Академический Документы
Профессиональный Документы
Культура Документы
Russell
D.
Fiber-reinforced comp
Jamison^
is necessary to confront t
Procedure
R E F E R E N C E : Jamison, R. D . , " O n the Interrelationship Between F i b e r F r a c t u r e
a n d Ply C r a c k i n g In G r a p h i t e / E p o x y L a m i n a t e s , " Composite
Fracture,
ASTM
Materials:
F"lf'<^<'"d
Philadelphia, 1986, p p . 2 5 2 - 2 7 3 .
A B S T R A C T - The present w o r k examines i n a systematic way the development o f m i c r o damage i n several laminates o f g r a p h i t e / e p o x y material subjected to both quasi-static tensile
loading and tension-tension f a t i g u e . Emphasis is placed upon d i s c r i m i n a t i n g and q u a n t i f y i n g
m a t r i x and f i b e r microdamage. Penetrant-enhanced standard- and stereo X - r a y radiography
were used along w i t h edge replication to f o l l o w the progression o f m a t r i x damage. I h e
recently developed technique o f laminate deply was used to map the development o f f i b e r
fracture.
.
. ^ .
The most s i g n i f i c a n t result, c o n f i r m e d i n b o t h the quasi-static and fatigue tests, was the
dominant role played by o f f - a x i s p l y cracks i n the fracture o f fibers i n adjacent load-bearing
plies B y direct observation o f f i b e r fractures in situ, it was established in b o t h cases that
f i b e r fractures do not occur i n a r a n d o m pattern at elevated loads/cycles, but instead occur
i n narrow bands adjacent to o f f - a x i s p l y cracks. I t is the action o f these crack tips upon
adjacent f i b e r s w h i c h may govern the wearout or overload o f laminates under these con''"BTdirect accounting o f f i b e r fractures i n deplied laminae taken f r o m damaged but u n f a i l e d
laminates
Specimens f o r fatigue
the relationship between the density o f f i b e r fractures and the number o f cycles
IQ-t
S-'.
at one c y c l i c stress level is reported. The relationship between f i b e r fracture density and
. . . .
emission
252
253
Procedure
Specimens for fatigue testing were fabricated f r o m N A R M C O T300/5208
prepreg material. Specimens for static testing were fabricated f r o m Fothergill
and Harvey T300/Code 91 prepreg. The Code 91 resin is similar in composition
and properties to the 5208 resin but has a lower cure temperature. Specimens
measuring 203 m m long and 25 m m wide with stacking sequences [ 0 , 0 ] , and
[0,902], were used f o r both fatigue and static test series. Fiber volume fractions
were approximately 6 1 % for both layups, and specimen thicknesses were approximately 0.6 and 0.9 m m , respectively. Fatigue testing was performed only
on the [0,902], laminates in a tension-tension mode at 10 Hz in a sinusoidal f o r m
and at a stress ratio R = 0 . 1 . M a x i m u m stress amplitude was 70% o f the mean
static ultimate strength ( ^ u , , ) . Static tension testing was accomplished on both
[0,902], and [ 0 , 0 ] , laminates in a displacement-controlled mode at a strain rate
of 1.7 X 10-" S - ' .
Fatigue specimens were cyclically loaded to various stages of their anticipated
fatigue lifetime based upon a reduction in stiffness damage analog described in
a previous paper [ 7 ] , Static specimens were loaded to various fractions o f the
mean ultimate strength and unloaded immediately. In both cases damage was
characterized by the nondestructive techniques of X-ray radiography and edge
replication and the destructive techniques of sectioning and deply. These methods
are described in detail in Refs 2,3, and 4.
For the static testing o f both unidirectional and cross-ply specimens, acoustic
emissions were monitored. The system consisted o f a wide-band piezoelectric
(PZT) transducer having a contact diameter o f 25.4 m m . The transducer output
passed through a 60-dB preamplifier which contained a 100-Khz to 1-Mhz bandpass filter into a modified A E T C Model 203 amplitude analyzer. This system
covered an amplitude range f r o m 20 fxV to 20 m V at the sensor (26 to 86 dB
with respect to 1 |xV at the sensor). The analyzer divided this range equally into
50 channels each of amplitude width 1.2 d B . The threshold o f the instrument
could be raised above 20 | j l V and channels eliminated to filter some low-amplitude
254
noise. I n the present work channels 0 - 3 were excluded in this way, giving an
effective threshold of approximately 30 |xV at the sensor.
The A E T C system operated also as an event counter capable of distinguishing
one event f r o m another provided there existed a dead time of at least 100 |xs
between successive events. For a typical pulse of 400 [xs the system as modified
was intrinsically capable o f registering 2000 events per second. Static testing
ms accomplished at a l o w strain rate to stay w i t h i n this limitation and avoid the
O S S of information near failure when large count rates are generally observed.
lesults and Discussion
atigue Damage
Assessment
255
I.OOr
0.99 P
097 h
Q96
10
20
30
40
THOUSANDS OF CYCLES
F I G . 1 T r a n s v e r s e crack
[0,90j],
laminate.
density'
and normalized
secant
modulus
versus
fatigue
cycles
for
Static Damage
Assessment
256
F I G . l^Radiograph
enlargement
of a fatigue
damaged
[0,90,],
laminate.
ply cracks were the first damage observed with the first cracks developing at
loads above 30% of the mean ultimate strength. Figure 4 shows the relationship
between tension load and the observed density of cracks in |0,902],, specimens.
These are instantaneous measurements at specified loads. Measured crack densities did not achieve the equilibrium saturation spacing associated with the characteristic damage state. The crack density at 100% S was 38% lower on average
than the CDS prediction based on a one-dimensional shear lag analysis [14].
This difference is attributable to the viscoelastic behavior of the resin matrix.
F I G . 3 S c h e m a t i c of damage
localization
257
pattern.
Even at very low strain rates there was insufficient time for complete crack
development to occur at a given load. A t a constant load of 70% 5||. for example,
approximately 30 s was required for a stable crack density to develop. Because
one goal o f the present work was to establish a correlation between transverse
ply cracking and acoustic emission measured during quasi-static tension loading,
the power-law relationship shown in Fig. 4 was taken to represent real-time crack
development under tension loading of these laminates.
I2h
lOh
Load
F I G . 4 T r a n s v e r s e crack
density
versus
(7oSuit)
static
tension
load for
[0.90,].
laminates.
258
F I G . 5Radiograph
enlargement
of a statically
damaged
[0,90,],
laminate.
Fiber
259
Fracture
The state o f fiber fracture in fatigue and quasi-static tension loading was
examined next. The method o f laminate deply [4] was used to "unstack" the
0-deg plies f r o m laminates loaded cyclically for various fractions of anticipated
fatigue lifetimes, or statically to specified fractions o f the mean ultimate stress.
Portions o f these plies were examined in a scanning electron microscope (SEM)
and the average fiber fracture density was calculated by counting broken fibers
in a number o f randomly chosen representative areas o f dimensions approximately
3.5 by 4.9 m m . The process o f deplying, i f done with care, produces very few
fiber fractures. Average fiber fracture densities for virgin specimens were calculated and used as a basis f r o m which the true relationship between load or
cycles and fiber fracture could be determined.
Figure 6 shows this relationship for tension fatigue loading o f [0,902] > laminates
at a stress amplitude o f 70% 5. Fiber fracture density is seen to increase rapidly
during eariy cycles. Fiber fracture continues to occur with increasing cycles but
at a substantially diminished rate. Although the available data were not sufficient
to completely characterize this relationship, the existence o f a plateau as shown
in F i g . 6 was suggested by examination o f post-failure specimens. I n these
specimens the fiber fracture density away f r o m the fracture site was not observed
to be significantly larger than in prefailure measurements o f similar specimens
at advanced life. While at the failure site rapid growth o f fiber fractures miist
occur, there is no evidence at this time to suggest that such a phenomenon occurs
on a global scale.
Fiber fracture density in quasi-static tension loading is shown in Fig. 7. A t
loads below 70%
there are essentially no broken fibers in excess o f those
found in virgin specimens. A t loads above this level, however, fiber fracture
density increases rapidly in an approximate power-law f o r m . Thus, while a load
of 70% j'ui, is not sufficient to produce a significant number o f fiber fractures in
ICQ
fracture
density
versus fatigue
cycles for
[0.90,],
laminates.
260
JAMISON
FIG.
Micrograph
ON
of fiber
FIBER
fractures
FRACTURE AND
in a fatigue-damaged
PLY
CRACKING
[0,90,],
261
laminate.
their associated resin damage zones in controlling fiber fracture. It is the combination o f global stress and local stress concentration due to adjacent ply cracks
which dictates the density and distribution of fiber fractures in the load-bearing
plies of these laminates.
Fiber fractures in [ 0 , 0 ] , laminates subjected to static tension loading were also
examined. I n this case the fiber breaks which occurred were distributed in an
apparently random pattern. The density of these fiber breaks as a function of
load is shown in Fig. 11. Results for the statically loaded [ C W j ] , laminates are
repeated f o r comparison. Significant numbers o f fiber fractures did not occur at
loads below approximately 70% 5. The fiber fracture density then increased
rapidly with increasing load through laminate failure.
Inspection of F i g . 11 would suggest that the density of fiber fractures near
failure is substantially higher in the cross-ply laminate. While this may in fact
be true, the difference would be less than that shown. The method of counting
fiber fractures has undoubtedly exaggerated whatever real difference there may
be. In the cross-ply laminates fiber fractures were counted in the fiber layer
adjacent to the 0/90-deg interface and hence nearest to the crack tip. It has been
shown by the author in previous work [ 2 , 5 ] that the density o f fiber fractures is
markedly diminished in fiber layers away f r o m the 0/90-deg interface70%
fewer only one fiber diameter removed f r o m the interface in the case of fatigue.
262
JAM ISC
[ 0 , SOg]
[0,
0],
0.)
20
F I G . I I F i b e r fracture
F I G . 9 M i c r o g r a p h of fiber
fractures
in a statically
damaged
[0,90,],
laminate.
dem
Although it is possible to
application and careful ren
vations required to produce
thickness accounting. It is j
fracture density observed a
unidirectional laminates.
Acoustic
d=
ADJACENT
PLY CRACK
F I G . 1 0 S c h e m a t i c of the localized
pattern
Analysis
SPACING
of fiber
Emission
fractures.
263
Load ( % S u i , )
FIG. 11Fiber
fracture
density
versus
static
tension
load for
[0,0],
and [ 0 , 9 0 , ] ,
laminates.
Acoustic
Emission
Analysis
One N D E method which would seem to hold promise for such a purpose is
acoustic emission. Therefore, acoustic emissions recorded during tension loading
of [0,902],, and [ 0 , 0 ] , laminates were analyzed to seek correlations between
attributes o f the A E pulses and the physical picture of damage development
described i n the previous sections.
Figure 12 shows the cumulative acoustic emissions recorded during the tension
264
40
o
32
F I G . n^Cumulative
AE counts
versus
load for a [ 0 , 9 0 , ] ,
laminate.
1
2
3
4
5
6
L O A D R A N G E ( % 5ui,)
0-30
30-60
60-70
70-80
80-90
90-prefailure
265
CHANNELS INCLUDED
1
2
3
4
5
6
4- 9
10-15
16-21
22-27
28-33
34-50
A t l o w loads there are few total counts and the interpretation is not meaningful.
But as the first significant damage begins to occur at loads between 4 and 6 k N ,
the share o f events in the lowest channel range begins to diminish. In this range
(channels 4 - 9 ) are included spurious, low-amplitude background events which
occur independently o f the load as well as a sizable number o f the actual damagegenerated acoustic events. As channel range 1 decreases its share o f the total,
channel ranges 2-5 increase their relative share o f the total cumulative events!
This trend becomes significant at loads around 8 k N (approximately 65% 5|,)
when ply cracking becomes increasingly active. Significantly, the proportion o f
events in the highest channel range is seen to begin to diminish at loads above
10 k N and to continue to diminish through failure. This is the load range over
which nearly all fiber fractures are occurring.
Such behavior is inconsistent with the assumption that fiber fracture is the
source o f higher-amplitude acoustic emissions. Indeed as the rate o f fiber fracture
IS greater than the rate o f ply cracking in the higher load ranges, it could be
concluded that transverse ply cracking is the principal contributor o f the higheramplitude events and, in becoming proportionately a smaller fraction o f the total
cumulative damage as load is increased, its higher-amplitude contributions to the
cumulative acoustic emission would likewise become proportionately smaller.
266
FRACTURE
20
Load Level 1
16
12
-1
10
20
Chann i
30
40
50
Numb* r ,
12
20
Load Level 5
16
12
S 8
a:
10
20
Chmnn!
F I G . 1 3 A m p l i t u d e distribution
30
40
50
Numl)r.
of AE events
laminate.
268
FRACTURE
(b)
FIG.
range
\4^Percent
I; (b) excluding
share
of AE events
channel
range
I.
in range
for a [ 0 , 9 0 , ] , laminate,
(a) Including
channel
F I G . 1 5 A m p l i t u d e disliilnitions
269
laminate.
These patterns were identified repeatedly in similar plots for other specimens
tested. In all cases, without exception, the proportion o f events occurring in the
highest-amplitude range recorded during tension loading to failure o f [0,902],
laminates diminished as the specimen neared failure.
Figure 15 presents the amplitude distribution data f o r a typical [ 0 , 0 ] , specimen.
The load ranges are the same as previously defined. Events occurring at low load
ranges are seen to principally populate channels in the range 4 to 22. In fact f o r
this and all other unidirectional specimens loaded to failure, event amplitudes
were generally below channel 25 f o r all load ranges. The shape o f the distributions
showed little change with increasing load, and no shift to higher amplitudes was
observed. In comparison with the evolution of the amplitude distributions o f
cross-ply laminates with increasing load (Fig. 13), the results f o r the unidirectional
specimens are seen to be more stationary. Whereas the cross-ply amplitude
distribution shifts to higher-amplitude channels with increasing load and then
retreats slightly to lower channels at loads near failure, the unidirectional specimens can be characterized by a single, low-amplitude-dominated distribution
across the load-to-failure spectrum. Such a result is consistent with the operation
of two damage mechanisms in the cross-ply laminate and only one in the unidirectional.
Figure 16 shows the cumulative percentage share of events in various channel
ranges as defined previously. The virtual absence of events in channels 2 2 - 5 0
is notable. The fraction o f events in channels 16-21 is essentially unchanged
270
FRACTURE
Load I n Newtons/10^
(a)
25^
Load In Newtons/10'
(b)
F I G . \6Percent
(b) excluding
channel
share
range
of events
I.
iaminate.
(a) Including
channel
range
271
through failure. The share o f events in channels 10-15 increases at loads near
failure. Inasmuch as fiber fractures are the dominant source o f acoustic emission
near failure, this channel range may best characterize the associated acoustic
emission. It is significant that this pattern observed in the amplitude distribution
of unidirectional laminates and attributable to fiber fracture alone is consistent
with the conclusion reached for fiber fracture in cross-ply laminates: For this
material system, fiber fracture is on average a lower-amplitude A E event than
transverse ply cracking.
The association of fiber fracture with low-amplitude acoustic emission is counter
to much o f the conventional wisdom for acoustic emission in composite materials.
It is frequently assumed that fiber fracture in any material system w i l l release
large amounts o f strain energy relative to other damage mechanisms and w i l l
therefore be the source of the highest-amplitude acoustic emissions. This assumption is so common, perhaps because it seems so intuitive, that the lack
of substantiating evidence o f its general validity is sometimes unquestioned. It
is indeed possible that for a particular fiber/resin system and a particular laminate
type, fiber fracture may be the source of high-amplitude acoustic emission.
However, as with nearly all aspects of A E analysis, the association o f an attribute
of the acoustic emission pulse, such as amplitude, with an attribute o f the source
of the pulse, such as energy, is complicated. First, given the dispersive nature
of the composite media, there is no assurance that pulses originating with given
relative amplitudes w i l l aiTive at the transducer with the same relative amplitudes.
Second, fiber fracture occurring in surface plies may yield fundamentally different
A E pulses at the receiver than the same fractures occurring in imbedded plies;
that is, the laminate geometry and stacking sequence have an important influence
on the characteristics o f the received pulse. Third, and perhaps most important
in the present context, the average energy released by a particular event is
intimately related to the environment in which that event occurs. In the case of
fiber fractures for example, it is commonly assumed that because they occur
typically at elevated loads, the strain energy released must be greater than for
events occurring at lower loads. By invoking this violin string analogy, the reality
of fiber fracture is obscured. When fibers which are imbedded in moderately
ductile resin matrices break at elevated loads, the strain energy released per unit
volume is relatively large. However, the volume of fibers f r o m which this energy
is released is quite small i f the fiber-matrix interface remains intact away f r o m
the fracture site. It has been shown in woven glass-reinforced plastic laminates
for example that the energy released by the fracture of a fiber disbonded over a
length o f 2 m m is approximately 200 times greater than the fracture energy f o r
an undisbonded fracture [15]. The present results for unidirectional graphite/
epoxy laminates and prior work reported by other investigators [16,17] indicate
that fiber-matrix disbonding in this system is minimal or nonexistent. It would
not be unexpected then that the net energy released by a single fiber fracture,
even at loads near failure, might be small relative to transverse ply cracking.
Discrimination between ply cracking and fiber fracture based on analysis of
272
the amplitude produced requires that quite subtle distinctions be made. These
distinctions in the evolution of the amplitude distribution with increasing load
are nonetheless clear and repeatable. Identification of damage modes based solely
upon attributes of the amplitude distribution of the associated acoustic emission
is d i f f i c u l t A n d when several modes are occurring simultaneously (albeit at
different rates) the discrimination problem is formidable. In the present instance,
discrimination between ply cracking and fiber fracture, perhaps the most physically distinguishable pair imaginable, requires a conclusion that ply cracking is
a slightly higher-amplitude event on average than fiber fracmre. Recent work on
a similal graphite/epoxy system by other investigators [18] yielded the same
Journal
of Fatigue,
V o l . 4 , 1982, pp
l ~ n ' s .
M. in Composite
Materials: Testing and Design i ^ ^ - ^ o ^ ' e ^ t ^ ASTM
1R7 Am'erican
A m e r i c a n Society
Society ff o
o rr Testing
Testing and
and Materials,
Materials, Philadelphia,
Philadelphia, 1982,
1982, pp.
pp. 55 00 --6622^.
[5J amison R D Schulte, KK. ,^ Reifsnider,
R ^ n i d e . KK. . LU. , and
a t . S.t m
m cchhccoommb ,l . W . W . in ^ e r t
TcZposite
Materials, ASTM STP 836, A m e r i c a n Society f o r Testing and Materials,
STP
S^irTa'Se,
[ 9 ] K o r c z n s k y j , J. and M o r l e y , J. G . ,
Journal
' 9 7 9 , V o l . 3, pp^ 2 2 7 - 2 9^
of Materials
Science,
1795
' ^
il3]
[14]
273
775
[JS]
[16]
[17]
[18]
[19]
Oct. 1980, pp
STP
209-218