Sei sulla pagina 1di 40

COMPOSITE MECHANICS

&
LAMINATE THEORY

Robin Olsson

Department of Aeronautics, Imperial College, 2006

Updated and corrected at Swerea SICOMP, March 2016b

1
1 Introduction ................................................................................................................ 3
1.1 Definitions ............................................................................................................................ 3
1.2 History and everyday applications ....................................................................................... 3
1.3 Advanced composites .......................................................................................................... 4

2 Materials and manufacturing ..................................................................................... 5


2.1 Fibre materials...................................................................................................................... 5
2.2 Matrix materials ................................................................................................................... 6
2.3 Fibre reinforced composites ................................................................................................. 6
2.4 Manufacturing techniques .................................................................................................... 8

3 Micromechanics of a ply ............................................................................................ 9


3.1 Volume and weight fractions ............................................................................................... 9
3.2 Models for elastic properties of a ply................................................................................... 9
3.3 Models for strength of a ply ............................................................................................... 12

4 Macromechanics of a ply ......................................................................................... 14


4.1 Stress and strain vectors ..................................................................................................... 14
4.2 Material classes .................................................................................................................. 15
4.3 Orthotropic ply under plane stress ..................................................................................... 16
4.4 Effect of hygrothermal strains............................................................................................ 19

5 Macromechanics of a laminate ................................................................................ 20


5.1 Forces and moments........................................................................................................... 20
5.2 Strains and curvatures ........................................................................................................ 20
5.3 Force-deformation relations ............................................................................................... 20
5.4 Some layup types and their effect on coupling .................................................................. 22
5.5 Estimates of laminate stiffness ........................................................................................... 23
5.6 Apparent engineering properties of laminates ................................................................... 23

6 Failure criteria and design ....................................................................................... 24


6.1 Peculiarities in failure of composites ................................................................................. 24
6.2 Ply failure criteria............................................................................................................... 25
6.3 Effect of ply failure in laminates........................................................................................ 28
6.4 Design methods .................................................................................................................. 30

7 Introduction to advanced topics ............................................................................. 32


7.1 Effect of stiffness gradients in laminates ........................................................................... 32
7.2 Notched strength of composite laminates .......................................................................... 32
7.3 Interlaminar stresses ........................................................................................................... 35
7.4 Delamination growth.......................................................................................................... 37

2
1 Introduction
1.1 Definitions
Composite materials are materials which on a microscopic level consist of two or more separate
materials. Sometimes it is suggested that the properties of the composite should be better than that
of the individual materials, but this may be an unnecessary limitation.

A B
+
=

Particulate Short fibre Continuous fibre Laminated

Examples
(Concrete) (Plastic boat hulls) (Reinforced concrete) (Plywood)

Fig. 1.1 Overview of different classes of composite materials

In some sense even alloys may be considered as composites, but normally we refer to materials
where reinforcing materials are embedded in a clearly separate bulk material called matrix. The
main function of the matrix is to transfer load between the reinforcements.

One way to classify composites is with respect to the geometry of the reinforcing material, e.g.
particulate composites, flake composites, fibre composites and laminar composites.

Composite laminates are obtained by bonding several composite laminae (or plies) on top of each
other. Each lamina (or ply) may be made of any type of composite, although fibre composites are
the most common. In the following we will use the terms laminate and ply to avoid the confusion
between laminate and lamina.

1.2 History and everyday applications


Nature has used composites since life emerged. Wood is a natural composite composed of cellulose
fibres in a lignin matrix. The first man made composite was probably straw-reinforced clay used to
build houses. Here straws with good tensile strength have been combined with clay, which has good
compressive properties and protects from wind and rain. A more modern example is steel reinforced
concrete.

Wooden laminates were used in the Second World War (e.g. on Spitfire) to obtain a stiff and light
structure. Today everyday applications of composite materials include fishing rods, skis, tennis
rackets, plastic boats and many parts on modern cars.

3
1.3 Advanced composites
Advanced composites has become the accepted term for materials with at least 50% of long or
continuous high modulus/high stiffness fibres embedded in a compatible matrix. Advanced
composites are typically used in fighter aircraft wings, pressure vessels, racing cars, boat hulls,
artificial limbs and other elements which need high stiffness and strength at a low weight.

There are several motivations for using thin long fibres as reinforcement. Thin fibres have less
defects and therefore a higher strength. The manufacturing by pulling thick fibres to gradually
smaller dimensions often introduces a beneficial orientation of molecules which increases stiffness
and strength. The load in a fibre increases gradually from the fibre end under the action of stresses
transferred by the matrix. Long fibres are more effective since the relative influence of this partially
ineffective end region is smaller.

Ineffective
length

Fig. 1.2 Stress transfer between matrix and fibres

4
2 Materials and manufacturing
2.1 Fibre materials
Glass fibres were introduced as reinforcement for plastic materials in the sixties, and is still the
most commonly used reinforcement. Several types of glass are used. E-glass (electrical) has low
alkali content and good electrical mechanical and chemical properties. C-glass (chemical) has
particularly good chemical resistance and is often used for surface protection of E-glass. S-glass
(strong) and R-glass have high mechanical properties and are mainly used in aerospace. Glass fibres
are cheaper than carbon fibres and have a similar strength but have significantly lower stiffness,
chemical resistance and fatigue properties. Typical fibre diameters are about 10 m.

Carbon fibres are produced by controlled oxidisation and carbonisation of cellulose and
polyacrylonitrile materials known as precursors. The use of temperatures up to 2600C produces a
high strength (HS) fibre, subsequent heating up to 3000C converts the fibre into a high modulus
(HM) fibre. Carbon fibres are, in contrast to glass fibres, anisotropic with a significantly lower
radial stiffness. Stiffness, chemical resistance and fatigue properties are excellent but the impact
resistance of carbon fibre composites is not so good. Typical fibre diameters are 5-10 m.

Boron fibres have properties similar to carbon fibres but their larger diameter (~100 m) results in
better compressive properties. The high price makes them unattractive for most applications.

Aramid fibres fibres have better strength/weight ratio than glass and an excellent abrasion
resistance. The most common trade name is Kevlar. Aramid fibres have a moderate price and a
tensile strength comparable to carbon fibres, but the compressive strength is low. The high tensile
failure strain makes them attractive in ballistic protection. A common fibre diameter is 12 m.

Other non-metallic fibres in use include polyester, nylon, cotton, sisal, jute and asbestos.

Metal fibres are often used in high temperature ceramic composites. The stiffness is similar to the
bulk materials but strength usually better due to fewer flaws. A comparison between the specific
strength and stiffness of some fibre composites and some common metals is given below.
Specific strength and stiffness along fibres
(60% glass, aram id and carbon in epoxy)
1200
Specific strength Y/

1000 Aramid HS Carbon


R-glass
800
[J/kg]

600 HM Carbon
E-glass
400

200 Steel
Aluminium
0
0 50 100 150
Specific stiffness E / [J/kg]

Fig. 2.1 Comparison of specific stiffness and strength for some composites and metals

5
2.2 Matrix materials
Thermosetting resins are low viscosity polymers, which become rigid by formation of molecular
cross-links during curing and do not soften upon subsequent re-heating. The most common
thermosets are polyester and epoxy resins. Other thermoset resins include vinyl esters, furanes and
phenolic resins which have certain advantages concerning chemical resistance and fire resistance.

Polyester resins combine low cost and ease of handling with good mechanical properties and
reasonable chemical resistance. Common applications include boats, storage tanks and roofings.

Epoxy resins are more expensive but have high stiffness and thermal stability as well as good
resistance to solvents and alkalis. Unmodified epoxy is fairly brittle but toughened epoxies have
been obtained by the addition of various toughening agents (e.g. rubber particles).

Thermoplastic resins soften upon re-heating. This simplifies manufacturing and recycling of
material, but the low stiffness makes them less suitable for structural applications.

Ceramic matrices and metal matrices are used in certain high temperature applications.

2.3 Fibre reinforced composites


The high stiffness to weight ratio of fibre reinforced plastics (FRP) makes them the most common
composites in aircraft. They are frequently classified with respect to the fibre material, e.g. glass
(GFRP), carbon (CFRP), Kevlar (KFRP) or boron (BFRP). In some contexts it is however more
practical to classify them with respect to the matrix material, e.g. Polymer Matrix Composite
(PMC), Metal Matrix Composite (MMC) and Ceramic Matrix Composite (CMC). When a
specification of the resin is needed it is common to write the fibre material followed by a slash and
the matrix, e.g. carbon/epoxy. A specific material is described by fibre/matrix quality, e.g. IM6/937
(intermediate modulus carbon fibres in a 937 epoxy resin).

Fibres may appear unidirectional (UD) or woven. Weaves are easier to handle in manufacturing, but
have lower stiffness and strength, especially in compression, due to the fibre undulation and stress
concentrations in the weave. Non-crimp fabrics have much better properties than traditional weaves,
as they essentially consist of straight fibre bundles (tows) sparsely stitched together.

Fibre composite laminates are obtained by stacking plies of different fibre orientation on top of each
other. The orientations may be tailored for expected load directions etc and are given with respect to
a global laminate x-y coordinate system, where the x-axis normally corresponds to the major load
direction. For each ply the properties must be specified in a local coordinate 1-2 system, where the
1-axis corresponds to the fibre direction in unidirectional plies.

Fig. 2.2 Fibre direction in a global coordinate system

The layup sequence of laminates with identical plies of different orientation is conventionally
specified within brackets where each ply direction is separated from the others by a slash. A

6
subscript is used to indicate the number of plies in a given direction. When the laminate is
symmetric with respect to the middle plane, only half of the layup is given, followed by a subscript
s. The subscript as is sometimes used to designate anti-symmetric layups. The notation is best
explained by a few examples:

[0/45/-45] means the layup 0/45/-45.


[02/45] means the layup 0/0/45/-45.
[0/45]s means the layup 0/45/-45|-45/45/0 (vertical line indicates symmetry plane).
[0/45]as means the layup 0/45/-45|45/-45/0 (vertical line indicates anti-symmetry plane).
[30/90]2 means the layup 30/-30/90/30/-30/90

Unidirectional laminate

Cross-ply laminate

+

Angle ply laminate
+

Symmetric laminate

Anti-symmetric laminate

Fig. 2.3 Overview of some layup types

7
2.4 Manufacturing techniques
There exist a large number of manufacturing methods for fibre composites which all have different
applications, depending on required product performance and geometry as well as cost, flexibility
and number of units to be produced. Manufacturing is done using two basic approaches: moulding
of material with pre-impregnated fibres or impregnation of dry fibres.

The use of pre-impregnated fibres allows a reliable positioning of the fibres and a low void content
in the matrix. The most common manufacturing method in high performance aerospace applications
is still autoclave moulding. This method involves manual or semi-automatic stacking of partially
cured sheets of pre-impregnated fibres (prepreg) followed by final curing under high temperature
and pressure in an autoclave. The drawbacks are high costs, extensive material waste and limited
abilities to obtain complex shapes. Filament winding is a related method primarily used for
cylindrical products, where pre-impregnated tows of fibres are wound on a rotating mandrel using
local heating for curing and a significant fibre tension to reduce void content.

Sheet Moulding Compounds (SMC) consist of pre-impregnated sheets of short fibres with random
in-plane orientation which are formed to selected shape under temperature and pressure.

Impregnation of fibres during manufacturing usually results in higher void content and significant
flow induced movement of the fibres. The advantages are better ability to form complex shapes and
reduced costs due to increased manufacturing speed and reduced material waste. Different fabric
concepts are frequently used to reduce the movement of fibres during manufacturing and to ease
handling prior to impregnation. The drawback is a larger fibre waviness and stress concentrations
caused by intersecting fibres. A similar product is the Chopped Strand Mat (CSM), which is made
from compacted short fibres with fairly low efficiency. The most common impregnation method in
low cost applications has been wet layup, where the resin is distributed manually over the fibres by
use of a roller. In recent years closed systems for Resin Infused Moulding (RIM) have been
introduced, which improves manufacturing quality and reduces health risks for workers.

8
3 Micromechanics of a ply

3.1 Volume and weight fractions


A two-component composite with volume Vc, mass Mc and density c, may be may be separated in
the corresponding quantities for matrix (subscript m) and filler/fibre (subscript f).

= +

Mc , Vc M m, V m Mf , Vf

Fig. 3.1 Mass and volume of a composite and its components

c = M c Vc m = M m Vm f = M f Vf (3.1)

We may now define fibre volume fraction Vf and matrix volume fraction Vm as follows:

Vf = V f Vc Vm = V m Vc (3.2)

Volume fractions are frequently used in calculations of the mechanical properties of the composite.
We are now able to calculate the density c of the composite

Mm + M f Vm m + V f f
c = = = Vm m + V f f (3.3)
Vc Vc

In manufacturing it is more practical to work with weight fractions wf and wm, which are obtained
from the volume fractions through the following relations:

wf = M f Mc = Vf f ( ) (Vc c ) = V f f c
(3.4)
wm = M m M c = (Vm m ) (Vc c ) = Vm m c

3.2 Models for elastic properties of a ply


Micromechanical models may be used to predict the elastic properties of the composite material
from known properties of the component materials. Such models have several applications:

Guide in development of new materials


Checking if measured or assumed properties are reasonable
Help in understanding of ply failure mechanisms

There are many sources of error in the predictions by such models:

The properties of fibres and matrix are usually not fully known
The material properties on micro-level may differ from those in larger specimens
The microscopic geometry contains irregularities

9
For example, it is easy to imagine the problems in measuring the stiffness of a fibre under axial
compression or radial tension. Curing or cooling of polymer matrices may locally cause a different
orientation of polymer chains in the presence of fibres. Finally, in practice fibres are non-uniformly
distributed, not perfectly aligned and slightly irregular in shape and size.

Here we will only use the simplest micromechanical models, where the two materials components
are modelled as parallel layers. Consider first tension along the fibres (1-axis), which causes an
equal strain 1 in the fibres and matrix.

L 2

1 1
1

L(1 + 1 ) B(1 + 2 )

Fig. 3.2 Model for longitudinal loading of ply

The average stress 1 is obtained by summing the load contributions from each material component
and dividing by the total cross sectional area, where the contribution from each component is given
by the product of its area, modulus and strain. The longitudinal modulus E1 is obtained by dividing
the average stress 1with the longitudinal strain 1:

1 F Am Em1 + A f E f 1 Am Af
E1 = = = = Em + Ef (3.5)
1 Ac1 Ac1 Ac Ac

For a composite with parallel fibres the ratio of cross sectional areas of fibres and matrix equals the
corresponding volume fractions. Thus:

E1 = 1 1 = Vm E m + V f E f (3.6)

This model is called the parallel model as the materials are assumed to act in parallel. It is also
possible to derive an expression for the transverse Poissons ratio 12. The average transverse strain
2 is obtained by summing the strain contributions from each component and dividing with the
specimen width:

Bm m + B f f
2 = = Vm m + V f f = Vm m 1 V f f 1 (3.7)
B

By definition the Poissons ratio is given by 12= 2/1. Thus:

12 = Vm m + V f f (3.8)

10
To calculate the transverse Youngs modulus we consider tension transverse to the fibres. In our
simple layer model this causes equal transverse stress 2 in the fibres and matrix.

2 2

B(1 + 2 )
1

Fig. 3.3 Model for transverse loading of ply

The average strain 2 is obtained by summing the extensions in each material component and
dividing by the total width, where the extension of each component is the product of its width and
the strain, given by the ratio between stress 2 and modulus.

Bm m + B f f Vm 2 Vf2
2 = = Vm m + V f f = + (3.9)
B Em Ef

The transverse modulus E2 is given by E2 = 2/2. Thus:

1 V Vf
= m + (3.10)
E2 Em E f

Similarly we may calculate the shear modulus in the 1-2-plane by considering shearing along the
fibres, which will result in an equal shear stress 12 in the fibres and matrix.

B12
2 12

12
B

1
12

Fig. 3.4 Model for shear loading of ply

The average shear strain is obtained by summing the shear contributions in each component and
dividing by the total width:

11
Bm m + B f f Vm 12 V f 12
12 = = Vm m + V f f = + (3.11)
B Gm Gf

The shear compliance 1/G12 is found by dividing the average shear strain with the shear stress 12:

1 V Vf
= m + (3.12)
G12 Gm G f

A comparison of predicted and measured properties for a composite with Ef=73 GPa, f=0.22,
Em=3.5 GPa and m=0.35, as studied by Greszcuk (1971) is given in Figure 3.5.

Predicted and measured properties


60
E1
50
Modulus [GPa]

40

30
E2
20

10
G12
0
0.4 0.5 0.6 0.7 0.8
vf
Fig. 3.5 Comparison of predicted and measured properties for glass-epoxy

The poor agreement for the transverse modulus E2 and shear modulus G12 is a result of the highly
non-uniform stress distribution in the transverse direction, which violates the assumed uniform
stress distribution in the model. In contrast, the strain in the longitudinal direction of the fibres is
highly uniform, which causes a good agreement between model and experiments for E1.

3.3 Models for strength of a ply

The limitations of simplified micromechanical models become even more apparent when we
attempt to predict ply strength, as failure is governed by peak stresses or fracture mechanics rather
than by average stresses or strains. The most reliable predictions are obtained for longitudinal
tensile failure, where the strain field is highly uniform and the failure mechanism is simple. For
composites with strong fibres, ultimate failure occurs at the fibre tensile strain utf. Thus:

1ut = E1 utf = V f E f utf + Vm E m utf = V f utf + Vm utf E m E f (3.13)

This expression may also be written

1ut = (V f + Vm E m E f ) utf V f utf (3.14)

12
In longitudinal compression, various failure mechanisms have been observed. The type of failure
depends on the relative flexibility of the fibre and matrix, as well as on the fibre strength in shear
and compression. For comparatively inflexible fibres with a low shear strength failure occurs as a
pure shear failure along the plane of maximum shear stress, which is inclined 45 to the load
direction. For strong flexible fibres failure typically occurs by buckling of the fibres, which leads to
flexural failure of the fibres, often termed fibre kinking.
2 2

1 1 1
1

1
1

Fig. 3.6 Compressive failure modes in the fibre direction

For the shear type failure the stress in the principal direction is obtained by the parallel model:

( )
1uc = 2 u 45 o = 2(Vm um + V f uf ) (3.15)

For the buckling type failure von Rosen (1965) showed that two failure modes are possible. For low
fibre volume fractions ( V f<0.2) fibre buckling occurs by tension and compression of the matrix
perpendicular to the fibres. For higher volume fractions, as seen in most commercial materials the
buckling is associated with shearing of the matrix. The solution for higher volume fractions ( V
f>0.2) may be approximated by the following expression:

1uc Gm Vm (3.16)

In practice fibre buckling is observed at significantly lower stresses, which has been attributed to
the fibre misalignment present in most real materials.

For transverse tensile/compressive failure the simplified serial layer models predict that the failure
stress equals the matrix failure stress, which is usually the weakest link. In practice the stress state
between fibres is highly three-dimensional with significant stress concentrations. The presence of
material flaws suggests that a fracture mechanics based strain energy criterion may be more suitable
for prediction of failure. Figure 3.7 shows possible initiation sites in transverse tension.

Max von Mises stress (>2)

2ut 2ut

1
Max dilatational strain energy

Fig. 3.7 Possible failure sites in transverse tension

In spite of this the observed transverse failure stress and shear failure stress are often of the same
order as the matrix failure stress.

13
4 Macromechanics of a ply
4.1 Stress and strain vectors

u z
dz
z z
zz z dz
u y
z
zy
zx u z
yz dz dy
y
xz
yy
xy yx dy u y
xx y uz dy
uy y

y
x
Fig. 4.1 Definition of all stresses and example of strains in the y-z plane

Figure 4.1 shows a three-dimensional stress state in a Cartesian system, and an example of
displacements used to define strains in the y-z plane. The first index of the stress component
indicates the normal direction of the plane where the stress is acting, while the second index
indicates the direction of the stress. Tensorial strains e and engineering strains are defined by:

u u j u i u i u j
eij = 12 i + ii = ij = ij = + (4.1)
x j xi xi x j xi

For convenience, double-indices (e.g. zz and xx) are usually replaced by a single index (e.g. z and
xx). A three-dimensional stress state is given by the stress vector and engineering strain vector :

= [ 1 2 3 23 31 12 ] T (4.2)
= [ 1 2 3 23 31 12 ] T
(4.3)

where the shear components are given in order of their surface normal vectors. In contracted
notation the elements are often labelled i and i where i=1,2..6. To simplify coordinate
transformations we use the tensorial strain vector e, which obeys the same coordinate
transformation rules as the stress vector.

e = [ 1 2 3 23 31 12 ]
1 1 1 T
2 2 2 (4.4)

Transformation between tensorial strains and engineering strains is obtained by Reuters matrix R

1 0 0
I 0
= Re where R = and I = 0 1 0 (4.5)
0 2 I 0 0 1

14
The generalised Hookes law for an elastic material may be written

= S or = C (4.6)

where the stiffness matrix C is the inverse of the compliance matrix S. Both matrices contain 6x6 =
36 elements, but symmetry reduces the maximum number of unknown material constants to 21.

4.2 Material classes


Anisotropic materials (21 independent constants) have no axes of symmetry and have the following
compliance matrix

S11 S12 S13 S14 S15 S16


S S 22 S 23 S 24 S 25 S 26
12
S S 23 S33 S34 S35 S36
S = 13 (4.7)
S14 S 24 S34 S 44 S 45 S 46
S15 S 25 S35 S 45 S55 S56

S16 S 26 S36 S 46 S56 S66

Orthotropic materials (9 independent constants) have three orthogonal planes of symmetry. For
coordinate axes along the axes of symmetry we obtain the following compliance matrix

S11 S12 S13 0 0 0


S S 22 S 23 0 0 0
12
S S 23 S33 0 0 0
S = 13 (4.8)
0 0 0 S 44 0 0
0 0 0 0 S55 0

0 0 0 0 0 S66

Transversely isotropic (quasi-isotropic) materials (5 independent constants) have one axis of


symmetry. If the 1-axis is the axis of symmetry we obtain the following compliance matrix

S11 S12 S12 0 0 0


S S 22 S 23 0 0 0
12
S S 23 S 22 0 0 0
S = 12 (4.9)
0 0 0 2(S 22 S 23 ) 0 0
0 0 0 0 S55 0

0 0 0 0 0 S55

15
Isotropic materials (2 independent constants) have identical properties in all directions and the
following compliance matrix

S11 S12 S12 0 0 0


S S11 S12 0 0 0
12
S S12 S11 0 0 0
S = 12 (4.10)
0 0 0 2(S11 S12 ) 0 0
0 0 0 0 2(S11 S12 ) 0

0 0 0 0 0 2(S11 S12 )

The compliance matrix of an orthotropic material expressed in engineering properties becomes

1 E1 21 E 2 31 E3 0 0 0
E 1 E2 32 E3 0 0 0
12 1
E 23 E 2 1 E3 0 0 0
S = 13 1 (4.11)
0 0 0 1 G23 0 0
0 0 0 0 1 G31 0

0 0 0 0 0 1 G12

where the strain in any direction i is obtained by dividing the stress i with the modulus Ei . The
resulting (negative) strain in any perpendicular direction j is obtained by multiplying with the
Poissons ratio ij (remember action from i to j). Symmetry of the compliance matrix implies

jk E j = kj E k
(4.12)

4.3 Orthotropic ply under plane stress


In many engineering structures, e.g. plates, beams and shells, the out-of-plane stresses are usually
small in comparison to the in-plane stresses. Classical plate theory assumes zero out-of-plane
stresses ( 3 = 13 = 23 = 0 ), i.e. plane stress. With a local coordinate system along the material
axes, and considering the symmetry condition in Eq. (4.12), the strain-stress relation is given by

1 1 E1 12 E1 0 1
= E 1 E2 0 2 (4.13)
2 12 1
12 0 0 1 G12 12

The stress-strain relation is obtained by inversion of the compliance matrix:

1 E1 12 E 2 0 1 Q11 Q12 0 1
= 1 E E2 0 = Q Q22 0 2
2 1 12 2 2 12
12 12 21
0 0 G12 (1 12 21 ) 12 0 0 Q66 12
(4.14)
or briefly l = Q l

16
where Q is the reduced stiffness matrix of the ply in plane stress. A laminate is composed of several
plies of different orientation.

y
2
1

Fig. 4.2 Global coordinate system and local coordinate system of ply

The transformation of stresses from a global coordinate system to stresses l in the local material
system is obtained from equilibrium for the grey triangles in Fig. 4.2, which yields

1 = x cos 2 + y sin 2 + xy 2 sin cos


2 = x sin 2 + y cos 2 xy 2 sin cos (4.15)
(
12 = ( x + y )sin cos + xy cos 2 sin 2 )
This may be expressed as l = T, where the transformation matrix T is given by:

c2 s2 2 sc c 2 s2 2 sc
2 1 c = cos
T = s c 2
2 sc T = s 2 c2 2 sc where
s = sin
sc sc c 2 s 2 sc sc c 2 s 2
(4.16)

Note that T 1 may be obtained by performing a negative rotation, i.e. by substituting for + in T.
These transformations also apply for strains provided tensorial strain is used. Thus

1 0 0
l = T and l = R el = RTe = RTR 1
where R = 0 1 0 (4.17)
0 0 2
(R is called Reuters matrix)
1 1 T T 1 T
Straightforward algebra shows that RTR = [T ] = [T ] T , where the second equality sign
follows from basic matrix theory and the common symbol T T has been introduced for brevity.

Thus, in the global system the stress-strain relation for the ply is given by

= T 1 l = T 1Q l = T 1QRTR 1 = T 1QT T = Q (4.18)

The quantity Q is called the transformed reduced stiffness matrix of the ply:

Q11 Q12 Q16



Q = Q12 Q22 Q26 (4.19)
Q16 Q26 Q66

17
where the elements of the transformed reduced stiffness matrix are given by:

Q11 = Q11c 4 + 2(Q12 + 2Q66 )c 2 s 2 + Q22 s 4


Q22 = Q11s 4 + 2(Q12 + 2Q66 )s 2 c 2 + Q22 c 4
(
Q12 = (Q11 + Q22 4Q66 )s 2 c 2 + Q12 s 4 + c 4 )
c = cos (4.20)
Q16 = (Q11 Q12 2Q66 )sc 3 (Q22 Q12 2Q66 )cs 3
s = sin
Q26 = (Q11 Q12 2Q66 )cs 3 (Q22 Q12 2Q66 )sc 3
(
Q66 = (Q11 + Q22 2Q12 2Q66 )s 2 c 2 + Q66 s 4 + c 4 )
The stress-strain relations in the local and global coordinate system may be summarised by the
following transformation diagram:

T -1
l

Q Q =T -1QT -T

R T R-1
l el e

T -T

Fig. 4.3 Transformation diagram for stresses and strains in local and global coordinates

18
4.4 Effect of hygrothermal strains
A temperature change T of an unloaded ply will cause thermal strains, which for the local material
coordinate system of an orthotropic ply are given by

Tl = [ T 1 T 2 T 12 ] T = [1 2 0] T T (4.20)

where 1 and 2 are the coefficients of thermal expansion in the principal directions 1 and 2.

Similarly a humidity change H of an unloaded ply will cause hygroscopic strains, which for the
local material coordinate system of an orthotropic ply are given by

Hl = [ H 1 H 2 H 12 ] T = [1 2 0] T H (4.21)

where 1 and 2 are the coefficients of thermal expansion in the principal directions 1 and 2.

The thermal strains T and hygroscopic strains H in the global coordinate system are easily found
by considering the previous transformation diagram:

T = (T T ) Tl = T T Tl = T T [ 1 2 0] T T = [ x y xy ] T T
1

(4.22)
H = (T )
T 1
Hl = T Hl = T [1 2 0] H = [ x
T T T
y xy ] H
T

where T and H are the changes in temperature and humidity and x, y, xy and x, y, xy are the
transformed coefficients of thermal and hygroscopic expansion in the global coordinate system. The
total strain on the ply tot is the sum of the strains due to applied loads , the thermal strains T and
the hygroscopic strains H:

tot = + T + H (4.23)

To calculate the applied stress on the ply, the hygrothermal strains must be subtracted from the total
strain:

= Q = Q ( tot T H ) (4.24)

Evidently prevention of thermal expansion and hygroscopic swelling (i.e. tot = 0) for T, H > 0
will result in compressive stresses on the ply.

19
5 Macromechanics of a laminate
5.1 Forces and moments
z dy
=
dx My y

x Myx Ny
Mx Nxy Nyx
Mxy
Nx

Fig. 5.1 Forces and moments acting on a plate

Figure 5.1 defines the forces and moments acting on a plate. Note that the twisting moments are
defined opposite to the common right hand screw rule. The forces and moments may be collected in
a membrane force vector N and moment vector M

[
N = Nx Ny N xy ]
T
M = Mx [ My M xy ] T
(5.1)

The forces and moments are also called stress resultants as they may be obtained from the
integrated action of stresses:

N = dz M = z dz where = [ x y xy ]T (5.2)

5.2 Strains and curvatures


The deformation of the laminate middle plane may be expressed by the membrane strain vector 0
and curvature vector

0 = [ 0 x 0 y 0 xy ]T = [ x y xy ]T (5.3)

The so called Kirchhoff assumption is fundamental in classical plate theory. It states that planes
which initially are normal to the middle plane of the plate remain plane and normal to the middle
plane after deformation. This implies that out-of-plane shear deformations are zero, i.e. infinite
shear stiffness. Neglect of shear deformations is justified when out-of-plane strains are small in
comparison to the in-plane strains caused by moments and membrane forces. With the assumption
of plane sections the strains vary linearly through the thickness, and may be obtained from

= 0 + z (5.4)

5.3 Force-deformation relations


By combining the previous expression with the stress-strain relation the stress resultants become

N = dz = Q dz = Q ( 0 + z ) dz (
M = z dz = Q z dz = Q z 0 + z 2 dz ) (5.5)

By collecting the stress resultants in one single 6x6 matrix we obtain the expression:

20
N Q Q z 0 A B 0
M = 2
dz = (5.6)
Q z Q z B D

where Q is the transformed stiffness matrix, which may differ for each ply through the thickness.
The A, B, D are 3x3 matrices given by the following relations:

A = Q dz = Qk ( z k z k 1 ) = Qk t k
k k

B = Q zdz = Qk 12 (
z k2 z k21 )= Q t z k k k
(5.7)
k k

D = Q z dz =
2
Qk 13 (
z k3 z k31) = Q t (z k k
2
k + 12
1 t2
k )
k k

where zk+1= z k +tk/2, zk-1= z k -tk/2 and Qk , tk and z k are the transformed reduced stiffness matrix,
ply thickness and middle plane coordinate of ply k. The last term in the D matrix may be neglected
when the number of plies is large, i.e. when tk2/12<<zk2 for most plies.

Ply k

zk-1 zk zk
x

Fig. 5.2 Thickness coordinates of a ply

When all the elements in the load-deformation matrix are written we obtain the expression

N x A11 A12 A16 B11 B12 B16 0 x


N
y A12 A22 A26 B12 B22 B26 0 y
N xy A16 A26 A66 B16 B26 B66 0 xy
= (5.8)
M x B11 B12 B16 D11 D12 D16 x
M y B12 B22 B26 D12 D22 D26 y

M xy B16 B26 B66 D16 D26 D66 xy

The elements Xij and Xji (X=A,B,D) have been written with a common index to stress the symmetry
of the stiffness matrix. Note that the B matrix imposes a usually undesirable coupling between
curvature and in-plane loading, as well as between membrane deformation and bending loads.
Similarly A16 and A26 impose a coupling between in-plane shearing and stretching while D16 and D26
impose a coupling between bending and twisting. This is usually undesirable, but may be used in
some applications, e.g. for aeroelastic tailoring. Figure 5.3 illustrates various coupling phenomena.

21
Stretching-bending coupling (Bij0)

Stretching-shearing coupling (A16 and A260)

Bending-twisting coupling (D16 and D260)

Fig. 5.3 Examples of coupling of deformations

5.4 Some layup types and their effect on coupling


Unidirectional laminates contain only 0 plies or only 90 plies.
Cross-ply laminates contain 0 and 90 plies.
Angle-ply laminates contain plies in other directions, i.e. + or .

Symmetric laminates have plies placed symmetrically with respect to the middle plane, for example
30/30/30/30. Such laminates have B=0, since each zQ term is cancelled by a zQ term.

Balanced laminates contain equal numbers of + and plies. Such laminates lack stretching/shear
coupling (A16=A26=0) since each Qi 6 ( ) term is cancelled by a Qi 6 ( ) term. Unidirectional and
cross-ply laminates are always balanced, since each half of any +90 plies is balanced by an equal
amount of 90 plies.

Specially orthotropic laminates are defined as lacking coupling between bending/stretching (Bij=0),
stretching/shear (A16=A26=0) and bending/twisting (D16=D26=0). The need for D16=D26=0 requires
that Qk z k2 terms cancel, which implies an anti-symmetric layup. The need for Bij=0 requires that
Qk z k terms cancel. This condition is fulfilled by anti-symmetric layups where the sublaminates on
each side of the middle plane are individually balanced. These sublaminates may then be considered
as two orthotropic plies with equal membrane properties Q , so that the terms Qk z k cancel.

In design estimations many laminates are approximated as specially orthotropic, as the relative
influence the coupling terms decreases fairly rapidly with the number of plies. It should, however,
be stressed that such simplifications can be disastrous for design against buckling, as this will lead
to an overestimation of the buckling load.

22
5.5 Estimates of laminate stiffness
Calculation of the ABD stiffness matrix is tedious and is best done by computers with laminate
theory software. It is, however, important to be able to estimate the elements of the stiffness matrix
in preliminary design, for checking computer outputs, or when a computer with suitable software is
not available. In most fibre composites Q11E1 as E1>> E2, which implies

( )
Q11 0 o E11 ( )
Q11 90o E 22 ( )
Q11 ( 45) E11 cos 4 45o = E11 4 (5.9)


The average reduced stiffness terms Q11 and Q22 of a laminate with n0 0 plies, n90 90 plies, n45
45 plies and with a total of ntot plies, are then estimated from:


Q11 (n0 E11 + n90 E 22 + n 45 E11 4 ) ntot
Q22 (n90 E11 + n0 E 22 + n 45 E11 4 ) ntot (5.10)

The corresponding laminate stiffness terms Aii and Dii are estimated from:

Aii Qii h (
Dii Qii k z k2 t ply Qii h 3 12 for large numbers of plies ) (5.11)

Equations (5.10)-(5.11) are only suitable crude estimation of symmetric balanced laminates.

5.6 Apparent engineering properties of laminates


The apparent engineering properties of orthotropic laminates are obtained from the relation between
average stresses and applied strains. The apparent membrane stresses m and peak bending stresses
b may be calculated from the stress resultants as follows:

mij = N ij h bij = 6 M ij h 2 (5.12)

The strain-stress relation is obtained by inverting the laminate stiffness matrix:

1
A B N a b N ah bh 2 6 m S mm S mb m
= B D M = b d M = 2 = S bb b
(5.13)
bh dh 6 b S bm

The elements of the compliance matrix Smm are the inverted values of the corresponding membrane
moduli, see Eq. (4.11). Thus the engineering membrane moduli and Poissons ratios are obtained
from the terms 1/(haij), where a is the upper left quarter of the inverted ABD-matrix. If B=0 and
A16=A26, as for a symmetric balanced laminate, the submatrix aij i,j=1 or 2 is obtained by direct
inversion of the submatrix Aij i,j=1 or 2. This allows us to express the engineering properties
directly in the elements A11, A22 and A12. Note that 1-A122/(A11A22) for the laminate then corresponds
to the factor 1 12 21 in Eq. (4.14) for a homogeneous orthotropic ply.

Engineering property General laminate Symmetric balanced laminate


xy a12 a11 A12 A22
yx a12 a 22 A12 A11
Ex 1 (ha11 ) [1 A 2
( A22 A11 )]A11 h
[1 A ( A22 A11 )]A22
12
Ey 1 (ha22 ) 2
12 h
G xy 1 (ha66 ) A66 h

23
6 Failure criteria and design
6.1 Peculiarities in failure of composites
Conventional materials, such as metals, are more or less isotropic and have a strength which is
fairly independent of direction. Metals usually fail by shear yielding along crystal slip planes, which
results in a similar strength in tension and compression. Furthermore, in most applications these
materials are fairly independent of strain rate and temperature. Note, however, that cold rolling of
metals may introduce significant strength anisotropy.

Polymers are often fairly isotropic with similar strength in all directions. The microscopic failure
phenomena differ from metals, which causes differences between the tensile and compressive
strength. Furthermore, the failure of polymers is strongly dependent on strain rate, temperature and
hydrostatic pressure.

Composites are usually highly anisotropic, and as a result the strength is strongly dependent on
direction. Differences in the failure phenomena cause significant differences between tensile and
compressive strength. The rate and temperature sensitivity of fibres and matrix are also reflected in
the failure of the composite.

Failure criteria for conventional isotropic materials have traditionally been based on stress
calculations. Many failure criteria for composites are based on extension of stress based criteria for
isotropic materials. Such approaches are convenient for use in existing FE-codes, where stress is a
standard output. However, plotting of stresses is very inconvenient since stresses and strengths in
multidirectional laminates fluctuate violently through the thickness, Fig 6.1a. In contrast strains are
continuos through the thickness for laminates without cracks, Fig. 6.1b. Furthermore, failure strains
are not strongly dependent on fibre direction. For these reasons strain based failure criteria are more
suitable for composite materials.

a) Stresses and failure stresses b) Strains and failure strains

Fig. 6.1 Stress and strain variation through the thickness of a laminate

In practice, strength of fibre composites is tested by loading in the principal material directions
parallel and perpendicular to the fibres. For these reasons, failure criteria must be evaluated in the
ply coordinate system, which requires transformation of stresses and strains from the global
coordinate system, Figs 4.2-4.3.

It should be noted that the tensile and compressive strength of fibre composites usually differ (see
Sect. 3.3), and that compressive testing is associated with significant problems to avoid buckling
and stress concentrations. Furthermore, stress concentrations or superimposed normal stresses
significantly reduce the strength measured in most shear tests, Fig. 6.2.

24
Stress
concentration

Fig. 6.2 Sources for premature failure in shear tests

Finally it should be borne in mind that ply properties are usually measured on unidirectional
specimens several millimetres thick. The strength of thin plies in laminates may be significantly
higher as shown in Fig. 6.3. This effect is at least partly explained by Weibulls statistical strength
theory, which predicts an increase in strength with decreasing material volume and fewer
microscopic defects.

Transverse strength vs specimen area


Data from O'Brien & Salpekar (1993)
100

Bending test
80
Strength [MPa]

60

40

20 Typical coupons

0
0 5 10 15 20 25 30
Specimen area [cm2]

Fig. 6.3 Influence of size in determination of transverse ply strength

6.2 Ply failure criteria


As mentioned, ply failure criteria are normally evaluated in the local coordinate system of the ply,
where the axes coincide with the principal material axes of the ply. The simplest, and perhaps most
commonly used, failure criteria for composites are the maximum strain and maximum stress
criteria. These criteria state that failure occurs when the ply reaches any of the failure strains or
stresses obtained from uniaxial tests:

Maximum strain : Maximum stress :


1 < 1uc or 1 > 1ut 1 < 1uc or 1 > 1ut
(6.1)
Failure when 2 < 2uc or 2 > 2ut 2 < 2uc or 2 > 2ut
12 > 12u 12 > 12u

where 1uc and 1ut are the compressive and tensile failure strains in the i-direction, and 12u is the
corresponding shear failure strain. Stresses are labelled identically. Typical values for carbon-epoxy
composites are:

25
1uc = 1.0 1ut = 1.3 2uc = 3.0 2ut = 0.6 12u = 1.5 [%]
1uc = 1500 1ut = 2000 2uc = 250 2ut = 50 12u = 110 [MPa]

Plots of the maximum strain and stress criteria in the local strain and stress coordinates of the ply
are shown in Fig. 6.4, although the additional shear failure criterion is not illustrated. This figure
shows why these criteria often are called box criteria, and also illustrates why stress criteria are
inconvenient for graphical presentation.

2 2
2ut

2ut
1

1 1uc
1ut
1uc
1ut 2uc

2uc

Fig. 6.4 Maximum strain and maximum stress criteria

The major drawback of the maximum strain and maximum stress criteria is that they do not
consider interaction between different stresses, which results in a discontinuous failure envelope.
The most common interactive failure criterion for isotropic materials is the von Mises criterion.
This criterion assumes that failure is governed by the deviation from a uniform (hydrostatic) stress
state, i.e. that failure is associated with shear stresses. Calibrated for failure in uniaxial tension and
expressed in the principal stress space I-II-III the criterion becomes:

1
2
[( I II )2 + ( I III )2 + ( II III )2 Y2 = 1 ] (6.2)

For plane stress (III=0) the criterion simplifies to:

[ I
2
]
+ II 2 I II Y2 = 1 (6.3)

This is the equation for an ellipse in the I-II coordinate system, which is shown in Fig. 6.5.

Uniform tension

Pure shear

Fig. 6.5 von Mises failure criterion for isotropic material in principal stress coordinates

The ellipse crosses the I- and II-axes at I = Y or II = Y (uniaxial stress). Its major axis
coincides with II = I (uniform stress) while the minor axis coincides with II = I (pure shear).

26
By using the definitions of principal stresses Eq. (6.3) may also be expressed in an arbitrary 1-2
coordinate system:

[ 2
1 + 22 1 2 + 3 12
2
]
Y2 = 1 (6.4)

By solving for the failure stresses in pure shear or pure tension/compression we obtain u =Y, u=
Y 3 . Thus, Eq. (6.4) may be rewritten on the following form:

12 u2 + 22 u2 1 2 u2 + 12
2
u2 = 1 (6.5)

A generalisation of this criterion to orthotropic materials was suggested by Hill (1950). Azzi and
Tsai (1965) proposed the use of Hills criterion for composite materials. By assuming plane stress
and transversely isotropic ply properties they obtained the following special case of Hills criterion:

12 12u + 22 22u 1 2 12u + 12


2
12
2
u =1 (6.6)

This criterion has, somewhat incorrectly, become known as the Tsai-Hill criterion. The original
expression by Hill did not consider different properties in compression and tension. For composites
is has been suggested that tensile and compressive loads may be considered in the criterion by using
the corresponding strengths. Figure 6.6 plots an example of the criterion evaluated with equal and
different strength values in tension and compression. Note that the failure envelope using different
compressive and tensile strengths (dashed) is composed of four quarter ellipses.

2
2ut

1
1uc 1ut

2uc

Fig. 6.6 Tsai-Hills criterion in the ply stress coordinates

To consider the local stresses in an off-axis ply under a uniaxial global stress we consider Eq. (4.15)
for x0, y=xy=0. The global failure stress xu may then be obtained by assuming that the failure
stress has been reached in the ply coordinate system. Thus, for uniaxial stress the maximum stress
criterion may be written:

x 1u cos 2 x 2u sin 2 x 12u (sin cos ) (6.7)

where is the off-axis angle of the ply. Under the same assumptions the Tsai-Hill criterion becomes

(
cos 4 12u + cos 2 sin 2 1 12
2 2 4
)
u 1 1u + sin 2u = 1 xu
2 2
(6.8)

Figure 6.7 gives a comparison of predictions by the maximum stress criterion and the Tsai-Hill
criterion for a glass/epoxy.

27
Glass/Epoxy, data by Tsai (1968)
1200
Maximum stress
1000
Tsai-Hill
Strength [MPa]
800

600

400
Compression
200
Tension
0
0 10 20 30 40 50 60 70 80 90
Off-axis angle

Fig. 6.7 Comparison of predictions and test data for an off-axis test

The major advantage of the Tsai-Hill criterion is a smooth failure envelope defined by a single
equation. Furthermore, it usually provides a better fit to experimental data than the simple box
criteria, although this is not true for all material systems and load cases. In contrast to the maximum
stress and strain criteria the Tsai-Hill criterion and most other interactive criteria do not predict the
failure mode (i.e. fibre or matrix failure), which is a major disadvantage in failure simulations.

There is a wealth of other failure criteria suggested for composites. More complex criteria usually
provide a better fit to experimental data, at the expense of more extensive tests, which are
sometimes not even feasible. One more general interactive criterion is the Tsai-Wu tensorial
criterion, which for an orthotropic ply under plane stress simplifies to:

F1 1 + F2 2 + F6 6 + F11 12 + F22 22 + F66 62 + 2 F12 1 2 = 1 (6.9)

The major problem with this criterion is the coupling term F12, which is normally assumed or
determined from complicated biaxial tests. However, DeTeresa and Larsen (2003) recently claimed
that it could be estimated with good accuracy from the term F11.

6.3 Effect of ply failure in laminates


Matrix failure and fibre failure have quite different effects on the ply properties. Matrix failure in
the fibre direction has a marginal influence on the longitudinal modulus E1, since the stiffness of
fibres usually is much larger than the stiffness of the matrix. Fibre failure, on the other hand, leads
to a complete failure of the ply, i.e. E1=0.

For unidirectional laminates matrix failure transverse to the fibres leads to a similar complete failure
E2= G12=0, and uncouples axial and longitudinal strain, i.e. 12=0. For multidirectional laminates,
stresses are redistributed to surrounding undamaged plies, and matrix cracks saturate at a crack
spacing in the order of one ply thickness. Stresses and strains in the cracked plies gradually increase
with growing distance to the matrix cracks, due to the shear lag effect discussed in Sect. 1.3. As a
result, cracked plies in multidirectional laminates maintain a reduced average stiffness, Fig. 6.8.

28
E2o
E2d

Fig. 6.8 Local modulus variation in cracked ply in multidirectional laminate

The properties of failed plies in laminates may be summarised as follows:

Fibre failure E1d = 0


Matrix failure E 2 d = E E 2o G12d = G G12o 12d = 12o
(6.10)
i = 0 for unidirectional laminates
0 i 1 for multidirectional laminates

where indices o and d refer to the properties before and after failure.

Multidirectional laminates fail sequentially as a result of the different failure strains for different ply
directions. Typically, plies at 90 to the principal load fail first (first ply failure), followed by off-
axis plies, and finally by failure in the 0 plies (last ply failure). An example of a failure sequence
for uniaxial tensile loading of a quasi-isotropic layup is outlined in Fig. 6.9. Note, however, that
certain load cases or materials may result in a different failure sequence.

Last ply failure


(0 plies fail)

45 plies fail
Stress

First ply failure


(90 plies fail)

Strain
Fig. 6.9 Example of failure sequence in quasi-isotropic layup

To simulate the sequential failure of laminates a stepwise analysis scheme may be used, Fig. 6.10.
Such schemes usually rely on the maximum stress or maximum strain criterion, since the ply failure
mode (i.e. matrix failure or fibre failure) must be known in order to perform the necessary
degradation of failed plies.

29
Start

Increase load

Yes
Matrix failure?

Modify E2, G12, 12 in failed plies

Yes
Fibre failure?

Modify E1 in failed plies

No
All plies failed?

Stop

Fig. 6.10 Scheme for simulation of sequential failure in laminates

6.4 Design methods


Fibre composites provide possibilities for a highly efficient design. However, in order to perform a
rational design, several questions must be answered:

Design for optimum stiffness or optimum strength?


Single or multiple load cases?
Do we have to consider fatigue?
Are there any hygrothermal loads?
Are matrix cracks allowed?
What are the cost constraints?
What materials and manufacturing methods are available?

The choice of optimum stiffness (which is common in aircraft) or optimum strength is crucial for all
design calculations and may affect the choice of material. Single load cases (which may still involve
a single multiaxial loading) allow efficient designs and fairly simple calculations, while multiple
load cases require comparison of a large number of calculations.

Hygrothermal loads may introduce significant additional stresses, but even when such loads are
absent we often need to consider the residual thermal stresses caused by cooling during
manufacturing. Fatigue of composite laminates is an important topic which is beyond the scope of
this presentation. Matrix cracks usually occur at much lower loads than fibre failure, and do not
severely affect the static strength and stiffness of the laminate. Prevention of matrix cracks is a
severe limitation but is often advisable, since matrix cracks promote moisture absorption and
delamination. This may cause further crack growth under cyclic hygrothermal or mechanical loads.

30
In design there is a lot to gain by using repeated sublaminates. Sublaminates consist of a few plies
which form a part of a thicker laminate. The properties of the entire laminate may be obtained from
laminate theory by treating the sublaminates as super plies. For a symmetric laminate with 2r
sublaminates, each with thickness t and a stiffness matrix Aijo , Bijo , Dijo , the total stiffness matrix of
the laminate is given by:

Aij = 2rAijo Bij = 0 h = 2rt


[ ]
(6.11)
Dij = 2r Dijo + (r 1) tBijo + (r 1)(2r 1)t 2 Aijo 6

It is convenient to define a normalised bending stiffness Qb and membrane stiffness Qm [GPa]:

Qijb = 12 Dij h 3 Qijm = Aij h (6.12)

The ratio of bending stiffness to membrane stiffness for the laminate is now given by:

Qijb Qijm = 12 Dij (h Aij ) = [r


2 2
Qijbo Qijmo + r 2 (r 1)3Bijo (Aijo t )+ r 2
(r 1)(r 1 2)] (6.13)

It is evident that the laminate bending stiffness and membrane stiffness will be equal when the
number of sublaminates r is large. The influence of differences between bending and membrane
stiffness of the sublaminate decays rapidly (~1/r2) while the influence of sublaminate bending-
stretching coupling decays somewhat slower (~1/r).

Design for single load cases may be done by using the principal stress approach suggested by Tsai:

1) Determine direction and magnitude of the principal stresses


2) Align fibres in the principal stress directions in proportion to the magnitude of stresses
3) Add some plies in 45 angle to the principal stresses to account for unexpected loads

Note that the fibre angles are defined along the principal load axes, rather than by the length and
width of the laminates. Design for multiple load cases and for layups with plies limited to 0,90 and
45 is more complex, but may be done by using the ranking approach suggested by Tsai:

1) Use software to rank a selection of sublaminates with different amounts of 0,90 and 45
plies for each load case
2) Select the sublaminate (or sublaminate combination) with the best average performance
3) Select the necessary amount of sublaminates to carry the required loads

Multiple complex load cases, which are common in aircraft design, often tend to require relatively
quasi-isotropic layups, while single uniaxial loads may allow highly orthotropic layups. Once an
optimised layup has been obtained, there are several rules considered to be good design practice:

Include a minimum fraction of 90 plies to avoid failure by unexpected transverse loads


Do not use 0 plies on unprotected surfaces to limit effect of surface scratches
Use repeated sublaminates rather than thick plies, which are more susceptible to cracking
Use symmetric (and balanced) laminates to reduce buckling problems

where, as before, the directions 0 and 90 refer to the principal load directions.

31
7 Introduction to advanced topics
7.1 Effect of stiffness gradients in laminates
Neglect of stress concentrations at abrupt stiffness changes frequently causes unexpected failure in
structures. Such stiffness changes cause stress gradients, which induce additional stresses required
to maintain local equilibrium. Common causes for stress concentrations are cracks, holes, material
discontinuities, impact damage and thickness changes or ply drops. Fibre composite laminates
contain material discontinuities both at a microscopic level (fibre-matrix) and at a ply level (ply-
ply), and often contain macroscopic cracks.

7.2 Notched strength of composite laminates


The stress fields close to a crack tip and a hole in an elastic (brittle) material may be obtained from
elasticity theory, and are shown schematically in Fig. 7.1. Theoretically, the stress at a crack tip is
infinite, while the stress at the edge of the hole in an infinite plate is given by the stress
concentration factor KT, which equals 3 for an isotropic sheet. If the plate has a finite width the loss
of area in the notched section causes an average stress higher than the far field stress 0. This is
accounted for by use of a finite width correction factor Y 1.

Stress, x(0,y) Stress, x(0,y) Hole


Crack
0 YHKTU -
0
Brittle material
KTU - Brittle material
2b
Finite width U -
x y Finite width y
x

0 0
Infinite width
Infinite width
0 0

0 rp Distance, y -1 4 0 rp Distance, y -1

Fig. 7.1 Stress distributions at a crack tip and hole


For a completely brittle material the strength would be reduced to zero in a cracked sheet, and to
1/KT in a sheet with a hole. In practice materials allow some yielding after reaching the failure stress
U. The onset of local failure results in a stress redistribution close to the notch, with the peak stress
limited by the failure stress of the material and stresses further ahead of the notch slightly higher
than predicted by elastic (brittle) theory, Fig. 7.1. For a completely ductile (unlimited yielding) the
only effect of the notch is the loss of net section area, which for a finite width plate results in a
linear variation between notch size and strength, Fig. 7.2. The notched strength of real materials
depends on the notch size and falls between the strength of completely ductile and completely
brittle materials.

Fracture mechanics and notch failure criteria have been developed to predict the failure at cracks
and holes. A fundamental assumption linear elastic fracture mechanics and elementary notch failure
criteria is that the damage zone radius rp = rp b << 1 , which allows the assumption that stress field
outside the process zone remains unaffected by the local inelasticity (small scale yielding). This
condition is usually satisfied for cracks, but often violated for holes.

32
Strength Strength
Crack Hole
U 0
U
0
Completely 2b
Completely
ductile 2b ductile
x y
Real
x y
materia 0

Real material 0 Infinite width


U/3 Completely brittle

Finite width
Completely brittle

Cracked fraction, y/B 1 Holed fraction, y/B 1

Fig. 7.2 Strength versus defect size for plates with a crack and a hole

For a crack in an elastic isotropic plate the stress distribution ahead of the crack tip is given by:

x (0, y ) = Y 0 2( y 1) where y = y b (7.1)

For a circular hole in an elastic orthotropic plate the stress distribution ahead of the hole is given by:

x (0, y ) = Y
0
2
[2 + 1 y 2
(
+ 3 y 4 (K T 3) 5 y 6 7 y 8 )] (7.2)

The stress concentration factor KT for a circular hole in an orthotropic plate is given by:

K T = 1 + 2 E x E y + E x G xy 2v xy (7.3)

(
For quasi-isotropic plates KT simplifies to K T = 3 since E x = E y = 2 1 + v xy G xy . The finite width )
correction factor for a crack of length 2b in an isotropic plate of width 2B is approximated by:

YS = 1 cos( b (2 B )) (7.4)

The finite width correction factor for a hole width 2b in an isotropic plate of width 2B is
approximated by:

[ ]
YH = 2 + (1 b B )3 [3(1 b B )] (7.5)

These functions have been illustrated in Fig. 7.3. It has been found that the finite width correction
factors are only weakly dependent on orthotropy.

To address the effect of size effect on the failure at holes Whitney & Nuismer (1974) suggested two
failure criteria. The simplest one is called the Point Stress Criterion and states that failure occurs
when the elastic stress at a characteristic distance d0 reaches the failure stress U of the material. A
more sophisticated criterion is the Average Stress Criterion which assumes that failure occurs
when the average stress over a distance a0 equals the failure stress U of the material. The criteria
are illustrated in Fig. 7.4.

33
1.8

Finite width correction Y


1.7
1.6
1.5
Hole
1.4
1.3
1.2
1.1
Slit
1
-0.2 1E-15 0.2 0.4 0.6
Defect size/Width, b/B

Fig. 7.3 Finite width correction factors for a slit (crack) and a hole

x
x
K T 0 Peak elastic stress

0 U Fracture stress

0 Applied stress
2a

b+d0 b+a0
y
2b

Fig. 7.4 Finite width correction factors for a slit (crack) and a hole

The formal definition for failure with the Point Stress Criterion is as follows:

x (b + d 0 ) = U (7.6)

The formal definition for failure with the Average Stress Criterion is as follows:

1 b + a0 ( y )dy = U

a0 b x (7.7)

When applied to the stress field at a hole, Eq. (7.2), the Point Stress Criterion provides the
following ratio between notched strength N and unnotched strength U:

(
N U = (2 Y ) 2 + 12 + 314 (KT 3) 516 718 ( )) where 1 = b (b + d 0 ) (7.8)

The corresponding ratio for the Average Stress Criterion becomes:

(
N U = (2 Y )(1 2 ) 2 22 24 + (K T 3) 26 28 ( )) where 2 = b (b + a 0 ) (7.9)

It is worth noting that the characteristic lengths for composites in tension and compression differ, as
a result of differences in the micromechanical failure mechanisms. Also, notch criteria are only
applicable for laminates where the strength is governed by in-plane fibre failure.

34
7.3 Interlaminar stresses
In the previous section we discussed the effect of macroscopic in-plane stiffness discontinuities like
holes and cracks. We will now look at the stresses caused by the stiffness discontinuities through
thickness of the laminate. These interlaminar stresses are neglected in classical laminate theory,
which only considers the in-plane stresses within each ply (intralaminar stresses).

Differences in the local stiffness of composite materials causes a discontinuous stress state under
mechanical or hygrothermal strains. The resulting internal ply stresses must however vanish at a
stress free surface, which gives rise to additional interlaminar stresses close to free edges. Figure 7.5
shows the interlaminar stresses caused by differences in shrinking during cooling of 0 and 90
plies.

Initial stress free state

Hypothetical unconstrained deformation after


curing and stresses needed to keep actual
uniform deformation

zx
Actual final state with
interlaminar shear
stresses

Fig. 7.5 Interlaminar stresses caused by hygrothermal shrinking

The discontinuous structure of a laminate also causes interlaminar stresses under pure in-plane
loading. To illustrate this we will study uniaxial tensile loading of an angle-ply laminate and a
cross-ply laminate.

Tensile loading of angle-ply laminate

Consider a []s laminate under uniaxial tension x0. The normal stresses x and y in + and
will be equal since Q11 , Q22 and Q12 are even functions of , while the shear stresses will have
opposite sign since Q16 and Q26 are odd functions of , see Eq. (4.20). Thus, the stresses become:

x ( ) = x ( ) = Q11 0 x + Q12 0 y = x
y ( ) = y ( ) = Q12 0 x + Q22 0 y = 0 0 y = 0 x Q12 Q22 (7.10)
xy ( ) = xy ( ) = Q16 0 x + Q26 0 y

Using the relation between 0x and 0y obtained from the second equation we may express the tensile
and shear stress as follows:

xy ( ) = [Q16 Q26Q12 Q22 ] 0 x [ ]


x ( ) = Q11 Q122 Q22 0 x (7.11)

35
The ratio between the shear stress xy and the applied tensile stress x for a typical carbon/epoxy
with E1=120 GPa, E2=10 GPa, 12=0.25 and G12=4 GPa has been plotted in Fig. 7.6a. Angles close
to 45 (90 mismatch) cause the largest shear stresses within the ply. A free body diagram for the
upper ply demonstrates that equilibrium in the x-direction must be maintained by interlaminar shear
stresses xz , Fig. 7b. The interlaminar shear stresses xz have to be proportional to the intralaminar
shear stress xy, and thus reach peak values for large mismatch angles. Taking equilibrium for a strip
of width Ly from the edge demonstrates that the interlaminar shear stresses must increase towards
the free edge, as the surface carrying xz stresses is proportional to Ly, while the surface carrying xy
stresses remains constant and proportional to the ply thickness tply. As a consequence interlaminar
shear stresses will vanish at some distance from the edge, which is proportional to the ply thickness.

Angle ply-Interlam. stresses


0.60 z
0.50
xy
0.40
xz
xy
0.30
xy / x

y
0.20
y
xy
0.10

0.00
0 10 20 30 40 50 60 70 80 90
-0.10
Off-axis angle [Deg]

Figs 7.6a (left) Effect of off-axis angle on shear stress vs tensile stress in angle-ply layup
Fig 7.6b (right) Interlaminar stresses caused by shear stress in angle-ply layup

Tensile loading of cross-ply laminate

Consider a [0/90]s laminate under uniaxial tension x0. The absence of a resultant load in the y-
direction implies that y(90)= y (0). The intralaminar shear stress xy will be zero since
Q16 and Q26 are zero for =0 and 90, see Eq. (4.20). Thus, the stresses become:

( )
x 0o = Q11 0 x + Q12 0 y ( )
x 90o = Q22 0 x + Q12 0 y
y (0 ) = Q + Q
o
12 0 x 22 0 y y (90 ) = Q
o
12 0 x + Q11 0 y (7.12)
y (0 ) + (90 ) = 0
o
y
o
0 y = 0 x 2 Q12 (Q11 + Q22 )
xy (0 ) = (90 ) = Q
o
xy
o
16 0 x + Q26 0 y = 0

The ply stress y and the applied average stress x = [x(0)+x (90)]/2 are now given by:

[
x = (Q11 + Q 22 ) 2 2Q12
2
(Q11 + Q22 ) 0 x ] ( )
y 90 o = [Q22 2Q12 Q11 (Q11 + Q22 )] 0 x (7.13)

The ratio between the resulting ply stress y and the applied tensile stress x for a typical
carbon/epoxy with E1=120 GPa, E2=10 GPa, 12=0.25 and G12=4 GPa is now given by:

( )
y 90 o x = 0.032 y 0 o x = 0.032 ( ) (7.14)

36
We observe that the induced stress y is much smaller than xy for the angle-ply laminate.
A free body diagram for the upper ply demonstrates that equilibrium in the y-direction must be
maintained by interlaminar shear stresses yz, Fig. 7.7. In addition, moment equilibrium around the
lower left corner of the ply demonstrates that interlaminar tensile stresses z must be present. As in
the previous example the interlaminar stresses will increase towards the free edge and vanish at
some distance from the edge, which is proportional to the ply thickness.

z
y
yz z
y y
y

Fig. 7.7 Interlaminar stresses due to uniaxial tension of a cross-ply laminate

We may conclude that angle-ply layups generally cause larger interlaminar stresses than cross-ply
laminates, and that thick plies should be avoided, as this increases the size of the region with high
interlaminar stresses. Note that matrix cracks in a laminate are stress free and act as interior free
edges. Thus, matrix cracks may cause significant interlaminar stresses.

7.4 Delamination growth


High interlaminar stresses are prone to cause local delaminations. The question if these
delaminations will grow is answered by fracture mechanics. Fracture mechanics for metals is
usually based on stress intensity factors and is focused on Mode I growth (peeling), as growth
preferentially occurs in the direction of minimum resistance to growth. Crack growth criteria for
composites are more conveniently based on strain energy release rate, which is easily measured in
tests. In contrast, evaluation of stress intensity factors depends on the local stiffness of the material,
which is non-uniform and sometimes not fully known. In addition delamination growth directions
are constrained by fibres, which often results in mixed mode growth. To provide qualitative
understanding of the factors influencing delamination growth we will look on crude estimates of the
total strain energy for some simplified cases (homogeneous properties etc)

Strain energy release at a ply drop

Consider uniaxial loading of a homogeneous laminate with a through-width delamination at a ply-


drop, Fig. 7.8. We assume thickness t of the ply-drop is much thinner than the laminate thickness T
so that bending stresses can be neglected.

37
(
E = E 1 2 )
t

0 0
T a

Fig. 7.8 Uniaxial loading of a ply-drop

The strain energy per unit volume before and after delamination of the dropped ply is given by:

Delaminated : U 1' = 0 Undelaminated : U 0' = 12 02 E (7.15)

The change in strain energy is obtained by multiplying the difference in strain per unit volume with
the volume tdA created by the growth in delamination area dA:

dU = (U1 U 0 )tdA (7.16)

The strain energy release rate G (strain energy per unit crack area) is obtained by dividing with dA:

G = dU
dA
= (U 0 U1 )t = 12 t 02 E * (7.17)

This quantity has to be compared with the critical strain energy release rate Gc for the material.

Strain energy release at ply buckling

Consider uniaxial loading of a homogeneous laminate with a through-width buckle at a


delamination, Fig. 7.9. The stress in the buckled part is now, to the first approximation, equal to the
buckling stress B.


E = E (1 )
* 2

w t
0
a T

Fig. 7.9 Uniaxial loading of a buckling ply

With the same assumptions as before (t<<T) the strain energies per unit volume become:

Delaminated : U 1' = 12 B2 E Undelaminated : U 0' = 12 02 E (7.18)

The strain energy release rate G (strain energy per unit crack area) is now given by:

G = dU
dA
(
= (U 0 U1 )t = 02 B2 ) 1
2
t (
E * = 1 B2 02 ) 1
2
t 02 E * = G 12 t 02 E * (7.19)

38
where G is a normalisation factor between zero and unity, which is expresses the fraction of
available strain energy that actually is released during delamination growth. In Fig. 7.10 this factor
has been plotted versus applied stress for three different cases of buckling induced delamination
growth. This figure shows that the available strain energy is almost entirely released ( G 1 ) when
the applied stress is much larger than the buckling stress, i.e. for thin sublaminates.

1.4

1.2

1.0
G/G0

0.8

0.6

0.4

0.2

0.0
0 2 4 6 8 10

0/ B
Fig. 7.10 Normalised strain energy release rate for delamination buckling

A final issue to consider is the mode mixity in delamination growth. Figure 7.9 illustrates that the
mode I (peeling mode) component will be associated with the factor M x = B tw , while the mode II
(forward shear) component will be depend on the factor N x = Bt . The importance of the mode I
component is decreasing for increasing delamination lengths.

The mode I and mode II components have to be compared with the corresponding toughness values.
Values for some typical materials have been illustrated in Fig. 7.11. It is observed that the influence
of mode mixity is decrasing with increasing matrix ductility (toughness).

2000

AS4/PEEK (thermoplastic)
AS4/PEEK (termoplast)
1500
G [J/m ]
2

1000
c

HTA/6376C (medelseg epoxi)


500 HTA/6376C (toughened epoxy)

AS4/3501-6
AS4/3501-6 (sprdepoxy)
(brittle epoxi)
0
0 0.2 0.4 0.6 0.8 1
G /G
II tot

39
Fig. 7.11 Interlaminar toughness vs mode ratio for different composite materials

The mixed mode delamination toughness may be tested with the Mixed Mode Bend (MMB) test,
which is a superposition of a Mode I DCB test and a Mode II ENF test. The mode ratio in the MMB
test is varied by changing the lever arm c, Fig. 7.12.

PI

dI

+
d II
dh PII
=
d
dc dII = dc + dI /4 d I /2

d I /4 d I /2
P dc
c
L L L L

Fig. 7.12 Superposition of Mode I and Mode II loading in the MMB test.

40

Potrebbero piacerti anche