Sunteți pe pagina 1din 42

CHAPTER

FOUR

MICROMECHANICS

Micromechanics is the study of composite materials taking into account the interaction

of the constituent materials in detail. Micromechanics allows the designer to represent

a heterogeneous material (see section 4 . 1 . 3 ) as an equivalent homogeneous material,

usually anisotropic (see section 4 . 1 . 4 and Figure 4 . 1 ) .

Micromechanics can be used to predict stiffness (with great success) and strength

(with lesser success). There are several approaches, increasingly complex, to derive

micromechanics formulas. To fix ideas, the simplest and most intuitive approach (the

mechanics of materials approach) is introduced first. Then more accurate formulas

and comparisons with experimental data are presented.

Composite materials are the first class of materials that are designed concurrently

with the structure. When designing with metals, the only choices are the type of

alloy and the thermal treatment, followed by the design of the geometry. Composite

materials are produced as combinations of various fibers with various matrices. The

designer can change fibers, matrices, the relative amount of each constituent, and

the geometry of the part, simultaneously. The relative amount of fiber and matrix is

expressed in terms of fiber and matrix volume fractions, which can be varied over a

broad range during manufacturing.

4.1 BASIC CONCEPTS

Basic concepts and definitions used in the rest of this chapter are described in this

section.

61
62 INTRODUCT!ON TO COMPOSITE MATERIALS DESIGN

(a) heterogeneous material (b) equivalent homogenous

anisotropic material

Figure 4.1 Micromechanics process.

4.1.1 Volume and Mass Fractions

The properties of a composite are controlled by the relative volume of fiber and matrix

used. The fiber volume fraction is defined as

volume of fiber
V¡------­
total volume

Toe amount of matrix by volume is the matrix volume fraction defined as

volume of matrix
Vm =
total volume

Since the total volume is the sum of the fiber volume plus the matrix volume,

V¡+ Vm = 1 (4.1)

Toe amount of fiber by weight in the composite is the fiber weight fraction

weight of fiber
W¡=----­
total weight

Toe amount of matrix by weight (or mass) is the matrix weight fraction

weight of matrix
Wm=-----­
total weight

Since the total weight is the sum of the fiber weight plus the matrix weight,

W¡ + Wm = 1 (4.2)
MICROMECHANICS 63

The mass of any material (fiber, matrix, and composite) is equal to the product of

the density times the volume. Therefore, the density of the composite can be computed

as

Pe = Pf V¡ + Pm Vm (4.3)

The volume of any material (fiber, matrix, and composite) can be written as the

mass divided by the density. Therefore, the density of the composite can be computed

also as
1 W¡ Wm
- = - + ­ (4.4)
Pe Pf Pm

In design of composite structures, volume fractions are used because they enter

directly into the computations of stiffness, etc. But during processing, weight fractions

are used because it is much easier to weigh components to be mixed than to measure

their volumes. The conversion from one to another is simple if the density of the

omposite has been computed. In this case, by writing the definition of weight fraction

in terms of density times volume, the relationship between volume and weight fraction

becomes

W¡ = PJ V¡ (4.5)
Pe

Pm
Wm = -Vm
Pe

When more than two constituents enter in the composition of the composite

material, the previous formulas become

P e = ¿p;V;

i=l

_!_ = t W;

Pe i=I Pi


W; =-V; (4.6)
Pe

where n is the number of constituents. Toe computed composite density may differ

from the experimental value Pe,exp because of voids. Toe volume fraction of voids Vv

can be determined experimentally using ASTM D2734, or it can be estimated as

void volume Pe - Pe,exp


Vv= =---- (4.7)
total volume Pe

here Pe is computed with (4. 3 ) and Pe,exp is measured experimentally (ASTM D792).

The fiber weight fraction can be measured by weighting a composite sample, then

removing the resin and weighting the fibers. The resin can be removed by matrix
64 INTRODUCTION TO COMPOS!TE MATERIALS DESIGN

digestion (ASTM 0 3 1 7 1 ) , solvent extraction (ASTM C 6 1 3 ) , or by burning the resin

in an oven.

lt is also possible to determine the fiber volume or weight fraction analytically if

the fiber architecture is known. For a unidirectional material

r¡ TEX
V¡=----- (4.8)
1 0 , 000 PJ te

where r¡ is the number of tows per unit width (tows per cm) perpendicular to the fiber

direction, TEX is the weight of the tow (in g/km), p¡ is the density of the fibers in

g/cc ( c e = cm ' ) , and te is the thickness of the composite !ayer in mm.

For a continuous strand mat (CSM), textile, or chopped strand mat,

w
V¡=---- (4.9)
1 , OOOp¡te

2],
where w is the weight in grams of a square meter of mat [g/m p¡ is the density of

the fibers in g/cc, and te is the thickness of the composite layers in mm. Conversions

to U.S. customary system are given in Appendix B .

Example 4.1 Compute the fiber volume fraction of a unidirectional layer with 5 tows

per cm, each tow is E-glass of 1 1 3 yield, and the final layer thickness is 2.0 mm.

Using the conversion factors from the appendix,

g ] 496, 238 496, 238


TEX [ - = = = 4 3 9 1 glkm

km yield [�f] 113

From Table 2 . 1 , the density of E-glass is 2 . 5 glcc. Then, using (4.8),

(5)4391
V¡ = = 0.439 = 43.9%
1 0 , 0 00 ( 2 . 5 ) 2 . 0

Example 4.2 Compute the fiber volume fraction of a laminate 12.7 mm thick built
2
with 22 layers of double bias [±45] stitchedfabric (Hexcel DB243) having 2 3 . 7 oz/yd

nominal weight of E-glass. Neglect any CSM backing material.

First convert the weight per unit area to metric units

oz 2 8 . 3 5 g/oz 2
w = 23.7- = 23.7(33.9) = 8 0 3 . 5 glm
2 2
yd ( 0 . 9 1 4 4 m!yd)

Then, using (4.9),

803.5(22)
V¡ = = 0.556 = 55.6%
1000(2.5) 1 2 . 7
- ""-

MICROMECHANICS 65

(a) (b)

Figure 4.2 Typical RVEs for (a) rectangular packing array and (b) hexagonal packing

arra y.

4.1.2 Representative Volume Element (RVE)

The fiber volume fraction V¡ can take any value over a broad range allowed by

the particular manufacturing process (e. g . , by hand lay-up between 0% and 60%

by volume). To test for the properties (mechanical, thermal, etc.) of ali these possible

material combinations is impossible; thus, the importance of micromechanics becomes

apparent.

Micromechanics is used to predict the properties of the composite material based

on the known (tested) properties of the constituents (fiber and matrix). Microme­

chanics analyzes the material considering the matrix and the fiber properties and the

geometry of the microstructure. This is done once and for all, so that the designer is

not concerned with the microstructure during the structural design process.

To avoid considering ali the fibers included in a composite during the derivation

of the equations, micromechanics uses the concept of an RVE. An RVE (Figure 4.2) is

the smallest portion of the material that contains ali of the peculiarities of the material;

therefore, it is representative of the material as a whole. Toe stresses and strains are

nonuniform over the RVE because the composite is a heterogeneous material (see

section 4 . 1 . 3 ) . However, the volume occupied by the RVE can be replaced by an

equivalent homogeneous material (see section 4.1.4) without affecting the state of

stresses around the RVE (see Figure 4 . 3 ) . In fact, the state of stresses in the rest of

the structure will not change as long as the designer looks at a scale larger than the

dimensions of the RVE. Toe fiber spacing and the )ayer thickness are typical RVE

dimensions (Figure 4 . 2).

4.1.3 Heterogeneous Material

Heterogeneous materials have properties (mechanical, etc.) that vary from point to

point. Consider a cross section of a tree, where each growth ring is different from the

rest. The lighter rings (summer wood) are softer, and the darker rings (winter wood)

are stiffer. In contrast, a homogeneous material (e.g., steel) has the same properties

everywhere.
INTRODUCTION TO COMPOSITE MATERIALS DESIGN
66

(a)

equivalent homogeneous material

(b)

Figure 4.3 (a) RVE replaced by equivalent homogeneous material and (b) ali structure

but the RVE replaced by equivalent material.

4.1.4 Anisotropic Material

Isotropic materials (e.g., aluminum) have the same properties in any direction. Aniso­

tropic materials have properties (mechanical, etc.) that vary with the orientation.

Anisotropic materials may be homogeneous, but the properties change depending

on the orientation along which the property is measured. A typical example is the

modulus of elasticity of wood, which is higher along the length of the tree and lower

across the growth rings. Although wood is a heterogeneous material (each growth

ring is different from the rest; see section 4 . 1 . 3 ) , when looking at a large piece of

wood, the material is assumed homogeneous. That is, for simplicity, the peculiarities

of the growth rings are ignored. However, the differences in the modulus along various

directions must be recognized.

Toe stiffness of an isotropic material is completely described by two properties,

for example, the modulus of elasticity E and Poisson's ratio v. In contrast, up to 21

properties may be required to describe anisotropic materials [ 1 , p. 157].

4.1.5 Orthotropic Material

An orthotropic material has three planes of symmetry (Figure 4.4) that coincide with

the coordinate planes. A unidirectional fiber-reinforced composite may be considered

to be orthotropic. One plane of symmetry is perpendicular to the fiber direction,

and the other two can be any pair of planes orthogonal to the fiber direction and

among themselves. Only nine constants are required to describe an orthotropic ma­

terial.
MICROMECHANICS 67

ººººº
ººººº -- - - - - -

Figure 4.4 An orthotropic material has three planes of symmetry.

4.1.6 Transversely Isotropic Material

A transversely isotropic material has one axis of symmetry. For example, the fiber

direction of a unidirectional fiber-reinforced composite can be considered an axis of

symmetry if the fibers are randomly distributed in the cross section. In this case, any

plane containing the fiber direction is a plane of symmetry. A transversely isotropic

material is described by six constants.

4.1.7 Isotropic Material

The most common materials of industrial use are isotropic, like aluminum, steel,

etc. Isotropic materials have an infinite number of planes of symmetry, meaning that

· the properties are independent of the orientation. Only two constants are needed to

represent the elastic properties of isotropic materials. These two properties can be the

Young modulus E and Poisson's ratio v, but severa! other pairs of properties have to

be related to any other pair. For example, isotropic materials can be described by E

and G , but the shear modulus of isotropic materials is related to E and v by

E
G=--- (4.10}
2(1 + v)

4.2 STIFFNESS

In the mechanics of materials approach, both fibers and matrix are assumed to be

isotropic (see section 4 . 1 . 4 ) . The stiffness of an isotropic material is completely rep­

resented by two properties: the modulus of elasticity E and Poisson's ratio v. Using
68 INTRODUCT/ON TO COMPOSITE MATERIALS DESIGN

1.0

0.9

0.8

--o- E1/Et
0.7
� -v- E2/Et
Q)

a. 0.6 -<r- G12!Et


o
....
a.. --O- G 2 8f G 1 2
0.5
cij
·;::

0.4
!
� 0.3

0.2

0.1

o.o
o.o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Fiber Volume Fraction

Figure 4.5 Micromechanics predictions as a function of the fiber volume fraction [2].

micromechanics, the combination of two isotropic materials (fiber and matrix) is rep­

resented a s a n equivalent, homogeneous, anisotropic material (see section 4 . 1 . 4 ) . The

stiffness of the equivalent material is represented by five elastic properties:

E ¡ : modulus of elasticity in the fiber direction,

E2: modulus of elasticity in the direction transverse to the fibers,

G12: inplane shear modulus,

G23: out-of-plane shear modulus,

v ¡ z : inplane Poisson's ratio.

Ali of these are functions of the fiber volume fraction (Figure 4 . 5 ) .

4.2.1 Longitudinal Modulus

The longitudinal modulus, or modulus of elasticity .in the fiber direction, can be pre­

dicted very well by (4.15), called the rule of mixtures (ROM) formula. The main

assumption in this formulation is that the strains in the direction of the fibers are the

same in the matrix and the fiber. This implies that the fiber-rnatrix bond is perfect.

When the material is stretched along the fiber direction, the matrix and the fibers will

elongate the same, as shown in Figure 4.6. This basic assumption is needed to be

able to replace the heterogeneous material in the RVE by a hornogeneous one (Fig­

ure 4 . 3 ) while satisfying compatibility of displacements with the rest of the body. Or

vice versa, replace everything but the RVE by an equivalent material while satisfying

compatibility.
MICROMECHAN/CS 69

Poisson Effect

··········¡··············· . . . . · - · .----.

Matrix
� a;

;

... 1

---....J···-···-·-··---··---··-··-··--·-- ···--···-···-····-················ ···················· ···················-····· ················�

Figure 4.6 RVE subject to longitudinal unifonn strain.

By definition of strains,

b. L
E ¡ = -

Since both fiber and matrix are isotropic and elastic, their stress-strain law is

ºf = E¡E¡

The average stress cr¡ acts on the entire cross section of the RVE with area

A = A ¡ + Am

Toe total load applied is

P = cr ¡ A = cr¡A¡ + CTmAm (4.11)

Toen

cr¡ = E¡(E¡V¡ + z; Vm) (4.12)

where

V ¡ = A¡/A and Vm = Am/A

For the equivalent homogeneous material,

CT¡ = E¡E¡ (4.13)


70 INTRODUCTION TO COMPOSITE MATER/ALS DESIGN

Toen, comparing ( 4 . 1 2 ) w i t h ( 4 . 1 3 ) results in

E¡=E¡V¡+EmVm (4.14)

Finally, the ROM formula is obtained using (4 . 1 )

E ¡ = E¡V¡ + E m ( l - V¡) (4.15)

According to the ROM, the property E 1 depends linearly on V¡ and the properties

of the constituents, as shown in Figure 4.5 for an E-glass composite (E¡ = 72.3

GPa, Em = 5.05 GPa, v¡ = 0.22, Vm = 0.35). In most cases, the modulus of the

fibers is much larger than that of the matrix (see Tables 2.1 and 2.4-2.5). Toen

the first term dominates, making the contribution of the matrix to the composite

longitudinal modulus negligible. This indicates that the longitudinal modulus E1 is

a fiber dominated property. Also note that V¡ cannot reach all the way to 100% as

Figure 4.5 may suggest. To begin with, the best packing of cylindrical fibers is the

hexagonal array (Figure 4.2) with V¡ about 90%. Furthermore, processing conditions

usually limit V¡ to values much lower than that (e.g., for pultrusion to about 45%).

4.2.2 Transverse Modulus

In the determination of the modulus in the direction transverse to the fibers, the main

assumption is that the stress is the same in the fiber and the matrix. This assumption is

needed to maintain equilibrium in the transverse direction. Once again, the assumption

implies that the fiber-matrix bond is perfect. Toe RVE, subject to a uniform transverse

stress, is shown in Figure 4.7. Note that, for simplicity, a cylindrical fiber has been

replaced by a rectangular one. In reality, most micromechanical formulations (except

advanced formulations [2]) do not represent the actual geometry of the fiber at ali.

Since fiber and matrix are assumed to be linear elastic materials, the strains in

the fiber and the matrix are

a2
Ef = ­

These strains act o v e r a portion of the RVE, E¡ over V¡ W, and Em over Vm W,

while the average strain E2 acts o ver the en tire width W . Toe total elongation is

E2 W = Ef V¡ W + Em Vm W (4.16)

Cancelling W and using Hookes law for the constituents, which are assumed to

be isotropic, a¡ am
E2 = - V ¡ + -Vm (4.17)
E¡ s;

Since the stress is the same in the fiber and the matrix (a¡ = am = a2),
MICROMECHANICS 71

T
w

l
¡ ¡ (j2 ¡ ¡ ¡ ¡ ¡

Figure 4.7 RVE subject to transverse uniform stress.

Toen, comparing with Hookes law for the equivalent material (a2 = E2E2), the

transverse modulus is given by

1 Vm V¡
- = - + ­ (4.18)
E2 z; E¡

Equation (4 . 1 8 ) is known as the in verse ROM. Toe fibers do not con tribute ap­

preciably to the stiffness in the transverse direction unless V¡ is very high (see Figure

4.5), in which case the assumptions made in this section are not valid. Therefore, it

is said that E2 is a matrix dominated property. The prediction obtained with ( 4 . 1 8 ) is

a lower bound, and it is not accurate in the majority of cases. A lower bound means

that the property is underestimated. Toe inverse ROM is a simple equation to be used

for qualitative evaluation of different candidate materials but not for design calcula­

tions.

A better prediction can be obtained with the semiempirical Halpin-Tsai [3] for­

mula

1 + {r¡V¡ J
E2 = Em
[
1 - r¡V¡

(E¡/Em)-1
r¡ = (4.19)
(E¡/Em)+t
72 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

6 � Halpin-Tsai

- Periodic Microstructure

-9- Rule of Mixtures

CJ Experimental Data
E
w
..._ 4
C\J
w

o ...._._........._._..........._._...,._._...._._.._._................................_._......................_._...._._.._.__.__.................

o.o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 o.a

Fiber Volume Fraction

Figure 4.8 Comparison of predicted values of E2/ Em with experimental data for

glass-epoxy.

where { is an empirical parameter obtained by curve fitting with (4 . 1 9 ) the results

of an analytical solution (too complex to describe here). The value { = 2 usually

gives good fit for the case of circular or square fibers. For rectangular fibers, a good

estimate is { = 2a/b, where a and b are the dimensions of the rectangle in the

direction of loading and perpendicular to it, respectively. The periodic microstructure

model (PMM, [2]) leads to accurate formulas for E2 and also G 12, G23, v12, and E ¡ .

Since the PMM formulas are relatively complex, they have been prograrnmed in the

accompanying software. Comparison with experimental data is presented in [2] and

Figure 4.8 for glass-epoxy with E¡/Em = 2 1 . 1 9 , v¡ = 0:22, Vm = 0.38.

4.2.3 lnplane Poisson's Ratio

Any Poisson's ratio is defined as minus the quotient of the resulting strain over the

applied strain:

(4.20)

That is, in a test in which load is applied in the i-direction, strain is induced by

Poisson's effect on the perpendicular j-direction. The mechanics ofmaterials approach

leads to a ROM equation for the inplane Poisson ratio

V12 = v¡ V¡ + Vm Vm (4.21)
MICROMECHANICS 73

o
2
o
Fiber

0 0

't

(a) l n p l a n e shear O's (b) lnterlaminar shear (J4

Figure 4.9 Distinction between inplane and intralaminar shear stress.

An approximate prediction of Poisson' s ratio is usually sufficient in design. Since

Poisson's ratio for the matrix and the fibers are not very different, (4.21) predicts also

a similar value for the composite. Poisson's ratios are difficult to measure, mainly

that of the fiber. Sorne fibers like carbon fibers are not even isotropic. Finally, Poisson

effects are usually secondary effects. For ali these reasons, ( 4 . 2 1 ) is generally ade-

quate for design. Since Poisson's ratio is predicted by a ROM equation ( 4 . 2 1 ) , the "

plot as a function of fiber volume fraction is similar to the plot of the longitudinal

modulus.

4.2.4 Inplane Shear Modulus

3
The inplane shear stress ª6 = r12 = r21 deforms the composite as in Figure 4.9a. The

strength of material approach leads to an inverse ROM equation [4] for the inplane

shear modulus

_l_ = Vm + V¡
(4.22)
G12 o; G¡

Equation (4.22) predicts that the inplane shear modulus is a matrix dominated

property in the case of stiff fibers. Toe inverse ROM can be rewritten as

Gm
G12=------ (4.23)
Vm + V¡Gm/G¡

3
See section 5 . 1 for the definition of the material coordinate system.
74 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

If the fibers are much stiffer than the matrix (Gm « G¡), the inplane shear

modulus can be approximated by the matrix dominated approximation

rv Gm
G12=--­ (4.24)
l - V¡

Once again, the inverse ROM gives a simple but not accurate equation for the

prediction of the inplane shear modulus. The cylindrical assemblage model (CAM,

[5]) gives a better approximation

(1 + V¡) + (1 - V¡) G m / Gf J
G 1 2 = Gm (4.25)
[
(1 - V¡)+ ( 1 + V¡)Gm/G¡

Again, if (Gm « G¡), the matrix dominated approximation is

+ V¡) (4.26)
G12�Gm ---
(1
1 - V¡

Equation (4.26) deviates slightly from the results of (4.25) as shown in Figure

4 . 1 0 . The periodic microstructure model (PMM) [2] yields the formula

G12
.
= Gm
[
1 + V¡ ( 1 - G m / Gf)
( _J (4.27)
Gm/G¡ + S3 1 - Gm/G¡)

where

S3 = 0.49247 - 0.41603V¡ - 0.02748Vj (4.28)

Toen the matrix dominated PMM formula is

(4.29)

Equation (4.29) gives almost the same results as (4.27) and good agreement

with experimental data for most common materials. Experimental data and predicted

values of G 1 2 / G m are compared in Figure 4 . 1 0 for glass-epoxy with Em = 4.0 GPa,

Vm = 0 . 3 5 , E¡ = 72.3 GPa, v¡ = 0.22 (see Table 2 . 1 ) .

In this section, the matrix and the fibers are assumed to be isotropic. Therefore, the

shear modulus of the fiber and the matrix can be computed from known elastic modulus

and Poisson's ratio using the following formula, valid for any isotropic material:

E
G=--- (4.30)
2(1 + v)
MICROMECHANICS 75

6 __.,_ Cylindrical Assemblage

--9'- Periodic Microstructure

-- Rule of Mixtures

5 -<>- Matrix-dominated CAM

O Experimental Data

E 4
o
..._
(\J

c5' 3

o _._ _._ _._ _._�

o.o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 o.a

F i b e r Volume Fraction

Figure 4.10 Comparison of G 12/ Gm with experiments for E-glass composite.

4.2.5 Interlaminar Shear Modulus

The interlaminar stress a4 = r23 = r32 acts across the thickness of the composite
4
as shown in Figure 4.9b. The interlaminar shear modulus can be computed with the

semiempirical stress-partitioning parameter (SPP) technique (section 4.2.6) as [7]

V¡ + r123 ( 1 - V¡)

G23 = e: ( )
r123 1 - V¡ + V¡Gm/G¡

3 - 4vm + Gm/G¡
23 (4.31)
T/ = 4 ( 1 - Vm)

The PMM [2] that accounts for the exact geometry of the fibers gives similar

results to ( 4 . 3 1 ) for most common composites.

Toe interlaminar shear stress as = r13 = r31 introduces a shear deformation

through the thickness of the composite, which is similar to that in Figure 4.9a. There­

fore, it is usual to assume

G13 = G12 (4.32)

which is exact for a transversely isotropic material (section 4 . 1 . 6 ) with the axis of

symmetry coinciding with the fiber direction.

4
See section 5 . 1 for the definition of the material coordinate system.
76 INTRODVCTION TO COMPOSITE MATERIALS DESIGN

4.2.6 Stress Partitioning Parameter (*)

The stress partitioning parameter (SPP) is a typical example of an experimental cor­

rection for an otherwise not accurate formula. This approach is used often in the field .

of composite materials. When an accurate formula is not available, the obvious al­

ternative is to test all the possible volume fractions to be encountered during design

and fit the data with a curve fit. This curve-fitting approach is too expensive because

of the large amount of testing required. A more intelligent approach is to correct

the inaccurate formula using limited experimental information. In this semiempirical

method, all the good information that was built in the original formula is preserved

while enhancing it with experimental data. The SPP can be used to correct both the

transverse modulus E2 (4 . 1 8 ) and the inplane shear modulus G 12 (4.22).

In the derivation of the inverse ROM equation for E2 (section 4.2.2), it was

assumed that the stress in the fiber and the matrix were the same

That is only an approximation brought by modeling the fiber as a rectangle in

Figure 4.7, while actually the fiber is cylindrical. To obtain a better approximation for

E2, a cylindrical fiber is considered. For a cylindrical fiber, the stresses a¡ and ªm are

not uniform, but they can be represented by their average values. Note that in reality

the average stress in the matrix is lower than in the fiber, because Em < E¡. Without

actually computing the real stress field, the SPP value rJe can be defined as the ratio

of the average stress in the matrix to the average stress in the fiber

O < rJe :S 1 (4.33)

where the average stresses are defined by averaging the actual stress distribution over

the volume [6]

ci¡ = _.!.._ { a¡ d V
V¡ Jv1

'i'fm = _1_ { ªm dV

Vm Jvm

(4.34)

where V¡, Vm, and Ve are the volume of fiber, matrix, and composite, respectively.

The average stress in the composite can be decomposed as

_ V¡
O' e = - -
1 f Vm
a ¡ d V + - -
1 f O' m d V (4 . 3 5 )
Ve V¡ V¡ Ve Vm Vm
MICROMECHANJCS 77

and recognizing that efe = a2,

(4.36)

Using ( 4 . 3 3 ) , the relationship among the averages (4.36) becomes

(4.37)

Now, rewrite ( 4 . 1 7 ) using (4.33) and recognizing that in (4.17) a¡ and am are

actually average values cf¡ and cfm

af r/eª f
E2 = -V¡ + - - V m (4.38)
E¡ e;

Multiply and divide by the parenthesis in (4.37) to get

(4.39)

Now recognize a2 from (4.37) and divide by it to get the corrected ROM equation

for E2

1 E2 (V¡ r/ e V m ) 1
(4.40)
E2 = a2 = E¡+ Em ( V ¡ + r¡eVm)

In the strength of materials derivation (section 4.2.2) it was assumed that a¡ = am

or r/e = l . Consequently, (4.42) reduces to (4 . 1 8 ) .

Next, one experiment is performed at a fixed V¡ to experimentally measure E2.

From this experiment, the correction r/e can be calculated from (4.40), or writing r/e

explicitly as

V¡Em(E2 - E¡)
r/ e = - --'-- - - - - � (4.41)
VmE¡(E2 - Em)

The next step is to assume that the correction factor n¿ is independent of V¡ for

the tested material system. Then (4.40) or

V¡ + n, ( 1 - V¡)
E2 = Em -�----'-----'------- (4.42)
r/e (1 - V¡) + V¡ Em/E¡

can be used to accurately predict the transverse stiffness E2. Equation (4.42) matches

experimental data for values of V¡ other than the one at which r/e was obtained, as

long as the fiber, matrix, and process are not changed.

Similarly, performing one experiment at a fixed V¡ to experimentally measure

G 12, the shear-stress partitioning parameter tls can be calculated as

(4.43)
78 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

Toen the equation

V¡ + n, ( 1 - V¡)
G12 = e; ( (4.44)
n, 1 - V¡)+ V¡ Gm/G¡

can be used to accurately predict the inplane shear modulus G 1 2 . Finally, the SPP for

inplane shear can be estimated [7] as

1 ( Gm)
rJs = 2 1 + G¡

4.2. 7 Continuous Strand Mat (*)

A continuous strand mat (CSM) is a fiber system containing randomly placed, con­

tinuous rovings held together by a binder (section 2.1.2). CSM is used to obtain

bidirectional properties on pultrusion and other processes where unidirectional rov­

ings constitute the main reinforcement. Since CSM is produced with unidirectional

roving, it is more expensive than chopped strand mat, which is produced directly from

the glass furnace, thus saving the intermediate step of making a roving. Furthermore,

better quality E-glass is commonly used for CSM, while the chopped version is usually

made of lower quality glass. CSM has, in general, better mechanical properties than

chopped strand mat, but it is more expensive. Therefore, hand lay-up production of not

so critica! structures (e.g., small boats) may use chopped strand mat. The interlaminar

shear strength F4 and Fs (section 4.4.6) are usually lower in composites made with

lower quality glass. If these properties are critical, E- or S-glass should be used.

Toe elastic properties of both CSM and chopped strand mat can be predicted

assuming that they are random composites. A ]ayer of composite with randomly ori­

ented fibers can be idealized as a laminate with a large number of thin unidirectional

layers, each with a different orientation from Oº to 1 8 0 º . Toe properties of the random

composite are the average properties of this fictitious laminate. Each fictitious unidi­

rectional layer has a reduced stiffness matrix [ Q ] given by (5.23). The orientation of

each !ayer is accounted for by rotating the matrix [ Q ] to a common coordinate system

by using ( 5 . 4 8 ) . Toen the average is computed by

(4.45)

which leads to

CSM 3 1 3 1
Qll = 8Q¡¡ + 4Q12 + 8Q22 + 2º66

CSM 1 3 1 1
Q¡z = + + -
8Q11 4Q12 8Q22 2Q66

CSM 1 1 1 1
º66 = 8Q¡¡ - 4Q12 + 8Q22 + 2º66

_ QCSM _ Ü
QCSM (4.46)
16 - 26 -
MICROMECHANICS 79

Using (5.23) to write the coefficients of both [Q] and [QCSM], the isotropic

properties E, G, and v of the random composite can be obtained from the known

properties of a unidirectional material with the same fiber volume fraction

vE

2
1 - v

(4.47)

with ó = 1 - v12 v 2 1 - Solving the above three equations for E, G, and v results in

Ef + 4E¡ G 1 2 ó + 2E1E2 + 8 v 1 2 E 2 G 1 2 ó - 4vf2E? + 4E2G12ó + Ei


E = ��������������������������

ó(3E1 + 2v12E2 + 3E2 + 4G12ó)

E¡ - 2 v 1 2 E 2 + E2 + 4G12ó
G = �����������

E ¡ + 6v12E2 + E2 - 4 G 1 2 ó
V = �����������- (4.48)
3 E ¡ + 2v12E2 + 3E2 + 4G12ó

where ó = 1 - v12 vzi lt can be verified that the equation above satisfies the isotropic

relationship G = E/2(1 + v ) . These expressions can be approximated [8] by

3 5
E = - E ¡ + -E2
8 8

1 1
G = - E ¡ + -E2
8 4

E
V = - - 1 (4.49)
2G

where E 1 and E2 are the longitudinal and transverse moduli of a fictitious unidirec­

tional !ayer having the same volume fraction as the CSM !ayer. Toe approximation for

G i s obtained by assuming ó = 1 , v12 = 1/4 , and G 1 2 = } E z . The approximation for

E is obtained dividing the equation for E by ( 1 - v2) and using the same assumptions.

Comparison between experimental and predicted elastic modulus (using (4.15)

and ( 4 . 1 9 ) to obtain E¡ and E2) is presented in Table 4 . 1 for vinyl ester reinforced
2
with E-glass CSM (Owens Corning 457.5g/m ). E-glass properties are given in Table

2.1 and vinyl ester properties, including the matrix modulus Em, are given in Table

2.4. From the experimental data it can be seen that the reinforcement is not truly

random. The direction along the length of the CSM roll is called longitudinal. Along

the transverse direction, the stiffness of the composite is about 13% lower because

of a slightly preferential orientation along the length of the roll. The predicted values
80 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

Table 4.1 Comparison of predicted and experimental

values of E/ Em for vinyl ester matrix reinforced with

E-glass CSM

Fiber Volume Fraction Modulus E!Em

Predicted Experimental

Longitudinal Transverse

0.16 8.81 8.58 7.04

0.25 11.90 10.81 9.94

are higher than expected, perhaps because the predictions are based on the nominal

weight of the CSM and a variability of up to ± 6% is possible.

4.2.8 Sorne Restrictions on the Elastic Constants

Severa! restrictions on the possible values for the various elastic constants presented

in this chapter can be derived from elasticity theory. The most useful relationship is

(4.50)

This relationship is routinely used to compute Poisson's ratio v z ¡ . This and other

restrictions [ 4, p. 43) are also used to check the validity of experimental data, as in

the following example.

Example 4.3 The elastic properties E¡ = 19.981 GPa and v¡z = 0.274 were mea­

sured in a longitudinal test (fibers in the direction of loading) by using ™{O strain

gages: one longitudinal and one transverse. The elastic properties Ez = 1 1 . 3 8 9 GPa

and v21 = 0.153 are measured in a transverse tensile test (fibers perpendicular to

loading] in the same way. For the test procedure to be valid, all four data values E 1 ,

E2, v ¡ z , and vz¡ must conform to (4.50), within the experimental error.

First, compute both sides of (4.50)

E¡ 19.981
- = -- = 7 2 . 9 GPa
V¡z 0.274

E2 1 1. 3 8 9
- = -- = 74.4 GPa
Vz¡ 0.153

The difference is small taking into account that there may be sorne experimental

error. Therefore, the data are considered valid.


MICROMECHANICS 81

4.3 THERMAL AND MOISTURE EXPANSION (*)

The change in dimension b.L of a body as a result of mechanical strain E, change in

temperature ,6. T, and change in moisture content S. m , can be approximated by the

linear relationship

b.L = EL o + {Jb.m + a b. T (4.51)

where Lo is the initial length, fJ is the coefficient of moisture expansion, and a is the

coefficient of thermal expansion (see Tables 2 . 1 and 2.4-2.5).

4.3.1 Thermal Expansion

6
The thermal expansion coefficient of ali resins are positive (about 30 to 100 10- /ºC)
6
and higher than steel alloys ( 1 0 to 20 10- f°C). The coefficient of thermal expansion
6
of E-glass is low (5.04 10- /ºC). Carbon fibers have a negative expansion in the fiber
6
direction (-0.99 10- f°C) and a relatively large expansion in the transverse direction
6
(16.7 10- f°C). This means that depending on the arnount of fibers (fiber volume

fraction) it is possible to tailor the coefficient of thermal expansion of the composite

to the user' s needs. Especially, it is possible to produce a composite material with a

very low coefficient of thermal expansion, which is useful when dimensional stability

is required (see Example 4.4).

Composite materials have two coefficients of thermal expansion. In the direction

of the fibers, the thermal expansion behavior is dominated by the fibers, and the

coefficient is computed according to [7]

1
a ¡ = -- [a¡E¡(T)V¡ + <X m E m ( T ) V m ] (4.52)
E¡(T)

In the direction perpendicular to the fibers, the thermal expansion is dominated by

the matrix material. In this case the thermal expansion coefficient is computed from

[7, 27]

(4.53)

Equations (4.52) and (4.53) have been correlated to experimental values and they

are reported to be quite satisfactory for the prediction of thermal properties. It must be.

noted that carbon fibers have a different coefficient of thermal expansion in the longi­

tudinal and transverse directions. The former should be used in (4.52) and the later in

(4.53). Note that the elastic moduli used in both equations depend on the temperature.

Because of this, these formulas cannot be used over a large range of temperatures [7].

Prediction of coefficients of thermal expansion at constant temperature are shown in

Figure 4 . 1 1 .

Since a !ayer of random composite is isotropic in the plane of the !ayer, the

coefficient of thermal expansion aq of a composite with randomly oriented fibers can

be found following the approach described in section 4.2.7 [9]

a¡ + a2 a¡ - a2 E¡ - E2
<Xq = + --- -------- (4.54)
2 2 E ¡ + (1 + 2v21)E2
82 /NTRODUCTION TO COMPOS!TE MATERIALS DES!GN

-O- a ¡ / a m

4 ��lam

-o- 132 I 13m

� 3
Q)
-0-- K¡ / 1Cm
a.
o
....
a..
2
ca
·;::

Q)


o.o 0.1 0.2 o.3 0.4 o.s o.e o.7 o.a o.9 1.0

Fiber Vo lume Fraction

Figure 4.11 Prediction of transport properties.

where a ¡ , a2, E¡, E2, v21 are the properties of a fictitious unidirectional composite

with the same fiber volume fraction of the random composite, computed with the equa­

tions described earlier in this chapter. It must be noted that experimental values of aq

are very scattered and the formula for aq may underestimate or overestimate the ex­

perimental values. Experimental values can be obtained following standard procedures

such as ASTM E 8 3 1 and E289.

Example 4.4 Select the materials and processing technique to produce a square tube

that will bend the least under a temperature gradient 6. T . The top face of the tube

will be exposed to the sun, while the bottom and the sides will be in the shade. The

temperature differential between the top and bottom fiange is 6. T . Two fabrication

techniques are available: hoop filament winding which lays the fibers at 90º with

respect to the axis of the tube and pultrusion that aligns the fibers at Oº.

First, select the resin with the least thermal expansion that is compatible with

both processes. Using the tables in Chapter 2, an isophthalic polyester with a =


6
30 x 10- j°C is selected. Next, using (4.52), (4.53), compute a ¡ and a2 using various

fibers. The required values of E¡ and E2 were computed using the ROM. Two values

of fiber volume fraction are selected, which can be achieved with both processes. The

results are shown in Table 4.2, from which it is clear that the least thermal expansion

is in the fiber direction. So the fibers should be aligned along the axis of the tube

(pultrusion). Carbon fibers should be used for optimum performance since they yield

the lowest thermal expansion or contraction.


MICROMECHANICS 83

Table 4.2 Coefficient of thermal expansion in

Example 4.4

V¡= 0.45 V¡= 0.55

Fiber aq a2 a¡ a2
6
¡ 1 0 - 6 /ºC] ¡10- /ºCJ ¡ 1 0 - 6 /ºC] ¡10-6 /ºCJ

E-glass 6.74 23.7 6.31 20.4

S-glass 4.16 23.1 3.76 19.5

Kevlar49 -1.02 21.9 -1.33 17.6

CarbonT300 -0.06 22.5 -0.23 18.3

4.3.2 Moisture Expansion

Moisture is absorbed primarily by the polymer matrix with the exception that aramid

fibers can also absorb moisture. Once moisture is absorbed, it produces swelling of

the polyrner. Moisture is absorbed in a Fick:ian process, and the absorption rate is

controlled by the diffusivity coefficient. The diffusion of moisture in the composite is

much slower than thermal conduction. When exposed to a changing environment, the

material will reach thermal equilibrium much faster than moisture equilibrium.

The average moisture concentration in the composite is defined as

moisture weight
m = -------�
dry composite weight

Then moisture absorption can be easily obtained by measuring the weight of

samples. However, careful consideration should be given to the void content, since -

water going to fill up voids <loes not contribute to the swelling. If the void content is

known, and if it can be assumed that ali voids will be filled by water, the weight of

the water in the voids can be subtracted to get the weight of water really absorbed

by the polymer. The void content per volume Vv can be estimated by (4.7). Toen the

density p of the material can be found as the weight of a sample divided by its volume

corrected for void content, v = v 0 ( 1 - Vv), where v0 is the apparent volume.

While organic fibers (e.g., aramid) absorb water, inorganic fibers (e.g., glass,

carbon) do not absorb moisture. In the case of inorganic fibers, the coefficients of

moisture expansion in the longitudinal and transverse direction of the composite [7]

reduce to

e; J Pe
fh = fh = (1 + Vm) - - ( v 1 2 ) -f3m (4.55)
[ E¡ Pm
84 /NTRODUCT/ON TO COMPOSITE MATERIALS DESIGN

where Pe and Pm are the densities of the composite and the matrix, respectively, and

f3m is the moisture expansion coefficient of the matrix. Predictions using (4.55) are

shown in Figure 4 . 1 1 .

Toe coefficient of moisture expansion {3q of a composite with randomly oriented

fibers can be computed in the same way as the coefficient of thermal expansion in the

previous section; that is,

f31 + f32 {3¡ - f3i E¡ - E2


{3q = + ---------- (4.56)
2 2 E1 + ( 1 + 2 v 1 2 ) E 2

where {3 ¡ , f32, E¡, E2, and v12 are the properties of a fictitious unidirectional !ayer

with the. same fiber volume fraction of the random composite. This formula has not

been thoroughly validated with experimental data.

4.3.3 Transport Properties

Thermal and electrical conductivities and mass diffusivity all are computed by the same

equations. In this section k¡ and km indicate either thermal or electrical conductivities,

or mass diffusivity, of the fiber and matrix, respectively. In the fiber direction, the

conductivity (or diffusivity) can be computed by the ROM

k, = k¡ V¡ + km Vm (4.57)

and in the transverse direction by the Halpin-Tsai approximation [10]

1 + �17V¡ J
k2 = km
[
1 - 17V¡

(k¡ / km) - 1
11 = � - - - -
(k¡ / k m ) + �

1
� = l o g - -fi" I o g � (4.58)

where a/b is the aspect ratio of the cross-sectional dimensions of the fiber, a along

the direction of heat (electrical) conduction, and b transverse to it. For circular fibers,

a/b = 1 and � = l . These formulas predict well the thermal and electrical conductiv­

ities of carbon-epoxy composites for V¡ up to 60%. Predictions using these formulas


6
are shown in Figure 4 . 1 1 for a composite with f3m = 0 . 6 , a¡ = 5 . 4 1 0 - / º C , am = 60
6
' 10- 1°c, PJ = 2.5 g/cc, Pm = 1.2 g/cc, Kj = 1.05 W/rn/ºC, Km = 0.25 W/rn/ºC,

E ¡ = 7 2 . 3 GPa, Em = 5.05 GPa, v¡ = 0.22, Vm = 0 . 3 5 , and � = l .

4.4 STRENGTH

Predictive models for composite strength are not nearly as successful as sorne of the

stiffness models presented earlier. The reasons are many, and research is very active in

this area. There are two issues of concern. One is how to establish ultimate strengths
MICROMECHANICS 85

70

60

� 50
n,

e.
(/) 40
(/)

(!)
....
éi5 30
....
ctl
(!)

.e:
(/) 20
• Gage #1

O Gage#2

-- Data Fil
10

o
o.o 0.5 1.0 1.5 2.0 2.5 3.0

% strain

Figure 4.12 Inplane shear stress-strain behavior of unidirectional carbon-epoxy (ex­

perimental data according to ASTM D5379).

(strains) using micromechanics, which is the purpose of this section. The other is how

to use the strength values in conjunction with stress analysis, which is the objective

of Chapter 7.

4.4.1 Longitudinal Tensile Strength

The simplest model for tensile strength of a continuous fiber-reinforced composite is

derived by assuming that ali the fibers have the same tensile strength. -Actually, the

tensile strength of fibers is not uniform. Instead, it is well approximated by a Weibull

distribution [13]. But as a first approximation, it is assumed that ali the fibers have

the same strength, equal to the average of the distribution, which is the fiber average

strength u¡ a .

Toe second assumption is that both the fibers and the matrix behave linearly up to

failure. This is not true for most polymer matrices that exhibit either elastic nonlinear

(Figure 4 . 1 2 ) or plastic behavior after a certain elongation. Toe behavior of polymers

is further complicated by their load-rate dependency. That is, polymers are viscoelastic

or viscoplastic (see section 2 . 2 . 3 ) .

The third assumption is that the fibers are brittle with respect to the matrix. As it

can be seen in Tables 2 . 1 and 2.4, E-glass has actually more elongation to failure than

most resins, invalidating this assumption. However, such large elongations to failure

for fibers hold only in ideal conditions, with sorne fibers breaking at much lesser strain

levels.
86 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

Fiber Failure

Matrix Failure
<J;,,u

E
Eta tmu

Figure 4.13 Micromechanics of strength assuming uniform fiber strength.

Toe fourth assumption is that the fibers are stiffer than the matrix. This is valid

for ali cases except for ceramic matrix composites (CMC).

Under the assumptions listed, the composite will usually break when the stress

in the fibers reaches their strength ªfa. After the fibers break, the matrix is unable to

carry the load. Therefore, the composite elongation to failure Ecu is equal to the fiber

elongation to failure Efu. At this strain leve!, the matrix has not failed yet because it is

more compliant and can sustain larger strains (Figure 4 . 1 3 ) . Under these conditions, it

can be assumed that the longitudinal tensile strength is controlled by the fiber strength

and is represented by

F1r = ªfa Vf + a ; ( l - Vf) (4.59)

where the stress in the matrix at failure (see Figure 4 . 1 3 ) is

(4.60)

Finally, the tensile strength is

F1r = ªfa [Vf + !; ( 1 - Vf)] (4.61)

This equation assumes that the strain in the matrix and the fibers are the same,

which is true if the fiber-matrix bond is perfect. The ultimate strain or stress of the

matrix is not realized, because the fibers are more brittle ( i . e . , fail at a lower strain,
88 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

been obtained from tests on single fibers in ideal conditions, usually represent the

maximum attainable rather than the typical strength and, therefore, cannot be used in

design.

Since fibers are damaged during processing, it is more advisable, and often done,

to use (4 . 6 1 ) with experimental average ten sil e strength values af a . These average

values (afa) have been back calculated, using (4.61), from experimental data on

composites rather than from tests on single fibers. Therefore, the data does represent

the fiber material in the real application. If sorne testing can be afforded, an average

fiber strength value can be determined far a given Vf. Then it can be used for other

values of Vf as long as the processing method is not changed. Experimental values

of tensile strength of composites can be obtained following AS1M D3039.

Example 4.5 Compute the tensile strength of a Ti-alumina metal matrix composite

(MMC) with a fiber volume fraction of 27%. The titanium matrix is Ti 6-4, which

at 427ºC has Em = 87.5 GPa, Vm = 0.3. The alumina fibers have a diameter of

df = 140 microns, and at 427ºC have Ef = 379 GPa, Vf = 0.27, and average

strength ªfa = 2614 MPa.


Using the strength of materials approach,

F11 = ªfa Vf + a ; ( l - Vf)

* Em 87.5
ªm = ª f a - = 2614-- = 603 MPa
Ef 379

F1t = 2614(0.27) + 6 0 3 ( 1 - 0.27) = 1 1 4 6 MPa

4.4.2 Longitudinal Compressive Strength

Compressive strength of continuous fiber-reinforced polymer matrix composites (PMC)

is lower than the tensile strength, about one half or less. Toe mode of failure is usually

triggered by fiber microbuckling, when individual fibers buckle inside the matrix (Fig­

ure 4 . 1 5 ) . The buckling process is controlled by fiber misalignment, shear modulus,

and shear strength of the composite.

Fiber rnisalignment measures the waviness of the fibers in the composite. Fiber

waviness is always present to sorne extent, even when great care is taken to align

the fibers during processing. Waviness occurs because of severa! factors. Toe fibers

are wound in spools as soon as they are produced, which induces a natural curvature

in the fibers. Then fibers tend to curl when stretched on a flat mold. Furthermore,

many fibers are wound together over a spool to form a tow or roving during fiber

production. The fibers wound on the outside of the spool are longer than those wound

in the inside. When the tow is stretched, the longer fibers are loose and microcatenary

is formed.

Microcatenary can be shown by stretching a tow horizontally and letting the

longer fibers hang under their own weight. Toe longer fibers hang in a catenary shape,

just as electrical power lines do. In the final composite part, microcatenary appears as

fiber misalignment.
MICROMECHANICS 89

Fiber

Fiber

buckling

(a) ln-phase (shear) mode (b) Out-of-phase mode

Figure 4.15 Modes of fiber microbuckling.

Finally, there is the shrinkage of the polymer. During cure, most thermoset poly­

mers shrink 3 to 9% by volume, that is, 1 to 3% on any direction. Along the fiber

direction, it means that the fiber, which <loes not shrink during cure, must accommo­

date the strain in the form of waviness.

The amount of fiber misalignment varíes from fiber to fiber, and it can be accu­

rately measured by the procedure given in [ 1 4 ] . The measurement technique consists

of cutting a sample at an angle cp with respect to the fiber orientation as illustrated

in Figure 4 . 1 6 . After polishing the sample and looking at it with a microscope, each

fiber shows as an ellipse. Toe longer dimension of the ellipse b is proportional to the

misalignment angle a through the equation

. d¡

sm(w) = b

a = w - cp (4.64)

wheEe q is tbe jj!JeE dúme/e¡; Tbo average, DI mean vaJut/ of aJJ the measured angles

will be zero since fibers are misaligned equally at positive and negative angles with

respect to the nominal fiber orientation. The experimental data of fiber angle can

be fitted accurately with a normal distribution [12]. The standard deviation of the

distribution can be computed as [ 1 3 ]

n L7=t a¡,- (I:7=1 a¡)2

Ll = (4.65)
n(n - 1)

where n is the number of observations of cq . Toe normal probability density p(a)

shown in Figure 4 . 1 7 represents the observed data and indicates the probability of

finding a particular value of misalignment a in the cross section of the composite.


90 INTRODUCTION TO COMPOSITE MATERIALS DESJGN

. a .

composite

sample �

Figure 4.16 Misalignment measuring technique.

0.15 ,-------------;::::==========�
- Probability Density

¡zz:;¡ Frequency Data


Q.

� 0.10
·¡¡;

e:
Q)


zs
_g o.os

e
a.

-2.0 -1.5 -1.0 -0.5 o.o 0.5 1.0 1.5 2.0

Misalignment Angle r,JQ

Figure 4.17 Probability density function p(a), p r o p o r t i o n a lto the number of fibers

having a misalignment angle a.


MICROMECHANICS 91

The first formula for compressive strength was proposed by Rosen [ 1 6 ] , recogniz­

ing the fact that compression failure is triggered by fiber microbuckling. When fibers

buckle, they can ali buckle in phase (shear mode, Figure 4 . 1 5 a ) or out of phase (ex­

tension mode, Figure 4 . 1 5 b ) . It can be demonstrated that the shear mode will always

occur at a lower compressive stress for PMCs having practica! values of fiber volume

fraction.

The first approximation to the problem is to assume that the buckling load of

the fibers is the limiting factor for compressive strength. To obtain the buckling load,

Rosen [ 1 6 ] performed a stability analysis of straight fibers laterally supported by the

matrix. Furthermore, the shear stress-strain law of the composite was assumed to be

linear, i . e . , T = G 12 y . Actually, PMCs have a nonlinear stress-strain law, as shown

in Figure 4 . 1 2 . Such behavior is better represented by the relationship

(4.66)

Even though fibers are wavy, they are assumed to be straight in Rosen's model

because this is the simplest approach to stability problems. The stability analysis is

performed on what is called the perfect system (straight fibers) [26]. The buckling

load is found for the perfect system, and it is assumed that the load capacity of the

imperfect system (wavy fibers) is lower than the buckling load of the perfect system.

The buckling stress is then

Gm �
0-CR = --- = G12 (4.67)
1 - V¡

This formula overpredicts compression strength as much as 200%. According to

this model, the buckling stress is numerically equal to the matrix dominated approxi­

mation of the composite shear modulus (4.24). An empírica! correction of (4.67) can

be made by adding a factor k, which is determined from one test at one particular

value of the fiber volume fraction, and assuming that the value of the factor k does

not vary with the fiber volume fraction

kGm
ªCR = --- � k G 1 2 (4.68)
1 - V¡

This approach has been validated for boron-epoxy composites [ 17]. Toe main

problem with using the empirical correction is that testing must be done, and testing

for compressive strength is very difficult. In fact, no single test method can be used for

ali materials, or even for various thicknesses of samples of the same material. Further­

more, composites from actual production have vastly different values of compressive

strength than composites made in the lab, even if the same materials are used. Also,

samples cannot be machined out of production parts without introducing significant

damage that affects the results. Therefore, there is a strong motivation for deriving a
92 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

1.0

o.a

-O- Buckling Stress

-- Applied Stress

0.6 -'9--- F(a)


N
,....
CJ

-t)
0.4

0.2

o.o
o 2 3 4 5 6

Misalignment Angle aIO

Figure 4.18 Buckling stress and applied stress as a function of fiber misalignment.

formula that can be related to parameters that can be easily measured on composites

manufactured in the shop rather than from samples made in the lab.

A better estímate of the buckling stress can be made by taking into account the

initial misalignment of the fibers and the nonlinear stress-strain law given by (4.66).

The buckling stress of a fiber bundle in which ali the fibers have the same misalignment

is given by [ 1 2 ]

2 2
F6 (,/2 - l ) ( e ../2 g - e g) + (,/2 + l)(e< +../2)g - 1 )

u ( a , y) = 2(y + a) 1 + e2g + e../2g + e<2+../2)g

g = y G 1 2 / F6 (4.69)

where G12 and F6 are the inplane shear modulus (section 4.2.4) and inplane shear

strength (section 4 . 4 . 5 ) , respectively.

Toe stress in the composite is a function of the misalignment a and the shear

strain y , induced when the fibers buckle in phase (Figure 4 . 1 5 a ) . The maximum value

of (4.69) is the buckling stress u(a) of a material in which fibers have the same

misalignment a (Figure 4 . 1 8 ) . However, in a real composite not ali the fibers have the

same value of fiber misalignment. Instead, a normal probability distribution (Figure

4 . 1 7 ) gives the probability of finding a fiber with any angle a.

A formula for the compressive strength of the composite can be derived by taking

into account that the fibers with larger value of misalignment will buckle first during

the loading process. Once a fiber has buckled, it carries negligible stress, thus over-
MICROMECHANICS 93

loading those fibers that have lower misalignment. Toe unbuckled fibers are able to

carry the extra load only up to a certain point. With reference to Figure 4 . 1 8 , buckling

of the fibers proceeds from large values of misalignment, both positive and negative,

toward lower values of misalignment. Toe number of unbuckled fibers is proportional

to the area under the probability density curve (Figure 4 . 1 7 ) , which is given by

F(a) = L: p(a')da' = erf(z)

tx

z=-- (4.70)
vrin

where erf(z) is the error function [18, 19] and Q is the standard deviation of fiber

misalignment given by (4.65). At any given point in the loading process, the applied

stress (see Figure 4.18) is equal to the product of the actual stress carried by the

unbuckled fibers (and surrounding matrix) times the area of the unbuckled fibers (and

surrounding matrix)

a= u(a)F(a) (4.71)

where the buckling stress u(a) is shown by Figure 4 . 1 8 . The maximum stress that

can be applied is the compressive strength of the material

F1c = max(a) (4.72)

which is given in [23] as the explicit formula

F1c X a
=
G12 24rr b

103961
2 4 3
a = -10979.6 - 8432.03x - 19037.205 x - 124.653 x - � x

2 3
x2
+ (12191.01 + 1881.87 x + 176.286 x + 7979.978x) + 2.356x
2

3 2
x2
b = (-7.146x - 41.298x + 5.608x) +2.356x
2

2
2
+ ( 10.106 x + 34.594x - 6 1. 3 8 9 ) ( x2 + 2.356 x ) (4.73)

with the dimensionless group x given by

(4.74)
94 INTRODUCTION TO COMPOSITE MATER/ALS DESIGN

Table 4.3 Comparison of predicted and experimental values of

compressive strength for various composite materials

Material V¡ G12 F6 n X
Exp, (4.75)

[%] [MPa] [MPa] [deg] F1c/G12 F1c/G12

XAS/914C 60 5133 125 1.01 0.724 0.352 0.357

AS4/PEEK 61 5354 157.5 1.53 0.908 0.310 0.315

AS4/E7K8 60 7923.5 90.95 1.18 1.793 0.213 0.211

Glass-vinyl ester 43 4223 54.77 3.3 4.441 0.124 0.118

Glass-polyester 40.2 3462 40.57 3.45 5.148 0.138 0.107

where Q is the standard deviation of fiber misalignment in radians, G 12 is the shear

modulus, and F6 is the shear strength of the composite.

Taking into account that the dimensionless compressive strength can be modeled

in terms of the dimensionless number x, the following formula is proposed:

(4.75)

where a = 0.21 and b = -0.69 are two constants chosen to fit the explicit equation

(4.73). Note that (4.75) is not an empirical equation, since no experimental data of

compressive strength has been used in its derivation. Comparison between experimen­

tal data and values obtained with (4.75) is shown in Table 4 . 3 . Note that (4.75) and the

procedure explained in this section do not apply to MMC or when the misalignment

is nearly nonexistent, as it is in the case of sorne boron-fiber composites, because in

those cases the mechanism of compression failure is vastly different.

Experimental values of compressive strength can be obtained using a number

of different fixtures, including those described in ASTM D3410, D695 and SACMA

modified D695 among many others. An extensive evaluation of most of the existing

compression test methods can be found in [20, 2 1 ] .

Example 4.6 Estímate the effect of misalignment on the compressive strength of a

carbon-PEEK composite fabricated by hand lay-up of prepreg tape, vacuum bagged

and oven cured. Sample A was fabricated under controlled laboratory conditions,

resulting in a standard deviation of misalignment QA = 1 . 0 1 degrees. Sample B was

fabricated under normal processing conditions, resulting in S1s = 1 . 4 1 degrees. The

fiber is carbon AS4-D (V¡ = 0 . 5 ) and the matrix is epoxy with V¡ = 0.6.

From Tables 2.1 and 2.4, the properties of the carbon AS4-D fiber and the epoxy

matrix (9310/9360 at 23ºC) are E¡ = 241 GPa, v¡ = 0.2 (assumed), Em = 3 . 1 2 GPa,

Vm = 0 . 3 8 (assumed), Umu = 7 5 . 8 MPa. The compressive strength is given by (4.74)

and (4.75). Therefore, the inplane shear stiffness and strength must be determinedfirst.

!
·[
MICROMECHANICS 95

Assuming both constituents to be isotropic, which is an approximationfor carbonfiber,

the shear moduli are

241
G¡ = = 100 GPa
2(1 + 0.2)

3.12
Gm = = 1 . 1 3 GPa
2(1 + 0.38)

Using (4.25), the inplane shear modulus is

+ 0.6) + ( 1 - 0 . 6 ) 1. 1 3 / 1 0 0 ] G
G 1 2 = 1. 1 3 = 4 . 34 Pa
[(l
1 - 0.6) + (1 + 0.6)1.13/100

The shear strength can be estimated by (4.79), but the transverse modulus E2

needs to be evaluatedfirst. Using (4.19),

241/3.12 - 1
9 62
n = 241/3.12 + 2 = º·

1 + 2 x 0.962 x 0 . 6 ] GP.
E2 = 3 . 1 2 = 1 5 .9 a
[
1 - 0.962 X 0.6

Then, using (4.79), assuming Vmu = O"mu and no voids (Cv = 1.0),

11·�)
F6 = 75.8 [1 + ( o . 6 - Jo.6 ) ( 1 - J = 62.7 MPa

Far sample A, using (4.74) and (4.75), the compression strength F1c is

4, 340 X liO X 1.01


X = = 1.22
62.7

1.22 )-0.69
F1c = 4, 340
(
- + 1 = 1, 155 GPa
0.21

Far sample B, X = 1 . 7 0 and F1c = 945 GPa. These values are clase to typical

experimental data.

4.4.3 Transverse Tensile Strength

The tensile strength in the direction perpendicular to the fibers F21 is controlled by the

matrix strength, the fiber-matrix interface strength, and defects in the matrix phase,

such as voids and microcracks. The transverse strength of the composite can be higher

or lower than the tensile strength of bulk matrix. It can be higher because many

defects, such as voids and cracks, are present in bulk matrix, thus making the bulk

matrix appear as a brittle and weak material. When the composite is made, defects
96 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

in the matrix between the fibers can be less than in bulk because the small space

between the fibers <loes not leave space for nucleation of voids. But, depending on

the manufacturing conditions, defects can still be present. Since the fibers are usually

much stiffer than the matrix, and they may be very close or even touching each other,

they induce stress concentrations in the matrix that may cause premature failure. In

this case, the strength of the composite can be lower than that of the bulk matrix.

None of the analytical models in the literature can predict the transverse strength

accurately. Two empirical formulas are presented here because of the lack of more

accurate formulas. The empirical formula proposed by Nielsen [24] can be further

modified with a reduction coefficient Cv to account for voids

(4.76)

where Umu is the tensile strength of the bulk matrix, E2 is the transverse modulus of

the composite, and Em is the elastic modulus of the matrix. The reduction coefficient

is also empirical [ 1 O]

e; = 1 - (4.77)
n (1 - V¡)

where Vv is the void volume fraction. Another empirical formula, proposed by Chamis

[22], can also be modified to account for voids using the same empirical coefficient

Cv, resulting in

(4.78)

where E¡ is the elastic modulus of the fiber. Both equations usually predict values

lower than the strength of the matrix, and (4.76) yields lower values for most materials.

The effect of voids is very detrimental to the transverse strength and this is reflected

by the empirical formulas. The results provided by these formulas can be used for

preliminary design, but experimental data are usually required if transverse strength is

the controlling mode of failure of the component. Experimental values can be obtained

by ASTM 03039.

4.4.4 Transverse Compressive Strength

The same comments made for transverse tensile strength (section 4.4 .3) apply in this

case. Only empirical formulas are available and these may not yield accurate predic­

tions. For preliminary design, the compressive strength in the direction perpendicular

to the fibers F2c may be estimated using the same equations as the previous section

(4.4.3), replacing the bulk tensile strength of the matrix Umu by the bulk compres­

sive strength of the matrix O-mue [28]. Transverse compressive strength values are

higher than tensile strength for both matrix and composite. Experimental values can

be obtained by ASTM 0695, 0 3 4 1 0 , etc.


MICROMECHANICS 97

(a)

(b)

(e)

Figure 4.19 Shear components and their relationship to shear failure modes.

4.4.5 lnplane Shear Strength

Clear distinction between the various shear stress components, illustrated in Figure

4.19, must be made. The first subscript of a shear stress indicates the direction of

the normal to the plane on which the stress acts. The second subscript indicates the

direction of the stress. In Figure 4. l 9a, the normal to the plane considered is along

the 1-axis. Both r12 and r13 would have to shear off the fibers to produce failure,

which is very unlikely to occur. In Figure 4 . 1 9 b , the normal to the plane considered is

along the 3-axis. Both r31 and r32 produce splitting of the matrix without shearing off

any fibers. In Figure 4 . 1 9 c , both r21 and r23 produce splitting of the matrix without

shearing off the fibers.

Since the stresses r21 are always equal to r12 to satisfy equilibrium, both stresses

are called ª6· The corresponding shear strength, called inplane shear strength, is F6.

The composite fails when the inplane stress r21 reaches its ultimate value F6 = r21u­

At this point matrix cracks appear along the fiber direction.

To illustrate further this situation, consider a tensile test of a composite as shown

in Figure 4.20, with the fibers oriented at 45° with respect to the loading direction. lt

is well known that at 45° the shear stress is half of ax. Along the planes AB and CD,
98 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

Figure 4.20 Tensile test of [ -45] !ayer illustrating the most likely plane of shear

failure.

the material fails when 1 0'6 1 > r21u. Along AD and BC, the material would fail at a

higher load when 1 0'6 I > r12u.5

Toe same comments made for transverse tensile strength (section 4.4.3) apply in

this case. Only empirical formulas are available, and these may not yield accurate

predictions. For preliminary design, the inplane shear strength F6 may be estimated

using an equation similar to (4.78), replacing the bulk tensile strength of the matrix

O'mu with the bulk shear strength of the matrix !mu as follows

(4.79)

5
1 1 denotes absolute value, which is used because the shear strength is independent of the sign of

the shear stress in the material coordinate system. Interaction between U6 = ox /2 and normal stresses

cr¡ = cr2 = ox /2, which are also present, will be addressed in Chapter 7.
MICROMECHANICS 99

Lacking experimental values for Tmu, (4.79) works reasonably well assuming

Tmu = <Jmu· Of course, (4.79) can be used to back calculate an effective value of Tmu

from composite shear strength data F6 at a particular value of fiber volume fraction V¡.

Then (4.79) can be used to predict F6 for other values of V¡ as long as the processing

conditions do not change significantly. Experimental values can be obtained by AS1M

0 3 5 1 8 , 05379, and 04255.

4.4.6 Interlaminar Shear Strength

Following the discussion in section 4.4.5, the shear strengths F4 and F5 are discussed

in this section. As it is clear from Figure 4 . 1 9 , the shear stresses r23 and r32 cause

splitting of the matrix without shearing off any fibers. Since, by equilibrium, the shear

stresses r23 and r32 are numerically equal, both are called <J4. The corresponding shear

strength is F4.

Toe interlaminar shear strength F4 is a matrix dominated property because the

shear acts on a plane perpendicular to the fiber direction. In this case the fibers do not

resist shear. On the contrary, the cross sections of the fibers can be viewed as circular

inclusions creating stress concentrations in the matrix, thus debilitating the composite.

It can be seen in Figure 4 . 1 9 that the shear stress r3¡ tends to split the matrix

along the fiber direction. This is the usual mode of failure under stress <J5. On the

other hand, the fibers would have to be sheared off by stress r¡3 to produce failure,

which is unlikely. Since, by equilibrium, the shear stresses rrs and r3¡ are numerically

equal, both are called <J5. The corresponding shear strength is F5, which is numerically

equal to the ultimate value of r31u. The shear stress <J5 applies shear along the fiber

direction. Therefore, the interlaminar shear strength Fs is affected by the fiber-matrix

bond strength.

Because of the lack of predictive formulas, both values (F4 and Fs) are approxi­

mated in preliminary design by the shear strength of bulk matrix. Experimental values

can be obtained by AS1M 02344, 04475, 0 3 9 1 4 , 03846, and 05379.

When the whole thickness of the composite is unidirectional and homogeneous,

there should be no difference between Fs and F6. This is true for a single !ayer of uni­

directional prepreg for a unidirectional pultruded material. However, when transverse

shear (<J4 or <J5) is applied to a laminate with distinct interfaces between layers (such

as prepreg lay-up), the resin rich interfaces may fail first. In this case, the interlaminar

shear strength values F4 and Fs may be lower than the inplane shear strength F6.

4.4. 7 Chopped and Continuous Strand Mat

Chopped strand mat and continuous strand mat (CSM) are considered randomly ori­

ented composites, even if in practice there is sorne preferential orientation of the fibers.

The resulting composite is considered to be quasi-isotropic (see section 6 . 3 . 5 ) . That

is, the properties are the same along any orientation on the surface of the layers. The

tensile strength Fcsm-t may be estimated by the empirical formulas [25]


100 INTRODUCTION TO COMPOSITE MATERIALS DESIGN

Fcsm-t = --
4aF21 [
1 +
1
-In
(
-
F11 )] for a <
2
- -
n 2 a F21

Fcsm-t = --
4F21 !fii
--
1 for Cl >
!fii1
-
n F21 F21

(4.80)

where F¡1, F21, and F6 are the longitudinal tensile, transverse tensile, and inplane

shear strength of a fictitious unidirectional material containing the same fiber volume

fraction as that of the CSM material. The formula compares well with experimental

data for sorne PMC with fiber volume fraction up to 20% [25].

There are no formulas available for compressive, inplane shear, and interlaminar

shear strengths of random composites. The accompanying software allows the user to

input experimental values, but as a default it assumes the following: the compressive

strength is assumed to be egua! to the tensile strength, the inplane shear strength is

taken as one-half of the tensile strength, and the two interlaminar shear strengths are

approximated by the shear strength of the matrix.

EXERCISES

4.1 Draw a representative volume element (RVE) for an epoxy matrix filled with cylindrical

steel rods arranged in a rectangular array. Compute the volume fraction V¡ as a function

of the fiber diameter d¡ and the spacing between the centers of the fibers ax and ay in

the two directions (center to center).

4.2 Write a definition for representative volume element. Draw a set of fibers in a hexagonal

array: six fibers at the vertices of an hexagon plus one in the center, with the fibers

touching each other. Draw the smallest RVE for this arrangement. Compute the fiber

volume fraction.

4.3 Consider the following material called carbon-epoxy with a fiber volume fraction of

70% and E¡ = 379 GPa, v¡ = 0.22, Em = 3 . 3 GPa, and Vm = 0 . 3 5 . Compute E¡,

E 2, v 1 2 , and G12 using both the rule of mixtures (ROM) and more accurate formulas

recommended in this chapter. Consider the results and comment on which values are

lower or upper bounds to the real value of each elastic property and what seems to be

the best model in each case.

4.4 The amount of fibers in a composite sample can be determined by a bum-out test. The

bum-out eliminates ali the resin and only the fibers remain. A composite sample plus

its container weighs 5 0 . 1 8 2 grams befare bum-out and 49.448 grams after bum-out.

The container weighs 47 .650 grams. Compute the fiber weight fraction WF and matrix

weight fraction WM .
MICROMECHANICS 101

4.5 Compute E¡, E2, G12, and V¡z given E¡ = 230 GPa, Em = EJl50, GJ = EJ/2.5,

Gm = Em/2.6, VJ = 0.25, Vm = 0.3, and VJ = 40%. The fibers have circular cross

section. Assume there are no voids.

4.6 Compute ali the elastic properties ( E 1 , E2, G 1 2 , G23, V12) for a unidirectional lamina

for the following material combinations:

(a) E-glass-polyester(isophthalic),

(b) S-glass-:-epoxy(93 l 0),

(e) Carbon(T300)-vinyl ester,

with VJ = 0.55. If actual data of Poisson's ratio for either fiber or matrix is not

available, take VJ = 0.22, and/or Vm = 0 . 3 8 . Compare the results using the strength of

materials (ROM) formulas (except for G23) and other formulas recommended in this

chapter.

4.7 Estímate F¡1 and F21 of AS4-D/PEEK with VJ = 0.5 and negligible void content.

4.8 Estímate the tensile strength of Kevlar-epoxy composite with VI = 0.5 at two tempera­

tures, 2 1 ºC and 149ºC, using the strength of materials approach. Use data from Tables

2.1 and 2.4 and section 2 . 1 . 1 . Assume that Em decreases linearly with temperature up

to 75% reduction at 1 7 1 º C . The void content is negligible.

4.9 With reference to Example 4.6, s e l e c t a different matrix in arder to achieve a compressive

strength of at least 1 , 100 MPa. Assume that the standard deviation of fiber misalignment

remains constant at the production value of 1 . 4 1 degrees. The fiber type and fiber volume

fraction remain unchanged.

4.10 Toe following data have been obtained experimentally for a composite based on an

unidirectional carbon-epoxy prepreg (MR50 carbon fiber at 63% by volume in LTM25

epoxy). Determine if the restrictions on elastic constants are satisfied:

E¡ = 156.403 GPa E2 = 7.786 GPa

V¡z = 0.352 Vz¡ = 0.016

G12 = 3.762 GPa

4.11 S e l e c t a m a t ri x , fiber, and fiber volume fraction to obtain a material with E 1 > 3 0 GPa

E1/E2<3.5.

4 . 1 2 Give an approximate value for the compressive strength of an Al-boron composite with

VJ = 0.4. Aluminum 2024 has E = 71 GPa, Poisson = 0.334, G = 26.6 GPa, yield

strength = 76 MPa, fatigue limit = 90 MPa, elongation = 22%. Boron fibers have

E = 400 GPa, ultimate strength = 3 . 4 GPa, Poisson = 0.25. Note that boron fibers

tend to be perfectly straight and that alumin um has a linear shear stress-strain plot

befare it yields.
102 INTRODUCT/ON TO COMPOSITE MATERIALS DESIGN

4.13 Compute approximate values of E 1 , E2, G12, v 12 for the composite of Exercise 4 . 1 2 .

Compare the results obtained using the strength of materials approach with the results

obtained using more accurate formulas which are programmed in the accompanying

software. Comment on the results.

4.14 Compute the tensile strength for the composite of Exercise 4 . 1 2 assuming that ali the

fibers have the same tensile strength.

REFERENCES

[l] Mase, G. E., and Mase, G. T., Continuum Mechanics for Engineers, CRC Press, Boca Raton, FL

(1992).

[2] Luciano, R., and Barbero, E. J . , Formulas for the stiffness of composites with periodic microstruc­

ture, l. J. Solids Structures, 3 1 , 2933-2944 (1994).

[3] Halpin, J., and Tsai, S. W., Effects of Environmental Factors on Composite Materials, Air Force

Materials Lab-Technical Report 67-423, Department of Defense, USA (1969).

[4] Jones, R. M . , Mechanics of Composite Materials, Taylor & Francis, Washington, D.C. (1975).

[5] Hashin, Z., and Rosen, B . W., The elastic moduli of fiber-reinforced materials, J. Applied Me­

chanics, June, 223-230 (1964).

[6] Aboudi, J., Mechanics of Composite Materials, Elsevier, Amsterdam ( 1 9 9 1 ) .

[7] Tsai, S. W., and Hahn, H. T., Introduction to Composite Materials, Technomic, Lancaster, PA

(1980).

[8] Hull, D., An Introduction to Composite Materials, Cambridge University Press, Cambridge

(1981).

[9] Halpin, J . , and Pagano, N. J., The laminate approxirnation for random oriented composites, J.

Composite Materials, 3, 720-724 (1969).

[10] Agarwal, B. D., and Broutman L. J . , Analysis and Performance of Fiber Composites, 2nd Ed.,

John Wiley & Sons, New York (1990).

[11] Kelly, K., and Barbero, E. J., The effect of fiber damage on the longitudinal creep of CFMMC,

L J. Solid Structures, 30, 3417-3429 ( 1 9 9 3 ) .

[12] Barbero, E. J., and Tomblin, J. S., A damage mechanics model for compression strength of

composites, l. J. Solids Structures, 33, 4379-4393 (1996).

[13] Hahn, G. G . , and Shapiro, S . S., Statistical Models in Engineering, John Wiley, New York (1967).

[14] Yurgartis, S. W., Measurement of small angle fiber misalignment in continuous fiber composites,

Composite Science and Technology, 30, 279-293 (1987).

[15] Rosen, B . W., The tensile failure of fibrous composites, AIAA J . , 2, 1 9 8 5 - 1 9 9 1 (1964).

[16] Rosen, B . W., Fiber Composite Materials, American Society for Metals, Metals Park, OH (1965).

[17] Lager, J. R., and June, R. R., Compression strength of boron-epoxy composites, J. Composite

Materials, 3, 48-56 (1969).

[18] Spiegel, M. R., Math Handbook, Schaum's Series, McGraw-Hill, New York (1968).

[19] Erdelyi, A., Higher Transcendental Functions, Vol. 2, McGraw-Hill, New York ( 1 9 5 3 ) .

[20] Haberle, J. G., Strength and Failure Mechanisms of Unidirectional Carbon Fibre-Reinforced

Plastics, Thesis, Imperial College, London ( 1 9 9 1 ) .

[21] Daniels, J. A,, and Sandhu, R. S . , Evaluation of compression specimens and fixtures for testing

unidirectional composite laminates, pp. 103-123 in Composite Materials: Testing and Design,

Vol. 1 1 , ASTM STP 1206, American Society for Testing and Materials, Philadelphia, PA (1993).

[22] Chamis, C. C., Simplified composite micromechanics equations for hygral, thermal, and mechan­

ical properties, SAMPE Quarterly, April, 14 (1984).

[23] Barbero, E. J., Prediction of compressive strength of polymer matrix composites, J. Composite

Materials, 32, 483-502 (1997).

[24] Nielsen, L. E., Mechanical properties of particulate-filled systems, J. Composite Materials, 1,

1 0 0 - 1 1 9 (1967).
MICROMECHANICS 103

[25] Hahn, H. T., On approximation for strength of random fiber composites, J. Composite Materials,

9, 316-326 (1975).

[26] Godoy, L. A., Theory of Stability: Analysis and Sensitivity, McGraw-Hill, New York (1999).

(27] Schapery, R. A . , Thermal expansion coefficients of composite materials based on energy princi­

pies, J. Composite Materials, 2, 380-404 ( 1 9 6 8 ) .

[28] Chamis, C. C., Simplified composite micromechanics equations for mechanical, thermal, and

moisture-related properties, in Engineer's Guide to Composite Materials, J. W. Weeton, D. M.

Peters, and K. L. Thomas, eds., American Society for Metals, Metals Park, OH (1987).

S-ar putea să vă placă și