Anda di halaman 1dari 15

Effect of microcracking on the micromechanics of fatigue crack

growth in austempered ductile iron


J. O R T I Z * , A . P. C I S I L I N O a n d J. L . O T E G U I
Welding and Fracture DivisionINTEMA, Faculty of Engineering, Universidad Nacional de Mar del PlataCONICET, Mar del Plata, Argentina
Received in final form 27 April 2001

A B S T R A C T The effect of microcracking on the mechanics of fatigue crack growth in austempered

ductile iron is studied in this paper. The mechanism of fatigue crack growth is modelled
using the boundary element method, customized for the accurate evaluation of the
interaction effects between cracks and microcracks emanating from graphite nodules.
The effects of nodule size and distribution and crack closure are considered, with
deviation bounds of computed results estimated through weight-function analyses. A
continuum approach is employed as a means of quantifying the shielding effect of
microcracking on the dominant propagating crack, due to the reduction of stiffness of
the material in the neighbourhood of the crack tip. Although the results obtained may
not yield actual numbers for real cases, they are in accordance with experimental
observations and demonstrate how the main factors affect the crack growth of the
macrocrack.
Keywords austempered ductile iron; boundary element method; fatigue crack growth;
microcracking; numerical modelling.
NOMENCLATURE

Ac =surface of the process area at macrocrack tip


Anod =cross-sectional area of nodules
ADI=austempered ductile iron
a=length of macrocracks
c=length of microcracks emanating from nodules
C=constant in fatigue-crack propagation law
d=average minimum distance between nodules
da/dN=crack propagation rate
Km =crack-tip stress intensity factor for microcracks
Kmax , Kmin , Kop =maximum, minimum and opening level of crack-tip stress intensity
factor
Ktip =crack-tip stress intensity factor for macrocracks
E=Youngs modulus
M=number of microcracks per unit volume
m=exponent in fatigue-crack propagation law
Narea =area nodule count
Nvol =volume nodule count
N=load cycles
r=average radius of graphite nodules
rc =radius of penny-shaped microcracks
S=strain density factor

Correspondence: A. P. Cisilino, Welding and Fracture Division


INTEMA, Faculty of Engineering, Universidad Nacional de Mar
del PlataCONICET, Av. Juan B. Justo 4302 (7600), Mar del
Plata, Argentina.
E-mail: cisilino@fi.mdp.edu.ar
*On leave from Universidad Privada del Norte, Trujillo, Peru.

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

591

592

J. O RT I Z e t a l .

U=Elbers crack closure ratio


Vnod =Volume of nodule
a=microcrack nucleation parameter
b=microcrack density
Da=fatigue extensions of macrocrack
Dc=fatigue extensions of microcrack
DK=crack-tip stress intensity factor range
n=Poissons coefficient
DKth =threshold value of crack-tip stress intensity factor range
h=crack propagation angle
s=stress
e=strain

INTRODUCTION

Austempered ductile iron (ADI) belongs to the family of


spheroidal graphite cast irons. ADI combines good
elongation and toughness with high tensile strength; a
combination that increases the resistance to wear and
fatigue when compared with other ductile irons. The
material has a wide range of industrial applications, as
in the case of chain wheels, lines of cement mills, railroad
wheels, gears and automotive crankshafts. The application of ADI will continue to grow as the design
engineer becomes familiar with its properties.1 The
outstanding properties of ADI are a consequence of its
matrix microstructure, obtained by a thermal treatment
called austempering. ADI microconstituents are reacted
austenite (enriched in carbon), retained austenite (unreacted) and acicular ferrite. Minor amounts of martensite
and carbides may also be present. The quantity and size
of graphite nodules, matrix phases formed during thermal treatment, and alloy content influence this microstructure, denominated ausferrite.2
The quantity, size and shape of graphite nodules are,
respectively, characterized by the area nodule count Narea
(nod/mm2 ), the nodule size and nodularity.2 Figure 1(a)
shows the schematic of a standard micrograph of ADI
60100%. This means 60 nodules per mm2, with 100%
nodularity (that is, all graphite in the form of roughly
equiaxed nodules). The mechanical properties of ADI
vary over a wide range of values, mostly controlled by
microstructural factors.3 Some of these factors depend
on heat treatment, as in the case of phases present
(quantity, size and distribution). Others may be related
to solidification, such as graphite nodules (number, size
and shape), and defects (porosity, inclusions, segregated
elements, second phase particles or unwanted eutectics).
In general, the microstructural factors that contribute to
a loss of toughness are: reduction in percentage nodularity (related to the quantity of degenerated graphite),
high percentage or continuity of intercellular or interdendritic carbides, and microporosity.

Fig. 1 (a) Micrograph of spheroidal graphite cast iron


microstructure. White, matrix; black, graphite nodules. Note the
sharp edges in spheroidal nodules (approximate magnification
100). (b) Enlarged 500 micrograph showing the fatigue crack
propagation mechanism in ADI.5

Fracture toughness KIC for ADI is within the range


of 5986 MPa m, clearly higher than those of other
ductile irons, and similar to most of the quenched and
tempered AISI 4140/4340 type steels.4 Fatigue properties

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

FAT I G U E C R A C K G R O W T H I N A U S T E M P E R E D D U C T I L E I R O N

of ADI are also better than those of other cast irons and
wrought steels, as measured by SN curves of polished
specimens. Fatigue properties can be further improved
by means of surface mechanical and thermal treatments.
The relationship between fatigue crack growth and
matrix microstructure is the focus of previous work by
one of the authors.5 A quantitative study of the morphology of fatigue crack growth proved that the crack
path preferentially intersects graphite nodules, and that
a microcracking process takes place in the region of
high-stress concentration around the tip of the macroscopic crack. Graphitematrix interfaces are extremely
irregular, with sharp corners that in some cases constitute
imminent microcracks that emanate from the nodules.
Ultimately the macroscopic crack advances by interaction
and coalescence of the microcracks, as shown in Fig. 1(b).
The authors propose that as microcracks simultaneously
propagate besides the main crack, the available elastic
energy for the propagation of the main crack is lowered
mainly because of the creation of a larger crack surface.
This reduces the general rate of advance and in some
cases causes the premature arrest of crack growth. The
above-mentioned mechanism provides evidence to
explain the relatively low propagation rates and high
effective propagation threshold values for this material.
This paper presents numerical and fracture mechanics
modelling of the mechanism of fatigue crack growth in
ADI, in order to provide further understanding of the
phenomena. The numerical tool for the analysis is based
on the boundary element method (BEM), customized
for the accurate evaluation of the interaction effects
between cracks, microcracks and graphite nodules.
Deviation bounds of the BEM results are estimated
through weight-function analyses. The shielding effect
of microcracking on the (macroscopic) main crack is
studied using a continuum mechanics approach.
NUMERICAL MODELLING

Method of analysis
Numerical modelling of fatigue crack growth requires
the capability of predicting the direction and amount of
crack growth as well as the robustness to update the
numerical model to account for the changing crack
geometry. The dual boundary element method (DBEM)
is a well-established numerical technique in this area of
fracture mechanics, as it eliminates the remeshing problems, which are typical of domain methods and other
boundary element formulations.6 General mixed-mode
crack problems are solved with the DBEM in a single
region formulation, in which the crack growth process
is efficiently simulated with an incremental analysis in
which crack extensions are modelled by adding new

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

593

elements ahead of the crack tips. Details of the DBEM


formulation are given in the Appendix.
Accurate evaluation of crack tip stress intensity factors
K is most important for the effective analysis of crack
propagation problems. For the kind of problems tackled
in this work, where propagation of close crackmicrocrack arrays is analysed, accurate assessment of interaction effects is a key factor. Crack tip stress intensity
factors are computed in this work by using the so-called
one-point displacement formula, in which the relative
displacements of the crack surfaces calculated from the
DBEM model are used in the near-crack-tip stress field
equations to obtain the local mixed-mode K-values.7
This technique was preferred to path and domain
integral methodologies. Although very accurate and
efficient,8,9 these methodologies are not the best option
for the case of close interacting cracks, as the definition
of appropriate integration paths or domains could in
some cases be difficult. On the other hand, the efficiency
of the one-point displacement formula strongly depends
on the accuracy of the displacements calculated on the
crack surface. This is ensured in this work by using
special crack-tip elements that exhibit the correct r
variation for the displacement field. The proposed
discretization strategy proved to be accurate for the
evaluation of the interaction effects between cracks and
microcracks. Details of its implementation and performance can be found in Ortiz et al.10
The incremental analysis of crack extension assumes
a piece-wise linear discretization of the unknown crack
path. For each increment of crack extension, the DBEM
is applied to carry out a stress analysis and computation
of the crack tip stress intensity factor K. The magnitude
and direction of crack increments are then computed for
each crack. Then the crack is extended accordingly by
adding new elements ahead of the previous crack tips.
The above steps are repeated sequentially until a specified number of crack-extensions are reached.
Among the several available criteria for computing the
local direction of crack growth, the minimum strain
energy criterion due to Sih11 is chosen in this work.
This criterion states that the direction of crack growth
at any point along the crack front is towards the region
with the minimum value of the strain density factor S.
The strain density factor S can be written in terms of
the stress intensity factors KI and KII as follows:
2
S(h)=a11 (h)K I2 +2a12 (h)KI KII +a22 (h)K II

(1)

where a11 , a12 , and a22 are functions of the propagation


angle h, which indicates the direction of the crack
extension Da relative to the current crack direction. The
propagation angle h can be obtained for each of crack
extensions by replacing the computed mixed-mode
K-values in Eq. (1) and comparing the values of S(h) at

594

J. O RT I Z e t a l .

stationary points d2S(h)/dh2 >0. This is done by using


the bisection method numerically.
As a continuous criterion, the minimum strain energy
criterion does not take account of the discreteness of the
crack extension procedure. In other words, the crack
path is always defined locally in the same direction
whatever length of crack extension Da is considered.
Therefore, the propagation direction of the crack predicted by the above procedure must be corrected to give
the direction of the actual crack-extension increment.
With this purpose a predictorcorrector algorithm due
to Portela et al.8 is employed. This algorithm ensures
that a unique final crack path is achieved regardless of
the crack-extension length Da.
The fatigue propagation formula due to Klesnil and
Lukas12 was chosen to correlate the crack propagation
rate da/dN, with the crack tip stress intensity factor
range DK=Kmax Kmin , as it accounts for the nearthreshold regime
da
m
)
=C(DK m DK th
dN

(2)

where C and m are material constants and the term DKth


accounts for the threshold value below which cracks do
not propagate.
An important factor to be considered when assessing
fatigue crack propagation is crack closure, given by early
crack-face contact at the crack tip. Many mechanisms
have been identified for fatigue crack closure, such as
plasticity, residual stresses in the crack-tip plastic wake,
and roughness-induced closure and stress-induced metallurgical transformations. The effect of closure on crack
propagation rate is found by considering an effective
crack tip stress intensity factor range
DKeff =Kmax Kop

(3)

where Kop corresponds to the value of K at which the


crack tip opens. Closure data is usually presented in
terms of Elber13 closure ratio U=DKeff /DK.
Finally, the magnitudes of crack-extensions can be
computed using Eq. (2) in an incremental form:
m
m
DK th
)DN
Da=C(DK eff

(4)

This methodology is robust and allows the DBEM to


effectively model general edge or embedded crack problems; crack tips, crack-edge corners and crack kinks. For
further details on the DBEM formulation and implementation the reader is referred to the work by Portela et al.8
Numerical results
Simple models consisting of a macrocrack and a microcracked nodule were considered first in order to study

the effect of crack closure on the crackmicrocrack


interaction mechanism. Crack closure is a relevant factor
when assessing the mechanism of fatigue crack propagation in ADI as it behaves differently for macrocracks
and microcracks.5 In this sense it is worth noting that
although closure levels can be significant for macrocracks, microcracks are mostly closure free.14
The geometry and discretization of the model are
shown in Fig. 2 together with the resulting propagation
paths for three closure levels. The length of the macrocrack was initially set to be 40 times that of the
microcracks. Microcracks were placed to coincide with
the equator of the nodule, where the maximum principal stresses develop. As for all models presented in this
work, graphite nodules were assimilated to circular voids.
This assumption implies a material with 100% nodularity, and the neglect of the mechanical response of
graphite when compared with that of the metal matrix.
The metal matrix is assumed to be isotropic and linear
elastic. Material constants for the propagation law were
obtained from data of DKth tests performed according
the ASTM E-647,15 and using the load-shedding technique. Results are C=4.431010 , m=2.85, DKth =
5 MPa m.5 Closure levels were selected as U=1 (no
closure), U=0.6 and U=0, the last one corresponding
to a limiting case for which the main crack does not
propagate. Loading configuration corresponds to remote
tension, with a load level set in such a way that initial
DK-values for the microcracks are close to the propagation threshold DKth . Growth paths were left to develop
naturally.
Figure 3 illustrates the evolution of DK with load
cycles N for the three closure levels chosen. Stress
intensity factor ranges DK are normalized with respect
to DKth , in such a way that ratios greater than one
represent propagating cracks, and values below one stand
for non-propagating cracks. Note that as the macrocrack
approaches the microcrack emanating from the nodule,
interaction effects cause a substantial increase in DK at
crack tip A, which propagates in the opposite sense to
the macrocrack growth direction until joining it. As soon
as the macrocrack and the first microcrack coalesce,
microcrack B on the opposite side of the nodule becomes
dominant, taking the role of macrocrack tip. The above
mechanism validates the theoretical model proposed by
Greno et al.5 As shown in Fig. 2, the effect of closure is
to delay the process, as the macrocrack propagation rate
slows down and it takes longer for the crack to become
close enough to the first microcrack.
A more general situation is illustrated in Fig. 4, where
a macrocrack propagates into an array of randomly
distributed nodules with equatorial microcracks, labelled
from A to L. The results obtained allow the extension
of the propagation mechanism of the previous example,

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

FAT I G U E C R A C K G R O W T H I N A U S T E M P E R E D D U C T I L E I R O N

Fig. 2 Effect of crack closure level on the


interaction mechanism between the main
crack and a microcracked graphite nodule.

Fig. 3 Evolution of DK values with load cycles in crackmicrocrack array of Fig. 2.

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

595

596

J. O RT I Z e t a l .

Fig. 4 Evolution of propagation paths for a general nodule array ahead of a long propagating fatigue crack.

as the tips B, D, F and H successively take the role of


macrocrack tip. At the same time, microcracks A, C, E
and G propagate towards the macrocrack tip, to finally
become dormant due to load shielding effects.
Microcracks I, J, K and L do not take part in the main
propagation path; however, they present the same general behaviour as the other microcracks. In this case
more than one microcrack propagates simultaneously
towards the tip of the dominant crack, justifying the
presence of those bifurcations observed during experiments in the encounters of the crack with nodules.5
It is worth mentioning that the observed behaviour of
the macrocrack in propagating intersection of the graphite nodules is not only in accordance with the experimental observations in Ref. [5] but with theoretical results
reported by Petrova et al.16 In this study16 the authors
applied the methods of singular integral equations and
a small parameter method to the problem of a crack and
a system of small holes, and they also observed the
tendency of the crack to propagate towards the damaged region.

Effect of nodule size and distribution


A major feature in the above analyses is that only a
limited number of nodules were included in the models,
whereas nodules are actually distributed over the entire
model domain. The effect of nodule size and distribution
on the propagation mechanism is studied in this section,
through the analysis of the stress fields in the ADI
microstructure. The ratio r/d, was chosen as the characteristic parameter, see Fig. 5(a), where r is the average
nodule radius and d the average minimum distance
between nodule centres. The results of a statistical
analysis of measurements performed on standard ADI
micrographs using image-processing software17 are
shown in Table 1. It can be seen that for a wide range
Table 1 Results of statistical analysis of measurements performed
on standard micrographs
Narea
nodules/mm2

Nodularity
(%)

Average r/d

Standard deviation
(%)

60
100
150
600

100
100
100
100

0.27682
0.26175
0.25294
0.25625

31.18
37.56
35.43
32.62

Fig. 5 BEM model of ADI microstructure:


(a) characteristic dimensions of the problem;
(b) locations of internal-point arrays: (A)
vicinity of nodules, (B) lines oriented
perpendicular to the applied load.

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

FAT I G U E C R A C K G R O W T H I N A U S T E M P E R E D D U C T I L E I R O N

of nodular counts, ranging from 60 to 600 nodules/mm2 ,


the average r/d and its standard deviation are almost
constant, with values r/d=0.25, standard deviation 35%.
BEM models were carried out on a series of randomly
generated geometries with r/d-values ranging from 0.1
to 0.4. A typical model discretization is illustrated in
Fig. 5(b), where schematics with the locations of the
internal-point arrays used for stress computations are
also shown. The locations of the internal points were
selected to evaluate the characteristics of the stress fields
in the vicinity of nodules and along lines orientated
perpendicularly to the applied load [locations A and B
in Fig. 5(b)]. The results are employed in assessing the
effects of nodules on the growth path of microcracks
and macrocracks, respectively.
Analyses of the results show that the average stress
fields in the vicinity of nodules correspond to that of an
isolated hole under remote tension s0 , with standard
deviations increasing with r/d. Results for r/d=0.25 are
shown in normalized form in Fig. 6(a), where the vari-

Fig. 6 (a) Variation with distance to the


nodule surface of the mean value of the
maximum principal stress (bars indicate the
standard deviation); (b) standard deviation
in K values of microcracks as a function of
microcrack length.

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

597

ation of the average value of the maximum principal


stress s1 /s0 is presented together with its standard
deviation as a function of x/r, the normalized distance
from the nodule surface. It can be appreciated how the
standard deviation remains at a constant value of around
10% for distances from the nodule surface shorter than
the average nodule radius (x/r<1), and starts to increase
up to 45% for x/r=2. This behaviour somehow reflects
the extent of the area within the influence of the nodule
stress field. As we move further away from the nodule
surface the effect of the neighbouring nodules becomes
more relevant, leading to larger deviation levels. The
effect of deviations in the stress field on the K-values
was evaluated by a weight function analysis. Stress intensity factor values were calculated through numerical
integration of the weight function for the case of two
cracks emanating from a circular hole18 subjected to the
stress field of Fig. 6(a). The resultant deviation bounds
for K-values as a function of the normalized crack length
c/r are plotted in Fig. 6(b). Results are normalized with

598

J. O RT I Z e t a l .

respect to K0 , the stress intensity factor value corresponding to a crack emanating from an isolated circular
hole and subjected to a remote constant stress field.
Deviation bounds in K-values are around 10%, provided
the microcrack length is less than the nodule radius
(c/r<1). The deviation grows up to 20% for microcrack
lengths c=2r.
As mentioned above, the effect of nodules on macrocracks was studied through the analysis of the stress fields
on lines orientated perpendicularly to the remote applied
load. Figure 7 illustrates a typical distribution of the
maximum principal stress field s1 , along a line of length
equal to 25 nodule distances, x/d=25. The normalizing
parameter for s1 is the mean stress on the line sm .
Segments of the curve with zero stress correspond to
nodule positions. Spectral analysis of the stress distribution was performed using a Fast Fourier Transform
(FFT) methodology. The results are illustrated in Fig. 8
for r/d=0.2, 0.25 and 0.30. It can be seen that there is
a clear periodicity in the stress field, given by the
dominant peaks at T/d=1 for the three cases.
Consequently, the period T corresponds to the average
minimum nodule distance d. It is worth noting that as
r/d increases the dominant peak becomes more marked,
and the noise in the spectral data is reduced. This
behaviour is an indication that as r/d increases, and
consequently a more compact array of nodules is
obtained, the periodicity of the stress field is more
marked.
As with microcracks, the effect of variations in the
stress field on K-values was estimated through a weight

function analysis. The weight function corresponding to


a finite crack of length 2a embedded in an infinitely thin
plate was numerically integrated, with a piece-wise
applied load modelling the periodic stress field. Two
load models were considered, as illustrated in the detail
of Fig. 9. For case I, a constant piece-wise stress is
applied on segments of length equal to the ligament
between nodules, whereas for case II a stress distribution
reproducing the stress concentration at the nodule surface is employed. Computed deviations in K-values with
respect to a crack in an homogeneous material are shown
in Fig. 9 as a function of a/d, for the two approaches. It
can be seen that K levels for a crack in a homogeneous
material are higher than those corresponding to the
same crack embedded in an ADI microstructure, but the
difference reduces as the crack length increases. The
reason for this is that the size of the nodules is diminishing with respect to the size of the crack. In order to
ensure a difference of less than 10%, crack lengths
2a>40d are necessary. Deviations can be reduced to 7%
if crack lengths 2a>100d are considered.
FRACTURE MECHANICS MODEL FOR CRACK TIP
SHIELDING DUE TO MICROCRACKING

Continuum approach
Microcracking in regions near the tip of a crack can
have a shielding effect on the crack tip, redistributing
and reducing the average near-tip stresses. There are
two sources for the redistribution of stresses in the near-

Fig. 7 Distribution of maximum principal


stress along a line perpendicular to the
remote applied load direction (path B in
Fig. 5).

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

FAT I G U E C R A C K G R O W T H I N A U S T E M P E R E D D U C T I L E I R O N

599

Fig. 8 Spectral analysis of the stress field


for different r/d values.

tip stress field. One is due to the reduction in the


effective elastic moduli resulting from microcracking.
The other is the strain arising from the release of
residual stresses when microcracks are formed. The
residual stresses in question could develop in the solidification process due to thermal mismatches between
phases and thermal anisotropy in the material. A study
of the shielding effect resulting from the reduction of
elastic moduli due to microcracking in ADI is conducted
in the following paragraphs, in the spirit of the work
carried out by Hutchinson.19 The effect of the release
of residual stress is discussed later.
When a continuum approach is considered, it is
assumed that a typical material element contains a cloud
of microcracks. The strainstress behaviour of the

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

element is obtained as an average over many microcracks.


It will be also assumed that microcracking ceases or
saturates above some applied stress ss . Thus, it is tacitly
assumed that there exists a zone of nominally constant
reduced stiffness surrounding an even smaller fracture
process zone within which the microcracks ultimately
link up. The microcrack region Ac , surrounding the
crack tip has reduced elastic moduli, which are uniform
and isotropic. Such moduli reduction depends on microcrack density and orientation, the former given by
some nucleation criteria. The crack is semi-infinite with
a remote stress field specified by the applied stress
intensity factor K. The near-tip fields have the same
classical form but the stress intensity factor, Ktip , is
different. Following Hutchinson,19 the ratio Ktip /K

600

J. O RT I Z e t a l .

Fig. 9 Deviations of K values for a long crack in an ADI microstructure with respect to a crack in an homogeneous material.

which is a function of the moduli differences and the


shape of Ac , but not on its size:
Ktip /K=f (d1 , d2 , shape of Ac )

(5)

where d1 and d2 are two special combinations of the


moduli:
d1 =

C D

G
1
1
1n G
9

(6)

Application to ADI
Figure 10(a) shows a schematic of the problem, when
applied to stationary and steadily growing cracks in ADI.
The shape of Ac results from considering that microcracks nucleate with no preferred orientation, and that
nucleation occurs when the maximum principal stress s1
reaches a critical tensile value sc , i.e.
Microcrack density b=0 (s1 )max <sc
b>0 (s1 )max sc

and
d2 =

1
G
n: n
1n
G
9

(7)

where G
9 is the effective shear modulus of the microcracked zone and n: is the effective Poissons ratio. It
is of interest to give some explicit deduction from Eq. (5).
In this sense the following result is exact to the lowest
order in d1 and d2 :

B A

5
3
Ktip
d1 + k2 +
d2
=1+ k1
K
8
4

(8)

where k1 and k2 are functions that depend on the shape


of Ac only. It is worth noting that Eq. (8) also holds for
some non-uniform microcrack distributions. In particular, of interest for this work is the case in which the
elastic moduli vary continuously within Ac , from their
saturation values in the close neighbourhood of the
macrocrack tip to those of the uncracked material.

(9)

where (s1 )max stands for the maximum value attained


over the history.19
As stated in the previous section, the computation of
crack tip shielding depends on the reduced moduli
parameters G
9 and n: , which are functions of the crack
density parameter b. A convenient measure of microcrack
density has been suggested by Budiansky and
OConnell:20
b=

G H

2M A2
p
P

(10)

where M is the number of embedded cracks per unit


volume, A is the area of a crack and P is its perimeter.
If, as illustrated Fig. 10(b), it is assumed that pennyshaped microcracks with radius rc nucleate from nodules,
the number of cracks per unit volume can be computed
using
M=aNvol

(11)

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

FAT I G U E C R A C K G R O W T H I N A U S T E M P E R E D D U C T I L E I R O N

601

Fig. 10 (a) Zones of microcracked material


Ac for stationary and steadily growing cracks
with nucleation criteria given by maximum
principal stress (from ref [18]); (b) a pennyshaped microcrack emanating from a
nodule; (c) a penny-shaped microcrack
under annular stress.

where Nvol stands for the nodule count per unit volume
(nod/mm3 ) and a nucleation parameter 0a1
accounts for the fact that only a fraction of nodules
actually nucleate cracks. Following Hilliard21 volume
and area nodule counts are related through the ratio
between the average nodule cross-section area and the
nodule volume, A
9 nod /V
9 nod , which results in
Nvol =Narea

A
pr2
3 Narea
9 nod
=Narea 4 3 =
V
pr
4
r
9 nod
3

(12)

Equation (12) can be further simplified since nodule


counts and r are not independent, as the volume fraction
of graphite nodules must remain constant with the
nodular count. In the case of ADI with a typical carbon
content of 3%, the volume fraction of graphite nodules
is around 9%. Thus,

A B

3 Narea 4 3
pr =Narea r2 p=0.09
Nvol V
9 nod =
4 r
3

M=a

5
3/2
p N area
2

(14)

The formation of microcracks from the nodules


increases the compliance of the material within Ac ,
which experiences an increase in strain due to the volume
of the opened microcracks. For each microcrack this
extra volume can be assimilated to that of a pennyshaped microcrack under annular stress distributions [see
Fig. 10(c)]. Thus, the increase in strain due to a component of stress acting normal and tangential to the plane
of the microcrack can be derived as in Ref. [19] from
results given in Tada et al.22
Denn =

(13)

which together with Eq. (12) allows one to rewrite

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

Eq. (11) as

Dent =

C A BD

snn
V E

C A BD

snt
V E

r
16
(1n2 ) 1
3
rc

r
16 (1n2 )
1
3 (2n)
rc

3/2 3
rc

3/2 2
rc

(15)

(16)

602

J. O RT I Z e t a l .

where V stands for unit volume. Note that these contributions to the strain are also based on the assumption
that the interaction between a microcrack and its neighbours can be ignored. Results in Eqs (15) and (16) can
be assimilated to those obtained in Ref. [19] for a pennyshaped crack under constant load provided an equivalent
microcrack radius is defined as
req =rc3

SC A B D
r
1
rc

(17)

2
Considering now that A=preq
and P=2preq together
with the result in Eq. (14), then Eq. (10) for the crack
density parameter b yields

b=

A B SC A B D

rc
27
a
400p
r

r
rc

(18)

which depends on a and rc /r only. Finally, the reduced


moduli parameters G
9 and n: within Ac can be obtained
from
32 (1n)(5n)
G
=1+
b
G
45
(2n)
9
V
9 =1

16n (3n)(1n2 )
b
15
(2n)

(19)
(20)

These lower estimates, which ignore microcrack inter-

action, agree with the dilute limit estimates that approximate interaction.20
The values of k1 and k2 for the situations depicted in
Fig. 10, when microcracks nucleate at a critical maximum
normal stress, have been obtained in Ref. [19] which
after substitution in Eq. (8) results in

A B
A B
Ktip
K

stationary

Ktip
K

growing

=10.547d1 +0.674d2
=10.673d1 +0.822d2

(21)

(22)

for the stationary and steadily growing crack, respectively. A related result worth considering corresponds
to anisotropic microcracking, in which microcracks
nucleate perpendicular to the maximum principal stress
s1

A B
Ktip
K

anisotropic

1
1+5.83b

(23)

Equations (21) and (22) give the results shown in


Fig. 11, wherein Ktip /K are plotted against the microcrack length rc /r for several values of the nucleation
parameter a. According to experimental data n=0.28
was employed.5 Maximum microcrack length was considered to be equal to half the average distance between

Fig. 11 Shielding effect Ktip /K for


stationary and steadily growing macrocracks
due to isotropic microcracking for different
values of the nucleation parameter a.

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

FAT I G U E C R A C K G R O W T H I N A U S T E M P E R E D D U C T I L E I R O N

nodules as rc /r=2. According to results from Effect of


nodule size and distribution, above, this length corresponds to the position of the lowest stress at the microcrack tip. Larger microcracks could not become
dormant. It can be observed that Ktip /K is reduced by
increases in microcrack lengths. Shielding is 40% larger
in the case of the growing crack. Results for the stationary crack, when the microcracks are nucleated only
perpendicular to the maximum principal stresses given
by Eq. (23) are plotted in Fig. 12. These results present
the same general behaviour as those in Fig. 11; however,
the increase in shielding is larger by a factor of two
when compared with the isotropic case.
DISCUSSION

There is growing evidence of the shielding effect that


microcracking has on the tip of a dominant macrocrack,
redistributing and reducing the average near-tip
stresses.16,19,23 For this mechanism to operate it is essential
that the microcracks arrest and be highly stable in the
arrested configuration. Ultimately the dominant crack
advances though a mechanism of interaction and coalescence of the microcracks. The existence of such a mechanism
in ADI for fatigue crack growth was found in experimental
observations5 and corroborated through numerical modelling in Numerical modelling, above. The estimates of the
deviation ranges for DK-values due to the nodules show

Fig. 12 Shielding effect Ktip /K for


stationary macrocrack due to anisotropic
microcracking for different values of the
nucleation parameter a.

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

603

the dependence of the propagation mechanism upon the


load level. For applied loads that induce DKtip levels close
enough to DKth , deviation ranges can easily situate DK
levels under the propagation threshold, inducing the crack
growth mechanism to stop. On the other hand, as the
difference between the applied DK and DKth increases, the
effect of deviation bounds in DK is less important. Although
microcracked nodules could affect the general propagation
rate, the probability of the mechanism to stop diminishes.
In the near threshold region the applied load amplitudes
are small, and the probability of the initiation of microcracks starting from the nodules is lower. This was verified
in fractographic observations, which showed a decrement
in the frequency of appearance of crack bifurcations at low
levels of applied stress intensity factors DK.5
The continuum approach of the previous section provides a means to quantify the shielding effect on the
macrocrack tip. It is worth noting the important role the
shielding effect has on the propagation rate, since da/dN
varies approximately with the cube of DK. If for example
DKtip /DK=0.9 the propagation rate decreases 30%,
whereas for DKtip /DK=0.8 it halves. Although calculations
in the previous section may not yield actual numbers for
real cases, they indicate how the main factors affect the
crack growth of the macrocrack. Results show that the
shielding contribution is larger for steadily growing cracks
and for anisotropic nucleation of microcracks. At issue are
both the stress dependence of nucleation and the orien-

604

J. O RT I Z e t a l .

tation distribution of microcracks, as can be seen from the


comparison of the results in Eqs (21) and (22). In this way
more observational data such as in Ref. [5] and a deeper
understanding of the microcracked zone are needed in
order to obtain accurate quantitative estimates of the shielding effect.
As pointed out in previous sections, an extra contribution
of the microcracks to the shielding effect arises from the
release of residual and transformation stresses when microcracks are formed. This contribution to the macrocrack
tip shielding was not considered in the analysis of the
previous section as static residual and transformation
stresses would not alter the fatigue propagation rate.
However, a significant contribution could arise from these
phenomena through the augmentation of closure levels.
The relevance of closure in delaying the propagation mechanism was shown in the section Numerical modelling
above. Although it would be difficult to quantify, closure
of the highly branched fatigue cracks in ADI can be
increased by mechanisms related to their large surface
roughness and the relatively large mode II crack opening
displacements that are generated in the inclined crack
segments between nodules. At the same time, theoretical
results exist that consider microcracking itself to be a cause
of closure. Results reported by Petrova et al.16 show that
under certain loading conditions and depending on the
configuration of the microcracks, the macrocrack could be
fully or partially closed, whereas microcracks could become
fully open or fully closed. Therefore, although unknown
beforehand, contact zones could emerge in the process of
the crack growth.
Finally, it is important not to forget the three-dimensional nature of the problem. Long and short cracks have
different behaviours due to the relative size of their crack
fronts. The effect of irregularities is more attenuated for a
wide front than for a reduced one. Therefore, it can be
said that the increment in the applied DK that was calculated for the main crack tip exerts only a local effect that
would be almost imperceptible if only the presence of a
single nodule next to the crack front is considered. The
propagation rate of the main crack corresponds to an
average of all the phenomena that simultaneously take
place along the whole extension of the crack front.
CONCLUSIONS

The properties of ADI are the consequence of its matrix


microstructure. Fatigue crack growth paths preferentially
intersect graphite nodules, and microcracking at graphitematrix interfaces takes place around the tip of the
macroscopic crack. Ultimately the macroscopic main
crack advances by interaction and coalescence with some
of the microcracks, which successively take the role of
the macrocrack tip. Yet some other microcracks propa-

gate towards the macrocrack tip but do not take part in


the main propagation path, and become dormant due to
load shielding effects. This mechanism explains the
relatively low propagation rates and high effective propagation threshold values for this material. Numerical and
fracture mechanics modelling of the above micromechanism have been provided in this paper, based on the
BEM and using a continuum approach.
The effects of nodule size and distribution on the
propagation mechanisms were considered by assessing
the deviation bounds in DK levels for cracks and microcracks, through the analysis of the stress fields
in the ADI microstructure. A clear periodicity in the
stress field was found in the case of long cracks, resulting
in applied DK levels lower than those in a continuous
material. On the other hand, the average DK for microcracks emanating from graphite nodules corresponds to
that of a microcrack emanating from an isolated hole.
Redistribution of stresses in the near-tip region due to
microcracks emanating from graphite nodules was correlated to the reduction in the effective elastic moduli
using a continuum approach. Stationary and steadily
growing cracks in ADI were considered, with isotropic
and stress-orientated microcrack clouds. Load shielding
effects on the applied DK at the main crack tip were
found to be more important in the case of a steadily
growing crack, and when the microcracks are considered
to nucleate perpendicular to the maximum principal
stresses. The continuum approach allowed a qualitative
description of the shielding effect, although a deeper
understanding of the microcracked zone is needed in
order to obtain accurate quantitative estimates. In this
sense the release of residual and transformation stress
constitute interesting topics to be addressed, as they
could contribute to enhance closure levels of the main
crack, thus further delaying the propagation mechanism.
Acknowledgements
This work was financed by grant PICT 1204586 of
Agencia Nacional de Promocion Cientfica de la Republica
Argentina. The authors wish to thank CONICET and the
Organization of American States (OAS) for additional
funding. The authors are also grateful to Professor
M. H. Aliabadi for providing DBEM software and
Dr G. Rivera and M. Chapetti for useful discussions.

REFERENCES
1 Authors. (1990) Ductile iron data for design engineers, Ductile
Iron Group of QIT-Fer & Titane, Canada.
2 Gundlach, R. B. and Janowak, J. F. (1991) A review of austempered ductile iron metallurgy. In: Proceedings AFS (American
Foundrymens Society, Inc.) 1991 World Conference on Austempered

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

FAT I G U E C R A C K G R O W T H I N A U S T E M P E R E D D U C T I L E I R O N

6
7
8

10

11
12

13

14

15

16

17
18
19
20
21

22
23

Ductile Iron, (Edited by the American Foundrymens Society, Inc).


American Foundrymens Society, Des Plains, Illinois, pp. 110.
Faubert, G. P., Moore, D. J. and Rundman, K. B. (1991) ADI
Part I. microstructure and mechanical properties of high alloy
heavy section castings. In: Proceedings AFS (American Foundrymens
Society, Inc.) 1991 World Conference on Austempered Ductile Iron,
(Edited by the American Foundrymens Society, Inc). American
Foundrymens Society, Des Plains, Illinois, pp. 1016.
Kobayashi, T., Yamamoto, H. and Yamada, S. (1991) On the
toughness and fatigue properties of austempered ductile cast
iron. In: Proceedings AFS (American Foundrymens Society, Inc.)
1991 World Conference on Austempered Ductile Iron, (Edited by the
American Foundrymens Society, Inc). American Foundrymens
Society, Des Plains, Illinois, pp. 567577.
Greno, G. L., Otegui, J. L. and Boeri, R. E. (1999) Mechanisms
of fatigue crack growth in austempered ductile iron. Int. J.
Fracture 21, 3543.
Aliabadi, M. H. (1997) Boundary element formulations in
fracture mechanics. Appl. Mech. Rev. 50/2, 8396.
Aliabadi, M. H. and Rooke, D. P. (1994) Numerical Fracture
Mechanics. Kluwer Academic Publishers, London, UK.
Portela, A., Aliabadi, M. H. and Rooke, D. P. (1993) Dual
boundary element incremental analysis of crack propagation.
Computers Struct. 46/2, 237247.
Cisilino, A. P., Aliabadi, M. H. and Otegui, J. L. (1998) Energy
domain integral applied to solve centre and double-edge crack
problems in three dimensions. Theoret Appl. Fracture Mech.
29, 181194.
Ortiz, J., Cisilino, A. P. and Otegui, J. (2001) Boundary element
analysis of fatigue crack propagation micromechanisms in ductile iron. Engng Anal. Boundary Elem. 25, 467473 (in press).
Sih, G. C. (1991) Mechanics of Fracture Initiation and Propagation.
Kluwer Academic Publishers, Dordrecht.
Klesnil, M. and Lukas, P. (1972) Influence of strength and stress
history on growth and stabilisation of fatigue cracks. Engng.
Fract. Mech. 4, 7792.
Elber, W. (1971) The significance of fatigue crack closure. In:
Damage Tolerance Aircraft Structures, ASTM STP 486. American
Society of Testing and Materials, Philadelphia, PA, pp. 230247.
Leis, B. N., Hoper, A. T. and Ahmad, J. (1986) Critical review
of the fatigue growth of short cracks. Engng. Fract. Mech.
23, 883898.
E 647-88, Standard Test Method for Measurements of Fatigue Crack
Growth Rates (1988) Annual Book of ASTM Standards,
Section 3, Metals Test Methods and Analytical Procedures,
American Society of Testing Materials, Philadelphia, 1988.
Petrova, V., Tamuzs, V. and Romalis, N. (2000) A survey of
macromicrocrack interaction problems. Appl. Mech. Review
53, 117146.
Image-Pro Plus. Media Cybernetics Inc., USA.
Wu, X. R. and Carlsson, A. J. (1991) Weight Functions and Stress
Intensity Factor Solutions. Pergamon Press, Silver Spring, MD.
Hutchinson, J. W. (1987) Crack tip shielding by micro-cracking
in brittle solids. Acta Metallurgica 35, 160519.
Budiansky, B. and OConnell, R. J. (1976) Elastic moduli of a
cracked solid. Int. J. Solids Structures 12, 81.
Hilliard, J. E. (1968) Measurement of volume in volume. In:
Quantitative Microscopy, (Ed. by R. T. DeHoff and F. N. Rhines).
McGraw-Hill Book Company, New York, pp. 4576.
Tada, H., Paris, P. C. and Irwin, G. R. (1985) Handbook for
Stress Analysis of Cracks. Del Research, St Louis, MO.
Laws, N. (1992) The effect of microcracks on energy density.

2001 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 24, 591605

605

In: Mechanics and Physics of Energy Density. (Edited by G. C. Sih


and E. E. Gdoutos), Kluwer Academic Publishers, Dordrecht,
pp. 195201.

APPENDIX

Boundary element modelling considerations


The mathematical degeneration of the boundary element
method when applied to crack problems, where the two
crack surfaces are considered coplanar is well known.7
Some special techniques have been devised to overcome
this difficulty. Among these, the most general is the dual
boundary element method (DBEM) introduced by
Portela et al.8 that has been used in this work. The
DBEM incorporates two independent boundary integral
equations, with the displacement equation applied for
collocation on one of the crack surfaces and the traction
equation on the other.
In the absence of body forces and assuming of the
displacements at a boundary point x, the boundary
integral representation of the displacement components
uj is given by
cij (x)uj (x)+

Tij (x, x)uj (x) dC=

Uij (x, x)tj (x) dC

(A1)

where i and j denote Cartesian components; Tij (x, x)


and Uij (x, x) represent the Kelvin traction and displacement fundamental solutions, respectively, at a boundary
point x.
Assuming continuity of both strains and traction at x
on a smooth boundary, the traction components tj are
given by:
1
tj (x)+ni (x)
2
=ni (x)

Tkij (x, x)uk (x) dC

Ukij (x, x)tk (x) dC

(A2)

where Tkij (x,x) and Ukij (x,x) contain derivatives of


Tij (x,x) and Uij (x,x), respectively, and ni (x) denotes the
component of outward normal to the boundary at x.
The boundary integral equations [Eqs (A1) and (A2)]
constitute the DBEM.
For the sake of efficiency, and to keep the simplicity
of the standard boundary element, the DBEM uses
discontinuous quadratic elements for the crack modelling, whereas the remaining boundary of the model is
discretized using continuous elements. This simple strategy is robust and allows the DBEM to effectively model
general edge or embedded crack problems; crack tips,
crack-edge corners and crack kinks do not require special
treatment, as they are not located at nodal points where
collocation is carried out.

Anda mungkin juga menyukai