Int. J. Fatigue Vol. 20, No. 7, pp.

517–530, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0142–1123/98/$19.00

PII: S0142-1123(98)00020-6

An experimental study of the effect of a flaw

at a notch root on the fatigue life of cast Al
319
A.A. Dabayeh, A.J. Berube and T.H. Topper
Department of Civil Engineering, University of Waterloo, Waterloo, Ontario,
Canada N2L 3G1
(Received 24 November 1996; revised 4 July 1997; accepted 8 February 1998)

Smooth and notched specimens of a 319 cast aluminum alloy containing casting defects were fatigue
tested in the as cast and hipped conditions. Hipping is a process in which the material is subjected to
a high pressure at high temperature to eliminate flaws and then slowly cooled. The materials were
tested under fully reversed constant amplitude loading and under a variable amplitude load history
consisting of underloads followed by constant amplitude small cycles. The stress ratio and the number
of constant amplitude small cycles following an underload were adjusted so that the crack did not
close and remained fully open for all the small cycles. Three notch sizes of 1.0 mm, 3.0 mm and
6.0 mm in diameter, with a natural flaw at the center of the notch root, were examined under constant
and variable amplitude loading. In another series of tests the natural flaws in the as cast material were
modeled in the hipped material by a notch made with a circular drill at the notch root. The fatigue
lives observed showed the following:
1. A notch size effect is observed for notched as cast and hipped Al 319 specimens having a 0.6 mm
diameter flaw at notch root under constant and variable amplitude loading. The fatigue limit stress
ranges for hipped Al 319 under constant amplitude loading were 117 MPa, 90 MPa, and 80 MPa for
1.0 mm, 3.0 mm, and 6.0 mm diameter edge notches respectively. The corresponding values under
variable amplitude loading were 25 MPa, 18 MPa, and 8 MPa.
2. The effect of natural flaws or casting defects, on fatigue strength, in the 319 cast aluminum
alloy can be modeled by a drilled hole of the same size.
A crack growth analysis based on a fracture mechanics approach was used to model the fatigue
behavior of the cast aluminum material under variable amplitude loading. In calculating the fatigue life
of the cast aluminum, the flaw was modeled by a spherical cavity having the same diameter as the
natural flaw. The size and location of the flaw at the notch root were varied to examine their effect
on fatigue life. Fatigue life predictions were in good agreement with the experimental results.  1998
Elsevier Science Ltd. All rights reserved

(Keywords: cast aluminum; artificial flaws; natural flaws; constant amplitude loading; variable amplitude
loading; flaw at notch root)

INTRODUCTION a stress concentration and high stresses. It is important

The casting process consists of pouring a totally molten
metal into a mold cavity which has the desired shape.
Upon solidification the metal takes the shape of the
mold. The solution of gases and the number of gas
pores left in the cast metal increases with increasing
temperature and with time at temperature. Therefore
casting is always done at the lowest possible tempera-
ture and with the least possible delay, to decrease the
defect size. If during the cooling and solidification
evolving gasses are trapped at the solid–liquid inter-
face, dendritic arms and gas pores are formed which
weaken the cast metal. Fatigue cracks initiate at these
casting defects, under repeated loading, if they are
located in places such as a notch root where there is
to consider the effect of these inhomogeneities when
cast alloys are used for components that are subjected
to fatigue loads.
Couper et al.1 studied the fatigue behavior of a
strontium-modified heat-treatable Al–7Si–0.4Mg cast
aluminum alloy. Their material had shrinkage pores up
to 140 mm in diameter and they found that in nearly
all samples tested the interdendritic pores close to or
at the specimen surface were the site of crack initiation.
One of their aims was to specify a minimum pore size
for their material below which classical crack initiation
would determine the fatigue life. They stated that a
full quantitative study of the initiation mechanism
should include notched samples in which there is a
chance of a pore being close to the notch region, and

517
518 A.A. Dabayeh et al.

that until such an analysis is done it is not possible

to specify the minimum pore size. Later Heuler et al.2
studied manufacturing defects in the near surface region
of 2.25% Cr 1% Mo castings. The defects examined
were micropores, pinholes, slag, oxide layers,
inclusions, hot tears and small microcracks in welds.
They found that the largest defects were hot tears
having a mean surface area of 11 mm2, and that only
inclusions and blowholes have large notch radii with a
range of 400–500 mm. Surface defects such as pinholes
(piping), slag inclusions and small surface defects in
welded areas proved to be the most effective crack
initiators while internal defects such as hot tears and
micropores, although they were large, served in less
than 40% of the cases as crack initiators. They con-
cluded that surface and near surface defects initiate
cracks sooner than defects embedded in the specimen.
They added that defects having small notch radii such Figure 2 A schematic illustration of the size distributions of the
as surface cracks at welds, pinholes, slag and oxide largest pore at the specimen surface7
layers initiate cracks more frequently than, for example,
hot tears, inclusions, blowholes and micropores. Thus Table 1 Chemical composition of 319 aluminum alloy
it can be seen that the probability of crack initiation (percentage by weight)
is higher from large than from small defects and that
Si Cu Cr Mn Mg Fe Ni
for the same defect size cracks initiate faster from
defects on or near a surface than from embedded
5.9 3.4 0.07 0.38 0.25 0.91 0.07
defects.
Modeling of defects simplifies the study of the effect
of the size of casting defects and flaws on fatigue life.
Murakami and Endo,3 in studying the effect of small materials containing defects or flaws is lower than that
defects on fatigue strength in low and medium carbon of a defect free material. If a defect is located at a
steel, modeled the defects by drilling a small artificial stress concentration region, such as a notch root, an
hole, the diameter of which varied from 40 to 200 mm. even lower fatigue strength might be expected. It is
The depth of the drilled hole was the same as its the purpose of this paper to determine the reductions
diameter. A fractographic examination by Skalleurd et in fatigue life caused by a defect at a notch root under
al.4 on a cast aluminum alloy showed that cracks the application of constant and variable amplitude
growing from irregular defects of a given type quickly fatigue loads.
approached a spherical or ellipsoidal shape. An FEM
approach for calculating K and da/dN in irregularly EXPERIMENTAL PROCEDURES
shaped cracks investigated by Smith and Cooper5 and Material
Soboyejo et al.6 also supported these simplified shapes.
In the present paper a fractographic examination The materials used in this study are 319 cast alumi-
showed that the defects in the cast aluminum alloy num alloys in the as cast and hipped conditions. Hip-
were either gas pores or shrinkage cavities. Gas pores ping is a process in which the cast material is subjected
are almost spherical while shrinkage cavities are irregu- to a high pressure at high temperature and then slowly
lar in shape. In modeling, the irregular shape of a cooled to eliminate flaws. A sand cast process was
shrinkage cavity was simplified to a spherical cavity used to cast the 319 aluminum alloy flat plates. The
which had the same shape as the gas pore shown in as cast 319 alloy had the number and size of the
Figure 1. gas pores artificially increased (high gas) compared to
It is generally accepted that the fatigue strength of material cast in the normal manner (low gas). Only
the material in the high gas condition is used in this
study. A schematic illustration of the distribution of
pore sizes and of the distribution of the largest pore
at a specimen surface is shown in Figure 27. The
chemical composition of the as cast and hipped 319
aluminum alloys is shown in Table 17.
A tensile test was performed for each of the as
cast and hipped 319 aluminum alloys. The mechanical
properties obtained are shown in Table 2. Cyclic stress–
Table 2 Mechanical properties

Material 0.2% yield Ultimate tensile Fracture strain

stress (MPa) stress (MPa) (%)

As cast Al 319 145 193 1.08

Hipped Al 319 162 232 1.65
Figure 1 Defect geometry and its simplification
 An experimental study of the effect of a flaw at a notch root on the fatigue life of cast Al 319
Figure 3 Cyclic stress–strain curves for as cast and hipped 319
aluminum alloy
strain curves for both the as cast and hipped 319
aluminum alloys are shown in Figure 3. The hipping
process slightly increased the ultimate strength and the
fracture strain of the material.
Specimens
Flat fatigue test specimens with the dimensions
shown in Figure 4, were machined from 6.35 mm
thick plates of the material. The smooth and notched
specimens were prepared in accordance with ASTM
standard E606 for constant amplitude low-cycle fatigue
tests. The preparation included hand polishing of the
gauge section in the loading axis direction with emery
paper of grades 400 and 600. An artificial flaw 0.6 mm
in diameter was drilled at the notch root of the notched
hipped specimens. The notched specimens tested had
1.0 mm, 3.0 mm and 6.0 mm diameter circular edge
notches. The position and shape of the artificial flaw
Figure 4 Specimen configurations
519
drilled at the notch root are shown in Figure 5. To
prepare specimens with natural flaws at notch root, the
cast plates were sent out to be examined using ultra-
sonic waves to accurately locate large flaws, their size
and their depth below the surface of the cast plate.
Specimens were then cut and the required notch was
machined so that the natural flaw was located at the
notch root. There was some variation in the accuracy
of positioning of the flaw at the notch root. Specimens
were gripped by inserting both ends between two
clamping wedges. The wedges were then tightened
so as to prevent any movement or rotation at the
gripping area.
Theoretical stress concentration
The theoretical surface stress concentration factor,
Kt, and the stress field ahead of the notch for the three
different notch sizes, has been calculated using finite
element analysis. The ABAQUS8 finite element analy-
sis package was used to construct the finite element
mesh and to perform the analysis. The finite element
mesh used was very finely spaced in the vicinity of
the notch as shown in Figure 6. Eight-noded plane
strain elements were used in the mesh which consisted
of 256 elements. The gripping of the specimen was
modeled in the finite element code by preventing any
rotation and displacement at the ends, except for the
displacement in the loading direction which was
allowed. The analysis gave gross stress concentration
factors, Ktg, of 3.205, 3.109, and 3.011 for the 6.0 mm,
3.0 mm, and 1.0 mm diameter edge notches respect-
ively. The stress distribution ahead of the notch was
also measured experimentally for the 6.0 mm diameter
edge notch by gluing six strain gauges along the
specimen width in front of the notch. The experimental
stress field was compared with the finite element results
and there was a good agreement as shown in Figure 7.
EQUIPMENT AND TEST TECHNIQUES
A digital process control computer was used to output
both constant amplitude small cycles and periodic
underloads in the form of a sinusoidal loading wave.
Figure 5 Position and shape of the artificial flaw
520 A.A. Dabayeh et al.
Figure 6 The finite element mesh in the vicinity of the 6.0 mm
diameter notch
Figure 7 Experimental and finite element results for the stress field
ahead of a 6.0 mm diameter edge notch
All fatigue tests on the smooth specimens were conduc-
ted under strain control at frequencies between 1 Hz
and 45 Hz, while all fatigue tests on the notched
specimens were conducted under load control at fre-
quencies between 1 Hz and 150 Hz depending on the
stress and strain amplitude of the fatigue cycles using
FLEX control software9.
A block loading sequence was used to study the
effect of variable amplitude loading on the fatigue life.
Each block consisted of one underload followed by N
small cycles, where N for a given test was chosen so
that the crack opening stress (or strain) at the end of
the underloads did not reach the minimum stress (or
strain) of the small cycles. The small cycles are then
fully effective. The value used for N was based on a
crack opening stress build-up model proposed by the
present authors10 and fitted to test data for the present
material (see Appendix A). The variation in crack
opening stress is shown schematically in Figure 8.
Crack opening stress
To estimate the crack opening or closing stress levels
an empirical model proposed by DuQuesnay11, fitted
to measured crack opening data, was used to calculate
the steady-state crack opening stress levels, and a
model suggested by the present authors10, also fitted
to measured data, was used to calculate the crack
opening stress build-up after underloads (see Appen-
dix A).
Crack growth rate
A load history containing underloads was applied to
the 319 cast aluminum alloy to obtain the effective
stress intensity versus crack growth rate curve. The
minimum and maximum stress levels of the underload
cycles were −82.8 and 41.4 MPa respectively. The
number of small cycles between underloads, N, was
300. The minimum and maximum stresses of the
underload and the number of small cycles were chosen
so that underload crack growth represents only a small
fraction of the total crack growth during a block of
loading history and does not exceed 5 × 10−10 m/cycle.
The crack growth rate, (da/dN)t, for the block loading
history was determined, then the crack growth rate for
the underload cycles alone was determined during the
block test by inserting a group of underload cycles
and measuring their growth rate, (da/dN)U.L., between
measurements of block growth rates. The crack growth
rate for the small cycles in the block test, (da/dN)s,
was obtained by subtracting the growth per block due
to the underload cycle, (da/dN)U.L., from the total
growth per block, (da/dN)t, and dividing by the number
of cycles in the block as follows:
(N + 1)(da/dN)t − (da/dN)U.L.
(da/dN)s = (1)
N
The experimental crack growth data plotted in terms
of crack growth rate versus the effective stress intensity
are shown in Figure 9. The effective threshold stress
intensity factor range obtained was 0.98 MPa√m.
Figure 8 Variations in crack opening stress
 An experimental study of the effect of a flaw at a notch root on the fatigue life of cast Al 319
Figure 9 Crack growth rate in terms of ⌬Keff for cast Al 319
FATIGUE LIFE PREDICTION MODEL
Theoretical basis
The fatigue life prediction model used by Dabayeh
et al.7,12 has been modified in this paper to account
for the variation in the flaw size and its position with
respect to the notch root.
To calculate local stresses and strains at a notch an
analysis based on Neuber’s rule13 can be used:
Kt = √K␴K⑀
where K␴ is the local stress concentration and K⑀ is
the local strain concentration. Solving simultaneously
Neuber’s rule with the cyclic stress–strain curve equ-
ation gives the local stress and strain corresponding to
a given nominal stress. However, since in this case
we are dealing with a crack emanating from a flaw at
a notch root, the theoretical stress concentration factor
Kp for the notch root is replaced by the elastic stress
concentration factor Kp for a crack in a flaw at the
notch root. The elastic stress concentration factor Kp
accounts for the increase in crack tip stress due to a
flaw in a notch. It can be calculated from the following
formula if the stress intensity factor range, ⌬K, is
known:
⌬K
Kp =
F⌬S√␲l
where ⌬K is the stress intensity factor range, F is a
geometric factor that accounts for crack front shape
and finite specimen size, Kp is the elastic stress concen-
tration factor, ⌬S is the nominal stress range, and l is
the crack length.
An approximate stress intensity factor solution for
the present specimens with a defect at a notch root
has been calculated using the stress intensity factor
solution for a circular crack emanating from a spherical
cavity by Murakami et al.14 in combination with the
stress intensity factor solution for embedded semi-
elliptical cracks in finite notched plates under tension
521
by Neuman15 as shown in Appendix B. The Kp was
then used together with the cyclic stress–strain curve
equation in Neuber’s rule to obtain the local stresses
and strains in the vicinity of the crack tip for inelastic
conditions. Local strains were used to calculate the
strain intensity factor for inelastic local strains as will
be explained in the following section. The procedure
for calculating the local stresses and strains is as fol-
lows:
1. Neuber’s rule is used to derive the relation between
local and nominal stresses and strains:
K 2p = K␴K⑀ (4)
⌬␴ ⌬⑀
K 2p = (5)
⌬s ⌬e
k2p⌬s⌬e = ⌬␴⌬⑀ (6)
Kp√⌬s⌬e = √⌬␴⌬⑀ (7)
2. The material’s cyclic stress–strain equation is used
to derive and plot the relation between √⌬␴⌬⑀
and ⌬⑀.
3. Local strains in the vicinity of the crack tip are
obtained by entering the value of Kp√⌬s⌬e in the
plot described in (2) and reading the value of ⌬⑀.
Strain intensity factor
The intensity factor solutions used in this paper are
strain based intensity factor equations which are suit-
(2) able for plastically strained short cracks and which
converge to the conventional linear elastic stress inten-
sity for long cracks and elastic nominal stresses. This
strain based intensity factor replaces the conventional
stress intensity factor which does not apply to small
cracks. The strain based K-solution used has the follow-
ing form:
⌬K = F⌬⑀EQ⑀√␲l (8)
where F is a geometric factor that accounts for crack
front shape and finite specimen size, ⌬⑀ is the local
strain range, E is Young’s modulus, l is the crack
length measured from the notch root and Q⑀ is a
surface strain concentration factor16 that accounts for
short crack behavior. The surface strain concentration
factor can be expressed as:16
(3) ⌬⑀
Q⑀ = = 1 + 5.3e−(␣l/D) (9)
⌬e
where D is the grain size in the crack growth direction,
and ␣ is a factor that accounts for the ease of cross
slip in the material.
The surface strain concentration factor, Q⑀, decays
rapidly with crack length and Equation (8) converges
to the long crack strain intensity factor equation. The
effective strain intensity factor was calculated by
replacing the total strain range ⌬⑀ in Equation (8) with
the effective portion of the strain cycle. The steady
state and build-up crack opening stress formulas in
Appendix A were used to estimate the effective stress
522 A.A. Dabayeh et al.

ranges. Effective strain ranges were calculated using

effective stress ranges and the elastic Young’s modulus
if the local stresses and strains were completely elastic.
For inelastic local stresses and strains, the crack open-
ing strain was calculated using the crack opening stress
and the cyclic stress–strain curve. The effective strain
was then obtained by subtracting the crack opening
strain from the maximum strain. After substituting the
effective strain range in Equation (8) it takes the
following form:
⌬Keff = F⌬⑀effEQ⑀√␲l (10)

FATIGUE LIFE PREDICTION

Fatigue life prediction was carried out by a numerical
integration of Equation (10) along the reference closure
free crack growth curve between initial and final crack
lengths l0 and lf respectively as

冕
lf

da
Nf = (11)
f(⌬Keff)
l0

Basinski and Basinski17 and Hunsche and Neumann18

defined cracks as singularities in a crystal related to
surface intrusions and extrusions deeper than 3 ␮m
developed in persistent slip bands. The value of l0 was
assumed to be 3 ␮m while lf was determined by the
onset of fast fracture. The crack growth stages for a
crack around a flaw at the notch root and a crack
around a flaw at a distance from the notch root,
assumed in the model, are shown schematically in
Figure 10(a) and (b) and are summarized in the follow-
ing five stages:
For a flaw at the notch root
1. Fast crack growth of the crack around the flaw to
a semi-circular shape; i.e., no crack growth calcu-
lations were done and the crack around the flaw
was assumed to jump to a semi-circular shape.
2. The growth of the semi-circular crack under the
effect of the flaw and notch stress fields until the
crack leaves the flaw stress field.
3. The growth of the crack under the effect of the
notch stress field until the crack reaches the speci-
men edges. Note that in this case the crack growth
rates in the depth and surface directions are different
and the crack assumes a semi-elliptical shape.
4. Fast crack growth from a semi-elliptical crack shape
touching the specimen edges to a straight fronted
through crack.
5. Growth of the through crack under the notch stress
field until fracture.
For a flaw at a distance from the notch root
1. The growth of a crack around the flaw under the
effect of the flaw and notch stress fields until the
crack reaches the notch root edge.
2. Fast crack growth of the crack around the flaw to
a semi-circular crack; i.e., no crack growth calcu-
lations were done and the crack around the flaw
was assumed to jump to a semi-circular shape.
Figure 10 Schematic showing crack growth stages
3. The growth of the crack under the effect of the
notch stress field until the crack reaches the speci-
men edges. Note that in this case the crack growth
rates in the depth and surface directions are different
and the crack assumes a semi-elliptical shape.
4. Fast crack growth from a semi-elliptical crack shape
touching the specimens edges to a straight fronted
through crack.
5. Growth of the through crack under the notch stress
field until fracture.
EXPERIMENTAL RESULTS
Fatigue tests for smooth specimens
Strain–life constant amplitude tests at R = −1 were
conducted on the hipped 319 aluminum alloy. The
fatigue limit strain range was 0.26%.
Variable amplitude strain–life fatigue tests were also
conducted. A block loading history was used consisting
of an underload followed by small cycles, which had
the same maximum strain as the underload cycle. A
schematic diagram for the block loading history is
shown in Figure 11. The number of small cycles N was
chosen so that the following conditions were fulfilled:
1. The damage due to the underloads (according to
Miner) did not exceed 20%; i.e., number of
underloads in the expected life did not exceed 0.2
of the number of underloads only required to cause
 An experimental study of the effect of a flaw at a notch root on the fatigue life of cast Al 319
Figure 11 A schematic for the block loading history
failure. This ensured a fatigue test in which most
of the damage was done by the small cycles.
2. All small cycles between underloads were free of
closure. This was achieved by choosing N so that
the crack opening stress, as it built up during the
small cycles, did not reach the minimum stress of
the small cycles before the application of the next
underload, which reduced the crack opening stress.
Fatigue test results under constant and variable
amplitude loading are shown in Figure 12. Results are
plotted on axes of the strain range of the small cycles
versus the equivalent number of small cycles to failure
on a log scale. Miner’s linear damage rule was used
to calculate the equivalent number of small cycles to
failure. The total damage is given by the sum of the
underload damage, Dul, and the small cycle damage Ds:
Dul + Ds = 1 (12)
The damage caused by the underloads is calculated
as the number of underloads in a test, Nul, divided by
the number of underload cycles to failure, Nful, in
a constant amplitude loading history containing only
underload cycles:
Nul
Dul = (13)
Nful
Then an equivalent life for the small cycles is
obtained as
Figure 12 Constant and variable amplitude strain–life curves for
hipped Al319
523
Table 3 Fatigue strengths for constant and variable amplitude
loading
Strain Range (%) Constant amplitude Variable amplitude
Hipped Al 319 0.26 0.04
Ns
Nfseq = (14)
Ds
where Ns is the number of small cycles in a fatigue
test, and Ds is the damage done by small cycles = 1
− Dul.
Combining Equations (13) and (14):
Ns Ns
Nfseq = = (15)
1 − Dul Nul
1−
Nful
Since the crack was fully open throughout the small
cycles the underload block loading history causes the
largest possible amount of small cycle damage for a
given small cycle strain range. The underload strain–
life curve shown in Figure 12 therefore represent lower
bound strain–life curve for the hipped 319 aluminum
alloy. The constant amplitude fatigue strength of
smooth specimens was reduced by 85% when the
variable amplitude loading history was applied. The
fatigue limit test results are given in Table 3.
Fatigue tests of notched specimens
The notches used were circular edge notches,
1.0 mm, 3.0 mm and 6.0 mm in diameter. The speci-
mens were tested under constant and variable amplitude
loading. The fatigue curves shown in Figure 13 are
plotted in terms of the fatigue life versus the nominal
stress range multiplied by the gross surface stress
concentration factor, Ktg⌬S, to account for the different
Ktg values of the three notch sizes. The fatigue limit
test data listed in Table 4 are the nominal fatigue limit
stresses applied on the notched specimens.
After testing, the fracture surface of each specimen
was examined under a high power microscope to locate
the flaw at the notch root. Figure 14 shows two cases
in which the flaws were positioned accurately at the
Figure 13 S–N curves for notched 319 cast aluminum alloy having
natural flaw at notch root
524 A.A. Dabayeh et al.

Table 4 Fatigue limit results for 319 cast aluminum alloy

Loading Constant amplitude Variable amplitude

Condition Natural flaw at Natural flaw at Natural flaw at Natural flaw at Natural flaw at Natural flaw at
1 mm D notch 3 mm D notch 6 mm D notch 1 mm D notch 3 mm D notch 6 mm D notch

Fatigue limit stress 90 82 57 32 18 12

range (MPa)

Figure 14 Two cases of flaws positioned accurately at the notch root

notch root. Figure 15 shows a case in which the
positioning of the flaw was less accurate as well as a
case in which the size of the flaw was less than the
largest average flaw at the specimen surface, which
was about 0.9 mm in diameter. This variation in size
and location of the flaws resulted in a considerable
range of scatter for the fatigue data as can be seen in
Figure 13. From the fracture surface examinations, it
was concluded that the average size of the flaws found
at the notch roots of the specimens tested was about
0.6 mm in diameter.
To model the natural flaw a hole, as shown in
Figure 5, of 0.6 mm in diameter and 0.6 mm in depth,
which was of the same size as the average flaw in the
319 cast aluminum alloy specimens, was drilled at the
notch root of the hipped 319 aluminum alloy. Fatigue
tests of the hipped material under constant and variable
amplitude loading were then conducted for specimens
at three notch sizes with these artificial flaws at their
notch roots. The fatigue life versus Ktg⌬S curves are
shown in Figure 16 and the fatigue limit test data are
listed in Table 5. The stress range values listed in
Table 5 are the nominal stresses applied on the
notched specimens.
DISCUSSION
The ultimate strength of the as cast 319 aluminum
alloy is 193 MPa but the hipped material has an ulti-
mate strength of 232 MPa. Since a material’s fatigue

Figure 15 Variations in accuracy of positioning and size of flaws. (a) Flaw size is less than 0.9 mm. (b) Flaw is located away from the
notch root
 An experimental study of the effect of a flaw at a notch root on the fatigue life of cast Al 319
Figure 16 S–N curves for notched 319 hipped aluminum alloy
having a hole drilled at the notch root
limit varies with its ultimate strength the constant
amplitude S–N curves of the three notch sizes examined
having a 0.6 mm drill at the notch root were multiplied
by the ratio of the hipped ultimate stress to the as cast
ultimate stress and then compared to the constant
amplitude S–N curves of the as cast material having a
natural flaw at the notch root. The comparison is
shown in Figure 17. From the figure it can be seen
that the data for natural and artificial flaws fall together
for each notch size which suggests that a natural flaw
in the as cast 319 aluminum alloy can be modeled by
an equivalent drilled hole of the same size.
The same procedure was applied to the fatigue test
data for variable amplitude loading and the results
obtained are plotted in Figure 18. The figure shows
that the 0.6 mm diameter drilled hole and the natural
flaw did not give similar fatigue lives near the fatigue
limit region. As shown in Figure 18 the difference
between the natural and artificial flaw at the notch root
results, in terms of fatigue limit stress ranges multiplied
by the notch stress concentration factor, was highest
for the 1.0 mm diameter notch, at 33 MPa, and lower
for the 3.0 mm and 6.0 mm diameter notches, at
12 MPa and 17 MPa, respectively. An examination of
the fracture surface of the broken specimens with
natural flaws at their notch roots revealed that there
was a variation in the sizes of the flaws that caused
failure and in their location in the notch root. This
variation explains the difference in fatigue limit
between the as cast and hipped specimens having a
natural and an artificial flaw at the notch root respect-
ively. The difference between fatigue limit for as cast
and hipped materials, due to the variation in flaw size
and flaw location, is larger for small notches than for
large notches. This is expected because notch stress
Table 5 Fatigue limit results for hipped 319 aluminum alloy
Loading Constant amplitude
Condition Artificial flaw at Artificial flaw at
1 mm D notch 3 mm D notch
Fatigue limit stress 117 90
range (MPa)
525
fields decay more rapidly for small notches than for
large notches, and a given variation in flaw size or
distance from the notch root will cause a larger change
in the combined notch–flaw stress field for a small
notch than for a large notch.
The fatigue life prediction model was used to predict
the fatigue life for the hipped Al 319 material with a
drilled hole at the notch roots of the 1.0, 3.0 and
6.0 mm diameter edge notches under variable amplitude
loading. Experimental fatigue lives are compared with
the model fatigue life predictions in Figures 19–21.
The model predictions agree well with the experimental
results. The model was also used to predict a fatigue
life band which covers the range of different flaw sizes
and flaw locations in the notch root, found in the
microscopic observations of the fractured surfaces of
the fatigued specimens. One limit was taken as a
0.6 mm flaw positioned accurately at the notch root,
which was the most severe case observed, and the
other limit was taken as a 0.4 mm flaw positioned
0.4 mm away from the notch root, which was the
least severe case. Figures 22–24 show the fatigue life
prediction bands together with the cast Al 319 material
fatigue test results. Most of the data fall within the
predicted fatigue life band.
The stress field ahead of the notch is increased if a
flaw is introduced at the notch root. If the flaw dimen-
sions are much smaller than the notch dimensions the
stress concentration factor, Kt, at the flaw is expected
to be the product of the stress concentration factor of
the notch, Ktg, and the stress concentration factor of
the flaw, Ktf, i.e. Kt = Ktg × Ktf, and the maximum
decrease in the fatigue strength for a flaw at a notch
root is expected to be equal to the smooth specimen
fatigue limit divided by the factor (Ktg × Ktf). The
observed results under constant amplitude loading
showed that ratio of the fatigue limit stress range, due
to a 0.6 mm diameter drill at the notch root of 6 mm
diameter, to that for smooth specimens is 3.3 which
is higher than Ktg = 3.205 and lower than the maximum
theoretical value of Kt = 3.205 × 2.03 = 6.5 for a
small flaw at the root of a large notch. Fatigue limit
data for notched specimens with flaws under variable
amplitude loading showed a decrease from the smooth
specimen variable amplitude fatigue limit of a factor
of 4 for the 0.6 mm diameter drill at the root of a
6 mm diameter notch, which is again higher than the
Ktg value of 3.205 and lower than the maximum theor-
etical value of Kt = 6.05.
DuQuesnay et al.19 studied the notch fatigue behavior
of a 2024-T351 aluminum alloy under variable ampli-
tude loading. The variable amplitude loading sequence
consisted of a periodic compression–tension overload
of yield magnitude followed by smaller high stress
Variable amplitude
Artificial flaw at Artificial flaw at Artificial flaw at Artificial flaw at
6 mm D notch 1 mm D notch 3 mm D notch 6 mm D notch
80 25 18 8
526 A.A. Dabayeh et al.
Figure 17 Comparison of constant amplitude S–N curves for
notched as cast and hipped 319 aluminum with a flaw or a hole
drilled at the notch root. (a) Notch diameter is 1.0 mm. (b) Notch
diameter is 3.0 mm. (c) Notch diameter is 6.0 mm
ratio cycles. They found that the notch size effect that
is typically observed under constant amplitude load-
ing20 is reduced under variable amplitude loading as
shown in Figure 25. A similar observation was made
by MacDougall and Topper21 for an SAE 1045 steel.
In their results, as shown in Figure 26, there was no
significant difference in the fatigue limit stress range
Figure 18 Comparison of variable amplitude S–N curves for
notched as cast and hipped 319 aluminum with a flaw or a hole
drilled at the notch root. (a) Notch diameter is 1.0 mm. (b) Notch
diameter is 3.0 mm. (c) Notch diameter is 6.0 mm
for different notch sizes tested under the same variable
amplitude loading. For the case of a constant flaw size
at a notch root, a notch size effect was observed under
both constant and variable amplitude loading. The
fatigue test results of Figures 13 and 16 for constant
and variable amplitude loading respectively show that
as the size of a notch with a constant size flaw
increases, the fatigue limit decreases. The fatigue
strengths multiplied by Ktg, for a 0.6 mm diameter hole
drilled at the root of 1.0 mm, 3.0 mm and 6.0 mm
diameter notches respectively, were 352 MPa, 280 MPa
and 256 MPa for constant amplitude loading and
75 MPa, 57 MPa and 25 MPa for variable amplitude
loading.
 An experimental study of the effect of a flaw at a notch root on the fatigue life of cast Al 319
Figure 19 Variable amplitude fatigue life data and fatigue life
prediction for hipped Al 319 with a hole drilled in the root of a
1.0 mm diameter notch
Figure 20 Variable amplitude fatigue life data and fatigue life
prediction for hipped Al 319 with a hole drilled in the root of a
3.0 mm diameter notch
Figure 21 Variable amplitude fatigue life data and fatigue life
prediction for hipped Al 319 with a hole drilled in the root of a
6.0 mm diameter notch
527
Figure 22 Variable amplitude fatigue life data and fatigue life
prediction for as cast 319 aluminum with a flaw in a 1.0 mm
diameter notch
Figure 23 Variable amplitude fatigue life data and fatigue life
prediction for as cast 319 aluminum with a flaw in a 3.0 mm
diameter notch
Figure 24 Variable amplitude fatigue life data and fatigue life
prediction for as cast 319 aluminum with a flaw at 6.0 mm diam-
eter notch
CONCLUSIONS
A notch size effect was observed under constant and
variable amplitude loading for a 0.6 mm diameter flaw
at the notch root. For the hipped material under con-
stant amplitude loading, the fatigue limit multiplied by
the gross stress concentration factor decreased from
352 MPa to 256 MPa as the notch size increased from
1 mm to 6 mm in diameter. The corresponding values
for variable amplitude fatigue were from 75 MPa to
25 MPa.
528 A.A. Dabayeh et al.
Figure 25 Notch size effect under: (a) constant amplitude loading;
(b) variable amplitude loading
Figure 26 Notch size effect for 1045 steel specimens subjected to
variable amplitude loading21
Natural flaws in the as cast 319 aluminum alloy
can be modeled by an equivalent drilled hole of the
same size.
The fatigue limit of notched 319 cast aluminum
specimens having a small notch radius is more affected
by a given variation in flaw size and flaw position
than specimens having large notch radius.
REFERENCES
1 Couper, M.J., Nesson, A.E. and Griffiths, J., R. Fatigue Farct.
Eng. Mater. Struct., 1990, 13(3), 213.
2 Heuler, P., Berger, C. and Motz, J., Fatigue Fract. Eng. Mater.
Sruct., 1992, 16(1), 115.
3 Murakami, Y. and Endo, T., Int. J. Fatigue, 1980, 2, 23.
4 Skalleurd, B., Iveland, T. and Härkegård, G., Eng. Fracture
Mechanics, 1993, 44(6), 857.
5 Smith, R.A. and Cooper, J.F., Int. J. Press. Vess. Piping, 1989,
36, 315.
6 Soboyeji, W.O., Kishimoto, K., Smith, R.A. and Knott, J., F.
Fatigue Fract. Eng. Mater. Struct., 1989, 12, 167.
7 Dabayeh, A.A., Xu, R.X., Du, B.P. and Topper, T.H., Int. J.
Fatigue, 1996, 18(2), 95.
8 ABAQUS User’s Manual. Hibbitt, Karlsson & Sprenson, Provi-
dence, RI, 1995.
9 Pompetzki, M.A., Saper, R.A. and Topper, T.H., Can. Metall.
Q., 1988, 252, 181.
10 Dabayeh, A.A. and Topper, T.H., Int. J. Fatigue, 1995, 17(4),
261.
11 DuQuesnay D. L., Ph.D. thesis, University of Waterloo, Water-
loo, Ontario, 1991.
12 Dabayeh, A. A., Xu, R. X. and Topper, T. H., In Proceedings
of the Sixth International Fatigue Congress, Berlin, Germany,
6–10 May 1996 (Eds G. Lütering and H. Nowack), vol. 1, pp.
123–128.
13 Neuber, H., ASME J. Appl. Mech., 1961, 28, 544.
14 Murakami, Y., Norikura, T. and Yasuda, T., Trans. Japan. Soc.
Mech. Engrs., 1982, 48(436), 1558.
15 Newman, J.C., ASTM STP, 1992, 1149, 6.
16 Abdel-Raouf, H., Topper, T.H. and Plumtree, A., Scr. Met.
Mater., 1991, 25, 597.
17 Basinski, Z.S. and Basinski, S.J., Acta Metall., 1985, 33, 1307.
18 Hunsche, A. and Neumann, P., Acta Metall., 1986, 34, 207.
19 DuQuesnay, D.L., MacDougall, C., Dabayeh, A. and Topper,
T.H., Int. J. Fatigue, 1995, 17(2), 91.
20 DuQuesnay, D. L., M.A.Sc. thesis, University of Waterloo,
Waterloo, Ontario, 1986.
21 MacDougall, C. and Topper, T. H., Int. J. Fatigue, 1997,
19(5), 389.
APPENDIX A
THE STEADY STATE CRACK OPENING STRESS
MODEL
An empirical model, proposed by DuQuesnay11 for
the steady-state crack opening stress under constant
amplitude loading was used. The model, which has
two experimentally determined constants ␪ and ␸, takes
the following form:
冋 冉 冊册
Sop = ␪␴max 1 −
␴max
␴y
2
+ ␸␴min
where ␴max and ␴min are the local maximum and
minimum notch root stresses in notched specimens or
the nominal maximum and minimum stresses in smooth
specimens, and ␴y is the cyclic yield stress. Crack
opening stress measurements were done using a high
magnification power microscope (900 × ) and a value
for ␪ = 0.55 and for ␸ = 0.2 was obtained. The values
for ␪ and ␸ were used in the model to determine the
steady-state crack opening stress of the cast alumi-
num alloy.
 An experimental study of the effect of a flaw at a notch root on the fatigue life of cast Al 319 529

THE CRACK OPENING STRESS BUILD-UP

MODEL
Another empirical model, proposed by the present
authors10, was used to obtain the change in crack
opening stress level after the application of an
underload. The model was obtained from extensive
crack opening stress measurements and has the follow-
ing form:
(Sop − Sopul)
= 1 − ␺ exp[ − b(N/N0.8)d]
(Sopss − Sopul)
where Sop is the instantaneous crack opening stress;
Sopul is the post-underload crack opening stress; Sopss
is the steady-state crack opening stress of the cycles
following the underload; N is the number of cycles
following the underload; N0.8 is the number of cycles
following the underload at which the normalized reco-
vered stress (Sop − Sopul)/(Sopss − Sopul) reaches 80% of
its steady-state level; and ␺, b and d are material Figure 28 The position of the cavity at the notch root
constants derived from curve fitting and have the values
of 1.9, 3.0 and 0.75 respectively.
N0.8 is given by a simple linear curve as
notch root as shown in Figure 28. It is assumed that
N0.8 = 1.158(Sopol − Sopss)1.331 the stress, ␴, applied to the flaw at the notch root is
equal to a uniform stress, ␴uf, instead of the non-linear
notch stress field. This assumption is made because
the stress intensity factor solution for a crack emanating
APPENDIX B from a flaw by Murakami is for a uniform remote
applied stress. The uniform stress, ␴uf, is obtained by
dividing the area, A, under the notch stress field by
the crack length as shown in Figure 29. To check the
APPROXIMATE STRESS INTENSITY FACTOR accuracy of the uniform stress assumption, ␴uf, a simple
SOLUTION FOR A DEFECT AT NOTCH ROOT case for a semi-elliptical surface crack in a smooth
finite width plate was tested. The stress intensity factor
The dimensionless stress intensity factor solution, FI,
at the deepest point of the semi-elliptical surface crack
for a circular crack emanating from a spherical cavity
was calculated using a weight function. The calculation
given by Murakami et al.14 is
was done first using a non linear notch stress field and
KI then repeated using the uniform stress field assumption
FI = explained earlier. The results showed that the difference
1.1215␩␴√␲␭
between the two solutions varies with the crack aspect
in which ␭ is the crack length, and for spherical ratio and that the uniform stress solution is smaller by
cavities ␩ = 27/14. This solution is shown in Figure 7–15%. The uniform stress solution used was multi-
27. In the problem under study the cavity occurs at a plied by a factor of 10%.
After the crack had grown out of the stress field of
the cavity a notch stress field using a Newman15 stress
intensity factor solution for a semi-elliptical surface
crack located at the center of a semi-circular edge
notch was assumed. This stress intensity factor solution,

Figure 27 Dimensionless stress intensity factor for a circular crack

emanating from a spherical cavity in an infinite solid body Figure 29 Schematic illustration for the determination of ␴uf
530 A.A. Dabayeh et al.

M3 = 0.29/[0.23 + (a/c)3/2]
g1 = 1 − [(a/t)4(2.6 − 2a/t)1/2/(1 + 4a/c)]cos ␾
g2 = [1 + 0.358␭ + 1.425␭2 − 1.578␭3 + 2.156␭4]/(1
+ 0.08␭2)
␭ = 1/[1 + (c/r)cos(0.9␾)]
g3 = 1 + 0.1(1 − cos ␾)2(1 − a/t)10
g4 = KT[0.36 − 0.032/(1 + c/r)1/2]
where KT is the elastic stress concentration factor.
g5 = 1 + (a/c)1/2[0.003(r/t)2 + 0.035(r/t)(1 − cos ␾3]
− 0.35(a/t)2(1 − 0.5a/c)3cos ␾
fw = 1 − 0.2␥ + 9.4␥2 − 19.4␥3 + 27.1␥4
␥ = (a/t)1/2(c + r)/w
Figure 30 Schematic for a semi-elliptical crack located at the center
of a semi-circular edge notch
f␾ = [(a/c)2cos2 ␾ + sin2 ␾]1/4
For a/c > 1
for the geometry shown in Figure 30, is given by the M1 = (c/a)1/2(1.04 − 0.04c/a)
following equations: f␾ = [(c/a)2sin2 ␾ + cos2 ␾]1/4

K=S 冪 冉
␲a a a c c r r
F , , , , , ,␾
⍀ c t r w t w 冊 When the surface-crack length, 2a, reaches the sheet
thickness, 2t, the stress intensity factors for a through
crack emanating from a semi-circular notch subjected
where F is the boundary correction factor. to remote uniform stress or uniform displacement is

冉 冊
The shape factor ⍀ is given by
c c r
⍀ = 1 + 1.464(a/c)1.65 for a/c ⱕ 1 K = S√␲cFn , ,
w r w
⍀ = 1 + 1.464(c/a)1.65 for a/c > 1
The boundary correction factor Fn is
The boundary correction factor equation for a semi- Fn = f1g4fw
elliptical surface crack located at the center of a semi-
circular edge notch subjected to remote uniform stress where g4 and fw are as given before. The function f1
or uniform displacement is is given by
F = [M1 + M2(a/t)2 + M3(a/t)4]g1g2g3g4g5f␾fw f 1 = 1 + 0.358␭ + 1.425␭2 − 1.578␭3 + 2.156␭4

M1 = 1
M2 = 0.05/[0.11 + (a/c)3/2] ␭ = 1/(1 + c/r)