You are on page 1of 91

Composite Mechanics Handouts 2008.

Table of content:

1- Introduction: the constituents

2- Basic Micromechanics

3- Stress-Strain relationship within a layer

4- Classical Lamination Theory

5- CLT with thermal terms

6- Laminate failure analysis

7- Fabrication techniques

8- FEM for composites

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. Constituents and manufacturing techniques

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)

fig. 1.1: Example of a laminate.

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.

1.2.1. Glass fibre


Glass fibre is the most popular type of fibre in the composite world. Glass fibre is
stronger than metals in bulk form by virtue of the fact that its structure contains
fewer flaws or defects. The ultimate tensile strength of a freshly drawn single
glass filament (diameter 9-15 m) is about 3.5 GPa. Fibreglass filaments became
commercially available in 1935. Glass reinforced polyesters were used in the
construction of aircraft radomes in the early 1940s. In the 1950s these materials
were applied widely in plastic boat hulls, car bodies and truck cabs.
All glass fibres are based on Si-oxydes. Addition of other oxides, like, B, Al, Na,
K, Ca and Mg for different applications. Production of glass fibres starts from
melt around 1300-1500C (higher for S-glass). The molten glass is driven by
gravity through small orifices. It is rapidly cooled down and pulled at high
velocity (65m/s) to form a fibre. Surface micro-damages, which have a negative
effect on strength, are induced during processing. A rubber roller applies a
subsequent coating with a protective aqueous size solution in order to reduce the
effect of surface micro-damages. This treatment also contains a coupling agent,
which promotes the chemical compatibility of the matrix and the fibre. The
diameter of the obtained glass fibre (for reinforcement) is typically 9 to 15m.
The internal structure of glass is a 3D long network of silicon, oxygen and other
atoms arranged in random fashion. Glass fibres are therefore amorphous and
isotropic.
The most used fibre is E-glass, originally developed for electronic applications
(good isolation). Others are C-glass, which has better chemical resistance. S-glass
has a higher strength, but is costly to produce. D-glass is an example of a tailor
made fibre, and it finds applications in radars.
Summarising in a few words:
- The most popular
- Isotropic properties
- Elasticity modulus lower than carbon or aramid.
- Brittle fibre, its surface contains lots of microdefects.

1.2.2. Carbon (and Graphite) fibre


Edison produced the first carbon fibres (1879) for electric lamps, by careful
carbonisation of cellulose strands. Bamboo or cotton was used as a starting
material, leading to extremely brittle filaments. The second-generation carbon
fibres (from 1963 onwards) are miles apart from their predecessors and have
exceptionally high strength and stiffness. Carbon fibres are based on
polyacrylonitrile (PAN) or pitch (by-product of petroleum refining). The PAN
precursor is stretched to obtain molecular and structural orientation.

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

The fibre is obtained by


letting the carbon-containing
precursor fibre going through
different thermal treatments:
oxidisation at around 200-
400C in air, carbonisation in
N2 around 1000C (carbon
fibre), and possibly
graphitisation in Ar from fig. 1.2: Microstructure of carbon fibres on basis of PAN
1500C (graphite fibre).
Carbon fibres have a carbon
content of 95%, graphite fibres at least 99%. The fibre can be subsequently
oxidised to promote adhesion to matrix resins, and coated with a sizing resin to
protect them during handling. The typical diameter of a carbon fibre is around 5 to
7m. It is important to note that the structure obtained is a mix of amorphous
carbon and graphitic carbon. This last form where the atoms are arranged in
parallel plan provides the high tensile modulus. A strong covalent bond exists
within these planes, but the planes are held together with weak van der Waals-type
forces (see fig. 1.2). This results in highly anisotropic properties.
Summarising in a few words:
- High specific elasticity modulus
- Anisotropic properties
- Negative coefficient of thermal expansion
- Electrically conductive
- Structures made with carbon are generally sensitive to impact and
therefore used in combination with aramid fibre.

1.2.3. Aramid fibre


The aramid fibre is one of the organic fibres. Aramids are highly crystalline
aromatic polyamides. The aramid fibre (Kevlar from DuPont as introduced in the
early 1970s, Twaron from Akzo, now Teijin) is obtained by spinning the polymer
into thin threads, which are then stretched to increase the stiffness. The stretching
results in a high molecular orientation in the longitudinal direction. Their
properties are therefore highly anisotropic. The fibre diameter is around 12m.
Summarising in a few words:
- Low density (1450kg/m3) and high specific strength.
- Not recommended in structures with high compressive loads.
(compression strength is 20% of tensile strength)
- Difficult to process
- Good thermal stability (decomposition at 450oC)
- Ductile fibre. Often used in applications where good energy
absorption is necessary.

1.2.4. Polyethylene fibre


The PE fibre is a second example of an organic fibre. Ultra-high molecular weight
polyethylene is dissolved in a solvent and then spun through small orifices
(spinneret). Successively, the spun solution is solidified by cooling, which fixes a
molecular structure which contains a very low entanglement density of molecular

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)

1.2.5. A few fibre properties


The table below presents some typical properties of selected fibres. The properties
can vary from one type to another, depending on the actual manufacturing
process.

Fibre Typical Density Youngs Tensile Strain to Coeff. of


Diameter Modulus Strength Failure Thermal
Expansion
(m) (kg/m3) EfL/EfT (GPa) max(GPa) max (%) (10-6/C)

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

table 1.1: Typical fibre properties

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.

1.3. Matrix materials


No matter in what form the fibre is used, it is hardly possible to use it as a single
component in any structure. A matrix is necessary to
1) transfer the load to the fibres and
2) protect the fibre from the environment (chemical-mechanical
aggression).
The matrix plays a minor role in the load carrying capacity of the composite in the
fibre direction. A typical elasticity modulus of an average matrix material
compared with that of an average Carbon fibre is 3GPa against 230GPa. However,
the matrix plays a major role for the in-plane shear properties of the laminate
(important for torsion) as well as its interlaminar shear properties (important for
bending properties). In addition, the physical interaction between the fibre and the
matrix (mostly called interface) plays an important role in the transfer loading as
well as for the way the laminate absorbs energy.
In general three types of plastics are distinguished:
1. Thermoplastics are linear or branched macromolecules, which are held
together by weak secondary bonding forces. They soften when they are
heated.
2. Thermosets are cross linked by many valency bonds, leading to a rigid
(and unsoluble) three-dimensional network which will not be affected by
heat until the temperature is high enough to break up the molecule. Usually
the formation of such network polymers requires heat.
3. In elastomers the molecules are linked together by a small number of
valency bonds, leading to a loose network behaving as a rubber-like
material with more elastic properties than a simple thermoplastic.

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.

The polymer synthesis reactions are based on combining reactive (functional)


groups. The functionality of a molecule is defined as the number of reactive
groups. A functionality of more than 1 is required to produce polymers. Appendix
A summarises a number of molecular groups in polymer matrix materials.

The synthesis reactions of polymers can be classified as:


1. Polymerisation: a chemical reaction, generally carried out in the presence
of a catalyst, which combines small molecules (monomers), containing a
double bond, into long chain molecules. The double bond is opened up,
thereby making valency bonds available for linking with its neighbouring
monomer molecule. No by-products are produced.

Example: polyethylene

H H H H

n C C C C

H H H H
n

2. Polycondensation: a chemical reaction between two similar or dissimilar


basic units which have at least two functional groups. It gives rise to the
elimination of small, low molecular weight by-products such as water,
hydrochloric acid, etc.

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

Example: linear polyester


OH OH

(n+1) HO (CH2)2 OH + n C (CH2)4 C

O O

HO (CH2)2 O C (CH2)4 C (CH2)2 OH + n H 2O

O O n

3. Polyaddition: two different types of molecules with reactive groups are


brought together. No by-products are eliminated, but hydrogen atoms
migrate from their positions in the functional group leaving combinable
free valencies.

Example: epoxide resin


R OH
O

R H + C C C C

The typical fibre reinforced plastics have a thermoset or a thermoplastic matrix


(we will leave elastomers aside, although car tyres and reinforced rubber tubing
meet the definition of a polymer matrix composite). Thermosets are mostly
synthesized in a multistep process of subsequent condensation and addition
reactions. The formation of a cross-linked 3D molecular network requires more
than two functional groups in at least one of the reactants. A major difference with
thermoplastics is that once solidified, subsequent heating cannot reshape the
thermosetting matrix. Thermoplastics can be repeatedly softened by increasing the
temperature and hardened by decreasing the temperature. The next points put the
emphasis on some advantages/disadvantages of the two types of matrix.
Melt viscosity: Very low for the thermosets, high for most thermoplastics.
Low viscosity is important for a good fibre impregnation. The use of
thermoplastics therefore requires fabrication methods designed specifically for
this purpose.
Processing cycles: Longer for thermosets than for thermoplastics. The final
hardening of a polyester based composite can take days or weeks.
Fracture toughness: Higher for thermoplastics. This item was, together with a
potential reduction in processing cycle, the argument to investigate the use of

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

thermoplastics as polymer matrix materials. Thermosets have improved, due to


combination with other phases (rubber, thermoplastics).
Fabrication costs: Depends on the type of composites with thermosets (from
high in aerospace application to low in for example glass fibre pipes or tanks).
Are potentially low for thermoplastic. However, efficient production
techniques still need to be developed.
Chemistry: Complex for the thermosets. Reproducibility in material properties
is therefore difficult. Comparably straightforward for thermoplastics
Residual stresses: Residual stresses are inherent to any composite structures.
In thermosets, their origin is three-fold: thermally, hygroscopically (absorption
of water) or chemically (when cross-linking) induced. No chemical shrinkage
occurs when producing with thermoplastics, but crystallisation can play a role
in semi-crystalline materials.

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

4 CH2 CH R CH CH2 + H2N R' H2N

O OH OH O

CH2 CH R CH CH2 CH2 CH R CH CH2

N R' N

CH2 CH R CH CH2 CH2 CH R CH CH2

O OH OH O

A functionality greater than 2 in the reactant enables the synthesis of a cross-


linked network, when the parent resin is such a diepoxide.

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

cross-linking of a polyester-styrene network

temperature, leading to a storage life of several months or even years. Heating


up starts the process of cross-linking. During this curing reaction the styrene
reacts with the carbon-carbon double bonds of the polyesters to build up a
three-dimensional (cross-linked) polyester network. Accelerators can also
induce cure at room temperature. As for the epoxy resins, the properties of
polyesters depend on the cross-link density; they are available in a wide range.
Unsaturated polyesters are often applied in hand lay-up, sheet and bulk
moulding compounds (SMC and BMC respectively). Advantages compared to
epoxies are their low viscosity, fast curing time and lower cost. However, their
material properties are generally lower. The major drawback is their chemical
shrinkage. The volumetric shrinkage of polyester during the curing process
can be up to 8%. Apart from this chemical shrinkage, thermal shrinkage can
occur when heat is used during curing. It is worth adding that even for cold-
cure resin, the curing process can produce heat and therefore can induce extra
local shrinkage of the matrix. In recent years, the emission of styrenes during
processing puts a heavy restraint on processing polyesters.

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

Epoxy Vinyl Ester Resins


EVERs (or Vinyl Esters) consist of an epoxy backbone with vinylester
end groups.
epoxy backbone
ester group
vinyl group O CH3 O

CH2 O CH2 CH CH2 O C O CH2 CH CH2 O C C CH2


C C

CH3 OH CH3 n OH CH3

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%.

Further thermoset resins


Phenolics were developed in the early 20th century. These brittle materials are
suitable for an elevated temparture range (up to 200C) and offer excellent
flame retardance. Bismaleimides and other thermoset polyimides offer again a
higher thermal resistance: up to 370C continuous temperature for certain
Polymerised Monomer Reactants. Cyanate esters also offer a high temperature
resistance, in addition to low moisture absorption and excellent hot-wet
performance.

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.

Polypropylene is the most-used thermoplastic matrix material, in combination


with long (50mm) or continuous glass fibres, in Glass Mat Thermoplastic
(GMT) and commingled composites respectively. The PP/glass composites
can mostly be found in automotive applications.

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.

Polyetheretherketone is another semicrystalline thermoplastic, in this case with an


outstanding thermal stability. This thermoplastic, reinforced with carbon, can
be regarded as the prime high performance thermoplastic composite,
introduced as APC-2 (Aromatic Polymer Composite) by ICI-Fiberite in the
late 1980s.
O

O O C

PEEK has a density of 1300 kg/m3. It has a glass transition temperature of


143C and a melting temperature of 334C. It can be safely used up to 250C.
Suitable forming temperatures are around 400C. The strength and solvent
resistance are excellent. However, the material is notoriously expensive.

Polyphenylenesulfide is a third semicrystalline thermoplastic applied in composite


materials. Reinforced with glass or carbon it is applied in aerospace
applications such as the Airbus A340 and the new A380 for structural parts.

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.

Polyetherimide is an amorphous thermoplastic in the high temperature range (as


other polyimide resins). Due to its low flammability with low smoke emission
it is very suitable for aerospace interior parts, reinforced with glass or carbon.
Being an amorphous material, PEI is more susceptible to solvents than its
semicrystalline counterparts. On the other hand, the fracture toughness is very
good, certainly when compared to PPS.

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.

Density Youngs Tensile stress Tensile strain Coeff. of


Modulus Thermal
Matrix max(MPa) max (%) Expansion
(kg/m3) E (GPa) (10-6/C)
Yield Failure Yield Failure

Epoxy 1150-1200 3.5 50 - 70 24 90


80 + 1406 1150-1200 3.5 60 - 80 35 60
120 + 160 1200-1300 3.5 80 - 90 45 65

Polyester 1150-1250 2.4 4.6 40 - 85 1.2 4.5 80 150

PEEK 1320 3.6 100 90 5 35 45


PPS 1370 3.7 80 50
PEI 1270 3 105 85 6 60 55
PES 1370

table 1.2: Properties of a selection of polymeric materials.

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

1.4. Manufacturing processes


With any material, the manufacturing process can greatly influence the
performance of the obtained structure. Think for example about molecular
orientation in injection moulded plastic parts. For composites, the fabrication
technique will influence the fibre content, entrapment of air bubbles (voids),
degree of crystallinity for semi-crystalline thermoplastics, positioning of the
reinforcement
Different basic manufacturing techniques will be explained briefly. As pointed out
earlier, the method used depends to a great extent on the type of matrix used, with
in general a large difference in viscosity between the uncured thermosetting resin
and the (highly viscous) thermoplastic melt. The choice of a fabrication technique
will also depend on the shape complexity, size and amount of product to be
fabricated, and the performance required.

1.4.1. Reinforcement architectures


The form in which the reinforcement is applied can vary depending on the
performance required. Only the reinforcement forms used for laminates will be
described here. Most reinforcements are based on rovings (bundles) of continuous
fibres. They can be aligned in an unidirectional way, or weaved to form a cloth
which can have different patterns as shown in fig. 1.4. Weaving techniques also
allow 3D types of reinforcement. Continuous fibres can also be randomly aligned
(think of a plate of spaghetti). Chopped fibres are generally used in high-volume
applications.

Unidirectional Plain weave Twill 2x2 Satin 8

Basic triaxial Basic basket Random Chopped


weave Weave

fig. 1.4: Different reinforcement architectures

It is worth noting that the impregnation of the fibres is often deliberately


performed before the composite is put into shape. For unidirectional or woven

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

fabric, this prepackage is called a prepreg and is generally aimed at high


quality-high fibre content applications. A similar technique is used for chopped
fibre with unsaturated polyester: Sheet Moulding Compound and Bulk Moulding
Compound (SMC and BMC). Preimpregnation reduces manual procedures and
improves for example the fibre distribution. In the case of thermosets, the resin is
partially cured and the prepreg needs to be stored at low temperature and used
fairly rapidly.

1.4.2. Hand laminating


This easy and moreover flexible way uses
an open mould, and is used for example for
prototype fabrication or when producing large
components (boat). Both resin and
reinforcement are applied by hand using
simple tools as shown in fig. 1.5. Cold cure
thermosets are mostly used with this
technique. Reinforcement forms are woven,
random and chopped. Stiffening ribs, inserts,
and core material can be fitted during the fig. 1.5: Hand laminating
laminate build-up. An alternative to the hand
lay-up is to spray a mix of resin and chopped fibres. Emission of styrenes or other
toxic gases more and more limit the use of hand lamination in the composite
industry.

1.4.3. Filament winding


Another open mould technique to
produce simple hollow shapes
(tubes, pressure vessels). The
negative mould is rotating while a
fibre guide moves parallel to the
mandrel longitudinal axis. The
fibre is impregnated in
thermosetting resin before it is
guided to the mandrel through a
guide. The product is mostly cured
in an autoclave after winding.
It is worth adding that some
applications are based on
thermoplastic matrix. In this case, a
prepreg tape is used.

fig. 1.6: Filament winding

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

1.4.4. Vacuum bag, pressure bag and


autoclave
Two closed mould techniques,
which are aimed to improve the
hand lay-up technique. Resin and
reinforcement are still applied by
hand in a simple open mould. Then
vacuum is applied via a rubber bag.
This induces a processing pressure
of 1 bar, whilst reducing air
entrapment. This technique is also
used for bonding sandwich cores to
existing laminates.
The pressure bag technique is
similar, but allows to apply
pressures up to 3,5 bar. This is fig. 1.7: Autoclave technique
favourable for the consolidation
and for fibre content. This puts higher demands on the stiffness of the mould. The
combination of pressure with the possibility of adding temperature to the process
makes it possible to use prepregs. This reduces the hand lay-up step.
The autoclave technique combines earlier named bag techniques. The vaccum bag
assembly is then used within a pressurised vessel (up to 10bar). Both vacuum and
pressure ensure a high quality product.

1.4.5. Hot press, Stamping,


Rubber press
On basis of prepreg / SMC,
hot pressing uses higher Rubber mould

temperature and pressure (up


to 100bar) to increase
productivity and to ensure
good consolidation of the
laminate. Thermoplastic
matrices require a high
Flat laminate Infrared oven Steel mould
temperature to enable forming
or draping (up to 400C,
depending on the matrix fig. 1.8: Rubber pressing
material), thermosets require
acceleration of the curing process. It is clear that different thermoset composition
with a higher viscosity is required. Reorientation of the reinforcement is inherent
to the high pressure. This induces variability in the thermo-mechanical properties
and makes it difficult to obtain good dimensional accuracy. Another consequence
is high tooling cost due to the pressure.
For automotive applications of thermoplastic composites two steel mould halves
are used, as in conventional deep-drawing of metals. The prepreg is preheated
above Tmelt (for example in an infra-red oven), transported to a high-speed press
and rapidly formed. This stamping process dramatically reduces the cycle time (to

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

the order of 1 minute). Rubber pressing is a similar process for thermoplastic


composite forming, now between a metal mould and a rubber mould. When
compressed, the rubber mould generates a nearly hydrostatic pressure, enabling
good consolidation throughout the laminate, also in products with sharp
curvatures. The lifetime of the rubber mould is of course significantly shorter than
the steel moulds used for stamping.

1.4.6. Liquid Composite moulding


LCM is a technique where the
reinforcement is pre-formed in the
general shape of the final product.
These pre-forms can be
manufactured by e.g. stitching
layers of fabric, compaction of
fabrics with a binder powder or
braiding around a mandrel. The
preform is then placed in a mould,
where a low-viscosity thermoset
resin is injected under pressure.
fig. 1.9: Resin Transfer Moulding
LCM can be applied in many
forms. A variant is called Resin Transfer Moulding, where a rigid mould (matched
metal tooling) is used (see figure). Other variant is the Resin Infusion in Flexible
Tooling, where the resin is impregnated by vacuum. In this technique, the preform
is layed onto a rigid mould. The matched metal tool found in RTM is replaced by
a formable vacuum bag. Due to the low pressure applied to the preform, the rigid
mould can be made of cheap and easy to form material, like PS foam. A know
application of RIFT are wind mill blades.

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.

fig. 1.10: Pultrusion

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

Appendix 1 Nomenclature of common polymer matrix


materials

molecular groups combinations


C O C
CH3 methyl ether
(R O R')
R C O R'
CH3 CH2 ethyl ester
O

CH3 CH3 CH2 propyl monomers


CH2 CH vinyl CH2 O formaldehyde

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.

2.2. Weight and volume fraction.


The relative content of the constituents is mostly quoted as a volume content or
fraction (of fibres for example). Whatever the amount of constituents, we have:
n
Vi
v
i =1
i = 1 with vi =
Vc
(2.1)

where vi is the volume fraction of the ith constituent,


Vi is the volume of the ith constituent,
Vc is the total volume of the composite.
The constituents considered are fibres (subscript f), matrix (m) and voids (v).
Although micromechanics equations are mostly based on volume content,
measurement of constituent content is often based on mass fractions. The mass
content of matrix and fibres (weight of voids is neglegted) follow a similar rule as
the volume fractions:
n
M

i =1
mi = 1 with mi = i
Mc
(2.2)

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

fig. 2.3: Square packing of fibre.

It is fairly straightforward to find an expression between the fibre volume fraction


vf of such a square packing, the fibre diameter d, and the distance between fibre s:
d
2

vf = (2.6)
4s
An other typical packing is the triangular (fig. 2.4)

fig. 2.4: Triangular packing of fibre.


SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.3

R
M

OU
CO
2. Micromechanics - Laurent Warnet & Remko Akkerman.

P
TE
UN I

EN
VE
R

W
SI T
TY OF

The fibre volume fraction can then be written as:


d
2

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%.

2.3 Mechanics of material approach


As was quoted in the introduction, this insight into micromechanics for
composites will be restricted to unidirectional layers. The mechanics of material
approach provides the simplest micromechanical equations for the in-plane
elasticity moduli E1, E2 and G12. No attention is paid to the type of fibre packing.
The composite is therefore considered as homogeneous. A block of composite
containing fibre and matrix as show in fig. 2.5 is simplified to block containing
two volumes. These two
volumes are connected
together and represent f
m
the matrix (m) and the
3
fibre (f) with their
2
respective properties
1
and volume fractions.
An elasticity modulus is fig. 2.5: Representative matrix and fibre volume elements
then obtained by
performing a simple experiment, where the two representative volumes are
subjected to an average stress. Poisson effects are neglected.

2.3.1 Longitudinal modulus


The longitudinal modulus is obtained by
considering the matrix (m) and fibre (f) 1 f 1
representative element in parallel, and by m
applying an average stress along the 1-axis as
shown in fig. 2.6 (model of Voigt). It can be 3
2
shown from strain energy approach that the strain
induced 1 is uniform for both fibre and matrix 1
volume elements. If the areas of the fibre and fig. 2.6: Basic experiment for the
matrix volumes on which the stress is applied are longitudinal modulus.
(Voigts model)
called Af and Am respectively, the following
relation can be written:
1 A = m Am + f A f (2.8)
Using equation (2.1) applied to areas, we obtain:
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.4

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.

2.3.2 Transverse modulus


The transverse modulus is obtained by 2
considering the matrix (m) and fibre (f)
representative element in series. An average f
stress is applied along the 2-axis as shown in
m
fig. 2.7 (model of Reuss). The total 3 2
deformation is the sum of the deformations 2
occurring on the two volumes. This can be 1
expressed in terms of strain: fig. 2.7: Basic experiment for the
transverse modulus.
(Reuss model)

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.

The Reuss model can be improved by simply 2


adding a matrix volume in parallel to the series
model as shown in fig. 2.8. This geometrical
model is an approximation of a square fibre m
packing. This result in the following equation: f
3 2
2
1
fig. 2.8: Parallel-series model for
the transverse modulus.

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

2.3.3 Shear modulus and Poissons ratio.


For the shear modulus, a similar approach similar to that, which led to the
expression of the transverse modulus, is used. Assuming equal shear stresses in
fibres and matrix, the following equations is obtained.
1 v v
= m + f (2.17)
G12 Gm G f 12
For the same reasons as for the transverse modulus, this equation is not very
accurate. The Poissons ratio can be derived form the following rules of mixture:
12 = m vm + f 12 v f (2.18)
This equation is known to be accurate enough for design purposes.

2.4 Semi-empirical models


These models are mostly based on a mechanics of material solution. However
they contain curve-fitting parameters in order to match experimental results or
elasticity solution. The Tsai-Hahn equations for example, use similar assumptions
as the inverse rule of mixture (2.15), but adds a "stress partitioning factor " in
order to take into account the mismatch in stress in the fibre and in the matrix. The
Tsai-Hahn equation for the transverse modulus is:
SI T ES G
O
Composites Course 2008 University of Twente, Eng. & Tech. 2.6

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.

See also on the Production Technology web-site for a program (U20mm)


calculating the thermo-mechanical propoerties of unidirectional layers, as well as
different fabric types. Not only mechanics of materials based procedures are
included, but also elasticity solutions (CCA). A document reviewing this method
is also available on this site.

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:

Ef1 Ef2 Gf12 Gf23 f12 f23 f1 f2 f


Carbon
(GPa) (GPa) (GPa) (GPa) (m/m.oC) (m/m.oC) (kg/m3)

230 141 9 4 0.2 0.25 -0.7x10-6 5.6x10-6 1760

PEI Em Gm
m m m
(GPa) (GPa) (/oC) (kg/m3)

3 1.1 0.35 57 x 10-6 1270

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 E1 (GPa) E2 (GPa) G12 (GPa)


Experimentally: 120 7.8 3.5
table 2.2: Measured unidirectional Carbon-PEI in-plane properties

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. Stress-strain relationships of a composite layer

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.

3.2 Stress-strain relationship


The stresses at a point in a material 3
can be described by 9 components ij
as shown in fig. 3.2, where i,j = 1,2,3. 33
According to this notation, ij are 32
normal stresses when i=j, shear 31 23
stresses when ij (also noted ij). It 22 2
13 12 21
can be shown that only 6 of these
stress components are independent, 11
i.e. ij = ji. Similarly, there are 6 1
strain components ij to describe the
deformation at this point. The fig. 3.2: State of stress in a material
principle is similar as for the stress
components, although the engineering shear components ij are twice the shear
deformation tensor components.

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

table 3.1: Conversion tensor notation-contracted 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

or, in matrix form:


{ } = [S ] { } (3.3)
where [S] is the compliance matrix. There are 36 (6 x 6) compliance constants.
The inverse of the compliance matrix is called the stiffness matrix ([C]=[S]-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

Isotropy reduces the amount of independent constant to 2, which gives the


following set of equations:
1 S11 S12 S12 0 0 0 1
S11 S12 0 0 0
2 2
3 S11 0 0 0 3
= (3.9)
4 Sym 2( S11 S12 ) 0 0 4
5 2( S11 S12 ) 0 5

6 2( S11 S12 ) 6
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.4

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

The reduction in the amount of independent constants in the compliance (and


stiffness) matrix is also due to the relation between the engineering constants E
and G:
E = 2 G (1 + ) (3.10)
The relation between the compliance coefficients Si and the engineering constants
will be developed in the next alinea.

3.3 Engineering constants


It is convenient to express the compliance (or stiffness) matrix components into
engineering constants such as the Youngs modulus E. These engineering
constants are defined by considering simple experiments. A stress situation is
defined where only one stress component is different from zero, i.e.:
k 0; i = 0 for i k ; k = 1,...,6 (3.11)

If we consider the kk -strain component related to this k-stress component, it is


possible to define a modulus as:
Rk = k / kk (3.12)

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

From the symmetry of the [S]-matrix we find:


kl lk
= for k , l = 1,..,3; l k (3.16)
E k El

3.4 Plane stress assumption


A simplification of the stress-strain relations for an orthotropic layer can be made
in case of plane state of stress, i.e. 3=4=5=0. If only the in-plane deformations
are considered, equation (3.15) reduces to:
1 21
0
1 E1 E2 1
21
2 =
1
0 2 or

{ } = [S ] { } (3.17)
E 2 E2
6 1 6
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

3.5 In-plane rotation


Up to now, the stress-strain relationships x2
x2
for a composite layer with the material x1
coordinate corresponding with the plate
coordinate system were considered. The
expressions for a rotated coordinate
x1
system will be derived here for a rotation x3= x3
in the 1-2 plane (axis 3 is fixed).
fig. 3.4: Rotation from (x1, x2, x3) to (x1, x2, x3)
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.6

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

Returning to the composite layer problem, 1*

1*,2*,3: layer CS
1,2,3: material CS (at )

fig. 3.5: material and layer CS

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

In matrix form, the transformation of the strain components in plane stress


becomes:
1* 1* 1 1
* * 1 1
2 = [R ] 2 = [R ] [T ] 2 = [R ] [T ] [R ] 2
1
(3.25)
* 1 *' 1
6 2 6 2 6 6

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 )

Similarly, it is possible to formulate an expression for the transformed compliance


matrix [S*]:
1* 1 1 1*
* *
2 = [R ] [T ] [R ] 2 = [R ] [T ] [R] [S ] 2 = [R ] [T ] [R ] [S ] [T ] 2
1 1 1 1 1 1

* *
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

From this transformation, an expression for the engineering constants of a layer in


any coordinate system can be written. For example, apply a stress 1* along the axis
1*. The strain along the same axis is then:
1* = S11* 1* (3.30)
Which means that E1* = 1 / S11* . Expanding (3.29) gives:
1 1 1 2 12 1
= cos 4 + ( ) sin 2 cos 2 + sin 4 (3.31)
E1* E1 G12 E1 E2

3.6 Stress-strain relationship of a layer including hygrothermal


effects
The anisotropic behaviour of the deformation of a layer is not only evident when a
mechanical stress is applied, but is also subject to changes in environmental
conditions. The most important are change in temperature and moisture content.
This can affect the composite both during fabrication as during its end-use. It is
therefore important to include them in the analysis.
The strain induced by a change in temperature iT or in moisture content iH can be
expressed as:
T i = 1, 2, 3 c i = 1, 2, 3
iT = i for ; iH = i for ; (3.32)
0 i = 4, 5, 6 0 i = 4, 5, 6
where: i is the coefficient of thermal expansion (CTE)
T is the change in temperature (T-T0), T0 the temperature for which
iT =0.
i is the coefficient of hygroscopic expansion (CHE)
c is the moisture concentration, with a moisture free-state as a
reference.
These hygrothermal strain terms are valid for temperature and moisture content
ranges where the coefficient of thermal expansion and the coefficient of
hygroscopic expansion can reasonably be assumed constant. For an orthotropic
material, the stress-strain relationship in the material coordinate system can be
written as:
1 S11 S12 0 1 1 1

2 = S 21 S 22 0 2 + 2 T + 2 c
(3.33)
0 S 66 6 6
6 0 6
or:
{ } = [C ] ({ } { } T { } c ) (3.34)
The transformation of the hygrothermal strain from the material coordinate system
to an arbitrary coordinate system is performed as for the strain components, i.e.
for the thermal expansion components:
1* 1* 1 1
* * 1 1
2 = [R ] 2 = [R ] [T ] 2 = [R ] [T ] [R ] 2
1
(3.35)
* 1 *' 0 0
6 2 6
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 3.9

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
*
.

3.2 Produce a graph of E1* , E2* , G12* , 12* , 16* and 26


*
as a function of for a
Carbon-PEI having orthotropic properties in its material coordinate system
with:
E1 = 120GPa, E2 = 8GPa, G12 = 3.5GPa, 12 = 0.3.

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. Classical lamination theory

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.

4.2. Laminated beams in pure bending


Considering laminated beams under pure flexural loading gives an introduction to
the more general laminated plate theory. Elementary principles of mechanics of
material applied for beam calculations will be used here. We will focus on the
main differences occurring with isotropic beams.
A rectangular laminated beam of thickness h and width b is subjected to a bending
moment M, which is constant over the whole length of the beam (pure bending).
The beam is made of N layers. As
shown in fig. 4.1, the position of a
layer k in the beam is defined by
the distances zk-1 and zk. It is worth
emphasising that these distances
are taken from the middle axis of
the beam and not from the neutral
d
axis.
The following assumptions are
taken:
z0 1
a. The plies are perfectly bonded z
together. M 1 2 h/2 -p l a
ne
m id
b. Plane cross-sections which are zk zk-1 M
initially perpendicular to the h
x
longitudinal axis of the beam k
remain plane and normal n
during flexure. z
c. Each ply behaves linear
elastically with no in-plane
dx
shear coupling ([C16]=0)
We will add for simplicity that the
beam has geometrical and property fig. 4.1: Composite beam bending with layer
numbering system
symmetry about the middle axis.

fig. 4.1 shows a beam element of length dx subjected to a moment M and


therefore having a radius of curvature and an angle between the normals to the
beam axis d. From assumption (b), an expression for the longitudinal strain at a
distance z from the middle axis is:
SI T ES G
O
Composites Course 2007 University of Twente, Eng. & Tech. 4.2

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)

Using this expression, it is possible to obtain expressions for the deflection of


laminated beams from elementary mechanics of material.
Expression for the stress in the kth layer xk can be written by eliminating in
relations (4.2) and (4.5):
M z E xk
x =k
(4.7)
I y E xl
This relation for the stress is similar to the one used for isotropic 1
beam, corrected by the dimensionless term in bracket. The stress 2
is therefore a discontinuous function of the beam depth, in 3
contrast to the stress in an isotropic beam. An example of a stress x
profile of a symmetric 6 layers beam subjected to a bending
moment, where E x2 > E 1x = E x3 is given in fig. 4.2.
fig. 4.2: Example of a
stress profile of a beam
beam subjected to a
bending moment.
SI T ES G
O
Composites Course 2007 University of Twente, Eng. & Tech. 4.3

R
M

OU
CO
4. Classical lamination theory - Laurent Warnet & Remko Akkerman.

P
TE
UN I

EN
VE
R

W
SI T
TY OF

4.3. Theory of laminated plates


The more general case of a z0=-1/2 h 1
laminated plate under plane
stress condition will now be zk-1 1*
analysed. In-plane loading (axial
and shear) will be considered as
loading, as well as moments 2*
k x
0
(bending and torsion). As for the 2
laminated beam, the layers are y
assumed bonded together. No
restriction is set on the lay-up zn=1/2 h
used, which means that various
coupling effects will be present. n

Coupling effects mean that an 3


in-plane stress applied to a 3* z
laminated plate may result in a
complex combination of
extensional, flexural and x,y,z: Laminate CS
torsional deformations. The (at mid-plane)
1*,2*,3: Layer CS
different notations are defined in
1,2,3: Material CS (at )
fig. 4.3. This figure also defines
the different coordinate systems.
fig. 4.3: The different coordinate systems used in the
The position of the layers in the laminated plate theory
normal direction is defined with
the mid-plane as a reference (and
not the neutral plane).
The assumptions relevant for the analysis are similar to the one used for the beam
analysis and can be translated to:
a. The displacements u, v, w corresponding to the directions x, y and z are small
compared to the plate thickness h. The displacement u and v are a linear
function of the depth z (plane cross-sections which are initially perpendicular
to the longitudinal axis of the plate remain plane and normal during
deformation)
b. Transverse shear strains and normal strain are negligible.

4.3.1. Strain in a layer k


According to assumption (a), in-plane displacements are a linear function of the
depth z. Both can be expressed with the displacement in the middle surface (u0
and v0) as reference:
u = u 0 ( x, y ) + z f u ( x, y )
v = v 0 ( x , y ) + z f v ( x, y ) (4.8)
w = w 0 ( x, y ) = w( x, y )
Normal strain is negligible (assumption (b)) and the normal (or transverse)
displacement w is constant for any coordinate (x,y). According to the same
assumption, transverse shear strains are negligible and can be used in the strain-
SI T ES G
O
Composites Course 2007 University of Twente, Eng. & Tech. 4.4

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.

4.3.2. Stress-strain relationship of a layer k


An expression for the stress in the kth layer as a function of the mid-plane strain
and the plate curvature is obtained by combining (4.10) with the relation (3.27)
obtained in chapter 3. In matrix form, this can be written as:
{ *}k = [C*]k { 0 } + z[C*]k { } (4.12)

4.3.3. Laminate loading-deformation relations


External forces and moments acting on a laminated plate can be related to the
stress in the layer, and then to the laminate deformation. For example, the axial
forces Nx per unit width can be obtained by summing the axial stresses x acting
on each layer:
N zk
Nx = ( 1* ) k dz
z
(4.13)

k =1
k 1
where (1*)k is the stress in the kth layer in the 1* direction (layer coordinate
system). A similar expression can be written for the normal force in the
y-direction as well as for the in-plane shear force Nxy. Substituting (4.12) in the
force resultants gives in matrix form:
SI T ES G
O
Composites Course 2007 University of Twente, Eng. & Tech. 4.5

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

4.3.4. Inversion of the laminate loading-deformation relations


Inversion of relation (4.20) is often necessary, as the loading is generally known,
and not the other way around. This can be done by simply inverting the [ABD]
matrix as a whole. However, performing the inversion by inverting the sub
matrices A, B and D can be interesting for particular laminate lay-up. The result is
then of the form:
0 a b N
= b T
d M
(4.21)

For example, it can be shown that the extensional compliance matrix [a] can be
written:

[ 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

x0 a11 a12 a16 b11 b12 b16 N x


0
y
a a 22 a 26 b21 b22 b26 N y
21
xy0 a 61 a 62 a 66 b61 b62 b66 N xy
=
x b11 b21 b61 d 11 d 12 d 16 M x
y b12 b22 b62 d 21 d 22 d 26 M y

xy b16 b26 b66 d 61 d 62 d 66 M xy
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 5.1

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. Classical lamination theory-2

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.

5.2. Engineering constants of a laminate


Laminate engineering constants can be obtained by performing simple
experiments on the laminate. For example, the longitudinal elasticity modulus of
the laminate Ex can be obtained by applying a force per unit width Nx. From the
laminate force-deformation relation:
x0 a11 a12 a16 N x
0
y = a 21 a22


a26 N y or

{ }= [a ]{N }
0
(5.1)
0 a a61 N xy
xy 61 a61

With [a] the laminate extensional compliance matrix. Ex is then:


x Nx /h 1
Ex = = = (5.2)
x a11 N x h a11
0

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

And the shear modulus of the laminate is:


1
Gxy = (5.5)
h a66

5.3. Hygrothermal effects in the theory of laminated plates


The stress-strain relationship of an orthotropic layer with hygrothermal effects
was developed in 3.6. It was based on the assumption that strain having
mechanical, thermal or moisture origins can be treated separately and then added
using superposition. The strain in a layer k subjected to a mechanical stress, a
temperature change T and a moisture concentration c is therefore written as
(3.33):
{ }k = [S ]k { }k + { }k T + { }k c (5.6)
With the thermal expansion vector {} and the moisture swelling vector {} as
defined in (3.32). The resulting stresses are therefore:
{ }k = [C ]k ({ }k { }k T { }k c ) (5.7)
In the global laminate coordinate system, relation (5.7) becomes:
{ *}k = [C *]k ({ *}k { *}k T { *}k c ) (5.8)
According to (4.12), this expression can be written as a function of the mid-plane
strain 0 and the plate curvature :
{ *}k = [C *]k ({ 0 }+ z { } { *}k T { *}k c ) (5.9)
Following a similar procedure as in 4.3.3, external forces and moments acting on a
laminated plate can be related to the stress in the layer, and then to the laminate
deformation. For example, the axial forces Nx per unit width can be obtained by
summing the axial stresses x acting on each layer:
N zk
Nx = ( 1* ) k dz
z
(5.10)

k =1
k 1
where (1*)k is the stress in the kth layer in the 1* direction (global coordinate
system). A similar expression can be written for the normal force in the y
direction as well as for the in-plane shear force Nxy. Substituting (5.9) in the force
resultants gives in matrix form:
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 5.3

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

fig. 5.1: 1D macroscopic residual strain field model of a cross-ply laminate.


The residual stress profile on the macroscopic scale can be obtained from a simple
strain compatibility equation:
T
u90 m
u90 u0T u0m
+ = +
L L L L
T
where ui (with i being 0 or 90 stands for the ply orientation) are the thermal
displacements due to the ply free expansion only, and uim are the mechanical
displacements necessary to compensate the free expansion of the different plies.
This equation can then be rewritten as:
90
r
r
90 T + = 0 T + 0
E90 E0
SI T ES G
O
Composites Course 2007 University of Twente, Eng. &Tech. 5.5

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

5.5. Summary of the calculation of strain and stresses in a


laminated plate.

Input Ply level Laminate level


k-ply engineering constants k-ply stiffness in ply local CS
(E1)k, (E2)k, (G12) k, (12) k, [C]k, []k , []k (3.18, 3.32)
(1)k, (2) k, (1) k, (2) k

k-ply orientation k-ply stiffness in ply global CS


k [C*]k , []k , []k (3.28, 3.35)

Ply lay-up Laminate stiffness


[A], [B], [D] (4.16-19)

Laminate compliance
[a], [b], [d] (4.22-23)

Loading Laminate deformation


(mechanical, thermal, moisture) [0], [] (5.18)

k-ply strains in ply global CS


[*]k (4.10)

k-ply stresses in ply global CS


[*]k (3.27, 5.9)

k-ply stresses in ply local CS


[]k (3.21)
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.1

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

6.2. Damage characterisation.


A short list of failure mechanisms will be given here.

6.2.1. Transverse cracking (also called matrix cracking)


This is often the first damage mechanism to occur. This type of crack grows parallel
to the fibre and in the thickness direction of the laminate. The formation of
transverse cracks does rarely mean the total fracture of a laminate, as it does not
affect the load carrying capacity of the fibres. However, transverse cracking
influences the mechanical and thermal properties of the laminate. Most importantly,
this type of cracking forms a trigger for further damage mechanisms, and provide
paths for the diffusion of surrounding fluids leading to accelerated environmental
degradation.
A simple illustration of a transverse crack is given in fig. 6.1, where a [90/0]s beam
is subjected to a three point bending test. In this case the crack initiates at the free
surface of the 90o layer, and grows in the thickness direction until it meets the fibres
of the 0o layer. On a macroscopic scale, the crack can be approximated by a plane,
which runs perpendicular to the principal stress. Microscopically, an example of a
micrograph (fig. 6.2) shows that the crack runs between fibre interfaces, also
through matrix region when locally the matrix fraction is high.
b

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..

The prediction of the occurrence of transverse crack is often performed using


mechanics of material failure criteria, as described in 6.3.

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.

The prediction of the initiation of a delamination is often related to the occurrence


of transverse cracking. The growth of a delamination is preferably described using
fracture mechanics energy criteria, as discussed in 6.5.

6.2.3. Fibre related failure


It is obvious that the fracture of fibres decreases significantly the load carrying
capacity of a laminate. Chronologically, matrix related cracks (transverse crack +
delamination) are mostly present in the laminate, as shown in fig. 6.5.

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

6.3. Mechanics of material failure criteria.

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.

6.3.1. Maximum Stress -or strain- Criterion.


The Maximum Stress Criterion compares the stress occurring in a layer (in the layer
local coordinate system) to the corresponding lamina strength. No failure occurs if
the following set of inequalities is satisfied:
s1c < 1 < s1t
s2 c < 2 < s 2 t (6.1)
s6 > 6
No interaction between the stress or strain components are taken into account.
The Maximum Strain Criterion is an alternative to the stress criterion. The
inequalities are:
e1c < 1 < e1t
(6.2)
e2 c < 2 < e2 t
e6 > 6

The limits described by these criteria (failure surface) are shown in a 1-2 space in
the following figure:
2
1

Maximum Stress Criterion


Maximum Strain Criterion
fig. 6.7: Maximum stress and strain criterion in 1-2 space
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.7

R
M

OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.

P
TE
UN I

EN
VE
R

W
SI T
TY OF

6.3.2. Quadratic interaction criteria.


Quadratic interaction criteria include terms which allow for the interaction between
the different stress components. For example, the Tsai-Hill criterion in 2D plane-
stress predict failure if this inequality is fulfilled:
12 1 2 22 62
2 + 2 + 2 >1 (6.3)
s12 s1 s2 s6
In this form, the difference in tensile and compression strength properties of the
basic lamina is not taken into account. The failure surface is therefore a symmetric
ellipse. Taking into account the difference in tensile and compressive strength is
done with Tsai-Hill by using the appropriate values of s1 or s2 according to the
considered quadrant of the failure surface. This is shown in fig. 6.8.

2
1

Maximum Stress Criterion


Tsai-Hill Criterion
fig. 6.8: Maximum stress and strain criterion in 1-2 space

A general form of the Tsai-Hill criterion is the Tsai-Wu criterion. It is a tensor


failure criterion, which takes more cross-terms into account.

6.3.3. Example of a laminate strength analysis.


A typical laminate analysis will be developed here, where the initiation of failure is
predicted using the mechanics of material failure criteria presented earlier. A stress
analysis will first be performed. This will summarise some of the previous chapters.
A Carbon reinforced Polyetherimide laminate of total thickness h is made of
unidirectional layers. The lay-up considered is quasi-isotropic [0/45/-45/90]s.
Thermal stresses are not taken into account. The laminate is loaded by a tensile
force in the x-direction inducing a global stress x. The following material data is
available:

E1 E2 G12 12 1 2
Stiffness data
(x109 Pa) (x109 Pa) (x109 Pa) (m/m.oC) (m/m.oC)
Carbon-PEI

120 7.8 3.5 0.32 1x10-6 32x10-6

Strength data s1t s1c s2t s2c s6


(x106 Pa) (x106 Pa) (x106 Pa) (x106 Pa) (x106 Pa)
1400 1200 35 150 60
table 6.1: Measured unidirectional Carbon-PEI in-plane properties
Calculate the force (or the stress) at which the first layer (and which one) fails. Use
for this purpose both simple stress and Tsai-hill criteria.
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.8

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:

All stiffness components in x109 Pa

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

0o 45o 45o 90o


1 0 0 12 1
2 1 12 1
2 1 0 1 0
0 1 0; 1 1
1; 1 1
1 ; 1 0 0 (6.10)
2 2 2 2
0 0 1 12 1
2 0 12 12 0 0 0 1

This results in the following stresses components:


0o 45 o 45 o 90 o
1

x 2.64 0.924 0.924 0.7906
2 (6.11)
= 0.0004; 0.078 ; 0.078 ; 0.1541
x 0 0.101 0.101 0
6

x

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.

h/2 0o h/2 0o h/2


0o
45 o
45o 45o
-45o
90o 1 (x 108 N/m2) -45o
90
o 2 (x 108 N/m2) -45o
90o 6 (x 107 N/m2)
-2 0 2 4 6 -2 0 2 4 -10 -5 0 5 10

-h/2 -h/2 -h/2

fig. 6.9: Stress profiles in the layer global coordinate system 1*, 2*, 3

0o h/2 o h/2 0o h/2


0
45 o
45
o 45o
-45o
1 (x 108 N/m2) 2 (x 107 N/m 2) -45o
6 (x 107 N/m2)
o
-45
90o 90o 90o

-2 0 2 4 6 -2 0 2 4 -2 0 2 4

-h/2 -h/2 -h/2

fig. 6.10: Stress profiles in the layer lobal coordinate system 1, 2, 3


SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.12

R
M

OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.

P
TE
UN I

EN
VE
R

W
SI T
TY OF

6.3.4. Limitations of the mechanics of material approach


As pointed out in the introduction of this alinea, mechanics of material strength or
strain criteria are not based on any physical theory of failure. Besides this, it is
worth adding that the 5 lamina strength mentioned in 6.3 are no material properties.

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.

Example: determination of the strength transverse to the fibres s2t


Let us consider the ultimate stress transverse to the fibres and in tension s2t. The
usual approach to measure this value is a tensile test. Typically, the force-strain
diagram in such a test will grow linearly until a crack initiates in a direction
transverse to the fibres. At this point the crack mostly grows through the whole
laminate as no obstacles are present (like a 0o layer) and fully fractures it. As an
example, tests performed on five [906c]s carbon-polyetherimide tensile coupons
(150 mm x 25 mm x 2 mm) gave an average axial stress at fracture s2t of 35 MPa.
As shown earlier, this value can be used in any of the stress-based criteria for
predicting the failure of a laminate. However, testing the same [906c]s under three
point bending conditions already shows the dependency of s2t on the type of test
used. Under these conditions, a crack will initiate and fracture the specimen at aa
average stress of 70MPa. This shows that s2t certainly cannot be considered as a
material property.
An explanation for the difference found between these two tests can be suggested
when assuming that the formation of a crack is due to the presence of a defect,
mostly due to fabrication. The problem becomes statistical, and it can be assumed
that in any material there exists a certain defect distribution. In a tensile test, the
axial stress is constant over the whole cross-section and the full length of the
laminate. The worst defect present in the laminate will therefore act as the
weakest link and will lead to the formation of the crack. In a three point bending
test, the bending stress is not constant. It varies from compressive to tensile over the
cross-section, and is maximum at half the beam length. The probability that the
location of the weakest defect leading to the crack coincides with the location
where the stress is maximum is small. This can explain why the first crack occurs at
a lower average stress in a tensile test than in a bending test.

Example: Variation of geometry.


An other example is a well-studied subject in the seventies. Several authors have
considered the (multiple) failure of the 90o layer within a laminate ([0/90n/0])
subjected to a tensile force. The strain at which a 90o lamina ([90n]) fails is
independent of the amount of layers n used. In the laminate however, they found
that the strain at which cracking occurs is inversely proportional to the thickness of
the 90 layer. This means that in this case the use of a maximum strain criteria is no
valid. The behaviour described is illustrated in fig. 6.11, where the strain at the
onset of cracking is set against the thickness of the 90o layer. This effect is referred
to as the constraining effect of the 0 layer, but can also be explained in terms of
statistics as in the last example.
SI T ES G
O
Composites course 2008 University of Twente, MechEng 6.13

R
M

OU
CO
6. Laminate failure - Laurent Warnet & Remko Akkerman.

P
TE
UN I

EN
VE
R

W
SI T
TY OF

Strain at onset of crack [90] lamina

[0/90n/0] laminate

Thickness of 90o layer

fig. 6.11: Constraining effect of the 0o layer in a [0/90/0] under tension

6.4. Some notion of fracture mechanics


Fracture mechanics based criteria mainly differ from mechanics of materials criteria
in one major aspect. It is assumed that materials contain defects, which are mostly
introduced during manufacturing. A fracture mechanics based failure criterion
assumes the presence of a defect/crack and studies the growth/no growth of the
considered crack. A crack will grow when a certain amount of energy per unit area
is available. This critical amount of energy is called the critical energy release rate
Gc, in J/m2 and is determined experimentally. It is well recognised that a crack can
grow following a combination of the three crack propagation modes illustrated in
fig. 6.12.

fig. 6.12: Constraining effect of the 0o layer in a [0/90/0] under tension

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

fibre resin isolation

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

product distortions geometry based model for the


thermomechanical properties of woven
autoclave fabric reinforced composites
RTM
hot pressing
rubber moulding lightweight part
... low density
high stiffness spring forward
CONTROL
high strength optimise process by inverse analysis
45o usually thin walled
good corrosion resistance
damage development in
cross-ply beams assembly

MEASURE
lightweight structure

micromechanical
inservice loading
stress analysis

impact and compression after impact


of curved panels
MODEL mechanically induced stresses structural deformation
crack initiation & crack growth and failure

optimal properties of constituents & interface

Composites 2007delamination growth during


compression after impact
CONTROL damage resistant & damage tolerant designs
optimise for impact energy absorption

Composite fabrication techniques:


Thats about Impregnating the fibres
Wetting

Composites 2007

1
Composite fabrication techniques:
Impregnating a porous medium

Needs: Low viscosity resin

Uncured Thermoset
+ pressure
Thermoplast above Tm

Composites 2007

Direct methods:
Direct impregnation

Fibre Yarn / Fabric


Resin
Product
Layer after layer

Liquid resin Thermosetting

Product ready for finish

Composites 2007

Direct method:
Hand lay-up

Cold cure thermoset


(Polyester)

High labour costs


Low tooling costs
Low performance
Emission

Small series
Prototyping
large parts
Low control!

Composites 2007

2
Direct method:
Filament winding

Thermoset

High labour costs


Low tooling costs
Medium performance

Low-Medium series
Large parts
Oven for curing

Composites 2007

Direct method:
Filament winding:

Resin bath Other geometries

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

Similar as Pultrusion but tailor-made lay-up!!

Composites 2007

Semi-Direct methods:
Direct impregnation of a dry preform

Fibre Yarn / Fabric


Preform
Resin
Product
On basis of dry preform

Liquid resin Thermoset

Product ready for finish

Composites 2007

Semi-direct method:
Liquid Composite Moulding (LCM)

Preform in a mould is filled with resin

Several variants:

Resin Transfer Moulding (RTM)

Resin Infusion under Flexible Tooling (RIFT)

Composites 2007

4
Semi-direct method:
Resin Transfer Moulding (RTM)

Preform is filled under pressure

Preform in mould Resin injection Resin cure Part removal

Medium labour costs


High tooling costs
High performance

Medium series
Small medium size
Composites 2007

Semi-direct method:
Resin Transfer Moulding (RTM)

Fibre content determined by cavity size

Mould costs high (Pressure, control)

Composites 2007

Resin Transfer Moulding (RTM)

Composites 2007

5
Semi-direct method:
Resin Infusion under Flexible Tooling (RIFT)

Preform is filled under Vacuum


Vacuum outlet

Bagging film

Sealant tape
Resin inlet

Medium-high labour costs


Low tooling costs
Medium performance

Medium series
Composites 2007 Medium Large size

Semi-direct method:
Resin Infusion under Flexible Tooling (RIFT)

Fibre content determined by pressure

Mould costs low (low pressure)

Composites 2007

Semi-direct method:
Liquid Composite Moulding (LCM)

Impregnation: A long way to go!

Composites 2007

6
Semi-direct method:
Liquid Composite Moulding (LCM)

Material properties to control filling:

Viscosity
Permeability=f(Fibre structure /
orientation / concentration)

K P
Darcys law: Q= (Macro)
x

Composites 2007

Meso flow dominated

Semi-direct method:
Liquid Composite Moulding (LCM)

Flow on a mesoscale:

Micro flow dominated

Composites 2007

Semi-direct method:
Liquid Composite Moulding (LCM)

Flow on a meso-scale: Void formation

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

Variants for thermosets and thermoplastics

Composites 2007

Multi-step methods:
Autoclave curing

Mainly thermosetting matrix


High labour costs
High tooling costs
High performance
Long cycle time
Small-large products
Function integration

Composites 2007

8
Multi-step methods:
Autoclave curing

Constraints on prepreg:

Flexibility
Drapability
Tack

Thermoset

Simple geometries possible


with thermoplastics

Composites 2007

Multi-step methods:
Autoclave curing

Vacuum:

Compressing the layers


(Debulking)

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

Interaction Fabrication method - design


Deformation due to forming:

Fibre reorientation
S S
S

S
S S S

S S

S S
S

Composites 2007

Interaction Fabrication method - design


Deformation due to forming :

Fibre / Matrix concentration

Corner behaviour = f (fabrication technique)

ex: RTM or Rubber pressing, Autoclave

Composites 2007

10
Choice of a fabrication techniques:
Parameters playing a role:

Amounts of products, cycle time, costs


Material (Matrix)
Performance
Vf
Fibre distribution
Void content
Control of the process
Reproducibility
Shape complexity
Assembly

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

Only for Thermosets Function Complex


integration geometry

Technical
Performance
Simplicity

Composites 2007

12
Multi-step methods:
Hand lay-up
Autoclave Productivity
5

4
Series size 3 Part size / cost

Only for Thermosets Function Complex


integration geometry

Technical
Performance
Simplicity

Composites 2007

Multi-step methods:
Hand
Rubber lay-up
Pressing Productivity
5

4
Series size 3 Part size / cost

Only for Thermosets Function Complex


integration geometry

Technical
Performance
Simplicity

Composites 2007

Multi-step methods:
Hand lay-up
Tape layer Productivity
5

4
Series size 3 Part size / cost

Only for Thermosets Function Complex


integration geometry

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 - Finite Elements for Composites

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.

8.2 Finite Element Formulation


The finite element method can be based on the minimum of potential energy [1]. For an
elastic medium, the potential energy can be represented by:

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)

The strains can be written in terms of the displacement field as

kl =
1
( uk ,l + ul ,k ). (7.3)
2

or

{ } = [B] {u} (7.4)


in which kl are the strains which can be written as a function of the displacement gradient.
The finite element matrix [B] contains the first derivatives of the shape functions1.
Substituting equation (7.4) into (7.2), minimised with respect to the displacements results in:

[B] [D] [B] d {u} = { f },


T

(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,

[ K ] sys { u} = { F } sys , (7.9)

where {u} is now the global vector of displacement degrees of freedom.

8.2.1 Plate and Shell Elements


Fibre reinforced composite products are often thin-walled, h/L 1, where h is the laminate
thickness and L is a characteristic length. Hence, very often layered plate or shell elements
can be used for finite element analyses of these types of structures. In general, plate elements
will employ the Kirchhoff theory, which we already know from the CLT. The midplane
strains and the curvatures completely describe the deformation within an element, leading to
stresses and, integrated over the thickness, to forces and moments. Extensions can be made to
include transverse shear (important for sandwich structures, for instance) as in Mindlin type
elements.
Composites Course University of Twente, Eng&Tech 8.3
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.

Figure 7.1 ANSYS SHELL99 Linear Layered Structural Shell element

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

h A 1 223 1323 1223 (7.13)


B D B
v i t i d 0 ( ) ,

Composites Course University of Twente, Eng&Tech 8.4
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.

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)

Simplifying equation (7.13) gives:

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.

8.2.2 Solid Elements


When the stress analysis has to include three-dimensional aspects (for instance at free edges
or connections), plate theory may be an over-simplification. One can go into full three-
dimensional analyses with one or more elements per layers. These elements then have to be
capable of simulating orthotropic properties. In this case, the values for Ex, Ey, Ez, Gxy, Gxz,
Gyz, xy, xz, yz (in the principal material directions) need to be supplied by the user, as well as
the orientation of the principal material axes. This can be an orientation with respect to the
global coordinate system, or with respect to the element axes, or with respect to some user-
defined coordinate system. The amount of data required easily leads to mistakes in the input
file for FE-analyses. This requires step-by-step generation of any model with meticulous step-
by-step verification!
With multi-layered structures, of course, it is easy to create a massive three-dimensional FE-
model requiring corresponding CPU-time. To this end, layered solid elements were
developed, similar to the shell elements discussed earlier. In this case, the plane stress
assumption will be replaced by another approximation of the through-thickness stress-strain
relations. Inplane, the same logic is followed as in the CLT type theories: the inplane strains
are assumed to be continuous over the thickness, so the inplane stresses will be integrated
over the thickness to find the inplane reaction forces. Out-of-plane, however, a different logic
is more applicable. The transverse shear and normal stresses are more likely to be continuous,
and hence the layer strains will be integrated to find the total transverse strain components.
Consider the three-dimensional stress-strain relation for transversely isotropic plies in their
principal coordinate system (123-coordinates),
Composites Course University of Twente, Eng&Tech 8.5
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.

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 ( )

and the Reuter matrix defined as


1 0 0 0 0 0
0 1 0 0 0 0

0 0 1 0 0 0 .
[R] = (7.20)
0 0 0 2 0 0
0 0 0 0 2 0

0 0 0 0 0 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

Figure 7.2 Large deflection of a clamped beam


Composites Course University of Twente, Eng&Tech 8.7
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.

When the simple relation


u
x = (7.24)
x
is used in this case, this will lead to an incorrect large inplane strain, as the length of the strip
is practically unchanged under perfect bending. This can be corrected by adding nonlinear
terms to the strain definition,
u 1 w
2

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

Buckling Mode 1 Buckling Mode 2

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.

8.3 Analysis of strength


Having performed a stress analysis, it is fairly straightforward to assess failure criteria in
terms of stress or strain components, such as a maximum stress or strain criterion, a Tsai-Hill
or a Tsai-Wu criterion. An interesting review of failure criteria is given in [4]. Again, the
more complicated criteria require many input parameters, which, most likely, will not be
readily available. Further, the use of these criteria requires meticulous step-by-step
verification by the user to ensure reliable results.

8.3.1 Continuum Damage


Having performed a strength analysis, the finite element simulation can be carried a step
further. Continuum damage theories (such as Chang Chang [5]) assume some stiffness
degradation (softening) after a failure criterion has been satisfied. Typically, one or more
damage parameters are used, with which the current stiffness is determined as
E (d ) = (1 d ) E0 , (7.26)
with d as the damage parameter (zero for perfect material, one for complete failure) and E0 as
the stiffness of the undamaged material. Theories for anisotropic materials describe different
stiffnesses, with possibly different damage parameters for the different failure modes (e.g.,
matrix or fibre fracture). Simulations of softening are susceptible to mesh-dependent results: a
Composites Course University of Twente, Eng&Tech 8.9
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.

proper analysis must, however, give convergent results with mesh refinement (to be checked
by the user).

8.3.2 Fracture Mechanics


As an alternative, fracture mechanics methods can be applied in FE simulations. Initially,
only the crack growth criteria were evaluated. Special crack tip elements effectively
incorporate the stress singularity around a crack tip. Using those, stress intensity factors can
be evaluated and compared to the corresponding critical values. More generally, energy
criteria can be evaluated. The Virtual Crack Extension method evaluates the elastic energy
stored under a certain loading condition with an initial crack length a and with a slightly
extended crack a+da, under the same loading conditions. Comparing both energies leads to a
G-value (energy release rate, in J/m2),
U (a + da ) U (a )
G , (7.27)
da
which can be compared to an experimentally determined critical energy release rate Gc. A
simulation of this type will only predict whether crack growth will occur under certain
loading conditions. Note, that two elastic analyses have to be performed for this prediction.
As an alternative, also the nodal forces around the crack tip can be evaluated. Sethuraman [6]
and Raju [7] provide a good reference for this Virtual Crack Closure method. In this case,
only one elastic analysis has to be performed, and the method automatically provides a
separation of the three basic crack growth modes. Analyses of cracks between dissimilar
materials (or dissimilar orientations) have to be interpreted with care, however.

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.

delamination growth rate (mm/cycle)

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.

Appendix Ch.8: Finite Element Discretisation

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)

i-2 i-1 i i+1 i+2

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)

A discrete approximation can be used again, now within an element:


m

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

as illustrated in Figure A7.2.


Composites Course University of Twente, Eng&Tech 8.14
8. FEM for composites 2008 - Laurent Warnet & Remko Akkerman.

f (x)

integral of the
linear approximation

i-2 i-1 i i+1 i+2

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.

9.2. Interlaminar stresses in a cross-ply laminate


Let us consider a [0/90]s laminated strip of finite width which is loaded in tension
as shown in fig. 8.1(a).
From a cross-section, first
Fx x
consider the free body
o
diagram of half of the top 0
layer (fig. 8.1(b)).
At the left side of the free (a)
body diagram, the CLT y
predicts the existence of a Fx z
tensile transverse stress y.
This stress is due to the Free-body diagram
mismatch in elasticity moduli ty
and Poissons ratio of the 0o 0o y
x tyz
and 90o layers. This stress (b) 90o tz
0o
cannot exist at the right side
of the free body diagram, as z
this side is free. Force
equilibrium in the y-direction ty
0o y
can only be satisfied with a x
(c) 90o
yz stress. Moment 0o tz
equilibrium provides a stress,
which equilibrate the moment z
of y on the left side of the
free-body diagram. It can be fig. 8.1: Free body diagrams at the interface of a 90 and a
shown [Pipes & Pagano 0 layer in a cross-ply laminated strip loaded under
1970] that the sign of z tension laminated beam in a three point bending
changes near the edge as loading situation.
shown in qualitatively in fig.
8.2. In fig. 8.1(c), a free-body
diagram including the right halves of both top 0o and 90o layers gives the situation
at the 0o/0o interface. In this case, only normal interlaminar stresses do occur. Note
also that z is zero far away from the edges. The region where the 3D stresses are
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 9.2

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

9.3. Consequences for the design of laminates.


It has been reported in the literature that high values of z (tensile), xz (showing
up in angle-ply laminates) and yz cause free-edge delaminations. These premature
failure cannot be predicted using the Classical Lamination Theory. However,
taking interlaminar stresses into account can be important when designing a
laminate, especially when it comes to choose the stacking sequence.
Take as a simple example the [0/90]s cross-ply laminate subjected to a Nx tensile
force. It was shown in fig. 8.2 that at the 0o/0o interface, tensile normal
interlaminar stress z occurs, which can lead to an edge delamination mechanism.
Choosing for a [90/0]s layup does not affect the extensional stiffness properties
(the [A] matrix) but would change the sign of the z stresses at the free edges from
positive to negative, therefore reducing the tendency to delaminations. An other
classical example is the one given by Pagano and Pipes (1971), where the
behaviour of a [15/-15/45/-45]s is compared with that of a [15/45/-45/-15]s. The
qualitative results concerning the normal interlaminar stress in fig. 8.4 show that
the latter lay-up should be stronger than the former one. Similar observations can
be drawn for interlaminar shear stresses.
This latter example gives us a simple rule which can be used when choosing the
lay-up of a laminate.
Avoid when possible stacking up layers of similar orientations in order to
reduce interlaminar shear and normal stresses at the edges.
Choose for example [15/45/-45/-15]s instead of [15/-15/45/-45]s
It can be shown that large orientation difference between adjacent layers tends
to increase the level of interlaminar stresses.
SI T ES G
O
Composites Course 2008 University of Twente, Eng. &Tech. 9.3

R
M

OU
CO
9. Interlaminar stresses - Laurent Warnet & Remko Akkerman.

P
TE
UN I

EN
VE
R

W
SI T
TY OF

Choose for example [0/45/90/-45]s instead of [0/90/45/-45]


More generally, the lecture 4 dedicated to
the stiffness matrix [ABD] of the laminate z [15/45/-45/-15]s
makes us aware of the possible coupling [15/-15/45/-45]s
effect induced when choosing a certain
lay-up. An obvious unwritten rule is to
avoid non-symmetrical lay-up. Translated
mathematically in the stiffness coupling z
matrix [B], non-symmetry induces for
example curvatures when used at a
temperature different than the one at
which it was manufactured. One should
also be aware of other coupling effects,
which can lead to surprising effects, like fig. 8.4: Interlaminar normal stress for
bending-torsion coupling translated in D16. two lay-up having the same
extensional compliance matrix