Determination of Stress Intensity Factors

i i
“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 102 — #1

i i
5
Determination of Stress Intensity
Factors
5.1 Introduction
The stress intensity factor (SIF) plays the most pivotal role in the application of
linear elastic fracture mechanics (LEFM) principles to practice. It is useful in the
assessment of safety or reliability of a machine or structural component with a
crack. It enables the calculation of crack growth rate through a component under
fatigue loading, stress corrosion, etc. For the safety assessment, two things are
needed: the SIF corresponding to the loading on component and the fracture
toughness of its material. The latter is a material data obtained through
experiment. The former is obtainable in some situations from handbooks of SIFs
(Sih 1973a; Rooke and Cartwright 1976; Murakami et al. 1987; Tada, Paris and
Irwin 2000), while in others it has to be determined using either an analytical
method, or a numerical method, or an experimental technique. The analytical
techniques include complex stress function based approaches, boundary
collocation method, integral transform technique (Sneddon and Lowengrub
1969; Sneddon 1973), Green’s function method, weight function method, etc. The
numerical methods have been very widely employed for their versatility and
capability for handling complex geometry easily. The three important numerical
techniques are: finite element method (Wilson 1973; Atluri 1986), boundary
element method (Aliabadi, Rooke and Cartwright 1987; Cruse 1996;
Mukhopadhyay, Maiti and Kakodkar 2000; Rabczuk 2013), and meshless method
(Belytschko et al. 1996; Atluri and Zhu 1998). The experimental techniques
include strain gauge based method (Dally and Sanford 1987), photoelasticity
(Kobayashi 1975; Dally and Riley 1991; Ramesh 2000), and method of caustics
(Theocaris and Gdoutos 1976; Theocaris 1981; Rosakis and Zehnder 1985). A very
good account of the analytical and finite element based methods is given in
compilations by Sih (1973b) and Atluri (1986).
Downloaded from https:/www.cambridge.org/core. New York University, on 20 Feb 2017 at 16:36:15, subject to the Cambridge Core terms of use, available at
https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316156438.006
i i
i i
i i

i i
Determination of stress intensity factors 103
In this chapter, some important analytical methods, numerical technique based

on finite element method, strain gauge based technique, and photoelasticity are
only discussed.
5.2 Analytical Methods

In the analytical methods, the SIFs are calculated using the following relations for
Mode I, II, and III, respectively, provided the crack-tip stress field is given in terms
of r and θ.
√
K I = lim 2πr σy θ =0
(5.1)
r →0
√
K I I = lim 2πr τxy θ =0
(5.2)
r →0
√
K I I I = lim 2πr τyz θ =0
(5.3)
r →0
In all these cases, origin is at the crack-tip. x axis (θ = 0) is aligned with the crack
plane and it points towards the direction of crack extension (Fig. 5.1).
If the Westergaard stress function Fi (z), i = I or II for a problem is known, the
SIF can be obtained from
q
Ki = lim 2π (z − a) Fi (z), i = I or I I (5.4)
z→ a
The Westergaard stress functions are generally defined with centre of the crack
(size 2a) as the origin, and the crack-tip is located at z = a.
Figure 5.1 Plate with angled crack and crack-tip coordinates.
i i
i i
i i

i i
104 Fracture mechanics
If the stresses are given in terms of two Williams stress functions F (z) and χ(z),
which are generally defined with the crack-tip as origin, Modes I and II SIFs are
given by
√
K I − iK I I = lim 2 2πz F 0 (z) (5.5)
z →0
If the stresses are given in terms of two analytic functions, with crack centre as the
origin and the crack-tip is located at z = z1 , the SIFs are given (Bowie 1973) by
q
K I − iK I I = lim 2 2π (z − z1 ) F 0 (z) (5.6)
z → z1
If conformal mapping, z = ω (ζ ), is used to map the given problem to a convenient

geometry like a circle in the mapping plane ζ (zeta) and the crack-tip is located at
ζ = ζ 1 , the SIFs for mixed mode problem in two dimensions are given by
F 0 (ζ ) dF
q
K I − iK I I = lim 2 2π [ω (ζ ) − ω (ζ 1 ) ] 0
, F 0 (ζ ) = (5.7)
ζ →ζ 1 ω (ζ ) dζ
Problem 5.1
Determine K I and K I I for the case when stresses along the x axis (Fig. 5.2) are given
(Maiti and Smith 1983) by s
P + iQ a2 − b2
σy + iτxy = (5.8)
π ( x − b) x 2 − a2
where P and Q are specified in N/m.
Solution
Using Eq. (5.1)
Figure 5.2 Point loading on crack edges.
i i
i i
i i

i i
√
K IB = lim 2πr σy θ =0
r →0
q
= lim 2π ( x − a) σy θ =0
( x − a)→0
s
P a2 − b2
q
= lim 2π ( x − a)
x→a π ( x − b) x 2 − a2
r r
P 1 1 P a+b
q
= 2π ( a2 − b2 ) lim =√
π x→a x − b x+a πa a−b
r
P a+b
Therefore, K IB = √ (5.9a)
πa a−b
r
P a−b
Similarly, K I A =√ (Ans.). (5.9b)
πa a+b
Further,
√ q
K I IB = lim 2πr τxy θ =0
= lim 2π ( x − a) τxy θ =0
r →0 ( x − a)→0
s r
Q a2 − b2 Q a+b
q
= lim 2π ( x − a) 2 2
=√ (5.10a)
x→a π ( x − b) x − a πa a−b
It can be shown that

r
Q a−b
KI I A = √ (Ans.). (5.10b)
πa a+b
Problem 5.2
Westergaard stress function for uniform pressure
loading on a crack of size 2a in an
z
infinite plate is F (z) = p √ − 1 , where p is pressure. Determine the SIF.
z − a2
2
Solution
From Eq. (5.4)
q q
z √
a
K I = lim 2π (z − a) F (z) = lim 2π (z − a) p √ − 1 = p 2π √
z→ a z→ a 2
z −a 2 2a
√
= p πa (Ans.).
i i
i i
i i

i i
Problem 5.3
Given the stress functions for the case shown (Fig. 5.3) in x − y cordinates (Maiti
and Smith 1983):
Figure 5.3 Crack edge loading.
s
P(sin α − i cos α ) a2 − b2
F10 (z) = ψ10 (z) = ,
2π (z − b) z2 − a2
h i
where σy + σx = 2 F10 (z) + F10 (z)
h i
0
σy − σx + 2 i τxy = 2 (z − z) F100 (z) − F10 (z) + ψ1 (z) .
Determine the SIFs.
Solution
0
Along the crack line, that is, y = 0 or θ = 0, σy + i τxy = F10 (z) + ψ1 (z) ; z = z.
s s
P sin α a2 − b2 P cos α a2 − b2
Therefore, σy = 2 2
, and τxy =
π ( x − b) x − a π ( x − b ) x 2 − a2
r
P sin α a+b
q
KI B = lim 2π ( x − a) σy θ =0 = √ (Ans.). (5.11)
( x − a)→0 πa a−b
r
P cos α a+b
q
Similarly, K I I B = lim 2π ( x − a) τxy θ =0 = √ (Ans.). (5.12)
( x − a)→0 πa a−b
Problem 5.4
Airy stress function for the infinite plate
(Fig. 5.4) in the mapping plane ζ, where
a 1
the mapping function, z = ω (ζ ) = ζ+ , is given by
2 ζ
i i
i i
i i

i i
1 − e2 i α
F (ζ ) = −σa .
4ζ
Determine the SIFs.
Figure 5.4 Loading on edges of infinite plate.
Solution
Since the crack-tip is located at z = ±a, through the mapping function the
corresponding location in the ζ-plane is ζ 1 = ± 1. Further, ω 0 (ζ 1 ) = 0 at the
crack-tip.
Using Eq. (5.7)
F 0(ζ )
q
K I − iK I I = lim 2 2π [ω (ζ ) − ω (ζ 1 )]
ζ →ζ 1 ω 0(ζ )
F 0(ζ 1 + h)
q
= lim 2 2π [ω (ζ 1 + h) − ω (ζ 1 )] , ζ = ζ 1 + h.
h →0 ω 0(ζ 1 + h)
Expanding ω (ζ 1 + h), ω 0(ζ 1 + h), and F 0(ζ 1 + h) about ζ 1 by Taylor’s series, it can
be seen that
F 0(ζ )
q
lim ω (ζ ) − ω (ζ 1 ) 0
h →0 ω (ζ )
2
"r #
h2 F 0 (ζ 1 ) + hF 00 (ζ 1 ) + h2 F 000 (ζ 1 ) + . . . . . .
= lim hω 0 (ζ 1 ) + ω 00 (ζ 1 ) + . . . . 2
h →0 2 ω 0 (ζ 1 ) + hω 00 (ζ 1 ) + h2 ω 000 (ζ 1 ) + . . . . . .
F 0 (ζ 1 )
=p , noting that ω 0 (ζ 1 ) = 0
2ω 00 (ζ 1 )
i i
i i
i i

i i
√ F 0 (ζ 1 )
K I − iK I I = 2 π p (5.13)
ω 00 (ζ 1 )
Substituting the values

" # √
√ σa (1 − e2 i α )ζ −2 σ πa
K I − iK I I = 2 π p = [(1 − cos 2α) − i sin 2α]
4 a ζ −3 2
ζ =ζ 1
Therefore
√
K I = σ πa sin2 2α, (5.14a)
√
K I I = σ πa cos2 2α (Ans.). (5.14b)
5.2.1 Boundary Collocation Method

This method has been mentioned in Chapter 3 earlier. In many cases, the stresses
can be expressed in terms of two analytic functions, which can be in the form of
finite series. Particularly when Williams’ eigenfunction expansions are used, each
term satisfies the stress-free conditions at the crack edges (Fig. 5.1). To get the
coefficients of the finite series, it is just necessary to fit or collocate the stress
boundary conditions at a few selected number of stations over the remaining part
of the boundary. This is why the method is known as the collocation method. The
method does not guarantee any convergence with an increase in number of terms
of the finite series. Nevertheless, it has been exploited to get some solutions
(Hartranft and Sih 1973).
5.2.2 Green’s Function Approach

In this method, known solutions for concentrated load on the crack edges are
employed to get the solution for distributed loading on the crack edges.
For a crack loaded as shown in Fig. 5.5, the SIF can be easily determined using
the solution for the standard case shown in Fig. 5.2 as the Green function. The
Green functions [see Eqs. (5.9) and (5.10)] for the two crack-tips corresponding to
Mode I (load P) and Mode II (Q) are as follows.
r r
P a−b P a+b
KI A = √ , K IB = √
πa a + b πa a − b
r r
Q a−b Q a+b
KI I A = √ , K I IB = √
πa a + b πa a − b
i i
i i
i i

i i
Figure 5.5 Uniform pressure loading on crack edges.
For the case (Fig. 5.5), which is symmetric, the SIFs at the two tips are the same and
is given by
√
Z a r
1 a−x
KI = √ p dx = p πa
−a πa a + x
The above integral can be easily integrated by making a substitution x = a sin2 θ.

In case the loading is specified as outer boundary loading rather than explicit
crack edge loading, it is possible to obtain first the loading on the crack-line in the
corresponding crack-free configuration. The required SIF is then given by
Z a r
1 a−x
KI A = − √ f ( x ) dx (5.15)
−a πa a + x
Z a r
1 a−x
KI I A = − √ g( x ) dx (5.16)
−a πa a + x
where f ( x ) is the crack-line normal stress distribution, and g( x ) is the crack-line

shear stress distribution.
5.2.3 Method of Superposition

If we have to find the SIF for the case (Fig. 5.6(b)), we can consider a plate
(Fig. 5.6(a)) without any crack. In this case, the SIF at A or B is zero. The two cases
(Figs. 5.6(b) and (c)) together are equivalent to the case (Fig. 5.6(a)). This means,
cases (Figs. 5.6(b) and (c)) are complementary. Therefore,
( a) (b) (c) (b) √ (b) √

KI = 0 = K I + K I = K I + p πa, that is, K I = − p πa
In case the loading on crack edges in the case (Fig. 5.6(b)) is in the opening mode,
(b) √
K I = p πa .
i i
i i
i i

i i
Figure 5.6 (a) Crack-free plate. (b) and (c) Two cases of complementary crack loadings.
5.2.4 Weight Function Method

This method makes use of the existence of a common crack edge profile in the case
of a family of crack problems (Wu and Carlsson 1991). The family may consist of
symmetrically located internal crack, or edge crack. A function of this type can be
used to calculate the SIF for an unknown problem by knowing the stresses on the
crack-line in the corresponding crack-free geometry. They were introduced by
Bueckner and were later shown to be independent of loading on the outer
boundary by Rice (1972). They can be derived as follows. The discussion is limited
to Mode I loading only.
Figure 5.7 Edge crack under opening load.
For a plate of unit thickness, as the edge crack extends from size 0 to a (Fig. 5.7),
the work done to extend the crack is given by
i i
i i
i i

i i
Z a
1
W= p ( x, 0) u dx (5.17)
2 0
where p( x, 0) is crack-line normal stress distribution in the corresponding crack-

free body and u is the total crack opening at x from the origin, that is, u = 2v. The
energy release rate G as the crack opens up is given by
Z a
dW 1 dū
G= = p ( x, 0) dx
da 2 0 da
Under external load σ on the boundary parallel to x axis, p( x, 0) = σ in the

corresponding crack-free body. Noting that G = K2I /E,
Z a
E dū
KI = p ( x, 0) dx (5.18)
0 2K I da
For a class of problems (Rice 1972), with the same type of symmetry, the quantity
within the square brackets in Eq. (5.18) is a constant. This is known as weight
function. Hence, the SIF K ∗I for the case of a different geometry but with the same
type of symmetry under Mode I loading can be determined, provided the crack-
line opening stress p∗ ( x, 0) in the corresponding crack-free geometry is known.
That is,
Z a Z a
∗ ∗ E dū
KI = p ( x, 0) dx = p∗ ( x, 0) m( x, a)dx
0 2K I da 0
where m( x, a) is the weight function. An illustrative example is given

subsequently (Fig. 5.8). The opening mode displacement for the case (Fig. 5.8(b))
2σ p 2
is given by v = a − x2 . Therefore,
E
4σ p 2
ū = a − x2
E
E dū 2a 1 √
Hence, m ( x, a) = = √ √ , using K I = σ πa for the case
2K I da πa a2 − x2
(Fig. 5.8(b)).
Since the two cases shown in Fig. 5.8 have similar symmetry, the SIF for the case
(Fig. 5.8(a)) is given by
Z a Z a
2a 1 √
K ∗I = p ( x, 0) m ( x, a) dx = p√ √ dx = p πa (Ans.).
0 0 2
πa a − x 2
i i
i i
i i

i i
Figure 5.8 Internal crack with two different loadings but similar symmetry. (a) Uniform pressure
on crack edges. (b) Uniform tension on plate edges.
Problem 5.5
The weight function for an edge crack in finite plate of width w (Fig. 5.9) is given
by (Parker 1981)
" #
a−x a−x 2

2
m ( x, a) = p 1 + m1 + m2 ,
2π ( a − x ) a a
m1 = A1 + B1 r2 + C1 r6 , m2 = A2 + B2 r2 + C2 r6 ,
Figure 5.9 Edge crack geometry.
i i
i i
i i

i i
a
for 0 ≤ r ≤ 1, where r = .
w
A1 = 0.6147 B1 = 17.1884 C1 = 8.7822
A2 = 0.2502 B2 = 3.2899 C2 = 70.0444
(a) Calculate K I for remote end loading σ and r = 0.45.

(b) Calculate K I for remote bending load M and r = 0.5.
Assume thickness as unity.
Solution
(a) For r = 0.45, m1 = 4.1682, and m2 = 1.4980.
" #
a−x 2
Z a
a−x

2
KI = σp 1 + m1 + m2 dx, since p( x, 0) = σ.
0 2π ( a − x ) a a
Substituting a − x = z, the integral can be easily integrated to give

√
K I = 2.4206σ πa (Ans.).
12 M
(b) For this case, p( x, 0) = ( a − x ), assuming the beam thickness as unity.
w3
For r = 0.5, m1 = 5.0490, and m2 = 2.1671.
" #
a−x 2
Z a
a−x

12M 2
KI = (a − x) p 1 + m1 + m2 dx
0 w3 2π ( a − x ) a a
√ 6M
Upon integration, K I = 1.4879 σmax πa, where σmax = (Ans.).
w2
5.3 Numerical Technique: Finite Element Method

The numerical technique like the finite element method (FEM) has been very
widely applied for the determination of the SIFs. It has been so because of its
applicability to a wide range of problems, irrespective of complexity of loading
and geometry of components.
For finite element analysis of a given domain, it is discretized into a convenient
number of elements of finite dimensions (Zienkiewicz and Taylor 2000; Cook et al.
2002). The elements are interconnected at their nodal points. Field variables (e.g.,
displacements in the case of displacement finite element formulation) are
associated with each of these nodes. A distribution, which may be linear,
quadratic, cubic, and so on, of the variables is assumed within each of the
i i
i i
i i

i i
elements and a functional such as potential energy functional is then derived in

terms of these nodal variables. When the restrictions are imposed to make the
derivative of the functional zero with respect to each of the variables, a set of
simultaneous equations is obtained. These are nothing but the global equilibrium
equations. The solution of this set gives the displacements at the nodal points
corresponding to the specified boundary and loading conditions.
Through gradual refinements of the discretization, convergence to the exact
solution can be guaranteed, provided the assumed displacement field meets the
convergence criteria. In particular, the assumed displacement field should satisfy
the rigid body mode, the constant strain condition, and the compatibility at the
inter-element boundary.
The application of FEM to different areas involve routine steps: idealization of
the geometry, discretization, element stiffness calculation and assembly, insertion
of boundary conditions, solution of simultaneous equations, and calculation of
output data, which may include element stresses, strains, total strain energy, etc.
A large number of element choices, for example, triangular, quadrilateral,
triangle with curved boundaries, quadrilateral with curved boundaries, etc., are
available for the discretization. The choices of elements to a large extent depend
on the particular application. Generally, the geometric configurations, local high
stress gradient, if any, in the domain, accuracy required, and available
computational facilities are some of the issues, which influence the selection. In
the early stages of application of FEM to fracture mechanics, the conventional
elements like the constant strain triangles, linear strain triangles, 4-noded
quadrilaterals, 8-noded quadrilaterals, and their analogues in the three
dimensions, were widely utilized to discretize the body, and special techniques
were adopted to obtain the SIFs. Later, it was realized that for any problem
involving stress singularity, the convergence rate is dominated by the singular
nature of the solution (Tong and Pian 1973). Therefore, the convergence can be
enhanced by employing elements that can approximate the singular field
properly. Now there are a large number of special or singularity elements
available, which can be constructed at the crack-tip to facilitate a faster
convergence rate (Tracey 1971; Byskov 1970; Henshell and Shaw 1975; Barsoum
1976; Tracey and Cook 1977; Stern 1979; Dutta, Maiti and Kakodkar 1990; Maiti,
1992 a, b and c. etc.). These help to eliminate the need for a very fine crack-tip
discretization, and some formulations even permit direct and accurate
determination of SIFs (Rao, Raju and Krishna Murthy 1971; Tong and Pian 1973;
Atluri, Kobayashi and Nakagaki 1975; Atluri 1986).
Consider a problem with eight elements and nine nodes (Fig. 5.10). The
external load is acting at 8 and the plate does not move in the y direction at the
i i
i i
i i

i i
Figure 5.10 (a) Typical FE discretization and (b) u-displacement surface.
nodes 1 and 3. Under the action of the load, all the nodes move in both x and y
directions due to deformations except the ones under constraint. If we plot the
displacement u in the vertical direction, we get a surface 13AB as shown
(Fig. 5.10(b)). This is u-surface. Similarly, we can get v-surface by plotting v
displacements. The portion of the surface above a typical element, say 4, is a
‘triangular’ surface. It can be approximated by a plane. As the size of the element
reduces, the accuracy of approximation as a plane increases. Similarly, the
displacement surface v above element 4 can also be approximated by a plane.
These two surfaces can be written as follows.
u = α1 + α2 x + α3 y, (5.19a)
u = α4 + α5 x + α6 y (5.19b)
where α1 to α6 are known as generalized coordinates. These coordinates can be

eliminated by noting that these surfaces pass through the nodal points. That is,
u i = α1 + α2 x i + α3 y i + α4 0 + α5 0 + α6 0
v i = α1 0 + α2 0 + α3 0 + α4 + α5 x i + α6 y i
u j = α1 + α2 x j + α3 y j + α4 0 + α5 0 + α6 0
v j = α1 0 + α2 0 + α3 0 + α4 + α5 x j + α6 y j
u k = α1 + α2 x k + α3 y k + α4 0 + α5 0 + α6 0
v k = α1 0 + α2 0 + α3 0 + α4 + α5 x k + α6 y k
i i
i i
i i

i i
where i = 3, j = 6, and k = 5 for element n = 4. Alternatively,

    

 ui  1 xi yi 0 0 0 
α1 



 v 
i

 
 0 0 0 1 x i y  
i  α2 





uj 1 x y 0 0 0
    
j j
 α 3
=  0 0 0 1 x j y j  α4 , {u}n = [ A]n {α}n (5.20 a and b)


 vj 

   
uk   1 xk yk 0 0 0  
  
α 
 5

 
 
  
vk 0 0 0 1 xk yk
  
α6
Therefore, {α}n = [ A]− 1

n {u}n
Substituting {α}n in Eq. (5.19)
 
 ui 
 
    vi 
u Ni 0 Nj 0 Nk 0

 
 
u

j
=  = [ N ] {u}n (5.21)
 vj 
v n 0 Ni 0 Nj 0 Nk 
  
u 

 
 k

 
vk

Ni , Nj , and Nk are the shape functions or interpolation functions and are given as
follows.
a i + bi x + c i y a j + bj x + c j y a + bk x + c k y
Ni = , Nj = , Nk = k (5.22)
2∆ 2∆ 2∆
∆ = area of triangle n with vertices i, j, and k or 3, 6, and 5.
ai = x j y k − x k y j bi = y j − y k , ci = x k − x j
a j = x k yi − xi y k b j = y k − yi , c j = xi − x k
a k = xi y j − x j yi bk = y i − y j , c k = x j − xi (5.23)
Thus, it is possible to express an element displacement field in terms of

displacements of its nodes. The strain field within the element n is given by
∂u 1 ∂v 1
εx = = [ bi u i + b j u j + b k u k ] , ε y = = [ ci vi + c j v j + c k v k ]
∂x 2∆ ∂y 2∆
∂u ∂v 1
γxy =+ = [ c i u i + bi v i + c j u j + b j v j + c k u k + b k v k ] ,
∂y ∂x 2∆
   
 εx  b 0 b j 0 bk 0
1  i
εy = 0 ci 0 c j 0 ck {u}n , {ε}n = [ B]n {u}n (5.24 a and b)
2∆
c i bi c j b j c k b k
 
γxy
i i
i i
i i

i i
Since bi , b j , bk , ci , c j , and ck are constants for an element, the strains within the
element remain constant. That is why this element is known as constant stain
triangular (CST) element.
The stresses and strains in two dimensions are related by the following
relationship for an isotropic elastic material.
1 ν 0
    
 σx  E ν 1  εx 
σy = 0  ε for plane stress condition,
y
1 − ν2 1−ν  
 
 
τxy 0 0 γxy
2
1−ν 0
  
ν
E  εx 
=
 ν 1 − ν 0
 ε y for plane strain condition. (5.25)

(1 + ν) (1 − 2ν) 1 − 2ν  

0 0 γxy
2
In short, {σ }n = [ D ]n {ε}n (5.26)
where [ D ]n is material property matrix, and {σ}n is stress matrix. Substituting the
value of strain matrix in terms of element nodal displacements{u}n ,
{σ }n = [ D ]n [ B]n {u}n (5.27)
Since elements of the [ D ]n matrix are all constants, the stresses in the element are all
constants. In general, therefore, for an element of finite dimensions, the field given
by the above relations indicates a constant stress−strain field. The strain energy
stored Un in an element is given by
1 1
Z
Un = {ε}nT {σ}n h dA = {ε}nT {σ}n hAn {u}n
An 2 2
1 1
= {u}nT [ B]nT [ D ]n [ B]n hAn {u}n = {u}nT [k]n {u}n (5.28)
2 2
where h is plate thickness, An is area of element n and [k ]n is its stiffness matrix. Its
dimension is 6×6. The element displacement field is given by three us and vs of
the three corner nodes.
The total strain energy U in the plate is obtained summing up Un over all the
eight elements in the present case. U can be written as follows.
ne
1
U= ∑ Un = {u}T [K ] {u}
2
(5.29)
1
i i
i i
i i

i i
Note that [K] is a matrix of size 18×18 and {u} is a column matrix of size 18, and
ne is the total number of elements. In the present case ne = 8. The work done by
external forces
W = {u}T { P} , (5.30)
where {u} is global displacement vector and { P} is global nodal load vector. The
potential energy π of the discretized system
1
π = U−W = {u}T [K ] {u} − {u}T { P} (5.31)
2
The deformed system is in equilibrium under the action of the external forces.
Hence, the potential energy of the system is minimum. By minimizing the
potential energy with respect to the displacements ui , i = 1, 2, 3 . . . . . . 18, a set of
simultaneous equations is obtained.
[K ] {u} = { P} (5.32)
These are the global nodal equilibrium equations. This set of equations can be
solved for after introduction of the displacement boundary conditions, u1 = v1 =
u3 = v3 = 0, and, thereby, the displacement field of the whole plate is obtained.
If displacement variation within the element is considered to be quadratic or
cubic, the number of nodes per element will have to be increased to six or nine,
respectively. Consequently, the strain field within the element will be linear and
quadratic, respectively. The element stiffness matrix for the element will have to be
obtained through Gauss quadrature or numerical integration.
After obtaining the global displacement vector, the element stresses or Gauss
point stresses can be calculated through Eq. (5.27). Based on these displacements
and stresses, the SIFs can be calculated through the displacement method, stress
method, J integral technique, stiffness derivative procedure, crack closure integral
(CCI) technique, and so on. These are discussed subsequently.
5.3.1 Displacement and Stress-Based Methods for Extraction of SIFs

Displacement method For Mode I loading, the x and y displacements are given
by
r
2r
u = KI ũ (θ ), (5.33a)
π
r
2r
v = KI ṽ (θ ) (5.33b)
π
i i
i i
i i

i i
The exact forms of ũ (θ ) and ṽ (θ ) are given in Chapter 3 [Eqs. (3.39) and (3.43)].
Generally, values of v are more significant than u around the crack-tip in this mode.
By selecting a corner node closer to the crack-tip, hence, small r on the crack face,
or θ = π, the SIF K I can be obtained using the above relation for v. It is better to
avoid the first corner node, since the FE solution is not that accurate around the
crack-tip because of the high stress and strain gradients. It is possible to improve
the accuracy of the results by using the appropriate singularity element like the
quarter point singularity element (Barsoum 1976; Henshell and Shaw 1975) around
the crack-tip.
By selecting a number of nodes on the crack face, a set of K I can be obtained.
Thereby, a variation of K I with r can be plotted (Fig. 5.11). By extrapolating this
variation back to r = 0, K ∗I can be determined (Wilson 1973). This appears to be the
most accurate value for the Mode I SIF. This graphical procedure helps to avoid the
inaccuracy in the SIF due to error in the FE solution close to the crack-tip and the
inadequacy of the first term solution of the eigen function expansion away from
the crack-tip.
The same procedure can be followed for Mode II problem. In this case, it is more
appropriate to use u displacement as the basis for the extraction of the SIF. The
extrapolation technique can be employed to enhance the accuracy of the results
further.
Stress method Stresses around the crack-tip are given by
KI
σij = √ σ̃ij (θ ) (5.34)
2πr
Figure 5.11 Extrapolation method.
i i
i i
i i

i i
The exact form of the functions σ̃ij (θ ) is given in Chapter 2. In Mode I problems,
σy stresses are more dominant than σx .
At the end of FE computations, stresses can be determined at the element Gauss
point locations or the nodal locations. More often, the nodal stresses are obtained
through extrapolation of the Gauss point data. By selecting a Gauss point on a
radial line θ close to, but less than 90◦ , and comparing σy stress, the SIF can be
extracted through the above relation. The accuracy can be improved further by
resorting to the extrapolation technique. Since the accuracy of stresses is less than
displacement in the displacement FE formulation, the results obtained by the stress
method are less accurate than the results obtained by the displacement comparison.
In a Mode II problem, τxy stresses are more significant around the crack line. By
selecting θ close to 0◦ and a small r, the SIF can be extracted. Again, the
extrapolation method can be employed to improve accuracy of the results.
5.3.2 Energy-based Methods for Determination of SIFs

Consider a plate with an edge crack a under Mode I loading (Fig. 5.12). Any crack
growth will therefore occur in-plane or in a self-similar manner. Under the action
of loading, it is possible to calculate the strain energy U1 stored in the plate. It is
possible to calculate the strain energy U2 corresponding to the same external load
and an extended crack of size a + ∆a. The strain energy release rate G is given by
U2 − U1 ∆U
G = lim ≈ for small ∆a. (5.35)
∆a →0 ∆a ∆a
where thickness of the plate is taken as unity. Thus, it is possible to calculate the
strain energy release rate by two finite element runs. By selecting ∆a about 0.1% a
(Maiti 1990) and doing computations in extended precision, it is possible to obtain
good accuracy. The FE approximation errors in the computation of U1 and U2 are
of the same order of magnitude because of small ∆a. This helps to obtain G, in turn,
the SIF, with very good accuracy through this method, which involves (U2 − U1 ).
In a sense, this accuracy is obtained at the cost of two separate FE runs.
The requirements of two separate FE runs can be avoided by resorting to the
variant of the energy method, for example, J integral method, stiffness derivative
procedure, and CCI technique.
J integral method J integral given below can be computed by selecting a contour

joining the two points lying on the opposite crack flanks.
Z
∂ui
J= Wdy − Ti dS (5.36)
S ∂x
i i
i i
i i

i i
Figure 5.12 Shifting of crack-tip to accommodate small crack extension.
Since J gives the potential energy release rate associated with an in-plane crack
extension, it is possible to evaluate the SIF through its computation. It is sometimes
convenient to consider the path passing through the element Gauss points. The
integration can be carried out through numerical integration by splitting the whole
contour S into a number of small segments. The splitting can be done by noting
the locations of Gauss points, marked by * in Fig. 5.13, and assuming the values of
stresses σij , strains ε ij , strain energy density W, and tractions Tj at a Gauss point 1 to
be valid over the sector mn. Similarly, the values at Gauss point 2 can be assumed
to be valid over the span nr. For the shown contour passing through elements 9 to
16, there are 16 segments. Therefore
16 16
∂u ∂ui ∂vi
J=∑ W dy − Ti i dS = ∑ {Wk ∆yk } − T1k + T2k ∆Sk , (5.37)
k =1
∂x k k =1
∂x k ∂x k
where, assuming that all elements around the crack-tip are 8-noded isoparametric
elements,
1 ∂u ∂v
W= (σx ε x + σx ε x + τxy γxy ), Ti = T1 + T2 ,
2 ∂x ∂y
T1 = lσx + mτxy , T2 = lτxy + mσy
i i
i i
i i

i i
∂u ∂N ∂N2 ∂N8
= u1 1 + u2 + . . . . . . . . . · · · + u8 ,
∂x ∂x ∂x ∂x
∂v ∂N ∂N2 ∂N8
= v1 1 + v2 + . . . . . . . . . · · · + v8 (5.38)
∂x ∂x ∂x ∂x
∆yk = yn − ym for a typical Gauss pointk, dSk = r dθ
l, m = direction cosines of the outer normal to the contour S at the Gauss point k
r = radius of the contour S.
Figure 5.13 Crack-tip shift to accommodate small ∆a and contour for J integral.
By using the appropriate relationship between J and K I , the SIF can be determined.
For pure Mode I and Mode II problems, only one-half of the body can be analysed
and J can be obtained by integration over elements, for example, 9 to 12 or 13 to 16,
and by multiplying by a factor 2.
For a mixed mode problem, J gives the total potential energy release rate. There
are schemes whereby the Mode I and Mode II parts of the total energy release rate
can be separated (Kitagawa, Okamura and Ishikawa 1976).
Domain integral based evaluation of J Rather than evaluating J through line

integral, J can be computed through domain integration (Shih, Moran and
i i
i i
i i

i i
Nakamura 1986; Moran and Shih 1987; Nikishkov and Atluri 1987; Raju and
Shivakumar 1990). This conversion is possible using the Green’s theorem/
formula in two dimensions and the divergence theorem of Gauss in three
dimensions (Shih, Moran and Nakamura 1986; Moran and Shih 1987). The
domain integral based evaluation of J (Raju and Shivakumar 1990) through
two-dimensional finite element analysis is discussed subsequently.
When J is calculated along the contour S0 or S1 (Fig. 5.14) by Eq. (5.36), it gives
energy flow through the contour in the x direction. Similarly, the energy flow
through the contour in the y direction too can be calculated through the following
relation. Z
∂u
Jy = −Wdx − Ti i dS (5.39)
S ∂y
The two energy expression can be given in the form of a single relation.
Z Z Z
∂ui ∂ui
Jx k = Wnk − Ti dS = Wnk − σij n j dS = [ Q] dS, say. (5.40)
S ∂xk S ∂xk S
Z Z
Jx k = 1 [ Q] dS − 0 [ Q]dS
S1 So
Z Z Z
= 1 [ Q] dS + 1 [ Q] dS + 1 [ Q] dS
ABC CO OA
Z Z Z
−0 [ Q] dS − 0 [ Q] dS − 0 [ Q] dS (5.41)
FED OF DO
Figure 5.14 Contour for domain integration.
i i
i i
i i

i i
Introducing a function H, which has value 1 along the contour ABC or S1 and 0
along the contour FED or S0 , the above relation can be written in the following
form.
Z Z Z Z
Jx k = [ Q] H dS + [ Q] dS + [ Q] dS − [Q] H dS
ABC CO OA FED
Z Z Z Z
=− [ Q] H dS − [ Q] H dS + [ Q] dS + [ Q] dS
CBA AF CO OA
Z Z Z Z
− [ Q] H dS − [ Q] H dS + [ Q] H dS + [ Q] H dS
FED DC AF DC
Z Z Z
=− [ Q] H dS + [ Q] dS + [ Q] dS
FEDCBAF CO OA
Z Z
+ [ Q] H dS + [ Q] H dS (5.42)
AF DC
Z
∂ui
Jx k = − [ Q] H dS + ( Jxk )line , Q = Wnk − σij n j (5.43)
FEDCBAF ∂xk
Considering only the first part,

Z Z
∂ui
− [ Q] H dS = − W H nk − σij n j H dS
FEDCBAF FEDCBAF ∂xk
Z
∂ ∂ ∂ui
=− (W H ) − H σij dA
A ∂xk ∂x j ∂xk
Z Z
∂H ∂H ∂ui ∂W ∂ ∂ui
=− W − σij dA − H − H σij dA
A ∂xk ∂x j ∂xk A ∂xk ∂x j ∂xk
Z Z
∂H ∂H ∂ui ∂W ∂ ∂ui
=− W − σij dA − H − H σij dA
A ∂xk ∂x j ∂xk A ∂xk ∂x j ∂xk
Z Z
∂H ∂H ∂ui ∂W ∂ε ij
=− W − σij dA − H − H σij dA (5.44)
A ∂xk ∂x j ∂xk A ∂xk ∂xk
∂σij
since, in the absence of body forces the equilibrium equation is given by =0,
∂x j

1 ∂ui ∂u j
and ε ij = + . Finally, the first part of Jxk is as follows.
2 ∂x j ∂xi
i i
i i
i i

i i
Z Z
∂H ∂H ∂ui ∂W ∂ε ij
Jx k =− W − σij dA − H − H σij dA
Part I A ∂xk ∂x j ∂xk A ∂xk ∂xk
(5.45)
In matrix form
     
  W ∂u1 ∂u2 
∂H   ∂H 
 Jx1 

Z 
 ∂x   ∂x1 ∂x1  σ11 σ12  ∂x1 
  
1
=− 
 −      dA
J  A  ∂H   ∂u1 ∂u2   ∂H 
x2 Part I W
 
 σ21 σ22  

∂x2 ∂x2 ∂x2 ∂x2
   
 ∂ε 11 ∂γ12 ∂ε 22 
 H ∂W

 σ11

Z   ∂x 
  ∂x1 ∂x1 ∂x1   

1
− − H  dA (5.46)
   
  σ12
A 
 H ∂W
  ∂ε ∂γ ∂ε 22   

 
 11 12 
σ22
∂x2 ∂x2 ∂x2 ∂x2
1
where W = [σ11 ε 11 + σ22 ε 22 + σ12 γ12 ] for linear elastic materials. For non-linear
2 R
elastic material W = σij dε ij . Both the area/domain integrals and the line
integrals involved in Jx1 and Jx2 can be evaluated by numerical integrations. In the
evaluation of the domain integrals, the order in integration can be kept the same
as used in the evaluation of the element stiffness matrices.
The first part of expression for integral Jxk (Eq. (5.42)) consists of the domain
integral and the second part includes the line integrals. Under pure Mode I and
Mode II loading and load-free crack edges, lines integrals involved in Jx1 and Jx2
are 0. In the case of a general loading, this is not so. For a general mixed mode
loading, Mode I and II energy release rates are given by
p
Jx1 = J I + J I I , Jx2 = − 2 JI JI I (5.47)
1 hq q i2 1 hq q i2
JI = Jx1 − Jx2 + Jx1 + Jx2 , J I I = Jx1 − Jx2 − Jx1 + Jx2 (5.48)
4 4
For evaluation of the pure Mode I and Mode II parts of J integral, it is better to
decompose displacements and stresses into symmetric and anti-symmetric parts
and compute only the domain integrals. The decomposition helps to make the line
integrals 0.
Considering two points P and P’ that are close to the crack-tip and are
symmetrically placed about the crack line (Fig. 5.15), the following relations can
be written.

u1P u1S u1AS u1P0 u1S −u1AS
= + , = + (5.49)
u2P u2S u2AS u2P0 −u2S u2AS
i i
i i
i i

i i
Figure 5.15 Symmetric and anti-symmetric displacements and stresses about crack plane.
These relations give the symmetric and anti-symmetric displacements of P and P’

as given below.

u1 1 u1P + u1P 0 u1 1 u1P − u1P 0
= , = (5.50)
u2 S 2 u2P − u2P 0 u2 AS 2 u2P + u2P 0
Similarly, the stresses too can be expressed as follows.

       
σ11 σ11P + σ11P 0  σ11 σ11P − σ11P 0 
1  1 
σ = σ22P + σ22P 0 , σ22 = σ22P − σ22P 0 (5.51)
 22 2 2
σ12P − σ12P 0 σ12P + σ12P 0
   
σ12 S σ12 AS
The symmetric and anti-symmetric displacements and stresses can be used in

Eq. (5.45) or (5.46) to evaluate the four integrals JSx1 , JSx2 , J ASx1 , and J ASx2 . It gives
i i
i i
i i

i i
rise to JSx2 = 0 and J ASx2 = 0 because of the symmetric and anti-symmetric nature
of the stress-displacement fields, respectively. Finally,
J I = JSx1 , J I I = J ASx1 (5.52)
For the evaluation of the domain integral, a typical zone consisting of elements 1
to 8 or 9 to 16 (Fig. 5.16) can be considered. The function H can be defined to be
unity at the inner boundary S1 of the zone and 0 at the outer boundary S0 and can
be considered to have a linear variation along ξ direction for a constant η. While
dealing with integration over each element, the Gauss quadrature can be utilized.
The derivatives of H are given by
   ∂x ∂x −1  ∂H 
 ∂H  1 2  

 ∂x   
1
  ∂ξ ∂ξ   ∂ξ 
= , x = x and x2 = y. (5.53)
 1
 
  ∂x1 ∂x2    ∂H 
 ∂H 

   
∂x2 ∂η ∂η ∂η
Figure 5.16 Typical arrangement of crack-tip elements for the evaluation of domain integral and
shape of typical element in mapping plane ξ − η .
Assuming that all elements in the discretization are 8-noded sub-parametric

elements, a typical element, for example, 5, involves eight shape functions for the
displacement interpolations and four shape functions associated only with the
four corner nodes (1 to 4) for the geometric interpolations. These geometric shape
i i
i i
i i

i i
1g
functions are Ni = (1 + ξξ i ) (1 + ηηi ), where ξ and η are the two natural
4
coordinates, and (ξ i , ηi ) are the coordinates of the corner nodes. Therefore,
4 4 4 ∂N g 4 ∂N g
∂x ∂x
∑ ∑ =∑ =∑
g g i i
x= Ni xi , y= Ni yi , xi , xi ,
i =1 i =1
∂ξ i =1
∂ξ ∂η i =1
∂η
4 ∂N g4 ∂N g
∂y ∂y
=∑ i
yi , =∑ i
yi (5.54)
∂ξ i =1
∂ξ ∂η i =1
∂η
where ( xi , yi ) are the coordinates of the associated corner nodes. In order to

∂W
compute the derivatives, , k = 1 and 2, it is necessary to express the variation
∂xk
of strain energy density W in an appropriate form. For each subparametric
element a 4-point Gauss quadrature is good for the evaluation of the element
stiffness matrix and all domain integrations. The strain energy density can be
evaluated at the four Gauss points. The variation of W within the element can be
conveniently considered to be given by the interpolation of the values at the four
Gauss points as follows.
4
W= ∑ Ni Wi (5.55)
i =1
where Wi are the values of strain energy density at the four Gauss points
(Fig. 5.16). The same variation, when extrapolated, can also be assumed to be
valid in the remaining regions of the element beyond the Gauss point locations.
Noting that the element Gauss points are off the natural coordinate axes by ± √13 ,
the interpolation functions Ni are given by

3 1 1 3 1 1
N1 = √ −ξ √ − η , N2 = √ +ξ √ −η ,
4 3 3 4 3 3

3 1 1
N3 = √ +ξ √ +η
4 3 3

3 1 1
N4 = √ −ξ √ +η (5.56)
4 3 3
The strain energy derivatives are given as follows.
   ∂x ∂y −1  ∂W 
 ∂W   
4 4

 ∂x   
  ∂ξ ∂ξ   ∂ξ  ∂W ∂Ni ∂W ∂Ni
=

∂x ∂y

∂W 
,
∂ξ
= ∑ ∂ξ
Wi ,
∂η
= ∑ ∂ξ
Wi (5.57)
 ∂W 

i =1 i =1

  
 
  
∂x2 ∂η ∂η ∂η
i i
i i
i i

i i
The derivatives of x and y with respect to the natural coordinates ξ and η are given
by Eq. (5.54). Finally,
       
 Jx1  8 4 Q11 Q21
= − ∑ ∑  +  det [ Jcb ] [ωGP ] 
i i (5.58)
J  n =1 i =1 Q Q
   
x2 Part I 12 22 i n
where
     
∂H  ∂u1 ∂u2 " ∂H 
W
 
Q11 
  #
 
 ∂x1   ∂x1 ∂x1  σ11 σ12  ∂x1 
=  −   (5.59)

Q12
 
W
 ∂H 


 ∂u1 ∂u2  σ21 σ22 

 ∂H 



∂x2 ∂x2 ∂x2 ∂x2
    
∂W ∂ε 11 ∂γ12 ∂ε 22  
H
 
Q21 
 
 ∂x1 ∂x1 ∂x1  σ11 
 
  ∂x1 
= −H
 ∂ε ∂γ ∂ε  σ12 
  (5.60)
 H ∂W


Q22
 
 

 11 12 22 σ22
∂x2 ∂x2 ∂x2 ∂x2
 ∂x ∂y 
 ∂ξ ∂ξ 
det [ Jcb ] = Determinant of Jacobian matrix = det  
 ∂x ∂y  (5.61)
∂η ∂η
and ωGP is the weightage to be associated with Gauss point i. For a 2 × 2 integration
scheme, the weightage is unity at all the four integration points.
Stiffness derivative procedure This method was proposed by Parks (1974).

Considering a crack of size a and loaded as shown in Fig. 5.13, the potential
energy of the system is given by
1 T
π= u K u − uT P (5.62)
2
where u stands for global displacement vector, K is the global stiffness matrix, and
P is the global load vector. Since
dπ
G=−
da
assuming a situation with constant load P and noting that Ku = P,
i i
i i
i i

i i
du T du T

1 dK du
G= P− Ku + u T u + uT K
da 2 da da da
du T 1 dK 1 dK
P=− [Ku − P] − uT u = − uT u
da 2 da 2 da
1 K2 − K1
= − uT u (5.63)
2 ∆a
where K1 and K2 are the global stiffness matrices corresponding to crack lengths a
and a + ∆a, respectively. The crack length a + ∆a can be very easily accommodated
by shifting the crack-tip as shown (Fig. 5.13). ∆a is recommended as 0.1% a of the
initial crack size a (Maiti 1990). It is also recommended that the computation of the
product be done in double precision.
It has been shown (Maiti 1990) that it is not necessary to calculate the global
stiffness matrices corresponding to the two crack sizes. It is possible to calculate G
from the total strain energies of the eight elements 1 to 8 surrounding the crack-tip.
That is,
8 8
∑ (Ui )a − ∑ (Ui )a+∆a
G= 1 1
, (5.64)
∆a
1 T
(U i ) a = ( u ) ( k i ) a ( ui ) a ,
2 i a
1 T
(U i ) a+∆a = (u ) (k i ) a+∆a (ui ) a
2 i a
where ui is displacement matrix and k i is the stiffness matrix of the element i. Note
that k i has to be evaluated separately for the two crack sizes. Since the displacement
is not going to change appreciably due to the extended crack length in the presence
of the same external load, the displacements corresponding to the extended crack
length a + ∆a can be approximated by those for the original crack size a.
Crack closure integral technique Consider a simple discretization with 16

elements and crack size a (Fig. 5.17). The external load is P. The crack extends
in-plane, or in a self-similar manner, to size a + ∆a under constant load. The crack
0
opening is v2y at the original crack-tip position 2. If the crack closure force just
before the onset of crack extension is P2y , according to Irwin, the crack closure
work is given by
1 0
W= P2y v2y (5.65)
2
i i
i i
i i

i i
Figure 5.17 Typical crack-tip discretization.
P2y can be obtained from the nodal forces of the elements 3 and 4 or 1 and 2. For
0
infinitesimally small crack extension ∆a, v2y can be approximated by v1y (Fig. 5.17).
Therefore
1
W= P2y v1y (5.66)
2
Thereby, the crack closure work can be calculated through a single finite element
run. The potential energy release rate is given by
W P2y v1y
G= = (5.67)
B ∆a 2 B ∆a
i i
i i
i i

i i
With a fine discretization around the crack-tip and ensuring that ∆a is also very
small, G can be very accurately obtained through this method. Again by using the
relation between G and K I , the SIF can be computed.
For pure Mode II loading, closure force at node 2 in the x direction and the
corresponding sliding displacement at node 1 must be considered. That is,
P2x v1x
GI I = (5.68)
2 B ∆a
To determine the corresponding SIF, the relationship between G I I and K I I can be
utilized.
When elements around the crack-tip are 8-noded quadrilaterals (Fig. 5.18(a)),
the potential energy release rate can be computed through single finite element
run using the following relationship.
W P3y v1y + P4y v2y

G= = (5.69)
B ∆a 2 B ∆a
As before, closure forces P3y and P4y can be obtained from the element nodal forces
of elements 3 and 4 or 1 and 2 (Fig. 5.18). An equivalent relation for Mode II can
be written replacing the forces and displacements parallel to the crack plane or x
direction.
If quarter point singularity elements (Fig. 5.18(b)) are employed around the
crack-tip (Henshell and Shaw 1975; Barsoum 1976), the crack closure forces (P3y
and P4y ) are to be obtained at the nodes 3 and 4, and the corresponding
displacements are obtained from the opening displacement at node location 1.
Computation of the opening displacement to be associated with P4y is done using
1
the fact that displacement along the crack edges varies as r 2 , where r stands
r for
3
radial distance from the crack-tip. That is, the required displacement v = v1y ,
4
3
which is equal to the opening displacement at r = ∆a.
4
To facilitate CCI calculations, in general, element sizes ahead and behind the
crack-tip are kept the same. The closure work can also be computed by
considering simultaneous opening of the crack over a length more than
one-element span ∆a. This will involve computing the closure forces and
corresponding opening displacements at a number of nodal points.
Practical problems may involve in addition to remote loading, crack edge
pressure loading, for example, fluid pressure, thermal loading, and residual
stresses. The calculation of CCI and SIFs for such situations is given by Maiti
(1992a).
i i
i i
i i

i i
Figure 5.18 (a) Quadratic and (b) quarter point singular elements around crack-tip.
Local smoothing for improving accuracy of CCI-based calculations Generally,

the crack closure forces and opening displacements obtained through finite
element calculations do not give rise to variations as stipulated by the crack-tip
singularity field. It is established (Krishnamurthy et al. 1985; Ramamurthy et al.
1986; Sethuraman and Maiti 1988) that by local smoothing these two quantities
utilizing the output data on nodal forces and displacements, the accuracy of
determining the strain energy release rates can be improved. Ramamurthy et al.
(1986) assumed a variation of the crack closure forces different from the variation
assume by Sethuraman and Maiti (1988). The final relations obtained by
Sethuraman and Maiti are simpler.
The derivations based on Sethuraman and Maiti (1988) are given subsequently.
In keeping with the displacement-based finite element formulations, they
assumed a variation of the crack opening displacements as dictated by the
crack-tip elements. If the crack-tip elements are quadratic, the displacements vary
quadratically; if the crack-tip elements are quarter point singularity elements, the
displacements vary directly with square-root of distance from the crack-tip; and
so on. The closure stresses are assumed to vary linearly and inversely with
square-root of distance from the crack-tip in the two cases respectively.
In the case of the quarter point singularity elements around the crack-tip, the
crack opening displacements is assumed to vary as the square-root of distance | x |
from the crack-tip 3 (Fig. 5.19a).
i i
i i
i i

i i
v ξ 0 = v2 (1 + ξ 0 )

(5.70)
∆a 2
noting that 1 + ξ 0 = ∆a − x, where v2 stands for y displacement of node 2,
4
and ξ 0 is a natural coordinate with values 0 at node 2, −1 at node 3, and 1 at node
1. The closure stresses over span 345 (Fig. 5.19(a)) ahead of the crack-tip 3 are
√
assumed to vary as 1/ x. It can be written in the form
B1
σy (ξ ) = Bo + (5.71)
(1 + ξ )
∆a
where (1 + ξ )2 = x and ξ is a natural coordinate with values 0 at node 4, −1
4
at node 3, and 1 at node 5. The arbitrary constants B0 and B1 are obtained by
equating the work done by the distributed opening stresses σy on y displacements
over the span 345 with the work done by nodal forces on the corresponding nodal
displacements.
Z ∆a
F3y v3 + F4y v4 + F5y v5 = v σy dx
0
∆a
Z 1
B1
= [ N3 v3 + N4 v4 + N5 v5 ] Bo + (1 + ξ ) dξ (5.72)
−1 1+ξ 2
Figure 5.19 Crack-tip element arrangements. (a) Quarter point singularity elements.
(b) Quadratic elements.
i i
i i
i i

i i
√
Note that v varies as x along span 345. F3y , F4y , and F5y are the closure forces in y
directions at nodes 3, 4, and 5, respectively and are given by
   
F3y Z 1 N3 B1

∆a

F = N4 Bo + (1 + ξ ) dξ (5.73)
 4y −1   1+ξ 2
F5y N5
ξ ξ
where N3 = − (1 − ξ ), N4 = 1 − ξ 2 and N5 = (1 + ξ ). Thereby, the two

2 2
arbitrary constants are obtained.
3 6
Bo = ( F4y − 2 F3y ), B1 = F3y (5.74)
2 ∆a ∆a
Finally, strain energy release rate in the opening mode is obtained through
Z ∆a
1
σy (ξ ) vy ξ 0 dx

G I = lim
∆a→0 2 ∆a 0
Z ∆a
1 B1
v2y 1 + ξ 0 dx

= lim Bo +
∆a→0 2∆a 0 1+ξ
√ !
2 √
Z ∆a
1 ∆a B1
= lim Bo + √ v2y √ ∆a − x dx
∆a→0 2∆a 0 2 x ∆a
v2y
= [ F4y + (1.5π − 4) F3y ] (5.75)
∆a
where v2y is the total crack opening at the nodal location 2 behind the crack-tip and
∆a is assumed to be very small. ∆a can be taken as 3−4% of given crack size a.
Similarly, for the shearing mode or Mode II problem,
v2x
GI I = [ F4x + (1.5π − 4) F3x ] (5.76)
∆a
where v2x indicates the total crack sliding, or crack ‘opening’, in x direction at the
nodal location 2 and F3x and F4y are the closure forces acting in x direction at nodes
3 and 4, respectively.
The expressions due to Ramamurthy et al. (1986) are as follows.
1
GI = (C11 F3y + C12 F4y + C13 F5y v2y + (C21 F3y + C22 F4y + C23 F5y )v1y ] (5.77a)
2∆a
1
GI I = [(C11 F3x + C12 F4x + C13 F5x ) v2x + (C21 F3x + C22 F4x + C23 F5x )v1x ] (5.77b)
2∆a
i i
i i
i i

i i
π π π π
where C11 = 33 − 52, C12 = 17 − 21 , C13 = 21 − 32, C21 = 14 − 33 ,
2 4 2 8
π 7 π
C22 = 21 − , C23 = 8 − 21 .
16 2 8
For 8-noded quadrilaterals (Fig. 5.19(b)) y displacement varies quadratically along
321 and opening stresses vary along 345 linearly.
1
v = v2 1 − ξ 2 + v1 ξ (1 + ξ )

(5.78)
2
σy = Bo + B1 ξ (5.79)
Following the steps as given in the case of the quarter point singularity elements,
the following results are obtained.
3
Bo = F4y , (5.80a)
2 ∆a
3 6
B1 = F4y − F3y (5.80b)
2 ∆a ∆a
1 1
GI = ( F3y v1y + F4y v2y ), G I I = ( F3x v1x + F4x v2x ) (5.81)
2 ∆a 2 ∆a
These results are the same as those obtained by Krishnamurthy et al. (1985).
In the case of 4-noded quadrilaterals, displacement has a linear variation along
a crack face, and the closure stress is constant along the crack line ahead of the
crack-tip. This gives
1 1
GI = F3y v1y , G I I = F3x v1x (5.82)
2 ∆a 2 ∆a
where F3i and v1i , i = x or y, stands for closure force at node 3 and total opening at
nodal location 1, respectively.
Problem 5.6
A square plate 80 mm× 80 mm of uniform thickness 1 mm contained a central
crack of size 8 mm. The plate boundary was maintained at 100◦ C and the crack
edges were at 0◦ C. Through FE analysis of the problem with arrangements of
elements (51−54) and nodes (177−191) around crack-tip A as shown (Fig. 5.20),
the following forces and displacements were obtained: F191y = −0.10048 ×10−2 N,
F186y = −0.43123 × 10−2 N, v185 = 0.011003 mm, and v190 = 0.005560 mm.
Calculate the SIF by the displacement method, CCI technique, and CCI technique
i i
i i
i i

i i
with local smoothing. Use modulus of elasticity E = 1 MPa and Poisson’s ratio
ν = 0.3. Assume plane strain condition.
Figure 5.20 (a) Thermo-elastic problem and (b) crack-tip discretization.
Solution
(a) Displacement method
Using v for θ=180◦ and v185
r
K I 2r
0.011003 = (1 − ν )
µ π
Noting that r = 0.2 mm, h = 1 mm, and µ = E/[2(1 + ν)], E = 1 MPa, and ν = 0.3,
K I = 0.01694 Nmm−3/2 (Ans.)
(b) CCI technique r

3
P191y v185 + P186y v185
4
GI =
h∆a
r
3
0.0010048 × 0.011003 + 0.0043123 × 0.011003
4
= N/mm
0.2
= (1 − v2 )K2I /E
This gives K I =0.01692 Nmm−3/2 (Ans.).
i i
i i
i i

i i
(c) CCI technique with local smoothing

Using relation (5.75)
2 v190y
GI = F186y + (1.5π − 4) F191y
∆a
2 × 0.00556
= [0.0043123 + (1.5π − 4)] 0.0010048] = (1 − ν2 )K2I /E
0.2
This gives K I = 0.01752 Nmm−3/2 (Ans.).
5.4 FEM-Based Calculation of G Associated with Kinking of Crack

An angled crack in a plate (Fig. 5.21) under tension does not lead to in-plane or self-
similar extension when loaded to a critical level. To understand such an extension,
it is necessary to calculate the potential/strain energy release rate with such out-
of-plane extension. It is possible to calculate this sort of energy release rate by
applying the methods, which are available for the in-plane extension (Maiti 1990).
Some of the methods are discussed below.
It is possible to calculate the strain energy Ua (Fig. 5.21(a)) of the system with
the given crack a and loading. It is then possible to calculate the strain energy Ua+l
corresponding to an extended crack a + l with an extension l in θ direction. The
required energy release rate Gθ is given by
Ua + l − Ua U − Ua
Gθ = lim = a+l for small l. (5.83)
l →0 l l
Figure 5.21 (a) Angled crack. (b) Discretization around tip and knee.
i i
i i
i i

i i
assuming thickness to be unity. This procedure will require two FE runs. While
computing Ua+l , it is possible to consider a branch length l up to 4% of a.
It is also possible to calculate Gθ by considering J integral procedure, stiffness
derivative method, CCI technique, and the continuation argument. As per the
continuation argument, the energy release rate associated with an out-of-plane
extension in the direction θ is equal to the energy release rate associated with the
in-plane extension of the branch l when the branch length l → 0. This means that
it is necessary to analyse a problem with the given crack plus a branch of small
length l in θ direction and compute the energy release rate associated with
extension of the branch l in its own plane. While analysing the kinked crack
geometry with crack a + l, the discretization must be done, recognizing the fact
that there is a stress singularity at the knee. The order of singularity depends on
the knee angle θ (Williams 1957). This calls for a very refined discretization over
the span l, square-root singularity elements at the crack-tip, and variable order
singularity elements at the knee (Maiti 1992b) to take care of the variation in the
kink angle. Alternatively, it is possible to use multi corner variable order
singularity element (Dutta, Maiti and Kakodkar 1990; Maiti 1992c) between the
crack-tip and the knee, square-root singularity elements at the crack-tip, and
variable order singularity elements at the knee. The calculation of Gθ can be done
by adopting the steps that are given earlier for J integral, or stiffness derivative
procedure or CCI technique for the in-plane extension (Maiti 1990).
5.5 Other Numerical Methods

Other numerical methods requiring computer-based solution include boundary
element method (BEM), extended FEM (XFEM), and meshless/meshfree methods.
In the BEM, the boundary alone is discretized and the nodal degrees of freedoms,
displacements, and tractions, associated with nodes, are interconnected by the
reciprocal theorem of elasticity. These relations can also be obtained by the
Galerkin method of weighted residuals. This step gives rise to a set of
simultaneous equations. After solving these equations, the SIFs can be obtained
through the comparison of displacements and stresses, and energy methods or
their variant like the CCI method. Since in this method, displacements and nodal
tractions are treated as independent nodal variables, the use of special singularity
elements to ensure strain singularity does not automatically guarantee the stress
singularity. This calls for use of separate shape functions for displacements and
tractions (Mukhopadhyay, Maiti and Kakodkar 2000). The methods to extract the
SIFs with good accuracy are given in references (Maiti, Mukhopadhyay and
Kakodkar 1997; Mukhopadhyay, Maiti and Kakodkar 1998).
i i
i i
i i

i i
In the FEM, the accuracy of computed SIFs can be increased by representing the
crack-tip field over some region around the crack-tip in terms of a few of the terms
of the Williams’ eigenfunction expansion. The strain energy for this region can be
calculated in terms of the arbitrary constants associated with this expansion. The
energy for the remaining part of the body can be calculated in terms of the nodal
displacements. Thus, the strain energy for the whole body is obtained. After
obtaining the potential energy of the body, it can be minimized with respect to the
displacements and the arbitrary constants. This will give rise to a set of linear
simultaneous equations. Solving these equations, the SIFs can be obtained
directly. The details of the method can be found in references (Moës, Dolbow and
Belytschko 1999; Stazi et al. 2003; Liu, Xiao and Karihaloo 2004; Xiao and
Karihaloo 2007; Rabczuk 2013). This method is termed as XFEM.
To overcome the problems associated with discretization in the FEM and BEM,
the meshless methods have evolved (Belytschko et al. 1996; Atluri and Zhu 1998;
Rabczuk 2013). The method formulation can be based on different forms of the
classical Galerkin method. The two important forms are Petrov-Galerkin and
element-free Galerkin methods. In this case, too, the SIFs can be extracted directly
by any of the method, for example, the displacement technique, stress method,
domain integral method, CCI method (Muthu et al. 2014), and so on, as given
earlier in the case of FEM.
5.6 Experimental Methods

5.6.1 Strain Gauge Technique
The technique based on the measurement of strain can be utilized to determine
the SIF experimentally. Because of the high stress/strain gradient close to the
elastic crack-tip, reduction in the influence of the singularity term of the Williams’
stress function expansion away from the crack-tip, and finite dimension of the
strain gauge, it is difficult to obtain good accuracy unless some care is exercised in
placing the gauges. Dally and Sanford (1987) started with a crack-tip stress field
representation using truncated Williams’ eigenfunction expansions for the two
functions associated with the Airy stress function and proposed a scheme
whereby the Mode I SIF could be determined with a good accuracy using a single
strain gauge (Fig. 5.22(a)). They ingeniously selected the orientations α and radial
line positioning θS for the gauge to reduce the number of unknown parameters in
the two truncated series. The orientation α, assuming a state of plane stress, is
given by
1− ν
cos 2α = −κ = , (5.84)
1+ ν
i i
i i
i i

i i
Figure 5.22 Strain gauge arrangement for (a) Mode I and (b) mixed mode.
where ν is Poisson’s ratio. The radial line orientation θ S is given by
θS
tan = −cot 2α. (5.85)
2
The measured strain ε m by the strain gauge (Fig. 5.22(a)) is related to the Mode I
SIF by the following relation.

KI θS 1 3θ S 1 3θ S
2 µ εm = √ κ cos − sin θS sin cos 2α + sin θS cos sin 2α , (5.86)
2πr 2 2 2 2 2
ν
where µ is rigidity modulus. Replacing ν by in Eq. (5.84), it is possible to
1−ν
obtain a relationship valid for a state of plane strain. Sarangi, Murthy and
i i
i i
i i

i i
Chakraborty (2010a, 2010b, 2013) recommend that the minimum radial distance r
should be greater than half the thickness of the sheet and less than certain
maximum, depending on the geometric configuration of the specimen. A distance
greater that half the thickness is recommended to exclude the effects of crack-tip
local plasticity. They provide a methodology (Sarangi, Murthy and Chakraborty
2010b) for determining the maximum permissible distance r for the location of
strain gauge without affecting the accuracy of the SIF obtained.
The problem of mixed mode was first examined by Dally and Berger (1986). For
this case, six gauges 1 to 6 (Fig. 5.22(b)) are required. The gauges are placed
symmetrically about the crack line. Angles α and θS are again given by Eqs. (5.84)
and (5.85), respectively. The gauges are not required to be spaced equally over the
span 1 to 3 or 4 to 6. However, radial distances of the gauges 1, 2, and 3 should be
equal to the distances of gauges 4, 5, and 6, respectively. Sarangi, Murthy and
Chakraborty (2012) again recommend that minimum radial distance r should be
greater than half the thickness of the sheet and less than certain maximum. They
provide a methodology for finding the maximum permissible distance for the
furthest strain gauges (3 and 6 in Fig. 5.22(b)) from the crack-tip.
5.6.2 Photoelasticity
Photoelasticity (Dally and Riley 1991) is a well-known method for finding elastic
stress distribution. In this method, model conforming to the given geometry is
made out of birefringent material and tested in circular polariscope to obtain the
fringe pattern. The fringe pattern can be analysed to obtain the required stresses at
a point. In the case of an object with a crack, the stress data around the crack-tip
can be processed to determine the SIF. A typical fringe pattern around the crack-tip
is shown (Fig. 5.23). Each fringe indicates the locus of a point of constant shear
stress, whose magnitude is known. That is,
σ1 − σ2
τmax =
2
1 h
2 2
i1/2
= √ (K I sin θ + 2K I I cos θ ) + (K I I sin θ ) , (5.87)
2 2πr
where σ1 and σ2 are the principal stresses at a point. Angle θ is measured with
respect to the crack axis x. Selecting a particular fringe, it is possible to locate
the points where the fringe intersects the x and y axes. Thereby, say, r1 and r2 ,
respectively are obtained. This gives,
KI I 1
τmax = √ = √ [K2I + K2I I ]1/2 (5.88)
2πr1 2 2πr2
i i
i i
i i

i i
Figure 5.23 Typical isochromatic fringes around crack-tip.
Thus, both the SIFs can be determined in a mixed mode problem.

The fringes get very crowded near the crack-tip because of extremely high
stress gradient due to stress singularity. It is very difficult to identify the
individual fringes. To overcome this problem, far field fringe data in conjunction
with multi parameter approximations of the crack-tip stress field (Sanford and
Dally 1979) or the digital photoelasticity techniques (Ramesh 2000) can be
employed for an accurate determination of the SIFs.
Exercise
5.1 Solve Problem 5.6 assuming all data to correspond to plane stress condition.
[Ans. 0.01542 and 0.01614 Nmm−3/2 ]
5.2 Determine the SIF at the right hand crack-tip at x = a for the loading shown.
√ √
p represents load per unit thickness. [Ans. K I = 0.7086p πa, K I I = P2 πa]
Figure Q.5.2
i i
i i
i i

i i
5.3 Determine the level of safety for the high strength low alloy steel (HSLA)
plate (Fig. Q.5.3) with a crack (2a = 15 mm) at the end of a rivet hole as
√
shown. Given K IC = 47.7 Pa m, 0.2% proof stress σY = 1640 MPa, and plate
thickness = 25 mm. [Ans. 3.275]
Figure Q.5.3
5.4 Determine the SIFs for the case shown. Crack at each end of the hole is of size
2 mm. If you make any assumption to obtain the solution, indicate it.
√ √
[Ans. K I = 0.210σ MPa m, K I I = 0.133σ MPa m; σ in MPa]
Figure Q.5.4
5.5 Determine the SIFs at the two crack-tips for the four cases shown. In case
Fig. Q.5.5c loading p acts over a span a around the centre of crack.
√ √
[Ans. (a) K I , right = 0.7086p πa, K I I , right = 0.4092p πa. (b) K I I ,
√ √ P P√
left = 0.4092p πa, K I I , right = 0.0908p πa, (c) K I = √ + πa,
2B πa 6
2σ0 a π

(d) K I = √ −1 ]
πa 2
i i
i i
i i

i i
Figure Q.5.5(a) Figure Q.5.5(b)
Figure Q.5.5(c) Figure Q.5.5(d)
5.6 Determine the SIFs at the crack-tip for the case shown (Fig. Q.5.6) when the
crack size is AB = 20 mm, σ0 = 200 MPa, and crack inclination with x axis
√
is 45◦ . [Ans. K I = K I I = 25.06 MPa m]
Figure Q.5.6
i i
i i
i i

i i
References
5.1 Aliabadi, M.H., D.P. Rooke and D.J. Cartwright. 1987. ‘An Improved
Boundary Element Formulation for Calculating Stress Intensity Factors:
Application to Aerospace Structures.’ Journal of Strain Analysis 22: 203−07.
5.2 Atluri, S.N., ed. 1986. Computational Methods in the Mechanics of Fracture.
Amsterdam: Elsevier Science Publishers B. V.
5.3 Atluri, S.N., A.S. Kobayashi and M. Nakagaki. 1975. ‘An Assumed Hybrid
Finite Element Model for Linear Fracture Mechanics.’ International Journal of
Fracture 11: 257−71.
5.4 Atluri, S.N. and T. Zhu. 1998. ‘A New Meshless Local Petrov-Galerkin
(MLPG) Approach in Computational Mechanics.’ Computational Mechanics
22(2): 117−27.
5.5 Barsoum, R.S. 1976. ‘On the Use of Isoparametric Finite Elements in Linear
Fracture Mechanics.’ International Journal of Numerical Methods in Engineering
10: 25−37.
5.6 Belytschko, T., Y. Krongauz, D. Organ, M. Flemming and P. Krysl. 1996.
‘Meshless Methods: An Overview of Recent Developments.’ Computer
Methods in Applied Mechanics and Engineering 139: 3−47.
5.7 Bowie, O.L. 1973. ‘Solutions of Crack Problems by Mapping Techniques.’ In
Methods of Analysis and Solutions of Crack Problems, Mechanics of Fracture, Vol. 1,
ed. Sih, G.C., 1−55. Leyden: Noodhoff International Publishing.
5.8 Bueckner, H.F. 1970. ’A novel principle for the computation of stress intensity
factors’; Zeitschrift Angewandte für Mathematik und Mechanik, 50, 9, 529-546.
5.9 Byskov, E. 1970. ‘The Calculation of Stress Intensity Factors Using the Finite
Element Method with Cracked Elements.’ International Journal of Fracture 6:
159−67.
5.10 Cook, R.D., D.S. Malkus, M.E. Plesha and R.J. Witt. 2002. Concepts and
Applications of Finite Element Analysis. New York: John Wiley and Sons.
5.11 Cruse, T.A. 1996. ‘BIE Fracture Mechanics Analysis: 25 Years of
Developments.’ Computational Mechanics 18: 1−11.
5.12 Dally, J.W. and J.R. Berger. 1986. ‘A Strain Gage Method for Determining Ki
and Kii in a Mixed Mode Stress Field’, 603−12. Proceedings of the 1986 SEM
Spring Conference on Experimental Mechanics, New Orleans, LA.
5.13 Dally, J.W. and W.F. Riley. 1991. Experimental Stress Analysis. Singapore:
McGraw Hill.
i i
i i
i i

i i
5.14 Dally, J.W. and R.J. Sanford. 1987. ‘Strain Gage Methods for Measuring the
Opening Mode Stress Intensity Factor.’ Experimental Mechanics 27: 381−88.
5.15 Dutta, B.K., S.K. Maiti and A. Kakodkar. 1990. ‘On the Use of One Point and
Two Points Singularity Elements in the Analysis of Kinked Cracks.’
International Journal of Numerical Methods in Engineering 29: 1487−99.
5.16 Hartranft, R.J. and G.C. Sih. 1973. ‘Alternating Method Applied to Edge
Crack and Surface Crack Problems.’ In Methods of Analysis and Solutions of
Crack Problems, Mechanics of Fracture, Vol. 1, ed. Sih, G.C., 179−238. Leyden:
Noodhoff International Publishing.
5.17 Henshell, R.D. and K.G. Shaw. 1975. ‘Crack-tip Finite Elements are
Unnecessary.’ International Journal of Numerical Methods in Engineering 9:
495−507.
5.18 Kitagawa, H., H. Okamura and H. Ishikawa. 1976. ‘Application of J-integral
to Mixed Mode Crack Problems.’ Transaction of Japan Society of Mechanical
Engineers 760: 46−48.
5.19 Kobayashi, A.S., ed. 1975. Experimental Techniques in Fracture Mechanics, SESA
Monograph No. 2. Iowa: Iowa State University Press.
5.20 Krishnamurthy, T., T.S. Ramamurthy, V. Vijyakuimar and B. Duttaguru. 1985.
’Modified Crack Integral Closure Integral Method for Higher Order Finite
Elements’, 89−900. Proceedings of the International Conference on Finite Elements
in Computational Mechanics, FEICOM, Mumbai, India.
5.21 Liu, X.Y., Q.Z. Xiao and B.L. Karihaloo. 2004. ‘XFEM for Direct Evaluation of
Mixed Mode SIFs in Homogenous and Bi-materials.’ International Journal for
Numerical Methods in Engineering 59: 1103−18.
5.22 Maiti, S.K. 1990. ‘Finite Element Computation of the Strain Energy Release
Rate for Kinking of a Crack.’ International Journal of Fracture 43: 161−74.
5.23 ———. 1992a. ‘Finite Element Computation of Crack Closure Integrals and
Stress Intensity Factors.’ Engineering Fracture Mechanics: 41(3): 339−48.
5.24 ———. 1992b. ‘A Finite Element for Variable Order Singularities Based on
the Displacement Formulation.’ International Journal of Numerical Methods in
Engineering 33: 1955−74.
5.25 ———. 1992c. ‘A Multi corner Variable Order Singularity Triangle to Model
Neighbouring Singularities.’ International Journal of Numerical Methods in
5.26 Maiti, S.K., N.K. Mukhopadhyay and A. Kakodkar. 1997. ‘Boundary
Element Method Based Computation of Stress Intensity Factor by Modified
Crack Closure Integral.’ Computational Mechanics 19: 203−10.
i i
i i
i i

i i
5.27 Maiti, S.K. and R.A. Smith. 1983. ‘Theoretical and Experimental Studies on
the Extension of Cracks Subjected to Concentrated Loading Near their Faces
to Compare the Criteria for Mixed Mode Brittle Fracture.’ Journal of Mechanics
and Physics of Solids 31: 389−403.
5.28 Moës, N., J. Dolbow and T. Belytschko. 1999. ‘A Finite Element for Crack
Growth Without Remeshing.’ International Journal for Numerical Methods in
5.29 Moran, B. and C.F. Shih. 1987. ‘A General Treatment of Crack-tip Contour
Integrals.’ International Journal of Fracture 25: 295−310.
5.30 Mukhopadhyay, N.K., S.K. Maiti and A. Kakodkar. 1998. ‘BEM Based
Evaluation of SIFs Using Modified Crack Closure Integral Technique under
Remote and/or Crack Edge Loading.’ Engineering Fracture Mechanics 61:
655−71.
5.31 ———. 2000. ‘A Review of SIF Evaluation and Modeling of Singularities in
FEM.’ Computational Mechanics 25: 358−75.
5.32 Murakami, Y. (Editor-in-Chief). 1987. Stress Intensity Factors Handbook, Vols. I
and II. Oxford: Pergamon Press.
5.33 Muthu, N., B.G. Falzon, S.K. Maiti and S. Khoddam. 2014. ‘Modified Crack
Closure Integral Technique for Extraction of SIFs in Meshfree Methods.’ Finite
Element Analysis and Design 78: 25−39.
5.34 Nikishkov, G.P. and S.N. Atluri. 1987. ‘Calculation of Fracture Mechanics
Parameters for an Arbitrary 3-Dimensional Crack, by the Equivalent Domain
Integral Method.’ International Journal of Numerical Methods in Engineering 24:
1801−21.
5.35 Parker, A.P. 1981. The Mechanics of Fracture and Fatigue − An Introduction, 60.
London: E. & F.N. Spon Ltd.
5.36 Parks, D.M. 1974. ‘A Stiffness-derivative Finite Technique for Determination
of Crack-tip Stress Intensity Factors.’ International Journal of Fracture 10(4):
487−502.
5.37 Rabczuk, T. 2013. ‘Computational Methods for Fracture in Brittle and Quasi-
Brittle Solids: State-of-the-Art Review and Future Perspectives.’ ISRN Applied
Mathematics, Article ID 849231, http://dx.doi.org/10.1155/2013/849231.
5.38 Raju, I.S. and K.N. Shivakumar. 1990. ‘An Equivalent Domain Integral
Method in the Two-Dimensional Analysis of Mixed Mode Crack Problems.’
Engineering Fracture Mechanics: 37: 707−25.
i i
i i
i i

i i
5.39 Ramamurthy, T.S., T. Krishnamurthy, K. Badri Narayana, V. Vijyakuimar and

B. Duttaguru. 1986. ‘Modified Crack Closure Integral Method with Quarter
Point Elements.’ Mechanics Research Communications 13: 179−86.
5.40 Ramesh, K. 2000. Digital Photoelasticity: Advanced Techniques and Applications,
1st edn. Heidelberg: Springer.
5.41 Rao, A.K., I.S. Raju and A.V. Krishna Murthy. 1971. ‘A Powerful Hybrid
Method in Finite Element Analysis.’ International Journal of Numerical Methods
in Engineering 3: 389−403.
5.42 Rice, J.R. 1972. ‘Some Remarks on Elastic Crack-tip Stress Fields.’
International Journal of Solids and Structures 8(6): 751−58.
5.43 Rooke, D.P. and D.J. Cartwright. 1976. Compendium of Stress Intensity Factors.
Her Majesty’s Stationery Office.
5.44 Rosakis, A.J. and A.T. Zehnder. 1985. ‘On the Method of Caustics: An Exact
Elastic Analysis Based on Geometrical Optics.’ Journal of Elasticity 15: 347−67.
5.45 Sanford, R.J. and J.W. Dally. 1979. ‘A General Method for Determining Mixed
Mode Stress Intensity Factors from Isochromatic Fringe Patterns.’ Engineering
Fracture Mechanics 11: 621−33.
5.46 Sarangi, H., K.S.R.K. Murthy and D. Chakraborty. 2010a. ‘Radial Locations
of Strain Gages for Accurate Measurement of Mode I Stress Intensity Factor.’
Materials and Design 31: 2840−50.
5.47 ———. 2010b. ‘Optimum Strain Gage Location for Evaluating Stress
Intensity Factors in Single and Double Ended Cracked Configurations.’
Engineering Fracture Mechanics 77: 3190−3203.
5.48 ———. 2012. ‘Optimum Strain Gage Locations for Accurate Determination
of the Mixed Mode Stress Intensity Factors.’ Engineering Fracture Mechanics
88: 63−78.
5.49 ———. 2013. ‘Experimental Verification of Optimal Strain Gage Locations
for Accurate Determination of Mode I Stress Intensity Factors.’ Engineering
5.50 Sethuraman, R. and S.K. Maiti. 1988. ‘Finite element based computation of
strain energy release rate by modified crack closure integral.’ Engineering
5.51 Sih, G.C. 1973a. Handbook of Stress Intensity Factors. Pennsylvania: Lehigh
University.
5.52 ———, ed. 1973b. Methods of Analysis and Solution of Crack Problems, Mechanics
of Fracture, Vol. 1. Leyden: Noodhoff International Publishing.
i i
i i
i i

i i
5.53 Shih, C.F., B. Moran and T. Nakamura. 1986. ‘Energy Release Rate along a
Three-dimensional Crack Front in a Thermally Stressed Body.’ International
Journal of Fracture 30: 79−102.
5.54 Sneddon, I.N. 1973. ‘Integral Transform Methods.’ In Methods of Analysis and
Solutions of Crack Problems, Mechanics of Fracture, Vol. 1, ed. Sih, G.C., 315−67.
Leyden: Noodhoff International Publishing.
5.55 Sneddon, I.N. and M. Lowengrub. 1969. Crack Problems in the Classical Theory
of Elasticity. Glassgow: Wiley International.
5.56 Stazi, F.L., E. Budyn, J. Chessa and T. Belytschko. 2003. ‘An Extended Finite
Element Method with Higher-Order Elements for Curved Cracks.’
Computational Mechanics 31: 38−48.
5.57 Stern, M. 1979. ‘Families of Consistent Conforming Elements with Singular
Derivative Fields.’ International Journal of Numerical Methods in Engineering 14:
409−21.
5.58 Tada, H., P.C. Paris and G.R. Irwin. 2000. The Stress Analysis of Cracks
Handbook, 3rd edn. New York: ASME Press.
5.59 Theocaris, P.S. 1981. ‘Elastic Stress Intensity Factors Evaluated by Caustics’.
In Mechanics of Fracture, Vol. VII, ed. Sih, G.C. The Netherlands: Sijthoff and
Noordhoff.
5.60 Theocaris, P.S. and E.E. Gdoutos. 1976. ‘Surface Topography by Caustics.’
Applied Optics 15 (6): 1629−38.
5.61 Tong, P. and T.H.H. Pian. 1973. ‘On the Convergence of Finite Element
Method for Problems with Singularity.’ International Journal of Solids and
Structures 9: 313−21.
5.62 Tracey, D.M. 1971. ‘Finite Elements for Determination of Stress Intensity
Factors.’ Engineering Fracture Mechanics 3: 255−65.
5.63 Tracey, D.M. and T.S. Cook. 1977. ‘Analysis of Power Type of Singularities
Using Finite Elements.’ International Journal of Numerical Methods in
5.64 Williams, M.L. 1957. ‘On Stress Distributions at the Base of a Stationary
Crack.’ Journal of Applied Mechanics, Transactions of ASME 24: 109−14.
5.65 Wilson, W.K. 1973. ‘Finite Element Methods for Elastic Bodies Containing
Cracks.’ In Methods of Analysis and Solutions of Crack Problems, Mechanics of
Fracture, Vol. 1, ed. Sih, G.C., 484−515. Leyden: Noodhoff International
Publishing.
i i
i i
i i

i i
5.66 Wu, X.-R. and A.J. Carlsson. 1991. Weight Functions and Stress Intensity Factor
Solutions. Oxford: Pergamon Press.
5.67 Xiao, Q.Z. and B.L. Karihaloo. 2007. ‘Implementation of Hybrid Crack
Element on a General Finite Element Mesh and in Combination with XFEM.’
Computer Methods in Applied Mechanics and Engineering 196: 1864−73.
5.68 Zienkiewicz, O.C. and R.L. Taylor. 2000. The Finite Element Method, Vols 1 and
2. Oxford, Butterworth–Heinemann.
i i
i i

Determination of Stress Intensity Factors

Uploaded by

Copyright:

Available Formats

Determination of Stress Intensity Factors

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Determination of Stress Intensity Factors

Uploaded by

Copyright:

Available Formats

i i

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 102 — #1

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 103 — #2

Determination of stress intensity factors 103

In this chapter, some important analytical methods, numerical technique based

5.2 Analytical Methods

Figure 5.1 Plate with angled crack and crack-tip coordinates.

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 104 — #3

104 Fracture mechanics

If conformal mapping, z = ω (ζ ), is used to map the given problem to a convenient

where P and Q are specified in N/m.

Figure 5.2 Point loading on crack edges.

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 105 — #4

Determination of stress intensity factors 105

It can be shown that

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 106 — #5

106 Fracture mechanics

Figure 5.3 Crack edge loading.

Determine the SIFs.

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 107 — #6

Determination of stress intensity factors 107

Determine the SIFs.

Figure 5.4 Loading on edges of infinite plate.

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 108 — #7

108 Fracture mechanics

Substituting the values

5.2.1 Boundary Collocation Method

5.2.2 Green’s Function Approach

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 109 — #8

Determination of stress intensity factors 109

Figure 5.5 Uniform pressure loading on crack edges.

The above integral can be easily integrated by making a substitution x = a sin2 θ.

where f ( x ) is the crack-line normal stress distribution, and g( x ) is the crack-line

5.2.3 Method of Superposition

( a) (b) (c) (b) √ (b) √

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 110 — #9

110 Fracture mechanics

5.2.4 Weight Function Method

Figure 5.7 Edge crack under opening load.

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 111 — #10

Determination of stress intensity factors 111

where p( x, 0) is crack-line normal stress distribution in the corresponding crack-

Under external load σ on the boundary parallel to x axis, p( x, 0) = σ in the

where m( x, a) is the weight function. An illustrative example is given

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 112 — #11

112 Fracture mechanics

Figure 5.9 Edge crack geometry.

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 113 — #12

Determination of stress intensity factors 113

A2 = 0.2502 B2 = 3.2899 C2 = 70.0444

(a) Calculate K I for remote end loading σ and r = 0.45.

Assume thickness as unity.

Substituting a − x = z, the integral can be easily integrated to give

5.3 Numerical Technique: Finite Element Method

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 114 — #13

114 Fracture mechanics

elements and a functional such as potential energy functional is then derived in

“Chapter*5˙Edited” — 2015/7/7 — 13:18 — page 115 — #14

Determination of stress intensity factors 115