Вы находитесь на странице: 1из 13

International Journal of Fracture 94: 357369, 1998.

1998 Kluwer Academic Publishers. Printed in the Netherlands.


J-Q characterization of propagating cracks
A. TRDEGRD

, F. NILSSON and S. STLUND


Department of Solid Mechanics, Royal Institute of Technology, S-100 44 Stockholm, Sweden;

e-mail: annika@hallf.kth.se
Received 15 December 1997; accepted in revised form 12 September 1998
Abstract. An investigation is performed to determine to what extent the state at a growing crack tip vicinity can
be characterised by J and Q calculated from FE analyses of successively stationary crack tip positions. FE models
in two dimensions of single edge notch bend and double edge cracked panel specimens with several different crack
lengths are used to cover a range of load and constraint levels. The stress and strain elds are compared between
different specimens keeping J- and Q-values equal. A remeshing technique in the commercial FE-code ABAQUS
is used to enhance the efciency of the analysis. The results show that the J-Q-theory provides reasonably accurate
crack tip characterization also for growing cracks. This leads to the conclusion that FE analyses of successive
stationary cracks rather than full FE propagation analyses are sufcient. The limit of validity for propagation is
similar to the validation limit for the stationary case, although somewhat more restrictive.
Keywords: Fracture, elastic-plastic fracture, constraint, two-parameter characterization, stable crack growth, nite
element method, remeshing.
1. Introduction
In recent years much effort has been spent on the question of how to characterize the state
at stationary crack tips in elastic-plastic materials when the loading is so elevated that a
one-parameter characterization is no longer possible. The characterization may be made for
instance by the J-integral. Some consensus has been reached that, at least approximately,
considerable improvement in the characterization capacity can be obtained by the so called
Q-parameter (cf. ODowd and Shih, 1993), which provides a way to quantify the loss of
constraint at a crack tip, which in turn is caused by increasing plastic ow at the crack tip. The
corresponding characterization problem for a propagating crack has, on the contrary, not been
very much studied. The crack growth considered here is stable and quasi-static.
It is well known from theoretical and numerical analyses that the stress and strain elds
at the growing tip are considerably different from those of a stationary crack tip. Thus it is
for instance well known (cf. Nilsson (1992) and others) that the J-integral becomes path-
dependent for moving cracks and its near tip value approaches zero. The results in (Nilsson,
1992) and other investigations indicate that by taking the integration path sufciently remote
from the crack tip a value is obtained that usually coincides with the so called deformation
type J. This quantity is dened as the J-value that would have been obtained for a stationary
crack tip at the current position of the moving crack tip and subjected to the same remote
loading. In the following the term J should be understood in this sense.
The question of whether J can characterize the state of moving crack tips was also studied
for some cases in (Nilsson, 1992). The conclusion was that J provides a poor characterization
if the load is elevated much above the ASTM E-399 limit for linearly elastic fracture mech-
195143.tex; 27/07/1999; 8:17; p.1
[Pdf] [Corrected] [Disc/Cp] (Kb. 6) INTERPRINT: Shirley [Frac 4485] (frackap:engifam) v.1.1
358 A. Trdegrd et al.
anics. The goal of the present investigation is to study if Q can improve the characterization
of the region near the crack tip if it is understood in a corresponding sense as deformation
J. There is no theoretical result for a moving crack tip indicating that a description similar
to the one for a stationary crack should exist and correspondingly it is unlikely that a eld
corresponding to the Q-value should prevail. However, it is plausible that the Q-value calcu-
lated for a stationary crack tip at the current position of the moving crack tip and subjected
to the same outer loading, should have some connection to the actual state at the moving
tip. If laboratory experiments of propagating cracks are analysed with FEM (Finite Element
Method) or other numerical techniques, this is mostly done with successive stationary cracks
and in the present study the validity of this approach is examined. This approach was for
instance used by Faleskog (1994) when analysing experimental results for a pressure vessel
steel. The relevance of the J-Q-description for growing cracks was conjectured by Faleskog
(1994) and it is the object of the present study to substantiate this conjecture. An investigation
of whether a J-Q description is possible for propagating cracks in nite geometries does not
seem to have been performed before even though Varias and Shih (1993) have investigated
steady state growth under assumed Modied Boundary Layer conditions thus imposing J and
Q. Dodds et al (1995) considered propagating cracks in nite specimens with the object to
study the inuence of crack extension on the tip elds. These authors did however not address
the question whether a J-Q description is possible or not.
To study whether the two parameters J and Q evaluated from an analysis of a station-
ary crack can be used to characterize the state of the corresponding growing crack, some
comparisons are made of the stress and strain elds for different two-dimensional analyses of
propagating cracks. By different is here meant that the analysed cases are different with respect
to geometry and remote loading but exhibit the same values of J and Q as evaluated from
analysis of the corresponding stationary crack. The question whether the crack tip elds will
then coincide is the central issue of the present investigation. The authors wish to emphasise
that the object is not a comparison of the elds of a propagating crack with those of successive
stationary cracks. As remarked above the crack tip elds are not even of the same functional
type for a propagating and a stationary crack tip, respectively, and there is no purpose in
comparing them.
In order to investigate the problem at hand, accurate and efcient numerical schemes are
needed. A conventional FE analysis suffers from some drawbacks when analysing problems
of this kind. A difcult problem is simulating crack propagation in, for instance, an elastic
plastic material. Nodal relaxation is an often used technique. This method tends to lead to
other problems. Focused meshes are of limited value as the propagating crack tip soon grows
out of the region to which the mesh is focused. Hence, great resolution is given to a region no
longer containing large gradients, while the crack tip has propagated into a coarser region of
the mesh which is unable to provide the desired accuracy. One way of handling this problem
is to create an element mesh with a ne resolution over the entire region where the crack is
assumed to propagate. This kind of analysis leads to satisfactory resolution although unneces-
sary computer time is consumed in regions where only small gradients occur. It should also
be noted that this method requires the crack path to be known in advance.
An alternative method to avoid these problems is to reform the mesh by a so called remesh-
ing. This can be done during the phase of the crack loading when the tip is blunted as well as
during consecutive crack propagation. Remeshing is now appearing as a standard feature in
several commercial FE codes which also makes it a viable alternative.
195143.tex; 27/07/1999; 8:17; p.2
J-Q characterization of propagating cracks 359
2. Denition of crack tip quantities and problem formulation
Numerous investigators have discussed the two-dimensional state at a stationary crack tip in
an elastic-plastic material. Following ODowd and Shih (1993) it is assumed that the state
in the vicinity of a stationary crack tip under plane strain conditions can be represented to a
sufcient degree of accuracy by

ij
= (
ij
)
SSY
+
ij
= (
ij
)
SSY
+ Q
0

ij
. (1)
The term (
ij
)
SSY
is the stress eld obtained from a standard two-dimensional plane strain
small scale yielding (SSY) analysis, and thus directly connected to the value of J. The yield
stress is denoted by
0
and Q is the deviation in hydrostatic stress from the SSY-solution,
scaled by
0
. It has to be realized that the adopted form of the crack tip eld is not a math-
ematically exact statement. The non-singular part is in fact not a constant but rather a slowly
varying eld. This requires a predened way of evaluating Q and here this will be taken as
Q =

)
SSY

0
, at = 0, r = 2J/
0
, (2)
where r and denote polar coordinates centered at the crack tip in deformed state. The stress

is subsequently circumferential stress.


The description (1) breaks down when the loading becomes sufciently high. In order to
judge when the description is applicable it has been suggested by for example ODowd and
Shih (1993) that the value of

must not vary too much over a certain distance ahead of


the crack tip. Dene a normalized radius as r = r/(J/
0
). If the mean gradient of

over
1 < r < 5 is

( r = 5)

( r = 1)
4
, (3)
then

< 0.1, (4)


provides a reasonable limit for the applicability of the Q-characterization.
As remarked earlier a characterization of the state of a moving crack tip analogous to (1)
is not known and there is in fact no reason to assume that this is even possible. In order to
investigate the use of analyses of stationary cracks to describe the state of propagating cracks,
stresses, strains and crack opening angle COA are here compared between corresponding FE-
models of propagating cracks. As already mentioned in the introduction, the strategy is to
compare the states at a propagating crack tip between different geometries at the same value
of J and Q dened in the way described above. It should be noted that during the propagation
phase up to the point of comparison the values of Q do not necessarily coincide. In order to
obtain reasonably realistic crack growth histories, a crack growth law of the following form is
assumed
J
J
0
= +

a
d

, (5)
195143.tex; 27/07/1999; 8:17; p.3
360 A. Trdegrd et al.
Figure 1. Geometries considered in the analyses including FE meshes.
where
J
0
= (d
2
0
)/5E. (6)
With the choice of J/J
0
= 1 the load level corresponds approximately to the ASTM E-
399-limit for linear elastic fracture mechanics, d is a characteristic length of the specimen
geometry as dened in Figure 1. When the crack length (and ligament length) is d/2 the
ASTM E-399-limit is exactly fullled at J/J
0
= 1. The Youngs modulus of the material is
denoted E. The dimensionless constants and have been assigned different values in this
study. In all analyses equals 1.25 while is chosen as 62.5 or 353 depending on the target
value of the loading. These values may be compared to the condition suggested as a limit for
the J-characterization by Hutchinson and Paris (1979)
=
(d/2)
J

dJ
da

1
, (7)
where
1
10 is considered sufcient. Strict adherence to this condition is however not
critical in the present context since the inclusion of the Q-parameter is intended to improve
the characterization capacity. It should be noted that the resistance curve usually is affected
by the constraint level, see for instance Hancock et al. (1993) and Faleskog (1995).
Consider now two arbitrary different geometries (1 and 2) for which J
1
(P, a), Q
1
(P, a)
and J
2
(P, a), Q
2
(P, a), respectively, are known either from numerical analyses of successive
195143.tex; 27/07/1999; 8:17; p.4
J-Q characterization of propagating cracks 361
stationary crack tip positions or from references such as ODowd and Shih (1993). P here
denotes the value of the chosen load parameter. In the calculations performed here, the loading
was imposed by prescribing the displacement at the boundary. A suitable pair of target
values for J and Q is chosen so that J
1
= J
2
and Q
1
= Q
2
and from this the crack length at
which the comparison is to be performed can be determined. In general this crack length will
be different for the two geometries. By the assumed propagation law (5), the crack growth
increment a is then also given. What remains to be determined are the loading histories
during the propagation phase of the two geometries and these are obtained from the J(P, a)
relation for each geometry so that the propagation law (5) is satised.
The material assumptions are those of an elastic-plastic material obeying von Mises ow
criterion with an associated ow rule and isotropic power law hardening. The calculations
have been performed under nite strain assumptions and the constitutive law can be written in
the form

ij
=

E
ijkl

E
1 +

s
ij
s
kl
(1+)
E

2
e
3

2
h +
2
3

2
e


kl
, (8)
where

ij
is the Jaumann stress rate and
kl
the rate of deformation tensor. The elastic modulus
tensor components are denoted by E
ijkl
, the stress deviator components by s
ij
, the effective
stress by
e
,
p
e
the effective plastic strain and h = d
f
/d
p
e
is the plastic hardening modulus.
Plastic loading yields = 1 whereas for elastic loading or any unloading = 0. The ow
stress, and thus the hardening is given by the relation

p
e
=

0
E


f
E
for
e
>
0
. (9)
In this study a value of the hardening exponent n = 5 has been assumed, also E/
0
= 500
and = 0.3.
The geometries considered are those shown in Figure 1, that is a single edge notch bend
(SEN(B)) specimen and a double-edge cracked panel (DECP). In addition, comparisons with
a so called modied boundary layer (MBL) model were also made. This is a model where the
outer region remains elastic and very large compared to the size of the plastic zone around the
crack with boundary conditions according to rst and second terms of the series expansion of
stresses for a crack tip in a linearly elastic material
3. Analysis
In the present computations of propagating cracks, the load is rst applied to the structure with
the crack tip being stationary. When J reaches the value for initiation, the crack is advanced
by node release through a small part of the structure. This part of the structure has a mesh
specially designed for propagation as shown in Figure 2b. All FE analyses in this study are
made assuming two-dimensional plane strain conditions and nite strain effects are accounted
for.
As the loading is applied to the FE mesh with the initially stationary crack the tip, blunting
causes the mesh to distort due to large strains. When the FE mesh of the blunted crack is
considered to be too distorted to provide reliable results or when the tip of a propagating
195143.tex; 27/07/1999; 8:17; p.5
362 A. Trdegrd et al.
(a) (b)
Figure 2. Crack tip meshes. (a) Initial mesh for stationary analysis. (b) Mesh after some amount of crack growth
using remeshing.
Figure 3. Remeshing of crack tip, stationary crack. The meshes shown are: (1) Initial mesh. (2) Mesh becomes
distorted due to loading. (3) New mesh with the same outer shape as previous mesh. (4) Mesh deformed due to
further loading.
195143.tex; 27/07/1999; 8:17; p.6
J-Q characterization of propagating cracks 363
Table 1. Summary of comparisons performed.
Comparison Geometry J/(d
0
) (J/J
0
) Q
a
w
1 SEN(B), DCP, MBL 0.001 2.5 0.19 0.02
2 SEN(B), DCP, MBL 0.008 20 0.29 0.30
3 SEN(B), DCP, MBL 0.0358 90 0.88 0.25
crack has grown through the ner part of the mesh, the analysis is temporarily stopped and
then continued with a new mesh, as shown in Figure 3. Stresses and strains are transferred to
the new mesh and the analysis continues until it is appropriate to change to yet another mesh.
The procedure is described in more detail by Trdegrd et al. (1998). The experience from this
study is that when remeshing is performed sufciently often, negative effects such as smearing
the stress eld in regions with strong stress gradients when interpolating the stress eld onto a
new mesh, are negligible. Another possibility to deal with crack propagation analyses would
for instance be the Boundary Element Method, which still has problems with nonlinearities
such as plastic ow and nite strains.
The commercial FE code ABAQUS (1993) was used for the computations. The main
advantage of the method of remeshing is that each step in the analysis starts with a new
undistorted and well focused mesh. The remeshing technique is associated with transfer of
data between the subsequent meshes and some loss of accuracy due to interpolation of data.
The latter is strongly dependent on the similarity of the consecutive meshes and on how often
remeshing is performed. For further details, see (Trdegrd et al., 1998).
In principle, two different crack tip meshes were used. One is for the loading of the sta-
tionary crack, see Figure 2a, and the other is used in the subsequent crack propagation, see
Figure 2b. The length l
0
in Figure 2 is the minimum element length in the current mesh. A
radius of the initial notch was introduced to enhance the numerical treatment. The radius has
to be sufciently small so that any effects from this on the nal results are negligible. The ratio
of CTOD at the time of obtaining results over initial notch radius was in all cases larger than
10. In fact, remeshing has been used during blunting to avoid too severe a mesh distortion, see
Figure 3.
Three different comparisons for the J-Q-characterization were made, ranging from a level
somewhat over the ASTM E-399-limit and up to a highly elevated load level. These load
levels are chosen in order to bracket possible behaviours and is not thought to represent
any particular material. The load levels are given in Table 1 in terms of J/(d
0
) and J/J
0
.
The distance w in Table 1 equals the distance d in the analyses (see Figure 1 for denition
of d).
Comparison 1, that is the comparison of the lowest load level, was made with the two
geometries DECP and SEN(B) with different crack lengths: a/d = 0.55 in DECP and a/d =
0.2 in SEN(B). The load is only slightly above the ASTM E-399-limit. The propagation phase
develops over almost 300 elements.
The comparison with a more elevated load level, comparison 2, was also made with the
two geometries DECP and SEN(B) with crack lengths a/d = 0.9 in DECP and a/d = 0.37 in
SEN(B). The load is here quite far above the ASTM E-399-limit. The propagation takes place
over about 300 elements.
195143.tex; 27/07/1999; 8:17; p.7
364 A. Trdegrd et al.
The comparison with the highest load, comparison 3, was made with a/d = 0.5 for both
DECP and SEN(B). This load is far above the ASTM E-399-limit for linearly elastic fracture
mechanics. The relation to the limit for non linear fracture mechanics according to ASTM
E-813 is here 1.8 times to high for the bend specimen SEN(B). The corresponding value for
the tensile geometry DECP is 14.
The MBL model was subjected to the same loading in terms of J and Q as the nite geo-
metries. The crack tip was also forced to propagate a corresponding distance a. The reason
for performing these analyses was to see whether a large plastic zone in a nite geometry has
any effect on the results. In an MBL analysis the plastic zone is always small compared to the
outer radius.
Originally, the constants and in the propagation law (5) were chosen in accordance
with Nilsson (1992) to have the values 1.25 and 62.5, respectively. But as the load was very
high in comparison 3 this choice would have caused the crack to propagate through the whole
ligament before reaching the desired load level. The slope (i.e. ) of the propagation law
was therefore increased to = 353. This gives a crack propagation a of 25 percent of the
specimen width. The propagation takes place over almost 300 elements.
The dimensionless constants and have, as mentioned above, been assigned different
values in the different comparisons. This should have no great effect on the solution as long
as it is observed that in all the comparisons the crack tip has propagated well through the
area close to the zone affected by the blunting of the original crack tip. If the crack were to
stop at a relatively small distance after initiation, for instance at about one crack tip opening
displacement, the results might have been affected by the severe plastic deformation caused
by the blunting. With these values in equation (7) ranges from 141 to 1.97 in comparison 3
and 25 to 12.5 in comparison 1. The effect of changing was to studied to some extent in the
MBL model.
The analyses were performed following the same procedure of remeshing and node relax-
ation as in the previously discussed performance study Trdegrd et al. (1998). The number
of elements in the analyses of the stationary crack for comparisons 1, 2 and 3 are 2407, 2007
and 1815 elements respectively. In the analysis of the growing crack, the number of elements
were 2969, 2664 and 2286 elements for the comparisons 1, 2 and 3, respectively. The ratio
of specimen width d and minimum element length l
0
was also varied. In the analysis of the
stationary crack this ratio falls in the range of 2 10
3
7 10
4
for comparison 3, 2 and 1 with
the coarsest mesh in comparison 3. The reason why the ratios differ so much in the analysis
of stationary cracks is that the minimum element length is related to the crack tip opening
displacement and thereby to the load level. In the analysis of the growing crack the ratio of
(d/l
0
) is about 1.5 10
4
for comparison 1 and around 1 10
3
and for both comparison 2 and
3. The reason why the ratios are almost the same in the last two comparisons is that the crack
growth increment a is almost the same, and the number of elements over which growth
occurs is also the same. The number of elements used in the MBL analyses was 3904 and
the crack growth proceeded over 90 elements. This is less than in the analyses of the nite
geometries, but the accuracy is still considered to be sufcient.
4. Results
For each of the comparisons the normal stress perpendicular to and along the prospective
crack plane is shown for the case where the crack tip is stationary all the time at the chosen
195143.tex; 27/07/1999; 8:17; p.8
J-Q characterization of propagating cracks 365
(a) (b)
(c)
Figure 4. Comparison 1 (lowest load level) J/J
0
= 2.5 and Q = 0.19. (a) stresses ahead of stationary crack
tip, (b) stresses ahead of propagated crack tip, (c) total strains ahead of crack tip.
comparison position and subjected to the nal loading (Figure 4a, Figure 5a and Figure 6a).
For the cracks that have propagated to this position the same stresses are shown (Figure 4b,
Figure 5b and Figure 6b) together with the corresponding total strains (Figure 4c, Figure 5c
and Figure 6c). In all these gures the results from the nite geometries are shown while
for the MBL analyses results are shown only for the stress at propagation for the lowest and
highest load, respectively.
Considering rst the results for stationary cracks (Figure 4a, Figure 5a and Figure 6a) it is
seen that the J-Q description is adequate for the DECP geometry for all three loading levels
and condition (4) is certainly satised. For the SEN(B) geometry this is also the case for the
two lowest load levels. The condition (4) is for instance satised since here |

/
0
| = 0.04.
A large discrepancy is however evident for the highest load level. It is clearly seen in Figure 6a
that the stress in the SEN(B) specimen does not satisfy the condition (4) for the difference
stress since |

/
0
| = 0.5. What is seen in Figure 6a is actually the global bending stress
of the SEN(B) specimen and thus it is not expected that the stresses of the propagating crack
should agree at this load level. The strains show even less agreement, as is evident in Figure 6c.
The variations Q vs. crack tip radius for the geometries DECP and SEN(B) are described in
greater detail by ODowd and Shih (1993).
195143.tex; 27/07/1999; 8:17; p.9
366 A. Trdegrd et al.
(a) (b)
(c)
Figure 5. Comparison 2 (intermediate load level), J/J
0
= 20 and Q = 0.29. (a) stresses ahead of stationary
crack tip, (b) stresses ahead of propagated crack tip, (c) total strains ahead of crack tip.
For the propagating cracks, the comparisons 1 and 2 at lower loads are encouraging (Fig-
ure 4b and Figure 5b) since the stress agrees well for the SEN(B), the DECP geometry and
the MBL results. For instance, at a normalized radius r of 2 the stresses differ 1.5 percent for
the lowest load. The strains shown in Figure 4c also show good agreement. The difference in
strains that can be seen in Figure 4c at normalized radius r = 1 emanates from the nature
of the nite element mesh used in the propagation analyses and has no physical background.
The reason is a cluster of elements by the end of the rened zone as can be seen in the mesh
Figure 2b.
In comparison 2, the stresses of the propagating cracks (Figure 5b) are not identical but
show good agreement. At a normalized radius r of 2 the stresses are almost identical. The
strains shown in Figure 5c also show good agreement.
In comparison 3, at the highest load level certain deviations are however evident. That the
result for the SEN(B) differs considerably from the other two geometries is not surprising
in view of the substantial disagreement for the stationary crack. More interesting is that the
DECP results also deviate from the MBL results even though the difference is moderate. It is
thus found that for this case the J-Q-description is not satisfactory, since a deviation occurs
even though the results for a stationary crack are almost identical. Thus the J-Q-description
195143.tex; 27/07/1999; 8:17; p.10
J-Q characterization of propagating cracks 367
(a) (b)
(c)
Figure 6. Comparison 3 (highest load level), J/J
0
90 and Q = 0.88. (a) stresses ahead of stationary crack
tip, (b) stresses ahead of propagated crack tip, (c) total strains ahead of crack tip.
Table 2. Crack opening angle for different comparisons.
COA/[rad] DECP SEN(B) MBL
Comparison 1 0.0326 0.0314 0.0519
Comparison 2 0.0274 0.0300 0.0328
Comparison 3 0.0278 0.0200 0.0488
probably has a somewhat smaller region of applicability in the case of a growing crack than
for the case of stationary cracks.
In summary it appears as the agreement between results for the two geometries and the
MBL results is good as long as the yielding is not too high.
The crack opening angle (COA) has also been evaluated and a comparison is shown in
Table 2 below. The crack opening angle varies slightly with the distance from the crack
tip. The angle COA is here dened as the angle between half crack opening measured at
normalized radius r = 2 and the distance from the crack tip, i.e. COA = arctan(u/r) where u
denotes the crack surface displacement. As can be seen in Table 2 there is no tendency for a
higher value of COA of the SEN(B) specimen compared to the DECP specimen or vice versa.
195143.tex; 27/07/1999; 8:17; p.11
368 A. Trdegrd et al.
Figure 7. Stresses ahead of the crack tip, MBL for different values of Q and .
But it is clear that the difference between the values within each comparison increases with
increasing load level. In comparison 1 the value for SEN(B) is 96 percent of the value from
the DECP specimen. In comparison 2 the value for DECP is 91 percent of the value from the
SEN(B) specimen. In the comparison 3, the highest load level, the value for SEN(B) is only
72 percent of the value from the DECP specimen.
For reasons mentioned above, the value of the parameter , i.e. the slope of the resistance
curve was set much higher in comparison 3 than in comparison 1 and 2. To investigate this,
the effect of is also examined within the MBL model whereby possible interaction effects
between the plastic zone and the boundaries are eliminated. The effect of variation in for
the different MBL computations is shown in Figure 7 both for the case when Q = 0 and
when Q = 0.88. As can be seen from Figure 7, change of has a moderate effect so
that a decrease of tends to lower the stress. The effect tends to decrease with increasing
crack growth increment as expected. The solution ought to approach a steady state if the crack
growth is driven sufciently far away from the initiation site. Varias and Shih (1993) have
presented results for the case of steady state growth. They nd that the stress for the growing
crack is substantially higher than the SSY-solution for a stationary crack. In our case the
difference in stresses from stationary crack to steady state growth is not very signicant and
thus our stresses deviate from the results of Varias and Shih (1993). At r = 2, for instance, the
stress

is 6 percent lower than the stresses presented by Varias and Shih (1993). The same
type of difference is also evident in the results of Dodds et al. (1995), who in their analysis
obtain lower stresses than in (Varias and Shih, 1993). These authors did, however, consider a
nite geometry so a direct comparison with our MBL results is not wholly relevant. Possibly
the convergence to the steady state requires a substantially larger amount of transient crack
growth.
195143.tex; 27/07/1999; 8:17; p.12
J-Q characterization of propagating cracks 369
5. Conclusions
The results of the present investigation do not contradict the hypothesis that the J-Q-theory
can be used to characterize the state at a propagating crack, also at load levels far above the
ASTME-399-limit. The limit of validity for propagation is most likely similar to the validation
limit for the stationary case although our results for the highest load indicate that the range of
validity may be somewhat more limited. This points to the conclusion for practical purposes
it is sufcient with FE analyses of successive stationary cracks, not full FE analyses of the
propagating crack, to characterize the state at a propagating crack. Of course an extensive
analysis including many cases is needed to verify the hypothesis more thoroughly but the
presented results are encouraging.
For a bend specimen, the global bending stress at load levels above the limit load becomes
signicant and destroys the Q-characterization as the stresses close to the crack tip differ too
much from the SSY-solution. This effect is not as apparent in a tensile geometry, since it is
much further away from general yield in the examples considered.
Even though the stresses in comparison 3, which do not all full the condition (4), differ
considerably between the specimen DECP and SEN(B) also at propagation, they show coher-
ence at small radii, close to the crack tip. Over which distances the states have to coincide
cannot be answered by this type of analysis but depends on the micro-structural distance
inherent in the real problem under consideration.
The corresponding MBL analysis shows, however, that if the size of the plastic zone be-
comes of the order of the specimen size, it will also have an effect on the stress state. In such a
case J and Q alone cannot be regarded as describing the state at the crack tip. This difference
is somewhat more pronounced for growing cracks than for stationary cracks.
References
ABAQUS (1993). Users and theory manuals, version 5.3. Hibbitt, Karlsson and Sorensen, Inc., (HKS),
Pawtucket, RI, USA.
Dodds, R.H., Tang, M. and Anderson, T.L. (1995). Numerical modelling of ductile tearing effects on cleavage
fracture toughness. ASTM STP 1244 (Edited by M. Kirk and A. Bakker), American Society for Testing and
Materials, West Conshohoken, Pa, 100133.
Faleskog, J. (1995). Effects of local constraint along three-dimensional crack fronts - a numerical and experimental
investigation. Journal of the Mechanics and Physics of Solids 43, 447493.
Faleskog, J. (1994). An experimental and numerical investigation of ductile crack growth characteristics in surface
cracked specimens under combined loading. International Journal of Fracture 68, 99126.
Hancock, J.W., Walter, G.R. and Parks, D.M. (1993). Constraint and toughness parameterized by T . Constraint
Effects in Fracture, ASTM STP 1171 (Edited by E.M. Hackett, K.-H. Schwalbe, and R.H. Dodds). American
Society for Testing and Materials, West Conshohoken, Pa, 2140.
Hutchinson, J.W. and Paris, P.C. (1979). Stability analysis of J-controlled crack growth. ASTMSTP 668 (Edited by
J. D. Landes, J. A. Begley and C. A. Clarke). American Society of Testing and Evaluation, West Conshohoken,
Pa, 3764.
Nilsson, F. (1992). Numerical investigation of J-characterization of growing crack tips. Nuclear Engineering and
Design 133, 457463.
ODowd, N.P. and Shih, C.F. (1993). Two-parameter fracture mechanics: Theory and applications. NUREG/CR-
5958, CDNSWC/SME-CR-16-92, United States Nuclear Regulatory Commission.
Trdegrd, A., Nilsson, F. and stlund, S. (1998). FEM-remeshing technique applied to crack growth problems.
Computer Methods in Applied Mechanics and Engineering 160(12), 115131.
Varias, A.G. and Shih, C.F. (1993). Quasi-static crack advance under a range of constraints steady state elds
based on a characteristic length. Journal of the Mechanics and Physics of Solids 41, 835861.
195143.tex; 27/07/1999; 8:17; p.13

Вам также может понравиться