seems to dictate a need for better
characterization of structures of this type.
Nomex aramid mechanical paper is a synthetic material composed of aramid fibers
stabilized with a phenolic resin. Sheets are then expanded and bonded together in the
honeycomb shape. Nomex has been shown to be very resistant to heat and moisture [7,8].
Polymers properties are generally a function of temperature; these effects are often
undesirable for an application. Aerospace composites are usually exposed to large
temperature ranges that significantly affect their mechanical behavior. Johnson and
Butkus [9] examined the compound effects of temperature and environmental aging on
aerospace-bonded joints. These conditions can affect both fracture toughness and fatigue
curves. Moreover, it is important to examine how environmental factors, especially
temperature, affect aerospace composites.
1418 C. K. BERKOWITZ AND W. S. JOHNSON
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
EXPERIMENTAL PROCEDURES
Materials and Specimen Preparation
The specimens used in this research were constructed from six layers (plies) of woven
carbon fiber-epoxy prepreg, all oriented squarely to the rectangular specimens. Each
six-ply facesheet had a total cured thickness of -1.5 mm. A thin (relative to a ply) film
adhesive was placed between the prepreg and core, and it was co-cured with the prepreg.
The honeycomb core had a cell size of 3.2 mm (1/8 in.), nominal density of 48 kg/m
3
(3.0 lb/ft
3
), and thickness of 9.5 mm (3/8 in.). The prepreg and honeycomb were Hexcel
products W3F282-T60
//
-F593 and HRH-10-1/8-3.0, respectively. The adhesive was Cytec
Metlbond 1515-3M.
The specimens were made from the manufactured rectangular panels, which had the
approximate dimensions of 610 mm483 mm11.9 mm (24 in. 19 in. 0.47 in.). The
panels were constructed from their layers of constituents and vacuum bagged, before being
cured in an autoclave at 177

C (350

F) for 2 h at an absolute pressure of 4 atm. The co-


curing of the facesheets and adhesive film created a blended interface, where there was
not a clear boundary between the two. Their viscosities during curing allowed a certain
amount of mixture to occur. Since the facesheets and adhesive were cured as they were
being bonded to the core under the autoclaves pressure, the core became partially
embedded in the facesheet/adhesive layer. This manufacturing technique likely leads to a
stronger joint and certainly affects interface properties.
Each specimen, which was tested in a double cantilever beam (DCB) configuration, was
cut from a panel to a size of 152 mm102 mm11.9 mm (6 in. 4 in. 0.47 in.), with the
ribbon direction of the honeycomb aligning with the longer dimension; the 102 mm (4 in.)
width was necessary to increase the loads to a reasonable level for the load cell. Piano
hinges were riveted to the specimen to facilitate loading, and one edge was painted white to
ease visual crack measurement. A starter crack was cut with a saw between one facesheet
and the core, 20 mm past the load line; the crack grew in the ribbon direction. A top view
photograph of a specimen is shown in Figure 1 (upper half of hinge is not shown to reveal
the rivets), and a side view sketch is shown in Figure 2.
Mechanical Testing
Identical specimens were used for the fracture and fatigue testing; only the loading and
data recording schemes differed. A servo-hydraulic universal mechanical testing machine
was used, and mechanical wedge grips were used to grasp the free halves of the hinges
used to load the specimens. Since no standards existed for fracture toughness testing
of sandwich structures, ASTM D5528-94a was used only as a guideline when appropriate
[10]. The specimens were loaded via the piano hinges at a constant displacement rate of
1.27 mm/min (0.05 in./min). Loads and displacements were recorded by the testing
software, and visual crack measurements were taken. The actuator was programmed to
displace -0.5 mm (0.020 in.) past the point of nonlinearity, which produced crack growth.
Then, the specimen was unloaded, and the test was repeated 1520 times per specimen,
each giving a fracture toughness value.
Fatigue testing was performed at 4 Hz and at an R-ratio (P
min
/P
max
) of 0.1. The goal
was to obtain da/dN versus G curves for the parameters studied. Displacement control
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure 1419
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
was used because the loads were relatively small for control feedback. The load was
allowed to drop 10% before the test was stopped and restarted. At the beginning and end
of each segment of cycles, peak loads and compliance were measured for use in analysis,
and periodic visual measurements were taken.
Testing was performed at hot (77

C (170

F)), room (21

C (70

F)), and cold (54

C
(65

F)) temperatures, mimicking the extremes a commercial aircraft should see in


service. The temperature-controlled environment was created with an insulated chamber
mounted to the test machine, having both resistance heater and liquid nitrogen capabilities
and an electronic controller. Photographs of a specimen in the testing machine are shown
in Figure 3.
ANALYSIS
Fracture Toughness
The critical strain energy release rate, G
c
, was used to characterize fracture toughness
according to:
G =
P
2
2B
dC
da
, (1)
Figure 1. Photograph of DCB specimen (top view).
Core
Piano hinges Debond
P
P
11.9mm (0.47in) Facesheets
152mm (6.0in)
Figure 2. Sketch of DCB specimen (side view).
1420 C. K. BERKOWITZ AND W. S. JOHNSON
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
where P is the applied load, B is the specimen width (102 mm), C is the compliance, and a
is the crack length. The critical, nonlinear load (P
c
) corresponds to G
c
. Equation (1)
characterizes the energy consumed to create an increment of new crack area and applies to
most cases of LEFM with constant specimen width. In this case, nominal area was taken;
no consideration was given to the geometry of the honeycomb.
Since no exact solution exists for this geometry due to asymmetry and the cores high
orthotropy, dC/da was computed from a very good power law correlation that resulted
from compliance versus crack length data for a number of specimens. A plot of the data
used to develop this correlation is shown in Figure 4. This technique is often called the
Berry Method [11]. This correlation is expressed as:
C = Da
m
, (2)
where D and m are fitting parameters. An equation of this form is easily differentiable
for use in Equation (1). The correlation for the specimens used here was found to be:
C = (2:01 10
6
)a
2:69
, (3)
where C is in units of mm/N and a in mm. Equation (2) can be inverted, such that a
measured compliance can estimate crack length. Utilizing that fact, Equations (1) and (2)
can be combined as:
G =
P
2
2B
mD
C
D

1=m
" #
m1
, (4)
Figure 3. DCB specimen loaded in test machine.
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure 1421
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
which reduces G to a function of machine-measured parameters and known constants.
This implies that the manual measurement of crack length is not needed once Equation (3)
is known, since compliance and crack length are 1-to-1. This assumption is validated by
the quality of the simple model of Equation (2) in fitting the actual visual measurements.
Many other models exist to predict and correlate compliance, but a more complex model
was not warranted here.
It may be unusual to use measured compliances in lieu of measured cracks in determining
applied G. It is usually preferable to apply Equation (2) with compliance as the dependant
parameter. However, for these particular experiments, it was found that achieving
repeatable visual measurements of crack lengths and crack extensions (a) was difficult
due to the mode of failure, specimen width, and experimental apparatus. The fracture line
of the crack was difficult to characterize, although an approximate crack length was always
obtainable, allowing the compliance correlation to be developed from many specimens.
Sensitivity and repeatability are important experimental issues with fracture and fatigue
testing, due to the strong power dependencies of the resulting properties. Since compliance
measurement was found to have much higher relative resolution than crack measurement,
compliance was preferred as the input to data reduction. Moreover, compliance was much
more repeatable and reliable for use in G calculations, whereas visual crack measurements
seemed to cause unnatural scatter in the data, again resulting from the visualization issues.
Therefore, the numerical compliance calibration approach described here can be a useful
alternative in certain experimental circumstances.
Fatigue Crack Growth
CRACK GROWTH RATE CALCULATIONS
Fatigue crack growth per cycle (da/dN) can be characterized by a version of Paris
equation [12]:
da
dN
= c G ( )
n
, (5)
Compliance vs. crack length: calibration curve
C = (2.01 x 10
6
) a
2.69
R
2
= 0.965
0
0.1
0.2
0.3
0.4
0.5
10 25 40 55 70 85 100 115
Crack length (mm)
C
o
m
p
l
i
a
n
c
e

(
m
m
/
N
)
Composite visual data
from 21 DCB fracture
toughness specimens
Figure 4. Compliance calibration.
1422 C. K. BERKOWITZ AND W. S. JOHNSON
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
where c and n are fitting parameters, and G=G
max
G
min
. In this research G,
expanded in Equation (6), was approximately 99% of G
max
because an R-ratio (P
min
/P
max
)
of 0.1 was used: (P
max
)
2
(0.1 P
max
)
2
=0.99 P
max
. Therefore, there is no significant
distinction in this case between G and G
max
, as shown by:
G =
P
2
max
P
2
min
2B
dC
da
- G
max
: (6)
Evaluating G utilizing the compliance calibration technique was performed identically as
in the monotonic fracture testing, and the applied G was calculated using Equation (6).
P
max
and P
min
were obviously used instead of the critical nonlinear P
c
, which results in
quasi-static fracture failure.
Perhaps more useful than using compliance to correlate with crack length is using
compliance changes to correlate with crack growth. This is particularly valuable when
visual measurements of cracks are difficult, and it also allows for greater automation. The
use of compliance changes to calculate da/dN is fairly straightforward. Compliance was
measured frequently during the fatigue testing to ensure that incremental increases were
captured. Compliance was measured at monotonic test speeds, because this allowed for
much greater resolution. Loads were chosen to easily avoid P
c
for a given condition.
Small increases in compliance allows for estimation of small changes in crack length to
be performed. The measuring of small changes in crack length is important because it
generates more data per specimen and lessens load drop effects, which are described later
in this paper. Equation (7) can be used to express a change in crack length during a
segment of cycles, from cycle N
1
to N
2
, in terms of compliance increase from C
1
to C
2
.
a
12
= (C
2
=D)
1=m
(C
1
=D)
1=m
, (7)
where D and m are defined in Equation (2). da/dN is simply evaluated with Equation (7)
and using the machine count of cycles, N=N
2
N
1
.
It is worth noting that this technique estimates a change in crack length from a change
in compliance, not the crack length itself from compliance measurements as in the
monotonic case. This fact decreases the sensitivity of da/dN to possible crack measurement
errors from the compliance calibration expression. There are many potential numerically
more sophisticated ways of performing these calculations, although they were not
necessary for the work presented here. Numerical differentiation for these purposes are
further documented in ASTM E647-00 [13].
LOAD SHEDDING EFFECTS AND EXPERIMENTAL STRATEGY
As displacement-controlled testing results in load shedding, care must be taken to
ensure that the applied range of G is approximately constant for each da/dN data point. G
drops dramatically with load, so frequent monitoring of load and compliance is prudent;
however, it is not clear how many cycles should pass between each measurement of a
da/dN data point. If this is performed too often, machine noise could become noticeable
and interfere with the desired material property. If this is performed too infrequently, G
might reduce dramatically during segments of cycles used in da/dN calculations, which
would make the data difficult to interpret. The G at the beginning and ending of a segment
of crack should be calculated and the consequences of assumptions considered.
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure 1423
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
There is some question as to how much the G should be allowed to drop in a
displacement-controlled test before taking measurements. It is impractical to take
measurements constantly, which would be time consuming and create noise in da/dN
data. Obviously, a resolvable growth in crack length must be allowed to take place over
time as G falls. It is best to understand the falling driving force behavior and
incorporate it into the experimental strategy. The effect of changes in G can be
thought of on a loglog plot of Equation (5), which is linear with slope n. A change in
log (G) is equal to log {(G
1
/G
2
)}, where G
1
occurs at the beginning of a cycle
segment and G
2
after load is shed. If n is assumed to be 4 (this can be checked after
some data is measured), which is reasonable for many composites [9], a value of 0.0833
for log {(G
1
/G
2
)} would result in 1/3 decades of change in log (da/dN). This seems
somewhat reasonable, and scatter bands are often this large for general da/dN
data. Therefore, this amount of drop in G was tolerated before taking measurements.
Log (P
2
1 max
=P
2
2 max
) is approximately proportional to log {(G
1
/G
2
)}, so this was
allowed to reach 0.0833 before each segment of cycles was stopped for measurement; i.e.,
(P
1max
/P
2max
) _1.10.
Additionally, a minimum compliance increase of 5% (C
2
/C
1
_1.05) was prescribed to
occur between cycle counts used to calculate da/dN. This prevented excessive noise in the
data by ensuring a resolvable crack extension had occurred. All data points were checked
against these criteria. The load drop and compliance increase restriction imply there is
only a window of cycles during which accurate data can be gathered. If approximate
growth rates can be estimated, specific cycle counts between necessary measurements can
be approximated.
The largest (beginning) G was associated with measured da/dN growth for a given set
of cycles. This may seem slightly nonconservative, but it is significantly more accurate
than averaging because of the power nature of the Paris Law. Conservatism can be added
to the data after it is accurately calculated by adding simple safety factors. In this
case, shifting the da/dN data 1/3 decade upward may make sense to create a worst-case
boundary for the measured data.
Due to the unusual nature of the analytical approach presented here, an examination of
the results of a hypothetical test may help evaluate the method for a particular use. If the
compliance calibration and crack growth empirical constants, as shown in Equations (2)
and (5), can be estimated, then the researcher can simulate the fatigue crack growth
process and following data reduction and analysis. This can be done to aid in specimen
design and to refine experimental strategy before an expensive and time consuming
experimental fatigue crack growth task is begun. For instance, one common pitfall is an
excessively compliant specimen, which can result in very high deflection not conducive
to high-frequency fatigue testing. A fatigue simulation using the empirically derived
constants measured for this research is shown in Table 1. Figures 5 and 6 aid the
visualization of several parameters throughout the experiment.
The simulation is of a single specimen, loaded in a sequence of 3 blocks. Each of these
blocks of fatigue cycles begins at a chosen load that produces reasonably high crack
growth rates. Then, as load is shed, the growth rates near an assumed threshold rate. Then
the test is reset, increasing the load (and displacement) for another period of crack
growth. Several snapshots in time are shown during the fatigue test with associated
parameters, including cycle count (time). The compliance and da/dN are calculated via
the assumed empirical constants, and the other parameters are derived according to
the previously mentioned analysis. Note the total time would be -57 days at 5 Hz.
1424 C. K. BERKOWITZ AND W. S. JOHNSON
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
Table 1. Simulation of DCB fatigue crack growth test.
Block P
max
(N)
max
(mm) C (mm/N) a (mm) dC/da G
max
(J/m
2
) da/dN N (cycles) log(N)
1 350 2.28 0.00651 20.0 0.00088 530 8.4E-04 1 0.0
1 280 2.28 0.00814 21.7 0.00101 390 3.1E-04 5487 3.7
1 224 2.28 0.01018 23.6 0.00116 288 1.2E-04 21,340 4.3
1 179 2.28 0.01272 25.6 0.00134 212 4.4E-05 67,153 4.8
1 143 2.28 0.01590 27.8 0.00154 156 1.7E-05 199,539 5.3
1 115 2.28 0.01988 30.2 0.00178 115 6.3E-06 582,102 5.8
1 92 2.28 0.02485 32.8 0.00204 85 2.4E-06 1,687,616 6.2
1 73 2.28 0.03106 35.7 0.00235 62 8.9E-07 4,882,280 6.7
2 190 5.90 0.03106 35.7 0.00235 418 3.9E-04 4,882,280 6.7
2 152 5.90 0.03882 38.7 0.00271 308 1.5E-04 4,903,257 6.7
2 122 5.90 0.04853 42.1 0.00311 227 5.5E-05 4,963,874 6.7
2 97 5.90 0.06066 45.7 0.00358 167 2.1E-05 5,139,043 6.7
2 78 5.90 0.07582 49.6 0.00412 123 7.8E-06 5,645,239 6.8
2 62 5.90 0.09478 53.9 0.00475 91 2.9E-06 7,108,020 6.9
2 50 5.90 0.11847 58.6 0.00546 67 1.1E-06 11,335,099 7.1
3 120 14.22 0.11847 58.6 0.00546 387 3.1E-04 11,335,099 7.1
3 96 14.22 0.14809 63.6 0.00629 285 1.1E-04 11,379,035 7.1
3 77 14.22 0.18511 69.1 0.00723 210 4.3E-05 11,506,000 7.1
3 61 14.22 0.23139 75.0 0.00833 155 1.6E-05 11,872,899 7.1
3 49 14.22 0.28924 81.5 0.00958 114 6.1E-06 12,933,147 7.1
3 39 14.22 0.36155 88.5 0.01103 84 2.3E-06 15,997,004 7.2
3 31 14.22 0.45193 96.2 0.01269 62 8.6E-07 24,850,797 7.4
Load, strain energy release rate, and displacement vs. cycles
0
100
200
300
400
500
0.0E+0 2.0E+6 4.0E+6 6.0E+6 8.0E+6 1.0E+7 1.2E+7 1.4E+7 1.6E+7 1.8E+7 2.0E+7
N (cycles)
L
o
a
d

a
n
d

G

(
l
b

a
n
d

J
/
m
2
)
0
3
6
9
12
15


(
m
m
)P
G
d
Figure 5. Load, strain energy release rate, and displacement during fatigue simulation.
Crack length and crack growth rate vs. cycles
0
30
60
90
120
0.0E+0 2.0E+6 4.0E+6 6.0E+6 8.0E+6 1.0E+7 1.2E+7 1.4E+7 1.6E+7 1.8E+7 2.0E+7
N (cycles)
C
r
a
c
k

l
e
n
g
t
h

(
m
m
)
1.0E-07
1.0E-06
1.0E-05
1.0E-04
1.0E-03
C
r
a
c
k

g
r
o
w
t
h

r
a
t
e
(
m
m
/
c
y
c
l
e
)
av.N
da/dN v. N
Figure 6. Crack and crack growth rate during fatigue simulation.
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure 1425
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
A researcher could use this type of simulation to examine trade-offs among priorities and
potential problems, and their causes.
RESULTS
The test matrix of the experiments performed is shown in Table 2. Eleven quasi-static
(monotonic) fracture toughness tests were performed and seven fatigue tests. The
experiments were performed at three different test temperatures: hot, cold, and room
temperatures, at the combinations described in the matrix.
Fracture Toughness
The numerous G
c
measurements from each test condition were averaged, with each
specimen being weighted equally, and they are shown in Table 3. Many fracture toughness
measurements for each specimen were obtained using the loadcrackunload technique
previously described. A typical specimens loaddisplacement plot obtained for four load
unload cycles is shown in Figure 7.
Many materials exhibit a crack length dependence on G
c
. This is generally attributed to
a change in plastic zone shape with crack growth [14]. However, no dependence on crack
length was found for fracture toughness for any condition studied in this work. This
suggests that the failure can be more easily characterized by the single parameter of strain
energy release rate. The lack of crack length dependence is often referred to as a flat
R-curve. A plot of G
c
versus crack length is shown in Figure 8.
Clearly, the test temperature had an effect on fracture toughness. The cold temperature
tests resulted in the highest fracture toughness, and the hot temperature tests had the
Table 2. Test matrix.
Fracture
toughness Fatigue
Test temperature (monotonic) crack growth
Room temperature (21

C) 6 2
Cold temperature (54

C) 3 2
Hot temperature (77

C) 2 3
Total 18
Table 3. Fracture toughness results.
Specimens Values
G
c,avg
(J/m
2
)
Room temperature (21

C) 6 56 1180
Cold temperature (54

C) 3 62 1620
Hot temperature (77

C) 2 42 1160
1426 C. K. BERKOWITZ AND W. S. JOHNSON
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
lowest fracture toughness. The cold temperature data is obviously separated from the
room temperature data; however, the hot data seems less significantly different from the
baseline.
The mode of failure for all toughness testing was in the honeycomb core; the adhesive
film never failed, even when a Teflon strip was tried as a starter. Figure 9 presents a
photograph of a typical core failure surface on a specimen well past an experimentally
reasonable crack length. Saw cuts were used as starter cracks for most specimens and gave
similar results as Teflon, and data near the artificial crack tip was disregarded. This failure
mode suggests that only core properties were measured, and this mode could be described
as tearing of the individual fibers and the binder material that compose the paper-like
Load vs. displacement
0
100
200
300
400
500
0 1 2 3 4 5 6 7 8 9
(mm)
P

(
N
)
a = 27.0mm
a = 29.1mm
a = 30.1mm
a = 31.5mm
Room temperature (21

C)
Figure 7. Typical loaddisplacement curves.
Critical strain energy release rate vs. crack length
0
400
800
1200
1600
2000
0 10 20 30 40 50 60 70 80
a (mm)
G
c

(
J
/
m
2
)
RT (21C)
CT (-54C)
HT (77C)
Fracture toughness
11 DCB specimens
Figure 8. G
c
dependence on crack length.
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure 1427
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
ribbons of the honeycomb. There was likely not a homogeneous plastic zone, and the
failure could have been more dominated by the aramid fibers strength than by any stress
concentrating effect within the Nomex material; however, this idea would need further
validation. This may be an explanation for the increase in fracture toughness at the
cold temperature; for colder temperatures generally increase a materials strength, while
generally decreasing fracture toughness.
In most applications, the core failure mode would be desirable and designed, since that
would imply a high quality bond at the adhesive interface, which is important to a
sandwich structures integrity. It is worth noting that the mode of failure and fracture
toughness values can be strongly dependent on processing, in addition to the material
constituents. For instance, another sandwich structure of identical facesheets, core, and
adhesive as the one studied here, except with precured facesheets secondarily bonded to
the core, could have considerably different failure behavior than a co-cured structure. The
presence (or not) of an adhesive film may also have a large effect; and clearly the curing
cycle, tooling, and constituents themselves are important parameters. Due to these
considerations, little can be generalized for other composite sandwich structures based on
these results. In fact, these issues warrant further research.
Fatigue Crack Growth
The fatigue crack growth results are presented by temperature in Figure 10. The
resulting trends from the fatigue data are consistent with those found from the quasi-static
fracture toughness tests, and the mode of failure was the same as in the fracture case.
Figure 9. DCB fracture surface.
1428 C. K. BERKOWITZ AND W. S. JOHNSON
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
The linear shape of the data on Figure 10 verifies the validity of the Paris crack growth
model that was applied. The data agrees with the model well in the range of crack growth
rates between -5 10
6
and 5 10
3
mm/cycle. A line has been sketched on the figure
showing that an approximate Paris exponent of 3.2 fits the data well in this regime. This
line is not intended to fit any particular set of data, but is added for visual purposes.
A few slower growth rate data points suggest the Paris relationship is not at all
valid below 1 10
6
mm/cycle, which is quite normal for many materials as threshold
is approached. Although a true threshold was not verified, a somewhat arbitrary con-
vention for composite fatigue is to assume 1 10
6
mm/cycle is significant, and a line
Fatigue crack growth rate vs. strain energy release rate
1.0E-07
1.0E-06
1.0E-05
1.0E-04
1.0E-03
1.0E-02
10 100 1000 10000
G (J/m
2
)
d
a
/
d
N

(
m
m
/
c
y
c
l
e
)
RT (21C)
CT (54C)
HT (77C)
Approx. Paris fit
Freq=4 Hz
R=0.1
da/dN=1.6E-12 (G)
3.2
Threshold
G
IC
Figure 10. Fatigue crack growth data.
Fatigue crack growth rate vs. strain energy release rate
1.0E-06
1.0E-05
1.0E-04
1.0E-03
100 1000
G (J/m
2
)
d
a
/
d
N

(
m
m
/
c
y
c
l
e
)
RT (21C)
CT (54C)
HT (77C)
Freq=4 Hz
R=0.1
Figure 11. Fatigue crack growth data zoomed-in.
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure 1429
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
corresponding to this is sketched on the figure at -70 J/m
2
. The data shown in this slow
growth regime supports that a threshold effect exists. Another line at 1200 J/m
2
is shown
on the figure; this is an approximation for fracture toughness.
Figure 11 shows a zoomed-in view of the data in Figure 10. In this plot, the data points
can be more easily distinguished and compared. Although the overlap of the data shown in
the figure lessens the certainty with which conclusions can be drawn, the cold temperature
resulted in slower crack growth rates than room temperature. This would be expected
since the cold temperature increased toughness. As shown in the figure, the hot
temperature data seems to be highly overlapped with the room data. The point averages
would suggest that the hot temperature resulted in faster growth, but the lack of
separation of the scatter bands on the figure causes any conclusions to be unclear. Similar
to the fracture toughness results, the cold temperature had a more obvious effect on the
fatigue results than the hot temperature, relative to the baseline of room temperature.
CONCLUSIONS
A fracture mechanics-based approach has been utilized for characterizing a sandwich
structure common to commercial aviation. Fracture toughness and fatigue crack growth
testing of a particular sandwich system were performed on double cantilever beam style
specimens. These specimens were tested at each of the three temperatures: hot (77

C),
room (21

C), and cold (54

C). Compliance versus crack length correlating was shown as


an effective tool in deducing both applied strain energy release rates and crack growth
rates from test data. Using compliance to estimate crack length can be a useful technique
when a high resolution in compliance change with crack extension is achievable, especially
in cases where crack tip visualization is problematic. When the necessary empirical
constants can be estimated, a fatigue test can be simulated. This can be a valuable tool for
the design of specimens and experimental strategy when a compliance crack length
calibration can be utilized.
The temperature effects on the particular sandwich system studied were significant. The
cold temperature increased toughness and reduced fatigue crack growth rates, relative
to room temperature. The hot temperature had relatively little impact; however, any
marginal effects were the opposite trends as those caused by the cold. A Paris crack growth
model fit the data well for most growth rates between near-threshold and near-G
c
, and
a flat R-curve behavior was exhibited by the material.
The failure of all specimens tested was in the core material. This suggests that core
properties were being measured. The mode of failure is possibly due to manufacturing
techniques, as much as the constituents. The core was weaker than the bonding of the
facesheets to the core. The mechanisms of failure and resulting data suggest that core
strength is the dominant factor in the fracture mechanics properties presented in this
paper.
REFERENCES
1. Zenkert, D. (ed.) (1997). The Handbook of Sandwich Construction, EMAS Ltd, UK.
2. Falk, L. (1994). Foam Core Sandwich Panels with Interface Disbonds, Composite Structures,
28(4): 481490.
1430 C. K. BERKOWITZ AND W. S. JOHNSON
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from
3. Carlsson, L.A. and Prasad, S. (1993). Interfacial Fracture of Sandwich Beams, Engineering
Fracture Mechanics, 44(4): 581590.
4. Avery, J.L. and Sankar, B.V. (2000). Compressive Failure of Sandwich Beams with Debonded
Face-Sheets, Journal of Composite Materials, 34(14): 11761199.
5. Ratcliffe, J. and Cantwell, W.J. (2001). Center Notch Flexure Sandwich Geometry for
Characterizing Skin-Core Adhesion in Thin-Skinned Sandwich Structures, Journal of Reinforced
Plastics and Composites, 20(11): 945970.
6. Ratcliffe, J. and Cantwell, W.J. (2000). A New Test Geometry for Characterizing Skin-Core
Adhesion in Thin-Skinned Sandwich Structures, Journal of Materials Science Letters, 19(15):
13651367.
7. Hentschel, R.A.A. (1975). Nomex

Aramid PapersProperties and Uses, In: TAPPI Paper


Synthetics Conference, Atlanta, GA, October 68, pp. 189213.
8. Shafizadeh, J.E. and Seferis, J.C. (2000). The Effects of Long Time Water Exposure on the
Durability of Honeycomb Cores, In: 45th International SAMPE Symposium and Exhibition,
Covina, CA, May 2125, pp. 380388.
9. Johnson, W.S. and Butkus, L.M. (1998). Considering Environmental Conditions in the Design
of Bonded Structures: A Fracture Mechanics Approach, Fatigue and Fracture of Engineering
Materials & Structures, 21(4): 465478.
10. ASTM D5528-94a (1994). Standard Test Method for Mode I Interlaminar Fracture Toughness
of Unidirectional Fiber-Reinforced Polymer Matrix Composites, In: Annual Book of ASTM
Standards, American Society for Testing and Materials, Philadelphia, PA.
11. Berry, J.P. (1963). Determination of Fracture Energies by the Cleavage Technique, Journal of
Applied Physics, 34(1): 6268.
12. Paris, P.C. and Erdogan, F. (1963). A Critical Analysis of Crack Propagation Laws, Journal of
Basic Engineering, 85: 528534.
13. ASTM E647-00 (2000). Standard Test Method for Measurement of Fatigue Crack Growth
Rates, In: Annual Book of ASTM Standards, American Society for Testing and Materials,
Philadelphia, PA.
14. Anderson, T.L. (1995). Fracture Mechanics: Fundamentals and Applications, 2nd edn, CRC Press
LLC, Boca Raton, FL.
Fracture and Fatigue Tests and Analysis of Composite Sandwich Structure 1431
at WAYNE STATE UNIVERSITY on July 11, 2014 jcm.sagepub.com Downloaded from

Das könnte Ihnen auch gefallen