Professional Documents
Culture Documents
L. N. McCartney
Materials Division, National Physical Laboratory
Teddington, Middlesex, TW11 0LW, UK
1 Introduction
Engineers responsible for the design and maintenance of composite struc-
tures will usually be involved with some form of finite element analysis
(FEA) so that the stress distributions within and the deformation of the
Axial
direction (x1 )
Through-thickness
Fibre direction direction (x3 )
Transverse
direction (x2 )
When referred to the coordinates (x, y, z), the stress-strain relations are
written in the form
1 νyx νzx σxy
εxx = σxx − σyy − σzz + αxx ΔT, εxy = ,
Exx Eyy Ezz 2μxy
νxy 1 νzy σxz
εyy = − σxx + σyy − σzz + αyy ΔT, εxz = , (2)
Exx Eyy Ezz 2μxz
νxz νyz 1 σyz
εzz = − σxx − σyy + σzz + αzz ΔT, εyz =
Exx Eyy Ezz 2μyz
Analytical Methods of Predicting Performance… 195
Using the compact notation to be used in this paper, involving only inde-
pendent thermoelastic constants, the stress-strain relations are written:
1 νA νa τA
εA = σA − σT − σt + αA ΔT, γA = ,
EA EA EA μA
νA 1 νt τa
εT = − σA + σT − σt + αT ΔT, γa = (3)
EA ET ET μa
νa νt 1 τt
εt = − σA − σT + σt + αt ΔT, γt =
EA ET Et μt
The subscripts A, T and t attached to stresses, strains and properties indi-
cate parameters associated respectively with the axial, in-plane transverse
and through-thickness directions of the lamina. It should be noted that
the upper case subscripts A and T are associated only with in-plane direc-
tions and parameters, while the lower case subscripts are associated with
the through-thickness direction and parameters. The above three sets of
stress-strain relations are equivalent only if:
and
νa ν13 ν31 νxz νzx
= = = =
EA E11 E33 Exx Eyy
Et
so that ν13 = νxz = νa , ν31 = νzx = νa ,
EA
νt ν23 ν32 νyz νzy
= = = =
ET E22 E33 Eyy Ezz
(5)
Et
so that ν23 = νyz = νt , ν32 = νzy = νt ,
ET
νA ν21 ν12 νyx νxy
= = = =
EA E22 E11 Eyy Exx
ET
so that ν12 = νxy = νA , ν21 = νyx = νA
EA
It should be noted that
γA = 2ε12 = 2εxy , γa = 2ε13 = 2εxz , γt = 2ε23 = 2εyz ,
(6)
τA = σ12 = σxy , τa = σ13 = σxz , τt = σ23 = σyz
Et = ET , νa = νA , μa = μA , αt = αT and ET = 2μt (1 + νt )
where
EA E
ĒA = 1 − νt2 t ,
Λ ET
ET E
ĒT = 1 − νa2 t ,
Λ EA
Et 2 ET
Ēt = 1 − νA ,
Λ EA
ET E
ν̄A ĒT = νA + νa νt t , (8)
Λ ET
Et
ν̄a Ēt = (νa + νt νA ) ,
Λ
Et ET
ν̄t Ēt = ν t + νa νA ,
Λ EA
Et Et 2 ET Et
Λ = 1 − νa2 − νt2 − νA − 2νa νt νA
EA ET EA EA
By comparing (12) with (7) using (8) and (9), it follows that the non-zero
components of the C matrix are related to the elastic constants defined by
the stress-strain relations (3) as follows
EA Et ET Et
C11 = 1 − νt2 = ĒA , C12 = νA + νa νt = ν̄A ĒT ,
Λ ET Λ ET
Et ET Et
C13 = (νa + νt νA ) = ν̄a Ēt , C21 = νA + νa νt = ν̄A ĒT ,
Λ Λ ET
ET Et Et ET
C22 = 1 − νa2 = ĒT , C23 = νt + νa νA = ν̄t Ēt ,
Λ EA Λ EA
Et Et ET
C31 = (νa + νt νA ) = ν̄a Ēt , C32 = νt + νa νA = ν̄t Ēt ,
Λ Λ EA
Et 2 ET
C33 = 1 − νA = Ēt , C44 = μt , C55 = μa , C66 = μA
Λ EA
(13)
with
Et Et 2 ET Et
Λ = 1 − νa2 − νt2 − νA − 2νa νt νA
EA ET EA EA
The non-zero components of the U vector are related to the thermoelastic
constants defined by the stress-strain relations (3) as follows
U1 = ĒA ᾱA , U2 = ĒT ᾱT , U3 = Ēt ᾱt (14)
The inverse matrix form of (12) is of the form
⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
εA S11 S12 S13 0 0 0 σA αA
⎢ εT ⎥ ⎢ S21 S22 S23 0 0 0 ⎥ ⎢ σT ⎥ ⎢ αT ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ εt ⎥ ⎢ S31 S32 S33 0 0 ⎥⎢ σt ⎥ ⎢ αt ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥+⎢ ⎥ ΔT, (15)
⎢ γt ⎥=⎢ 0 0 0 S 0 0 ⎥ ⎢ τt ⎥ ⎢ 0 ⎥
⎢ ⎥ ⎢ 44 ⎥⎢ ⎥ ⎢ ⎥
⎣ γa ⎦ ⎣ 0 0 0 0 S55 0 ⎦⎣ τa ⎦ ⎣ 0 ⎦
γA 0 0 0 0 0 S66 τA 0
and a comparison with (3) shows that
1 νA νa
S11 = , S12 = − , S13 = − ,
EA EA EA
νA 1 νt
S21 = − , S22 = , S23 = − ,
EA ET ET
(16)
νa νt 1
S31 = − , S32 = − , S33 = ,
EA ET Et
1 1 1
S44 = , S55 = , S66 =
μt μa μA
Analytical Methods of Predicting Performance… 199
1 N
Vfi Vm
= + m , (18)
kTeff m
+ μT i
k + μT
i=1 T
m kT + μm
T
N
Vfi kTi νA
i
Vm k m ν m
+ m T A
i
k + μT
i=1 T
m kT + μm T
eff
νA = , (19)
N
Vfi kTi Vm kTm
+ m
i=1
kTi + μm
T kT + μm T
⎛ 2 ⎞
eff 2 f(i) f(i)
4kTeff νA μm
N
⎜ f(i) 4kT νA μm ⎟
eff
EA + eff
= Vfi ⎝EA + f(i) ⎠
kT + μm i=1 kT + μm (20)
4k m ν 2 μm
+ Vm Em + mT m ,
kT + μm
1 N
Vfi Vm
= + m , (21)
μeff
A
m
+ μA i
μ + μA
i=1 A
m μ A + μm
A
1 N
Vfi Vm kTm μm
= + m , where μ∗m = T
, (22)
μeff
T
∗
+ μm i
μ + μm
i=1 T
∗ μT + μ∗m kTm + 2μmT
eff
eff eff 4νA (αeff eff eff
T + νA αA ) = V E f αf + V E α
EA αA + 1 1 f A A m m m
eff
+
kT μm
f (23)
4νA (αfT + νAf f
αA ) 4ν (α + νm αm )
+Vf 1 1 + Vm m 1 m 1 ,
+ m +
kTf μm kT μm
200 N. McCartney
N
Vfi kTi αiT + νAi i
αA Vm kTm (αm m m
T + νA αA )
+
i=1
kTi + μm
T kTm + μm
T
αeff eff eff
T + νA αA = , (24)
N
Vfi kTi Vm k m
i m + m Tm
k + μT
i=1 T
kT + μT
eff 2 2
1 1 1 (νA ) νTeff 1 1 eff
4(νA )
= + + , = − − (25)
ETeff 4μeff
T 4kT
eff E eff
A ETeff 4μeff
T 4kT
eff EA
eff
x1 Through-thickness
direction (x3 )
x3
Transverse
φ direction (x2 )
x2 Global axes
Local axes
⎡ ⎤ ⎡ ⎤
εA ε11
⎢ εT ⎥ ⎢ ε22 ⎥
⎢ ⎥ ⎢ ⎥
⎢ εt ⎥ ⎢ ε33 ⎥
⎢ ⎥≡ ⎢ ⎥
⎢ γt ⎥ ⎢ 2ε23 ⎥
⎢ ⎥ ⎢ ⎥
⎣ γa ⎦ ⎣ 2ε13 ⎦
γA 2ε12
⎡ ⎤⎡ ⎤ ⎡ ⎤
S11 S12 S13 0 0 S16 σA V1
⎢ S12 S22 S23 0 0 S26 ⎥⎢ σT ⎥ ⎢ V2 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ S13 S23 S33 0 0 S36 ⎥⎢ σt ⎥ ⎢ V3 ⎥
=⎢
⎢
⎥⎢
⎥⎢
⎥+⎢
⎥ ⎢
⎥ ΔT,
⎥
⎢ 0 0 0 S44 S45 0 ⎥⎢ τt ⎥ ⎢ 0 ⎥
⎣ 0 0 0 S45 S55 0 ⎦⎣ τa ⎦ ⎣ 0 ⎦
S16 S26 S36 0 0 S66 τA V6
(26)
where on setting m = cos φ and n = sin φ it can be shown that the co-
efficients in these stress-strain relations are related to the thermoelastic
202 N. McCartney
and
U1 = m2 ĒA ᾱA + n2 ĒT ᾱT ,
U2 = n2 ĒA ᾱA + m2 ĒT ᾱT ,
(43)
U3 = Ēt ᾱt,
U6 = mn ĒA ᾱA − ĒT ᾱT
(P ) (P )
(P ) 1 ν̃A ν̃a (P )
ε̃A ≡ εA + λA γA = σ −
(P ) A
σ −
(P ) T (P )
σt + α̃A ΔT,
ẼA ẼA ẼA
(P ) (P )
(P ) ν̃A 1 ν̃t (P )
ε̃T ≡ εT + λT γA = − σ +
(P ) A
σ −
(P ) T (P )
σt + α̃T ΔT, (47)
ẼA ẼT ẼT
(P ) (P )
(P ) ν̃a ν̃t 1 (P )
ε̃t ≡ εt + λt γA = − σ −
(P ) A (P )
σT + (P )
σt + α̃t ΔT,
ẼA ẼT Ẽt
where
1 1 2 1 1 1 2 1
(P ) (P )
(P )
= (P )
− λA (P )
, (P )
= (P )
− λT (P )
,
ẼA EA μA ẼT ET μA
1 1 2 1 (P )
ν̃t
(P )
νt
(P ) (P )
λ t λT
(P )
(P )
= (P )
− λt (P )
, (P )
= (P )
+ (P )
,
Ẽt Et μA ẼT ET μA
(P ) (P ) (P ) (P ) (P ) (P ) (P ) (P ) (48)
ν̃A νA λA λT ν̃a νa λt λA
(P )
= (P )
+ (P )
, (P )
= (P )
+ (P )
,
ẼA EA μA ẼA EA μA
(P ) (P ) (P ) (P ) (P ) (P ) (P ) (P )
α̃A = αA + λA αS , α̃T = αT + λT αS ,
(P ) (P ) (P ) (P )
α̃t = αt + λt αS
The relations (47) are known as the reduced stress-strain relations for the
angled lamina as they have exactly the same form as three of the stress-
strain relations (3) which apply when φ = 0.
(P ) (P ) (P ) (P ) (P )
σA = Ω11 εA + Ω12 εT + Ω13 σt + Ω16 γA − ω1 ΔT,
(P ) (P ) (P ) (P ) (P )
σT = Ω12 εA + Ω22 εT + Ω23 σt + Ω26 γA − ω2 ΔT,
(P ) (P ) (P ) (P ) (P ) (49)
εt = −Ω13 εA − Ω23 εT + Ω33 σt − Ω36 γA + ω3 ΔT,
(P ) (P ) (P ) (P ) (P )
τA = Ω16 εA + Ω26 εT + Ω36 σt + Ω66 γA − ω6 ΔT,
206 N. McCartney
where
(P ) (P ) (P ) (P )
(P ) ẼA (P ) ν̃ Ẽ (P ) ν̂a
Ω11 = (P )
, Ω12 = A (P T) , Ω13 = (P ) ,
Ψ Ψ Ψ
(P ) (P ) (P ) (P )
(P ) ẼA (P ) ν̃A ẼT (P ) (P ) ẼT
Ω16 = (P ) λA + λ T , Ω 22 = ,
Ψ Ψ(P ) Ψ(P )
(P ) (P ) (P ) (P )
(P ) ν̂ (P ) ν̃ Ẽ (P ) Ẽ (P )
Ω23 = t(P ) , Ω26 = A (P T) λA + T(P ) λT , (50)
Ψ Ψ Ψ
(P ) (P )
(P ) Λ (P ) (P ) ν̂a (P ) ν̂t (P ) (P )
Ω33 = , Ω 36 = λ + λ + λt ,
Ψ(P ) Ẽt
(P ) Ψ(P ) A Ψ(P ) T
(P ) (P )
(P ) ẼA (P ) (P ) ẼT (P ) (P ) (P )
Ω66 = λ λ̂ + (P ) λT λ̂T + μA ,
Ψ(P ) A A Ψ
(P ) (P ) (P )
(P ) ẼA (P ) ν̃ Ẽ (P )
ω1 = (P )
α̃A + A (P T) α̃T ,
Ψ Ψ
ẼT (P )
(P )
(P ) (P ) (P )
ω2 = α̃T + ν̃ A α̃A ,
Ψ(P )
(51)
(P ) (P )
(P ) ν̂a (P ) ν̂ (P ) (P )
ω3 = α̃ + t(P ) α̃T + α̃t ,
Ψ(P ) A Ψ
(P ) (P )
(P ) ẼA (P ) (P ) ẼT (P ) (P ) (P ) (P )
ω6 = λ̂ α̃ + (P ) λ̂T α̃T + μA αS ,
Ψ(P ) A A Ψ
and where
2 Ẽ (P )
(P )
Ψ(P ) = 1 − ν̃A T
(P )
,
ẼA
(P ) (P ) (P )
ν̂a = ν̃a(P ) + ν̃t ν̃A , (52)
(P )
(P ) (P ) (P ) ẼT
ν̂t = ν̃t + ν̃a(P ) ν̃A (P )
ẼA
(P ) (P )
Ẽt (P ) Ẽt (P )
Λ(P ) = Ψ(P ) − ν̃a(P ) ν̂ (P ) − ν̃t
(P ) a
ν̂
(P ) t
ẼA ẼT
2 Ẽ (P ) 2 Ẽ (P ) 2 (P )
ẼT
(P ) (P )
= 1 − ν̃a(P ) t
(P )
− ν̃ t
t
(P )
− ν̃ A (P )
(53)
ẼA ẼT ẼA
(P )
(P ) (P ) Ẽt
−2ν̃a(P ) ν̃t ν̃A (P )
,
ẼA
(P )
(P ) (P ) (P ) (P ) ẼT (P ) (P ) (P ) (P )
λ̂A = λA + λT ν̃A (P )
, λ̂T = λT + λA ν̃A (54)
ẼA
Analytical Methods of Predicting Performance… 207
x1
h
x1 = L
Plane External
symmetry surface
h1 h2 hi hn
Axial
direction
Through-thickness
direction
x3
(1) (i−1) (i) (n−1) (n)
0 x3 x(2)
3 x3 x3 x3 x3
Figure 3. Schematic diagram of geometry for one half of a general sym-
metric laminate
n
h= hi . (55)
i=1
⎡ ⎤ ⎡ ⎤
εA ε11
⎢ εT ⎥ ⎢ ε22 ⎥
⎢ ⎥ ⎢ ⎥
⎢ εt ⎥ ⎢ ε33 ⎥
⎢ ⎥ ≡⎢ ⎥
⎢ γt ⎥ ⎢ 2ε23 ⎥
⎢ ⎥ ⎢ ⎥
⎣ γa ⎦ ⎣ 2ε13 ⎦
γA 2ε12
⎡ ⎤
(L)
S11 S12
(L) (L)
S13 0 0 S16 ⎡
(L) ⎤ ⎡ (L) ⎤
σA V1
⎢ (L) (L) ⎥
⎢ S12 (L)
S22
(L)
S23 0 0 S26 ⎥⎢ ⎥ ⎢ (L) ⎥
⎢ (L) (L) ⎥⎢
σT ⎥ ⎢ V2 ⎥
⎢S (L) (L)
S36 ⎥ ⎢ ⎥ ⎢ (L) ⎥
=⎢
S23 S33 0 0 ⎥⎢ σt ⎥+⎢ V3 ⎥ ⎥ΔT,
⎥ ⎢
13
⎢ 0 0 ⎥⎢⎥ ⎥
⎥ ⎢
(L) (L) τt
⎢ 0 0 S44 S45 ⎢ 0 ⎥
⎢ (L) (L) ⎥⎣ τa ⎦ ⎢
⎣ 0 ⎦
⎣ 0 0 0 S45 S55 0 ⎦
(L) (L) (L) (L) τA V6
(L)
S16 S26 S36 0 0 S66
(56)
where the superscript (L) is used to denote effective thermoelastic constants,
Analytical Methods of Predicting Performance… 209
and where
(L) (L) (L)
(L) 1 (L) νA (L) νa (L) λA
S11 = (L)
, S12 = − (L)
, S13 = − (L)
, S16 = − (L)
,
EA EA EA μA
(L) (L)
(L) 1 (L) νt (L) λT (L) 1
S22 = (L)
, S23 = − (L)
, S26 = − (L)
, S33 = (L)
,
ET ET μA Et
(L)
(L) λt (L) 1 (L) (L) 1 (L) 1
S36 = − (L)
, S44 = (L) , S45 = Φ(L) , S55 = (L) , S66 = (L)
,
μA μt μa μA
(L) (L) (L) (L) (L) (L) (L) (L)
V1 = αA , V2 = αT , V3 = αt , V6 = αS
(57)
It can be shown that the laminate properties appearing in (57) are calculated
using the following relations
2 2
(L) (L)
1 1 λA 1 1 λT
(L)
= (L)
+ (L)
, (L)
= (L)
+ (L)
,
EA ẼA μA ET ẼT μA
2
(L)
1 1 λt (L)
ν̃t λt λT
(L) (L)
(L) (L)
(L)
= (L)
+ (L)
, νt = ET (L)
− (L)
,
Et Ẽt μA ẼT μA
(L) (L) (L) (L) (L) (L)
(L) (L) ν̃A λA λ T (L) ν̃a λ t λA
νA = EA (L)
− (L)
, νa(L) = EA (L)
− (L)
,
ẼA μA ẼA μA
(L) (L) (L) (L) (L) (L) (L) (L)
αA = α̃A − λA αS , αT = α̃T − λT αS ,
(L) (L) (L) (L)
αt = α̃t − λ t αS ,
(58)
where
2 2
(L) (L)
Ω12 (L)
Ω12 Ω12
(L) (L) (L) (L) (L)
ẼA = Ω11 − (L)
, ν̃A = (L)
, ẼT = Ω22 − (L)
,
Ω22 Ω22 Ω11
(L)
(L) (L) (L) (L) (L) (L) ẼT (L)
ν̃a(L) = Ω13 − ν̃A Ω23 , ν̃t = Ω23 − ν̃A Ω ,
(L) 13
ẼA
2 2
(L) (L) (L) (L)
1 ν̃a (L) ν̃a ν̃t (L) ν̃t (L) (L)
(L)
= (L)
Ω11 + 2 (L) (L) Ω12 + (L)
Ω22 + Ω33 ,
Ẽt ẼA ẼA ẼT ẼT
(59)
(L) (L)
(L) 1 (L) ν̃A (L) (L) 1 (L) ν̃A (L)
λA = Ω −
(L) 16
Ω , λT
(L) 26
= Ω −
(L) 26 (L)
Ω16 , (60)
ẼA ẼA ẼT ẼA
210 N. McCartney
(L) (L) (L) (L) (L) (L)
λt = Ω36 − Ω13 λA − Ω23 λT , (61)
2 2
(L) (L) (L) (L) (L) (L) (L) (L) (L)
μA = Ω66 − Ω11 λA − 2Ω12 λA λT − Ω22 λT (62)
(L) (L)
(L) 1 (L) ν̃A (L) (L) 1 (L) ν̃A (L)
α̃A = ω
(L) 1
− ω , α̃T
(L) 2
= ω
(L) 2
− (L)
ω1 , (63)
ẼA ẼA ẼT ẼA
1 1
n n
(L) (i) (L) (i)
Ω11 = hi Ω11 , Ω12 = hi Ω12 ,
h i=1 h i=1
1 1
n n
(L) (i) (L) (i)
Ω22 = hi Ω22 , Ω33 = hi Ω33 ,
h i=1 h i=1
(65)
1 1
n n
(L) (i) (L) (i)
Ω16 = hi Ω16 , Ω26 = hi Ω26 ,
h i=1 h i=1
1
n
(L) (i)
Ω66 = hi Ω66 ,
h i=1
1 1 1
n n n
(L) (i) (L) (i) (L) (i)
Ω13 = hi Ω13 , Ω23 = hi Ω23 , Ω36 = hi Ω36 , (66)
h i=1 h i=1 h i=1
1 1
n n
(L) (i) (L) (i)
ω1 = hi ω 1 , ω 2 = hi ω 2 ,
h i=1 h i=1
(67)
1 1
n n
(L) (i) (L) (i)
ω3 = hi ω 3 , ω 6 = hi ω 6
h i=1 h i=1
These relations are derived by applying the following expressions for effective
stresses and strains to the relations (50) that apply to individual plies
1 1 1 1
n n n n
(i) (i) (i) (i)
σA = hi σA , σT = hi σT , τA = hi τA , εt = hi ε t
h i=1 h i=1 h i=1 h i=1
(68)
Analytical Methods of Predicting Performance… 211
The effective properties characterising the out-of-plane shear of the laminate
are given by
1
n
(L) m2i n2i 1
S44 = hi (i)
+ (i) = (L) ,
h i=1 μt μa μt
1
n
(L) 1 1
S45 = h i mi n i (i)
− (i) = Φ(L) , (69)
h i=1 μ
a μt
(L) 1 n
mi 2
ni2
1
S55 = hi (i)
+ (i) = (L) ,
h i=1 μa μt μa
where
1 1 1
n n n
(L) (i) (L) (i) (L) (i)
S44 = hi S44 , S45 = hi S45 , S55 = hi S55 (70)
h i=1 h i=1 h i=1
where
1 1 νA 1
a11 = , a22 = , a12 = − , a66 = , a = a26 = 0 (72)
ET EA EA μA 16
212 N. McCartney
Putting
a22 2a12 + a66
γ= , δ= , (85)
a11 2a11
it follows from (78) that
s2 = −δ ± δ2 − γ (86)
It can be shown that for both GRP and CFRP, the values of s 2 are real
and negative. The following distinct pure imaginary roots are, therefore,
obtained
s1 = i δ − δ 2 − γ, s2 = i δ + δ 2 − γ (87)
The other two roots of (78) are given by s̄1 = −s1 , s̄2 = −s2 . The stress and
displacement representation automatically satisfies the equilibrium equa-
tions and the stress-strain relations (71) for any analytic functions φ(z) and
ψ(z) of the complex variable z. They are now assumed to take the following
form:
)t2
1 1
φ(z) ≡ w(t)ρ̂(t) ln dt, (88)
2πi z−t
t1
)t2
1 1
ψ(z) ≡ w(t)σ̂(t) ln dt (89)
2πi z−t
t1
The density functions ρ̂(t) and σ̂(t) are assumed to be polynomials and
t 2 − t1
w(t) ≡ 1/2
(90)
[ (t − t1 )(t − t2 ) ]
214 N. McCartney
)t2 )t2
w(t)ρ̂(t)dt = 0, w(t)σ̂(t)dt = 0 (91)
t1 t1
)t2
1 w(t)ρ̂(t)
φ (z) ≡ dt, (92)
2πi t−z
t1
)t2
1 w(t)σ̂(t)
ψ (z) ≡ dt (93)
2πi t−z
t1
indicating that the stress field arising from the representation (79)-(81) has
zero net force applied at infinity.
The algebra may be simplified by changing variables to
ζ = ξ + iη, (95)
where
2z − (t1 + t2 )
ζ = ζ(z) = , (96)
t 2 − t1
x−a y−b
ξ= , η= (97)
c c
The crack is then described by −1 < ξ < 1, η = 0 and it follows from (92)
and (93) that
)1
1 ρ(s) ds
φ (z) ≡ √ , (98)
π 1−s 2 ζ(z) −s
−1
)1
1 σ(s) ds
ψ (z) ≡ √ , (99)
π 1 − s ζ(z) − s
2
−1
Analytical Methods of Predicting Performance… 215
where ρ(ξ) ≡ ρ̂(t), σ(ξ) ≡ σ̂(t). From (91) the crack closure conditions are
written
)1 )1
ρ(s)ds σ(s)ds
√ = 0, √ =0 (100)
1−s 2 1 − s2
−1 −1
d
Gn (ζ) = −nHn (ζ), n ≥ 1 (119)
dζ
N
An
φ(z) = −c Gn (ζ), (120)
n=1
n
N
Bn
ψ(z) = −c Gn (ζ) (121)
n=1
n
Let
ζ1 = ξ + s1 η and ζ2 = ξ + s2 η (122)
On substituting (106), (115), (120) and (121) into the representation (79)-
(83), the stresses and displacement components may be expressed
N
σxx = − (αn s1 s2 + βn (s1 + s2 )) Hn (ζ1 ) + s22 (αn s1 + βn )ΔHn (ζ1 , ζ2 ) ,
n=1
(123)
N
σyy = [αn Hn (ζ1 ) − (αn s1 + βn )ΔHn (ζ1 , ζ2 )] , (124)
n=1
N
σxy = [βn Hn (ζ1 ) + s2 (αn s1 + βn )ΔHn (ζ1 , ζ2 )] , (125)
n=1
N
1
u = c [(αn (a11 s1 s2 − a12 ) + βn a11 (s1 + s2 )) Gn (ζ1 )
n=1
n (126)
+ (a11 s22 + a12 )(αn s1 + βn )ΔGn (ζ1 , ζ2 ) ,
218 N. McCartney
N
1 s1 + s2 a22
v = c −αn a22 + βn a12 − Gn (ζ1 )
n=1
n s1 s2 s1 s2
+ a12 s2 + as22
2
(αn s 1 + β n ) ΔGn (ζ ,
1 2ζ ) ,
(127)
where
⎧
⎪
⎪ Hn (ζ1 ) − Hn (ζ2 )
⎪
⎪ , for s1 = s2 ,
⎪
⎪ s1 − s2
⎪
⎨ Hn (ζ1 ) − Hn (ζ2 )
ΔHn (ζ1 , ζ2 ) ≡ lim
⎪ s1 →s2
s1 − s2
⎪
⎪
⎪
⎪ n ζ
⎪
⎩ = −η
1
⎪ + 2 H(ζ1 ), for s1 = s2
(ζ12 −1 )
1/2
ζ1 −1
(128)
and
⎧
⎪
⎪ Gn (ζ1 ) − Gn (ζ2 )
⎨ , for s1 = s2 ,
ΔGn (ζ1 , ζ2 ) ≡ s1 − s2
⎪
⎪ Gn (ζ1 ) − Gn (ζ2 )
⎩ lim = −nηH(ζ1 ), for s1 = s2 ,
s1 →s2 s1 − s2
(129)
One limiting situation, s1 = s2 = i, occurs when the material is isotropic,
and in this case the expressions (123)-(127) coincide with those of McCart-
ney and Gorley (1987) for the case of parallel cracks. It should be noted
that
ΔGn (ζ1 , ζ2 ) → 0 , ΔHn (ζ1 , ζ2 ) → 0 as y → 0
It is deduced from (109) that the limiting values of Gn (ζ) on the crack faces
are given by
G±n (ξ) = Tn (ξ) ± i 1 − ξ Un−1 (ξ)
2 (131)
By considering the limiting distributions for the normal and tangential dis-
placements along the upper and lower surfaces of the crack, which are de-
noted by Vn+ , Vn− , Ut+ , Ut− , use can be made of (126) and (127) together with
(131) to obtain an expression for the displacement discontinuities across the
Analytical Methods of Predicting Performance… 219
crack:
√ αn
N
Δv(ξ) ≡ (Vn+ − Vn− )(ξ) = 4c a22 g 1 − ξ 2 Un−1 (ξ), (132)
n=1
n
N
√ βn
Δu(ξ) ≡ (Ut+ − Ut− )(ξ) = 4c a11 g 1 − ξ2 Un−1 (ξ), (133)
n=1
n
where
1 √
g= (2 a11 a22 + 2a12 + a66 ) (134)
4
In the isotropic limit, this result agrees with the corresponding result of
McCartney and Gorley (1987).
where Sn (ξ) and St (ξ) are the normal and tangential tractions acting on
ξ < −1, η = 0. It follows from (80), (81), (106), (108) and (115) that on
ξ < −1, η = 0 the tractions are given by
N
Tn (ξ)
Sn (ξ) + iSt (ξ) = − (αn + i βn ) 2 + Un−1 (ξ) (136)
n=1 ξ −1
√ N
KI1 + iKII
1
= πc (−1 )n+1 (αn + i βn ) , (137)
n=1
where Sn (ξ) and St (ξ) are the normal and tangential tractions acting on
ξ > 1, η = 0. For ξ > 1, η = 0,
N
Tn (ξ)
Sn (ξ) + iSt (ξ) = (αn + i βn ) 2 − Un−1 (ξ) (139)
n=1 ξ −1
220 N. McCartney
which are identical to the numerical estimates, thus confirming the validity
of the methodology based on orthogonal polynomials.
It should be noted that the relation (139) can be used to investigate
magnitude of the tractions on the crack surfaces. For the example considered
the tractions, which should be zero, have the order of 10−16 indicating the
very high accuracy of the methodology used.
Consider now a test example of three equally spaced vertically stacked
cracks of equal length embedded in an infinite isotropic material. Figure
Analytical Methods of Predicting Performance… 221
2c x
2c
2s
5 illustrates the geometry where the three cracks have length 2c and the
vertical separation of the cracks is s. A uniaxial stress σ is applied in a
direction normal to the crack planes. The crack problem is such that the
deformation at the tips of the central crack is mode I, and the deformation
at the other tips is mixed mode. The magnitudes of the model I and mode
II stress intensity factors for the upper and lower cracks are expected to be
the same.
For a unit applied stress and when c = 0.5 c0 , s = 0.5c0 and on selecting
N = 100, for any normalising crack length c0 , the methodology based on
222 N. McCartney
x
s 2c
It is seen that the mode I stress intensity factor for the central crack is less
than that of the upper and lower cracks. This illustrates the shielding effect
on the central crack because of the presence of the other two cracks. The
model II stress intensity of the lowest crack is seen to be negative because the
local shear stress is negative. For this example the tractions, which should
be zero, have the order of 10−16 again indicating the very high accuracy of
Analytical Methods of Predicting Performance… 223
2h 2h
x1 x1
x3 x3
2L
0 0
b b b b
2a 2a
(a) (b)
Figure 6. Representative volume elements for a cracked cross-ply laminate
where σij are the stress components. The plies are regarded as transverse
isotropic solids so that the stress-strain-temperature relations involve the
axial and transverse values of the Young’s modulus E, Poisson’s ratio ν,
shear modulus μ and thermal expansion coefficient α. Superscripts ‘0’ or
‘90’ will be used to denote the ply to which a stress, strain and displacement
Analytical Methods of Predicting Performance… 225
where the strain and displacement components are denoted by εij and
ui respectively. The subscripts A, T and t are attached to the proper-
ties to associate them respectively with the axial, in-plane transverse and
through-thickness directions of the lamina. It should be noted that the up-
per case subscripts A and T are associated only with in-plane directions,
while the lower case subscripts are associated with the through-thickness
direction. The relations (144) are either obtained by modifying directly the
relations (143) for the 0o plies, or by using the relations (27)-(40) to ro-
tate the ply by an angle ± 90o . The thermoelastic constants of individual
226 N. McCartney
2 = ±W εT ,
u02 = u90 on x2 = ±W, (147)
where εT is the in-plane transverse strain that is uniform everywhere in the
laminate when generalised plane strain conditions are imposed. The edges
x2 = ±W are assumed to have zero shear stresses so that
0
σ12 90
= σ12 0
= 0, σ23 90
= σ23 = 0, on x2 = ±W (148)
For the above boundary conditions, and because of the symmetric nature of
the laminate, there will be symmetry about x3 = 0 of the stress, strain and
displacement distributions such that the following conditions are satisfied
0 90 0 90
σ13 = σ13 = 0, σ23 = σ23 = 0, u03 = 0, onx3 = 0 (149)
When applying laminate edge conditions applied on planes normal to the
x1 -axis, two possible approaches can be made. Consider first of all the RVE
shown in Fig. 6(a) which can be used for undamaged laminates, and for
damaged laminates where a ply crack in the 90o ply is located at x1 = 0.
The edges x1 = ±L are such that in-plane axial displacement is uniform
having the following values
1 = ±LεA ,
u01 = u90 on x1 = ±L, (150)
where εA is the effective axial applied strain. The edges x1 = ±L are
assumed to have zero shear stresses so that
0
σ12 90
= σ12 = 0, 0
σ13 90
= σ13 = 0, on x1 = ±L (151)
Analytical Methods of Predicting Performance… 227
90
90
1 1 90 2 ET 1 1 90 2 ET
= 90 1 − (νA ) 90 , = 90 1 − (νA ) 90 ,
90
ẼA EA EA ẼT90 ET EA
90 90 90 90 90 90 90 90
ν̃a ν ν ν ν̃t ν ν ν
= a90 + t 90A , = t90 + a 90A , (155)
90
ẼA E A E A ẼT90 E T EA
90
90 90 ET 90
α̃90
A = αA + νA 90 αT , α̃90 90
T = αT + νA αA
90 90
EA
From a consideration of mechanical equilibrium the uniform ply stresses can
be used to define, for an undamaged laminate, the effective axial stress σ̂A
and the effective in-plane transverse stress σ̂T as follows
0 90
hσ̂A = bσ̂11 + aσ̂11 , (156)
228 N. McCartney
0 90
h σ̂T = bσ̂22 + aσ̂22 (157)
The ‘hat’ symbol is used to distinguish these effective stresses from those
that will result when the laminate is damaged. Corresponding to the uni-
form through-thickness stress σt , an effective through-thickness strain ε̂t
can be defined by the relation
hε̂t = bε̂033 + aε̂90
33 (158)
It should be noted that the value σt for the through-thickness stress of an
undamaged laminate corresponds to the effective value when the laminate
is damaged. It can be shown that on defining the constants
b 0 a b 0 0 a 90 90 b a
A = ẼA + ẼT90 , B = νA ẼT + νA ẼT , C = ν̃a0 + ν̃t90 ,
h h h h h h
b 0 a 90 b 0 a 90
F = ẼT + ẼA , G = ν̃t + ν̃a , (159)
h h h h
b 0 0 a b 0 0 a 90 90
P = ẼA α̃A + ẼT90 α̃90
T ,Q = Ẽ α̃ + ẼA α̃A ,
h h h T T h
the thermoelastic constants of an undamaged cross-ply laminate are given
by
(L) B2 (L) B2 (L) B
EA = A − , ET = F − , νA = ,
F A F
(L) BG (L) BC
νa = C − , νt = G − , (160)
F A
1 (L)
(L) (L) (L) 1 (L) ET
αA = (L) P − νA Q , αT = (L) Q − νA (L)
P ,
EA ET EA
such that the in-plane non-shear stress-strain relations for an undamaged
laminate are
(L) (L)
1 νA νa (L)
ε̂A = σ̂ −
(L) A
σ̂ −
(L) T
σ + αA ΔT,
(L) t
EA EA EA
(L) (L) (161)
ν 1 ν (L)
ε̂T = − A(L) σ̂A + (L) σ̂T − t(L) σt + αT ΔT
EA ET ET
It should be noted that
(L)
(L) b 0 a 90 (L) ET B b ν 0 Ẽ 0 + aνA90 90
ẼT
ẼA = A = Ẽ + Ẽ , νA == A T0 ,
h A h T (L)
EA A bẼA + a ẼT90
(L) b a 90 (L) B bν 0 Ẽ 0 + aνA 90 90
ẼT
ẼT = F = ẼT0 + Ẽ , νA = = A T0 ,
h h A F bẼT + aẼA 90
(162)
Analytical Methods of Predicting Performance… 229
where
(L) (L)
(L) EA (L) ET
ẼA = (L) (L) (L)
, ẼT = (L) (L) (L)
(163)
1 − (νA )2 ET /EA 1 − (νA )2 ET /EA
such that
(L) (L)
νa νt 1 (L)
ε̂t = − σ̂ −
(L) A (L)
σ̂T + (L)
σt + αt ΔT, (166)
EA ET Et
It can be shown that the same values result for the minor Poisson’s ratios
(L) (L)
νa and νt when using the relations (160) or (165).
1 0 ν̃a0 0 0
0 ET
ε011 = σ
0 11
− σ
0 33
+ α̃0
A ΔT − ν A 0 εT , (171)
ẼA ẼA EA
ν̃a0 0 1 0
ε033 = − 0
σ11 + 0 σ33 + α̃0t ΔT − νt0 εT , (172)
ẼA Ẽt
where
1 1 (ν 0 )2
= 0 − t 0 , α̃0t = α0t + νt0 α0T (173)
0
Ẽt Et ET
For the 90o ply
90
σ22 90 90
= νA σ11 + νa90 σ33
90
− EA
90 90 90
αA ΔT + EA εT , (174)
1 90 ν̃ 90 90
ε90
11 = 90
σ11 − t90 σ33 T ΔT − νA εT ,
+ α̃90 90
(175)
ẼT ẼT
ν̃t90 90 1 90
33 = −
ε90 t ΔT − νa εT ,
+ α̃90 90
σ11 + 90 σ33 (176)
ẼT90 Ẽt
where
1 1 (νa90 )2
= 90 − 90 , α̃90 90 90 90
t = αt + νa αA (177)
Ẽt90 Et EA
(184)
90 0
ν̃ ν̃
u03 (x1 , x3 ) = ε̂033 (x3 − a) + ε̂90 33 a + b 90 −
t a
0
(x3 − a) C(x1 )
ẼT ẼA
1 1 3 ab
+ b − (h − x 3 )3
+ (2h + b) C (x1 ),
6 Ẽt0 Ẽt90
(185)
b ν̃t90 b 1
90
u3 (x1 , x3 ) = C(x1 )x3 + C (x1 ) 3ah − x3 x3 + ε̂33 x3 (186)
2 90
a ẼT90 6a Ẽt90
It follows that
0 90 0 90
hσA = bσ11 (x1 , x3 ) + aσ11 (x1 , x3 ) = bσ̂11 + aσ̂11 = hσ̂A , (187)
so that the effective applied axial stress for damaged laminate σA is equal to
the effective axial stress σ̂A for the corresponding undamaged laminate. The
function A(x1 ) appearing in (183) and (184) is for the moment arbitrary.
The representation automatically satisfies the equilibrium equations (142)
and the required interface continuity conditions. In addition, all the stress-
strain relations (143) and (144), except for the two relations (143)1 and
(144)1 involving the axial strains ε011 and ε90 11 respectively, for any functions
C(x1 ) and A(x1 ). It is possible, however, to satisfy these axial relations after
they are averaged through the thickness of the 0o and 90o plies respectively,
as will now be described.
Assuming symmetry about the mid-plane x3 = 0, the average of any
quantities f0 (x1 , x3 ) and f90 (x1 , x3 ) associated with the 0o and 90o plies are
defined respectively by
)h )a
1 1
f¯0 (x1 ) = f0 (x1 , x3 ) dx3 , f¯90 (x1 ) = f90 (x1 , x3 ) dx3 (188)
b a
a 0
Analytical Methods of Predicting Performance… 233
where
)x1
C̄(x1 ) ≡ C(x)dx. (195)
0
1 ν̃ 0
ū01 (x1 ) = 0
C̄(x1 ) − a0 b2 C (x1 ) + ε̂A x1 , (196)
ẼA 6ẼA
b 1 2a + 3b ν̃t90 2
1 (x1 ) = −
ū90 C̄(x1 ) − b C (x1 ) + ε̂A x1 (197)
a ẼT
90 6b ẼT90
b
90
σ13 (L, x3 ) = C (L)x3 = 0, implying C (L) = 0, (203)
a
b a 90
90
σ11 (L, x3 ) = − C(L) + σ̂11
90
= 0, implying C(L) = σ̂ (204)
a b 11
On applying these conditions and on writing P = A + B, Q = A − B, the
parameters A and B must be selected so that
(p−q)L (p+q)L
a (p − q) tanh b a (p + q) tanh b
A=− 90
Λσ̂11 , B= 90
Λσ̂11 , (205)
b cosh (p+q)L
b
b cosh (p−q)L
b
where
1 (p + q)L (p − q)L
= (q + p) tanh + (q − p) tanh (206)
Λ b b
The only boundary condition for a damaged laminate that has not been
satisfied is given by (153)1 . It is clear from (183), (184) and (193) or (194)
that it is not possible for this boundary condition to be satisfied by the ap-
proximate solution derived. The boundary condition (153)1 is now replaced
by the following averaged condition
where
4Λpq (p + q)L (p − q)L
Φ= tanh tanh (209)
p −q
2 2 b b
On using (157) it can be shown using (180) and (181) that the effective
applied transverse stress σT defined by (167)2 is given by
0
ab 0 ET
σT = νA 0 − νA Φσ̂11
90 90
+ σ̂T (210)
Lh EA
236 N. McCartney
1 1 2 a Ẽ 90 Φ
(L) 90
= (L)
+ 1 − ν A ν A
T
, (214)
EA EA L E (L) Ẽ 0 ξ
A A
2 2
0 0 90
1 1 0 ET b a ẼA ẼT Φ
= 1 + νA 0 − νA 90
, (215)
ET (L)
ET EA h2 L Ẽ (L) E (L) ξ
A T
(L)
implying ET = ξET ,
νA
(L)
νA
E0
b a ẼT90 Φ
(L) 90
= (L)
+ 1 − ν A ν A ν 0
A
T
0 − ν 90
A (L) (L)
, (216)
EA EA EA hLE E ξ
A T
(L)
νa νa
= (L)
EA
EA
(L)
ν 90
(L)
ν 90 (L) 90
νa νt (L) 90 a ẼA ẼT Φ
+ (L)
− t90 + νA
90
(L)
− a90 1 − νA νA ,
EA ET ET EA L E (L) ẼA
0 ξ
A
(217)
(L)
νt νt
= (L)
ET
ET
(L) (L)
νa ν 90 νt ν 90 0
0 ET
90
90 b a ẼT Φ
− (L)
− t90 + νA
90
(L)
− a90 νA 0 − νA ,
EA ET ET EA EA h L E (L) ξ
T
(218)
Analytical Methods of Predicting Performance… 237
a Ẽ (L) Ẽ 90 Φ
(L) (L) 90 (L) 90 (L)
αA = αA + 1 − νA νA αA + νA αT − α̃90
T
A T
, (219)
L E (L) ẼA0 ξ
A
E0
b a ẼT90 Φ
(L) (L) 90 (L)
αT = αT − αA + νA αT − α̃90T
0
νA T
0 − νA
90
, (220)
EA h L E (L) ξ
T
where
2 2
0 0 90
0 ET b a ẼA ẼT
ξ = 1 − νA 0 − ν 90
A 2
Φ. (221)
EA h L Ẽ E (L)
(L)
A T
where
νa νa
(L) 90
90 (L) b a ẼT Φ
= (L)
+ Ω 1 − ν ν
A A , (224)
EA EA h L E (L) ξ
A
(L)
0
2 0 90
νt νt 0 ET b a ẼA ẼT Φ
= (L)
− Ω ν A 0 − ν 90
A 2
, (225)
ET ET EA h L Ẽ E (L) ξ
(L)
A T
$ %
(L) (L)
1 1 νa νt90 νt νa90 b a 90 Φ
= (L) + Ω − 90 + νA 90
− 90 Ẽ , (226)
Et Et E
(L) ET E
(L) EA hL T ξ
A T
ba Φ
(L) (L) (L)
αt = αt − Ω αA + νA
90
αT − α̃90
T Ẽ 90 , (227)
hL T ξ
and where
(L)
ν̃ 0 ν̃ 90 ν 0
0 ET
Ω = a0 − t90 − t(L) νA 0 − νA
90
(228)
ẼA ẼT ET EA
238 N. McCartney
The relations (217) and (218) are equivalent to the results (224) and (225)
because it can be shown that
(L) (L)
νa νt90 νt νa90 b ẼA 0
(L)
− 90 + νA90
(L)
− 90 = Ω (L)
(229)
E ET E EA h Ẽ
A T A
When these ply properties are used in conjunction with the formulae
(214)-(220) and (224)-(227), for a set of ply crack densities in the range
0 – 4 cracks/mm, the results shown in Fig. 7 are obtained. The results
shown assume the following identifications:
It is noted that for ply crack densities exceeding 2/mm, the effective
properties no longer depend on the crack density. Also, it is seen that the
effective in-plane transverse modulus ET is hardly affected by ply cracking,
and that the effective axial thermal expansion coefficient is affected a great
deal by ply cracking.
Similar situations arise for ply cracking in the 90o plies of general sym-
metric laminates for a variety of laminate configurations considered in the
WWFE III International Exercise (Kaddour et al., 2013a) concerned with
the assessment of damage models for composite laminates. Results anal-
ogous to those derived here for cross-ply laminates have been derived for
general symmetric laminates (McCartney, 2013a,b) and assessed/discussed
by the organisers of the Exercise (Kaddour et al., 2013b). However, the
author recommends that a great deal of caution is applied when consider-
ing the comparison models as the information presented by the organisers
Analytical Methods of Predicting Performance… 239
1
Normalised properties
0.9
EA
ET
0.8
Est
nuA
0.7
nusa
nust
0.6 alA
alT
0.5 alst
0.4
0 0.5 1 1.5 2 2.5 3 3.5 4
Ply crack density (/[mm])
Figure 7. Predictions of the normalised effective properties of a simple
cross-ply laminate as a function of the density of a uniform distribution of
ply cracks in the 90o ply
(Kaddour et al., 2013b) relating to the generalised plane strain model devel-
oped by the author is wholly misleading, and conclusions are not justified
by the information presented to the exercise by the author (see McCartney
(2013b)).
F (t)
Fibre Matrix
bundle
Vf Vm
σ(t) σm (t)
Eb (t) Em
ε(t) ε(t)
Ef
F (t)
failure process. The composite is subject to a fixed applied load F for all
times t > 0, where t = 0 corresponds to the time when the fixed load F is
first applied. Environmental defect growth in the fibres leads to progressive
fibre failure until the bundle collapses. It is assumed that bundle collapse
corresponds to the catastrophic failure of the composite, i.e. the matrix
strength is insufficient to maintain the load when all the fibres have failed.
The objective is to develop the parallel bar model of a composite so that
it can predict the dependence of composite life tf on the fixed applied load
F, and the dependence of the residual strength F *(t ) of the composite on
elapsed time t from the instant of first loading.
242 N. McCartney
Ab Am
Vf = , Vm = = 1 − Vf (230)
Ab + Am Ab + Am
The load applied to the fibre bundle at time t is denoted by Fb (t ), the stress
in each surviving fibre being denoted by σ(t). The cross-sectional area of
each of the fibres in the bundle is denoted by A, and the axial modulus of
each fibres is denoted by Ef which is assumed to be time independent. The
axial stress at time t in the matrix is denoted by σm (t). The modulus of
the matrix is denoted by Em which is assumed to be independent of time.
A time dependence could be included to account for visco-elastic effects, or
for time-dependence arising from matrix ageing.
The axial strain in all surviving fibres of the bundle and the matrix
has the same time dependent value that is denoted by ε(t). As thermal
expansion mismatch effects are neglected it follows that
σ(t) σm (t)
ε(t) = = (231)
Ef Em
The balance of forces in the parallel bar model leads to the equilibrium
relation
Fb (t) + Am σm (t) = F (232)
The number of surviving fibres in the bundle at time t is denoted by N (t )
so that the load applied to the bundle at time t may be written
Fb (t)
σb (t) = , (235)
Ab
and since the axial strain of the bundle and the individual fibres has the
value σ(t)
where use has been made of (233) and (235). Clearly the effective axial
modulus of the fibre bundle is given by
N (t)
Eb (t) = Ef (237)
N0
F
σapp = , (238)
Ab + Am
and it can be shown from (230) and (234), together with the fact that
Ab = N0 A, that
2 )t
KIc
X0 (t) = 2 ⎣σ n−2 (t) + (n − 2)λ σ n (τ )dτ ⎦ , (243)
y
0
Analytical Methods of Predicting Performance… 245
where
1 n−2 2
λ= CKIc y (244)
2
On using (240) it follows from (243) that the initial strength σi (t) of the
fibres, that fail at time t when their stress is σ(t), is given by
⎡ ⎤ n−2
1
)t
σi (t) = ⎣σ n−2 (t) + (n − 2)λ σ n (τ )dτ ⎦ (245)
0
The corresponding static strength for the composite is then obtained using
Fmax m
m+1
mσ̂max
= F̂max = σ̂max α + e−σ̂max = α m −1
, (250)
N0 A σ0 m σ̂max
which is consistent with the known result for a loose bundle Kelly and
McCartney (1981) when the limit α → 0 is taken.
The equation (249) governing the maximum fibre stress does not always
have a solution as is easily seen by examining the form of the LHS and RHS
m
of (249). On letting x = σ̂max the critical conditions defining the limit of
solutions to (249) may be written
If this condition is not satisfied then it is deduced that the fibres progres-
sively fail until there is just one surviving fibre which will then fail, i.e. the
bundle does not suddenly collapse. The value of the Weibull modulus m for
fibres of interest is usually such that the condition (253) is satisfied so that
bundle collapse is always expected in practice.
where
1 n−2 2 2
η = λσ02 =
CKIc y σ0 , (256)
2
and where use has been made of the definitions (248), which when applied
to the load sharing rule (234) lead to
F̂
N̂ (t) = −α (257)
σ̂(t)
⎣ 1 F̂ σ̂(t) dσ̂(t)
ln − σ̂ n−2 (t)⎦ = σ̂ n+1 (t) (259)
m F̂ − ασ̂(t) F̂ − ασ̂(t) d(ηt)
The stress σ̂f in the surviving fibres when the composite fails can thus be
determined using numerical methods without having to solve the differential
equation (259). It should be noted that when F̂ = F̂max the solution of (262)
248 N. McCartney
is given by σ̂f = σ̂max where σ̂max and F̂max are given by (249) and (250)
respectively. The time to failure for the composite is denoted by tf .
The number of surviving fibres just before composite failure is obtained
using (254) and is given by
N (tf ) m σi (tf )
= N̂f = e−σ̂i , σ̂i = (263)
N0 σ0
The transcendental equation (262), that usually must be solved numerically,
involves the dimensionless loading parameter F̂ in a complicated way. It
is useful to unravel the dependence on this parameter by using the load
sharing rule (257) to express (262) in terms of Nf as follows
n−1 n−m−2
1 N̂ f + α 1
m
F̂ n−2 = ln (264)
m N̂f N̂f
Having solved (264) to find N̂f using numerical methods, the normalised
failure stress is obtained, on making use of (234), from the relation
F̂
σ̂f = (265)
N̂f + α
)t
∗
2−n 2−n 1
(X ) 2
= (a∗c ) 2
+ C (n − 2)y n σ n (τ )dτ (267)
2
0
On using (240) the initial strength of the fibres that are critical at time t
when the fibre stress has the value s is denoted by si and is given, on using
(267), by the relation
)t
ŝn−2
i = ŝ n−2
+ (n − 2)η σ̂ n (τ )dτ, (268)
0
ŝn−2
i = ŝn−2 − k(t), (270)
where
n−2
m
σ̂(t)
k(t) = σ̂ n−2 (t) − ln (271)
F̂ − ασ̂(t)
The load applied to the composite Fs , when the fibre stress has the value
s, is obtained from (248) and (257) so that
Fs
= F̂s = ŝ α + N̂s , (272)
N0 σ0 A
where F̂s is the normalised applied load and where N̂s is the normalised
number of surviving fibres when the load on the composite is such that the
fibre stress has the value s. It follows from (246) that
m
N̂s = e−ŝi (273)
On using (270) the stress σmax (t) in the surviving fibres just before the
composite fails during a residual strength test is obtained from
It then follows from (270) and (274) that the residual strength of the com-
posite S (t ) is obtained using
S(t) m
= σ̂max (t) α + e−x (t) = Ŝ(t). (277)
σ0
When t = 0 it can be shown using (258) that k (0) = 0 in which case the
transcendental equation (275) reduces to the form (249) which needs to be
solved when calculating the static strength of the composite.
0.8
0.6
F/Fm
F/Fm
F ∗ /Fm = 0.2
0.4
F ∗ /Fm = 0.4
F ∗ /Fm = 0.6
0.2
0
-10 -5 0 5 10 15
log10 (ηt)
Figure 9. Schematic diagram of the parallel bar model of a unidirectional
composite for predicting effects of environmental exposure on axial compos-
ite properties
4. The level of loading applied axially to the composite where the model
assumes that the ratio F/Fm is given where F is the axial load applied
to the composite and Fm is the static strength, i.e. the strength of the
composite before environmental exposure. The value of F/Fm always
lies in the range 0 – 1.
The Euler-Richardson solution technique (Churchhouse, 1981) is used to
solve the ordinary differential equation (259) where the normalised dimen-
sionless time η t may be regarded as an unknown function of σ̂. In other
words, the differential equation can be used directly to determine an incre-
ment in the value of η t for any given increment in σ̂. The initial condition
is specified by (260) and (261) and the range s0 ≤ σ̂ ≤ σ̂f is subdivided into
100 equal intervals when solving the differential equation. The upper limit
σ̂f is determined by the relations (264) and (265). Figure 9 shows the result
of solving the differential equation (259) to find the normalised time η tf
for various values of the loading ratio F/Fm . The normalising parameter
η is defined by (256). It is seen that as F/Fm → 1 the lifetime tends to
zero. Figure 9 also shows predictions of the normalised residual strength Ŝ
defined by (277), as a function of the normalised time η t.
The principal conclusion to be drawn from the results presented is that
the time dependence of the axial properties of a unidirectional fibre rein-
forced glass composite subject to environmental exposure under fixed load
252 N. McCartney
can be predicted using a parallel bar model of the composite where interface
bonding is neglected. The model enables the prediction of the stress history
of the fibre stress in surviving fibres from the point of first loading to the
occurrence of catastrophic failure. Results not shown indicate that the fibre
stress is almost independent of the matrix properties, a situation that arises
because Em Ef .
The model can also be used to predict the time dependence of the resid-
ual strength of the composite, a property which does show some dependence
on matrix properties. However, results not shown indicate that, when the
residual strength is divided by the static strength, the resulting residual
strength ratio is virtually independent of the matrix properties. It is con-
cluded that the residual strength ratio for a unidirectional composite is
predictable (and therefore measurable) from the static strength of the com-
posite, and the time dependence of the residual strength of a loose bundle
of fibres.
6 Closing Remarks
A varied set of topics concerning the behaviour of composite materials has
been considered in this paper. They concern the estimation of the undam-
aged properties of plies in terms of fibre and matrix properties, the esti-
mation of the undamaged properties of general symmetric laminates, the
consideration of an elegant method of considering cracks in anisotropic ma-
terials using orthogonal polynomials, a detailed treatment of ply cracking
in a simple cross-ply laminate, and the modelling of the effects of envi-
ronmental exposure on the lifetime and residual strength of unidirectional
composites. Much of the work presented here has not been published be-
fore. For the analyses dealing with composite damage, example predictions
have been given to help readers understand the capabilities of the various
damage models.
It is hoped that readers of this paper will be convinced that analytical
modelling, which has been undertaken in some quite complex situations,
enables much deeper insight into the modelling of composite material sys-
tems than numerical solution methods permit, and provides opportunities
for convenient design methods based on relatively compact formulae rather
than on data tables and graphs that have to be generated when using nu-
merical methods such as finite element analysis.
Bibliography
W.R. Broughton and L.N. McCartney. Predictive models for assessing
long-term performance of polymer matrix composites. Technical Report
CMMT(A)95, NPL, Teddington, April 1998.
R.F. Churchhouse. Numerical methods. In W. Ledermann, editor, Handbook
of Applicable Mathematics, volume 3, pages 319–321, Chichester, 1981.
Wiley.
G.M.L. Gladwell and A.H. England. J. Mech. Appl. Math., 30:175, 1977.
Z. Hashin. Analysis of composite materials - a survey. Trans. ASME. J.
Appl. Mech., 50:481–505, 1983.
A.S. Kaddour, M.J. Hinton, P.A. Smith, and S. Li. The background to the
third world-wide failure exercise. J. Comp. Mater., 47(20-21):2417–2426,
2013a.
A.S. Kaddour, M.J. Hinton, P.A. Smith, and S. Li. A comparison between
the predictive capability of matrix cracking, damage and failure criteria
for fibre reinforced laminates: part a of the third world-wide failure
exercise. J. Comp. Mater., 47(20-21):2749–2779, 2013b.
A. Kelly and L.N. McCartney. Failure by stress corrosion of bundles of
fibres. Proceedings of the Royal Society of London, A374:475–489, 1981.
J.C. Maxwell. A Treatise on Electricity and Magnetism, volume 1. Claren-
don Press, Oxford, 1st edition, 1873.
L.N. McCartney. Time dependent strength of large bundles of fibres loaded
in corrosive environments. Fibre Science & Technology, 16:95–109, 1982.
L.N. McCartney. Theory of stress transfer in a 0-90-0 cross-ply laminate
containing a parallel array of transverse cracks. J. Mech. Phys. Solids,
40:27–68, 1992.
L.N. McCartney. Model of composite degradation due to environmental
damage. Technical Report CMMT(A)124, NPL, Teddington, September
1998.
L.N. McCartney. Maxwell’s far-field methodology predicting elastic prop-
erties of multiphase composites reinforced with aligned transversely
isotropic spheroids. Phil. Mag., 90:4175–4207, 2010.
L.N. McCartney. Derivations of energy-based modelling for ply cracking
in general symmetric laminates. J. Comp. Mater., 47(20-21):2641–2673,
2013a.
L.N. McCartney. Energy methods for modelling damage in laminates. J.
Comp. Mater., 47(20-21):2613–2640, 2013b.
L.N. McCartney and T.A.E. Gorley. Complex variable method of calculating
stress intensity factors for cracks in plates. In A.R. Luxmoore et al.,
editor, Proc. 4th Int. Conf. on Numerical Methods in Fracture Mechanics,
Swansea, 1987. Pineridge Press.
254 N. McCartney