Académique Documents
Professionnel Documents
Culture Documents
In this paper two modes of composite tensile failure are investigated. The failure loads
predicted by these analyses are significantly closer to experimental data than predictions
of other theories. The first study considers composites containing a planar array of parallel
fibers which exhibit a large number of isolated fiber breaks before failure. A statistical
analysis that includes the effects of stress concentrations is employed to describe the mechanics
of failure. The second study, which is also statistical in nature, considers monolayer and
multilayer unidirectional composites that do not display many isolated breaks before failure.
Failure criteria are established for each mode, and the implications for nondestructive test-
ing are discussed.
Downloaded by UNIVERSITY OF QUEENSLAND on June 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.4990
ture. As in Rosen's analysis, the matrix is considered to be mediately before failure is quite small. In addition, the
purely a medium for the transmission of shear stress be- probability of having an adjacent element break because
tween the fibers, and crack propagation through the matrix its strength is less than a has been ignored since this prob-
itself is not included in the study of this failure mode. How- ability is small and we are interested in the effects of fracture
ever, crack propagation in the matrix is a possible failure propagation due to load concentrations.
mechanism, and it is discussed later. Given that a single element is broken, the probability that
one, and only one, of the two adjacent fibers will break due
Determination of Ineffective Length to the load concentration is
The ineffective length is a measure of the portion of the - F(a)] - 2[F(Kl(r) - (5)
fiber which has a significant reduction in stress, and several
definitions have been proposed for it. In this paper the defi- The probability that both adjacent fibers will break simul-
taneously is
nition proposed by Friedman5 is used unless an experimental
value (as determined by photoelastic analysis, for example)
is available. Friedman's definition for the elastic case is
It is now assumed that only the two fibers immediately
d== (1) adjacent to a break are subjected to an overstress and that
where df is the fiber diameter, Ef the Young's modulus, Gb the all of the remaining fibers in the cross section have a stress
matrix shear modulus, and vf the fiber volume fraction. level equal to the nominal stress a. If one of the fibers ad-
jacent to a single fracture breaks, the fibers adjacent to the
Downloaded by UNIVERSITY OF QUEENSLAND on June 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.4990
Therefore, for the composite as a whole the expected number tive distribution function G(a). For a sample of N fibers
of groups of two or more broken fibers is (neglecting edge drawn from this population the probability density function
effects) for the strength of the weakest fiber is given by
E2 = MNp* (13) = Ng(a)[l - (23)
The associated probability of having at least one such group We now assume that the fiber strength can be character-
is ized by a Weibull distribution of the form 3
= 1 _ H n^MN (14) G(o-) = I - e~aL^ (24)
The probability that an element will fracture followed by which has the corresponding density
the breaking of at least two other adjacent fibers is equal to le-aLP (25)
the probability of at least two fractures, less the probability
that two, and only two, will break without further fractures [The Weibull distribution is also used for F(d) with L = 8.]
occurring. This probability is Substituting (24) and (25) into (23) and differentiating, the
- ps/2 - 4/2) (15) following expression is obtained for the mode of the weakest
fiber strength distribution:
The probability of having at least one group with three or
more fractures is ov. = [(0 - l)/NLap]l/f> (26)
The details of the failure mechanism for this mode are
P3 = 1 - (1 - Pz)MN
Downloaded by UNIVERSITY OF QUEENSLAND on June 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.4990
a)
b) d)
Fig. 2 Typical sequence of photographs of tensile failure specimen (taken from Ref. 3).
tures at low stress levels is possibly a result of damage to the fibers) given by Eqs. (2, 13, 17, and 20) are plotted for series
fibers during the fabrication of the specimens since glass is B and C in Figs. 5 and 6, respectively. It can be seen that
notoriously sensitive to handling as far as strength is con- not only do the curves for Ez, E8, and E* rise sharply in the
cerned. range of observed failures, but they are also close together,
The quantities Ei, E2, E^, and E4 (where Ei represents the indicating that fibers adjacent to initial fracture sites are
expected number of group of fractures having at least i broken expected to break in this stress range, and that when this
DECEMBER 1968 TENSILE FAILURE OF FIBROUS COMPOSITES 2329
GLASS EPOXY
REFERENCE 4
SERIES B
AVERAGE FAILURE
STRESS = 161 KSI
occurs there is a high probability that propagation of fiber which is the probability that at least one fiber adjacent to an
fractures will result. The failure loads predicted by the initial break will fracture. This quantity has been evaluated
theory of Ref. 3 are presented for comparison. using static and dynamic load-concentration factors for test
In order to assess the validity of the expressions for the series B of Ref. 4. The results are shown in Fig. 10. It can
expected number of multiple fractures, the numbers of such be seen that the dynamic curve is markedly higher than the
groups were counted on films of the tests. The results are static in the failure range, indicating the dynamic effects may
plotted in Figs. 7 and 8. The dashed curves in the figures be significant. The situation is further complicated by the
are calculated values of E%, which, it will be recalled, is the possibility of fiber debonding and crack propagation in the
expected number of groups having at least two adjacent matrix.
breaks. As such, E2 represents the expected number of The fracture propagation theory was also compared with
multiple-fracture groups. It can be seen that in general the results of two tests on continuous glass fibers in an epoxy
multiple breaks begin to appear in the stress range predicted matrix which were made by Friedman.5 The Weibull
by the theory. Furthermore, those multiple fractures that parameters for the fibers are ft = 4.0 and a~1^ = 137.7 in
occur at lower stress levels do not cause immediate failure. in.-kip units. The ineffective length was computed using
This is consistent with the theory, which predicts that frac- Eq. (1). The curves of Ei, EZj Es, and E* are presented in
ture propagation is not likely at these low stress levels. It Fig. 11. The experimental failure stresses and the prediction
can also be seen that the rate of formulation of multiple- of Ref. 3 are presented for comparison.
fracture groups is quite high in the failure range and the Since ft is an inverse measure of dispersion, these fibers
composites fail without the occurrence of a large number of had a much wider spread in failure stress levels than did
such groups. This is the type of behavior that the theory those of Ref. 4. This large dispersion is reflected in the wide
predicts since the probability of fracture propagation is high spread of the curves of EI, EZj Es, and E presented in Fig.
in this stress range. This fact is further demonstrated in 11. The separation of the Ez, Es, and E4 curves signifies a
Fig. 9, which presents Pi, P2, PS, and P, where Pi is the smaller probability of fracture propagation than those of
probability of having at least one group of i fractures. The test series B and C of Ref. 4. However, both failures do
expressions for these quantities are given in Eqs. (11, 14, occur in a region where multiple fiber breaks are expected.
16, and 19). On the basis of good agreement between theoretical pre-
Although the static failure analysis, which is based on the dictions and experimental observations it is suggested that,
accumulation of fractures, provides good agreement with for two-dimensional composite specimens exhibiting a signifi-
cant number of scattered fiber breaks before failure, a con-
servative prediction of the failure load is that load for which
the first multiple break is expected. That is, the composite
failure load <r* can be determined from the equation
pZ(CT
J-J /.T-#\)
_ i.i ^97^
\.it)
GLASS-EPOXY
REFERENCE 4
SERIES C
AVERAGE FAILURE STRESS = 152
Fig. 4 Number of breaks as a function of applied load Fig. 6 Expected numbers of groups of fractures as a
(Ref. 4, series C). function of applied load (Ref. 4, series C).
2330 C. ZWEBEN AIAA JOURNAL
GLASS-EPOXY FAILURE
REFERENCE 4
STRESSES
SERIES B
Bl GLASS EPOXY
x B2 REFERENCE 4
O B3 SERIES B
A 84
O B5 (PORTION OF SPECIMEN
ALONG EDGE BROKE
OFF AT 151 KSI)
u.
90 120 150 Fig. 9 Pi, P2, PS, and P4 as a function of applied load (Ref.
FIBER STRESS (KSI) 4, series B).
Fig. 7 Observed and predicted multiple-fracture groups
(Ref. 4, series B). 'where Ei is defined in Eq. (2). Using the occurrence of the
first break as a failure criterion, it is possible to estimate the
where E% is defined in Eq. (13). The selection of the occur- probability of failure at any stress by using Eq. (11).
Downloaded by UNIVERSITY OF QUEENSLAND on June 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/3.4990
THEORY OF REFERENCE 3
irz^*w!S5aETH