See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/226344296

Fracture mechanics analysis of a crack in a

residual stress field

Article in International Journal of Fracture · January 2000

DOI: 10.1023/A:1026574400858

CITATIONS READS

83 102

3 authors, including:

Yuebao Lei Noel O'Dowd

EDF Energy Nuclear Generation Ltd. University of Limerick
53 PUBLICATIONS 559 CITATIONS 150 PUBLICATIONS 3,160 CITATIONS

SEE PROFILE SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Noel O'Dowd
letting you access and read them immediately. Retrieved on: 29 August 2016
 International Journal of Fracture 106: 195–216, 2000.
© 2000 Kluwer Academic Publishers. Printed in the Netherlands.

Fracture mechanics analysis of a crack in a residual stress field

Y. LEI, N.P. O’DOWD and G.A. WEBSTER

Department of Mechanical Engineering, Imperial College, London, SW7 2BX, UK

Received 21 May 1999; accepted in revised form 24 April 2000

Abstract. The standard definition of the J integral leads to a path dependent value in the presence of a residual
stress field, and this gives rise to numerical difficulties in numerical modelling of fracture problems when residual
stresses are significant. In this work, a path independent J definition for a crack in a residual stress field is obtained.
A number of crack geometries containing residual stresses have been analysed using the finite element method
and the results demonstrate that the modified J shows good path-independence which is maintained under a
combination of residual stress and mechanical loading. It is also shown that the modified J is equivalent to the
stress intensity factor, K, under small scale yielding conditions and provides the intensity of the near crack tip
stresses under elastic-plastic conditions. The paper also discusses two issues linked to the numerical modelling
of residual stress crack problems-the introduction of a residual stress field into a finite element model and the
introduction of a crack into a residual stress field.

Key words: Crack, fracture mechanics, residual stresses.

1. Introduction

The effect of residual stress on structural integrity has been given special consideration in the
defect assessment of welded structures as residual stresses can have a detrimental influence on
the safety of such structures [1, 2, 3]. The combination of high tensile residual and operating
stresses can promote failure by fracture or may change the susceptibility to failure modes,
e.g., it can promote brittle cleavage failure rather than ductile tearing. When the cracked
body remains elastic, the linear stress intensity factor (SIF), K, may be used, as the material
response is linear, and the total value of K due to residual stress and mechanical load can
be obtained by superposition, e.g. [4]. However, outside the linear elastic fracture mechanics
regime K is no longer applicable and an appropriate elastic-plastic parameter must be used.
Although most current defect assessment procedures, such as R6 [1], BS 7910 [2] and EPRI
[3] adopt the J-integral [5] as the elastic-plastic fracture parameter, for a crack in a residual
stress field or the combination of mechanical and residual stresses, a general path-independent
J definition appears to be lacking. Furthermore, there has been little investigation into the issue
of J dominance under these loading conditions.
This paper addresses the modification of the Rice J-integral [5] to produce a path-indepen-
dent integral when residual stresses are present. The residual stress problem is treated as an
initial strain problem and the J integral proposed for this class of problems, [6], is used. The
situation when residual stress only is present is examined as is the case when mechanical
stresses are applied in conjunction with a residual stress. The interpretation of the J value
defined in this way is also discussed.
196 Y. Lei et al.

Figure 1. Contour integration path, 0, and domain, A, around crack tip.

2. General definition of J-integral

In a single parameter fracture mechanics approach, the value of the J-integral is a measure
of the non-linear elastic energy release rate and the intensity of the crack tip fields under
J-dominant conditions. For a two-dimensional body with a crack directed along the x1 axis
under quasi-static conditions, a general definition of the J-integral is e.g. [7]
Z  
∂uj
J = lim W δ1i − σij ni ds, (1)
0→0 0 ∂x1
where W is the strain energy density, σij and ui are components of stress and displacement in
Cartesian coordinates respectively, 0 is a curve surrounding the crack tip which begins at the
lower face of the crack and ends at the upper one, ni is the unit vector normal to 0 and ds is
the path length along 0 (see Figure 1). For a purely mechanical load, in the absence of a body
force and crack face traction, the integral in Equation (1) is path independent and J can be
estimated from fields remote from the crack tip. This is the result due to Rice [5]. However,
for many cases, in order to determine the true J value, the limiting process must be invoked
and the contour 0 shrunk to the crack tip. This can give rise to numerical difficulties due to the
inaccuracy of estimates of the crack tip fields in the area near the crack tip. In order to avoid
such numerical difficulties, Equation (1) is often modified to eliminate the limiting process.
A path-independent J expression for a thermally stressed body has been developed by a
number of authors (e.g., [6, 7, 8]) and is available in commercial finite element packages,
e.g., [9]. However, it appears there is currently no path-independent J definition available for
the general case when residual stress is present. The modification of Equation (1) into a path
independent integral for the case of residual stress will be addressed in the next section.

3. Path independent J-integral for a crack in a residual stress field

A residual stress field is a self balancing stress distribution in a body. Such stresses can result
from internal strains, due to, for example, non-uniform plastic deformation, non-uniform heat
expansion and local volume changes during phase transformations. If the internal strain field
results in an additional mechanical strain field (without external loading) residual stresses are
set up in the body. Such residual stress problems can then be treated as initial strain problems.
In this work, the initial strain is defined as the strain due to either non-mechanical loadings
such as temperature variations within a body or due to previous plastic deformation in the
body.
 Fracture mechanics analysis of a crack in a residual stress field 197

When initial strains are present, Equation (1) can still be applied. The total strain, ij , is
then written as the sum of mechanical strain, ijm , and initial strain, ij0 , i.e.,

ij = ijm + ij0 (2)

The initial strain, ij0 , remains constant during subsequent deformation and the mechanical
strain, ijm is related to the stress through the material constitutive law. In the absence of body
forces and crack face traction, following [6], a path independent J-integral equation can be
obtained from Equation (1):
Z   Z
∂uj ∂ij0
J = W δ1i − σij ni ds + σij dA (3)
0 ∂x1 A ∂X1
Here W is defined as the mechanical strain energy density,
Z ijm
W = σij dijm . (4)
0

A is the area enclosed by 0 (see Figure 1) and A → 0 as 0 shrinks to the crack tip. Implicit in
this definition for J is that the initial strains ij0 , are bounded at the crack tip.
When evaluating ijm which appears in Equation (4), it is important to separate the initial
strain from the total mechanical strain. For this reason we introduce the concept of an ‘ini-
tial state’. All plastic strains before the initial state must be treated as initial strains in the
subsequent analysis. Elastic strains, at the initial state, however, are recoverable and therefore
contribute to the mechanical strain. The mechanical strain is then defined as
p
ijm = (ije + ij ) − ij initial state ,
p
(5)
p p
where ije and ij are the total elastic and plastic strains, respectively, and ij |initial state is the
plastic strain at the initial state.

4. Determination of initial strains and mechanical strain energy density

As we have discussed, the numerical computation of J by Equations (3) and (4) requires
evaluation of the initial strain. In some cases, the initial strains can be evaluated directly – the
case of thermal loading, for example. More generally, as seen from Equations (2) and (5), the
initial strain can be determined from the total strain, ij and the elastic mechanical strain, ije at
the initial state, i.e.,
ij0 = (ij − ije )initial state . (6)
Here the ‘initial state’ at which the initial strain field is calculated must be chosen appropri-
ately and this will be discussed for specific cases in Section 5.
In a finite element program, the mechanical strain energy density is generally calculated
p
from the total elastic strains, ije and total plastic strains, ij and therefore only the total
strain energy density, W total is available. However, W total is not necessarily equal to W in
Equation (4) since the plastic strain will include plastic strains accumulated before the ‘initial
state’. Therefore W must be adjusted to account for the initial strain energy density, i.e.,

W = W total − W p initial state , (7)
198 Y. Lei et al.

where W p |initial state is the plastic work done before the initial state.
For the case of applying initial strains directly to a cracked body, Equations (6) and (7) still
apply. However, difficulties may be met in separating the plastic strains and plastic work due
to the stress concentration at the crack tip from those due to the initial strain.

5. Finite element procedures

5.1. D OMAIN INTEGRAL METHOD IN FINITE ELEMENT ANALYSIS

The domain integral J method, [10], provides a convenient method for evaluating J within a
finite element framework. The domain integral expression is obtained by introducing a smooth
function, q, which takes the value of unity at the crack tip nodes and zero on the contour 0
and then applying the divergence theorem to give
Z "  #
∂uj ∂q ∂ij0
J = σij − W δ1i + σij q dA. (8)
A ∂x1 ∂xi ∂x1

To demonstrate the finite element implementation of the J expression via Equation (8), the
4-noded isoparametric element will be considered. In the finite element method, the global
coordinates (x1 , x2 ) and displacements (u1 , u2 ) may be written as [11]

X
4 X
4
xi = Nk Xik , ui = Nk Uik , i = 1, 2, (9)
k=1 k=1

where Nk are the bi-linear shape functions, Xik are the nodal coordinates and Uik are the nodal
displacements. The variation of q within an element is written in a similar way,

X
4
q= Nm Qm , (10)
m=1

where Qm are the nodal values for the mth node and the gradient of q is given by

∂q X X ∂Nm ∂ηk
4 2
= Qm , (11)
∂xi m=1 k=1
∂ηk ∂xj

where ηk represents the local coordinates (η1 , η2 ) and ∂ηk /∂xj is the inverse Jacobian matrix
of the transformation in Equation (9).
In a similar way the initial strain in an element may be expressed as

X
4
ij0 = 0
Nm ijm , (12)
m=1

0
where ijm are the nodal values of initial strain for the mth node and the gradient of ij0 is given
by

∂ij0 X
4 X
2
∂Nm ∂ηk
= 0
ijm . (13)
∂xi m=1 k=1
∂ηk ∂xj
 Fracture mechanics analysis of a crack in a residual stress field 199

The domain integral expression of J in Equation (8) for the plane problem is then
("  #  )
XX
ntotal ng
∂uj ∂q ∂ij0 ∂xk
J = σij − W δ1i + σij q det wp , (14)
p=1
∂x1 ∂xi ∂x1 ∂ηk
p

where ntotal represents the number of elements in the area A and ng is the number of gauss
points per element. The quantities within { }p are evaluated at all gauss points in an element
and wp are the respective weights. A similar expression to Equation (14) has been provided
in [10] for the particular case of thermal strains.

5.2. I NTRODUCTION OF RESIDUAL STRESSES INTO A FINITE ELEMENT MODEL

Appropriate methods must be used to introduce a residual stress field into a finite element
model. In this work, we have examined two methods. One is via plastic deformation, requir-
ing knowledge of the history of the prior deformation in the structure (which is not always
available) and another by specifying the ‘initial stress’ which is based only on the current
condition in the structure.

5.2.1. Residual stress due to plastic deformation

The stress-strain relationship of a structural material generally consists of a linear (elastic)
part and a non-linear (plastic) part and the plastic strain is not recovered upon unloading.
When a structure of such a material is loaded beyond its elastic limit, if plastic strains are non-
uniform, a residual stress field remains in the structure after the load is removed. The following
illustrates the introduction of a residual stress field in a finite width beam by mechanical
loading in a finite element analysis.
A beam of width w is shown in Figure 2(a). The stress-strain relationship of the mater-
ial used in the analysis is described by Von Mises flow theory with an isotropic power law
hardening relation of the form,
 σ
= , for σ ≤ σ0
0 σ0
 n (15)
 σ
= , for σ > σ0
0 σ0
where n is the strain hardening exponent, σ0 is yield stress and 0 the yield strain defined as
0 = σ0 /E, where E is the Young’s modulus. In this analysis, n has been taken as 5, the
Poisson’s ratio, ν = 0.3 and E/σ0 = 500 to represent a typical steel. Four noded plane strain
elements have been used and the analysis has been performed using the commercial finite
element package, ABAQUS [9].
The specimen was loaded under four point bend up to a moment of 1.25 times the plane
strain limit moment based on σ0 and the load removed. Due to the non-uniform plastic strains
in the beam a self balancing residual stress, σyy (x) remains in the beam after the load has been
removed. The stress distribution along section y = 0 is shown in Figure 2(b). As discussed
in Section 3, the plastic strain remaining after the load has been released plays the role of the
initial strain for subsequent loadings and can be determined from Equation (6).

5.2.2. Initial stress method

An alternative approach is to introduce a residual stress distribution directly into a finite
element model as an initial condition. Note that the issue of uniqueness arises in this case
200 Y. Lei et al.

Figure 2. Introducing residual stress into a finite width strip by mechanical loading, (a) four point bend loading,
(b) y-direction residual stress distribution along section y = 0.

Figure 3. (a) Difference between initial stress input and residual stress obtained (b) schematic drawing of
proportional integral (PI) adjustment.
 Fracture mechanics analysis of a crack in a residual stress field 201

Figure 4. Comparison of required residual stress distribution with final iteration of PI method.

as a number of different loading histories could result in the same residual stress field. The
issue of uniqueness aside, it has furthermore been found that introducing a residual stress field
into a finite element model is not a straightforward method in practice. When a self balancing
residual stress distribution, such as that shown in Figure 2(b), is input into a region of the
stress-free beam the resultant stress distribution after an equilibrium step (no additional load
applied) differs from the expected one in both profile and amplitude (see Figure 3(a)). This
is because the residual stress field has been input only in a small area of the finite element
mesh. During the equilibrium step, these stresses are smoothed out to satisfy equilibrium
over the whole body giving rise to a discrepancy between the input and output stress fields.
(It should be pointed out that in practice, residual stress distributions are measured only at
selected positions and therefore a full distribution of the stress is often not available.)
A proportional integral (PI) adjustment, which is widely used in automatic control, can be
used to modify the input initial stress values to reproduce a desired residual stress distribution
in the finite element model. The adjustment equation used is


inp = σ (x)inp + β σ (x)targ − σ (x)out ,
σ (x)i+1 i i
(16)

where σ (x)inp is the initial stress distribution input into the finite element model, σ (x)targ is
the target stress distribution and σ (x)out the resultant residual stress distribution due to σ (x)inp
(see Figure 3(b)). The superscript, i, in the equation represents the ith adjustment and β is the
integral factor (taken to be 1 here.) The procedure starts with i = 0 and σ (x)inp = σ (x)targ .
The value of σ (x)out is then obtained from the finite element analysis, σ (x)inp is calculated
using Equation (16) for the next iteration and so on until the agreement between σ (x)out and
σ (x)targ is satisfied.
The PI adjustment method has been used to introduce the residual stress distribution shown
in Figure 2(b). Figure 4 shows the residual stress distribution along section y = 0 after four
iterations – the good agreement between this stress distribution and the target distribution
in Figure 4 is clear. Figure 5 shows residual stress contours of σyy over the whole beam.
Figure 5(a) is the result from the analysis of Section 5.2.1 where the residual stress was
generated by mechanical loading and unloading and Figure 5(b) gives the result for the case
when the initial stress is input in the region indicated. Comparing Figures 5(a) and 5(b), it
202 Y. Lei et al.

Figure 5. Residual stress contours (a) obtained by four point bend loading and unloading and (b) obtained using
the initial stress method.

can be seen that although the residual stress distributions along section y = 0 agree very well
for the two methods, the stress distributions in other areas throughout the specimens can be
totally different. However, as will be seen this difference is not of great significance and the
stress intensity factors corresponding to these two stress distributions are almost identical.
The above method can easily reproduce the known residual stress distribution in a finite
element model. Since residual stresses have been input, this suggests that the problem is an
initial stress problem. In fact the self balanced initial stress field is obtained by applying
initial strains in the finite element method. Figure 6 shows an example of a beam analysed
using ABAQUS [9]. The finite element mesh in the middle section of the specimen is shown
in Figure 6(a). The constitutive behaviour of the material used in the analysis is described
by Equation (15) and the material constants used are given in Section 5.2.1. When a self
balanced residual stress distribution in Figure 6(c) is input into a local area of the beam (the
fine square mesh area in Figure 6(a)), after equilibrium the deformation obtained is as shown
in Figure 6(b). The strain yy at a point corresponding to the deformation field of Figure 6(b)
e p
is not equal to the sum of elastic strain, yy , and plastic strain, yy , at that point. This indicates
 Fracture mechanics analysis of a crack in a residual stress field 203

Figure 6. Initial strain corresponding to an initial stress distribution, (a) Original mesh, (b) Deformed mesh after
the introduce of the initial stress and (c) The initial stress and initial strain distribution along A-A section.
204 Y. Lei et al.

Figure 7. Plastic strain left behind the crack tip during crack introduction in a residual stress field (dark contours
correspond to largest plastic strain).

Figure 8. Method of introducing a crack by changing boundary conditions, (a) inputting residual stress into the
uncracked bosy and (b) changing boundary conditions to form a crack.

that there is an additional strain field, iju , which does not obey the stress-strain relationship of
the material, i.e.,
p
ij = ije + ij + iju . (17)

The residual stress problem is therefore an initial strain problem such that the strain field ij
results in a compatible deformation field and an equilibrium stress field, σij which in general
will, as discussed in the previous paragraph, differ from the applied ‘initial’ stress. Note also,
u
as shown in Figure 6(c) that the residual stress, σyy , and the strain field yy have opposite
sign, i.e., the applied strain is not related to the residual stress field through the stress-strain
p
relationship of the material. In the case presented in Figure 6, ij = 0 so the initial strain,
ij0 = iju as discussed in Section 3.

5.3. I NTRODUCTION OF A CRACK INTO A RESIDUAL STRESS FIELD

Having established a method to set up a residual stress field, the next step is to introduce a
crack into this field.
 Fracture mechanics analysis of a crack in a residual stress field 205

Figure 9. CCP specimen with a residual stress distribution: (a) specimen geometry, (b) uncracked body residual
stress distribution along crack line section.

In a finite element analysis, a crack may be introduced into a residual stress field by releas-
ing nodes sequentially until the desired crack length is reached. This method may be relevant,
for example, when a crack is formed by fatigue or creep. However, when the residual stresses
are large, a plastic wake is left behind the moving crack tip during the crack introduction
process (see Figure 7) which leads to difficulties in separating the plastic strains due to the
final crack tip position from the total strains, when the J calculation method discussed in
previous sections is used. In this work, the crack is therefore introduced by a change of
boundary conditions – after the residual stress is introduced into the uncracked body, the
original boundary conditions are modified to form a crack of the desired location and size (see
Figure 8). In this method, the nodes on the crack faces are released simultaneously and there
is no plastic wake left behind the crack tip. When the residual stress is very large such that the
node releasing process can not be completed in a single step in the analysis, a plastic wake
may be found behind the crack tip. However, this plastic wake is generally in a very small area
near the crack tip. Outside this area, J can be evaluated by Equation (3) without significant
effect on the result.

5.4. S PECIMEN DESCRIPTION

Edge cracked beams of width w in bend and tension and centre cracked plates in tension
(Figure 9a) are analysed. As before, the material is assumed to obey Von Mises flow theory
with isotropic hardening and the uniaxial stress-strain behaviour of the material is described by
Equation (15). A range of hardening exponent values, n, have been examined in the analyses
and these are indicated where necessary.
Two dimensional plane strain finite element analyses have been performed. A typical finite
element mesh used in the analysis is shown in Figure 10. A fine mesh has been used in the
whole area near the symmetry plane, y = 0, to allow for an accurate description of the residual
stress distribution along this section.
Residual stresses were introduced into the specimens by the two methods described in
Section 5.2. The specimen with residual stresses introduced by mechanical loading (four point
bend) will be referred to as specimen A and that with residual stress introduced as an initial
stress will be referred to as specimen B. Both specimens A and B have identical σyy residual
206 Y. Lei et al.

Figure 10. Typical finite element mesh used in the analysis.

stress distribution along the section y = 0 (see Figure 2). In addition, a centre cracked panel
(CCP) specimen with the residual stress distribution shown in Figure 9(b) introduced using
the same method as for specimen B will be analysed. This CCP specimen will be referred to
as specimen C.
The crack has been introduced by changing the boundary conditions, as discussed in Sec-
tion 5.3, after the residual stresses have been introduced. The resulting crack length (a) to
specimen width (w) ratio for all specimens examined is a/w = 0.2. Specimens A and B
have been analysed under superimposed applied mechanical bending and tension loading.
The bending moment applied is in the opposite direction to that used to introduce the residual
stress field in specimen (to give a positive normal stress at the crack face for specimens A and
B). For specimen C, only remote tension has been applied.
As discussed earlier, an ‘initial state’ must be defined in order to determine the initial strain
to be used in the J evaluation. For specimen A, the initial state is chosen as the state after
unloading of the beam and for specimens B and C, it is the state after the initial residual stress
is input and the equilibrium step performed. In the analysis, J values have been evaluated on
forty contours by the procedure described in this work and by the Rice J equation. The results
will hereafter be referred to as ‘Modified J’ and ‘Unmodified J’, respectively.
 Fracture mechanics analysis of a crack in a residual stress field 207

Figure 11. J and K for a crack in a residual stress field.

For the cases analysed, distributions of the near crack tip opening stress along the crack
ligament (y = 0) are presented and this stress is denoted as σyy .

6. Results and discussion

6.1. J- INTEGRAL RESULTS FOR RESIDUAL STRESS FIELD

Results are first presented for the case when no mechanical loading is applied, i.e., the J value
is due only to the residual stress field.
In Figure 11, normalised J values for specimens A and B obtained by the modified J
definition are plotted against domain number. It can be seen from Figure 11 that the J values
show good path-independence except for those calculated on the first several domains near
the crack tip. This result is similar to that observed in the case of mechanical loading alone
as the stress and strain fields in the area very close to the crack tip may not be as accurate as
those in the far field due to the high gradients there. Therefore, the J values evaluated on the
domains some distance away from crack tip are more reliable. This will be further discussed
later in the text.
Comparing the results obtained from specimens A and B in Figure 11, it is found that the J
values calculated in the two cases are almost identical. This is because of the identical residual
stress distribution along the crack line for the two specimens and the small scale yielding
conditions. As small scale yielding conditions hold for this case, K can be calculated using
linear elastic fracture mechanics methods, such as the weight function [12] or superposition
method [4] and converted to J via the relation
K2
J = (1 − ν 2 ). (18)
E
The J value obtained using the weight function method is also included in Figure 11 as a
long dashed line. It is clear that good agreement has been obtained between the modified J
expression and the weight function J. This, therefore, confirms that the physical meaning of
the modified J for a residual stress field is strain energy release rate and equivalent to K under
small scale yielding conditions.
208 Y. Lei et al.

Figure 12. Normalised J versus normalised domain radius, R obtained by modified J definition for specimens A
and B for mechanical load only, (a) four-point-bending, n = 5 (b) remote tension, n = 5, (c) four-point-bending,
n = 10 and (d) remote tension, n = 10.

6.2. J- INTEGRAL RESULTS FOR COMBINED LOADING

Results for the specimens under mechanical loading and combined loading are next presented.
The resultant normalised J values for n = 5 and 10 under four point bend and remote ten-
sion are plotted against normalised domain radius, R, in Figure 12 for mechanical load only
(without residual stress) and Figures 13 and 14 for specimens A and B, respectively. Figure 15
shows the results for specimen C with n = 5. In Figures 12, 13 and 14, the applied moment,
M, and remote tensile load, P, have been normalised by the limit moment, M0 , and limit load,
P0 respectively for a/w = 0.2 under plane strain conditions from [13],
2
w2
M0 = √ 1 + 1.7(a/w) − 2.7(a/w)2 (1 − a/w)2 σ0 (19)
3 4
 Fracture mechanics analysis of a crack in a residual stress field 209

and
2  
P0 = √ 1 − a/w − 1.2(a/w)2 + (a/w)3 + 22(a/w)3 (0.5 − a/w)2 wσ0. (20)
3
In Figure 15, P0 is the limit load for a CCP specimen under plane strain conditions [13],
4
P0 = √ (w − a)σ0 (21)
3
In Figures 12–15 the normalised J values are plotted against normalised distance, R/a,
from the crack tip. To demonstrate path-independence, the values of J obtained from the finite
element analysis could be plotted against domain number as in Figure 11. However, this can
be misleading for the case of a very fine mesh as a large number of domains can be chosen
in a small region around the crack tip and the J values obtained on these domains may only
show local path-independence rather than in the far field.
We first consider the results for mechanical load only. Figure 12 shows the J values cal-
culated by the modified J definition for mechanical load only. From Figure 12, good path
independence has been found except in the near crack tip region R/a < 0.02. This is believed
to be due to inaccuracies in the stress and strain fields in this region as discussed in the previous
section. Although not shown, the unmodified J results are identical to the results presented in
Figure 12. This is to be expected, as seen by examining Equation (3) – the second term in the
right hand side of Equation (3) disappears when residual stresses are absent as ij0 ≡ 0 in this
case and Equation (3) reduces to the Rice J for mechanical loading (note that the strain energy
density, W can still be evaluated from Equation (4)).
The path independence of the modified J for combined residual stress and mechanical
loading is next examined. From Figure 13, the J values obtained from the modified J definition
for R/a > 0.1 show good path-independence but those calculated by the unmodified J defini-
tion are clearly path-dependent. These trends remain unchanged with the increase of external
mechanical load. In the area of R/a < 0.1, the J values obtained by the two definitions are
similar. However, as was seen in Figure 12, in the small area very near the crack tip (about
R/a < 0.05), the J values calculated using either method deviate from these calculated on the
remote domains. Therefore, J should be evaluated on domains some distance away from the
crack tip, say, R/a > 0.05. It appears for this case, that the unmodified (Rice) J expression
provides a good approximation of J in the area of 0.1 > R/a > 0.05. However, the size of
this region may change with the loading level and is not easy to predict in practice. This issue
will be further discussed in Section 7.
In the case of Figure 14 (Specimen B) even in the region remote from the crack tip, the J
values calculated from the Rice definition are not very different from those by the modified J
definition. This is because of the low residual stress level introduced. The results in Figure 15
for specimen C, which has higher residual stresses, show strong path-dependence for the
unmodified J definition. In contrast, the modified J definition provides a path independent
J value for all contours except those very close to the crack tip. The path-dependence of J
evaluated by the unmodified definition can therefore be significant and the use of the Rice J
definition when residual stresses exist [14, 15] may be questionable without considering the
intensity of the residual stress field.
Comparing J values obtained from specimens A and B (see Figures 13 and 14), it is found
that the modified J values obtained from specimen A are lower than those from specimen B
(note that the residual stress distributions and the applied mechanical load are identical for the
210 Y. Lei et al.

Figure 13. Normalised J versus normalised domain radius, R obtained by modified and unmodified J definitions
for specimen A with a combination of residual and mechanical stress, (a) four-point-bending, n = 5, (b) remote
tension, n = 5, (c) four-point-bending n = 10 and (d) remote tension, n = 10.

two specimens.) It is also found that the path dependence of the unmodified J for specimen
A and B are different – it is stronger in specimen A than in specimen B. These issues will be
further discussed later in Section 7.

6.3. C RACK TIP STRESS FIELDS

In single parameter fracture mechanics, the elastic-plastic stress field is controlled by J in the
J dominant zone near the crack tip [16, 17]. If the modified J in the presence of residual stress
and for the combination of residual and mechanical stress is still a crack tip field controlling
parameter, elastic-plastic fracture mechanics concepts can still be applied when residual stress
exists.
Firstly, the near crack tip stress distribution for a residual stress field is considered. The
normalised opening stress, σyy /σ0 , for specimens A and B, corresponding to the J values
in Figure 11 is plotted against normalised distance in Figure 16 together with the normalised
 Fracture mechanics analysis of a crack in a residual stress field 211

Figure 14. Normalised J versus normalised domain radius, R obtained by modified and unmodified J definitions
for specimen B with a combination of residual stress and mechanical stress, (a) four-point-bending, n = 5, (b)
remote tension, n = 5, (c) four-point-bending n = 10 and (d) remote tension, n = 10.

HRR and K-field stress distributions for comparison. From Figure 16, in the zone very close to
the crack tip, the stress distributions for both specimen A and B are similar to the HRR field.
In the zone some distance away from the crack tip, the K-field stress distribution provides
the best representation of the stress field. Note that the stress distribution for specimen A in
Figure 16 is slightly different from that of specimen B. This is because the previous plastic
deformation in Specimen A has increased the yield stress of the material in the near tip area
under the assumption of isotropic hardening. This will be discussed later in Section 7.
Next we examine the near crack tip stress distribution for combined loading. The nor-
malised opening stress is plotted against normalised distance from the crack tip in Figures 17,
18 and 19 for specimen A, B and C, respectively. The stress distribution according to the HRR
field based on the total J values due to residual stress and mechanical loading obtained by
the modified method is plotted in the figures for comparison. From Figures 17, 18 and 19,
the stress distributions for a pure residual stress field and the combination of residual stress
212 Y. Lei et al.

Figure 15. Normalised J versus normalised domain radius, R obtained by modified and unmodified J definitions
for specimen C with a combination of residual and mechanical stress.

Figure 16. Crack tip opening stress distribution for the residual stress field in specimens A and B (r is the distance
from crack tip).

and mechanical stress are very close to the HRR field when the mechanical load is small.
This indicates that under small scale yielding conditions, the modified J is a crack tip stress
field controlling parameter. Figures 17, 18 and 19 also show that with increasing mechanical
load, the stress distributions due to the combination of residual stress and mechanical load
gradually deviate from the HRR field. However, comparing these distributions with the stress
distribution under pure mechanical load at the same load (see Figures 17, 18 and 19), the
combined stress field is close to that for pure mechanical loading. The deviation of the stress
distribution due to combined loading from the HRR field can be explained as due to constraint
loss, see e.g. [18]. It may be seen that in most cases the stress level, i.e., constraint, is increased
slightly, due to the residual stress field, though in Figure 19 the normalised stress fields for
the pure mechanical and combined mechanical and residual stress field are almost identical.
Similar trends have been reported in e.g. [19] and [20] where the effect of thermal stress on
constraint has been examined.
 Fracture mechanics analysis of a crack in a residual stress field 213

Figure 17. Near tip opening stress distribution for specimen A under combined residual and mechanical stress for
n = 5, (a) four-point-bending, (b) remote tension.

Figure 18. Near tip opening stress distribution for specimen B under combined residual and mechanical stress for
n = 5, (a) four-point-bending, (b) remote tension.

7. Discussion

A path independent J-integral expression for a crack in an initial strain induced residual
stress field or under the combination of residual stresses and mechanical stresses has been
proposed. When residual stresses are absent, the J definition is consistent with the standard
Rice J integral. Comparing Equation (3) with the standard J definition, it is seen that there
are two factors that affect the path-dependence of J : (i) in the standard J definition, the
contribution to J from the initial strain gradient, associated with the residual stress field, is
omitted and (ii) the strain energy density at the initial state is not accounted for in the standard
J definition. In the modified J definition (see Equation (3)), the plastic component of strain
energy density at the initial state, W p |initial state is subtracted from the total strain energy density
p
value, however, in the standard J definition, the total W (ije + ij ) is always used. The effect
of (i) increases with increasing stresses and, therefore, increases with applied load while the
214 Y. Lei et al.

Figure 19. Near tip opening stress distribution for specimen C under combined residual and mechanical stress of
n = 5.

effect of (ii) remains constant with load. Therefore, increasing the mechanical load level will
not completely eliminate the path dependence of the Rice J. Of course, the relative effect of the
residual stress field will decrease with increasing load as the total J value increases. Therefore,
at high mechanical load level, when J is evaluated on contours far from the crack tip where
the residual stress field is negligible, the standard J value may be a good approximation to
the modified (path-independent) J value. However, for the cases considered in this paper, such
contours were not found because the residual stress field extends across the specimen section.
It has been found from Figures 13 and 14 that in a small area near the crack tip, 0.1 >
R/a > 0.05, the J value calculated by the unmodified J is almost identical to that from the
modified J . Examining Equation (3), it is found that, firstly, when 0 is taken in an area very
close to the crack tip, the second term of the right hand side of Equation (3) tends to zero
as the area, A, becomes very small. Also, in the near crack tip region, the total strain energy
density, W, is much higher than W p |initial state, i.e.
p
W (ije + ij ) ≈ W (ijm ).

Therefore, J evaluated from the standard J definition is very close to that from Equation (3). It
should be pointed out that the size of this region may change with the residual stress intensity
and may not easily be predicted a priori.
In this paper, two methods have been used to introduce a residual stress field in a finite
element model, by plastic deformation (specimen A) and by the initial stress method (speci-
mens B and C). From Figures 13 and 14 the unmodified J (Rice J) values are path-dependent
for both specimens A and B. However, the degree of path dependence is different – that
of specimen A (Figure 13) is much stronger than that of specimen B (Figure 14). This can
again be explained by the factors which affect the path dependence of the standard J value.
The initial strain gradients in the two specimens are very similar, i.e., the effect of initial
strains on the unmodified J values are nearly identical. However, the plastic component of the
mechanical strain energy density, W p , in specimen A at the initial state, W p |initial state is much
larger than that in specimen B as the former has experienced large plastic deformation during
the introduction of the residual stress field. For specimen B, W p |initial state ≈ 0 and the path
 Fracture mechanics analysis of a crack in a residual stress field 215

Figure 20. Comparison and normalised J values from specimen A and B, (a) four-point-bending, n = 5, (b)
remote tension, n = 10.

dependence of Rice J shown in Figure 14 is mainly because of the initial strains. Of course,
even in specimen B, the degree of path dependence of the unmodified J value depends on the
intensity of the residual field as we have seen in Figure 15.
The modified J values for two of the cases shown in Figures 13 and 14 have been replotted
against normalised load in Figure 20. Examining Figure 20, it is found that the J values
obtained from specimen A and specimen B are nearly identical when the mechanical load
level is low. However, with increasing load, the difference in the J values becomes significant,
the J values obtained from specimen A being less than those from specimen B for a given
load level. The reason for this behaviour is that specimen A has been loaded beyond the
elastic limit and then unloaded to introduce a residual stress field. As isotropic hardening has
been used in the analysis, the yield point of the material which has deformed plastically has
been increased and when the specimen is subsequently reloaded the J values, are therefore
decreased compared with those obtained from specimen B (i.e., less deformation at the same
value of stress). If kinematic hardening is used in the analysis, the J values obtained for the
two specimens are expected to be closer since the stress levels on reloading from a residual
stress field will be similar in specimens A and B if kinematic hardening is assumed.

8. Concluding remarks

A modified J integral has been presented which can be used in the presence of residual
stresses. The integral is path-independent for a pure residual stress field and for a combination
of residual stress field and stress due to mechanical loading. This enables the evaluation of J
on domains away from the high stress gradient area near a crack tip.
The modified J for residual stress is equivalent to the stress intensity factor K under small
scale yielding conditions and provides the intensity of the crack tip stress field for elastic-
plastic conditions analogous to J for pure mechanical load. A similar loss of constraint is
observed with increasing mechanical load as has been observed under mechanical loading
alone.
216 Y. Lei et al.

Acknowledgements

Financial support for this work, provided by the IMC, HSE, EPSRC and DERA is gratefully
acknowledged. The authors would also like to thank Dr. R. A. Ainsworth from BE for his
invaluable assistance with this work.

References

1. Milne, I., Ainsworth, R.A., Dowling, A.R. and Steward, A.T. (1988). Assessment of the intensity of
structures containing defects. Int. J. Pres. Ves. & Piping 32, 3–104.
2. BSI (1999). Guide on methods for assessing the acceptability of flows in structures. Guide B57910-1999.
3. Kumar, V., Schumacher, B.I. and German, M.D.(1984). Development of a procedure for incorporation
secondary stress in the engineering approach, Section 7 in EPRI Report EPRI NP-3607.
4. Anderson, T.L. (1995). Fracture Mechanics Fundamentals and Applications, CRC Press, 2nd edition.
5. Rice, J.R. (1968). A path independent integral and the approximate analysis of strain concentration by
notches and crack, J. Appl. Mech. 35, 379–386.
6. Ainsworth, R.A., Neale, B.K. and Price, R.H. (1978). Fracture behaviour in the presence of thermal strains,
Proc. Int. Conf. on Tolerance of Flaws in Pressurised Components, pp. 171–178.
7. Shih, C.F., Moran, B. and Nakamura, T. (1986). Energy release rate along a three-dimensional crack front in
a thermally stressed body. Int. J. Fracture 30, 79–102.
8. Wilson, W.K. and Yu, I.W. (1979). The use of the J-integral in thermal stress crack problems. Int. J. Fracture
15, 377–387.
9. ABAQUS V. 5.6 (1996). Hibbitt, Karlsson and Sorensen Inc., Providence, RI.
10. Li, F.Z., Shih, C.F. and Needleman, A. (1985). A comparison of methods for calculating energy release rates.
Engng. Fracture Mech. 21, 405–421.
11. Crisfield, M.A (1994). Non-linear Finite Element Analysis of Solids and Structures Vol. 1, Wiley and Sons,
New York.
12. Buecker, H.F. (1971). Weight function for the notched bar, Z. Angewandte Mathemat. Mechan. 51, 97–109.
13. Miller, A.G. (1988). Review of limit loads of structures containing defects. Int. J. Pres. Ves. & Piping 32,
197–327.
14. Qi, D.M. (1992). Recommendations on the treatment of residual stress in PD6493 for the assessment of the
significance of weld defects. Engng. Fracture Mech. 41, 257–270.
15. Finch, D.M. and Burdekin, F.M. (1992). Effects of welding residual stresses on significance of defects in
various types of welded joint. Engng. Fracture Mech. 41, 721–735.
16. Rice, J.R. and Rosengren, G.F. (1968). Plane strain deformation near a crack tip in a power law hardening
material. J. Mech. Phys. Solids 16, 1–12.
17. Hutchinson, J.W. (1968). Singular behavior at the end of a tensile crack in a hardening material. J. Mech.
Phys. Solids, 16, 13–31.
18. O’Dowd, N.P. (1995). Applications of two parameter approaches in elastic-plastic fracture mechanics.
Engng. Frac. Mechanics 52, 445.
19. O’Dowd, N.P. and Sumpter, J.D.G. (1996). Effect of thermomechanical loading on near tip constraint. J. de
Physique IV, Colloque C6, Vol. 6, C6-539-548.
20. Hancock, J.W.(1999). Constraint based Failure Assessment Diagrams for Primary and Secondary Loading,
ASME Conference on Pressure Vessels and Piping, August 1-5 Boston.