Failure Criteria

Damage and fracture of composite materials may occur in a

variety of failure modes:
- fiber breaking: occurs mainly under tensile loads
- matrix crazing: the appearance of microscopic cracks in
the polymer matrix
- matrix cracking: similar to crazing but the cracks are
larger
- fiber debonding: occurs when the fiber-matrix bond
fails

Difficult to incorporate these many modes of failure into the

design. Thus a simpler alternative is to use empirical failure
criteria

 Failure criteria empirical equations with parameters adjusted

to fit experimental data (obtained on simple tests (Fig 6) of
failure of single-layer composite

Fig 6: Simple test configurations used to determine strength values

 F1t = longitudinal tensile strength (in the fiber direction)

F1c = longitudinal compressive strength
F1t = transverse tensile strength
F2c = transverse compressive strength
F6 = inplane shear strength
F4, F5 = interlaminar shear strength

A few popular criterion:

Maximum stress criterion

Maximum strain criterion
Tsai-Hill failure criterion
Tsai-Wu failure criterion
Truncated-maximum strain criterion

 To use any failure criteria more efficiently, the strength ratio

(similar to safety factor), R is defined : R = sultimate / sapplied
(strength over stress applied)
R > 1, stress level is below the strength of the material
R < 1, stress value is higher than the strength and failure is predicted
Since linear elastic behaviour is assumed, multiplying all the applied loads by a factor R
increases or reduces the stresses proportionally, so that the new value of strength ratio = 1
Failure load can be computed simply by analysing the structure with an arbitrary reference
load, then multiplying the reference load by the value of R

Maximum Stress Criterion

Predicts failure of a layer when at least one of the stresses in
material coordinates (s1, s2, s3, s4, s5, s6) exceeds the
corresponding experimental value of strength.

The criterion states that failure occurs if any of the following is

true:
s1 > F1t
if s1 > 0
abs (s1) > F2c if s1 < 0
s2 > F2t if s2 > 0
abs (s2) > F2c if s2 < 0
abs (s4) > F4
abs (s5) > F5
abs (s6) > F6
6

 F1t, F1c, F2t, F2c, F4, F5, F6 are ultimate values, can also be replaced by
allowable values; if ultimate values are used, a safety factor needs to be
used in the design, if allowable, already have safety factors built in
(allowable = ultimate / safety factor)
A graphical representation of the maximum stress criterion can be obtained
when only 2 stress components are different from zero.

Fig 7 case of s1 = s4 = s5 = 0
No failure as long as s2, s6 define a point (called design state) inside
the rectangle
Failure is predicted when s2, s6 reach any point on the rectangle
(failure envelope)
The rectangle limits the design space
If allowable strength values are used instead of ultimate strength, the
design space becomes the safe region

Fig7: The maximum stress criterion in stress space s2 - s6

 In terms of strength ratio, the maximum stress criterion reads

R1 = F1t / s1 if s1 > 0
R1 = - F1c / s1 if s1 < 0
R2 = F2t / s2 if s2 > 0
R2 = - F2c / s2 if s2 < 0
R4 = F4 / abs (s4)
R5 = F5 / abs (s5)
R6 = F6 / abs (s6)
The strength ratio for the layer is the minimum of all Ri
computed

Advantages: (i) gives information about the mode of

failure
(ii) detect overdesign

 Minimum Ri corresponds to a particular mode of failure the

laminate can be changed to reduce likelihood of failure mode in
that.

- Example: change fibre angles to reinforce the

against the identified mode of failure or
in the desired orientation

laminate
add more material

A large strength ratio Ri indicates the overdesign of the

laminate

10

 Its not conservative for stress states that are not dominated by
one component of stress
Point P in Figure 7; both s2 and s6 are positive, their value
are close but still lower than the corresponding strength
values F2t and F6, predicts no failure. But experimental data
from biaxial strength tests show that there are interactions
effects which produce failure when two or more stress
components are close to their limits
When s2 becomes negative (compression), the maximum stress
criterion maybe too conservative (point Q) or it may
overpredict the strength of the structure (point R) as it does in
this example for tensile transverse load

11

Example 7.1
Use the maximum failure criterion to compute the tensile strength ratio for
a layer of carbon-epoxy subject to s1 = 200 MPa, s6 = 100 MPa, s2 = -50
MPa, s4 = s5 = 0. The strength values for the composite are: F1t = 1.826 GPa,
F1c = 1.134 GPa, F2t = 19 MPa, F2c = 131 MPa, F6 = 75 MPa. Give a
justification for the result.
Ans:
Compute the values of strength ratio corresponding to each orientation.
Since the stress is tensile along the fibre direction:
R1 = F1t / s1 = 1826 MPa / 200 MPa = 9.13
In the transverse direction, the stress is in compression, so
R2 = - F2c / s2 = -131 MPa / -50 MPa = 2.62
Since there is only one shear stress different from zero,
R6 = F6 / s6 = 75 MPa / 100 MPa = 0.75
and R4 and R5 are not used. Finally, the minimum strength ratio for the layer
is R = R6 = 0.75 and the mode of failure is inplane shear. The layer will fail
when the applied load is only 0.75 of the reference value, but the laminate is
overdesigned for the applied load because it could carry 9.13 times the
applied longitudinal load.

12

Maximum Strain Criterion

Most popular failure criterion in industry
In terms of strength ratio, the maximum strain criterion reads
R1 = e1t / e1 if e1 > 0
R1 = - e1c / e1 if e1 < 0
R2 = e2t / e2 if e2 > 0
R2 = - e2c / e2 if e2 < 0
R4 = g4u / abs (e4)
R5 = g5u / abs (e5)
R6 = g6u / abs (e6)
where e1t, e1c , e2t , e2c , g4u , g5u , g6u are strains to failure

13

 If the material is linear elastic up to failure, the strains to failure

are directly related to the ultimate strength values by:
e1t = F1t / E1
e1c = F1c / E1
e2t = F2t / E2
e2c = F2c / E2
g4u = F4 / G23
g5u = F5 / G13
g6u = F6 / G12
In both criteria, the maximum stress and strain give close but
not identical predictions of failure (Fig 8 represents the failure
envelope in the space e1 , e2; g4 = g5 = g6 = 0 Maximum strain
criterion is a rectangle defined by the elongations to failure e1t,
e2t, e1c, e2c)
14

Fig 8: Failure envelopes using the maximum strain and maximum

stress criteria in strain space e2 - e6 for carbon-epoxy

15

 To illustrate some differences between maximum stress and

strain criteria, the maximum stress and strain is represented in
Fig 8, assuming that the material is linear up to failure
A material with a stress given by point P is predicted to fail by
the maximum strain criterion but considered not to fail by the
maximum stress criterion

One of the reason for the use of the maximum strain criterion
is the non-linear behaviour of the composite
( Most polymer matrices exhibit either elastic non-linear or
plastic behaviour after a certain elongation)

16

Tsai-Hill Criterion
The maximum stress and strain criteria separate the failure
modes into fibre modes, represented by the fibre-direction
strengths, and matrix modes, represented by the transverse
strengths ----- they look at failure based on one stress (or
strain) at a time, ignoring the interaction between two stress
components
Tsai-Hill criterion is not conservative when 2 stress (or strain)
components are close to their ultimate values and both
stresses hasten the same type of failure (e.g. matrix failure)
example, in Fig 9, the presence of transverse stress deteriorates
the shear strength and vice versa

17

Fig 9: Tsai-Hill failure envelopes in s6 s2 space compared with

experimental data in the first quadrant

18

 The quadratic interaction criteria use in this approach to

address the interaction. However, including interaction
between 2 unrelated modes of failure (e.g. fibre failure and
matrix cracking) may be counterproductive and lead to poor
results --- development of separated failure criteria for the
matrix and fibres (discuss later)
Failure criteria are nothing more than curve fits for
experimental data. Along this lines, the following equation (eq.
1) is proposed to fit the experimental data:

where the superscript f is used to indicate that any state of

stress (sf1, sf2, sf6, sf4, sf5) is a state of stress that produces
failure (on the failure envelope)
19

 All the failure states of stress obtained by combinations of

those five stress components generate a closed surface (the
failure envelope) that separate the no-fail region from the
failure region
When only 2 stress components are different from zero, the
failure envelope generate by equation above is an ellipse as
shown in Fig 9. Good comparison between the predicted and
experimental data of failure stresses--both stresses s2 and s6
cause the same type of failure (matrix dominated failure)

Interaction between the 2 inplane stresses is included explicitly

(clearly) in the second term of equation above, even though s1
and s2 correspond to remarkably different modes of failure
led to poor fit

20

 For a particular design situation, if the design point is inside the

failure envelope the material does not fail, outside means
failure and some changes are needed, but the magnitude of
the necessary changes is not known
In order to have a useful criteria, eq. 1 is rewritten using the
strength ratio, R:

where si are the components of stress computed by an applied

reference load (eq. 2)

21

 The strength values used in the equation of R above are not

specified to be tensile or compressive strengths because this
criterion cannot incorporate different behaviour in tension and
compression
The strength values that are used in the Tsai-Hill criterion have
to be chosen to be either the tensile or the compressive
strength values
If the state of stress is tensile only (as in the case of a thin
pressure vessel under internal pressure), the tensile values (F1 =
F1t, F2 = F2t) can be used and similarly compressive only, the
compressive strength can be used

22

 If the state of stress is combined, the lower values of strength

along each direction would have to be used for safety, which
may lead to a very conservative design

Experimental data of glass-epoxy hoop-wound tubes with Vf =

0.65 is used in Fig 10 to compare various failure criteria. The
state of stress is such that s2 and s6 are applied simultaneously
while s1 = s4 = s5 = 0. Tsai-Hill correlates well with the data as
long as s2 > 0, using experimental strength data F2 = F2t = 40
MPa and F6 = 60 MPa in eq.1. Using compressive strength, F2c =
148 MPa produces poor correlation in the second quadrant (s2)
and leads to non-conservative predictions in the first quadrant
(s6). It is clear that this criterion is not adequate for materials
having different tensile and compressive strengths

23

Fig 10: Various failure envelopes in s6 s2 space compared with

experimental data

24

 Disadvantages:
(i) the mode of failure is no longer identified as it was in the
case of maximum stress or strain criteria.
(ii) does not take into account different behaviour in tension
and compression (shown in experimental data previously)

25

For this example, the maximum stress criterion predicts the minimum value of
R is R2 = F2t / s2 = 19/25 = 0.76 corresponding to failure in the transverse
direction (tensile), while R1 = 1826/50 = 36.52. Such large R1 indicates that a
hoop fibre orientation is not ideal for a pressure vessel.
26

Tsai-Wu Criterion
This criterion uses a complete quadratic expression to draw a
failure envelope that attempts to fit the experimental data
The criterion is written as (eq.3):

27

With

(eq.4)

28

 In eq.3, sf1, sf2, sf6, sf4, sf5 are the components of stress at any
point of the failure envelope (Fig 10). That is, any such state of
stress corresponds to failure of the material
As it can be seen from the definition of coefficients (eq.4), TsaiWu criterion accounts for different behaviour in tension and
compression

The interaction between two normal stresses is accounted for

with an independent coefficient f12 that must be measured
independently of the remaining strength properties.

29

 Since a biaxial test is needed for measurement of the

interaction coefficient, experimental data are not easily
available. An approximation of the interaction coefficient can
be obtained by requiring the interaction term in eq.3 to be
numerically equal to the interaction term in Tsai-Hill equation
(eq.1), or

leading to

(eq.5)

30

 In other words, if the normalised interaction coefficient f*12 is

set to -1/2, and the behaviour of the material is identical in
tension and compression (or all the stresses have the same
sign), then the Tsai-Wu criterion yields identical results as the
Tsai-Hill criterion
The Tsai-Wu criterion is superior to the Tsai-Hill criterion
because the differences in tension and compression behaviour
of the material are accounted for automatically. Furthermore,
there is still possibility of measuring the interaction coefficient,
thus improving the fit to the experimental data.

31

 Another approximation for the interaction coefficient is given

by (eq.6):

Once an equation that fits the experimental data has been

defined, the Tsai-Wu criterion is rewritten in a way that is
convenient for design
Since linear elastic behaviour up to failure is assumed, the
components stress at failure sif can be replaced by the product
of the strength ratio times the nominal stress R si , obtained by
performing an analysis with a reference load, the Tsai-Wu
failure criterion become (eq.7):

32

which can be written as (eq.8):

Solving the roots and taking the positive value, the strength
ratio is (eq.9)

The Tsai-Wu criterion seems to fit well the available

experimental data (Fig 10). A single value of strength ratio is
obtained with a simple quadratic equation, and it accounts for
different behaviour in tension and compression. The maximum
stress or strain criterion could be computed simultaneously to
provide information about the most likely mode of failure,
which will guide the designer in the optimisation of the
laminate
33

eq.4 and eq.5

eq.9

34

Finally, the burst pressure is pb = R pref = 5.27 MPa. The increase

of burst pressure with respect to example 7.2 can be explained
as follows. The failure in example 7.2 was dominated by the
tensile strength F2t = 19 MPa, as it was indicated by the
maximum stress criterion. By changing the state of stress to a
compressive stress s2, a much higher compression strength F2c
= 131 MPa allows a higher burst pressure of vessel

35

Fibre-Matrix Failure Criteria

1. Tsai-Hill and Tsai-Wu criteria do not explicitly differentiate
matrix failure from fibre failure quadratic criteria fit well
experimental data under biaxial loading as long as 2 interacting
stresses affect the same failure mechanism (example in Fig 9,
both stresses produce matrix failure)
When the 2 interacting failure mechanisms are different, such
as longitudinal failure F1t and transverse failure F2t, the
quadratic criteria forces an artificially smooth transition from
one failure into the other (example 7.1, F1t of carbon-epoxy is
so high compared with F2t that failure is unlikely affected by
insignificant transverse stresses)

36

2. Failure prediction using quadratic criteria for combined states

of stress depends on strength data of completely unrelated
failure modes, since the quadratic criteria describe failure by
one equation that contains all the strength values (example,
tensile properties, F1t and F2t influence the failure predictions
for all s1 and s2, even in the compression-compression
quadrant (s1 < 0, s2 < 0) where real failure is unlikely to be
related to tensile properties)

37

3.Quadratic failure criteria model the failure envelope with a

smooth surface, represented as an eclipse (Fig 9) implies that
there are no abrupt changes in failure mechanism when the
various components of stress or strain change sign each
failure mechanism is assumed to have a gradually increasing
influence on the neighboring failure modes, proportional to the
values of the stress components and inversely proportional to
the strength values (eq.2) unlikely cause there are drastic
differences among the various failure modes

38

 To differentiate between fibre and matrix failure, the quadratic

expressions can be divided into 2 separate criteria. Using
notation of eq.4 and eliminating the interaction term, f12,
decomposes in 2 equations:
Fibre failure
(f11s12)R2 + (f1s1)R 1 = 0
Matrix failure
(f22s22 + f66s62 + f44s42 + f55s52 )R2 + (f2s2)R 1 = 0
Both of which can be solved for the strength ratio R. This
gives the designer the choice to evaluate the two major
modes of failure separately

39