Professional Documents
Culture Documents
Table of content:
2- Basic Micromechanics
7- Fabrication techniques
9- Interlaminar stresses
Laurent Warnet
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.1
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1.1. Introduction
A composite material is in its broadest definition a combination of two or more
constituents. In general, it combines the characteristics of these components in
order to obtain properties, which could not be obtained with the separate
constituents.
In this Composites Course we will restrict ourselves to continuous fibre reinforced
plastics. In these composite materials, the fibres are used for their high stiffness
and strength, while the matrix protects, binds the fibres together and transfers load
between the fibres.
The form in which fibres can be combined with a matrix is pretty infinite. Every
composite is tailored for a certain application. Even before a composite is used as
a component in a mechanical structure, it is already a structure. It would be
therefore more appropriate to talk about composite structure instead of composite
material.
Ply (a)
Laminate (b)
We will focus on composites that are made as a laminate form, i.e. a stacking of
(different) layers (or plies). This is illustrated in fig. 1.1 for layers made of
continuous fibre aligned in a unidirectional way. The laminate used in this
example is made of 3 layers where the fibres are oriented in different directions.
As far as constituents are concerned, we will focus on fibre-reinforced polymers,
which are used in laminate form.
The fibres can be organic as well as anorganic. They are manufactured by some
kind of drawing or spinning process, whereas the polymers are synthesised from
monomers.
1
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.2
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1.2. Fibres
A few types of fibre will be described next. The form in which it is used as a
reinforcement material will be dealt with in the manufacturing section of this
chapter.
2
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.3
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
3
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.4
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
chain. This structure gives an extremely high draw ratio and results in extremely
high strength. The highly drawn fibre (Dyneema from DSM, patented in 1979,
Spectra from Honeywell) contains an almost 100% crystalline structure with
perfectly arranged molecules, which promotes its extremely high strength,
modulus, and other excellent properties.
Summarising in a few words:
- Low density (950kg/m3). Further similar to aramid fibre but:
- Low upper-limit temperature (120oC)
- Low moisture absortion (1% compared to 5% for Kevlar)
Glass
E-glass 10 2540 72 / 72 3.451 4.81 5
S-glass 10 2490 87 / 87 4.31 51 3
PAN Carbon
2
T-300 7 1760 230 / 20 3.53 1.5 -0.7
3
IM-7 5 1780 300/ ?? 5.31 1.31
Pitch Carbon
4
P-55 10 2000 380 / ?? 1.9 0.5 -1.3
4
P-100 10 2150 690 / ?? 2.4 0.32 -1.4
Aramid
5
Kevlar 49 12 1450 131 / ?? 3.6 2.8 -2
Polyethylene
Spectra 900 38 970 117 2.6 3.5
1
Literature value of the bare fibre under ideal circumstances. Contact with any substrate reduces the
value by 50%.
2
Torayca
3
Hercules
4
Amoco
5
DuPont
4
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.5
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
3.50
3.00
PAN
IM7
Specific strength (x10 m /s )
2
Polyethylene
2.50
2
Aramide
6
2.00
PAN
S-Glass T300
1.50
E-Glass
1.00 Pitch
Pitch P1004
P55
Steel wire
0.50
0.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50
8 2 2
Specific modulus (x10 m /s )
fig. 1.3: Specific stiffness vs. specific strength for various fibres.
The history of synthetic polymers starts with Alexander Parkes (UK), who made
an ivory-similar material in 1860 but did not succeed in making his Ivoride or
Parkesine a commercial success. John Hyatt (USA, 1870) was commercially
5
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.6
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
successful with this Celluloid, using it as a coating for billiard balls (motivated
by the threat of a shortage of elephants). Currently, this early thermoplastic
material is mainly used for table-tennis balls. The thermosetting family of
polymers is stemming largely from the work of Leo Baekeland, with his first
patent on phenol-aldehyde plastics in 1907. He formed his General Bakelite
Company in the USA in 1910. Urea-formaldehyde (UF) compounds were
introduced in 1926, followed by melamine-formaldehyde (MF) materials in 1935.
Well-known thermoplastics were developed in the same decade: polystyrene (PS)
in 1930, polyvinylchloride (PVC) and polyamide (PA) in 1938, low density
polyethylene (LDPE) in 1938. More thermosets were introduced in the 40s:
unsaturated polyester (UP) in 1941, epoxide resins starting from a patent by
Farben in 1934, leading to CIBAs Araldite in 1946. In the 50s further
thermoplastics were developed: high density polyethylene (HDPE) in 1955,
polycarbonate (PC) in 1956 and polypropylene (PP) in 1957. Technical
thermoplastics are available since the seventies, such as Polyphenylene Sulphide
(PPS) in 1973, Polyetherimide (PEI) in 1982, Polyetheretherketone (PEEK) in
1987.
Example: polyethylene
H H H H
n C C C C
H H H H
n
6
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.7
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
O O
O O n
R H + C C C C
7
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.8
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1.3.1. Thermosets
Epoxy resins
Epoxy resin is a generic O O
term that covers a wide
C C CH2 CH CH2
range of crosslinked
polymers, which are based
on polymers containing
epoxide group and the glycidyl group
epoxide groups. Often,
epoxide resins are manufactured from glycidyl chloride (epichlorohydrin,
ECH). The epoxide resins will react, in general, with compounds containing
active hydrogen atoms. A typical example of the polymerisation of epoxy is
the reaction between a diepoxide (functionality 2) with a primary diamine
(functionality 4).
O O
O OH OH O
N R' N
O OH OH O
8
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.9
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
The nature of the epoxide resins and the reagents causes the molecular
structure of the crosslinked epoxy resins to differ. Consequently, the
mechanical properties and the useful temperature range of epoxies, as well as
processing viscosity can vary over a wide range. Curing conditions and
therefore the cross-link density also have an important effect on the
mechanical properties. For example, the complete reaction of the epoxide
groups can only be achieved after long time at high temperature. This is
mostly achieved by post-curing and is important for an increase in Youngs
modulus, chemical resistance and useful temperature range.
Polyester resins
The base for this thermoset is a linear unsaturated polyester resin dissolved in
styrene monomer with added catalyst (often a peroxide), accelerators and
inhibitors. Usually, the process of cross-linking is very slow at room
C CH CH C O R O + CH CH2
O O styrene
unsaturated polyester
CH
CH2
C CH CH C O R O
O O
9
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.10
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
The vinyl esters have a very good chemical resistance and a higher
temperature range than their polyester counterparts. They also offer a low
viscosity and fast curing, better than the epoxy systems. However, the
chemical shrinkage is higher than the epoxies, in the order of 5-10%.
1.3.2. Thermoplastics
The molecules in a polymer are of varying sizes. In a thermoplastic, these are so
tangled together that instead of a sharp melting point to an easy flowing liquid,
they soften over a much wider range of temperatures to form a melt which will
flow with difficulty. The macromolecules can be linear or branched. If the
polymeric molecules have few branches which increase the distance between the
main chains, then the parallel lengths of the chains can form crystalline regions.
Due to the limited mobility of the polymer chains, a thermoplastic will never
reach a fully crystalline state: the material is at most semi-crystalline. Multi-
branched molecules cannot be close together so there will be less crystallinity and
the density will be lower. Polymers with low crystallinity are called amorphous.
Only some of the thermoplastic matrices, which are used as a basis for continuous
fibre reinforced composites (high performance), are shortly reviewed here.
10
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.11
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
H H
C C
H CH3
n
Polypropylene is a semicrystalline thermoplastic, with a glass transition at 0C
and a melting temperature range from 158-168C. PP has a density of 900
kg/m3. It should not be used in air above 110C. Suitable forming
temperatures of PP/glass composites are around 200C.
O O C
With a glass transition at 90C and a melt temperature range from 280-288C
PPS is another high temperature thermoplastic. It can be used safely up to
240C. Suitable forming temperatures are around 350C. The material
crystallises rapidly and has an excellent solvent resistance. PPS is more brittle
than PEEK but also significantly cheaper.
11
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.12
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
O O
CH3
C C
N N
C O C O C
O CH3 O n
3
PEI has a glass transition temperature of 215C and a density of 1270 kg/m .
Typical forming temperatures are around 350C.
1.3.3. Properties
The table below presents some typical properties of selected matrices.
Properties of thermosets are indications, as they very much can vary according
to their compositions. The properties of the semi-crystalline thermoplastics
will vary with their degree of crystallisation.
6
Curing and post curing temperature
12
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.13
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
13
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.14
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
14
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.15
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
15
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.16
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1.4.7. Pultrusion
Pultrusion is a continuous process, which makes it possible to produce infinite (in
principle) length of composite having a certain profile. Bundles of fibres
impregnated with a thermosetting matrix are pulled through a (heated) die having
the required profile cross-section form. Layers of random mat or woven cloth can
also be added before the die, as shown in fig. 1.10.
16
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 1.17
R
M
OU
CO
1. Constituents and manufacturing techniques - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
phenylene OH phenol
O
epoxide CH CH2 styrene
CH2 CH
O
glycidyl
CH2 CH CH2
O
C
CH
N imide
CH
C
17
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.1
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
2. Micromechanics
2.1. Introduction
In the field of composite material, micromechanics
is about mechanics at the level of the constituents,
i.e. the fibre, the matrix and possibly the interface.
This can be used for example to study the stress-
strain situation around a fibre under different
loading conditions in order to understand the failure
behaviour of composites.
In our case, we will concentrate on fig. 2.1: Fibre-matrix scale
micromechanical models, which are built in order
to find a relation between the properties of the constituents and that of the
composite layer. These models allow the designer to evaluate the combination of
different constituents. Another important aspect is that of understanding
composites.
The first part of this chapter is dedicated to the relation between the constituents
volume and mass fraction. Then, two methodologies giving expressions for in-
plane mechanical properties are discussed. A numerical example is also given.
This chapter shows that although convenient and helpful, these models should be
used in conjunction with some experimental work. Experimental data of the
constituents is necessary as input to the models, but also to validate the model
results.
Substituting the product of density and volume for mass for each constituents
gives:
n
c = i vi (2.3)
i =1
The density can also be obtained as a function of mass by substituting the quotient
of mass by density for volume in equation (2.1), this gives:
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.2
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1
c = n
(m
(2.4)
i / i )
i =1
Using the same substitution, it is possible to calculate the void fraction from
equation (2.1):
M f Mm
+
f m
vv = 1 (2.5)
Mc
c
Typical values:
The packing (or distribution) of fibres within a
composite cannot be described. As shown in fig.
2.2, the distribution of fibres is unhomogeneous. In
order to build micromechanical models, simplifying
assumptions are made on the packing of fibres. The
most simple packing is the square packing as shown
in fig. 2.3.
fig. 2.2: Typical composite
cross-section micrograph
vf = (2.6)
4s
An other typical packing is the triangular (fig. 2.4)
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
vf = (2.7)
2 3s
Maximum packing is obtained in both packing models for d=s. It gives in the case
of square packing vf-max=0.79, and for the triangular packing 0.91.
In the practice, fibre volume fraction for composite based on unidirectional layers
can be found in the range 0.5 to 0.8.
An other remark concerns the void content. Typical autoclave (pressure+vacuum)
cured composite products have voids content varying from 0.1 to 1%. Pressure
bag (no vacuum) cured composites can have voids content in the order of 5%.
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1 = m vm + f v f (2.9)
It is now assumed that both fibre and matrix behave in a linear elastic way, and
that the matrix is isotropic and is characterised by the elasticity modulus Em. The
fibre is assumed orthotropic (important in case carbon or aramid fibre are
considered) and is therefore characterised by a longitudinal modulus Ef1 and a
transverse modulus Ef2. Poisson strains are neglected. With these assumptions, the
equation (2.9) can further be developed as:
E1 1 = Em m vm + E f 1 f v f (2.10)
It was earlier assumed that the strain was equal for both volumes and therefore:
E1 = Em vm + E f 1v f (2.11)
This relation is called the rule of mixture for the longitudinal modulus. As will be
shown later on basis of comparison with experimental data, the Voigts model
provides, and is generally known as an adequate prediction for predicting the
longitudinal modulus.
2 L = m Lm + f L f (2.12)
where Lm and Lf are the lengths of the matrix and fibre volumes. Using equation
(2.1) applied to the volume lengths, we obtain:
2 = m vm + f v f (2.13)
Here again we assume that the 1D Hookes law applies. Equation (2.13) can
further be developed as:
2 m vm f v f
= + (2.14)
E2 Em Ef2
From the series volume arrangement, the stress in matrix, fibre and the composite
are equal, an expression for the transverse elasticity modulus E2 is obtained:
1 v v
= m + f (2.15)
E2 Em E f 2
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.5
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
This relation is often called the inverse rule of mixture. Note again that the fibre
transverse modulus should be used. This equation is generally known as being
inadequate for predicting the transverse modulus. This is due to the fact the
assumption made on the equality of the stress in the matrix and the fibre in the
volume-in-series model is not valid in a real composite. This can be shown on
basis of strain energy approach. A second reason for the inaccuracy of models for
the transverse modulus composite based on orthotropic fibre (carbon and aramid)
is that the fibre transverse modulus is difficult to measure (and has actually never
directly been measured). Quoted values for the transverse modulus of fibres are
actually derived from the comparison between micromechanical model results and
experiments. The same actually applies for the shear modulus of these orthotropic
fibres.
v f Em E f 2
E 2 = (1 v f ) Em + (2.16)
v f E m + v f (1 v f ) E f 2
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1 1 v f vm
= + (2.19)
E2 v f + v m E f 2 Em
This equation gives for =1 the inverse rule of mixture. A similar equation can be
written for the shear modulus.
Other widely used equations are the Halpin-Tsai equations, which for the
transverse modulus are:
E 2 1 + v f E f 2 Em
= , with = (2.20)
Em 1 v f E f 2 + Em
where is the curve fitting parameter. It is worth noting that giving the value 0
gives the inverse rule of mixture (2.15). It was shown that using =2 gives similar
results as more complex elasticity solutions. Replacing Ei by Gi in equation (2.20)
gives the Halpin-Tsai equation for the shear modulus. A curve fitting parameter
of 1 is mostly used.
http://www.pt.ctw.utwente.nl/organisation/tools/
2.5 Example:
An example is given, based on a unidirectional carbon reinforced Polyetherimide
(PEI) having the following constituents properties:
PEI Em Gm
m m m
(GPa) (GPa) (/oC) (kg/m3)
table 2.1: Linear elastic thermomechanical properties of the Torayca T300 carbon fibre, and the
Ultem1000 PEI.
The matrix mass content of the composite obtained with these components is
mm = 41.4%, as measured by desolving the matrix of several composite specimen
in Choroform. The corresponding fibre volume fraction is vf = 51%, The measured
in-plane properties can be found in table 2.2. The measured longitudinal modulus
compares very well with the rule of mixture (119 GPa). The transverse modulus
according to different models are given in table 2.3.
1
Comes as well as Gf12 from R.F. Gibson's Principle of Composite Material Mechanics book.
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.7
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
carbon-PEI E2 (GPa)
Experimentally: 7.8
Inverse rule of mixture (2.15) 5.0
Parallel-series model (2.16) 7.7
Halpin-Tsai =2 6.5
Halpin-Tsai =10 7.85
table 2.3: Different models for the transverse modulus
The results obtained from the different models are set in graph form as a function
of the fibre volume fraction. The experimental value is also marked as a reference.
1.6E+10
1.4E+10
1.2E+10
experimental value
1.0E+10
E 2 (Pa)
8.0E+09
6.0E+09
4.0E+09
2.0E+09
0.0E+00
0 0.2 0.4 0.6 0.8 1
vf
Inverse rule of mixture Parallel-series model
Halpin-Tsai x=2 Halpin-Tsai, x=10
fig. 2.9: Composite transverse modulus as a function of the fibre volume fraction.
It is clear from fig. 2.9 that the inverse rule of mixture does not apply to the
carbon-PEI considered. In a less extent, a similar conclusion can be drawn for the
Halpin-Tsai having =2. The parallel-series model and Halpin-Tsai with =10
both give a good approximation of the composite's transverse modulus. As
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.8
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
discussed earlier, these models are dependent on the value chosen for the fibre's
transverse modulus. One should therefore be careful when drawing conclusion.
An other remark is that the parallel-series model and Halpin-Tsai with =10 do
not follow the same trend as a function of the fibre volume fraction.
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.9
R
M
OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
2.6 Problems:
2.1 Derive equation (2.16)
2.2 Find in the literature expressions for the longitudinal and transverse
coefficients of thermal expansion 1 and 2 as a function of de components
properties.
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.1
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
3.1 Introduction
In the first chapter, it was shown that
the properties of the basic
constituents of a composite material
fibre and matrix- are extremely
different. This implies that a
composite made of these two
constituents is heterogeneous
(properties vary from point to point)
when considered at a fibre matrix
scale as shown in fig. 3.1. When
coming to a composite layer, it can
be assumed that the anisotropic fig. 3.1: Fibre-matrix scale, ply scale.
properties are independent of the
position from which they are looked at. It makes it possible to assume a composite
as a plate having averaged anisotropic material properties. The stress-strain
relations of a composite layer in general will be developed next.
Tensor
11 22 33 23=23 31=31 12=12 11 22 33 23=223 31=231 12=212
notation
Contracted
1 2 3 4 5 6 1 2 3 4 5 6
Notation
Assuming small linear elastic deformations, the strain components are related to
the deformations by the following relations:
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.2
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
u1 u u
1 = ; 2 = 2 ; 3 = 3 ;
x1 x2 x3
(3.1)
u u u u u u
4 = 2 + 3 ; 5 = 3 + 1 ; 6 = 1 + 2 ;
x3 x2 x1 x3 x2 x1
where u1, u2 and u3 are the displacement along the 1, 2, 3 axis respectively.
The relation between stress and strain components describes the behaviour of the
material when loading is applied. We will consider linear elastic materials, and in
this case, each strain component i is a linear function of the stress components
j (j = 1,,6):
6
i = Sij j (3.2)
j =1
The stress-strain relationship just described is valid for any anisotropic elastic
material. It will be shown that the amount of compliance (or stiffness) constants
can be reduced.
Firstly, it can be shown that the compliance (and stiffness) matrix is symmetric.
The demonstration of this symmetry is based on the fact that the energy necessary
to induce deformation to the elastic (and hence reversible) material is independent
from the way the energy is produced. This leads to 15 equations of the form
i j
= (3.4)
j i
This means that the compliance matrix is symmetric (and therefore the stiffness
matrix), i.e.:
Sij = S ji ; Cij = C ji (3.5)
This leaves 21 of the 36 compliance constants to describe an anisotropic material.
3
Further simplifications can be made
by considering geometrical 2
symmetries in the composite layer
considered. Take as an example a
layer of unidirectional material as 1
shown in fig. 3.3. Let us define a
Cartesian coordinate system where fig. 3.3: A unidirectional layer and its material
the first axis (1) coincides with the coordinate system
direction along the fibres. The two
other directions are then perpendicular to the fibres (2 & 3), with the third axis
perpendicular to the layer. This coordinate system is called the material coordinate
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.3
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
system. It can be seen that there are 3 planes of material property symmetry (i.e.,
the planes defined by the coordinate system axis 1&2, 2&3, 1&3. A material
agreeing to these symmetry assumptions is called orthotropic. It can be shown
that this leaves a compliance matrix with 9 independent components.
1 S11 S12 S13 0 0 0 1
S 22 S 23 0 0 0 2
2
3 S 33 0 0 0 3
= (3.6)
4 Sym S 44 0 0 4
5 S 55 0 5
6 S 66 6
As an example, apply a stress 1 along the first axis (all other stress
components are zero). This means that the relation for the in-plane shear
reduces to
6 = S 61 1 (3.7)
It can be seen from fig. 3.3 that a stress applied along the fibres cannot
generate any in-plane shear deformation. Therefore the compliance matrix
component S61=0. Such simple experiments can be applied for the other zero-
components.
It is important to note that the concept of orthotropy is dependent on the
coordinate system chosen. For example, a normal stress applied in an arbitrary
direction will induce shear strains. This will be shown later on when rotating the
compliance (or stiffness) matrix.
A particular case of orthotropy is transverse isotropy. This assumption is taken
for unidirectional layers, where the properties are assumed isotropic in the 2-3
plane, i.e. perpendicular to the fibres. The compliance matrix contains now 5
independent constants and can be further reduced to the following set of relations:
1 S11 S12 S12 0 0 0 1
S 22 S 23 0 0 0 2
2
3 S 22 0 0 0 3
= (3.8)
4 Sym 2( S 22 S 23 ) 0 0 4
5 S 66 0 5
6 S 66 6
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
where the superscript k in the strain component kk reminds that it finds its origin
in a k -stress.
For k=1,2,3: Rk is the modulus of elasticity along the k-axis and will be
given the notation Ek.
For k=4,5,6: Rk is the shear modulus relating respectively to the 2-3, 3-1 and
1-2 planes and along the k-axis and will be given the notation
Gk.
If we now consider the ik k -strain component, a deformation ratio is defined as:
kl = lk / kk (3.13)
For k and l = 1,2,3, kl are the Poisson's ratios with a negative sign (kl). In the
case of an orthotropic layer, all other deformation ratios are zero, which is not true
for an anisotropic material in general. In this case, it is worth mentioning the
shear-extension coupling coefficients kl (Lekhnitskii) and the shear-shear
coupling coefficients kl (Chentsov). The earlier applies for k = 1,2,3 and l = 4,5,6
or k = 4,5,6 and l = 1,2,3. The later one concerns k and l = 4,5,6.
These definitions make it possible to fill the compliance matrix. For an orthotropic
layer, this gives:
Skk = 1 / Rk for k = 1,..,6;
lk lk kk kl (3.14)
Skl = = = for k = 1,..,3, l k
k kk k Rk
The equations are:
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.5
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1 21 31
E 0 0 0
E2 E3
1
12 1 32
1 0 0 0 1
E1 E2 E3
2 13 23 2
0
1
0 0
3 E1 E2 E3 3
= (3.15)
4 0
0 4
1
0 0 0
5
5
G23
1
6 0 0 0 0 0 6
G13
1
0 0 0 0 0
G12
The other strain components can be non-zero, but we will focus on the behaviour
of the in-plane strains. Inverting the compliance matrix gives:
E1 21 E1
0
1 1 12 21 1 12 21
1
21 E1
2 =
E2
0 2 or { } = [C ] { } (3.18)
1 12 21 1 12 21
6 0 0 G12 6
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
In general, the transformation of the coordinates of a point from the (x1, x2, x3)
system to the (x1, x2, x3) system (fig. 3.4) can be written as:
3
xi' = ij x j with ij = cos( xi' , x j )
j =1
(3.19)
i.e. 11 = 22 = cos ; 12 = 21 = sin
13 = 31 = 23 = 32 = 0; 33 = 1
The transformation of the Cartesian components of any tensor tij is then given by:
tij' = ik jl tkl (3.20)
In matrix form, the stress transformation gives:
1' m 2 n2 0 0 0 2mn 1
'
2 n
2
m2 0 0 0 2mn 2
3 0 0 3 m = cos
'
0 1 0 0
'= with
4 0 0 0 m n 0 4 n = sin
5' 0 0 0 n m 0 5
' 2
6 mn mn 0 0 0
m n 6
2
(3.21)
This transformation matrix will be called 3
2*
[T] in the following. 2
1
1*,2*,3: layer CS
1,2,3: material CS (at )
fig. 3.5 shows the chosen notations for the coordinate system: (1, 2, 3) for the
material CS as in fig. 3.3, (1*, 2*, 3) for the CS of the layer. The properties in the
(1, 2, 3) coordinate system are known, and need to be transformed into the layer
CS. Compared to the situation in fig. 3.4, it therefore implies to rotate with an
angle -, which means using [T]-1. In case of plane stress, the stress transformation
can be written:
1* cos2 sin 2 2 cos sin 1
2 cos sin 2 or { * }= [T ] { }
* 1
2 = sin cos2
2
(3.22)
* cos sin cos sin cos2 sin 2 6
6
The transformation of the strain components is performed similarly, but the
transformation described above based on tensors is valid for 12, and therefore 6/2
(see table 3.1), i.e:
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.7
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1* 1
* 1
2 = [T ] 2 (3.23)
1 * 1
2 6 2 6
This is usually corrected by introducing the Reuter matrix [R]. In plane stress
form, it can be written as:
1 0 0
[ ] 0 1 0
R = (3.24)
0 0 2
i.e.:
1* cos2 sin 2 cos sin 1
*
2 = sin cos2 cos sin 2
2
(3.26)
* 2 cos sin 2 cos sin cos2 sin 2 6
6
It is now possible to formulate a transformed stiffness matrix [C*]:
1* 1 1 1*
* 1 1 *
2 = [T ] 2 = [T ] [C ] 2 = [T ] [C ] [R] [T ] [R ] 2
1 1
* * (3.27)
6 6 6 6
=[C*]
This gives for an orthotropic material with its stiffness matrix written in its
material coordinate system (i.e. C16=C26=0):
C11* = C11 cos 4 + 2(C12 + 2C66 ) sin 2 cos 2 + C22 sin 4
*
C 22 = C11 sin 4 + 2(C12 + 2C66 ) sin 2 cos 2 + C22 cos4
C12* = (C11 + C22 4C66 ) sin 2 cos 2 + C12 (sin 4 + cos 4 )
(3.28)
C16* = (C11 C12 2C66 ) sin cos3 + (C12 C 22 + 2C66 ) sin 3 cos
*
C 26 = (C11 C12 2C66 ) sin 3 cos + (C12 C22 + 2C66 ) sin cos3
*
C66 = (C11 + C22 2C12 2C66 ) sin 2 cos 2 + C66 (sin 4 + cos 4 )
* *
6 6 6 6
=[S*]
(3.29)
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.8
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
R
M
OU
CO
3. Stress-strain relationships of a composite layer - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
3.7 Problems
3.1 Write expressions for E2* , G12* , 12* , 16* and 26
*
.
3.3 A filament wound cylindrical pressure vessel of mean diameter d=1m and
wall thickness t=0.02m is subjected to an internal pressure p. The vessel is
wound at an angle of 54o from its longitudinal axis and the glass-epoxy
material used has the following mechanical properties:
E1 = 40GPa, E2 = 10GPa, G12 = 3.5GPa, 12 = 0.25.
By the use of a strain gauge, the strain along the fibre direction is measured
and has at the pressure a value of 1=0.001m/m. Determine the internal
pressure in the vessel.
SI T ES G
O
Composites Course 2007 University of Twente, Eng. & Tech. 4.1
R
M
OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
4.1. Introduction
Chapter 2 and 3 focused on the behaviour of a single layer. The thermo-
mechanical properties of such a layer can be predicted, and the stress-strain
relationship is known. This will now be extended to the more general case of a
laminated plate. The most common type of analysis will be developed next, and is
known as the classical lamination theory. This plane stress theory makes it
possible to relate external loads (in-plane forces and moments) to the composite
plate deformations.
The analysis of laminates will first be introduced by considering the behaviour of
simple laminated beam under pure bending.
R
M
OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
x =
( + z ) d d
z
= (4.1)
d
According to assumption (c), the longitudinal stress at a distance z from the
middle axis becomes:
z
x = Ex (4.2)
Static equilibrium gives an expression for the bending moment:
h
2
M= x b z dz (4.3)
h
2
By taking into account the stress xk and elasticity modulus E xk in each layer k,
we obtain:
b z2
E x (zk 1 z k3 )
N zk
b N k 3
M = E xk dz = (4.4)
k =1 zk 1 3 k =1
The bending moment can also be expressed as a function of the elasticity modulus
of the laminated beam E xl :
E xl I y
M= with I y = b z 2 dz (4.5)
With Iy the moment of inertia with respect to the y-axis. From (4.4) and (4.5), an
expression for the elasticity modulus of the beam can be obtained:
E xl =
b N k 3
3I y k =1
E x z k 1 z k3 ( ) (4.6)
R
M
OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
R
M
OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
displacement relations in order to find an expression for fu and fv. The strain in a
layer and in its coordinate system (1*, 2*, 3) can be written as:
v w w w
4* = + = f v ( x, y ) + = 0 f v ( x, y ) =
z y y y
u w w w (4.9)
5* = + = f u ( x, y ) + = 0 f u ( x, y ) =
z x x x
Using the expressions for the in-plane displacements (4.8), the in-plane strains
are:
u u 0 2w 2w
1* = = z 2 = x0 z 2
x x x x
v v 0
w
2
2w
2* = = z 2 = y0 z 2 (4.10)
y y y y
u v u 0 v 0 2w 2w
6* = + = + 2z = xy0 2 z
y x y x xy xy
With x0 , 0y , xy0 the in-plane strains in the mid-plane. It can be shown from
geometrical considerations that:
2w 2w 2w
x = ; = ; = 2 (4.11)
x 2 y 2 xy
y xy
with the plate curvature (1/). The equation (4.10) relates the strain in a layer
with quantities related to the laminate: the mid-plane strain and the curvature.
R
M
OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
N zk
{N } = ([ C *] k { 0 } + z [C *] k { }) dz
z
k =1 k 1
N zk zk
{N } = [C *] k { 0 } dz + [C *] k { } zdz
(4.14)
k =1 z k 1 z k 1
N 1 N
{ N } = [C *] k ( z k z k 1 ) { 0 } + [C *] k ( z k2 z k21 ) { }
k =1 2 k =1
This is mostly rewritten in the following way:
Nx A11 A12 A16 x0 B11 B12 B16 x
Ny = A22 A26 y0 + B 22 B 26 y (4.15)
N sym A66 xy0 sym B66 xy
xy
The A-matrix is also called the laminate extensional stiffness matrix, is symmetric
and its components are defined as:
N
Aij =
k =1
( C ij *) k ( z k z k 1 ) (4.16)
The B-matrix is called the laminate coupling stiffness matrix, is symmetric and its
components are defined as:
1 N
Bij =
2 k =1
( C ij *) k ( z k2 z k2 1 ) (4.17)
A similar development can be performed for the moment resultants. This gives as
end result the following relations:
Mx B11 B12 B16 x0 D11 D12 D16 x
My = B 22 B 26 y0 + D 22 D 26 y (4.18)
M sym B66 xy0 sym D 66 xy
xy
The B-matrix is also present here. The D-matrix is called the laminate bending
stiffness matrix, is symmetric and its components are defined as:
1 N
Dij = (Cij *) k ( zk3 zk3 1 )
3 k =1
(4.19)
The relations (4.15) and (4.18) are often written in partitioned form:
N A B 0
=
D
(4.20)
M B
Although the three components of the 'ABD' matrix have similar appellations
(stiffness), they have distinct units. As loading is mostly expressed "per unit
width" (force resultants N in N/m and moments M in N), the A-components have
for unit N/m, the B-components in N and the D-components in Nm.
SI T ES G
O
Composites Course 2007 University of Twente, Eng. & Tech. 4.6
R
M
OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
[ a ] = [ A] 1 + [ A]1 [ B ] ([ D ] [ B ] [ A] 1 [ B ]) [ B ] [ A] 1
1
(4.22)
This relation reduces for to [A]-1 when the components of the [B]-matrix are zero
(laminate is symmetric). Components of the [b]- and [d]-matrix can be evaluated
as follow:
[ b ] = [ A] 1 [ B ] ([ D ] [ B ] [ A] 1 [ B ])
1
(4.23)
[ d ] = ([ D ] [ B ] [ A] [ B ])
1 1
Similarly to the [A]-matrix, the inverted bending stiffness [d] components can be
calculated by directly inverting the [D]-matrix only if the components of the [B]-
matrix are zero. It is worth adding that the coupling compliance matrix [b] is not
symmetric. The relation between the curvature vector {} and the force vector
{N} is defined by the transposed of the [b]-matrix.
4.4. Problems
4.1 Derive for the [ABD]-matrix for 2 laminates [-45/45]s (symmetric) and
[-45/45/-45/45] (antisymmetric) based on 1mm thick layers of Carbon-PEI
having the following layer properties:
E1=120GPa, E2=8GPa, G12=3,5GPa, 12=0.3
In which [ABD] terms do these two lay-up differ, and what does this mean?
4.2 Give on the next figure the physical signification of the following stiffness-
matrix terms:
A16 and A26; B11, B22 and B12, B16 and B26, B66, D16 and D26
SI T ES G
O
Composites Course 2007 University of Twente, Eng. & Tech. 4.7
R
M
OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
R
M
OU
CO
5. Classical lamination theory-2 - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
5.1. Introduction
The classical lamination theory as summarised in chapter 4 allows us to find a
relation between the laminated plate deformation on one side, and the loading on
the plate on the other side. The deformations considered were the mid-plane in-
plane strains ( x0 , y0 , xy0 ) and the plate curvatures ( x , y , xy ). The loading
studied were in-plane force (Nx, Ny, Nxy) and moments (Mx, My, Mxy).
From this relation, in-plane stresses and strains can be deduced in any layer of the
laminate.
From this displacement-loading relation, engineering constants of the laminate can
also be derived. As for the single layers in chapter 3, this will be done by
performing simple experiments.
As shown in chapter 3, anisotropy of composites is also present in hygrothermal
properties. It means stresses and strains are sensitive to changes in temperature as
well as in moisture contents. The influence of temperature is already noticeable
during fabrication, as most composites are consolidated above room temperature.
Taking hygrothermal loading into account in the deformation-loading relation
will be developed in the second part of this chapter. An example on the
calculation of thermal residual stresses in a laminated cross-ply beam ([0/90]s)
will also be given.
Using the same loading case, an expression for the in-plane Poisson's ratio can be
written:
0y a N a
xy = 0 = 21 x = 21 (5.3)
x a11 N x a11
Similarly, the transverse elasticity modulus of the laminate Ey is obtained by
applying a force Ny and is then:
1
Ey = (5.4)
h a 22
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 5.2
R
M
OU
CO
5. Classical lamination theory-2 - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
R
M
OU
CO
5. Classical lamination theory-2 - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
N zk
{N } = ([ C *] { 0 } + z[C *] { } T [C *] { *} c.[ C *] { *} ) dz
z k k k k k k
k =1 k 1
N 1 N
{ N } = [C *] k ( z k z k 1 ) { 0 } + [C *] k ( z k2 z k21 ) { }
k =1 2 k =1
N N
T [C *] k ( z k z k 1 ) { *} k c [C *] k ( z k z k 1 ) { *} k
k =1 k =1
(5.11)
This is mostly rewritten in the following way:
{N } = [A] { 0 }+ [B] { } {N Th } {N H } (5.12)
Where the [A]- and [B]-matrix are defined in (4.16) and (4.17). The components
of vectors {NTh} and {NH} are fictive thermal and hygroscopic forces respectively
defined as:
N
N iTh = T
k =1
( C ij *) k ( j *) k ( z k z k 1 ) (5.13)
And :
N
N iH = c ( C ij *) k ( j *) k ( z k z k 1 ) (5.14)
k =1
The term fictive emphasises that the components in {NTh} and {NH} are not
actual forces. It only means that for the deformations to be zero, {0}={0} and
{}={0}, a force vector {N} equal and opposite to {NTh} must be applied to the
system.
A similar development can be performed for the moment resultants. This gives as
end result the following relations:
{M } = [B ] { 0 } + [D ] { } {M Th } {M H } (5.15)
Where the D-matrix is defined in (4.19). The components of vectors {MT} and
{MH} are fictive thermal and hygroscopic moments respectively defined as:
N
1
M iTh =
2
T
k =1
( C ij *) k ( j *) k ( z k2 z k21 ) (5.16)
And:
N
1
M iH =
2
c
k =1
( C ij *) k ( j *) k ( z k2 z k2 1 ) (5.17)
Now the laminate stiffness matrix and the fictive force and moment vector are
known ([A], [B], [D], {NTh} ,{NH} ,{MTh} & {MH}), it is possible to express the
laminate deformations:
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 5.4
R
M
OU
CO
5. Classical lamination theory-2 - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
0 a b N N Th N H
= T + +
d M M Th M H
(5.18)
b
5.4. Example:
Thermal residual stress in a cross-ply laminate:
1D solution.
Roughly said, residual stresses are stresses, which are present in a material even
before it is loaded. These stresses will build up during fabrication, when the
matrix is consolidated to form the composite. The following example is meant to
illustrate the formation of thermal residual stresses in a simple laminate on a
macroscopic scale. The matrix used consolidates at a temperature higher than the
temperature at use. A cross-ply [90/0/90] beam is shown in fig. 5.1, where three
situations are depicted:
The situation at a temperature where the matrix starts to consolidate. In
the case of an amorphous thermoplastic material at the glass transition
temperature Tg. It is assumed that the layers have the same initial length.
The situation at a lower temperature, in the case the layers can freely
contract, i.e. with a strain i.T.
The situation at room temperature, in the most common case that the
layers were not able to freely contract due to the temperature change.
u90T
u0T
90o
: at Tg
0o
: i T 90o
u90m u0m
90o
: at RT
0o t0
90o t90/2
R
M
OU
CO
5. Classical lamination theory-2 - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
where i and Ei are respectively the coefficients of thermal expansion and the
Youngs moduli of the unidirectional 0 and 90 layers. The stress equilibrium
equation is:
90
r
t90 + 0r t0 = 0
with ti the layer thickness. This equation then leads to the following relation for
the thermal stress in 90 layer:
E t E ( 0 )T
90r = 90 0 0 90
t 0 E0 + t 90 E90
Laminate compliance
[a], [b], [d] (4.22-23)
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
6. Laminate failure
6.1. Introduction
It has been shown in the first lectures that composite laminates have anisotropic
thermo-mechanical properties. The behaviour of laminates at failure is also highly
directional in nature. The strength of the constituents plays an important role in this
anisotropy. For example, the strength of an E-glass fibre is around 1750 MPa,
compared to a value of about 50MPa for an epoxy matrix. This means for example
that a unidirectional layer will sustain a higher stress when loaded parallel to the
fibres than perpendicular to the fibres.
When studying the (potential) failure behaviour of laminates, the term strength
alone is not appropriate. It suggests that the problem is limited to a simple change in
state from non-failed to failed, and would not provide a realistic representation
of the complexity of laminate failure.
The presence of fibres makes a composite a structure having an infinite amount of
potential stress concentrations for a crack to initiate. More than this, the presence of
fibres and layers makes a composite a sum of a nearly infinite amount of surfaces,
all of which are a potential site for a crack to grow1. This means that even if a
certain amount of cracks occurs within a composite, it does not necessary mean the
complete fracture of the structure. There might still be enough cohesion between
fibre and matrix to carry load. This makes the qualitative description, i.e. the way
the laminate chronologically fails an important part of a composite failure study.
The quantitative prediction of failure (ultimate stress, strain or energy) is of use for
the designer. However, it will be seen that the prediction of failure is mostly
phenomenological (only the occurrence of failure is predicted, not the actual mode
of failure). They only represent the level of stress at which failure occurs.
Both qualitative and quantitative approaches can be performed at different scales
(microscopically: the interface between fibre and matrix fails, macroscopically: a
delamination separating two adjacent layers).
These lecture notes will first present an overview of in the literature often
mentioned failure mechanisms. Then, simple mechanics of material failure criteria
aiming at predicting the initiation of a failure history will be described. A typical
failure analysis of a laminate is then presented. A short insight in fracture
mechanics is then given.
1
This typical characteristic of composite is used in structures made to absorb impact energy: Such a
structure is designed as such that as many crack surfaces as possible are created.
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.2
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
h t0
y
45o t90 /2
x
z d
L
(a) (b)
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.3
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
fig. 6.1: Transverse cracking in a [90/0]s laminated beam in a three point bending loading situation.
fig. 6.2: Micrograph of cross-sectioned 90o layer in the region of a transverse crack..
6.2.2. Delamination
The term delamination is directly related to the process of laminating. Just in
the same way as producing a laminate is about bonding different layers to each
other, delaminating is about debonding the layers. It is therefore a crack, which
again runs in a plane parallel to the fibres, but at the interface between two layers
(can also within a layer).
Chronologically, it is recognised that a delamination mostly initiates from the tip of
a transverse crack. Using a similar illustration as for the transverse crack, the
transverse crack is shown to grow at the interface between the 0o and the 90o layer.
Another example of the combination transverse crack delamination is given on a
micrograph of a cross-section of a glass-reinforced epoxy tube subjected to an
impact test. Delamination will further affect the thermo-mechanical properties of
the laminate, for example its bending stiffness. However, the presence of
delaminations does not generally lead to the complete fracture of the laminate.
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.4
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
90o
0o
y 90o
x
Transverse crack Delamination
z
fig. 6.3: Transverse cracking in a [90/0]s laminated beam in a three point bending loading
situation.
fig. 6.4: Cross-section of an impacted composite tube illustrating transverse cracks and
delaminations.
fig. 6.5: SEM (Scanning Electron Microscope)picture showing delaminations and fibre fracture
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.5
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
Fibre related failure can also involve local buckling of fibre, or fibre pull-out, as
shown in fig. 6.6.
fig. 6.6: Micrograph showing a fibre bundle buckling (left), and a SEM showing fibre pull-out
on a fracture surface
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.6
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
It should first be made clear that failure criteria are not based on any physical theory
of failure. They are not concerned with the details of micromechanical failure
modes like the growth of a defect, fibre pull-out or interfacial failure. A mechanics
of material strength criterion merely defines a stress (or strain) limit upon which
failure occurs. This can also be done in a biaxial stress space (or triaxial), where
failure envelopes are defined. The stress limit or the envelope are related to basic
lamina strength values. These lamina strength are obtained by measurements on
unidirectional lamina, and in such a way that a uniform stress state is imposed on
the specimen. The envelopes describing a multiaxial state of stress are then defined
by "curve fitting" the basic experimental failure data.
Most of these criteria are based on existing criteria used for isotropic material. For
example, the Tsai-Hill criterion discussed later is based on the von Mises criterion,
which was originally meant for describing yielding.
In the following, we will use the 5 lamina strength:
s1t, s1c, s2t, s2c, s6,
where the subscript 1, 2, and 6 relate to the loading directions parallel to the fibres,
perpendicular to the fibres and to the in-plane shear respectively. The second
subscript t and c relates respectively to the tensile and the compression loading. In
most criteria found in the literature, all numerical values are assumed positive.
The limits described by these criteria (failure surface) are shown in a 1-2 space in
the following figure:
2
1
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
2
1
E1 E2 G12 12 1 2
Stiffness data
(x109 Pa) (x109 Pa) (x109 Pa) (m/m.oC) (m/m.oC)
Carbon-PEI
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
Stress analysis.
The stress analysis follows the procedure described in 5.4.
The laminate is built up from the same type of layer. The stiffness matrix of all
layers in their local coordinate system is therefore:
E1 21 E1
1 0
1 12 21
12 21
120.8 2.5 0
0 = 2.5 7.8 0 ( x 10 9 Pa )
E E2
[C ] = 21 1 (6.4)
1 12 21 1 12 21
0 0 G12
0 0 3 . 5
The stiffness matrix of each layer can then be rotated in their global coordinate
system. The values obtained for the different layer orientations are given in the
following table:
k *
C11 *
C 22 *
C12 *
C16 *
C 26 *
C66
0o 120.8 7.8 2.5 0 0 3.5
45o 36.9 36.9 29.9 28.2 28.2 30.9
90o 7.8 120.8 2.5 0 0 3.5
table 6. 2: Stiffness matrix components in the global coordinate system
The thickness of each layer is h/8. The laminate extensional stiffness matrix can be
written relative to the laminate thickness h:
0o 45 o 45 o 90 o
A11
= 14 120.8 + 14 36.9 + 14 36.9 + 14 7.8 = 50.6 x 10 9 Pa
h
A22
= 14 7.8 + 14 36.9 + 14 36.9 + 14 120.8 = 50.6 x 10 9 Pa
h
A12
= 14 2.5 + 14 29.9 + 14 29.9 + 14 2.5 = 16.2 x 10 9 Pa
h (6.5)
A16
= 14 0 + 14 28.2 + 14 (28.2) + 14 0 = 0 x 10 9 Pa
h
A26
= 14 0 + 14 28.2 + 14 (28.2) + 14 0 = 0 x 10 9 Pa
h
A66
= 14 3.5 + 14 30.9 + 14 30.9 + 14 3.5 = 17.2 x 10 9 Pa
h
Only in-plane forces are applied to the laminate and the bending stiffness matrix
[D] does not need to be evaluated. The laminate considered is symmetric, which
implies that the components of the coupling [B] stiffness matrix are all zero. The
[A] matrix can therefore be inverted directly to give an expression for the mid-plane
strain vector {0}.
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.9
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
1
1 50.6 16.2 0 0.022 0.007 0
[ A]
= h [a ] = 16.2 50.6 0
= 0.007 0.022 0 x 10 9 m 2 N
h (6.6)
0 0 17.2 0 0 0.058
The laminate strain vector is {0}=[a]{N} with {N}={hx 0 0}T, which gives:
x0 = a11 x h = 0.022 x 10 9 x
y0 = a 21 x h = 0.007 x 10 9 x (6.7)
0
xy = a 61 x h = 0
In-plane stresses can now be calculated for each layers. Expressions for these
stresses relative to the loading stress x are first given in their global coordinate
system:
1* 0 y0
= C11* x + C12* = C11* 0.022 x 10 9 C12* 0.007 x 10 9
x x x
2* * x
0
y0
= C 21 + C 22
*
= C 21
*
0.022 x 10 9 C 22
*
0.007 x 10 9 (6.8)
x x x
6* * x
0
y0
= C 61 + C 62
*
= C 61
*
0.022 x 10 9 C 62
*
0.007 x 10 9
x x x
This gives for the different layers:
0o 45 o 45 o 90 o
1*
x* 2.64 0.602 0.602 0.154
2 (6.9)
= 0.0004; 0.399; 0.399 ; 0.7906
x 0 0.423 0.423 0
6*
x
A useful check at this stage is to calculate the average of the stress vector
components over the thickness. For the loading case studied, the average of the 2*
and 6* components should be zero. The average of the 1* component should give
.
The stresses must now be rotated in their local coordinate system in order to apply
the failure criterion, i.e. {}k=[T]k{*}k
The transformation matrix are for the lay-up considered:
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.10
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
Failure criteria.
The Tsai-Hill criterion (6.3) can be written in relation to x:
12 1 2 22 s12 62 s12 s12
+ 2 2 + 2 2 = 2 (6.12)
x x x s2 x s6 x
2 2
And therefore:
12
12 1 2 22 s12 62 s12
x = s1 2 + 2 2 + 2 2 (6.13)
x s 2 x s6
2
x x
In the 0o layer, relation (6.11) shows tension along the fibres, hardly some tension
transverse to the fibres. In the modified Tsai-Hill relation, s1 is therefore replaced by
s1t and s2 by s2t:
12
2 2 s2 2 s2
x = s1t 12 1 2 2 + 22 12t + 62 12t
x x x s 2t x s6
12
(6.14)
1400 x 10 6
2
x = s1t (2.64) 2 (2.64)(0.0004) + (0.0004) 2
35 x 10
6
In the 45o layer, s1t and s2t can be filled in the equation:
2 2
2 1400 x 10 1400 x 10 6
6
x = s1t (0.924) (0.924)(0.078) + (0.078)
2
+ (0.1015) 2
35 x 10
6
60 x 10
6
(6.15)
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.11
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
The Tsai-Hill equation for the -45o layer is similar to the latest one.
In the 90o layer, the results given in (6.11) show compression in the 1-direction,
tension in the 2-direction requiring the use of s1c and s2t:
12
2
2 1200 x 10
6
x = s1c (0.7906) (0.7906)(0.1541) + (0.1541)
2
(6.15)
35 x 10
6
This gives:
0o 45 o 90 o
x x (6.16)
= 0.38; 0.25; = 0.19
s1t s1c
This means that failure will first occur in the 90o layer. This will occur at an
external stress x of:
x = 0.19 s1c = 0.19 (1200) = 228MPa (6.17)
Filling this value in relation (6.11) for the 90o layer shows that the failure occurs
transverse to the fibre.
Application of the maximum stress criterion shows that the first failure occurs in the
90o layer, transverse to the fibre direction (2-direction), at an external stress x of
2.27 x 108 N/m2. It is most plausible that the type of failure that occurs here is a
series of transverse cracks, which form planes perpendicular to the loading
direction. The stress profiles in the layers at this level of external stress are given in
the global coordinate system and the local coordinate system in respectively fig. 6.9
and fig. 6.10.
fig. 6.9: Stress profiles in the layer global coordinate system 1*, 2*, 3
-2 0 2 4 6 -2 0 2 4 -2 0 2 4
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
This will be illustrated with two examples. The first one illustrates the difficulty to
obtain values for the critical strength (or strain) values used in the different criteria.
The second will show the dependency of geometry on the results obtained with
mechanics of material criteria.
R
M
OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
[0/90n/0] laminate
The mode I is the opening mode, mode II the sliding mode, mode III the tearing
mode.
6.5. Problems
6.1 Use the example derived in 6.4. The stress at which the first ply fails has been
predicted at a level of 2.27 x 108 N/m2. This concerns the 90o plies. Calculate
the stress at which the next ply fails according to two different assumption:
a. All 90o ply stiffnesses are set to zero.
b. Since transverse cracking was the failure assumed to occur for this first
ply failure, set the elasticity moduli of the 90o layer related to the direction
perpendicular to the fibres to zero, i.e. E2 = G12 = 21 = 0
For this purpose, you need to re-evaluate the laminate extensional stiffness
matrix [A].
S I T ES G
Composites 2007
O
RO
M
CO
UP
Composite Production Methods
TE
U NI
EN
V
R
W
SI T
TY OF
Remko Akkerman
Laurent Warnet
Composites 2007
glass thermoplastic
carbon
...
thermoset
MEASURE heater
pump
water
cooling
CFRP
impregnation continuous fibre thermally induced distortions of single
curvature composites
reinforced plastic
distortion
prepreg measurement
MODEL
al
c
bri
on
fa
c ti
residual stresses
en
ire
production
ov
id
un
MEASURE
lightweight structure
micromechanical
inservice loading
stress analysis
Composites 2007
1
Composite fabrication techniques:
Impregnating a porous medium
Uncured Thermoset
+ pressure
Thermoplast above Tm
Composites 2007
Direct methods:
Direct impregnation
Composites 2007
Direct method:
Hand lay-up
Small series
Prototyping
large parts
Low control!
Composites 2007
2
Direct method:
Filament winding
Thermoset
Low-Medium series
Large parts
Oven for curing
Composites 2007
Direct method:
Filament winding:
Composites 2007
Direct method:
Pultrusion / Pull-winding / Pull-braiding
Thermoset
Continuous process!
Medium tooling costs
Medium-high performance
Medium-Large series
Small-medium cross-sections
Composites 2007
3
Direct method:
Pultrusion / Pull-winding / Pull-braiding
Composites 2007
Semi-Direct methods:
Direct impregnation of a dry preform
Composites 2007
Semi-direct method:
Liquid Composite Moulding (LCM)
Several variants:
Composites 2007
4
Semi-direct method:
Resin Transfer Moulding (RTM)
Medium series
Small medium size
Composites 2007
Semi-direct method:
Resin Transfer Moulding (RTM)
Composites 2007
Composites 2007
5
Semi-direct method:
Resin Infusion under Flexible Tooling (RIFT)
Bagging film
Sealant tape
Resin inlet
Medium series
Composites 2007 Medium Large size
Semi-direct method:
Resin Infusion under Flexible Tooling (RIFT)
Composites 2007
Semi-direct method:
Liquid Composite Moulding (LCM)
Composites 2007
6
Semi-direct method:
Liquid Composite Moulding (LCM)
Viscosity
Permeability=f(Fibre structure /
orientation / concentration)
K P
Darcys law: Q= (Macro)
x
Composites 2007
Semi-direct method:
Liquid Composite Moulding (LCM)
Flow on a mesoscale:
Composites 2007
Semi-direct method:
Liquid Composite Moulding (LCM)
Composites 2007
7
Semi-direct method:
Liquid Composite Moulding (LCM)
Common characteristics:
Function integration:
Composites 2007
Multi-step methods:
Composite processing based on prepreg
Fibre Yarn
/ Fabric Prepreg
Product
+ Resin
Semi-Product
Product
On basis of preimpregnated layer
Composites 2007
Multi-step methods:
Autoclave curing
Composites 2007
8
Multi-step methods:
Autoclave curing
Constraints on prepreg:
Flexibility
Drapability
Tack
Thermoset
Composites 2007
Multi-step methods:
Autoclave curing
Vacuum:
Withdrawing volatiles
Composites 2007
Multi-step methods:
Rubber pressing
Needs semi-product
Thermoplastic matrix
Low-medium labour costs
High tooling costs
High performance
Short cycle time
Small-medium products
Composites 2007
9
Multi-step methods:
Tape placement
Tailor-made
Thermoplastic
Thermoset (+autoclave)
Low labour costs
Medium-high tooling costs
High performance
Medium cycle time
Medium Large products
Composites 2007
Fibre reorientation
S S
S
S
S S S
S S
S S
S
Composites 2007
Composites 2007
10
Choice of a fabrication techniques:
Parameters playing a role:
Composites 2007
Direct methods:
Hand lay-up
Hand lay-up Productivity
5
4
Series size 3 Part size / cost
0
Function Complex
integration geometry
Technical
Performance
Simplicity
Composites 2007
Direct methods:
HandWinding
Filament lay-up Productivity
5
4
Series size 3 Part size / cost
0
Function Complex
integration geometry
Technical
Performance
Simplicity
Composites 2007
11
Direct methods:
Hand lay-up
Pull-Processes Productivity
5
4
Series size 3 Part size / cost
0
Continuous Process Function Complex
Thermoset integration geometry
Tailor-made lay-up
Technical
Performance
Simplicity
Composites 2007
Semi-direct methods:
Hand lay-up
Resin Transfer Moulding Productivity
5
4
Series size 3 Part size / cost
0
Function Complex
integration geometry
Only for Thermosets
Technical
Performance
Simplicity
Composites 2007
Semi-direct methods:
Hand
Resin Infusion lay-up
Flexible Tooling Productivity
5
Vacuum outlet
4
Bagging film Series size 3 Part size / cost
Sealant tape
Resin inlet
2
Technical
Performance
Simplicity
Composites 2007
12
Multi-step methods:
Hand lay-up
Autoclave Productivity
5
4
Series size 3 Part size / cost
Technical
Performance
Simplicity
Composites 2007
Multi-step methods:
Hand
Rubber lay-up
Pressing Productivity
5
4
Series size 3 Part size / cost
Technical
Performance
Simplicity
Composites 2007
Multi-step methods:
Hand lay-up
Tape layer Productivity
5
4
Series size 3 Part size / cost
Technical
Performance
Simplicity
Composites 2007
13
Composites Course University of Twente, Eng&Tech 8.1
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
8.1 Introduction
The CLT (Classical Lamination Theory, see Ch.4) provides a means for quick analyses of
composite laminate stiffness and strength. For complicated geometries, three-dimensional
stress analyses or large deflections one often reverts to finite element simulations. This lecture
will give a survey of the basics and some applications.
1
2
= ij ij d ui f j d v j t j d = (7.1)
,
1
2
ij Cijkl kl d ui f j d v j t j d
in which is the total volume and the boundary surface of the continuum. The first
volume integral of equation (7.1) contains the stiffness matrix and the displacement field of
the system, whereas the second represents the volume integral of the body forces per unit
mass, which is supposed to be zero. Finally, the integral over the boundary surface contains
the applied forces. In terms of virtual work / minimum potential energy this leads to
1
2
= ij ij d ui f j d v j t j d = 0 . (7.2)
kl =
1
( uk ,l + ul ,k ). (7.3)
2
or
(7.5)
14442444
3
K
1
N.B. Both FE and CLT literature use B as a de facto standard term. Here, the choice was
made to use [B] for the FE matrix with derivatives of the shape functions, and B for the CLT
(strain-curvature coupling) matrix.
Composites Course University of Twente, Eng&Tech 8.2
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
in which [D] is the material stiffness matrix and {f} contains the externally applied forces. In
simplified form this is the familiar elastic equilibrium equation for a single element:
[K ]e {u}e = {R}e . (7.6)
The stiffness matrix for the whole system can be obtained by recognising the displacements
are continuous over the element boundaries. Adding all element contributions for a node n
leads the nodal force equilibrium,
m (7.7)
{F} n = {R} ne ,
c =1
in which {F}n is the applied nodal force and {R}en is the contribution of element e to the
reaction force in node n, where m is the number of elements that belong to node n.
Substituting equation (7.7) into equation (7.6) results in:
m (7.8)
{F }n = [K ]ne {u}ne .
c =1
Assembly of equation (7.8) for every node (or, usually, for every element) results in the
equilibrium equation for the entire system,
Often, elements with 3 displacement and 3 rotation degrees of freedom per node will be used.
The inplane displacement components will be used to determine the midplane strains, the
inplane rotations (and sometimes the out-of-plane displacement) will be used to determine the
curvatures,
xx u0, x xx y , x
{ o } = yy = v0, y = yy = x , y (7.10)
u + v +
xy 0, x 0, y xy y , y x ,x
.
With those, the strain distribution follows from
{ (z ) } = { 0 }+ z { } , (7.11)
which can be substituted into the volume integral in (7.2), split into an integral over the
element surface and the element thickness,
=
1
2Ah {
0 + z } . {[Q ] { 0 + z }}dz dA vi ti d .
T
(7.12)
For the layered composite elements, [Q] is the stiffness matrix according to CLT, A the
elements surface, h the element thickness and z the thickness co-ordinate. Evaluating this
equation results in:
2 3 2
1
{ 0 }T [Q ] { 0 }dz + { 0 }T h [Q ] { }dz + { }T h
[Q ] { }dz + { } [Q ] { 0 }dz dA
h
2 A
= h{
T
where the virtual strains and curvatures are related to the virtual displacements and rotations
by means of the finite element [B]-matrix,
{ }= [B ] {u},
0
u
{ }= [B ] { }.
0
(7.14)
T A B d
[B ]
A B D
[B ] dA = {R } ,
d b
(7.15)
where we recognise the familiar ABD-matrix. Further, [B] contains both [Bu] and [B], and
{dd} and {d} contain the displacement and rotation degrees of freedom, respectively. Note
that in this way the resulting forces and moments over an element are obtained. Equation
(7.15) is usually solved numerically using Gauss quadrature. As in the CLT, the stiffnesses are
integrated over the thickness of an element. Hence only the mid-plane displacements and
rotations serve as degrees of freedom, keeping the FE-system matrix to a very moderate size,
regardless of the number of layers in the laminate.
1 12 12 0 0 0
E1 E1 E1
1 23 1
12 E 1 0 0 0
E2 E2
2 1
23 2
3 12 1 0 0 0 3
= E1 E2 E2
(7.16)
0
23 0 0 1 0 0 23
13 G23
13
0 0 0 0 1 0
12 G12
12
0 0 0 0 0 1
G12
or, symbolically,
{ 123 } = [S ] { 123 } { 123 } = [S ]1 { 123 } [Q ] { 123 } (7.17)
Transformed to the laminate xyz-coordinates (due to an inplane rotation ) we find
{ }= [T ] [Q ] [R] [T ] [R] { } [Q ] { },
xyz
1 1
xyz
*
xyz (7.18)
where the transformation matrix [T] is applied, given by
cos 2 ( ) sin 2 ( ) 0 0 0 2 sin ( ) cos( )
sin ( ) cos 2 ( ) 2 sin ( ) cos( )
2
0 0 0
[T ] = 0 0 1 0 0 0
(7.19)
0 0 0 cos( ) sin ( ) 0
0 0 0 sin ( ) cos( ) 0
sin ( ) cos( ) sin ( ) cos( ) 0 0 0 cos 2 ( ) sin 2 ( )
Hence, the three-dimensional off-axis stress-strain relation for a unidirectional layer is given
by
(i )
x (i ) Q11* Q12* Q13* 0 0 Q16* x
(i ) * *
y Q12 Q22
* *
Q23 0 0 Q26 (yi )
z Q13* Q23 * *
Q33 0 *
0 Q36 z
=
(i ) , (7.21)
yz 0 0 0 Q44*
Q45 0 yz
*
(i )
xz 0 0 0 Q45* *
Q55 0 xz
(i ) *
xy Q16 Q26 xy
* * *
Q36 0 0 Q66
Composites Course University of Twente, Eng&Tech 8.6
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
where the transformed stiffness matrix components Qij* are all a linear combination of sin(k),
cos(k) with k=2,4 and as the rotation with respect to the lamina principal axes. The
components denoted with superscript (i) can be discontinous functions across the layer
boundaries. The components z, xz, yz, x, y and xy will be continuous functions across the
thickness of the laminate.
A possible approach in a three-dimensional element can be to bring all layer variables to the
left hand side by matrix manipulation,
(i ) (i )
x M 11* M 12* M 13* 0 0 M 16* x
* * * *
y M 12 M 22 M 23 0 0 M 26 y
z M 31
* *
M 32 *
M 33 0 0 *
M 36 z
= (7.22)
yz 0 yz
* *
0 0 0 M 44 M 45
xz 0 0 0 *
M 54 *
M 55 0 xz
*
xy M 16
*
M 26 *
M 63 0 0 *
M 66 xy
and subsequently add (or integrate) the layer variables over the thickness in order to find the
element stiffness matrix. Alternatively, an effective transverse modulus can be computed by
inverting an averaged compliance, as in
H
E z = n (i )
eff
h
(7.23)
(i )
i =1 E z
which is then directly applied in a layer stiffness contribution as in (7.21), which can
subsequently simply be added to the full laminate stiffness [1]. A detailed discussion on
transverse strains can be found in [3].
The main conclusion is that certain approximations have to be made to find a reasonable
description of the transverse behaviour within acceptable CPU-time. Users should perform
patch tests to get an indication of the accuracy of the method employed in their specific
application.
8.2.3 Nonlinearit y
Sofar, only linear elastic behaviour has been discussed. The thin-walled composite structures
may, however, easily experience deflections far greater than the laminate thicknesses. In this
case, the relation between strains and displacements easily surpasses the linear region, and
geometric nonlinearity (GNL) has to be taken into account.
Consider the large deflections in the following beam due to bending:
x
x0
x
x = + , (7.25)
x 2 x
which can be easily verified by using Pythagoras theorem and subsequent linearisation.
GNL also plays an important role in buckling simulations, where bending is induced by
inplane loads. Consider for instance a simulation of a compression test of a rectangular plate.
The first and second buckling modes are depicted in Figure 7.3
Figure 7.3 The first two buckling modes for a simply supported plate under inplane compressive load.
Both modes of buckling affect the inplane stiffness significantly. The inplane load-
displacement predictions are depicted in Figure 7.4.
Composites Course University of Twente, Eng&Tech 8.8
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
20
in plane displacement (mm)
0
-1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2
-20
Mode 1
load (kN)
-40 Mode 2
straight
-60
-80
-100
Figure 7.4 Inplane loads versus deflection for the inplane compression test.
Further nonlinearity can arise from the material behaviour. During laminate production, the
matrix will turn from a fairly viscous state to approximately an elastic solid. Special purpose
material models are under development to capture these types of (physical) nonlinearity.
Transverse to the fibres the material behaviour is matrix-dominated. For certain matrix
materials, this can result in creep, another physically nonlinear phenomenon. Further
nonlinearities are of course found during damage growth.
proper analysis must, however, give convergent results with mesh refinement (to be checked
by the user).
Similar approaches are in use for fatigue analyses. For regular loading, the so-called Paris
Law can be applicable, relating the delamination growth rate da/dN [mm/cycle] to
da
cG n (7.28)
dN
with a as the crack length, N the number of load cycles, G as the maximum cyclic energy
release rate and c,n constants. An example of experimental results under mixed mode bending
is given below.
Composites Course University of Twente, Eng&Tech 8.10
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
Figure 7.5 Paris Law diagram for APC-2 from a Mixed Mode Bending test, log(c) = -3.59, n = 3.58
Knowing the appropriate critical energy relations, arbitrary load cases can be simulated. Up to
the year 2000, the applicability of these laws for random loading appears fairly limited,
however.
One step further is the use of special interface elements (with zero thickness). Generally, these
are inserted on a predetermined possible crack plane (e.g. the interface between two plies in a
laminate). Up to a certain load, the elements are very stiff, preventing separation of the solid
surfaces attached. After a maximum stress (the strength), the element opens, according to
some softening curve, meanwhile dissipating the critical energy release as found from fracture
mechanics experiments.
Composites Course University of Twente, Eng&Tech 8.11
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
Figure 7.6 Stress-displacement curve for the LUSAS interface elements for delamination growth simulations [8].
8.4 Conclusion
The Finite Element Method offers significantly more possibilities for stress analysis of
composite materials than simple laminate analysis. However, the complexity caused by
anisotropy and orientations, and the corresponding amount of input data required, imply that
such an analysis is by no means a trivial exercise and all results should be critically verified
before drawing conclusions.
8.5 References
[1] A.H. van den Boogaard, Aanvulling en handleiding bij Stijfheid en Sterkte 3 (115711), Universiteit
Twente, juli 1999.
[2] ANSYS Theory Reference, 000855 Eighth Ed. SAS IP, Inc
SOLID46 3-D Layered Structural Solid
[3] A.K. Noor, M. Malik, An assessment of five modelling approaches for thermo-mechanical stress
analyses of laminated composite panels, Computational Mechanics, vol.25, p.43-58 (2000).
[4] H. Thom, A review of the biaxial strength of fibre-reinforced plastics, Composites Part A, vol.29A,
p.869-886 (1998).
[5] Chang, F.K & Chang, K.Y., A progressive damage model for laminated composites containing stress
concentrations, Journal of Composite Materials, vol.21, p.834-855 (1987).
[6] Sethuraman, R. & Maiti, S.K., Finite element based computation of strain energy release rate by
modified crack closure integral, Engineering fracture mechanics vol.30/2, p.227-231 (1988).
[7] Raju, I.S., Calculation of strain-energy release rates with higher order and singular finite elements,
Engineering fracture mechanics, vol.28/3, p.251-274 (1987).
[8] FEA Ltd., LUSAS Modeller User Guide, v.13.0 (1998).
Composites Course University of Twente, Eng&Tech 8.12
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
A8.1 Approximation
A series expansion is a usual approximation of a function f(x):
f% ( x ) = ci H i ( x ) (A7.1)
i
with a set of constants {ci} and a set of basic functions {Hi(x)}, e.g. polynomials, sine series
etc.
In the special case of basic functions Ni(x) which are piecewise linear functions with a local
support, satisfying Ni(xj)=ij, the approximation is equivalent to linear interpolation:
xi +1 x x xi
f% ( x ) = f ( xi ) + f ( xi +1 )
xi +1 xi xi +1 xi xi x xi +1 (A7.2)
N i ( x ) f i + N i +1 ( x ) f i +1
In FE terms this approximation employs the shape functions Ni(x). As Ni(xj)=ij, this
representation automatically satisfies fi=f(xi).
f (x)
linear interpolation
continuous function
shape functions
1
Ni(x) Ni+1(x)
x
Figure A7.1 Piecewise linear approximation of a continuous function f(x).
Composites Course University of Twente, Eng&Tech 8.13
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.
A8.2 Differentiation
Differentiating the piecewise linear approximation in space leads to a constant value per
element [xi,xi+1]:
d % f ( xi +1 ) f ( xi )
f ( x) =
dx xi +1 xi
xi x xi +1 (A7.3)
d Ni ( x) d N i +1 ( x )
= fi + f i +1
dx dx
or, in general, expressed in the derivatives of the basic functions,
d % d Hi ( x)
f ( x ) = ci (A7.4)
dx i dx
In words: if the derivatives of the shape functions (in the global coordinates) are known, the
(approximated) derivative of the function can be determined easily, with the same set of
constants {ci} as the function approximation itself.
N.B. A linear element with displacement degrees of freedom will have a constant strain per
element and hence also a constant stress.
A8.3 Integration
An integral over a full domain can be split into element contributions:
xn n xi
f ( x ) dx =
x0 i xi 1
f ( x ) dx (A7.5)
f ( x ) dx W j f ( x j ) (A7.6)
x j
using a set of weight factors {Wj}, such as Gauss quadrature for which
m
W p ( x ) = p ( x ) dx
j
j k j k (A7.7)
x
is satisfied exactly for polynomials pk(x) of the degree k=2m-1 (in words: with two integration
points, m=2, a third order polynomial p3(x) can be integrated exactly). The weight factors
{Wj} and the specific locations, the integration points {xj} follow from this requirement.
For our linear approximation one integration point is sufficient, which is located in the centre
of the element, leading to
( )
xi
f% ( x ) dx = ( xi xi 1 ) f% xi 1 (A7.8)
2
xi 1
f (x)
integral of the
linear approximation
x
Figure A7.2 1-point Gauss quadrature of the linear approximation of a continuous function f(x).
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 9.1
R
M
OU
CO
9. Interlaminar stresses - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
9. Interlaminar stresses
9.1. Introduction
In the first lectures, the analysis of stresses and strains in a layer of a composite
laminate was performed assuming a state of plane stress (3= 4= 5=0). This
assumption is often valid, as long as the analysis is performed far away from
geometric discontinuities, like free-edges or holes. Close to these discontinuities, a
3D state of stress exists. Out-of-plane stresses, or interlaminar stresses, cause
delamination. Delamination is one of the ways composite laminates fail, where
different layers are separated.
It is not the intention to present here the complete 3D stress analysis, but to give
an insight into the mechanisms causing the interlaminar stresses. The interlaminar
stresses induced at the free edges of a laminate will be used as an illustration. This
will help critically considering the lay-up of a laminate.
R
M
OU
CO
9. Interlaminar stresses - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF
not any longer negligible is often called the boundary layer width. It was shown in
many publications that this width is in the order of magnitude of a laminate
thickness at most. A finite analysis result of the cross-ply laminated strip
considered in fig. 8.1. also shows the deformations involved at the edges.
y
y/b
0 1
fig. 8.2: Interlaminar normal stress near fig. 8. 3: Deformations at the edge of the cross-ply
the edge. strip considered in fig. 8.1
R
M
OU
CO
9. Interlaminar stresses - Laurent Warnet & Remko Akkerman.
P
TE
UN I
EN
VE
R
W
SI T
TY OF