Sie sind auf Seite 1von 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/303032034

On Failure Theories for Composite Materials

Chapter · May 2016


DOI: 10.1007/978-981-10-0959-4_21

CITATION READS

1 1,936

1 author:

Ramesh Talreja
Texas A&M University
182 PUBLICATIONS   4,301 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Failure of unidirectional composite under tensile loading View project

2nd Summer School on Fatigue and Damage Mechanics of Composite Materials View project

All content following this page was uploaded by Ramesh Talreja on 29 August 2017.

The user has requested enhancement of the downloaded file.


Composites Science and Technology 105 (2014) 190–201

Contents lists available at ScienceDirect

Composites Science and Technology


journal homepage: www.elsevier.com/locate/compscitech

Feature Article

Assessment of the fundamentals of failure theories for composite


materials
Ramesh Talreja ⇑
Department of Aerospace Engineering, Texas A&M University, College Station, TX 77843, USA
Department of Materials Science and Engineering, Texas A&M University, College Station, TX 77843, USA
Department of Engineering Sciences and Mathematics, Luleå University of Technology, Luleå 971 73, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: Failure theories for composite materials have been a subject of concern for almost half a century.
Received 16 June 2014 Numerous theories have been proposed over the years, but none of them has successfully predicted
Received in revised form 27 September the full range of observed behavior. This paper critically reviews some of the main theories in an effort
2014
to understand their deficiencies. Rather than seek their validity by comparison with test data, the
Accepted 14 October 2014
Available online 23 October 2014
assumptions and analyses underlying the proposed theories are examined. Certain remedies are
proposed to the current approaches to achieve a failure analysis procedure with a potential to succeed
in predicting the complex failure behavior of composite materials.
Keywords:
A. Polymer–matrix composites (PMCs)
Published by Elsevier Ltd.
B. Fiber–matrix bond
B. Strength
C. Failure criterion

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
2. The early failure theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
3. Failure theories with physical considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
4. Mechanisms based failure analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
4.1. Remedy #1: Multi-scale analysis of failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
4.2. Remedy #2: Analysis of constrained failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
4.3. Remedy #3: Analysis of manufacturing defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
4.4. Multi-scale analysis of failure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
4.5. Fiber/matrix debonding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.6. Cavitation-induced brittle matrix failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.7. Ductile matrix failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.7.1. Determining the governing failure mode in transversely loaded UD composite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.8. Analysis of constrained failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
4.9. Effects of defects on composite failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
5. A comprehensive strategy for composite failure analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

1. Introduction the early 1960s composite materials such as glass/polyester and


carbon/epoxy began emerging as promising materials of the future.
The mechanical behavior of monolithic materials (metals, It was natural for the scientific community then to apply and
ceramics and polymers) had been a fairly mature field when in extend concepts and analyses developed for the monolithic

⇑ Address: Department of Aerospace Engineering, Texas A&M University, College Station, TX 77843, USA.

http://dx.doi.org/10.1016/j.compscitech.2014.10.014
0266-3538/Published by Elsevier Ltd.
R. Talreja / Composites Science and Technology 105 (2014) 190–201 191

materials to composites. In the decades that followed, great suc-


cess was achieved in micromechanics estimates of effective elastic
properties, homogenization, laminate plate theory, etc. However,
theories for treating failure of composite materials did not succeed
to the same extent. In fact after numerous efforts extending over
approximately five decades many uncertainties and controversies
still remain in predicting composite failure.
The main concepts related to failure of monolithic materials, in
particular metals, are yielding and fracture, and the fields of
analyses to treat these failure modes are plasticity and fracture
mechanics. For polymer matrix composites (PMCs), which are the
Fig. 1. A metal sheet with preferred orientation resulting in three planes of
most common class of composite materials, concepts in both these
symmetry denoted by the rectangular coordinate axes x, y, z.
fields must be modified significantly. As far as yielding as initiation
of plastic flow is concerned, polymers as well as fibers of glass or
carbon, which are the common constituents of PMCs, do not have
shear-driven thresholds for inelastic behavior as in metals. Fur- noteworthy. These authors took their starting point in Hill’s theory
thermore, instability of crack growth, which is the basis of brittle of 1948 for yielding and plastic flow of anisotropic metals [4]. Hill
fracture in metals, is not the main concern in laminated PMCs, noted that at large plastic strains and in processes such as rolling,
where unstable matrix cracks seldom lead to failure as they are drawing and extrusion, metals develop a preferred orientation due
in most cases arrested at interfaces within the composite volume. to crystalline plane alignment. A metal sheet can then be viewed as
Most of the failure process in composites then consists of multiple having three orthogonal planes of symmetry as illustrated in Fig. 1.
crack formation, linkage of crack surfaces, extensive fiber failure, For such orthotropic materials, Hill proposed a straightforward
etc. These events may eventually separate a composite part in mathematical generalization of the von Mises yield criterion for
two or more pieces, but prior to that may cause significant changes isotropic metals as follows. Starting with the von Mises criterion
in the deformational response that induces failure to perform the expressed in the x–y–z Cartesian coordinates,
design function. These considerations for composites question      
the so-called ‘‘strength’’, which in monolithic materials is synony-
rxx  ryy 2 þ ryy  rzz 2 þ ðrzz  rxx Þ2 þ 6 r2xy þ r2yz þ r2zx
mous with ‘‘failure’’. More relevant to composites is the notion of ¼ 2A2 ð1Þ
‘‘criticality’’ associated with the function for which the composite
part was designed and manufactured. Thus, failure theories for where A is the yield stress in uniaxial loading in any direction, and
composites should aim at describing and predicting appropriate making two assumptions, namely, that there is no Bauschinger
criticalities. effect, i.e., the yield stress is same in tension and compression,
Over the years, numerous failure theories for composites have and that the plastic potential is not affected by superposition of
emerged. The common industry practice is to use only the theory hydrostatic pressure, Hill generalized the isotropic yield criterion
or model that has been validated by experimental evidence. to the following form.
Because of the uncertainty of validation of the failure theories an
ambitious exercise named worldwide failure exercise (WWFE) Fðryy  rzz Þ2 þ Gðrzz  rxx Þ2 þ Hðrxx  ryy Þ2 þ 2ðLs2yz þ Ms2zx
was conducted to objectively test how different theories fare
þ Ns2xy Þ ¼ 1 ð2Þ
against experimental data. The results of this useful activity have
been reported and assessed [1,2]. where F, G, H, L, M and N are material constants representing the
The aim of this paper is different from the validation exercise. direction dependent yielding. Denoting by X, Y and Z the yield
Although useful, the validation against test data cannot fully strength values in the principal orthotropic directions, x, y and z,
explain why a particular theory works, or does not do so. As it respectively, one obtains,
has turned out, not a single theory has the ability to agree with test
1
data in all cases examined. The effort made here is to complement
X2
¼ G þ H; Y12 ¼ H þ F; Z12 ¼ F þ G
the WWFE in trying to understand the reasons for the lack of val-
2F ¼ Y12 þ Z12  X12
idation. To begin, historical development of failure theories for ð3Þ
composites is reviewed, examining the foundation of each theory. 2G ¼ Z12 þ X12  Y12
Attempt is made to unravel the underlying assumptions – explicit 2H ¼ X12 þ Y12  Z12
or implicit – with the aim to gain insight into the limitations and
relevance in each case. Following that, a comprehensive scheme While in isotropic yielding with no Bauschinger effect the initiation
for failure assessment is proposed and its ingredients are of yielding in pure shear is related to the tensile yield strength, this
illustrated by examples. is not the case for anisotropic yielding. Hill [4] assumed three
It is noted that the objective here is not to provide a compre- independent yield strength values corresponding to pure shear in
hensive review of all failure theories proposed so far. For this the the three orthogonal symmetry planes, and denoting these by R, S
reader is referred to WWFE [1,2]. Rather, four theories are focused and T, obtained,
on to examine the fundamental assumptions. It is believed that this 1 1 1
close examination suffices to reveal the nature of the limitations 2L ¼ ; 2M ¼ ; 2N ¼ ð4Þ
R2 S2 T2
inherent in the approach to which these theories, and other later
ones, belong. Hill [4] further showed that if rotational symmetry about one of the
symmetry axes exists in a plane normal to the axis, then the six
yield strength constants will reduce to four independent constants.
2. The early failure theories The two relationships between the yield strength constants in case
of rotational symmetry about the x-axis are
Among the early attempts to develop failure theories for
composites the work by Azzi and Tsai [3], published in 1965, is H ¼ G; L ¼ G þ 2F ð5Þ
192 R. Talreja / Composites Science and Technology 105 (2014) 190–201

If the x-axis is aligned with the fiber direction in a unidirectional elastic response in each of the loading cases terminates at the
composite, then the y–z plane would be the plane of isotropy, i.e. respective strength values. The sketches next to the stress–strain
the composite will be transversely isotropic. Azzi and Tsai [3] plots in each loading case illustrate the implied assumptions that
assumed this to be the case. They further recognized that the unidi- the fibers and matrix have been homogenized and at the failure
rectional composites used commonly as plies in a laminate are thin, stress in each case the homogenized composite fails by separation
and therefore, placing the z-axis in the thickness direction, they into pieces. Let us examine these assumptions a bit closer.
neglected the stress components rzz, szx and szy. Using this At rxx = X, tensile (or compressive) failure is assumed to occur
condition in Eq. (2) gives, as a single event when the composite loses its load carrying
capacity in the fiber direction. In reality the failure processes under
ðG þ HÞr2xx þ ðF þ HÞr2yy  2Hrxx ryy þ 2Ns2xy ¼ 1 ð6Þ tension and compression along fibers are quite different and in
each case there is progression of failure involved that is likely to
From the first of Eq. (5), i.e. H = G, it can be seen that using Eq. (3) affect the assumed linear stress–strain behavior. Fig. 3 schemati-
gives Z = Y. Thus, this is the consequence of transverse isotropy, cally depicts the failure progression in tension along fibers. The
and it is obvious on physical grounds. Using this condition in Eq. fibers fail initially at random sites depending on the distribution
(6), and further using Eqs. (3) and (4), gives the familiar form of of weak points. With increasing load more fibers fails under the
the so-called Tsai–Hill failure criterion, stated below. influence of local stress concentration by the pre-existing fiber
r2yy s2xy breaks. Final failure results from a cluster of failed fibers in a
r2xx rxx ryy
2
þ 2
 2
þ ¼1 ð7Þ cross-section aided by axial splits connecting other cross-sections
X Y X T2 with failed fibers. This progression of failure varies depending on
It is clear from the review above that Eq. (7) is in all essence the factors such as fiber properties, fiber volume fraction and distribu-
yield criterion for thin metal sheets with one preferred orientation tion, and quality of the fiber/matrix interface. The point to note in
lying in the plane of the sheet. The adaptation of this criterion to the context of the failure theory under discussion is that the
composite materials by Azzi and Tsai [3] consists of regarding the applied stress rxx does not cause failure at a deterministic single
preferred orientation as the fiber direction in a thin layer of unidi- value rxx = X, but that the failure process is progressive with a
rectional composite. In the following we examine the implications significant stochastic nature. This is not the case in yielding of
of Eq. (7) as a composite failure criterion. metals under tension.
The criterion expressed by Eq. (7) implies that failure occurs The case of failure under axial compression is illustrated in
when linear elastic response in any combination of the stress com- Fig. 4. As illustrated there, the two idealized micro-buckling defor-
ponents rxx, ryy and sxy reaches the limit condition given by Eq. (7). mation modes are the extension mode and shear mode, assumed
The three basic failure modes underlying the failure criterion are by Rosen [5] in his pioneering work of 1965. The shear mode can
tension (or compression) failure in the fiber direction, i.e. rxx = X, lead to localization of failure in a kink band, also illustrated in
tension (or compression) failure normal to fibers, i.e. ryy = Y, and Fig. 4. Argon [6] in 1972, and many since then, analyzed the
shear failure in the x–y plane, i.e. sxy = T. A graphical representation formation of kink band. The physical evidence of failure under
of the failure modes is given in Fig. 2. As depicted there, the linear compression from kink bands has also been reported by many,

Fig. 2. Graphical representation of basic failure modes in the failure criterion, Eq. (7).
R. Talreja / Composites Science and Technology 105 (2014) 190–201 193

Fig. 3. Schematic illustration of the fiber failure process in a unidirectional composite under axial tension.

Fig. 4. Failure modes in unidirectional composite under axial compression.

e.g., for carbon/epoxy by Shikhmanter et al. [7]. It is clear that the involves unstable crack growth, then the failure criterion ryy = Y
failure mechanisms and their progression in compression are and the stress–strain behavior depicted for this case in Fig. 2 can
different from those in tension. Also, the critical applied stress to be reasonably regarded as valid.
failure in compression is different from that in tension and is much For compressive transverse loading, ryy < 0, the mechanism of
less influenced by statistical effects. failure in a unidirectional composite is far from clear and likely
Let us consider next the failure under tension (or compression) not the same as in tension, and the critical stress to failure is also
normal to fibers, ryy = Y. For tension, ryy > 0, the failure mecha- generally different from that in tension. Observations of failure
nisms have been a subject of extensive studies since 1967 when under compression are usually difficult to make and therefore little
Adams and Doner [8] reported the local stress enhancement due evidence has been reported. Some indication that failure occurs on
to the presence of fibers. The purpose at this point is not to review a plane inclined to the loading direction suggests that shear stress
those studies, but to assess the validity of representing the plays a role. In any case, in the context of the failure criterion,
strength by ryy = Y in the context of the failure criterion given by Eq. (7), it can be stated that the failure mechanisms and strength
Eq. (7). From the observations of failure under transverse tension values under transverse loading are not the same in tension and
and analyses of mechanisms involved it can be stated that failure compression.
initiates as microcracks in the matrix or at the fiber/matrix inter- Finally, let us consider the so-called in-plane shear strength,
face, followed by linking of these microcracks into macrocracks sxy = T, for a unidirectional composite. Azzi and Tsai [3] assumed
that grow unstably to failure. An illustrative example of the coales- this to be a single-valued strength as depicted in Fig. 2. This implies
cence of microcracks into a macrocrack is shown in Fig. 5 taken that the composite deforms linearly until it reaches sxy = T when
from Gamstedt and Sjogren [9]. Thus if the failure under ryy > 0 failure occurs. The question to ask is: Is the failure caused by
194 R. Talreja / Composites Science and Technology 105 (2014) 190–201

or shear), this slip is driven by shear. For isotropic metals, an


equivalent stress can be defined from the stress state at any point
to express the shear-governed slip that causes yielding. The two
common ways for defining the equivalent stress is by the use of
the Tresca or von Mises criteria for yielding. In the former case,
the equivalent stress is the maximum shear stress, while in the lat-
ter case it is given by the energy of distortion. When a given metal
is anisotropic, more than one equivalent stress (or yield function)
results due to the directional dependence of yielding. However,
all equivalent stresses relate to the same underlying mechanism
of shear-driven slip. In a unidirectional composite, even under
applied shear stress there seems to be no shear-driven failure
mechanism, as discussed above.
As a summary of the discussion above, it can be stated that
while the criterion expressed by Eq. (7) is rooted in the mechanism
of yielding, and therefore is appropriate for orthotropic metal
sheets, its adaptation to failure of a unidirectional composite raises
severe doubts about its validity because of the diverse failure
mechanisms that operate under different imposed stress states.
The next failure theory of interest for unidirectional composites
appeared in 1967 by Hoffman [11]. This author assumed, without
physical evidence, that all failure modes in composites were brittle
fracture. This implied that the linear elastic response in each stress
component terminated abruptly at a limiting value (fracture
strength) for that stress component and that the fracture strength
in combined stresses could be formally represented by a ‘‘yield
condition’’ in spite of the assumed brittle fracture. Although Azzi
and Tsai [3] did not explicitly state these assumptions, they
essentially implied them by renaming yield stresses in Hill [4] as
the composite strength values. The essential difference between
Hoffman [11] and Azzi and Tsai [3] in this respect is that the former
assigned different strength in tension and compression in each
principal composite direction (i.e. along fibers and normal to
fibers). In the context of Hill [4] this meant that the Bauschinger
effect, which Hill neglected, was now included. Accordingly,
Fig. 5. (a) Fibre–matrix debonds and (b) matrix crack formed by coalescence of
Hoffman [11] added three linear terms to Eq. (2), increasing the
debonds [9]. number of material constants from six in Hill’s version to nine.
For a lamina (a unidirectional fiber-reinforced sheet in plane
stress) the modified criterion led to the graphical representation
in the form of an ellipsoid with its center shifted from the origin
of the orthogonal axes rxx, ryy, and sxy.
The ellipsoidal representation of the ‘‘failure surface’’ (or the
yield surface without subsequent plastic flow) became the focus
of several future efforts. Notable among these is the work of Tsai
and Wu [12] in 1971 where they assumed a scalar-valued function
f of stress tensor components associated with the failure surface,
given by
f ðrij Þ ¼ F ij rij þ F ijkl rij rkl ¼ 1 ð8Þ
Fig. 6. Multiple cracks inclined to the fiber axis initiated by an applied ‘‘axial’’ shear
stress [10]. where Fij and Fijkl are second order and fourth order strength ten-
sors, respectively. For unidirectional composites in plane stress,
Eq. (8) reduces to a quadratic equation

shear? Closer examination of failure surfaces indicates that this is F p rp þ F pq rp rq ¼ 1 ð9Þ


not the case. As illustrated in Fig. 6 [10], showing failure under where the Voigt notation, p, q = 1, 2, 6, is used such that r1 = r11 =
in-plane shear stress sxy (also referred to as ‘‘axial shear’’ stress rxx, r2 = r22 = ryy and r6 = r12 = sxy. In analytical geometry of
to emphasize that the failure is on the plane parallel to the fiber quadratic surfaces, this equation describes an ellipsoid in the (r1,
axis), microcracks form with their planes inclined to the fiber axis. r2, r6) coordinate axes when the following conditions among the
Thus the failure is not caused by the shear stress itself but by the strength coefficients are satisfied.
transformed tensile stresses resulting on those planes. The inclined
microcracks connect when their fronts are diverted in the axial F 11 F 22  F 212 > 0
direction, forming an apparent ‘‘axial’’ crack. This fact is important F 22 F 66  F 226 > 0 ð10Þ
for the validity of the failure strength sxy = T in the context of the
F 11 F 66  F 216 > 0
failure criterion expressed by Eq. (7). To better appreciate this
point, consider yielding in metals caused by crystalline slip. The geometrical interpretation of the strength coefficients is as
Irrespective of the imposed loading mode (tension, compression follows. F1, F2 and F6 describe the center of the ellipsoid, F11, F22
R. Talreja / Composites Science and Technology 105 (2014) 190–201 195

and F66 represent the size of the ellipsoid, and F12, F16 and F26 the governing stress interactions should be expressed by quadratic
describe the inclination of the ellipsoid with respect to the r1r2, polynomials since linear terms only would generally be insufficient
r1r6 and r2r6 coordinate planes, respectively. while more terms than the quadratic ones would make the result-
As shown in Tsai and Wu [12], five of the nine strength coeffi- ing failure criteria impractical. In formulating the quadratic failure
cients in Eq. (9) are expressed in terms of the strengths in the prin- criteria Hashin resorted to the stress invariants for transversely
cipal material directions by the following relationships. isotropic symmetry arguing that the unidirectional composites
0 0 usually have randomly distributed fibers in the cross section.
F 1 ¼ XXXX0 ; F 2 ¼ YYYY0
ð11Þ For plane stress conditions in a unidirectional composite,
F 11 ¼ XX1 0 ; F 22 ¼ YY1 0 ; F 66 ¼ T12 Hashin’s criterion for tensile fiber mode, r1 > 0, takes the following
simple quadratic form.
where the primed symbols stand for strength values in compres-
r 2 r 2
sion. The strength coefficients F6, F16 and F26 vanish since the sign 1 6
þ ¼1 ð13Þ
of the shear stress r6 is immaterial. Finally, the strength coefficient X T
F12 cannot be determined uniquely and must satisfy the first of the
For compression along fibers, r1 > 0 Hashin simply set r1 = Y0 ,
inequalities in Eq. (10).
without attempting to formulate the effect of shear stress on this
It is noted that Eq. (9) is invariant with respect to coordinate
failure mode on the grounds of insufficient knowledge of the effect.
transformation and can describe unequal strengths in tension
For matrix failure modes, Hashin speculated that a failure plane
and compression. In this respect it is a more versatile curve-fitting
passing through the matrix (i.e. between fibers) existed such that
tool for unidirectional composite strength than Hill’s yield criterion
the critical normal and shear traction components on the plane
for orthotropic metals, given by Eq. (7), and Hoffman’s modifica-
governed its inclination. The concept of failure (decohesion) on a
tion [11] of this criterion to include the Bauschinger effect (i.e.
plane is in fact classical, having been proposed for failure in soils
unequal tensile and compressive yield strengths). However, the
by Mohr in 1905 and earlier by Coulumb in 1776. If the tractions
fact remains that the ellipsoidal representation of the strength of
on the failure plane are expressed by tensor transformation in
thin sheets of unidirectional composites in the in-plane stress com-
terms of the stress components rij, then the failure function for this
ponents is only a postulate that is not motivated or supported by
mode can be written as
any physical consideration of the failure mechanisms.
gðr22 ; r33 ; r23 ; r12 ; r13 ; hÞ ¼ 1 ð14Þ
3. Failure theories with physical considerations where h is the angle of inclination between the transverse axis x2
and the normal to the plane.
Hashin [13] in 1980 examined the Tsai–Wu criterion [12] for its For plane stress conditions in a unidirectional composite,
ability to represent failure in unidirectional composites. He noted Hashin derived the following criterion for the tensile matrix mode,
that the representation of failure in plane stress by a single ellipsoi- r2 > 0
dal surface, given by Eq. (9), was not appropriate in all situations. A
r 2 r 2
particular difficulty, as he pointed out, was related to the strength 2 6
þ ¼1 ð15Þ
coefficient F12 in Eq. (9). As proposed by Tsai and Wu [12], this Y T
coefficient can be obtained by a biaxial test. However, as noted
and for the compressive failure mode, r2 < 0
by Hashin [13], values of F12 resulting from such tests can be unac- " #
 2 2 r 2
ceptable on physical grounds. He pointed out that for a biaxial ten- r2 Y0 r2 6
sion test, Eq. (9) gives F12 that depends on compressive strengths in 0 þ 1 þ ¼1 ð16Þ
2T 2T 0 Y 0
T
the fiber and transverse directions. This is clearly not reasonable. In
fact the difficulties with F12 lie in that it does not have a unique where the prime on Y represents the numerical value of the
value, but depends on the stress state used to evaluate it. In fact compressive strength normal to fibers, and the prime on T is the
it can have any value restricted by the first of the conditions in shear strength in a plane across fibers (‘‘transverse shear strength’’),
Eq. (10), i.e. while the unprimed value is the usual shear strength in a plane
1 1 parallel to fibers (‘‘axial shear strength’’).
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi < F 12 < pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð12Þ The problem of identifying the strength constants in Eq. (16)
XX 0 YY 0 XX 0 YY 0 rendered the criterion impracticable.
where Eq. (11) has been used and the strength values in tension and Based on his earlier work in German, Puck [14] took Hashin’s
compression are expressed with the same sign. The restriction on proposed approach to matrix failure modes and developed it fur-
F12 expressed by Eq. (12) thus involves tensile as well as compres- ther. Renaming Hashin’s matrix failure mode terminology as
sive strengths in the fiber and transverse directions. There is no evi- inter-fiber failure (IFF) modes, Puck and Schürmann [15] proposed
dence to support this behavior in unidirectional composites. In fact an elaborate procedure for evaluating the inclination of the failure
the said restriction is purely a result of the requirement that the plane and the critical tractions on the plane.
ellipsoidal surface is closed. It is thus not rooted in any observed To understand the nature of the Puck theory, see Fig. 7, which
or expected physical behavior of unidirectional composites in plane shows the three traction components rn, snl and snt, acting on
stress. the failure plane. These traction components along with the incli-
Hashin [13] proposed to not require a single (continuous) nation angle hfp of the plane are the variables in the Puck theory.
smooth surface as a representation of unidirectional composite As described in Puck et al. [16], a function fE(h), called the stress
strength but to instead allow piecewise smooth surfaces, with each exposure factor or ‘‘risk of fracture’’ is defined based on sugges-
branch representing a distinct failure mode. This keeps interac- tions in Hashin [13] and fracture is assumed on the plane where
tions of the stress components restricted to the relevant failure this function takes its maximum with respect to the angle h.
modes and avoids the difficulties of the type noted above with The seven parameters (or material constants) in the Puck theory
F12 in the Tsai–Wu criterion. He then proceeded to examine indi- for IFF are three strength constants and four so-called ‘‘inclination
vidual failure modes. parameters’’. The latter are characteristic angles in the ‘‘master
Hashin [13] proposed to separate the fiber failure modes from curves’’, which are graphical depictions of the assumed quadratic
the matrix failure modes. For each failure mode Hashin argued that interactions between normal and shear tractions on the failure
196 R. Talreja / Composites Science and Technology 105 (2014) 190–201

nonuniform) and the nature of interfaces (weak vs. strong) do


not explicitly affect the critical tractions on the assumed failure
plane or its orientation. The seven material parameters in the Puck
theory are assumed to carry all these effects. Obviously, a different
set of assumptions regarding the failure plane than those based on
Coulomb and Mohr concepts utilized in the Hashin and Puck
approaches will result in different material parameters. The lack
of explicit accounting of the presence of fibers thus also induces
lack of uniqueness in the material parameters. Furthermore, the
assumed absence of any progression in the failure process intro-
duces another approximation that cannot be easily assessed.

4. Mechanisms based failure analysis

The failure theories discussed above attempt to capture the


Fig. 7. Shown is the failure plane of inclination angle hfp with respect to the x3-axis
occurrence of failure in composites by a set of expressions that for-
and the resolved normal stress rn and shear stresses snl and snt acting on the plane. mulate functions of stress (or strain) based on assumed criticality
From [15]. conditions of failure. While this classical approach has worked well
for materials where a single underlying failure mechanism is at
play, e.g. yielding in metals, it embodies inherent limitations when
plane (see Fig. 8). The interaction between tensile rn and either of different mechanisms operate leading to multiple failure modes.
the two shear tractions on the failure plane is assumed elliptical, as Although some remedy to this situation was sought in Hashin
in most other theories, and this interaction for the case of compres- [13] by separating the fiber and matrix modes, a host of assump-
sive rn is taken to be parabolic. The argument for the parabolic tions became necessary to capture the conditions of criticality of
assumption is that the presence of compressive normal traction the failure modes and their interactions. This has opened the door
impedes failure induced by shear on the failure plane. The inclina- to other failure theories with other assumptions ranging in degrees
tion parameters (denoted by p with subscripts referring to tension, of complexities and ease (or difficulty) of determining the resulting
compression or shear) are the angles with respect to the rn-axis failure related constants.
the tangents to the master curves make at points of junction In the following a new way forward is proposed that has the
between the elliptical and parabolic parts of the master curves, potential to avoid ambiguities and uncertainties in failure predic-
as illustrated in Fig. 8. tions arising from assumptions concerning governing conditions
Significant test data are needed to ‘‘identify’’ the seven for different failure modes. Rather than making such assumptions
constants in the Puck theory. To what extent this is justified with the proposed approach aims at developing analyses of observed
respect to other failure theories is assessed elsewhere [1,2]. The failure mechanisms and transferring the outcome of those to meth-
objective of this work is to examine the assumptions in the light odologies for failure assessment. The elements of the approach can
of the observed failure mechanisms in composites in order to be viewed as remedies needed to rectify deficiencies of the classical
evaluate validity of the failure theories. For this purpose, the composite failure theories. These are described briefly first below
following comments are offered. followed by more details in the next section.
Both the Puck failure theory and Hashin’s approach to develop-
ing failure criteria for unidirectional composites consider compos- 4.1. Remedy #1: Multi-scale analysis of failure
ites as homogeneous solids. The presence of fibers is accounted for
by distinguishing fiber failure modes from the failure modes that Composite materials are heterogeneous solids with distinct
are assumed to occur without affecting fibers. The latter called interfaces that act as weak planes under favorable local stress con-
matrix failure modes in Hashin [13], and inter-fiber failure modes ditions. Under general imposed loading on a composite, failure of
in Puck et al. [14,15], recognize fibers only in terms of their these planes triggers subsequent events of the failure process.
direction such that the assumed failure plane is parallel to fibers, Although the failure progression can take different paths depend-
i.e. it does not cut across fibers. The matrix in which the assumed ing on the fiber architecture and the nature of the imposed loading,
failure plane lies does not have an explicit influence from fibers, the sequence of the events on those paths often involve linkages of
e.g. through perturbation of the local stress field or by fiber/matrix the cracks formed at interfaces and in the matrix, and fiber failures
interface failure. Thus the distribution of fibers (uniform vs. in the later stages. The state of criticality in the performance of the

Fig. 8. Sections of the master failure surface for w = 90° and w = 0°. From Puck et al. [16].
R. Talreja / Composites Science and Technology 105 (2014) 190–201 197

composite depends on the designed function, and it can be reached averages of the generally nonuniform local fields. It must also be
before what is conventionally described as failure, i.e. separation of noted that the local stress fields are generally triaxial, even when
the material in parts seen at the macro level. The intensity of the the applied macro level stress is uniaxial.
distributed micro-level cracks can, for instance, degrade the defor- To illustrate the multi-scale (i.e. micro–macro) approach to fail-
mational characteristics, measured as stiffness properties at the ure initiation consider the simple case of a UD composite subjected
macro level. The composite structure in some applications would to a uniform macro-level tensile stress normal to the fiber direc-
be considered to have failed to perform when that occurs to an tion, i.e. ryy > 0, while all other macro-level stresses are zero. As
undesirable level. In other situations, the inability to carry imposed noted in Section 2 above, failure under this imposed stress condi-
loads will be seen as failure to perform the designed function. In tion appears to occur from unstable growth of a crack that has
any case, determining the criticality conditions associated with formed by coalescence of smaller cracks at the fiber/matrix inter-
failure requires analyses of the first events of failure at the face (Fig. 5). If it can be assumed that a single failure initiation
micro-level and their subsequent development leading to macro mechanism exists, then the criticality of that mechanism can be
level failure. Thus the failure prediction necessarily involves a captured in the macro-level failure condition, ryy = Y, such as that
multi-scale analysis. given by the Tsai–Hill criterion, Eq. (7), or the Hashin criterion, Eq.
(15), where ryy = r2. However, as pointed out by Asp et al. [17], two
4.2. Remedy #2: Analysis of constrained failure competing mechanisms for failure initiation, namely, fiber/matrix
debonding and matrix failure are possible due to the local triaxial
Composite materials are designed with selected fiber architec- stress state. By systematically examining the local stress states in
ture to meet the needs imposed by the service environment. Thus, transversely loaded UD composites different fiber packing arrange-
as an example, thin plies of UD composites with straight fibers are ments and fiber volume fractions they found that the energy den-
stacked in different orientations to create laminates. Other more sity for dilatation (related to failure from triaxial tension) was
complex fiber architectures are generated by using woven fabrics generally lower than the energy density for distortion (related to
instead of straight fibers, as dictated by cost and performance yielding) in the matrix within the composite. Also, the locations
requirements. In any case, the failure process is significantly in the matrix within the composite where the largest values of
altered by the presence of interfaces between layers containing the two energies occurred were different. More importantly, the
straight fibers or woven fabrics. Until these interfaces fail, i.e. sites where the dilatation energy density attained high values were
delamination occurs, the failure process within the layers is sub- found to be close to the fiber surface in the matrix. Asp et al. [17]
jected to what is described as a mutual ‘‘constraint’’ imposed on then concluded that what appeared to be fiber/matrix debonding
each other by the layers with differently oriented fibers. The was indeed cavitation-induced brittle failure of the matrix, leading
classical failure theories, described above, are for ‘‘unconstrained’’ to fiber/matrix interface failure. Based on this, they proposed the
UD composites. It is common to apply these theories in a so-called critical dilatation energy density as the condition for initiation of
‘‘ply-by-ply’’ failure analysis of laminates. This ignores the failure cavitation-induced brittle failure in glassy polymers and verified
progression induced by the ply constraints and is therefore a the criterion by testing three different epoxies [18].
source of significant errors in the resulting failure predictions. The microstructure in the cross section of a UD composite was
represented in Asp et al. [17] by unit cells with three different
4.3. Remedy #3: Analysis of manufacturing defects fiber-packing arrangements. In real composites, however, the
fibers are nonuniformy distributed. Bulsara et al. [19] treated the
As noted before, current failure theories for composite materials problem of characterizing the nonuniform (arbitrary) fiber distri-
are formulated on homogenized solids with account made of bution by statistical methods that use the cross-sectional images
anisotropy induced by fiber orientations. Real composites, of actual composites for estimating the statistical parameters. A
however, contain defects resulting from the particular manufactur- basic issue in this approach is how large the field of observed
ing process. These defects either initiate failure or affect the failure images needs to be in order to be representative of the microstruc-
initiation from weak sites such as interfaces between constituents ture. Bulsara et al. [19] proposed that this size, generally described
and between layers. Traditional approach to analysis of defects has as the size of a representative volume element (RVE), is not an
been to embed selected defects into the homogenized solid for absolute size, but is dependent on the property to be evaluated.
assessing their effects. This is inadequate for analyzing the interac- Thus the minimum RVE size for estimating elastic properties is
tions between the defects and the composite microstructure. not the same as that needed to estimate initiation of failure. For
In the following, the three remedies will be discussed individu- failure estimation, too, the RVE size depends on the failure mode.
ally, illustrating these with examples. This will be followed by Bulsara et al. [19] studied the RVE size of a ceramic matrix compos-
discussion of an integrative approach to failure assessment. ite and applied appropriate failure criteria for two failure modes,
viz. fiber/matrix debonding and radial cracking. They evaluated
4.4. Multi-scale analysis of failure the level of the applied thermal or mechanical loading associated
with initiation of a given failure mode and suggested selecting
As noted above, failure in a composite material – or in a heter- the RVE size that minimizes the coefficient of variation of the esti-
ogeneous solid, generally – initiates either in a constituent or at an mated applied failure stress.
interface. Thus, failure initiation is local, i.e. at the microstructure Having described the background of the possible failure modes
level. When and where precisely failure initiates, however, in a polymer based UD composite subjected to transverse tension
depends on the local stress field and the criticality condition for we shall now examine these individually and then consider how
the potential failure mode. Therefore, homogenizing the micro- their competition to attain criticality can be evaluated by a
structure will generally not allow capturing the event of failure ini- multi-scale approach.
tiation. For the special case when only one failure mode is
operative, the condition for its occurrence can be expressed in an 4.5. Fiber/matrix debonding
average sense in terms of the macro-level (averaged) stress field.
In a general case, when multiple competing failure modes are Initiation of this failure mode may be defined as breakage
possible, the conditions for the winning (first) failure event cannot of the bond between fiber and matrix or unstable growth of a
always be expressed in terms of stress fields that represent pre-existing flaw (crack) at the interface. Assuming no flaws exist
198 R. Talreja / Composites Science and Technology 105 (2014) 190–201

at the interface, and the fiber is perfectly bonded to the matrix, the microstructure. For transversely loaded UD composites, the micro-
initiation of debonding will be given by exceeding the bond structure is the cross-section in which fibers are distributed in a
strength by the local stress acting on the interface. In a commer- manner that appropriately represents the actual composite manu-
cially produced composite, flaws at the interface are likely, in factured by a given process. Realizing that the fibers in an actual
which case the failure of the interface must be assessed by an composite are nonuniformly distributed, various methodologies
energy based criterion. The interface failure properties, i.e. the have been developed to statistically simulate the distribution. This
bond strength or the interface toughness (critical energy release problem was addressed in Bulsara et al. [19], where the size of the
rate), cannot be determined accurately by theoretical means. Sev- RVE was also treated. Based on that work, Bulsara and Talreja
eral experimental methods have therefore been devised (for a (unpublished work) have examined the failure initiation in UD
review, see Zhandarov and Mader [20]). However, applying the composites under transverse tension considering the cavitation-
interface toughness criterion for evaluating initiation of fiber/ induced brittle failure and ductile matrix failure as two competing
matrix debonding faces difficulties due to uncertainty of knowing modes. Taking the same material properties and fiber volume frac-
the flaw size and its variability. tion of a glass/epoxy composite as in Asp et al. [17], an RVE (cross-
section of a UD composite) was generated with 40 fibers. Fig. 9
4.6. Cavitation-induced brittle matrix failure shows one realization of the RVE for illustration. Since the fibers
are long, a plane strain state is assumed in the finite element
As clarified by Asp et al. [21], cavity formation is possible in a (FE) analysis. The RVE is subjected to a thermal cooldown of
polymer if the stress state is equi-triaxial tension, or nearly so. 82 °C and a uniaxial transverse displacement, which is increased
Under this stress state the strain energy density equals the monotonically.
dilatation energy density, i.e. the energy of distortion vanishes, At every incremental step of 0.1% average mechanical strain in
resulting in a purely elastic behavior. Thus with no driving force the RVE, the local stresses in the matrix are calculated and con-
for inelasticity, the material behaves in a perfectly brittle manner. verted to principal stresses r1, r2 and r3. The dilatation energy
The growth of the formed cavity under this condition occurs by density, given by
increase of its surface, which for its expansion draws upon the
1  2m
strain energy from the surrounding material. It is reasonable to Uv ¼ ðr1 þ r2 þ r3 Þ2 ð17Þ
assume that a limit exists to the extent the cavity can expand 6E
before its growth becomes unstable. Asp et al. [17] assumed the where E and m are the Young’s modulus and Poisson’s ratio of the
value of the dilatation energy density corresponding to this unsta- matrix, respectively, is calculated and its maximum value occurring
ble state as the critical value for cavitation-induced brittle failure within the RVE is determined. In Fig. 10, the principal stress ratios
of the polymer. As shown by them, in a UD composite subjected at the locations of maximum Uv are plotted, and the ratio of the
to transverse tension, this failure mode tends to occur close to maximum principal stress to the mean stress is also indicated. As
the fiber surface, resulting in apparent fiber/matrix debonding. can be seen, there are only certain strain levels where the principal
stresses associated with the maximum dilatational strain energy
4.7. Ductile matrix failure density are nearly equal. Thus the favorable value of the applied
strain to induce cavitation falls in the range of 0.4–0.5%. The maxi-
In resin-rich regions of a composite, the stress state in the mum dilatation energy density in this range of applied strain was
matrix will tend to approach the applied remote stress. Thus if a found to vary between 0.146 and 0.195 MPa. The critical value of
uniaxial tensile stress is applied transverse to the UD composite, the dilatation energy density for cavitation, as determined by the
the matrix polymer will deform in a resin-rich region much like poker-chip test for this epoxy was 0.17 MPa (Asp et al. [18]). The
in a specimen under uniaxial tension. It will then yield at the yield experimental value of the uniaxial strain at which failure occurs
stress obtained by a standard tensile test. However, the presence of in the UD glass/epoxy composite in transverse tension also agrees
fibers enhances the stress in the matrix close to the fiber surface with the 0.4–0.5% range (Asp et al. [17]).
and renders the stress state triaxial. Under this stress state, yield- Fig. 11 shows how the absolute maximum values of the dilata-
ing initiates when a critical condition is reached. Asp et al. [18] tional strain energy density and the distortional energy density at
studied the yielding behavior of three epoxies under uniaxial, the fiber/matrix interfaces compared within the RVE. As seen, the
biaxial and triaxial stress states and found that yielding in these
polymers is driven by the energy of distortion and is influenced
by the energy of dilatation.

4.7.1. Determining the governing failure mode in transversely loaded


UD composite
Summarizing the discussion above, it can be stated that in a UD
composite subjected to transverse tension the three failure modes
will generally compete and the failure initiation will occur when
the condition for occurrence of any of the modes is met first. For
the case of fiber/matrix debonding it is difficult to know whether
this failure mode occurred by itself or if it was triggered by cavita-
tion-induced brittle failure of the matrix close to the fiber surface.
The studies by Asp et al. [17,18] suggested strongly that the latter
is the case at least for the three epoxies examined. Assuming there-
fore that the interface failure by cavitation-induced brittle cracking
and ductile matrix failure are the two competing failure modes, we
will describe below the process of analyzing failure initiation in
transversely loaded UD composites.
To simulate the local conditions that initiate failure the first Fig. 9. A representative cross-section of a UD composite constructed by randomly
step in the failure analysis is to construct a representative distributing forty fibers in the matrix.
R. Talreja / Composites Science and Technology 105 (2014) 190–201 199

The multi-scale analysis illustrated above provides failure initi-


ation conditions for a UD composite under in-plane macro-level
stresses. From failure initiation to final failure (separation) a pro-
gression in failure is expected. The significance of the progression
part will depend on the failure mode. For instance, only a small
increment in the applied load is needed after initiation of the
fiber/matrix debonding to induce linkup of the macro-level debond
cracks to form a macrocrack that grows unstably, causing final fail-
ure. This is due to the failure progression being largely under brit-
tle failure conditions. The ductile matrix failure mode, on the other
hand, has significant failure progression involving void growth and
coalescence with consequent stable crack growth.
When a UD composite is in the form of a ply (or a set of aligned
plies) within a laminate, it experiences in-plane stresses resulting
from loads on the laminate. The initiation of failure in a given
ply of a laminate can then be analyzed by the multi-scale failure
analysis by imposing these stresses on the RVE boundary. How-
Fig. 10. Principal stress ratios plotted against the applied uniaxial strain to the RVE ever, such analysis will only provide failure initiation conditions,
shown in Fig. 9. The ratio of the maximum principal stress to the mean stress is also as discussed above. The subsequent failure progression will be dif-
shown. Zero applied strain corresponds to thermal residual stress state due to ferent from what occurs in a UD composite in the free condition,
cooldown.
i.e. not bonded to differently oriented plies in a laminate. The ply
failure in a laminate is a constrained failure process whose analysis
will be discussed next.

4.8. Analysis of constrained failure

The constrained ply failure process in laminates has been stud-


ied experimentally and analyzed by a variety of models since the
beginning of 1970s. An excellent summary of this process is cap-
tured in Fig. 12 (Jamison et al. [22]), which qualitatively depicts
the stages in the progression of the process. Although it was devel-
oped based on observations of failure mechanisms in tension–ten-
sion fatigue of laminates, it is illustrative of the failure process in
quasi-static loading as well. Overall, the early part of the process
involves multiplication of cracks within the plies of a laminate,
leading to saturation of the crack density (number of cracks per
unit distance normal to the crack planes) in individual plies. The
end of this distributed cracking process has been called the charac-
teristic damage state (CDS) [23]. On continued loading, the ply
Fig. 11. Variations of the maximum values of dilatational and distortional strain cracks tend to divert into the ply interfaces, thereby connecting
energy densities occurring locally within the RVE shown in Fig. 9 with the applied with neighboring cracks. Further interface cracking leads to sepa-
transverse strain.
ration of plies (delamination), and the consequent fiber breakage
in plies resulting in final failure (separation).
maximum distortional energy density is consistently higher than
the maximum dilatational energy density at all strain levels. More
importantly, the critical value of the distortional energy density for
the epoxy considered here is 0.7 MPa (Asp et al. [18]), correspond-
ing to which the yielding in the RVE will initiate at the applied
strain of approximately 0.9%, which is much higher than the strain
that initiates cavitation-induced brittle failure.
The failure mode analysis described above was for UD compos-
ites under transverse tension, ryy > 0. This case was taken to illus-
trate the procedure for multi-scale failure analysis of UD
composites. In the more general case when the composite is sub-
jected to a biaxial stress state, i.e., when the macro-level stresses
rxx, ryy and sxy act simultaneously, the procedure will consist of
analyzing the operating failure modes corresponding to a given
stress component under the influence of the other stress compo-
nents. For example, the UD composite failure modes under trans-
verse tension discussed above must be analyzed by considering
the RVE with all three macro-level stress components as boundary
tractions. It can be expected that the fiber direction stress rxx will
have negligible effect on the fiber/matrix debonding and ductile
matrix failure modes. The effect of the shear stress sxy, on the other Fig. 12. A schematic depiction of progression of failure events in a general laminate
hand, will be significant. subjected to axial static or cyclic tension (from Jamison et al. [22]).
200 R. Talreja / Composites Science and Technology 105 (2014) 190–201

From the point of view of load bearing capacity (‘‘strength’’) the local failure (Chowdhury et al. [28]), where specific interactions
early stage of the failure process until CDS is of little consequence, between voids and fibers in a transversely loaded UD composite
as the cracked plies retain largely their share of the load. Instead, were examined. Other studies focused on examining the effects
the deformational response of the laminate, measured at the macro of voids on crack growth in resin rich areas of a composite in mode
scale, shows changes that may have bearing on the laminate per- I cracking (Ricotta et al. [29]) and due to out-of-plane delamination
formance, particularly if the laminate is designed for a purpose buckling (Zhuang and Talreja [30]) have also been performed.
where its deformational characteristics are critical (such as in In the early studies of the effects of defects it was common to
vibrational response). The post-CDS behavior is much more critical introduce artificial (simulated) defects into a composite to deter-
for the load bearing function, and its analysis may require knowl- mine the extent of degradation caused on failure related proper-
edge of the ply-cracking phase. In the following, therefore, we shall ties. Such studies provided evidence of the effects but could not
review the analysis of multiple ply cracking. be used to predict those effects in real composites. In recent years
The initiation of the first crack formation in a ply is governed by the actual defects and irregularities resulting from manufacturing
the local stress state and can be analyzed by the multi-scale can be observed and quantified, e.g. for voids (Lambert et al.
approach discussed above for the case of failure initiation in a free [31]; Scott et al. [32]) and fiber waviness (Wang et al. [33]). At
(unconstrained) ply. As noted above, if the failure initiation process the same time, the available computational power gives possibili-
in a UD composite is fiber/matrix debonding, then its subsequent ties of incorporating actual or statistically equivalent defect popu-
development to unstable crack growth is almost immediate. When lations in RVE based analyses of stress fields and consequent
the same UD composite is in a constrained environment within a failure initiation. The work begun in Chowdhury et al. [28] has
laminate, the failure initiation process will still be the same if the set the direction in which much more needs to be done in order
ply thickness is sufficient for the RVE to apply. The effect of con- to derive conditions of failure criticality at the macro level.
straint is to reduce the crack surface separation when a crack has
formed. Furthermore, the presence of ply interfaces arrests the 5. A comprehensive strategy for composite failure analysis
crack growth in the ply thickness direction, leading the further
crack growth to become unidirectional in the direction of fibers Fig. 13 depicts a comprehensive strategy for analyzing failure of
in the ply. This crack growth has in a part of the literature been composite laminates. It integrates the three remedies to the cur-
described as ‘‘tunneling’’. These tunneling cracks form initially at rent failure theories discussed above. The first steps in the failure
random locations in a ply, without interaction with one another, analysis consist of determining stress fields within appropriate
but once their mutual spacing reduces, interactions occur. The RVEs of UD composite microstructure and attendant defects in
interactive cracking process generates more cracks with reducing order to ascertain initiation of failure modes. These modes can be
average crack spacing, tending to a limiting value. The vast litera- separated into fiber failure modes and failure modes related to
ture related to the multiple ply cracking has been comprehensively matrix and fiber/matrix interface failures. The instabilities in the
reviewed in Talreja and Singh [24]. progression of the failure modes will provide the failure conditions
The failure theories for UD composites discussed in Section 2 for the UD composite. These conditions are expected to be certain
above have been applied to the multiple ply cracking in a rather relations in terms of the applied (uniform) tractions on the RVE
crude way. Since only one failure condition called ‘‘strength’’ can boundaries and appropriate material constants such as the critical
be treated, one applies the criterion of a given theory assuming a value of the dilatation energy density for the matrix polymer. For
constrained ply in a laminate to have ‘‘failed’’ when that failure fiber failure modes the relations are expected to involve statistical
criterion is satisfied. One then proceeds ‘‘ply-by-ply’’ in the fiber failure distribution parameters. Note that these conditions are
sequence in which plies meet the given failure criterion, after the not the same as analytical expressions (criteria) in the current
stress state in a ply under consideration has been recalculated by failure theories.
‘‘discounting’’ (i.e. removing one or more elastic properties) of
the already failed ply or plies. The failure of the last surviving ply
then determines the ‘‘strength’’ of the laminate. This so-called
‘‘ply-discount method’’ can be said to have no basis in the actual
failure process for laminates described above.
The load bearing capacity of a laminate is depleted for the most
part by the localized mechanisms occurring after CDS. Analysis of
these mechanisms is a complex fracture mechanics problem
involving interconnected cracks of different orientations.
Addressing this problem with a macro-level failure theory poses
a great challenge.

4.9. Effects of defects on composite failure

Practical composites cannot be manufactured without defects.


As reducing the manufacturing cost of composite structures
becomes increasingly imperative, the challenge for the failure ana-
lyst is to assess the effects of defects on failure in order to conduct
a cost/performance trade-off (Talreja [25,26]). It is obvious that the
failure theories reviewed in Section 2 above are incapable of
accounting for defects for the simple reason that they are formu-
lated on homogenized composites wherein no direct knowledge
of the defects is retained. It is possible only in a multi-scale analy-
sis to incorporate defects along with other microstructure details.
Analyses of this nature have been performed for voids in UD com-
posites to evaluate elastic properties (Huang and Talreja [27]) and Fig. 13. A comprehensive failure analysis scheme for composite laminates.
R. Talreja / Composites Science and Technology 105 (2014) 190–201 201

6. Conclusion [10] Gamstedt EK, Sørensen BF. Interaction of damage development and dissipation
in tensile fatigue loading of composite laminates. In: Miyano Y, Cardon AH,
Reifsnider KL, Fukuda H, Ogihara S, editors. Durability analysis of composite
This paper has performed a critical examination of phenomeno- systems. Lisse: A.A. Balkema; 2002. p. 213–20.
logical theories for failure of composite materials. By reviewing the [11] Hoffman O. The brittle strength of orthotropic materials. J Compos Mater
1967;1:200–6.
underlying assumptions of the theories in light of the current
[12] Tsai SW, Wu EM. A general theory of strength for anisotropic materials. J
understanding of the failure mechanisms it is found that the early Compos Mater 1971;5:58–80.
theories, such as Tsai–Hill and Tsai–Wu, do not capture the physi- [13] Hashin Z. Failure criteria for unidirectional fiber composites. J Appl Mech
1980;47:329–34.
cal nature of the mechanisms, while the later theories, exemplified
[14] Puck A. Calculating the strength of glass fibre/plastic laminates under
by Hashin and Puck, attempt to incorporate some knowledge of the combined load. Kunststoffe German Plast 1969;55:18–9 (German text p.
mechanisms, but are severely limited by the formulation of the 780–87).
theories being on homogenized composites. Another problem with [15] Puck A, Schürmann H. Failure analysis of FRP laminates by means of physically
based phenomenological models. Compos Sci Technol 1998;58:1045–67.
the theories lies in their inadequacy to apply for laminated [16] Puck A, Kopp J, Knops M. Guidelines for the determination of the parameters in
structures due to the lack of proper account of the constrained Puck’s action plane strength criterion. Compos Sci Technol 2002;62:371–8.
ply cracking and interlaminar failure modes. [17] Asp L, Berglund LA, Talreja R. Prediction of matrix initiated transverse failure in
polymer composites. Compos Sci Technol 1996;56:1089–97.
A way forward to a different failure analysis paradigm has been [18] Asp L, Berglund LA, Talreja R. A criterion for crack initiation in glassy polymers
proposed that makes use of current observation techniques to subjected to a composite-like stress state. Compos Sci Technol
delineate microscopic details of the failure mechanisms as well 1996;56:1291–301.
[19] Bulsara VN, Talreja R, Qu J. Damage initiation under transverse loading of
as to characterize manufacturing induced defects and the available unidirectional composites with arbitrarily distributed fibers. Compos Sci
computational power for stress analysis. Remedies for overcoming Technol 1999;59:673–82.
deficiencies of the failure theories have been offered and illustrated [20] Zhandarov S, Mader E. Characterization of fiber/matrix interface strength:
applicability of different tests. Approaches and parameters. Compos Sci
by examples. An integrative and comprehensive scheme for failure
Technol 2005;65:149–60.
assessment of composite laminates has then been proposed that [21] Asp LE, Berglund LA, Gudmundson P. Effects of a composite-like stress state on
has the potential to provide robust and reliable failure prediction. the fracture of epoxies. Compos Sci Technol 1995;53:27–37.
[22] Jamison RD, Schulte K, Reifsnider KL, Stinchcomb WW. Characterization and
It is noted that the proposed failure analysis paradigm, for
analysis of damage mechanisms in tension–tension fatigue of graphite/epoxy
which the work is ongoing, is not likely to result in expressions laminates. Effects of Defects in Composite Materials. ASTM STP 836; 1984. p.
of failure criteria that would replace the current failure criteria. 21–55.
Instead, the three remedies proposed are likely to become an inte- [23] Reifsnider KL. Some fundamental aspects of fatigue and fracture response of
composite materials. In: Proceedings of the 14th annual meeting. Society of
gral part of computational structural analysis that assesses failure Engineering Science, Lehigh University; 1977. p. 343–84.
by a multi-scale approach with proper account of the operating [24] Talreja R, Singh CV. Damage and failure of composite
failure mechanisms and the manufacturing induced defects. materials. Cambridge: Cambridge University Press; 2012.
[25] Talreja R. Defect damage mechanics: broader strategy for performance
evaluation of composites. Plast Rubber Compos 2009;38:49–54.
References [26] Talreja R. Studies on the failure analysis of composite materials with
manufacturing defects. Mech Compos Mater 2013;49:35–44.
[1] Hinton MJ, Kaddour AS, Soden PD, editors. Failure criteria in fibre reinforced [27] Huang H, Talreja R. Effects of void geometry on elastic properties of
polymer composites: the world-wide failure exercise. Amsterdam: Elsevier; unidirectional fiber reinforced composites. Compos Sci Technol
2004. 2005;65:1964–81.
[2] Soden PD, Kaddour AS, Hinton MJ. Recommendations for designers and [28] Chowdhury KA, Talreja R, Benzerga AA. Effects of manufacturing-induced
researchers resulting from the world-wide failure exercise. Compos Sci voids on local failure in polymer-based composites. J Eng Mater Technol
Technol 2004;64:589–604. 2008;130:021010–7.
[3] Azzi VD, Tsai S. Anisotropic strength of composites. Exp Mech 1965;5(283– [29] Ricotta M, Quaresimin M, Talreja R. Mode I strain energy release rate in
28):8. composite laminates in the presence of voids. Compos Sci Technol
[4] Hill R. A theory of the yielding and plastic flow of anisotropic materials. Proc 2008;68:2616–23.
Roy Soc A 1948;193:281–97. [30] Zhuang L, Talreja R. Effects of voids on postbuckling delamination growth in
[5] Rosen BW. Mechanics of composite strengthening. Fiber composite materials. unidirectional composites. Int J Solids Struct 2014;51:936–44.
In: Am. Soc. Metals Seminar, Metals Park (OH); 1965. p. 37–75. [31] Lambert J, Chambers AR, Sinclair I, Spearing SM. 3D damage characterisation
[6] Argon AS. Fracture of composites. Treatise Mater Sci Technol 1972;1:79–114. and the role of voids in the fatigue of wind turbine blade materials. Compos Sci
[7] Shikhmanter L, Eldror I, Cina B. Fractography of unidirectional CFRP Technol 2012;72:337–43.
composites. J Mater Sci 1989;24:167–72. [32] Scott AE, Sinclair I, Spearing SM, Mavrogordato MN, Hepples W. Influence of
[8] Adams DF, Doner DR. Transverse normal loading of a unidirectional composite. voids on damage mechanisms in carbon/epoxy composites determined via
J Compos Mater 1967;1:152–64. high resolution computed tomography. Compos Sci Technol 2014;90:147–53.
[9] Gamstedt EK, Sjogren BA. Micromechanisms in tension–compression fatigue [33] Wang J, Potter KD, Hazra K, Wisnom MR. Experimental fabrication and
composite laminates containing transverse plies. Compos Sci Technol characterization of out-of-plane fiber waviness in continuous fiber-reinforced
1999;59:167–78. composites. J Compos Mater 2012;46:2041–53.

View publication stats

Das könnte Ihnen auch gefallen