International Journal of Fatigue 62 (2014) 159–170

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

3D effects around notch and crack tips

Rafael Cesar de Oliveira Góes a,1, Jaime Tupiassú Pinho de Castro a,⇑, Luiz Fernando Martha b,2
a
Department of Mechanical Engineering, Pontifical Catholic University of Rio de Janeiro, PUC-Rio, Rua Marquês de S. Vicente 225, Rio de Janeiro 22451-900, Brazil
b
Department of Civil Engineering, Pontifical Catholic University of Rio de Janeiro, PUC-Rio, Rua Marquês de S. Vicente 225, Rio de Janeiro 22451-900, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Notches and cracks are usually treated as two-dimensional problems in most structural design and anal-
Received 11 January 2013 ysis applications, employing 2D limit solutions from plane elasticity theories to evaluate highly localized
Received in revised form 28 September 2013 stress/strain concentration effects around their tips. However, such effects are also associated to high
Accepted 10 October 2013
stress gradients, which cause three-dimensional stress/strain fields around those tips that can severely
Available online 23 October 2013
restrict local Poisson-induced transversal strains. Modeling of fatigue crack initiation and propagation,
estimation of plastic zone sizes and shapes, and localized constraint effects are typical problems affected
Keywords:
by such 3D effects, which may lead to non-conservative damage and life predictions if neglected. To
3D notch and crack tip fields
3D stress concentration effects
quantify how important they can be in practical applications, first traditional finite element techniques
Fatigue crack front curvature are used to simulate effects of component thickness and notch tip radius in the stress and strain fields
Stress gradient effects along and around such tips, and to evaluate their importance from the structural design point of view.
Then, versatile sub-modeling techniques are used to study similar effects along the fronts of short and
long cracks. Finally, a stepwise remeshing routine is used to show how an initially straight crack must
slightly curve its front during its propagation by fatigue, due to the unavoidable 3D effects that always
surround real crack tips.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction and many other numerical and experimental SCF have been ob-
tained (see e.g. Peterson et al. [3] and Savin [4]) for countless notch
It is a truism to affirm that the precise calculation of stress and geometries, but most of them model the notches as if they could be
strain fields around notch and crack tips is a very important prob- properly described by their two-dimensional (2D) approximations,
lem in structural engineering, particularly in fatigue applications. assuming plane stress (pl–r), plane strain (pl–e), or axisymmetric
Notches act as localized stress raisers that can serve as fatigue conditions. Creager and Paris [5] proposed a method to approxi-
crack initiation points, thus their effects must be properly quanti- mate SCF from the Stress Intensity Factor (SIF) of a similar crack,
fied in all structural integrity evaluation tasks. but, again, most listed SIF also assume plane geometries, see e.g.
Tada et al. [6].
1.1. Notch stress fields However, 2D models of notched components have important
limitations, even in very simple cases. Consider, for instance, a
For design purposes, the maximum stresses r0 that act at notch notched plate loaded by a uniform nominal stress ry = rn, see
tips are usually calculated by using a geometry-dependent, Linear Fig. 1. Far from the notch tip, the plate material is subjected to a
Elastic (LE) stress concentration factor (SCF) Kt to multiply the pl–r state, but the stress and strain fields that surround the notch
nominal stress rn that would act there if the notch had no effect tip are in fact 3D, due to the restriction of the Poisson contraction
on the stress and strain fields that surround it: induced by the stress/strain gradients that act there. A transversal
constraint factor Tz can be defined to quantify this restriction at
r0 ¼ K t  rn ð1Þ any point by the ratio between the transversal stress rz and the
sum of the in-plane stress components rx and ry. Under pl–r limit
Pioneer analytical solutions for SCF were obtained by Kirsch in conditions, the only non-null stress components are rx and ry, thus
1898 [1] and Inglis in 1913 [2], who studied infinite plate with cir- Tz = 0 in such cases, whereas under pl–e conditions, Hooke’s Law
cular and elliptical holes respectively. Since then, a few analytical leads to Tz = m. Therefore,

⇑ Corresponding author. Tel.: +55 21 3527 1642. 

E-mail addresses: [email protected] (R.C. de Oliveira Góes), jtcastro@ rz 0; pl  r
Tz ¼ ¼ ð2Þ
puc-rio.br (J.T.P. de Castro), [email protected] (L.F. Martha).
1
rx þ ry m; pl  e
Tel.: +55 21 3527 1639.
2
Tel.: +55 21 3527 1188.

0142-1123/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijfatigue.2013.10.014
160 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170

Nomenclature

a notch or crack length; elliptic hole or semi-elliptic notch Tz transversal constraint

major semi-axis W half plate width
b elliptic hole or semi-elliptic notch minor semi-axis r stress
B plate thickness e strain
H half plate height q notch tip radius
K stress intensity factor rn nominal stress
Kt 2D 2D stress/strain concentration factor ()0 subscript indicating ‘‘at the notch tip’’
Kr 3D 3D stress concentration factor ()mp subscript indicating ‘‘at the middle plane’’ (z = 0)
Ke 3D 3D strain concentration factor ()surf subscript indicating ‘‘at the free surface’’
SCF stress concentration factor
SIF Stress Intensity Factor

Very few analytical solutions are available for stress fields the notch tip Tz0 is maximal at the notch mid-plane and decreases
around 3D notches. Youngdahl and Sternberg solved the LE infinite to zero close to the free surface of the plate. Also, the constraint le-
solid with a 3D ellipsoidal cavity problem, and also obtained vel decays with the distance from the notch tip. The normalized
approximate solutions for the 3D stresses in an infinite plate of fi- constraint gradient Tzmp(x)/Tz0mp along the x-direction at the mid-
nite thickness with a circular hole [7,8], but numerical tools are al- plane (z = 0) of notched plates was shown to be independent of
most indispensable to study more complex 3D stress concentration the notch configuration and tip radius, being well fitted by the fol-
problems. For example, through extensive finite element (FE) mod- lowing equation [9]:
eling of various notch configurations, Guo et al. [9–11] investigated  
T zmp ðxÞ x 2  x 4
3D elastic fields on notched plates of finite thickness subjected to ¼ 1  4:35 1 þ 0:686  1 þ 0:686 ð3Þ
tension loads, coming to important conclusions summarized as T z0 mp B B
follows.
SCF along 3D notch tips depend on their configuration and on Moreover, unlike crack problems, finite notch tip radii cannot
their thickness-to-tip-radius ratio B/q. They observe that, although provide enough constraint to reach limit pl–e conditions along
the stresses along the notch front ry0 may vary significantly, the the notch fronts. Yang et al. [12] presented similar results and
correspondent ry(x, z0 )/ry0(z0 ) stress ratio distributions ahead of showed that stress and strain concentration effects are decoupled
the notch tips at any given z’ plane are almost z-independent. Par- along the thickness, even within the LE regime. Therefore, instead
ticularly, the rymp(x)/ry0mp ratio along the notched plate mid-plane of the single SCF Kt = rmax/rn = emax/en used in 2D analyses, inde-
z = 0 is almost insensitive to the plate thickness B and to the notch pendent Kr = rmax/rn and Ke = emax/en stress and strain concentra-
geometric configurations up to x/q ffi 0.75, and can be approxi- tion factors should be considered when analyzing 3D notch
mated by the 2D notch solution. Moreover, the 3D affected zone problems.
is somewhat independent of the notch configuration for notches
with a/q P 1, and it is limited to a distance of approximately 3B/ 1.2. Crack stress fields
8 from the notch tip. Concerning the transversal constraint, the
through-thickness variation of Tz0(z)/Tz0mp is also nearly indepen- The traditional description for singular LE stress fields ahead of
dent of the notch configuration. The transversal constraint along crack tips was independently obtained by Williams and Irwin in
1957 [2]. Williams proposed an asymptotic infinite series to de-
scribe 2D fields around crack tips, usually approximated by its first
term, the SIF. This well known one-parameter SIF-based approxi-
mated solution
pffiffiffiffiffiffiffiffiffi for a crack loaded in mode I is given by
rij ¼ ðK I = 2prÞg ij ðhÞ, where r and h are polar coordinates centered
at the crack tip and i = x, y.
Like for notches, 3D crack solutions are scarce as well. Concern-
ing the crack behavior at free surfaces, Bazant and Estenssoro [13]
showed how the angle b, with which the crack intersects the
free surface, is related with the singularity of the stress field at
this point. In the particular case of a crack under pure mode I,
they show that for b = p/2, KI must be zero at the free surface
and, for the crack to achieve a singularity proportional to r1/2,
the b – p/2 value solely depends on m.
Nakamura and Parks [14,15] presented valuable numerical re-
sults on 3D LE fields around an ideal crack tip with a straight front
within the SIF-dominated zone in a thick plate. The region close to
the crack tips was modeled as a disk of radius R centered at the
crack tip, considering the crack size a as long with respect to the
cracked plate thickness (a  B). The boundary of the disk (r = R)
was loaded by the displacement field generated by the 2D SIF KI
and KII applied on the plate, using the so called Boundary Layer ap-
Fig. 1. Notched plate under uni-axial load with Cartesian coordinate axes origin at proach. Strong 3D effects were observed within a distance r = B/2
the center of the notch tip. from the crack tip, with a 3D–2D transition occurring within a
 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170 161

distance B/2 < r < 3B/2 from the crack tip. Besides, the SIF was
shown to significantly vary along the crack front when compared
with the classical 2D prediction.
Parks and Nakamura’s Boundary Layer model was revisited by
She and Guo [16], who fitted expressions for their FE solutions
and proposed a 2-parameter description of the crack tip stress
fields from the 2D SIF Kfar that governs the LE fields far from the
crack tip and the transversal constraint factor Tz.
Although ingenious, Boundary Layer models have intrinsic con-
ceptual limitations that should not be disregarded for some practi-
cal applications. To start with, they assume that the crack is much
longer than the plate thickness. Hence, they may not represent
well a crack in its early stages of propagation, when its size a is
of the order of B or smaller, a most important problem for fatigue
life predictions. In addition, the cracked plate stress field is ob-
tained assuming that the plate is far-field loaded by the stresses in-
duced by the SIF applied on it. Therefore, all limitations associated
to K-field assumptions are incorporated by the boundary layer
model. For instance, K-description for the (assumed LE) stress
fields in cracked components is strictly valid only very close to Fig. 2. Mesh close to notch tip, for the elliptical hole with b/a = 0.5 and q/a = 0.25.
the crack tips, exactly where plasticity-induced perturbations tend
to spoil it. Such assumptions also fail to describe the situation
where the 3D affected zone surpasses the K-dominated region. mid-thickness and to the xz plane, using Fig. 1 notation. The EH
Moreover, since K-fields do not reproduce the nominal stresses models received additional symmetry with respect to the yz plane.
far from the crack tips, non-negligible effects induced by high rn The uniform load is applied as a uniform tensile stress on the supe-
cannot be accounted for by such models [17]. rior plate face y = H. The notch tip region is described by structured
Furthermore, ideally straight cracks are just a convenient math- meshes, see Fig. 2, with a maximum element size of 0.1q at the
ematical trick, as experiments show that they propagate with notch tip, where q = b2/a is the notch tip radius. The resulting lin-
curved fronts, a phenomenon known as crack tunneling. Extensive ear equation systems are solved using ABAQUS sparse solver. The
research on the tunneling phenomenon [18–20] shows that the dimensions of the analyzed notches are listed in Table 1.
curved front shape in a through-cracked plate can present a tun- Fig. 3 shows Kr and Ke distributions at the notch tip along the
neling depth (amax  asurf) up to 0.05B. This slight curvature is plate thickness for an EH with b/a = 0.5 and q/a = 0.25, to illustrate
shown to bring considerable impact on SIF calculations along the the typical results obtained from such analyses. Note in particular
crack front. Through a recursive remeshing algorithm, Wu [21] that Kr – Ke, and that for relatively thick plates (with B  q) their
was able to numerically obtain the shape of the crack front for a maxima values occur close to the plate surface, whereas for thinner
given KI distribution in a through-cracked plate, and associated a plates such values occur at the plate center [11,12].
crack tunneling depth around 0.025B to a uniform KI distribution Fig. 4 shows how Kr and Ke depend on B/q. For relatively thin
along the crack front. plates or blunt notches (those with low B/q ratios), the maximal
This work first revisits the most recent literature concepts on stress and strain rmax and emax occur at the middle plane of the
3D LE notch analysis by presenting several FE models results, and plate. For thicker plates and sharper notches, they dislocate to-
discusses the importance of 3D effects on notch design issues. Then wards the plate free surfaces (located at z/B = 0.5), in a slightly
it uses powerful sub-modeling techniques, which avoid the intrin- decoupled way. Note how different such Kr and Ke values can be
sic limitations of the K-field domination and the long crack hypoth- (up to 15% for the analyzed models). Also, the maximum value of
eses, to simulate 3D SIF distributions for large single edge cracked the SCF at the notch tip Krmax can be up to about 8% higher than
plates with several B/a ratios, and to evaluate their true influence the 2D SCF Kt, a non-negligible difference. Therefore, such results
on the KI distribution. On the sequence, the FRANC3D code [22] show that Kt values measured (as usual) from the strain fields ob-
is used in conjunction with the ABAQUS FE [23] solver to simulate served at the free surfaces may severely underestimate rmax at the
the 3D growth of an initially straight front crack with initial length notch root, indicating that 3D effects on the stress and strain fields
a0, assuming it can be described by the classical Paris rule, to eval- along notch roots may indeed be significant for some practical
uate how the crack front shape and the KI distribution change as applications.
the crack grows. Finally, the influence of a0/B and of Paris’ expo- Fig. 5 shows how Krmax, Krmp, and Krsurf, vary with B/q for the
nent n on the crack front shape transition is also discussed. EH, synthesizing many results recently published in the literature.
It reinforces that the error in Krmax predictions based on 2D models
can be up to about 8%.
2. 3D effects on stress fields around notch tips
Table 1
Elliptical and semi-elliptical notch parameters analyzed in the present FE models.
To evaluate notch-induced 3D stress concentration effects, sev-
eral Elliptical Holes (EH) and Semi-Elliptical (SE) notches with Notch b/a q/a B/q
semi-axes a and b in large plates of width W and height H under Elliptical 1 1 0.1, 0.2, 0.4, 0.6, 1, 1.5, 2, 3, 4, 6, 8, 10, 20
uniaxial load were simulated in ABAQUS, using W/a = H/a = 60 to 0.5 0.25 0.4, 0.8, 2, 2.8, 4, 6, 8, 12, 16, 32, 48
avoid boundary effects within 1% error. C3D20 full integration qua- 0.2 0.04 3, 6, 10, 15, 20, 30, 50, 75, 100
0.1 0.01 0.4, 0.6, 1, 2, 4, 6, 10, 20, 40, 60, 100
dratic solid element from the ABAQUS library was used in all sim-
ulations. To check similar previously published results, E = 200 GPa Semi-elliptical 1 1 0.2, 0.3, 0.5, 0.7, 1, 2, 3, 5, 7, 10, 20, 30
0.5 0.25 0.16, 0.24, 0.4, 0.8, 2, 4, 8, 12, 16
and m = 0.33 are used, although m = 0.29 would better match the
0.2 0.04 0.5, 1, 3, 10, 20, 30
chosen modulus, a typical value for steels. The models were built 0.1 0.01 0.6, 6, 10, 20, 40, 60, 100, 200, 400
with symmetry with respect to the xy plane at the plate
162 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170

Fig. 3. (a) Ke/Kten and (b) Kr/Ktrn distribution along the notch front, for an elliptical Fig. 4. (a) Krmax/Kt and Kemax/Kt variation with B/q for elliptical holes and (b)
hole with b/a = 0.5 and q/a = 0.25. position of Krmax and Kemax along the notch root.

The maxima stress and strain positions indicate the preferred

locations for crack initiation, whereas the stress gradients ahead
of such critical points affect how a short crack propagates from
them. According to the results presented above, if the cracks do
prefer to start at maxima stress and strain points, as usually as-
sumed in most damage models, they should do so in the center
of the thinner notched plates (z/B = 0) and closer to the free sur-
faces (z/B = 0.5) of the thicker ones. But the growth of such initially
small surface cracks is strongly dependent on the stress gradient
around the notch tip, as discussed elsewhere [24–26]. Since the
studied notches have much stronger stress gradients in the x than
in the z-direction [9], the short crack driving force decrease is shar-
per ahead than along the notch tip direction. Therefore, it may be
argued that cracks initiated at a notch tip should prefer to advance
first along it, trying to become a passing crack, then along the x-
direction, inwards the specimen. However, although reasonable, Fig. 5. Variation of Krmax/Kt, Krmp/Kt, and Krsurf/Kt with the thickness to root radius
such speculations certainly need further investigation. ratio B/q for the elliptical holes.

3. 3D effects on notch design procedures using the previously defined out-of plane transversal constraint
factor Tz, then rz is given by
Although the differences between 2D and 3D stress fields
around notch tips may be non-negligible, it is important to verify 8
< rTresca ¼ ry
as well the influence the studied 3D stress/strain fields may have qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rz ¼ T z ry ) ð4Þ
in the main failure criteria employed to design structural compo- : rMises ¼ ry 1  T z þ T 2z
nents, by comparing typical predictions based on them with tradi-
tional predictions made using simplified 2D hypotheses. As the
only non-null stress components along notch tips are ry and rz Therefore, the transversal constraint Tz does not affect rTresca
(since such tips are free surfaces), they are both principal but tends to decrease rMises, as it restricts distortion despite not
stress components, and have the same signal rn has. Since rTresca = affecting smax along the notch tip. Assuming that ry0 at the notch
(r1  r3) and rMises = {[(r1  r2)2 + (r2  r3)2 + (r3  r1)2]/2}1/2, root is calculated from a 2D solution (ry0,2D = Ktrn), ry0,2D contains
 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170 163

an intrinsic error with respect to the 3D solution. At the notch tip,

the corresponding errors in Tresca and Mises stresses are:

1 þ err Tresca ¼ ry0 =K t rn
ð5Þ
1 þ err Mises ¼ rMises =K t rn

The results presented in Section 2 show that ry0 > Ktrn when
the thickness-induced transversal constraint Tz is accounted for.
Therefore, 2D SCF predictions are intrinsically non-conservative,
although not that much. Fig. 6 complements such results by show-
ing the ratio rMises/Ktrn in function of Tz for several arbitrary errors
on ry0 calculated from 2D plane solutions. Hence, even though ry0
predictions based on SCF calculated assuming 2D conditions are
always non-conservative, the Mises stresses around notch tips
caused by the 3D conditions (those that actually act there) may Fig. 7. Ratios of the maximum, mid-plane, and Mises 3D stresses with respect to
be estimated conservatively or non-conservatively by the tradi- the 2D maximum stress calculated by Ktrn, respectively Krmax/Kt = rmax/Ktrn,
tional 2D estimate rMises ffi Ktrn, a somewhat non-intuitive result. Krmp/Kt = rmp/Ktrn, and KrMises/Kt = rMises/Ktrn, as a function of B/q.

Poisson’s ratio m is the upper bound limit to Tz, never reached for
notches with finite tip radii, no matter how thick the plate is. Guo
et al. proposed an estimate for the Inglis’ hole Krmax [10]: built with B/a ratios varying from 0.1 to 100. In-plane displacement
fields (ux, uy) from the global model solution were applied to every
K rmax =K t ffi 1 þ 0:01  exp½m=ð0:14364 þ 0:07b=aÞ ð6Þ node of the sub-models boundary surfaces, while their out-of-
plane displacements uz were left free. To maintain kinematic com-
p patibility between global and local solutions, it is necessary to en-
Since in this case K t ¼ 1 þ 2a=b ¼ 1 þ 2 ða=qÞ; and since this Kt
sure that the sub-model dimensions are large enough to allow the
can be used as a reasonable approximation for the SCF of many
3D stress field close to the crack tip to fade to the pl–r condition
other notches with size a and notch tip radius q, then the maxi-
assumed in the global model. In other words, the sub-model
mum SCF for other notches could be estimated by
dimensions Wsub/a, and Hsub/a must be chosen so that Tz presents
K rmax =K t ffi 1 þ 0:01  expfm=½0:14364 þ 0:14=ðK t  1Þg ð7Þ a smooth asymptotic behavior towards zero within the sub-model
limits. Since the size of the 3D affected zone for an arbitrary B/a
value was not known prior to simulation, the sub-models were
However, such approximated predictions are questionable for
arbitrarily built with both Wsub/a and Hsub/a > 5B/a. The results
Kt  3. For design purposes, 1.08Kt is a better assumption for the
presented following show that such limits were adequately
Krmax upper bound and, consequently, for Tresca’s analyses based
chosen.
on 2D SCF. Nevertheless, as mentioned above, the maximum value
Besides being numerically efficient, this procedure also has
of rMises around notch tips tends to be still less sensitive to such 2D
some non-negligible advantages over the Boundary Layer approach
approximations, see Fig. 7.
[14–16]. Its crack tip fields are calculated considering all the load
characteristics, since they are not restricted to the limitations of
4. FE models for cracks at the border of large plates SIF-based hypotheses. Hence, it recognizes e.g. the nominal stress
effects far from the crack tip, which are ignored when K-field load-
Several LE FE 3D analyses are performed to evaluate the influ- ing conditions are assumed to be valid. It also allows analyses of
ence of the thickness-to-crack-size B/a ratio on the crack tip fields cracks with high B/a ratios, thus of relatively shallow cracks com-
of large edge cracked plates under uniaxial loads. Sub-modeling pared to the plate thickness. It must be emphasized that for the
techniques [23] were employed to take advantage of the fact that estimation of fatigue crack lives in practical applications the
most of the plate is expected to respond in pl–r, with 3D stress behavior of such cracks is much more important than the behavior
state limited to the proximities of the crack. of long cracks.
A large global model for the plate was built using plane ele- The sub-models were built assuming symmetry with respect to
ments, with overall dimensions Wglobal/a = Hglobal/a = 1000, while xy and xz crack planes. Fifteen elements are used along their thick-
several 3D sub-models of the region surrounding the crack were ness (the z-direction) with sizes varying in geometric progression
from the middle-plane (coarser) to the free surface (finer), with a
progression ratio q = 1.3. The circumferential direction is divided
into 24 elements. In the radial direction, the elements are built
with size 0.003B at the very crack tip, coarsening in geometric pro-
gression with ratio q = 1.15. This refinement pattern, which is sim-
ilar to the one used in [16], is enough to guarantee numerical
convergence. Fig. 8 illustrates the models used in these analyses.
In other words, a 2D solution is applied to model this plate far
field conditions because (within the LE assumption) it is possible
to establish 3 distinct zones or domains for the stress and strain
fields in that cracked plate: (i) Very far from the crack tip the crack
existence is not even noted, hence the stresses and strains in this
domain behave as if the crack did not exist. Hence, if the plate is
large enough to fulfill this condition, this domain works under con-
stant stresses or constant displacements (or, more generally, under
constant nominal plane stress field conditions), and such a hypoth-
Fig. 6. rMises/Ktrn as a function of Tz0, for various ry0 values. esis may be used to model it and its contour conditions. (ii) In the
164 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170

Fig. 8. Global and sub-model concepts, and sub-model 3D mesh.

cracks. Comparison with previous results obtained for long cracks

[14,16] shows that as the cracks get longer, or as their B/a value de-
crease, the KI distribution along the crack front gets closer to the
results calculated assuming SIF-dominated far field boundary con-
ditions, the limit behavior expected from such simulations. How-
ever, they also indicate that shallow cracks behave differently
from long cracks. Their SIF distribution tends to the 2D solution
along most of the crack front, but it shows a peak close to the plate
free surface, which does not appear for the long cracks. This differ-
ence between the short and the long crack behaviors may affect
how they grow by fatigue, thus should not be neglected when ana-
lyzing them. Note once again that in the notation used here ‘‘long
cracks’’ mean ‘‘cracks large in comparison to the plate thickness B’’,
not to the plate width W. The analysis of very deep cracks, those
with large a/W ratios, must include the influence of the back face
plane on the LE fields ahead of the crack tip, a problem considered
beyond the scope of this work.
Fig. 9. KI/KI,2D distribution along the front of short and long cracks. Fig. 9 also shows that the SIF value drops in the vicinity of the
free surface. As the crack front is assumed straight and perpendic-
intermediate domain, closer to the crack tip but not within its ular to the free surface in such 3D analyses, KI in fact should be null
dominance zone, the crack influence is noted and affects the at that surface, where z/B = 0.5, as previously mentioned. However,
stress/strain fields, but such variable fields are still 2D, since there such a limit value could not be achieved with any reasonable level
is no restriction to force them to vary along the z direction. (iii) of mesh refinement. Anyway, its real importance is to force real
Very close to the crack tip, within its dominance zone, such a cracks to slightly curve their fronts during their propagation, as
restriction exists and 3D effects are clearly present, hence the studied later on.
stresses and strains vary along z inside it. Refs. [15,16,18,19] show Fig. 10 shows how the KImax/KI,2D and KImp/KI,2D ratios vary with
that this is true for very long cracks, and the intent here is to inves- B/a, and compares them with the long crack SIF-dominated limit
tigate how such restriction effects work for short cracks. Therefore, solution from [16]. This figure clearly shows how the maximum
3D elements are strictly only necessary in this 3rd zone, but it is SIF values tend to the long crack solution for cracks with very
not possible to precisely model its size prior to its simulation. Even low B/a ratios. It also shows how the SIF at the middle of the plate
so, the sub-modeling technique allows the 3D model frontier to be KImp tends to the limit 2D solution KI,2D as the B/a value becomes
reduced to a perimeter located somewhere within the 2nd domain large and the cracks get shorter. Moreover, the separation of the
without the need to assume a (questionable, to say the least) K- KImax/KI,2D curve from the KImp/KI,2D curve of the short cracks shows
dominance there, recognizing the nominal stress effects that are that their KImax value is higher than the reference KI,2D value, about
not accounted for by the SIF fields and significantly saving compu-
tational effort at the same time. Therefore, the simulation results
obtained from the sub-models can furnish more information than
would be possible to get from assuming a K-field around the crack
tip.

4.1. Influence of the thickness to crack size ratio B/a on the stress fields
around crack tips

Since the studied sub-models do not assume a relatively long

crack (compared to the plate thickness), they can be used to study
the behavior of small cracks and to identify how different it can be
from the long crack behavior. Fig. 9 shows how much the normal-
ized SIF distributions along idealized straight crack fronts KI/KI,2D
deviate from standard 2D solutions. Such distributions are pre-
sented for a wide range of B/a values, including short and long Fig. 10. KImax/KI,2D and KImp/KI,2D variation with the crack size in large plates.
 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170 165

3% higher for the Poisson coefficient m = 0.3 used in these numerical

simulations.
It is interesting to compare the long and short crack behavior
with that of similar shallow and deep elliptical notches with a very
sharp tip radius. The KI distribution along the crack tips shown in
Fig. 9 is notably similar to the Kr and Ke distributions along the
notch tips presented in Fig. 3. They all tend to concentrate stresses
more in the middle of thinner plates, and close to the free surface
of thicker plates. This similar behavior certainly indicates that
these numerical results are consistent. For instance, KImax/KI,2D
and KImp/KI,2D presented in Fig. 10 behave very much like
Krmax/Kt and Krmp/Kt after their peak, as shown in Fig. 5. Even
the fact that KI must be zero at the free surface is suggested in
Fig. 5, where Krsurf/Kt is shown to decrease as the notch sharpens.
In addition, the present model can clearly distinguish the K- and
the rn-dominance on the crack-induced stress field, a matter of
great importance for validating LEFM predictions [17]. Fig. 11 illus-
trates how the ratio rymp/rn varies along the normalized distance
x/a ahead of the crack tip in this large edge cracked plate. This
rymp/rn ratio is almost independent of the crack size. Indeed, note
in that figure that at relatively small distances from the crack tip,
say up to x/a ffi 0.2, the rymp(x) stress distribution is clearly domi-
nated by the idealized stress field induced by KI alone, and that
rymp/rn ? rn for distances x/a greater than about 2. Since this va-
lue is independent of the crack size (at least for the large crack size
range 0.1 6 B/a 6 100 studied here,) this result clearly indicates
that estimates for the LE stress field far from the crack tip in this
plate should include a rn correction for distances x/a > 0.2.
Fig. 12 illustrates how ry(z)/rn, the ratio between the normal
stress perpendicular to the crack faces and the nominal stress var-
ies with the relative distance x/a ahead of the crack tip at various
positions z/B along the crack front, for cracks sizes B/a = 0.4, 4,
and 40. This figure shows that the z/B influence on the stress gra-
dient is slightly more important close to the plate free surface.
Tz gradients ahead of the crack tip are a good indicative of how
much material is subjected to high transversal constraints due to
crack-induced perturbations on the (LE) stress/strain field that sur-
rounds it. In Section 2 it was shown that in notched plates the con-
straint factor Tz is maximal at the plate mid-plane (at z/B = 0), and
that as the notches become sharper Tz ? m, but it never reaches this
value since notch tips radii q remain finite. Fig. 13(a and b) shows
the behavior of the transversal constraint factor along the mid-
plane of the cracked plate Tzmp/m ahead of the crack tip, for cracks
with sizes in the large range 0.1 6 B/a 6 100. Note that at the mid-
dle point of the crack front (x ? 0, z = 0) the limit condition Tzmp = m
is always achieved, no matter the crack size. In other words, any

Fig. 12. ry(z)/rn distribution ahead of the crack tip in the large edge cracked plate
versus x/B for (a) B/a = 0.4, (b) B/a = 4, and (c) B/a = 40.

crack (with q ? 0) reaches the pl–e limit value for the transversal
constraint Tzmp at the mid-point of its tip, at least under LE condi-
tions. Note that in Fig. 13(a), x is normalized by B, whereas in
Fig. 11. rymp/rn distribution ahead of the crack tip in the large edge cracked plate. Fig. 13(b) it is normalized by a.
166 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170

yield higher values when measured in thinner specimens, and

the measured toughness decreases as the specimen thickness in-
creases until reaching a material-independent value, the so-called
‘‘plane strain fracture toughness’’ KIC. The traditional ASTM E 399
standard specifies B > 2.5(KIC/SY)2 as the minimum specimen thick-
ness required for a valid KIC value. Moreover, it somewhat oddly
also specifies a relatively small range 0.45 < a/W < 0.55 for (large)
crack sizes acceptable to validate such values. Much less stringent,
but similar requirements apply for J or for critical CTOD tests.
The presented results are obviously aligned with this concept. If
the toughness is related to the plastic work ahead of the crack tip,
it depends on the transversal restriction that acts there. The max-
imum restriction, which occurs under pl–e conditions, is reached at
some points by all cracks, due to the very high stress gradients
ahead of their tips. However, although the calculated values are
based in LE assumptions, they clearly indicate that the definition
of ‘‘dominant plane strain conditions’’ must be associated not only
to the capacity of reaching it at some point, but also to the amount
of material that is subjected to very high transversal constraints
associated to it. As expected, the size of the highly constrained
zone ahead of the crack tips is in fact dependent on the plate thick-
ness in the analyzed cases, growing with it until asymptotically
reaching a stable value at some B/a ratio. Fig. 14 shows the area
ahead of the crack tip where the ratio Tz/m falls within an arbitrary
value c, Tz/m 6 c, while Fig. 15 shows the boundary profile of the re-
gion ahead of the crack tip inside which Tz/m = 0.5 for several B/a
ratios. Note that it is necessary to have B/a > 60 for obtaining a uni-
form penetration of the largest transversal constraint Tz/m = 0.5
along 90% of the crack front, a surprisingly high value.
Moreover, the size of the transversally restricted zone ahead of
the crack tip is very small for relatively short cracks, indicating that
they should tend to have higher freedom to activate plasticity re-
lated phenomena, such as the toughness. Hence, based on these re-
sults, it can be said that short cracks tend to be less damaging not
only because they are small, but also because they induce less
transversal restriction ahead of their tips. As B/a increases, so does
the size of the restricted zone ahead of the crack tip. Therefore, its
effect on the toughness is expected to also increase, up to a critical
Fig. 13. Behavior of the transversal constraint along the mid-plane of the cracked
plate Tzmp/m in function of the relative position ahead of the crack tip: (a) versus x/B
B/a value, after which the restricted zone grows no further. At this
and (b) versus x/a. point, the toughness is expected to reach its minimum value and to
become independent on the crack size. Once again, although based
on LE calculations, this behavior can provide a reasonable explana-
These results show that for relatively long cracks, say those tion for why SIF, a purely LE concept, can in practice control this
with B/a < 6, the transversal constraint gradient at the plate mid-
plane Tzmp converges to a function of (x/B), better visualized in
Fig. 13(a), which agrees reasonably well with Eq. (3), fitted for
the long notches, and with the long crack solution [16]. However,
for shorter cracks, with higher B/a ratios, the results presented here
show that the transversal constraint effects decrease more rapidly
ahead of the crack tip. Essentially, this means that the size of the
restricted zone ahead of a long crack is proportional to the plate
thickness B. Meanwhile, shorter cracks induced restriction to a
shorter x/B distance, a behavior demonstrated by the steeper Tzmp
versus x/B gradients in Fig. 13(a). The short crack limit behavior
is better noted in Fig. 13(b): as B/a increases (say B/a P 20), Tzmp
tends to a function of x/a, meaning that the restricted zone induced
by a very short crack is in fact proportional to the crack length it-
self. However, keep in mind that under LE assumptions, the con-
cepts here referred as ‘‘long’’ and ‘‘short’’ crack are equivalent to
‘‘thin’’ and ‘‘thick’’ plate, as the only varying parameter is the B/a
ratio. Therefore, it is equivalent to say that in thin plates the size
of the restricted zone ahead of the crack tip is proportional to
the plate thickness and, in thick plates, to the crack length.
It is well known that the material toughness depends on the
Fig. 14. Relative areas of the transversally restricted zones ahead of the crack tip,
specimen geometry, particularly on its thickness, unless it is mea- for various c values (the maximum restriction occurs under pl–e limit conditions,
sured under ‘‘plane-strain conditions.’’ Toughness tests usually when Tz = m), for cracks with sizes 0.1 6 B/a 6 100.
 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170 167

 n
DK ji
Daji ¼ Dajmean ð8Þ
DK jmean

Note that as a result of the assumed (reasonable) hypotheses,

the SIF distribution along the entire crack front K(z), and conse-
quently the ratios between the SIF increments at each crack front
node and the mean SIF increment at each load step, DKji/DKjmean,
are all a uniquely function of ai(z), the crack length at each node
in that given step. Hence, Dajmean becomes an arbitrary analysis
parameter, dissociated from the number of load cycles and from
everything else. The length value at the plate free surface, asurf, is
adopted as a descriptive parameter of the crack overall length,
since it can be measured in physical experiments by optical meth-
ods, or by any other surface-dependent technique for measuring
Fig. 15. Regions ahead of the 3D crack tip where the transversal restriction Tz is at
crack length such as crack gages e.g.
least equal to half the maximum value it can achieve, Tz = m/2, for cracks with sizes After solving each particular crack propagation step, the ob-
0.1 6 B/a 6 100. tained crack front increment is smoothed and fitted by a polyno-
mial, in order to minimize the unavoidable numerical noise
and other plasticity-induced damage mechanisms, like in particu- associated with the KI(z) solution. A high order polynomial (7th de-
lar fatigue crack propagation [27].
gree) was chosen in order to capture the odd KI distributions char-
However, the 3D models studied so far are too simple to explain acteristics of shallow cracks. The simulated edge-cracked plates are
some features of the fatigue cracking problem, even when it is
built with the same overall dimensions H, B, and W. The initial edge
clearly K-controlled. Indeed, even if the crack could somehow orig- cracks are introduced with idealized straight fronts, but with dif-
inate with a perfectly straight front on the plate edge, experimental
ferent depths a0. Values of Dajmean between 0.002B and 0.05B were
observations show that real fatigue crack fronts are in fact curved. used along the crack propagation simulation, varied in order to
The next section explores this problem, and studies how an ideally
deal with convergence issues in the calculation of the crack front
straight crack front must curve by mechanical reasons as the crack in step j + 1. The plate models assume symmetric boundary condi-
grows by fatigue.
tion at the plate mid-plane z = 0 and are supposed tensioned by a
unitary uniformly distributed load at their upper and lower bound-
5. Simulation of the growth of edge cracks with initially straight aries. Fig. 16 shows the overall model used in these simulations, as
fronts in large plates well as the two possible curved crack fronts after the initially
straight crack propagates for a while. Table 2 presents the param-
FRANC3D was used to simulate the growth of large and small eters used in the various models.
edge cracks with an initially straight front in large plates. This ver- Fig. 17 shows how the SIF ratio between the 3D and 2D results
satile Fracture Mechanics code is equipped with specific routines KI(z)/KI,2D varies along the crack front with increasing values of asurf
to track the local propagation of each point along a 3D crack front for a0/B = 0.02 and a Paris’ exponent n = 2. It also shows the crack
and to obtain the crack front shape of the next propagation step, as front shape evolution along the plate thickness for the same crack
described elsewhere [22]. Within this work scope, the fatigue incremental stages, quantified by the (a(z)  amin)/B ratio. The con-
cracks are assumed to grow under LEFM conditions. Since the figurations assumed by the crack front while it grows from the ini-
solution of LE problems is unique and proportional to the imposed tially straight shape with a0/B = 0.02 show first an anti-tunneling
load P, the calculated SIF Kcal along the crack front can be inter- and then a tunneling effect, driven by the non-uniform KI(z) distri-
preted as a shape function. Therefore, the following equations bution along the crack front at each crack increment. This non-
apply to describe the various crack loading parameters: Kmax = PKcal, intuitive behavior occurs because the crack front naturally curves
Kmin = PRKcal, R = Kmin/Kmax, and DK = Kmax  Kmin. Without any itself looking for a more uniform SIF distribution along it. The
loss of generality, the analyses developed here considered R = 0 non-uniform SIF distribution along the initially straight crack front
and P = 1, thus in the propagation cases described below
DK = Kmax = Kcal.
Moreover, it is assumed that the local crack advance at any spe-
cific point of the crack front follows the simple Paris’ fatigue crack
propagation rule. Since the material is assumed isotropic and
homogeneous, its crack propagation behavior along the crack front
is also assumed as so. Hence, the crack growth steps are assumed
to depend only on the local value of their crack driving force DK,
since Kmax, the other FCG driving force, is maintained fixed during
these simulations.
From the provided Kcal values at the!crack front nodes in any gi-
ven step, a crack propagation vector da is obtained, which can be
described as the product of a local advance module Da multiplied
by a unitary vector ~ p that determines the local crack propagation
direction. The local direction vector ~
p must be parallel to the crack
plane (xz in this work notation) and normal to the crack front at
each point, as the model is symmetric with respect to the xz plane.
Since the (fatigue) crack growth increments are assumed to follow Fig. 16. (a) Single edge cracked plate studied in this work and (b) geometric
Paris’ rule, they can be described for any given node i at every jth parameters of the tunneling and of the anti-tunneling effects possibly observed
growth step in terms of a given advance parameter Dajmean by along the 3D crack fronts.
168 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170

Table 2 ing fixed load conditions, Paris’ rule usually fits well the phase II of
Parameters used to model the edge-cracked plates. the FCG curves of many structural alloys. Hence it has been suc-
Poisson’s ratio m = 0.3 cessfully used to model 1-D FCG supposing the crack growth is
Young’s modulus E = 200 GPa controlled by DK and uniform along their front in that phase,
Plate thickness B=5 neglecting that the fatigue damage process locally varies along
Plate width W = 4B
Plate height H = 2.5B
their fronts due to microscopic non-homogeneities. Corner and
Crack initial length a0 = 0.02B, 0.2B, and B surface 2D cracks can be modeled in the same way, if it is recog-
Paris rule exponent n = 2.0 and 4.0 nized that their SIF values vary along their fronts. Such macro-
scopic procedures are acceptable if the cracks are large enough
because the specimens used to measure FCG properties generate
tends to disappear after the crack propagates for a while and grad- da/dN curves that are obtained by fitting the average behavior of
ually assumes its characteristic slight curved front. Fig. 18 shows their entire crack front. The original results presented in this paper
similar results for a higher Paris exponent, n = 4. show how such a classical assumption can explain why fatigue
It should be noted that both for n = 2 and for n = 4 such short cracks like to propagate with a slight curved crack front as a con-
cracks start propagating close to the plate free surfaces, exactly sequence of their attempt to achieve an iso-KI regime along them.
where their SIF values along the crack front are maxima for high In this sense, such results explain as well why real cracks do not
B/a ratios, see Fig. 9. Hence, in this particular a0/B = 0.02 case, the grow by fatigue maintaining a straight front. Hence, albeit it is
crack first experiences the anti-tunneling behavior schematized not possible to simulate what is actually happening at every in-
in Fig. 16, in a more pronounced way for n = 4 than for n = 2. More- stant along a real crack front by using the same Paris’ constants
over, as the crack front advances and curves, the SIF distribution for every node along the modeled crack front, it certainly can be
along the crack front KI(z) grows flatter, until each reaches a steady shown where this (macroscopically reasonable) assumption leads
state value approximately 6% higher than the 2D solution for both to. In this way, the results obtained here also demonstrate that
Paris’ exponents. deeper tunneling effects should be associated with further details
Such results deserve some comments. Although to assume that not included in this model, like plasticity-induced crack closure,
FCG rates are controlled by the crack driving forces is a consensual for example. Indeed, if as expected closure effects vary significantly
hypothesis, there is some dispute on which are the actual FCG driv- along the crack front inducing non-negligible variations on the
ing forces. Some prefer DK and Kmax while others defend the use of DKeff values along it, and if DKeff is the actual driving force for
DKeff = Kmax  Kop, but this point is irrelevant for this work. Assum- FCG as assumed by the many supporters of the classical Elberian

Fig. 17. Evolution of the 3D to 2D SIF ratio along the crack front KI/KI2D and of the crack front shape as the crack grows from an initially straight front, for a0/B = 0.02 and n = 2.

Fig. 18. Evolution of the 3D to 2D SIF ratio along the crack front KI/KI2D and of the crack front shape as the crack grows from an initially straight front, for a0/B = 0.02 and n = 4.
 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170 169

model, then more pronounced tunneling effects could be expected

in such cases. However, such a fascinating argument cannot be fur-
ther pursued here due to space limitations.
Fig. 19 shows how the [a(z)  amin]/B ratio, achieved after the
transient from the initially straight crack fronts, varies along the
stable curved crack fronts for all studied cracks. Assuming that
the difference between the maximum and the minimum crack size
amax  amin can be taken as a descriptive parameter of the (curved)
steady state crack front, Fig. 20 shows that all studied cracks
achieve a steady state [amax  amin]/B ratio that is more or less inde-
pendent of a0 and n, after having grown for a distance of about
0.6[asurf  a0]/B. Note that the cracks that initiate at a deeper a0/
B achieve a steady state front shape sooner than the initially shal-
lower cracks. Moreover, cracks that propagate under n = 4 achieve
such stable regime shape sooner than cracks that propagate under
n = 2, as well. The deeper cracks with a0/B = 0.2 and a0/B = 1.0 pre- Fig. 20. Variation of the (amax  amin)/B ratio along the curved crack front as the
sented a more pronounced advance in the plate interior from the crack grows departing from an initially straight front with size a0.
beginning of their propagation from an initially straight front,
while the shallower ones with a0/B = 0.02 presented a higher ad-
vance close to their borders. This behavior is consistent, since the
shallower cracks tend to have KI peaks close to their free surfaces,
while the deeper ones have maximum KI values in the middle of
their fronts.
Fig. 21 shows how the ratio KImax/KI,2D varies as the cracks prop-
agate from their initially straight fronts with size a0. It is worth
mentioning that, except for the shortest studied crack with size
a0/B = 0.02, the cracks are well represented by the straight front
crack solution. The differences between this particular case and
the others are caused by the intense shape changes suffered by
the crack front along the early propagation stages, easily observed
in Figs. 19 and 20. The presented results show that the transient
crack propagation Datrans is closely related to a0/B and fairly inde-
pendent of Paris’ exponent n. Table 3 shows the approximate val-
ues obtained for Datrans. Fig. 21 also shows that results obtained
for long cracks with straight fronts previously reported in [13,15]
are reasonably good approximations for KImax.
Finally, the angle b is obtained from the derivative of the poly- Fig. 21. Variation of the KImax/KI,2D ratio as the crack propagates from an initially
straight front and looks for a stable curved shape.
nomial adjusted for every crack growth step at the intersection
with the free surface, and is shown in Fig. 22. In the present numer-
ical simulations, it was found that after a short propagation dis-
tance, b stabilized at a fixed value around 98°, for all simulated Table 3
cases. The obtained results are 27% lower than Bazant and Estenss- Transient distance from the straight to the steady state crack front.
oro predictions [13], in terms of their difference from 90°. As KI cal-
a0/B 0.02 0.2 1
culation was based in a correlation for plain-strain condition, the Datrans 0.25 0.1 0
solution is not expected to be accurate very close to the free sur-
face, where a plane stress state dominates. However, as the crack
front shape is obtained through a polynomial fit through the nodes
of the entire front, b is not expected to be very much affected by
this impropriety. On the contrary, it is expected to be dominated

Fig. 19. Stable and slightly curved front shapes achieved under steady growth Fig. 22. Variation of the incidence angle b at free surface as the crack propagates
conditions. and curves its initially straight front.
170 R.C. de Oliveira Góes et al. / International Journal of Fatigue 62 (2014) 159–170

by the overall crack front curvature development under the same allowing the use of FRANC3D in this research work. Castro, JTP
(Paris) propagation rule. and Martha, LF are grateful for scholarships granted by CNPq, the
Brazilian Research Council. Castro is also grateful for a grant pro-
6. Conclusions vided by ONR, the Office of Naval Research from the US Navy.

Several FE analyses were performed in order to observe 3D ef- References

fects on the stress/strain fields close to notch and crack tips. The
[1] Timoshenko S, Goodier J. Theory of elasticity. McGraw-Hill; 1969.
obtained results are aligned with the recent literature published [2] Sanford RJ. Selected papers on foundations of linear elastic fracture mechanics.
on the matter. It was observed that the stress and strain concentra- SEM; 1997.
tion along the notch tip is variable, but the ry gradient ahead of the [3] Pilkey W, Pilkey D, Peterson R. Peterson’s stress concentration factors. John
Wiley; 2008.
notch tip can be obtained from the plane 2D solution. Also, the very [4] Savin GN. Stress distribution around holes. NASA Technical, Translation; 1968.
same gradient causes an out-of-plane restriction on the material, [5] Creager M, Paris PC. Elastic field equations for blunt cracks with reference to
which tends to lead the notch tip to plane-strain condition as the stress corrosion cracking. Int J Fract 1967;3:247–52.
[6] Tada H, Paris PC, Irwin GR. The stress analysis of cracks handbook. 3rd
tip radius grows sharper (q ? 0). Concerning notched specimens
ed. ASM; 2000.
design, the present results show that, for notches with reasonably [7] Sadowsky MA, Sternberg E. Stress concentration around a triaxial ellipsoidal
low SCF (say Kt < 3), 8% is a safe margin to estimate the maximum cavity. J Appl Mech 1949;16:149–57.
non-conservative error committed in the maximum stress and in [8] Youngdahl CK, Sternberg E. Three-dimensional stress concentration around a
cylindrical hole in a semi-infinite elastic body. J Appl Mech 1966;33:855–65.
Tresca’s equivalent stress predictions, if 3D effects are ignored [9] Li Z, Guo W, Kuang Z. Three-dimensional elastic stress fields near notches in
(i.e., if the 2D solution is used). They also show that rMises at the finite thickness plates. Int J Solid Struct 2000;37:7617–32.
notch tip is much less sensitive to such 3D effects. [10] Yu P, Guo W, She C, Zhao J. The influence of Poissons ratio on thickness-
dependent stress concentration at elliptic holes in elastic plates. Int J Fatigue
Further on, sub-modeling techniques were used to examine the 2008;30:165–71.
3D effects present in cracks on the edge of tensioned large plates [11] She C, Guo W. Three-dimensional stress concentrations at elliptic holes in
with different thickness-to-crack-length (B/a) ratios. Crack tip LE elastic isotropic plates subjected to tensile stress. Int J Fatigue 2007;29:330–5.
[12] Yang Z, Kim CB, Cho C, Beom HG. The concentration of stress and strain in
stress/strain fields were obtained taking into account the full load finite thickness elastic plate containing a circular hole. Int J Solid Struct
description, not restricted to K-field limitations and long crack 2008;45:713–31.
assumptions. Also, the T-stress and the nominal stress influence [13] Bazant ZP, Estenssoro LF. Surface singularity and crack propagation. Int J Solids
Struct 1979;15:405–26.
were intrinsically computed. Both KI and T-stress were observed [14] Nakamura T, Parks DM. Three-dimensional stress field near the crack front of a
to vary along the crack front, presenting maxima KI values always thin elastic plate. J Appl Mech 1988;55:805–13.
higher than the 2D solution. The influence of the B/a ratio on the [15] Nakamura T, Parks DM. Three-dimensional crack front fields in a thin ductile
plate. J Mech Phys Solid 1990;38:787–812.
KI,mp was obtained, and describes a smooth transient from the long
[16] She C, Guo W. The out-of-plane constraint of mixed-mode cracks in thin elastic
crack solution presented before in the literature for B/a = 0.1) and plates. Int J Solid Struct 2007;44:3021–34.
the plane 2D solution (for B/a = 100). It was observed that KI,max [17] Sousa RA, Castro JTP, Lopes AAO, Martha LF. On improved crack tip plastic zone
is always higher than the 2D predictions. estimates based on T-stress and on complete stress fields. Fatigue Fract Eng
Mater Struct 2012;36:25–38.
The obtained stress gradients showed good agreement with K- [18] Lazarus V, Leblond JB. Three-dimensional crack-face weight functions for the
solution for plates with B/a between 0.1 and 100 up to x = 0.2a. semi-infinite interface crack-I: variation of the stress intensity factors due to
From this point on, non-negligible influence of nominal stress some small perturbation of the crack front. J Mech Phys Solid
1998;46:489–511.
was observed. Considerable differences were observed in the re- [19] Leblond JB, Mouchrif SE, Perrin G. The tensile tunnel-crack with a slightly
stricted zone caused by short and long cracks. For long cracks, wavy front. Int J Solid Struct 1996;33:1995–2022.
the reach of restricted zone ahead of the crack tip is proportional [20] Pindra N, Lazarus V, Leblond JB. Geometrical disorder of the fronts of a tunnel-
crack propagating in shear in some heterogeneous medium. J Mech Phys Solid
to the plate thickness (limited to x = 0.4B), while for short cracks, 2010;58:281–99.
the restricted zone is limited by the crack length (up to x = 10a). [21] Wu Z. On the through-thickness crack with a curve front in center-cracked
Finally, several crack growth simulations under Paris rule were tension specimens. Eng Fract Mech 2006;73:2600–13.
[22] Carter BJ, Wawrzynek PA, Ingraffea AR. Automated 3-d crack growth
performed. Initially straight cracks with different initial lengths a0 simulation. Int J Num Methods Eng 2000;47:229–53.
and Paris exponent n progressively curved their front during prop- [23] ABAQUS manual.
agation, simultaneously flattening the KI distribution along the [24] Meggiolaro MA, Miranda ACO, Castro JTP. Short crack threshold estimates to
predict notch sensitivity factors in fatigue. Int J Fatigue 2007;29:2022–31.
front. After some transient propagation, all cracks converged to
[25] Wu H, Imad A, Nouredine B, Castro JTP, Meggiolaro MA. On the prediction of
the same regime crack propagation front, with tunneling depth the residual fatigue life of cracked structures repaired by the stop-hole
close to 2.5% of B. method. Int J Fatigue 2010;32:670–7.
[26] Castro JTP, Meggiolaro MA, Miranda ACO, Wu H, Imad A, Nouredine B.
Prediction of fatigue crack initiation lives at elongated notch roots using short
Acknowledgments crack concepts. Int J Fatigue 2012;42:172–82.
[27] Durán JAR, Castro JTP, Payão Filho JC. Fatigue crack propagation prediction by
The authors are grateful to Professor Anthony Ingraffea, Dr. Paul cyclic plasticity damage accumulation models. Fatigue Fract Eng Mater Struct
2003;26:137–50.
Wawrzynek, and Dr. Bruce Carter (Cornell Fracture Group) for