Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

The effect of porosity on the fatigue life of cast aluminium-silicon

alloys∗
Y. X. GAO, J. Z. YI, P. D. LEE and T. C. LINDLEY
Department of Materials, Imperial College London, Prince Consort Road, London SW7 2BP, UK

Received in final form 18 February 2004.

A B S T R A C T Fatigue strength optimization of cast aluminium alloys requires an understanding of the


role of micropores resulting from the casting process. High cycle fatigue tests conducted
on cast A356-T6 show that the pore size and proximity to the specimen surface significantly
influence fatigue crack initiation. This is supported by finite element analyses (both elastic
and elastic–plastic) which demonstrate that high stress/strain concentration is induced by
pores which are both large and near to the specimen surface. A new pore-sensitive model
based on a modified stress-life approach has been developed which correlates fatigue life
with the size of the failure-dominant pore. The model prediction is in good agreement
with experimental data.

Keywords casting porosity; cast aluminium alloy; fatigue life prediction; finite element
analysis; stress/strain concentration.

NOMENCLATURE Ap = projected area of pore


α = scaling factor of pore effect
b = Basquin exponent
d 0 = diameter of specimen cross-section
d e = equivalent diameter of pore
ε/ε∞ = normalized maximum principal strain
εlocal , ε ∞ = local strain, far-field nominal strain
k0 , k1 , k2 = stress/strain concentration parameters
kε = strain concentration factor
kt , kσ = stress concentration factor in elastic case, in elastic–plastic case
kg = geometric mean values of kσ and kε
λ2 = secondary dendrite arm spacing
N f = fatigue life
s = proximity of pore to the specimen surface
σ /σ ∞ = normalized maximum principal stress
σ local , σ ∞ = local stress, far-field nominal stress
σ a, σ ∞a = stress amplitude, far-field nominal stress amplitude
σ f = fatigue strength coefficient

INTRODUCTION
minium alloys must be understood in order to optimize
Microstructure and defect population are important fac-
the fatigue performance of these alloys.
tors which can strongly influence the fatigue properties of
Because of the potential weight reduction and conse-
a material. The roles of casting porosity, silicon particles
quent improvement in fuel economy of vehicles, cast alu-
and secondary dendrite arm spacing (SDAS) in cast alu-
minium alloys are being used to replace traditional cast
iron and forged steel components in the automotive and
Correspondence: T. C. Lindley. E-mail: t.lindley@imperial.ac.uk aerospace industries. However, the relatively low fatigue

This paper is dedicated to the memory of Professor H. M. Flower. resistance of these cast alloys is known to be an obstacle


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570 559
560 Y. X . G AO et al.

in terms of cost-effective design and safety. Although fa- <0.02Zn, <0.01Ti and <0.01B. A wedge- shaped perma-
tigue behaviour depends in general on a number of me- nent mould was employed to generate varying solidifica-
chanical and microstructural factors,1,2 it is the presence of tion rates from the top to the bottom of the mould in
porosity, introduced by a combination of dissolved gas and casting. During the casting, a 0.02 wt% Sr modifier was
shrinkage, and other defects formed during the solidifica- added as Al-10 wt% Sr master alloy, together with 0.1 wt%
tion process that are mainly responsible for the reduction Ti grain refiner as Al-5Ti-B master alloy. The melt was
in fatigue resistance of cast alloys.1,3–12 Many experimen- first degassed and then a Ar +5% H2 mixture was bubbled
tal studies have shown that fatigue cracks are predom- through it to produce castings with a controlled hydrogen
inantly initiated from micropores7,10,13–16 and the most content. In all five casts produced under these conditions,
common finding was that the pores responsible for fatigue the dimensions were approximately 200 mm high, 150 mm
crack nucleation are generally located at or very near the in length and with a thickness of 45 mm at the top and
specimen surface. Seniw et al.13 examined the effect of the 13 mm at the bottom. All the wedge casts were subjected to
size, morphology and location of micropores on the fa- a standard T6 heat treatment directly after casting. Twelve
tigue properties of cast A356 aluminium by experimental specimens were cut from each wedge and divided into
observation, and their results indicated that fatigue cracks two groups with similar microstructure and defect popu-
initiate from pores which are large and are located close to lation. One group was cut from the top part of the wedge
the surface of the specimen. Stanzl-Tschegg et al.14 found where the secondary dendrite arm spacing (λ2 ) was found
that casting pores are the main factor which influences the to be in the range of 43–52 µm. The other group was
number of cycles leading to crack initiation in AlSi7Mg cut from the bottom wedge location where λ2 was in the
cast alloys. Caton15 examined the fracture surface of cast range 19–27 µm. They are henceforth denoted as group
W319-T7 aluminium fatigue specimens to identify the St and Sb, respectively. Further details about the cast-
crack initiation sites, and found similar results to those ing process and cutting position and orientation of spec-
reported by Seniw et al.13 Boileau16 also found a similar imens from the wedges can be found from our previous
situation in cast W319 and reported a scenario in which work.21
fatigue cracks initiated at the small pores located at the Light microscopy and scanning electron microscopy
surface rather than at larger pores located in the specimen (SEM) were used to characterize the material microstruc-
interior. tures and defects. High cyclic fatigue tests with sinusoidal
It is also generally known that stress/strain concentra- loading were carried out using an Amsler Vibrophore
tions can have a considerable influence on fatigue fail- (Dynamic Testing GmbH&Co. KG, Gottmadingen,
ure and the initiation and growth of a fatigue crack at Germany) fatigue machine following the standard testing
stress-concentration sites has long been a topic of prac- procedure of ASTM-E466.22 The applied load frequency
tical importance.1,2,17–20 In the present work, the role was approximately 70 Hz. The circular section specimens
of casting pores as micro stress concentrators is con- with a diameter of 5 mm were tested at a stress ratio of
sidered in aiding the evolution of fatigue damage. In 0.1, with a maximum stress of 120 MPa for group St spec-
order to characterize the microstructure and porosity, imens and 150 MPa for group Sb . These stress levels were
and to examine the fatigue life in relation to the po- selected based on the experimental S–N data previously
tency of fatigue crack initiators, experiments were carried established for the alloy. The fatigue lives of the specimens
out on cast A356-T6 alloy. Experiments were comple- were measured and the fracture surfaces were examined
mented by finite element analysis (FEA) for the config- under SEM to identify the defects responsible for fatigue
urations of both typical and idealized micropores so as crack initiation.
to estimate the degree of stress/strain concentration at
pores in relation to pore size and location. A new pore-
sensitive model was then developed which correlates fa- Microstructure and defects
tigue life with the size and location of the failure-dominant Metallographic examination showed that the basic con-
pore. stituents of the cast A356 comprise a primary Al ma-
trix together with an Al+Si eutectic located between the
secondary dendrite arms. The eutectic regions were also
E X P E R I M E N TA L S T U DY the favoured location for defects such as porosity. Fig-
ure 1(a) shows a light micrograph which gives an overview
Material and test method
of the microstructure and porosity, while Fig. 1(b)–(d) are
The material used in this work is a cast Al-7Si-Mg al- SEM images showing the detail of the microstructural
loy, designated A356, with composition (wt%) 7.25Si, constituents associated with porosity. Figure 2 shows the
0.32Mg, 0.06Fe, <0.01Cu, <0.01Mn, <0.01Cr, <0.01Ni, shape and spatial distribution of the micropores from (a)


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
THE EFFECT OF POROSITY 561

Fig. 1 Typical microstructure and porosity (a) Light micrograph overview; (b)–(d) details of the microstructural constituents and pores of
different shape found using the SEM. Note: A, pores; B, Si particles; C, secondary dendrite arm spacing and D, Al matrix.

optical imaging and (b) after image analysis to distinguish and (c) give details around the failure-dominant pores.
the pores. The pores are highly irregular in shape and The equivalent diameters of the failure-dominant pores
randomly distributed. The size of the pores is broadly in together with the fatigue lives of the specimens are listed
the range of several tens to hundreds of microns. To mea- in Table 1. The corresponding stress level and ratio ap-
sure the size of an irregular-shaped pore for the purpose plied in the fatigue tests are also presented in Table 1. The
pore of diameter d e is
of further modelling, an equivalent  dependence of the fatigue life on the equivalent diameter
used which can be defined as d e = 4 Ap /π with Ap as the of the failure-dominant pore is presented in Fig. 4.
projected area of the pore measured directly from met- In summary, the present experimental study confirms re-
allographic samples or fatigue fracture surfaces.21,23 The sults of previous studies showing that the micropores in-
equivalent diameters of the pores responsible for fatigue troduced by the casting processes are responsible for ini-
crack initiation were measured from the fatigue fracture tiating fatigue cracks and that the failure-dominant pores
surfaces and are in the range of 276–914 µm for the spec- are large and close to, or at the surface of the speci-
imen group St and 120–621 µm for Sb (Table 1). mens. A physically based model to explain the effect of
these pores on the fatigue failure mechanism will now be
developed.
Crack initiation site and fatigue life
The fatigue crack origins were identified from SEM stud-
F I N I T E E L E M E N T A N A LY S I S
ies of the fracture surfaces. As expected, under the present
casting and testing conditions all the fatigue cracks ini- As noted in the Introduction, since the fatigue damage
tiate from pores, most of them from large pores located and failure mechanisms of casting alloys are dependent
at, or close to, the specimen surface. These are termed on the local stress/strain concentration at the micro-
the failure-dominant pores. Similar results have been re- pores, finite element analysis (FEA) was carried out to in-
ported in studies of other cast aluminium alloys.13–16 A vestigate the relation between pore and associated local
typical fatigue fracture surface highlighting the pore as stress–strain fields. The commercial FEA software
the crack initiator is shown in Fig. 3(a), while Fig. 3(b) ABAQUS (ABAQUS, INC. Pawtucket, RI, USA)


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
562 Y. X . G AO et al.

Table 1 Fatigue life N f versus the pore equivalent diameter d e


(µm) and the corresponding stress amplitude σ ∞a (MPa) and stress
ratio R

St Sb
σ∞
a = 54, R = 0.1 σ∞
a = 67.5, R = 0.1

N f (×105 ) de N f (×105 ) de

1.08 914 1.01 621


1.28 813 1.25 358
1.38 548 1.33 487
1.61 732 1.41 380
1.76 629 1.42 418
1.80 644 1.46 490
1.91 686 1.67 351
2.18 534 1.69 405
2.72 590 1.75 414
2.74 579 1.78 319
2.78 569 1.95 356
2.95 506 2.03 232
3.29 380 2.05 328
3.29 338 2.16 322
3.45 388 2.18 335
3.51 406 2.22 358
3.53 353 2.24 328
3.55 328 2.25 239
3.68 474 2.25 339
3.71 355 2.40 251
Fig. 2 Morphology of pores in A356 (a) Light micrograph and (b)
3.74 383 2.70 270
after image analysis, showing that the pores are highly irregular in
3.76 321 2.76 256
shape and randomly distributed.
3.93 329 3.09 193
4.13 278 3.22 199
(ABAQUS/CAE Version 6.2) was used. The material is 4.17 325 3.36 201
assumed to be an elastic or elastic–plastic isotropic solid 4.38 352 3.41 258
with a Young’s modulus of 70 GPa and a Poisson’s ratio of 4.86 276 5.48 156
5.69 199
0.3. Experimentally determined values of yield stress were
7.63 120
used of 223 MPa for the specimen group St and 266 MPa
for Sb . The region surrounding the pores is modelled with
a refined mesh so as to increase the accuracy of the cal-
culation (see Fig. 5). Solutions for the local stress/strain For the purpose of the study of the effects of location
distribution around pores were calculated and the results and size of the pores and regardless of their irregular
were presented in terms of normalized maximum princi- shape, the pore is idealized as a sphere of equivalent
pal stress (σ /σ ∞ ) and strain (ε/ε ∞ ) where the superscripts diameter.
∞ represent the nominal far-field values. The stress and The stress/strain concentration fields have been calcu-
strain concentration factors due to a pore are denoted by kt lated for two irregularly shaped pores (Fig. 5) located re-
and kε respectively. When only elastic deformation occurs motely from the surface and subject to a far field tension
around the pores, kt and kε have the same value. But they stress in the horizontal direction. Figure 5 shows SEM im-
take different values once local plastic deformation oc- ages of the pores together with the corresponding global
curs when kσ is used to represent the stress concentration and refined meshes from FEA simulation. In these FEA
factor instead of kt . The influence of the size of casting simulations, two-dimensional plane strain is assumed with
pores and their shape and location in the specimen, on a far-field stress of 120 MPa and a yield stress of 223 MPa.
the stress/strain concentration factors is explored in the Figure 6(a) and (b) show the distribution of the normal-
following section. ized stress found by elastic analysis for the pore shapes
given in Fig. 5(a) and (b). Figure 6(c) shows the normalized
Idealization of pore shape
in-plane principal stress assuming elastic–plastic be-
Since the casting pores are highly irregular in shape (see haviour while Fig. 6(d) gives the equivalent plastic strain.
Figs 1 & 2) a simplified shape is used in the modelling. The calculated elastic stress concentration factors are 6.2


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
THE EFFECT OF POROSITY 563

Fig. 4 The dependence of the fatigue life on the equivalent


diameter of the failure dominant pore.

It must be noted that the stress concentration around a


pore is certainly dependent on the geometric shape of that
pore, particularly its local radius in elastic stress analysis
where ‘sharp’ features can cause a very high stress concen-
tration [see, Fig. 6(a) & (b)]. The apparently small effect
of irregular pore shapes on fatigue life can be rationalized
as a consequence of the maximum local stresses being de-
termined by ‘sharp’ pore features and being limited by
local plastic deformation. Figure 6(d) demonstrates the
equivalent plastic strain around the sharp features of the
pore shown in Fig. 5(a) while Fig. 6(c) gives the normal-
ized maximum in-plane principal stress with respect to
the far-field stress when local plastic deformation occurs.
It is clear that the local plastic deformation reduces the
stress concentration factor dramatically, from 6.2 to 2.1.
It is generally accepted that micromechanisms of fatigue
crack initiation involve local plastic deformation. Hence,
although the irregular pore shape is idealized as a sphere
in this study in order to evaluate the effects of location and
Fig. 3 Fracture surfaces and crack initiating pores (a) SEM image
size of pore, a further study is being undertaken which in-
of a typical fracture surface highlighting the crack initiation pore
(A), (b) and (c) two typical initiation pores and the detailed fracture corporates micro-plasticity effects in the analysis of the
surfaces around them. role of pore shape.

Effect of pore location and size


and 6.4 for the two pores analysed which is a small differ- Figure 7 is a schematic diagram showing (a) the geometric
ence in view of the very different shapes. This is consistent configuration of a specimen with an idealized pore close to
with other recent studies23,24 where it was reported that the specimen surface and (b) a cross-section showing the
the shape of pores and inclusions has a negligible effect dimensions of the specimen and the pore. The equivalent
on fatigue performance in cast aluminium alloys. Such ob- diameter d e and the proximity of pore to the specimen
servations support the assumption of idealization of pore surface denoted by s are employed to characterize the size
shape adopted in the present model. and location of the pore as shown in Fig. 7(b). Because


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
564 Y. X . G AO et al.

Fig. 5 (a), (b) SEM images of two typical pores of irregular shape. (Note: edges of pores are outlined.) (c) The global mesh and (d) local
refined mesh used in the two-dimensional FEA simulation.

of symmetry, only a quarter of the specimen is required Results from Table 2 and Figs 9, 10 and 11 clearly show
for FEA simulation as shown in Fig. 8(a). Figure 8(a) and that the closer the pore is to the specimen surface, the
(b) illustrate the global mesh of the domain and the re- higher the stress concentration. A pore that intersects the
fined mesh around the pore respectively. Figure 9 gives surface induces the highest stress concentration. It is also
the FEA result for a pore of size d e = 600 µm and loca- apparent that for pores either at the centre of the speci-
tion s = 10 µm, as an example to illustrate the distribution men, or far removed from the specimen surface, the pore
of the normalized maximum principal stress around the size over the range examined has a negligible effect on
pore. It is clear that the highest stress occurs in the nar- the stress concentration factor. In contrast, for pores ei-
row region between the pore and the specimen surface ther near or at the surface, the stress concentration factor
and here the stress concentration factor is 3.61. The ef- increases significantly with the pore size. These conclu-
fects of pore location and size were also examined using sions are generally in agreement with the experimental
ABAQUS. Figure 10 illustrates the effect of pore location observation that the pores responsible for crack initia-
in terms of the normalized maximum principal stress at a tion are located at or close to the specimen surface (see
pore with d e = 800 µm, located at the centreline of the Fig. 3). Since both the FEA and the experimental results
cylindrical specimen (s = 2100 µm) [Fig. 10(a) & (b)], and indicate that the large pores close to the specimen sur-
for the same size pore close to the specimen surface (s = face are the defects most likely to cause crack initiation
10 µm) [Fig. 10(c) & (d)]. For cases where the pore inter- in the present study, they are designated failure-dominant
sects the specimen surface (s becomes negative) the FEA pores.
simulation shows that the stress concentration is higher
than that for pores close to the surface. Figure 11 shows
the stress concentration for one of these cases of pore in-
Local plastic strain around failure-dominant pores
tersection, with s = −20 µm and d e = 600 µm. The FEA
solutions for the stress concentration factor kt are listed From Table 2, it can be seen that the local stress con-
in Table 2 for all the cases examined. centration at a failure-dominant pore is sufficiently high


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
THE EFFECT OF POROSITY 565

Fig. 6 Distribution of the maximum in-plane principal stress normalized with respect to the far field stress acting in a horizontal direction,
given by (a), (b) elastic and (c) elastic–plastic analyses, respectively. Fig. 6(d) shows the equivalent plastic strain around the pore
corresponding to Fig. 5(a) under far-field stress σ ∞ = 120 MPa and yield stress σ y = 223 MPa.

so as to exceed the yield stress. It is therefore necessary


to incorporate the local plastic strain in any model aimed
at correlating the pore-size effect quantitatively with fa-
tigue life. Assuming that the material behaves in an elastic–
perfectly plastic manner and taking a worst case scenario
of a pore with a fixed location very close to the specimen
surface (s = 10 µm), the local plastic deformation at a
pore is given in Fig. 12(a). The detailed equivalent plas-
tic strain distribution is shown in Fig. 12(b) including an
indication of proximity to yielding, and the strain concen-
tration (the total maximum principal strain normalized by
the far-field nominal value) in Fig. 12(c). It is seen that the
plastic zone is not very large but the strain concentration is
very high. Stress and strain concentration factors, kσ and
kε , determined using ABAQUS, are given in Table 3 for
Fig. 7 Schematic diagram showing (a) a specimen with a pore close failure-dominant pore sizes in the range of 100–1000 µm.
to the surface under uniform tensile stress σ ∞ and (b) the pore Since the far-field stresses σ ∞ and yield stresses σ y for the
location s and the equivalent diameter d e on a cross section. specimen groups St and Sb are different, and kσ and kε are


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
566 Y. X . G AO et al.

Fig. 8 (a) Symmetrical quarter cut from the specimen and the global mesh and (b) the refined mesh around the pore. Note: A, cross section;
B, the pore close to the surface; C, specimen surface; D, longitudinal section.

Fig. 9 Maximum principal stress around the pore close to the


surface, normalized by the far-field stress (d e = 600 µm, s = 10 µm).

dependent on them, they are included. From Table 3, it is


clear that both the strain and stress concentration factors
increase with pore size as expected. In the model devel-
oped in the next section, the geometric mean values of kσ

and kε , kg = kσ kε are considered. Figure 13 shows the
Fig. 10 Pore-location effect showing the stress concentration
parameter kg plotted against the pore equivalent diame- around the pore on a cross section: (a) global and (b) local views of a
ter while the data represented by circle (specimen group central pore (kt = 2.11, d e = 800 µm, s = 2100 µm); (c) global and
St ) and square (specimen group Sb ) symbols are calculated (d) local views of a pore close to the surface (kt = 4.07, d e = 800
from the kσ and kε values given in Table 3. The solid line µm, s = 10 µm).


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
THE EFFECT OF POROSITY 567

Fig. 11 (a) Global and (b) local views of the distribution of the maximum principal stress normalized with respect to the far-field stress,
around a pore that intersects the surface (kt = 5.23, d e = 600 µm, s = −20 µm).

Table 2 Stress concentration factor kt as a function of equivalent P O R E - S E N S I T I V E FAT I G U E L I F E


diameter d e (µm) and pore location s (µm) (The case of s = −20 PREDICTION MODEL
corresponds to the pore intersecting the specimen surface)
As already noted, experimental observations demonstrate
s that the failure-dominant pores are always close to, or, at
the specimen surface and the FEA analysis indicates that
de Central 1000 500 100 50 10 −20 these pores induce higher stress/strain concentration than
other pores. The stress-life approach is now employed in
100 2.06 2.06 2.07 2.10 2.14 2.42 2.94
order to relate the fatigue life N f to the equivalent pore
200 2.09 2.07 2.07 2.11 2.15 2.54 3.80
400 2.07 2.10 2.09 2.26 2.30 3.02 4.75 size, d e , based on both experimental data and FEA results.
600 2.10 2.11 2.11 2.42 2.70 3.61 5.23 This model, based on elastic behaviour, is then developed
800 2.11 2.11 2.15 2.65 3.05 4.07 6.79 further by incorporating the effects of local plasticity at
1000 2.13 2.13 2.28 2.94 3.40 4.63 7.22 the pore.
From constant-amplitude fatigue tests, the relation be-
tween the number of cycles to failure, N f , and the
applied stress amplitude, σ a , is given by the Basquin
equation:
in Fig. 13 illustrates the same relation in the form of an  1 /b
equation employed to fit the FEA data. The form of the 1 σa
Nf = , (2)
equation is selected as: 2 σf
 2
de de where σ f is the fatigue strength coefficient and b is the
kg = k0 + k1 + k2 , (1)
d0 d0 Basquin exponent. Both are material constants which are
determined from experimental S–N data. Strictly, Eq. (2)
because it gives a good fit with the data from FEA (see is applicable to initiation controlled fatigue. As shown
Fig. 13). In using Eq. (1), the specimen diameter, d 0 = 5000 schematically in Fig. 14, the failure-dominant pore in-
µm, is employed to normalize the equivalent diameters, duces a region of high stress in the material such that the
and k0 , k1 and k2 are stress/strain concentration param- local stress and strain, σ local and ε local , rather than the far-
eters, determined by least square regression of the FEA field nominal stress and strain, will control the evolution
data (see Table 4). of fatigue damage. The local stress and strain are given


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
568 Y. X . G AO et al.

Fig. 12 Plastic deformation and strain concentration around a pore close to the surface (d e = 600 µm, s = 10 µm, σ ∞ = 120 MPa, σ y = 223
MPa): (a) the plastic zone from a global view, (b) the local distribution of the equivalent plastic strain and (c) the total maximum principal
strain normalized with respect to the far-field nominal strain.

Table 3 Stress and strain concentration factors, kσ and kε , as a function of equivalent pore size d e (µm) for the failure-dominant pores

de

St 100 200 300 400 500 600 700 800 900 1000
σ ∞ = 120 MPa kσ 2.14 2.22 2.33 2.43 2.45 2.58 2.63 2.68 2.76 2.98
σ y = 223 Mpa kε 2.75 3.10 3.63 3.74 4.62 5.50 5.96 6.78 7.54 8.25
Sb
σ ∞ = 150 MPa kσ 2.10 2.19 2.33 2.43 2.60 2.97 3.01 3.02 3.14 3.32
σ y = 266 MPa kε 2.80 3.36 3.79 3.97 4.95 5.37 6.12 6.78 7.85 8.55

by: In the present work, FEA has shown that the greater part
 of the specimen is in the elastic state and the plastic zone
σ local = kt σ ∞ , elastic case around the pore is very small (Fig. 12). A stress-life relation
∞ (3) [Eq. (2)], is first employed to give the fatigue life N f (kt σ ∞
σ local
= kσ σ , ε local
= kε ε ∞ , elasto-plastic case a )


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
THE EFFECT OF POROSITY 569

Fig. 13 The dependence of kg on the normalized pore equivalent


diameter d e /d 0 .

Table 4 Fatigue strength coefficient σ f (MPa) and Basquin


Fig. 14 Schematic diagram illustrating the stress-life approach
exponent b and the stress/strain concentration parameters k0 , k1
employed to estimate the fatigue-life reduction due to casting pores.
and k2

Specimen α σ f b k0 k1 k2

St 0.2 726 −0.17 2.21 9.83 19.09


Sb 0.3 746 −0.15 2.16 12.31 17.87

but then this equation is modified to take into account the


local plastic deformation and strain concentration factor.
Taking kg σ ∞a , which includes both stress and strain con-
centrations, to replace σ a in Eq. (2) leads to a prediction
of fatigue life in the range of N f (σ ∞ ∞
a ) ∼ N f (kt σ a ) which
1,2,18
is a physically reasonable range. Based on notch sen-
sitivity studies17,18 , a fitted coefficient α (0 ≤ α ≤ 1) is
introduced to modify kg in terms of [1 + α(kg − 1)] and
consequently the total fatigue life is estimated from the
modified Eq. (2), i.e.:

1 [1 + α(kg − 1)]σa∞ 1/b Fig. 15 Comparison between the model predictions and
Nf = . (4)
2 σf experimental data.

The definition of α has the merit of providing a scaling


factor for the effect of the casting pore on fatigue life where
α varies from 0 (no pore effect) to 1 (full pore effect). of N f , the normalized equivalent pore size d e /d 0 and the
Eq. (4) can be rewritten as: applied remote stress amplitude σ ∞ a . The stress and strain
concentrations around the pore are separately character-

 2  1/b
1 de de σa∞ ized by k0 , k1 and k2 in terms of the FEA simulations.
Nf = 1 + α k0 + k1 + k2 −1 . More importantly, the pore size d e is correlated with the
2 d0 d0 σf
fatigue life N f by the stress/strain concentration factors.
(5)
In other words the microscopic parameter d e that charac-
In Eq. (5), it is noted that α, σ f
and b are material constants terizes the casting pores has been correlated to the fatigue
which can be determined by fitting the experimental data life, a macroscopic property, by Eq. (5).


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570
570 Y. X . G AO et al.

Based on the experimental data from A356 as listed in 4 Hyzak, J. M. and Bernstein, I. M. (1982) The effect of defects
Table 1 and the geometric constants k0 , k1 , k2 , given above on the fatigue crack initiation process in two P/M superalloys:
(see Table 4), and with d 0 = 5000 µm, the material em- Part I. Fatigue origins. Metall. Trans. 13A, 33–43.
5 Tanaka, K. and Mura, T. (1982) A theory of fatigue crack
pirical constants α, σ f and b in Eq. (5) are estimated using
initiation at inclusions. Metall. Trans. 13A, 117–123.
least-square regression (Table 4). Using these estimated 6 Lankford, J. and Davidson, D. L. (1983) Fatigue crack
values, Fig. 15 shows the comparison between the model micromechanisms in ingot and powder metallurgy 7xxx
prediction and the experimental data, where the linear re- aluminium alloys in air and vacuum. Acta. Metall. 31,
lation of Eq. (5) is also plotted by solid line. Good agree- 1273–1284.
ment is found, which implies that the equivalent size of the 7 Couper, M. J., Neeson, A. E. and Griffiths, J. R. (1990) Casting
failure-dominant pore is an important factor that domi- defects and the fatigue life of an aluminium casting alloy.
Fatigue Fract. Eng. Mater. Struct. 13, 213–227.
nates the fatigue life, and the pore-sensitive model [Eq.
8 Skallerud, B., Iveland, T. and Härkegård, G. (1993) Fatigue life
(5)] gives a satisfactory prediction of the dependence of assessment of aluminium alloys with casting defects. Eng. Fract.
fatigue life on pore size. Mech. 44, 857–874.
The relatively small degree of scatter in predicted fatigue 9 McLellan, D. L. and McLellan, M. M. (1996) American
life in this model (Fig. 15) supports the view that the size Foundrymen’s Society, Inc. 505 State St., Des Plaines, IL
of a failure-dominant pore is a major factor determining 60016–8399, USA.
fatigue life. However, other factors such as the pore shape, 10 Wang, Q. G., Apelian, D. and Lados, D. A. (2001) Fatigue
behavior of A356-T6 aluminium cast alloys: Part I. Effect of
silicon particles, SDAS and the interactions between them
casting defects. J. Light Met.. 1, 73–84.
must also be considered. These microstructural features, 11 Flemings, M. C. (1974) Solidification Processing. McGraw-Hill,
which potentially can influence the evolution of fatigue New York.
damage, are the subject of further study. 12 Lee, P. D., Chirazi, A. and See, D. (2000) Modeling
microporosity in aluminium-silicon alloys: A review. J. Light
Met. 1, 15–30.
CONCLUSIONS 13 Seniw, M. E., Fine, M. E., Chen, E. Y., Meshii, M. and Gray, J.
(1997) In: Relation of Defects Size and Location to Fatigue Failure in
Experimental observations show that casting pores are the
Al Alloy A356 Cast Specimens, Warrendale, PA, pp. 371–379.
defects primarily responsible for fatigue failure in the cast Minerals, Metals & Materials Soc (TMS).
A356-T6 alloy. 14 Stanzl-Tschegg, S. E., Mayer, H. R., Beste, A. and Kroll, S.
FEA analysis demonstrates that failure-dominant pores (1995) Fatigue and fatigue crack propagation in AlSi7Mg cast
which cause crack initiation are both large and near to the alloys under in-service loading conditions. Int. J. Fatigue 17,
specimen surface where they induce much higher stress 149–155.
concentration than other pores. 15 Caton, M. J. (2001) Ph.D. Thesis, University of Michigan.
16 Boileau, J. M. (2000) Ph.D. Thesis, Wayne State University.
A pore-sensitive fatigue life prediction model has been
17 Peterson, R. E. (1959) In: Metal Fatigue (Edited by G. Sines and
developed based on the experimental data and finite ele- J. L. Waisman); pp. 293–306. McGraw-Hill, New York.
ment analysis which shows good agreement with experi- 18 Peterson, R. E. (1974) Stress Concentration Factors. John Wiley &
mental data. Sons, New York.
19 Lukáš, P. (1996) In ASM Handbook 19: Fatigue and Fracture, pp.
96–109. ASM Interrnational, Materials Park, OH.
20 Petroski, H. (1996) Invention by Design: How Engineers Get From
Acknowledgements
Thought to Things. Harvard University Press, Cambridge, MA.
The authors gratefully acknowledge the financial support 21 Yi, J. Z., Gao, Y. X., Lee, P. D., Flower, H. M. and Lindley, T.
of the EPSRC (GR/R32376) and Aisin Takaoka Co. Ltd. C. (2003) Scatter in fatigue life due to effects of porosity in cast
A356-T6 aluminium-Silicon alloys. Metall. Trans. 34A,
1879–1890.
REFERENCES 22 In: Annual Book of ASTM Standards; Vol. 03.01 (Edited by R. A.
Storer). ASTM, West Conshohocken, PA, 1997.
1 Suresh, S. (1998) Fatigue of Materials. Cambridge University 23 Murakami, Y. and Endo, M. (1994) Effect of defects, inclusions
Press, Cambridge, UK. and inhomogeneities on fatigue strength. Int. J. Fatigue 16,
2 Frost, N. E., Marsh, K. J. and Pook, L. P. (1974) Metal Fatigue. 163–182.
Oxford University Press, Oxford. 24 Gall, K., Horstemeyer, M. F., Degner, B. W., McDowell, D. L.
3 Bowles, C. Q. and Schijve, J. (1973) The role of inclusions in and Fan, J. (2001) On the driving force for fatigue crack
fatigue crack initiation in an aluminium alloy. Int. J. Fract. 9, formation from inclusions and voids in a cast A356 aluminium
171–179. alloy. Int. J. Fract.. 108, 207–233.


c 2004 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 27, 559–570

You might also like