Professional Documents
Culture Documents
MATERIALS SCIENCE
C16: Composite Materials
Name............................. College..........................
Prof. T. W. Clyne
Lent Term 2014-15
2014 15
C16H1
Overview of Composites Usage. Types of Reinforcement and Matrix. Carbon and Glass Fibres. PMCs,
MMCs and CMCs. Aligned Fibre Composites, Woven Rovings, Chopped Strand Mat, Laminae and
Laminates.
Recap of Axial and Transverse Youngs Moduli for an Aligned Long Fibre Composite, derived using the
Slab Model. Errors for Transverse Loading and Use of Halpin-Tsai Equations. Derivation of Shear
Moduli and Poisson Ratios. Number of Elastic Constants for Systems with different Degrees of Symmetry.
Plane Stress Loading of a Uniaxial Lamina and the Kirchoff Assumptions. Off-axis loading of a Lamina.
Elastic Constants as a Function of Loading Angle. Tensile-shear Interactions and Lamina Distortions.
The Shear Lag Model for Stress Transfer. Interfacial Shear Stresses. The Stress Transfer Aspect Ratio.
Stress Distributions with Low Reinforcement Aspect Ratios. Numerical Model Predictions. Hydrostatic
Stresses and Cavitation.
Lecture 6 - Short Fibre & Particulate Composites Stiffness & Inelastic Behaviour
Load Partitioning and Stiffness Prediction for the Shear Lag Model. Fibre Aspect Ratios needed to
approach the Long Fibre (Equal Strain) Stiffness. Inelastic Interfacial Phenomena. Interfacial Sliding and
Matrix Yielding. Critical Aspect Ratio for Fibre Fracture.
Interfacial Bonding Mechanisms. Measurement of Bond Strength. Pull-out & Push-out Testing. Control
of Bond Strength. Silane Coupling Agents. Interfacial Reactions and their Control during Processing.
Axial Tensile Strength of Long Fibre Composites. Transverse and Shear Strength. Mixed Mode Failure
and the Tsai-Hill Criterion. Failure of Laminates. Internal Stresses in Laminates. Failure Sequences.
Testing of Tubes in combined Tension and Torsion.
Energies absorbed by Crack Deflection and by Fibre Pull-out. Crack Deflection . Toughness of Different
Types of Composite. Constraints on Matrix Plasticity in MMCs. Metal Fibre Reinforced Ceramics.
Modes of Failure in Compression. Kink Band Formation. The Argon Equation. Prediction of
Compressive Strength and the Effect of Fibre Waviness. Failure in Highly Aligned Systems. Possibility of
Fibre Crushing Failure.
Thermal Expansivity of Long Fibre Composites. Transverse Expansivities. Short Fibre and Particulate
Systems. Differential Thermal Contraction Stresses. Thermal Cycling. Thermal Residual Stresses.
Misfit Strains in Substrate-Coating Systems. Force and Moment Balances. Relationship between Residual
Stress Distribution and System Curvature. Curvature Measurement to obtain Stresses in Coatings.
Limitations of Stoney Equation. Sources of Misfit Strain. Driving Forces for Interfacial Debonding.
C16H2
Booklist
D.Hull & T.W.Clyne, "An Introduction to Composite Materials", Cambridge University Press,
(1996) [AN10a.86]
Web-based Resources
Most of the material associated with the course (handouts, question sheets, examples classes
etc) can be viewed on the web and also downloaded. This includes model answers, which are
released after the work concerned should have been completed. In addition to this text-based
material, resources produced within the DoITPoMS project are also available. These include
libraries of Micrographs and of Teaching and Learning Packages (TLPs). The following TLPs are
directly relevant to this course:
Mechanics of Fibre Composites
Bending and Torsion of Beams
C16H3
Fig.1.1 Data for some engineering materials, in the form of a map of Youngs modulus against
density
C16H4
Fig.1.2 Effect of heat treatment temperature on the strength and Youngs modulus of carbon
fibres produced from a PAN precursor
TWC - Lent 2015
C16H5
Glass Fibres
Polymeric Fibres
Fig.1.3 Structures of (a) cellulose & (b) Kevlar (poly paraphenylene terephthalamide) molecules
TWC - Lent 2015
C16H6
Other Reinforcements
C16H7
C16H8
Fig.2.1 Schematic illustration of loading geometry and distributions of stress and strain, and
effects on the Youngs moduli and shear moduli, for a uniaxial fibre composite and for
the slab model representation
C16H9
Halpin-Tsai Expressions
Fig.2.2 Predicted dependence on fibre volume fraction, for the epoxy-glass fibre system, of
(a) transverse Youngs modulus and (b) shear moduli of long fibre composites
TWC - Lent 2015
C16H10
Poisson Ratios
Fig.2.4 Predicted dependence on fibre volume fraction, for the epoxy-glass fibre system, of the
three Poisson ratios
TWC - Lent 2015
C16H11
Fig.2.5 Examples of how 2-D relative displacement components can represent different
combinations of shear and rigid body rotation
C16H12
Fig.3.1 Indication of the form of the Spq and Cpq matrices (matrix notation for Sijkl and Cijkl
tensors), for materials exhibiting different types of symmetry. All of the matrices are
symmetrical about the leading diagonal
C16H13
Fig.3.2 (a) Relationship between the fibre-related axes in a lamina (1, 2 & 3) and the coordinate system (x, y & z) for an arbitrary in-plane set of applied stresses.
(b) Illustration of how such an applied stress state !ij (!x, !y & "xy) generates stresses in
the fibre-related framework of !ij (!1, !2 & "12)
TWC - Lent 2015
C16H14
For a thin lamina, stresses and strains in the through-thickness (3) direction are neglected, so
that the 3, 4, and 5 components in matrix notation are of no concern. Therefore, when a lamina is
loaded parallel or normal to the fibre axis, the strains that interest us are given by
S11
!1
#1
! 2 !=! S ! # 2 !=! S12
" 12
$ 12
0
S12
S22
0
0 #1
0 ! #2
S66 $ 12
(3.1)
S11 !=!
"
"
1
!!!!!!!!!!S12 !=!! 12 !=!! 21
E1
E2
E1
S22 !=!
1
1
!!!!!!!!!!S66 !=!
!
E2
G12
The first step in establishing the lamina strains for off-axis loading is to find the stresses,
referred to the fibre axis (!1, !2 and "12), in terms of the applied stress system (!x, !y and "xy).
This is done using the equation expressing any second rank tensor with respect to a new
coordinate frame
in which aik is the direction cosine of the (new) i direction referred to the (old) k direction.
Obviously, the conversion will work in either direction provided the direction cosines are defined
correctly. For example, the normal stress parallel to the fibre direction !11, sometimes written as
!1, can be expressed in terms of the applied stresses !'11 (= !x), !'22 (= !y) and !'12 (= "xy)
The angle # is that between the fibre axis (1) and the stress axis (x). Referring to the figure, these
direction cosines take the values
a11 =!cos ! !(= c)!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!a12 !=!cos ( 90 " ! )!=!sin ! !(= s)
!1
!x
! 2 !=! T ! ! y !
" 12
" xy
(3.2)
where
c2
T !=! s 2
s2
c2
!cs cs
2cs
!2cs
(c
! s2
!1
!x
! 2 !=! T ! ! y !
!12
! xy
TWC - Lent 2015
C16H15
However, to use engineering strains ($xy = 2%xy etc), T must be modified (by halving the
elements t13 and t23 and doubling elements t31 and t32 of the matrix T ), so as to give
!1
!x
! 2 !=! T # ! ! y !
" 12
" xy
(3.3)
in which
c2
T ! !=! s
s2
cs
"2cs 2cs
(c
"cs
2
" s2
The procedure is now a progression from the stress-strain relationship when the lamina is
loaded along its fibre-related axes to a general one involving a transformed compliance matrix,
S , which will depend on #. The first step is to write the inverse of Eqn.(3.3), giving the strains
relative to the loading direction (ie the information required), in terms of the strains relative to the
"1
fibre direction. This involves using the inverse of the matrix T ! , written as T !
!x
!1
$1
! y !=! T # ! ! 2 !
" xy
" 12
in which
c2
"1
T ! !=! s 2
s2
"cs
cs
2cs "2cs
(c
" s2
Now, the strains relative to the fibre direction can be expressed in terms of the stresses in those
directions via the on-axis stress-strain relationship for the lamina, Eqn.(3.1), giving
!x
%1
$1
! y !=! T # ! S ! % 2 !
" xy
& 12
Finally, the original transform matrix of Eqn.(3.2) can be used to express these stresses in terms
of those being externally applied, to give the result
!x
%x
%x
$1
! y !=! T # ! S ! T ! % y !!=! S ! % y
" xy
& xy
& xy
(3.4)
C16H16
(3.5)
C16H17
Fig.3.3 Variation with loading angle # of (a) Youngs modulus Ex and shear modulus Gxy and
(b) Poisson ratio $xy (using equal stress model), for a lamina of epoxy-50% glass fibre
C16H18
Fig.3.4 Variation with loading angle # of the tensile-shear interaction compliance S16, for a
lamina of rubber-5% Al fibre, and photos of 4 specimens (between crossed polars) under
axial tension, lined up at the appropriate values of #, showing tensile-shear distortions
C16H19
Fig.4.1 Schematic depiction of the loading angle % between the x-direction (stress axis) and the
reference direction (#=0), for a laminate of n plies. Also shown is the angle #k between
the reference direction and the fibre axis of the k th ply (1k direction)
TWC - Lent 2015
C16H20
Stiffness of Laminates
Fig.4.2 Variation with loading angle % (between the stress axis and the reference (#=0)
direction) of the Youngs modulus of a single lamina and of two simple laminates, made
of epoxy-50% glass fibre. (The equal stress model was used to obtain the transverse
Youngs modulus of the lamina, E2.)
TWC - Lent 2015
C16H21
Fig.4.3 Variation with loading angle % (between the stress axis and the reference (#=0)
direction) of the interaction ratio, &xyx (ratio of the shear strain 'xy to the normal strain
(x) of a single lamina and of three simple laminates, made of epoxy-50% glass fibre.
(The equal stress model was used to obtain the transverse Youngs modulus of the
lamina, E2.)
TWC - Lent 2015
C16H22
Fig.4.4 (a) Predicted stresses within one ply of a loaded crossply laminate (epoxy-50%glass)
and (b) a schematic of these stresses for loading parallel to one of the fibre axes
C16H23
Fig.4.5 Elastic distortions of a crossply laminate as a result of (a) uniaxial loading and
(b) heating
C16H24
Fig.5.1 Schematic illustration of the basis of the shear lag model, showing (a) unstressed system,
(b) axial displacements, u, introduced on applying tension parallel to the fibre and
(c) variation with radial location of the shear stress and strain in the matrix
C16H25
Derivation of Equations
The model is based on assuming that the build-up of tensile stress along the length of the fibre
occurs entirely via the shear forces acting on the cylindrical interface. This leads immediately to
the basic shear lag equation:
d! f
"2# i
!=!
dx
r0
(5.1)
The interfacial shear stress, "i, is obtained by considering how the shear stress in this direction
varies within the matrix as a function of radial position. This variation is obtained by equating the
shear forces on any two neighbouring annuli in the matrix:
"1
r
!=! 2
"2
r1
The displacement of the matrix in the loading direction, u, is now considered. The shear strain
at any point in the matrix can be written both as a variation in this displacement with radial
position and in terms of the local shear stress and the shear modulus of the matrix, Gm
#r &
" i !% 0 (
$ r'
"
du
! !=
!=
!!!!!!!!and!! !=!
Gm
Gm
dr
It follows that an expression can be found for the interfacial shear stress by considering the
change in matrix displacement between the interface and some far-field radius, R, where the
matrix strain has become effectively uniform (du/dr ! 0).
uR
!u
r0
du !=!
!!" i !=!
(u
" i !r0
Gm
r0
dr
r
# ur0 Gm
$ R'
r0 !ln & )
% r0 (
(5.2)
The appropriate value of R is affected by the proximity of neighbouring fibres, and hence by the
fibre volume fraction, f. The exact relation depends on the precise distribution of the fibres, but
this needn't concern us too much, particularly since R appears in a log term. If an hexagonal array
of fibres is assumed, with the distance between the centres of the fibres at their closest approach
being 2R, then simple geometry leads to
2
! R$
'
1
#" r &% !=! 2! f ! 3 !(! f
0
Substituting for "i in the basic shear lag equation now gives
!=! ! f =
C16H26
"f
!
Ef
while the differential of uR, ie the far-field axial strain of the matrix, can be taken as the
macroscopic strain of the composite
du R
!!! "1 !
dx
Differentiating the expression for the gradient of stress in the fibre and substituting these two
relations into the resulting equation, with the shear modulus expressed in terms of Young's
modulus and Poisson's ratio [Em=2 Gm (1+'m)], leads to
d 2! f
n2
!=! 2 !(! f !"!Ef ! #1 )
dx 2
r0
(5.3)
n=
(5.4)
" 1%
Ef (1 + ! m )!ln $ '
# f&
This is a second order linear differential equation of a standard form, which has the solution
# n!x &
# n!x &
! f !=!Ef ! "1 !+!Bsinh %
( !+!D cosh %
(
$ r0 '
$ r0 '
and, by applying the boundary condition of !f = 0 at x = L (the fibre half-length), the constants B
and D can be solved to give the final expression for the variation in tensile stress along the length
of the fibre
$ n!x '
,*
,.
! f !=!Ef ! "1 ! +1 # cosh &
) sech ( n!s ) /
,,0
% r0 (
(5.5)
in which s is the aspect ratio of the fibre (=L/r0). From this expression, the variation in interfacial
shear stress along the fibre length can also be found, using the basic shear lag equation, by
differentiating and multiplying by (-r0/2),
! i !=!
# n!x &
Ef !n! "1
sinh %
( !sech ( n!s )
2
$ r0 '
(5.6)
!1
$ tanh ( n!s ) '
=! f !Ef &1 #
+ (1 # f )!Em
"1
n!s )(
%
(5.7)
C16H27
Fig.5.2 Predicted (shear lag) variations in (a) fibre tensile stress and (b) interfacial shear stress
along the axis of a glass fibre in a polyester-30% glass composite subject to an axial
tensile strain of 10-3, for two fibre aspect ratios
TWC - Lent 2015
C16H28
(a)
(b)
Fig.5.3 Photoelastic (frozen stress) models under applied axial load, showing the stress field
in the matrix around two stiff reinforcements having the same aspect ratio, with
(a) cylindrical and (b) ellipsoidal shapes
Fig.5.4 Predicted (finite element) hydrostatic stress fields around sphere and cylinder (s=5) of
SiC in an Al matrix, with an applied axial tensile stress of 100 MPa (and differential
thermal contractions stresses corresponding to a temperature drop of 50 K)
TWC - Lent 2015
C16H29
Lecture 6:
Short Fibre & Particulate Composites - Stiffness & Inelastic Behaviour
Shear Lag Model Predictions for Stiffness
C16H30
Fig.6.2 A set of four (rubber 5% Al fibre) photoelastic models under axial load, showing how
the stress field and the axial extension change as the aspect ratio and degree of
alignment of the fibres are changed
C16H31
Fig.6.3 Plots of the dependence of peak fibre stress, !f0, (at the onset of interfacial sliding or
matrix yielding) on the critical shear stress for these phenomena, "i*. Plots are shown
for different fibre aspect ratios, with n values typical of polymer- and metal-based
composites. Also indicated are typical value ranges for fracture of fibres and for matrix
yielding and interfacial debonding
C16H32
C16H33
C16H34
Fig.7.1 Predicted stress distribution around and within a single fibre, in a polyester-35% glass
long fibre composite, as a result of differential thermal contraction (T drop of 100 K)
C16H35
Fig.7.4 Depiction of the action of silane coupling agents, which are used to generate improved
fibre-matrix bonding for glass fibres in polymeric matrices. The silane reacts with
adsorbed water to create a strong bond to the glass surface. The R group is one which
can bond strongly to the matrix
C16H36
Fig.7.2 Schematic stress distributions and load-displacement plot during single fibre pull-out
testing. The interfacial shear strength, "*, is obtained from the pull-out stress, !0,*
TWC - Lent 2015
C16H37
Fig.7.3 Schematic stress distributions and load-displacement plot during the single fibre pushout test. One difference from the pull-out test is that the Poisson effect causes the fibre
to expand (rather than contract), which augments (rather than offsets) the radial
compressive stress across the interface due to differential thermal contraction
TWC - Lent 2015
C16H38
Fig.8.1 Schematic depiction of the fracture of a unidirectional long fibre composite at critical
values of (a) axial, (b) transverse and (c) shear stresses
Axial Strength
C16H39
Maximum stress:
! 1*
c2
!
! 2 = ! 2* !"!! # = 2*
s2
$
$ 12 = $ 12* !"!! # = 12*
cs
! 1 = ! 1* !"!! # =
Tsai-Hill:
!"
% c2 c2 # s 2
s 4 c2 s 2 (
='
+ 2 + 2 *
! 1*2
! 2* $ 12* *
'&
)
#1/2
Fig.8.2 Single ply failure stresses, as a function of loading angle: (a) maximum stress criterion,
for polyester-50%glass (!1*=700 MPa, !2*=20 MPa, "12*=50 MPa) and (b) maximum
stress and Tsai-Hill criteria, plus experimental data, for epoxy-50%carbon
(!1*=570 MPa, !2*=32 MPa, "12*=56 MPa)
TWC - Lent 2015
C16H40
Failure of Laminates
C16H41
Fig.8.4 Loading of the crossply laminate of Fig.4.4 parallel to one of the fibre directions:
(a) cracking of transverse plies as !2 reaches !2*, (b) onset of cracking parallel to fibres
in axial plies as !2 (from inhibition of Poisson contraction) reaches !2* and (c) final
failure as !1 in axial plies reaches !1*
TWC - Lent 2015
C16H42
C16H43
C16H44
Fig.9.1 Schematic load-displacement plots for 3-point bend testing of monolithic SiC and a SiC
laminate with (weak) graphitic interlayers
Fig.9.2 SEM micrographs showing the layered structures of (a) a mollusc and (b) a SiC laminate
TWC - Lent 2015
C16H45
Fig.9.3 Schematic representation of the advance of a crack in a direction normal to the fibre
axis, showing interfacial debonding and fibre pull-out processes
C16H46
fs 2 r# i*
$ f ' $ " r# i* ' $ L3 '
Ndx0
" rx02# i* = & 2 ) &
=
& )
L
3
% " r ( % L )( % 3 (
0
Gcpo !=! !
(9.1)
Fig.9.4 Schematic depiction of stress distribution, and associated probability of fracture, along a
fibre bridging a matrix crack, for (a) fixed fibre strength !* (m=!) and (b) strength
which varies along the length of the fibre, due to the presence of flaws (finite m)
TWC - Lent 2015
C16H47
(a)
(b)
There have been many attempts to produce ceramic-matrix composites with high toughness,
but with limited success. Probably the most promising approach is to introduce a network of
metallic fibres, and this is the basis of a commercial product (Fiberstone see Fig.9.5). The
fibres are often about 0.5 mm diameter, although finer fibres can be used. During fracture, fibres
bridge the crack and energy is absorbed by both frictional pull-out and plastic deformation - see
Fig.9.6. These mechanisms are likely to dominate any other contributions to the work of fracture.
Fig.9.6 Schematic representation of the fracture of Fiberstone, showing: (a) overall fracture
geometry, (b) fibres undergoing debonding, possibly fracture, and then frictional pullout and (c) fibres undergoing debonding, plastic deformation and then fracture
TWC - Lent 2015
C16H48
The work of fracture can thus be estimated by summing the energy absorbed via both
processes, assuming that a fraction g of the fibres bridging the crack plane undergo pull-out and
the remainder (1-g) undergo plastic deformation and rupture.
Gcnet = Gcpo + Gcfd
(9.2)
An expression for the fibre pull-out work was derived previously (Eqn.(9.1)), but the relationship
between N and f depends on fibre orientation distribution, and that treatment referred to aligned
fibres. For this type of composite, it can be taken as isotropic (random), in which case N is half
that for the aligned case (see EE Underwood, Quantitative Stereology. 1970, Addison-Wesley)
f
2! r 2
(9.3)
gfs 2 r! i*
Gcpo !=!
6
(9.4)
N=
leading to
where s is here the ratio of (, the (average) length of fibre extending beyond the crack plane, to
the fibre radius, r.
Fig.9.7 Data from tensile testing of single 304 stainless steel fibres, showing (a) a set of 10
stress-strain curves and (b) the distribution of corresponding work of deformation values
The work done during plastic deformation and rupture of fibres can be estimated by assuming
that interfacial debonding extends a distance x0 from the crack plane - see Fig.9.6(c). The energy
is obtained by summing the work done on each fibre, as if it had an original length 2x0 and were
being subjected to a simple tensile test
# f &
Gcfd = (1 ! g)2x0 NU fd = (1 ! g)2x0 %
W " r 2 = (1 ! g)x0 fWfd
$ 2" r 2 (' fd
(9.5)
where Ufd and Wfd are the work of deformation of the fibre, expressed respectively per unit length
(J m-1) and per unit volume (J m-3). The latter is given by the area under the stress-strain curve of
the fibre. Some such curves, for the fibres used in Fiberstone, are shown in Fig.9.7, together
with corresponding Wfd values. The value of ( is in this case given by the product of x0 and %*, the
fibre strain to failure, leading to
$"'
(1 ! g)srfWfd
Gcfd = (1 ! g) & ) fWfd =
#*
% #* (
(9.6)
C16H49
Use of Eqns.(9.2), (9.4) and (9.6) allows prediction of the composite fracture energy, although
it requires measurements or assumptions to be made concerning several parameters. In addition to
the single fibre work of deformation, Wfd, and the failure strain, %*, estimates are required for the
proportion of fibres undergoing pull-out, g, the interfacial frictional sliding stress, "i*, and the
(average) length of fibre extending beyond the crack plane, (, and hence the protrusion aspect
ratio, s (= (/r) Nevertheless, predictions can be made, based on experimental data or on plausible
assumptions, and compared with measured composite fracture energies. An example is shown in
Fig.9.8, where it can be seen that, even with the relatively low fibre content (~10-15%) that is
normally present, the work of fracture is both predicted and observed to be substantial. The
experimental Gc values were obtained by impact (Izod) testing.
Fig.9.8 Comparison between experimental data for the fracture energy of Fibrestone
composites, as a function of fibre volume fraction, and predictions obtained using
Eqns.(9.4) amd (9.6), for fine and coarse fibres
The value of s can be estimated from observation of fracture surfaces. However, its difficult
to be sure whether particular fibres have predominantly undergone pull-out, rather than plastic
deformation and rupture - of course, some fibres could deform plastically and then pull out. In
any event, very strong bonding may be undesirable, since this will tend to inhibit both pull-out and
plasticity, although very poor bonding may allow fracture to take place without the fibres being
significantly involved in the process. An intermediate bond strength is likely to give optimal
toughness.
It also worth noting that, for a given fibre protrusion aspect ratio, s (= (/r), both pull-out and
plastic deformation contributions increase linearly with the absolute scale (fibre diameter).
Composites reinforced with coarser fibres therefore tend to be tougher, particularly for this type of
composite. Its clear that refining the scale of the microstructure does NOT always give benefits!
C16H50
Euler Buckling
C16H51
Fig.10.1 Optical micrograph of an axial section of a carbon fibre composite after failure under
uniaxial compression, showing a kink band
C16H52
(a)
(b)
Fig.10.3 (a) Fragment of SiC monofilament extracted from a Ti-35%SiC composite after loading
under axial compression and (b) schematic of the crushing process
Fig.10.4 Stresses in Ti-35%SiC monofilament composite (average axial values for fibre, matrix
and composite) as axial strain is increased by external loading. At zero strain, stresses
in fibre and matrix are from differential thermal contraction. The matrix yields when the
stress in it reaches !mY. It is assumed that no matrix work hardening occurs during
plastic straining. Failure occurs when the fibre stress reaches the critical value !f*
Failure is expected when the fibre stress reaches !f*, taken to be a single, fixed value. The
composite strength !c* can readily be predicted, provided it can be assumed that the matrix yields
before composite failure and that matrix work hardening is negligible, since it is then given by
(10.1)
in which the composite moduli before and after matrix yielding are given by
TWC - Lent 2015
C16H53
E1c = fEf + (1 ! f ) Em
E1c' = fEf
Now, the strains at matrix yield and at final failure can be written as
! cmY =
! c* =
Substituting into Eqn.(10.1), and applying the residual stress force balance condition
f ! f"T + (1 # f ) ! m"T = 0
then leads to
! c* = f ! f * + (1 " f ) ! mY
A correction should be applied for the effect of misalignment in reducing the stress parallel to the
fibre axis, leading to
! c* =
f ! f * + (1 " f ) ! mY
cos 2 #0
(10.2)
This predicted strength is independent of the thermal residual stresses (whereas the strain at which
failure occurs will depend on them).
Fig.10.5 Experimental strength data, as a function of the initial angle between fibre and loading
axes, during compression of misaligned Ti-35%SiC specimens. Also shown are
predicted curves for failure by kink band formation and by fibre crushing, obtained by
substitution of the values shown into the kinking equation and Eqn.(10.2) respectively
TWC - Lent 2015
C16H54
Lecture 11:
Thermal Expansion of Composites & Thermal Residual Stresses
Thermal Expansivity Data for Reinforcements and Matrices
Fig.11.1 Thermal expansion coefficients for various materials over a range of temperature
C16H55
Fig.11.2 Schematic showing thermal expansion in the fibre direction of a long fibre composite,
using the slab model
C16H56
C16H57
Fig.11.4 Neutron diffraction data for an Al-5vol%SiC whisker (short fibre) composite, showing
lattice strains (& hence stresses) within matrix & reinforcement during unloaded
thermal cycling. (111) reflections were used for both constituents. The gradients shown
are calculated values for elastic behaviour, assuming a fibre aspect ratio of 10
C16H58
Fig.12.1 Schematic depiction of the generation of curvature in a flat bi-material plate, as a result
of the imposition of a uniform misfit strain, )(. The strain and stress distributions shown
are for the case indicated, obtained using Eqns.(12.10) & (12.11)
C16H59
!h+H$
M !=!P! #
" 2 &%
(12.1)
where h and H are deposit and substrate thicknesses respectively. Since the curvature, *,
(through-thickness strain gradient) is given by the ratio of moment, M, to beam stiffness, )
! !=!
M
"
(12.2)
P can be expressed as
P=
2!!!"
h+H
(12.3)
% h2
% H2
(
2(
! = b $ E ( yc )!y !dyc != b!Ed !h ' " h# + # * + b!Es !H '
+ H# + # 2 *
& 3
& 3
)
)
" H "#
2
c
(12.4)
where +, the distance from the neutral axis (yc = 0) to the interface (y = 0) is given (see Appendix
on p.64) by
h 2 Ed " H 2 Es
!=
2 ( hEd + HEs )
(12.5)
The magnitude of P is found by expressing the misfit strain as the difference between the
strains resulting from application of the P forces.
P
P
+
HbEs hbEd
(12.6)
Combination of this with Eqs.(12.3)-(12.5) gives a general expression for the curvature, *, arising
from imposition of a uniform misfit strain, +(
!=
(12.7)
Note that, for a given deposit/substrate thickness ratio, h/H, the curvature is inversely proportional
to the substrate thickness, H. This scale effect is important in practice, since it means that
relatively thin substrates are needed if curvatures are to be sufficiently large for accurate
measurement. Predicted curvatures, obtained using this equation, are shown in Fig.12.2.
Curvatures below about 0.1 m-1 (radius of curvature, R > 10 m) are difficult to measure accurately.
Biaxial Stresses
C16H60
In practice, there are often in-plane stresses other than those in the x-direction. For an isotropic
in-plane stress state, there is effectively another stress equal to !x in a direction at right angles to it
(z-direction); this induces a Poisson strain in the x-direction. Assuming isotropic stiffness and
negligible through-thickness stress (!y = 0), the net strain in the x-direction can be written
!x
E
=
= E'
" x (1 # $ )
(12.8)
and this modified form of the Youngs modulus, E, is usually applicable in expressions referring
to substrate/coating systems having an equal biaxial stress state.
A simplified form of Eq.(12.7) applies for coatings much thinner than the substrate (h << H).
The substrate stress becomes negligible and that in the deposit varies little as a result of curvature
adoption, so the misfit strain can be converted to a deposit stress. For an equal biaxial case,
!d =
Ed
$%
(1 " # d )
6h (1 # $ s )
%d
Es H 2
(12.9)
This is Stoneys equation, which is commonly used to relate (biaxial) stress to (biaxial) curvature
for thin coatings. The properties required (Es and 's) are only those of the substrate. This is
convenient, since these are usually more readily accessible than those of the coating.
Unfortunately, the Stoney equation is only accurate in a regime (h<<H) where curvatures are often
very small (and hence difficult to measure) - see below.
When the condition h << H does not apply, then stresses and stress gradients are often
significant in both constituents. Stress distributions are readily found for the simple misfit strain
case outlined above, from the values of P and *, using the expressions
!d
!d
!s
!s
y= h
y= 0
"P
+ Ed !# ( h " $ )
b!h
(12.10a)
"P
!"!Ed !# !$
b!h
(12.10b)
y= " H
y= 0
P
!"!Es !# !( H + $ )
b!H
P
!"!Es !# !$
b!H
(12.11a)
(12.11b)
TWC - Lent 2015
C16H61
The stress distributions in Fig.12.1, 12.3 and 12.4 were obtained using these equations. The
adoption of curvature can effect substantial changes in stress levels and high through-thickness
gradients can result. It may be seen from Eqns.(12.10) and (12.11) that (for a given value of h/H),
since P is proportional to H and * is inversely proportional to H, the stresses (at y=-H, 0 and h) do
not depend on H, ie the stress distribution is independent of scale. However, the curvature is not.
Substrates must be fairly thin if measurable curvatures are to be generated, although the maximum
thickness could be as small as 50 m, or as large as 50 mm, depending on various factors.
Fig.12.2 Predicted curvature, as a function of the fall in temperature, for four different
substrate/deposit combinations
C16H62
Fig.12.3 Predicted dependence on thickness ratio of (a) curvature and (b) stress in deposit
(coating), obtained using Eqns.(12.7), (12.10) and (12.11), and the Stoney equation
(Eqn.(12.9).) The Poisson ratios of substrate and deposit were both taken as 0.2
C16H63
Fig.12.4 Predicted effect of substrate thickness, H (for a fixed coating thickness h), on the average
stress levels in coating and substrate, created by a given misfit strain, and also on the
strain energy release rate for debonding (showing the contributions from stresses in
coating and substrate). The effect of curvature adoption on the stresses is neglected.
TWC - Lent 2015
C16H64
b # ! ( y )!dy!= 0
(12.,.1)
"H
can be divided into contributions from the two constituents and expressed in terms of the strain
h
#H
(12.,.2)
which can then be written in terms of the curvature (through-thickness strain gradient) and the
distance from the neutral axis
h
"H
(12.,.3)
Removing the width, b, and curvature, *, which are constant, this gives
0
# y2
&
# y2
&
Ed % ! " y ( +!Es % ! " y ( != 0
$2
'0
$2
'! H
* h2
* !H 2
) Ed , ! " h / + E s ,
!"H/ = 0
+ 2
+ 2
.
.
) " ( Ed h + E s H ) =
)" =
1
Ed h 2 ! E s H 2
2
h 2 Ed ! H 2 Es
2 ( hEd + HEs )
C16H65
Density
(Mg m-3)
Glass
2.45
Axial
Modulus
E1 (GPa)
Transverse
Modulus
E2 (GPa)
Shear
Modulus
G12 (GPa)
Poisson
Ratio
'12
76
76
31
0.22
Axial
Axial
Strength
CTE
!* (GPa) -1 (% K-1)
3.5
Transverse
CTE
-2 (% K-1)
Kevlar
1.47
154
4.2
2.9
0.35
2.8
-4
54
Carbon (HM)
1.94
385
6.3
7.7
0.20
1.7
-1
10
5
Carbon (HS)
1.75
224
14
14
10
Boron
2.64
1000
420
170
0.20
4.2
SiC
3.2
400
400
170
0.20
3.0
SiC
3.2
550
350
170
0.17
Al2O3
3.9
385
385
150
0.26
1.4
Al2O3
3.4
300
300
120
0.26
2.0
19.3
413
413
155
0.33
3.3
(monofilament)
(whisker)
(- continuous)
(( staple)
Matrices
Matrix
Density
(Mg m-3)
Epoxy
Polyester
1.25
1.38
Polycarborate
-1
1000
420
0.20
2.1
3.52
Diamond
415
0.20
Young's
Modulus
E (GPa)
Shear
Modulus
G (GPa)
Poisson
Ratio
'
Tensile
Strength
!* (GPa)
Thermal
Expansivity
- (% K-1)
1.4
0.37
0.07
45
3.5
3.0
1.27
1.1
0.38
0.37
0.04
0.04
58
150
PEEK
1.30
1.15
2.4
Polyurethane
Rubber
1.2
0.01
0.003
0.46
0.02
200
Aluminium
2.71
70
26
0.33
0.07
24
Titanium
Magnesium
Borosilicate
glass
1.74
0.9
0.33
0.33
0.06
0.19
70
4.51
45
115
7.5
26
44
0.33
0.24
10
2.23
64
28
0.21
0.09
3.2
C16H66
Question Sheet 1
[Can be attempted after lecture 8: property data on C16H65 can be used if necessary.]
1. Show that the Young's modulus of a composite lamina (having the elastic constants, referred
to the fibre axis, given below) falls by about 50% if it is loaded at 7 to the fibre axis,
compared with the on-axis value. What is the minimum Young's modulus that the lamina can
exhibit and at what loading angle does this occur ?
[E1 = 200 GPa, E2 = 7 GPa, G12 = 3 GPa, v12 = 0.3]
Fig.1 Stresses (ratios to !x) within one ply of an angle-ply laminate, as a function of the ply
angle, when subjected to biaxial loading, with !x=2!y.
Find the failure pressure using the Tsai-Hill failure criterion, which can be expressed as
2
" !1 % " ! 2 %
! 1! 2 " ) 12 %
$# ! '& + $# ! '& ( ! 2 + $# ) '& * 1
1*
2*
1*
12*
Explain any difference between this value and that obtained previously. Using the criterion
you consider most reliable in this case, obtain an approximate estimate of the ply angle that
would give the largest failure pressure.
{from 2014 Tripos}
TWC - Lent 2015
C16H67
# !Y
where !1, !2 and !3 are the principal stresses and !Y is the uniaxial yield stress. The latter
has a value of 150 MPa for the steel. The density of the steel is 7.8 Mg m-3.
The Tsai-Hill criterion for failure of a composite ply under plane stress conditions can be
expressed as:
2
" !1 % " ! 2 %
! 1! 2 " ) 12 %
+
(
+$
2
$# ! '& $# ! '&
'& * 1
!
)
#
1*
2*
1*
12*
where !1, !2 and "12 are the stresses parallel, transverse and in shear relative to the fibre
axis and !1*, !2* and "12* are corresponding critical values (measured respectively to be
900 MPa, 30 MPa and 40 MPa). The composite density is 1.8 Mg m-3.
The stresses within a lamina, subject to !x, !y and "xy, are given by
!1
!2
! 12
2cs
c2 s 2
2
2
= s
c
"2cs
"cs cs c 2 " s 2
!x
!y
# xy
where c = cos# and s = sin#, and # is the angle between x and 1 (fibre) directions.]
{from 2012 Tripos}
C16H68
Question Sheet 2
[Can be attempted after lecture 12; property data on C16H65 can be used if necessary.]
1. A strut is in the form of a hollow cylinder with an outside diameter of 25 mm and a bore of
20 mm. It is manufactured from MMC material composed of 70 vol.% SiC monofilaments in
a titanium alloy matrix, with the SiC fibres aligned approximately parallel to the axis of the
strut. However, the limitations of the manufacturing process are such that fibre misalignments
of up to 4 are present. The crushing strength of the SiC monofilaments is known to be about
8 GPa and the yield stress of the Ti alloy is 600 MPa, while the critical shear stress of the
composite, on planes parallel to the fibre axis, is measured to be about 200 MPa.
(i) Estimate the shear modulus of the composite and hence the stress for failure by kink band
formation. Would failure of this type occur under an axial compressive load of 25 kN?
(ii) Would any other type of failure or deformation be expected under this applied load?
(b) When the bonded pair is loaded in compression parallel to the fibre axis of the ply, it is
observed that the curvature(s) it exhibits starts to reduce. Account for this effect. Calculate
the applied stress at which the specimen would become flat and comment on whether this is
likely to be achievable.
[For glass fibres: E = 76 GPa, - = 5 / 10-6 K-1, ' = 0.22
for epoxy resin: E = 3.5 GPa, - = 58 / 10-6 K-1, ' = 0.40
for steel: E = 210 GPa, - = 11.4 / 10-6 K-1, ' = 0.26
For an aligned long fibre composite. axial and transverse thermal expansivities, , c, tr and , c, tr,
are given by the following (force balance and Schapery) expressions
! (1 " f ) Em + ! f fEf
! c, ax =! m
(1 " f ) Em + fEf
The curvature, *, exhibited by a pair of bonded plates, each of thickness h, when there is a
misfit strain +% between their natural (stress-free) lengths, is given by
12 "#
!= $
]
E1
E2 '
h &&
+ 14 + ))
E1 (
% E2
{from 2008 Tripos}
C16H69
3. (a) During formation of a coating on a substrate, its common for a misfit strain, +%, to be
created, representing the difference between the (stress-free) in-plane dimensions of the two
constituents. For example, this often arises during deposition and/or subsequent cooling. This
misfit creates stresses and stains in the coating (and possibly in the substrate). Show that the
relation between the stress and strain in the coating, in any (in-plane) direction, can be
expressed
!
E
=
"
(1 # $ )
where E is the Youngs modulus and ' is the Poisson ratio.
(b) The curvature, *, arising from a misfit strain, +%, between a coating (deposit) of thickness
h and a substrate of thickness H is given by
6Ed Es ( h + H ) h!H !"#
!= 2 4
3
Ed h + 4Ed Es h H + 6Ed Es h 2 H 2 + 4Ed Es h!H 3 + Es2 H 4
where Ed and Es are the Youngs moduli of deposit and substrate. Show that, in the limit of
h<<H, this reduces to the Stoney equation, giving the curvature in terms of the stress in the
deposit, its Poisson ratio, the Youngs modulus of the substrate and the thicknesses of the two
constituents.
(c) A glass sheet of thickness 3 mm has a 10 m layer of Al evaporated onto one side, to form
a mirror. The production process generates negligible stress in the coating. The sheet is
subsequently heated from room temperature (20C) to 170C. Calculate the curvature
exhibited by the sheet after heating, assuming that the system remained elastic.
(d) Decide, stating any assumptions, whether yielding is in fact likely to occur in the Al layer
during heating, given that it has a uniaxial yield stress at 170C of 100 MPa.
(e) Hence give an opinion as to whether any detectable distortion of the reflective
characteristics of the mirror is likely to be present after it has cooled to room temperature.
[For the glass: E = 75 GPa, - = 8.5 / 10-6 K-1,
for the Al:
E = 70 GPa, ' = 0.33, - = 24.0 / 10-6 K-1]
4. (a) Show that the curvature, *, of a beam (reciprocal of the radius of curvature, R) is equal to
the through-thickness gradient of the strain, with the strain being zero at the neutral axis.
{15%}
(b) A vibration-damped sheet material is made by bonding a 1 mm thick rubber layer
between two steel plates of thickness 1 mm. The sheet is pushed against the surface of a large
cylindrical former, which has a radius of 0.5 m. Sketch the through-thickness distributions of
strain and stress in the sheet, assuming that both the steel and the rubber remain elastic.
{25%}
(c) This forming operation is actually designed to generate plastic deformation, creating a
shaped component with a uniform curvature in one plane. Taking the steel to have a yield
stress of 300 MPa (in compression or tension), and assuming that the rubber remains elastic,
show that the above operation would in fact induce plastic deformation in outer layers of both
metal sheets and calculate the thickness of the layers that would yield in this way and the
plastic strain at the free surfaces.
{20%}
C16H70
(d) Show that, if the width of the sheet (length along the axis of the cylinder) is 0.5 m, then the
beam stiffness () = EI) of the sheet is 216.7 N m2 and the bending moment that would be
needed in order to bring the sheet into contact with the cylindrical former would be 433 N m,
assuming that the steel remained elastic. Calculate the required bending moment for the actual
case, with the steel undergoing plastic deformation at a yield stress of 300 MPa (but
neglecting any work hardening).
{40%}
[Steel:
Rubber:
5. John Harrison, the famous clock-maker credited with developing a time-keeping system
sufficiently reliable to establish longitude at sea, was reportedly the first to create a bi-metallic
strip (for compensation of the effects of temperature change), which he did by casting a thin
brass layer onto a thin steel sheet. Show that, if both layers have a thickness of 0.1 mm, and
the strip is 100 mm long, then the temperature change required to generate a lateral deflection
of 1 mm at its end is about 4.6 K, assuming that the system remains elastic.
Sketch the (approximate) through-thickness distributions of stress and strain within the above
strip, after it had been heated by 100 K. Give your view as to whether such heating would be
likely to cause plastic deformation within the strip, given that the yield stresses of both
constituents are expected to be of the order of 100 MPa.
[The curvature, *, of a bi-material strip comprising two constituents of equal thickness (h),
arising from a misfit strain of +( between them, is given by
12 "#
!=
$E
E '
h & 1 + 14 + 2 )
E1 (
% E2
where E1 and E2 are the Youngs moduli of the constituents. The relationship between
curvature, *, end deflection, y, and length, x, of a bi-material strip may be expressed as
( )
For steel: Youngs modulus, E = 200 GPa; thermal expansivity, , = 13 / 10-6 K-1
For brass: Youngs modulus, E = 100 GPa; thermal expansivity, , = 19 / 10-6 K-1]
C16H71
Examples Class I
[Property data on C16H65 can be used if necessary.]
1. (a) The components of the compliance tensor of an epoxy-glass fibre composite lamina,
referred to the fibre axis direction and the transverse direction, can be written
S11
! S !!=! S12
0
S12
S22
0
1 / E1
0
0 != !" 21 / E2
S66
0
!"12 / E1
0
1 / E2
0
0
1 / G12
Using information in the Data Book, show that the interaction compliance giving the shear
strain arising from a normal stress, when the lamina is loaded at an angle # to the fibre axis, is
S16 !=!( 2S11 ! 2S12 ! S66 ) c 3 s!!!( 2S22 ! 2S12 ! S66 ) cs 3
(b) Using the following measured values of elastic constants of the composite
E1 = 50!GPa,!!!E2 = 5!GPa,!!!12 = 0.3,!!G12 = 10!GPa
calculate the shear strain induced in the lamina when a normal tensile stress of 100 MPa is
applied at an angle of 30 to the fibre axis.
(c) The dependence of this interaction compliance on # is shown below for a different
composite. Sketch the corresponding plot for a 0/90 crossply laminate of the same material,
obtained by assuming that the laminate compliance, at any given #, can be taken as the
average of those for the constituent plies at their corresponding # values.
C16H72
3. On the page Failure of Laminates and the Tsai-Hill criterion, use the facility at the end to
create a polyester-50%glass angle-ply laminate (40). Taking this to be a filament-wound
tube, with the plies at 40 to the hoop direction, and a radius/wall thickness ratio of 20,
subjected to internal pressure, P, estimate the value of P at which failure will occur, according
to the Tsai-Hill criterion, given that !1* = 700 MPa, !2* = 20 MPa and "12* = 50 MPa. Using
analytical equations, carry out the same calculation for one of the two plies (ignoring the
presence of the other). Account for the difference between this value and the one you
obtained treating the laminate as a whole (using the numerical procedure in the TLP).
C16H73
Examples Class II
[Property data on C16H65 can be used if necessary.]
1. (a) For a small aircraft, a choice must be made between an Al alloy and a composite for the
fuselage material. The fuselage will approximate to a cylinder of diameter of 2 m and will
experience internal pressures up to 0.6 atm (0.06 MPa) above that of the surrounding
atmosphere, axial bending moments of up to 500 kN m and torques of up to 600 kN m. The
composite fuselage would be produced by filament-winding at 45 to the hoop direction. It
may be assumed that this is a strength-critical application, with the airframe stiffness expected
to be adequate in any event. Using the Tresca yield criterion (Al) and the Tsai-Hill failure
criterion (composite), and ignoring the issue of safety factors, estimate the minimum wall
thickness in each case and hence deduce which material would allow the lighter fuselage.
(b) Comment on the main sources of error in your calculation and also on whether there might
be a danger of any other type of failure.
[For the Al alloy, the yield stress in uniaxial tension = 250 MPa
For the composite, failure stresses for loading transverse and in shear relative to the fibre
axis are both 50 MPa: the possibility of failure by fracture of the fibres can be neglected.
Densities: Al = 2.70 Mg m-3, composite = 1.50 Mg m-3
The peak axial stress in a thin-walled cylinder subjected to a bending moment M is R M/I,
where R is the radius and I is the moment of inertia, which is given by "R3 t, where t is the
wall thickness.]
{from 2006 Tripos}
2. (a) A thick metal sheet was held at 1000C in air for several hours, after which time an oxide
film had formed (on both sides), with a thickness of 100 m. No significant stresses were
created in metal or oxide during this process. During subsequent cooling, much of this oxide
spalled off from the substrate when the temperature reached 300C. Estimate the fracture
energy of the interface between the metal and the oxide, stating your assumptions.
(b) The above thermal treatment was repeated on a different sheet of the same metal, in the
form of a relatively narrow strip of a thinner sheet and in a configuration such that only one
side of the strip was exposed to air. In this case, it was observed that spallation did not occur,
even after cooling to ambient temperature (20C), and that the strip exhibited noticeable
curvature at this stage. Give a qualitative explanation of the fact that spallation occurred in the
first experiment (part (a)), but not in the second.
(c) In the curved strip obtained after the above experiment (part (b)), would the oxidized side
be expected to be convex or concave? The residual thickness of the metal was found to be
1 mm. What magnitude of curvature would be expected? Is this significantly different from
the value that would be obtained if the Stoney approximation were used?
[The curvature, *, arising from imposition of a uniform misfit strain, +(, between the two
layers in a bi-layer system (with thicknesses h and H) is given by
6Ed Es ( h + H ) h!H !"#
!= 2 4
3
Ed h + 4Ed Es h H + 6Ed Es h 2 H 2 + 4Ed Es h!H 3 + Es2 H 4
where Ed and Es are the corresponding Youngs moduli (and biaxial versions of these apply if
the same misfit strain is also being created in the other in-plane direction).
Thermal expansivities: metal = 15 / 10-6 K-1 oxide = 7 / 10-6 K-1
Youngs moduli: metal = 100 GPa oxide = 200 GPa
Poisson ratios: metal = 0.3 oxide = 0.2]
C16H74
3. (a) Steel sheet of thickness 1 mm is given a thin protective layer of vitreous enamel. This
coating is created by adding glassy powder to the surface and holding at around 700-800C,
causing the powder to fuse and form a layer of uniform thickness. The sheet is then furnace
cooled, taking several hours to reach room temperature, such that the thermal misfit strain is
completely relaxed by creep down to about 220C, after which cooling is elastic. Assuming
that the coating / substrate thickness ratio, h/H, is sufficiently small for the Stoney equation to
be valid, estimate the elastic strain in the coating, stating your assumptions.
(b) The adhesion of the enamel to the steel is excellent, so the system is highly resistant to
debonding, but its found that, if the coated sheet is progressively bent in one plane (with the
steel undergoing plastic deformation), then through-thickness cracks appear in the enamel
layer (on the convex side) when the local radius of curvature reaches 60 mm. Assuming that
such cracking starts when the tensile strain in the enamel reaches a certain level, use this
information to estimate this critical strain.
(c) A fabrication procedure requires bending of the coated sheet to a radius of curvature of
50 mm. The suggestion is made that, instead of furnace cooling the sheet after formation of
the coating, it should be removed from the furnace and cooled more quickly, such that elastic
cooling occurs below about 420C (and stress relaxation is complete until this point). Would
you expect this measure to result in the elimination of through-thickness cracking during
bending of the sheet to this curvature?
(d) For the latter case (ie the rapidly cooled sheet), what are the principal stresses within the
coating, before and after the bending operation? (The deformation of the steel sheet can be
taken as entirely plastic.)
[Property data:
, = 14 10-6 K-1
Steel:
Enamel:
, = 5 10-6 K-1; E = 70 GPa; $ = 0.2
where , is the thermal expansivity, E is the Youngs modulus, and $ is the Poisson ratio]