Documente Academic
Documente Profesional
Documente Cultură
Title Stress intensity factors for elliptical surface cracks in tensile stepped shaft
Abstract
This work evaluates and analyses the mode I stress intensity factor (SIF) of semi-elliptical surface cracks in tensile
stepped shaft using the dual boundary element package of BEASY. Large specimen scope of 400 geometry cases is
covered. A notch shape factor fs for stepped shaft is defined by adopting the crack mid-point or the deepest point SIF
of the smooth shaft as the common denominator; results show this choice of factorization essentially annuls the
complexity of the variation of SIF in the smooth shaft as crack varies sans the geometrical irregularities of stepped
shaft, and thus allowing effective and insightful study on the notch effects of the stepped geometry. Guided by poring
the big data of 400 cases, it is concluded that a crack classification of deep, shallow, wide, and narrow is instrumental
to narrate fs as notch varies and as crack varies. As notch varies, the effect of fillet radius is secondary unless it is
relatively sharp. Deep cracks shows less sensitivity to notch effects, with deep and wide crack approaching notch
invariant first as depth increases. Shallow and narrow crack displays pronounced notch effect at the crack surface
point. As crack varies, the effect on fs at the surface point is more pronounced than at the deepest point, while the fs at
the deepest point is predominantly affected by crack depth with the exception of very wide crack. The extensive results
of the notch shape factor fs enables the estimation of stepped shaft SIF from solely obtaining the smooth shaft SIF of
the corresponding crack geometry, rendering it useful for failure analysis and the study of crack in general.
Keywords stress intensity factor; semi-elliptical surface crack; tensile stepped shaft; notch
shape factor
Order of Authors Sharon Teh, Haw Ling Liew, Judha Purbolaksono, Devi Chandra, Dhamodaran
Kumar, Andri Andriyana
To view all the submission files, including those not included in the PDF, click on the manuscript title on your EVISE
Homepage, then click 'Download zip file'.
Dear Professor,
Greetings!
I am enclosing the manuscript “Stress intensity factors for elliptical surface cracks
in tensile stepped shaft” for consideration of publication in Engineering Failure
Analysis.
This work is original, and this manuscript is not under consideration for
publication elsewhere.
On the behalf of all authors, I declare that we have no financial and personal
interest on this paper that could potentially cause conflict of interests.
Sincerely,
Haw-Ling Liew
Stress intensity factors for elliptical surface cracks in tensile stepped shaft
Graphical display of tensile stepped shaft SIF of 120 cases for semi-elliptical cracks.
Effective representation of SIF of stepped shafts by way of a notch shape factor fs.
The fs tends to more or less notch invariant at a smaller depth for deep and wide
crack.
The fs is very pronounced for the surface point of shallow and narrow cracks.
The effect of diameter ratio on fs is predominant over the fillet radius except for sharp
notches.
Padang, Indonesia.
d
LaFarge Malaysia Berhad, Rawang 48000, Malaysia
e
Department of Mechanical Engineering, Faculty of Engineering, University of Malaya, Kuala
Abstract
This work evaluates and analyses the mode I stress intensity factor (SIF) of semi-elliptical
surface cracks in tensile stepped shaft using the dual boundary element package of BEASY.
Large specimen scope of 400 geometry cases is covered. A notch shape factor f s for
stepped shaft is defined by adopting the crack mid-point or the deepest point SIF of the
smooth shaft as the common denominator; results show this choice of factorization
essentially annuls the complexity of the variation of SIF in the smooth shaft as crack varies
sans the geometrical irregularities of stepped shaft, and thus allowing effective and insightful
study on the notch effects of the stepped geometry. Guided by poring the big data of 400
cases, it is concluded that a crack classification of deep, shallow, wide, and narrow is
instrumental to narrate f s as notch varies and as crack varies. As notch varies, the effect of
fillet radius is secondary unless it is relatively sharp. Deep cracks shows less sensitivity to
notch effects, with deep and wide crack approaching notch invariant first as depth increases.
1
Corresponding author: H.L. Liew; E-mail: hlliew@um.edu.my, tel.: +603-79676840
1
Shallow and narrow crack displays pronounced notch effect at the crack surface point. As
crack varies, the effect on f s at the surface point is more pronounced than at the deepest
point, while the f s at the deepest point is predominantly affected by crack depth with the
exception of very wide crack. The extensive results of the notch shape factor f s enables the
estimation of stepped shaft SIF from solely obtaining the smooth shaft SIF of the
corresponding crack geometry, rendering it useful for failure analysis and the study of crack
in general.
Keywords: stress intensity factor, semi-elliptical surface crack, tensile stepped shaft, notch
shape factor
1. Introduction
and inevitable. The stepped geometries of these parts are mostly manufactured through
machining to attain the necessary dimensional precision. Machining flaws and fatigue
surface flaws or microcracks in such components frequently get induced at the location with
the highest stress concentration, namely the tapering circumferential area of the shoulder of
the stepped shaft. Left unchecked, these flaws become surface cracks and propagate
towards the core of the shaft and ultimately cause abrupt fracture that may incur serious
machine damage and other significant tangible costs. Manufacturers at large, as part of Best
Practices and Total Quality efforts, inspect and check components for defects such as
surface cracks and other defects before they are released for use. During service life,
scheduled inspections are also carried out to ensure the components’ safety fitness for
continued use. In these processes, accurate knowledge of stress intensity factors (SIFs)
requisite for crack growth using the damage tolerance approach would lead to an enhanced
understanding of the surface fatigue crack behavior, and is instrumental to ultimately make
2
Generally, fatigue failures in power shafts have origin on surface cracks that grow with
nearly semi-elliptical shape under cyclic bending, mode I, combined with steady torsion [1].
Large number of power plant systems run with a general steady torsion combined with cyclic
bending stress either due to the self-weight bending during the rotation or possible
misalignment between journal bearings [2]. In the case of power shafts such as those used
in electric power plants, propeller shafts of screw ships, or any other rotary load-transmission
devices, the lifetime spent between crack initiation and final fracture is of capital importance
to improve the inspection intervals and maintenance procedures [1] . In the recent years, in
documented fatigue failures of the roller support shaft for the rotary kiln used in the
observed in the fracture surface of the roller shaft support. The support shaft has multiple
zoned stepped geometry; one such is to facilitate the fitting of the roller using shrinkage. One
of the major failure locations reported is the stress concentrated curvature zone of the
As society at large, strive to transit from fossil fuel dependence towards renewable energy,
turbine use will increase in future. There have been reported cases of engineering failure
analysis on turbines from hydropower plants which fracture failure occurred during its service
life at the critical cross-sectional area where there was a shaft tapering or geometry
irregularity [4,5]. It is also reported that information about cracks and failures of rotor shafts
is generally kept confidential by the plant management and by the machine manufacturer;
therefore very few cases have been reported and analyzed in literature, especially in recent
years [5,6].
formidable complexity and challenges to the analytical treatment of the problem fracture
mechanics, or more specifically the problem of linear elastic fracture mechanics in relation to
the determination of the SIFs due to flaws. Nearly all such analysis in the literature has been
various kinds, e.g. circumferential notch of various types; and stepped geometry that
3
corresponds to a raised portion of the shaft with fillet to avoid abrupt corner. Stepped
applications include flywheel and the roller support shaft for rotary kiln and other devices. By
and large, the complexity attributed to such irregularity can be sized up by considering the
has two-fold symmetry of the stress field about the axial axis and the transverse plane; and
stress field for stepped geometry is asymmetrical about the transverse plane. In the
presence of a surface crack in the transverse plane, the symmetry about the axial axis is
lost; but common notches maintain one fold symmetry, and stepped geometry has
Literature on stepped shaft SIFs is extremely limited, and to the best of the authors’
knowledge, there are only two published articles – Hojfeldt and Ostervig [7] of pure bending,
and Thompson and Sheppard [8] of normal and torsional loading. Simulation and
experimental studies of SIFs and fatigue growth were performed by employing one stepped
geometry of similar size. On the other hand, SIFs of surface cracks in common notches and
threaded bars has received much more research attention. Carpinteri, et al. [9] computed
SIFs on a circular-arc circumferential notch in a round bar in both tension and bending for
specimens of various stress concentration factors. Guo et al. [10] investigated SIFs of
surface cracks in round bars with various types of notches. Lefort [11] presented semi-
empirical SIF results for circumferential crack emanating from a notch. Toribio et al. [12]
studied circumferential notch of different severity for solid and hollow round bars. The
evolution of the crack shape for circularly notched round bar was studied by Lin and Smith
[13,14]; however, SIF data wasn’t presented. Notches in the form of threaded bars were
studied by Nord and Chung [15], Toribio et al. [16,17], and James and Mills [18].
Of relevance and great interest is the research of surface crack SIFs in smooth bar, as the
stepped bar semi-elliptical crack SIFs in this work would be represented in the form that
modulo the corresponding smooth counterpart. Large amount of work are available in the
literature, we cite some of these work on normal mode SIF [9,18-34]. Among these, Daoud
4
et al. [21], Athanassiadis et al. [23], Wilhem et al. [24], Lorentzen et al. [26], Salah and
Lovegrove [22], Shin and Cai [33] contain experimental results. Most of the references in
[9,18-34] either tabulated or presented a fitted polynomial to depict the behavior of SIF as
crack depth varies for the point of maximum crack depth; and some with additional results
for the point of intersection of the crack with the free surface. Notably, Athanassiadis et al.
[23] includes crack front SIFs of a large set of crack geometry calculated using
The aim of this present work is to numerically investigate and analyze the mode I SIF for
considered. All simulation are performed using the dual boundary element method (DBEM)
of the fracture mechanics package of BEASY [35]; its validity has been variously confirmed
for problems described in [36-40]. The utilization of DBEM enables a large number of cases
to be simulated in a relatively short time with comparable accuracy to FEM, enabling the
collection of big data and qualitative establishment of trends and effects of different
2. Simulation
Stress intensity factors (SIFs) are the asymptotical value of the local stress field as the
material point 𝐱′ approaches the crack front. Adopting the coordinate system in Fig. 1, the
These SIFs are evaluated in BEASY by way of the J-integral concept of Rice [41] and
Cherepanov [42] which gives, for crack opening in the 𝑥𝑖 direction, a path-independent
5
energy integral of the form 𝐽 = ∫Γ �𝑊𝑛𝑖 − 𝑡𝑘 𝑢𝑘,𝑖 �dΓ over a surface Γ with outward normal n.
This concept was developed for linear elastic materials, and it was further extended to HRR
solutions [43] for materials with constitutive relationship in the form of Ramberg-Osgood.
Following the presentation in Cruse [44], and Mi and Aliabadi [45], and by using Green’s
functions 𝑈𝑖𝑗 for displacement and 𝑇𝑖𝑗 for traction, the strain energy density 𝑊(Γ) and the
work-conjugate of traction t and displacement u in J-integral are calculated using the dual
1 ′)
𝑡 (𝐱 + 𝑛𝑖 (𝐱 ′ ) ∫Γ 𝑇𝑖𝑗,𝑘 (𝐱 ′ , 𝐱)𝑢𝑘 (𝐱)dΓ(𝐱) = 𝑛𝑖 (𝐱 ′ ) ∫Γ 𝑈𝑖𝑗,𝑘 (𝐱 ′ , 𝐱)𝑡𝑘 (𝐱)dΓ(𝐱). (2)
2 𝑗
Leading to Eqn (1-2) is the Somigliana’s identity expressed as an integral equation involving
𝑇𝑖𝑗 and several other terms as its kernel. 𝑇𝑖𝑗 ’s singularity of 𝛰 �1�‖𝐱 � as 𝐱 → 𝐱′
− 𝐱 ′ ‖2
warrants regularization and treatment in the sense of Cauchy; and 𝛼𝑖𝑗 (𝐱 ′ ) is a term that
emanates from this process. Mi and Aliabadi [45] claimed to have presented an effective
numerical implementation of the dual boundary integrals of Eqn (1-2), and this technology is
the basis of the dual boundary element fracture mechanics package of BEASY.
As the fracture mechanics package of BEASY is relatively less used for research, we would
firstly validate the software by way of comparison with the published results of Newman and
6
Raju [28]. For this purpose, smooth cylinder specimens of 10 mm diameter, subjected to
uniform tensile stress of 100 MPa (below yield stress for steel) with semi-elliptical surface
crack of various crack depth ratio (a/d) and various crack aspect (a/c) were simulated on
BEASY. In order to investigate the influence of shaft geometry and semi-elliptical crack ratio
on the SIFs of surface crack on the crack front, twenty sets of stepped shaft geometric
parameters were modeled as listed in Table 1. Four different larger diameter, D over smaller
diameter, d ratio (D/d), were selected based on their practicality in shaft design and are
aligned to commonly found design chart values for stress concentration factor guides [46].
The value of d is kept constant at 10 mm for all specimens. For every D/d variation, five
different shoulder fillet radius, r ranging from r/d of 0.05 to 0.25 were explored as shown in
Table 1, generating a total of twenty different geometry specimen cases. The surface crack
applied on each shaft geometry comprised of twenty crack variations through selecting four
different crack depths with a/d varying from 0.025 to 0.1 and five aspect ratio or crack width
of a/c; the aspect ratio selected was guided by Corn [47]. Thus, when the 20 crack variations
were applied on the 20 specimen geometries, a total of 400 cases were generated for this
The model consists of a one zoned stepped shaft, with all surfaces defined to be outward
normal direction. The material has a modulus of elasticity of 200 GPa and a Poisson’s ratio
of 0.3. Two-dimensional quadratic elements were used to discretize all surfaces. The free
end of the smaller diameter (d) cylinder of the stepped shaft is applied with a nominal
tension stress of 100 MPa. The total length (L) of every specimen is 100 mm, giving a L/d
A “thumbnail” semi-elliptical crack was introduced on the root of the fillet radius as illustrated
in Fig. 2. This is justified as it is the location with the highest stress concentration as
concluded by Timoshenko [48], and this is further confirmed from our simulation of the stress
distribution for a flawless stepped shafts in tension. We acknowledge the more recent work
of Tipton [49] reveals that the maximum stress exact location is on the curvature zone,
7
slightly away from the root; the marginal differences are insignificant for the purpose of this
study.
Geometry of
Location of thumbnail thumbnail crack
surface crack 2c a
Clamped
Axial normal loading
z
x
Fig. 2. Geometry and boundary conditions of the meshed model with a thumbnail/semi-
elliptical crack on the cross-section with the maximum stress concentration
Table 1
8
Table 2
count of all the simulation cases is plotted on the ordinate against the ratio of average
distance between nodes on crack front over the total crack front length (Fig. 3). All cases
have the ratio of less than 4%, translating to at least 25 discretization nodes or more for all
crack fronts considered. At the point where the crack edge intersects the free surface, i.e.
the CPE, BEASY’s provides for a built-in algorithm for automated (without allowing user’s
Count
160 151
140
120
100
80
80
56
60 51 49
40 33
20
0
1.0-1.5% 1.5-2.0% 2.0-2.5% 2.5-3.0% 3.0-3.5% 3.5-4.0%
9
3. Numerical results and discussion
The value of normalized SIF, K is plotted on the ordinate against the crack front expressed in
the form of ratio of parametric angle (0 to 180 degrees). The parametric angle of the semi-
elliptical crack front, φ is defined as shown in Fig. 4, and a and c are respectively the
principal axes of the ellipse, giving the crack geometrical aspect ratio of a/c. We define the
two important points along the crack front, Point A represents the deepest point of the ellipse
(DPE) or the midpoint of the semi elliptical crack front and Point B represents either of the 2
endpoints which the crack intersects the free surface or also known as the corner point of
3.1 Smooth cylinder results and BEASY validation using NR’s results
The simulated SIF value, KBeasy is converted to the normalized SIF, K and is defined as
follows:
𝐾𝐵𝑒𝑎𝑠𝑦
𝐾 =
𝜎√𝜋𝑎
Finite element simulation results of surface cracks on tensile rod by Newman and Raju (NR)
[28] is used for validation of the lesser used fracture mechanics package of BEASY. In NR’s
work, the crack front profile is non-elliptical and is attained through conformal transformation
of a semi elliptical crack front in a rectangular bar. The crack ratio used is a/(s/2) (see Fig. 4)
where s/2 is one half arc crack length. Reported results were limited to a handful of profiles
only.
10
(A)
(B)
(DPE)
(CPE)
To obtain a proper comparison, we converted our crack ratio parameter a/c to the
corresponding ratio of crack depth over crack arc width, 2a/s (as illustrated earlier under Fig.
4). The comparison of the specimen with crack depth ratio a/d equal to 0.2 for three most
closely matched crack ratio cases is presented in Fig. 5 and shows close agreement and
similar plot trends between the two independent simulation works. Fig. 6 shows the
comparative crack front profiles of the cases compared for a better appreciation of the small
crack geometrical differences of the closest matched cases compared. We focus on the
maximum SIF values of the three sets of comparative cases, Point A (DPE) for the low crack
ratio case and Point B (CPE) for high crack ratio cases. For high crack ratio with significant
maximum SIF at Point B where crack ratio is 0.99 and 1.0 for BEASY and NR results
respectively, the SIF value is close at 0.843 and 0.803 respectively (Table 3). For BEASY
and NR crack ratio of 0.76 and 0.8 respectively, the maximum SIF value at Point B is very
close at 0.863 and 0.846 respectively. For low crack ratio of 0.57 and 0.6 for BEASY and NR
11
respectively, the maximum SIF is at Point A, with values of 0.998 and 1.031. The marginal
differences can be attributed to the different way the crack front shape is derived between
the BEASY and NR’s simulation (Fig. 6). We remark that the NR SIF data was manually
extracted.
Newman and Raju [28] reported that when 2a/s was 0.6, the maximum normalized SIF was
at maximum depth point and when crack ratio, 2a/s was 0.8, the SIF curves for all a/d
explored trended towards uniformity along the crack front. The simulation results herein
reflect a close agreement to this observation, whereby for 2a/s of 0.57 of the smooth shaft,
the SIF curve has a centre peak value, for 2a/s equal to 0.76, the SIF was uniform along the
crack front and for a/c equal to 0.99, the SIF curve shape has inverted to minimum value at
the centre of crack front, denoting that the curve shape had transited gradually with the
1.1 A
B
0.9
Noramlized K
0.8
0.7
0.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2 φ/π
Fig. 5. Comparison of mode I SIFs for semi-elliptical crack on round bar by BEASY and NR’s
FEM simulation [28] for crack depth ratio (a/d) equal to 0.2.
12
0.57
0.6
0.8 0.76
1.0 0.99 _____ BEASY
- - - Newman-Raju
Fig. 6. Comparison of crack front geometry of BEASY and NR’s FEM simulation [28] of
crack depth ratio equal to 0.2.
Table 3
Comparison of maximum and minimum mode I SIFs on round bar by BEASY and NR’s FEM
simulation [28] for crack depth ratio (a/d) equal to 0.2.
13
3.2 Stepped shaft results and its effects on SIF
For this section, the simulated SIF value, KBeasy is converted to the normalized SIF, K as
follows:
𝐾𝐵𝑒𝑎𝑠𝑦
𝐾 =
𝜎�𝑑/2
The results for stepped shafts will be presented as a plot of normalized SIF against the
normalized coordinate along the crack front length (ratio of crack front length of SIF point
over the total crack length), hence the two points at free surface (referred to in the earlier
sections as Point B) are 0 and 1 respectively on the abscissa in this representation and Point
A remains 0.5 as it is in the middle of the crack and half crack length from either free surface
point. This presentation of results would facilitate easy comparison of SIF values at Point B
and also to identify the trends of gradual transition on SIF curves from minimum SIF to
maximum SIF at Point B due to the changes of shaft geometry. The crack aspect ratio and
crack depth curves of this study are close in range with other authors, e.g. Carpinteri et
al.[9].
For ease of discussion hereon, the term “notch” shall refer to any geometry discontinuity or
which as a parameter is expressed as the shaft diameter ratio D/d and/or the fillet radius
applied at the shaft tapering, which is represented as a parameter of fillet radius ratio r/d. It is
acknowledged that the higher the shaft diameter ratio and the lower the fillet radius ratio, the
higher is the geometry irregularity and thus will be collectively termed as the ‘shaft notch
SIF value range for the stepped shafts is generally slightly higher than for the smooth
cylinder. Comparing for all stepped shaft geometry for the three cases of a/c crack ratio of
less or equal to unity (a/c < 1), shaft with the most gradual geometry change or least
irregularity (Case: D/d = 1.1, r/d = 0.25) displayed normalized SIF value curves closest in
shape and magnitude range to that of the smooth shaft for the corresponding a/c crack ratio
studied. On the other extreme end, the shaft with the sharpest geometry irregularity (Case:
14
D/d = 2.0, r/d = 0.05) showed the largest SIF curve shape and magnitude trend deviation
For stepped shafts applied with surface crack ratio a/c of 0.33, for smaller D/d shaft diameter
ratio of 1.1 and 1.2, the SIF curves follow the shape of the smooth cylinder, with maximum
normalized SIF value at the maximum crack point depth, where parametric angle, φ is equal
to 90 degree. The increase of D/d and the decrease of fillet radius ratio r/d are two
parameters that independently influence the SIF curves to trend towards more uniform
values along the crack front. The effect of SIF uniformity along crack front is more
pronounced as the crack size ratio a/d increases. In addition to that, the SIF values will also
see an increase in average magnitude for the smaller a/d ratio cases but less substantial SIF
15
Normalized SIF vs Normalized Crack
Length
for a/c =0.33
Legend
Fig. 7. Typical results of normalized K I vs normalized crack front length (a/c = 0.33)
16
For stepped shaft with surface crack aspect ratio a/c equal to 0.5, again the shaft notch
effect parameters (D/d increase, r/d decrease) influence the SIF curves in a similar trend as
shown in Fig. 8. For smooth shaft, large crack size ratio a/d equal to 0.1, the gradient of SIF
curve for surface crack ratio a/c equal to 0.5 is gentler than for a/c equal to 0.33. Hence, the
notch effect parameters influence the SIF curves to transcend quicker into uniformity and
subsequently invert the SIF curve shape into marginal maximum SIF values at free edges (φ
equal to 0 and 180 degrees) as the shaft geometry irregularity gets more intense. This is
especially apparent for large crack size ratio of a/d equal to 0.1. Generally, for all geometry
of stepped shaft in this work, for a/c equal to 0.5, all the SIF curves can be considered
relatively uniform with the range of each SIF curve limited to within 0.22 between the
maximum and minimum SIF value of a crack front. Comparing the most gentle notched shaft
of D/d equal to 1.1, with r/d of 0.25, to the most sharply notched shaft of D/d equal to 2.0,
with r/d equal to 0.05, for the case of small crack depth ratio of a/d equal to 0.025, the SIF
range for the former shaft geometry specimen is 0.42 to 0.52 and for the latter sharp
17
Normalized SIF vs Normalized Crack
Length
for a/c =0.50
Legend
Fig. 8. Typical results of normalized K I vs normalized crack front length (a/c = 0.5)
18
Fig. 9 shows the shaft notch effect parameters (D/d increase, r/d decrease) also exert the
same influence on the SIF curves for stepped shaft with surface crack aspect ratio a/c equal
to 1.0. For this smooth shaft case of a/c equal to unity, the SIF curves for all crack depth
ratio a/d display a minimum SIF value at centre of crack front, with 2 maximum SIF values at
the free edges of the crack front. Hence, in the same trend, the notch effect parameters
influence the SIF curve to amplify the maximum SIF values at free edges (φ equal to 0 and
180 degrees) as the shaft geometry irregularity gets more intense. Again, this is most
apparent for large crack size ratio of a/d equal to 0.1. Hence, for a/c of 1.0, the increase of
maximum SIF at the two free edges of crack and the increase of average SIF magnitude
increases with the crack depth ratio a/d when influenced more heavily by the two shaft notch
effect parameters. The same influence of shaft notch effect parameters on SIF curves are
seen for all stepped shaft geometries for crack aspect ratio a/c exceeding unity (a/c = 1.5
19
Normalized SIF vs Normalized Crack
Length for a/c =1.0
Legend
Fig. 9. Typical results of normalized K I vs normalized crack front length (a/c = 1.0)
20
Normalized SIF vs Normalized Crack
Length for a/c =1.5
Legend
Fig. 10. Typical results of normalized K I vs normalized crack front length (a/c = 1.5)
21
Normalized SIF vs Normalized Crack
Length for a/c =2.0
Legend
Fig. 11. Typical results of normalized K I vs normalized crack front length (a/c = 2.0)
In summary, the shaft notch effect parameters of D/d increase and r/d decrease, will drive
any given crack size or geometry (any a/d or a/c) to have SIF crack front values behavior
closer to that of a relatively sharper crack (crack with higher a/c ratio) and also have a higher
22
3.3 Notch Shape Factor (f s ) for Stepped Shaft
Based on the SIF results above, for any given crack geometry case (fixed a/c and a/d), the
SIF for the deepest point of crack (DPE or Point A) of the smooth specimen K A Smooth is used
We designate this f s as the notch shape factor for the reason to be explicated later. The
value of the common denominator of K A Smooth in Eqn (3) is listed in Table 4. The resulting SIF
plots for each stepped shaft geometry case would then be a representation of solely the
notch effects on the crack front SIF curve shape, independent of the changes in the
magnitude of K A Smooth that is caused by the variation of crack in smooth shaft sans the
geometrical irregularities of the stepped shaft; this allows for proper and effective analysis to
gain insights of the notch effects. Furthermore, the use of this common denominator
enhances the usefulness of this study as the plots can be used to estimate the SIF value of
any stepped shaft geometry (within the notch geometry range of this study) by just obtaining
the sole value of the crack front mid-point SIF of the smooth specimen for a wide range of
crack geometry of a/c between 0.33 to 2.0 and a/d between 0.025 to 0.1.
The graphical display of the effects of the stepped shaft on the notch shape factor f s are
presented in Fig. 12-Fig. 16 for crack ratio a/c of 0.33, 0.50, 1.0, 1.5 and 2.0 respectively for
all crack depth ratio a/d studied. As shown in these plots, for all a/c ratio of the smooth
It is observed that for all smooth shafts (i.e D/d = 1), all notch shape factor f s are but one
except for low a/c where some small divergence is seen at the small neighbourhood of the
CPE (Point B) region, suggesting appropriate and effective factorization choice that removes
nearly all crack geometry effects on the smooth shaft. As such, the designation of “notch”
The notch effects are greatest in the smallest crack size, a/d equal to 0.025 as seen from the
greatest deviation of the f s curve from the smooth shaft, for any given shaft geometry.
23
Consistent with the influence of notch factors on SIF reported earlier above, the f s curve also
deviates away from that of the smooth specimen with the increase of notch factors (D/d
increase and r/d decrease) of the shaft for all crack geometry studied. We will move beyond
the rudimentary exposition by considering the sensitivity of f s with respect to notch variation
and crack geometry variation in Sec 3.4 and Sec 3.5 respectively.
Table 4
Smooth specimen K A Smooth values used to obtain the notch shape factor. The nominal stress
K A Smooth (K at φ=π/2)
a/d a/c
MPa√mm
0.025 0.330 86.7758
0.050 0.330 121.7487
0.075 0.330 149.5671
0.100 0.330 171.5176
0.025 0.500 78.4399
0.050 0.500 111.4396
0.075 0.500 137.8716
0.100 0.500 158.8101
0.025 1.000 58.7855
0.050 1.000 83.5311
0.075 1.000 102.9630
0.100 1.000 119.9358
0.025 1.500 46.1643
0.050 1.500 65.5154
0.075 1.500 80.9943
0.100 1.500 94.4650
0.025 2.000 38.4916
0.050 2.000 54.5922
0.075 2.000 66.9218
0.100 2.000 77.6838
24
Crack ratio, a/c =0.33
Legend
25
Crack ratio, a/c =0.50
Legend
26
Crack ratio, a/c =1.0
Legend
27
Crack ratio, a/c =1.5
Legend
28
Crack ratio, a/c =2.0
Legend
29
3.4 Crack front Shape Factor (f s ) Behavior with Respect to Notch Variation
From the typical plots for shape factor (f s ) in Fig. 12-Fig. 16, we can now summarize the key
observations of crack front f s behaviour with respect to notch variation (discussed in this
section), and crack geometry variation (in the next section). From the viewpoint of
𝛿𝑓𝑠 ⁄𝛿(𝐷⁄𝑑 , 𝑟⁄𝑑 ) and 𝛿𝑓𝑠 ⁄𝛿(𝑎⁄𝑑 , 𝑎⁄𝑐 ) where 𝛿 is the variation notation.
For ease of discussion hereon, the Most Severely Notched (MN) shaft specimen (D/d = 2.0,
r/d = 0.05) and Least Notched (LN) shaft specimen (D/d = 1.1, r/d = 0.25), will be denoted by
MN and LN respectively.
We also introduce the expression f s,P Q where subscript P represents either Point A or Point
example, f s,A MN represents the shape factor for Point A of the most severely notched stepped
shaft specimen.
After careful study of results by way of multiple raster displays of all 400 f s plots and the plot
of f s values at DPE against f s values at CPE (f s,A ,vs f s,B ) shown in Fig. 17 which includes all
120 cases displayed graphically in Sec 3.3, we conclude that the behaviour requires a
description using four distinctive groups of crack. Fig. 18 gives an overview of the shape
dependence of f s on notch parameters. In this figure, the vertical axis represents the two
variations of a/d namely shallow (a/d < 0.05) and deep (a/d > 0.05). On the other axis, the
variation of a/c is divided into wide (a/c < 1.0) and narrow (a/c > 1.0). Hence, the four crack
To study the sensitivity of crack front SIF with respect to the notch factors, we correlate the
f s results of the four distinct groups of cracks (from Fig. 18) with the shaft notch effects(D/d,
r/d) by way of plotting the f s values against D/d for the various r/d shown in Fig. 19-Fig. 20.
In these plots, the corresponding f s value of the smooth round bar is indicated by a
horizontal dotted line, for an overview comparison of the deviation of f s of the notched
sensitivity.
30
Thus, the primary findings and their associated supporting evidence from Fig. 18 -Fig. 20
are:
i) Deep crack (high a/d) shows less sensitivity to notch effects (both D/d and r/d)
than shallow crack (low a/d). Hence, the shallower a crack, the higher the
a. This is deduced from comparing the deep crack plots to the shallow crack
plots in Fig. 19-Fig. 20. Table 5 summarizes pertinent f s values of these plots
behavior. For deep crack, f s at Point A (DPE) remains close to 1.0, which is
the value of f s of the smooth round bar. For shallow crack, f s at DPE shows
comparatively a little more increase from the f s of the smooth bar, indicating
higher notch sensitivity. When we refer to Table 5, the effect is even more
significant for f s at Point B (CPE), where for deep crack the range in increase
is between 0.20 to 0.52 (ratio of 1.29 to 1.33), while for shallow crack the
b. This is also supported by to Fig. 18(a) and (c), where for shallow crack, the
value of f s is observed to increase sizeably, while for deep crack shown in Fig.
ii) For deep and wide crack, f s appears to be nearly notch invariant at a depth of
0.1, as seen from the more or less uniform distribution over the D/d and r/d
domain (Fig. 19-Fig. 20). For other crack groups, the crack depth influenced by f s
is larger; and in accordance with the observations of other notch effects in James
and Mills [18], Lefort [11], and Nord and Chung [15], we intuitively expect that as
crack depth extends, f s becomes nearly notch invariant for all cracks.
31
iii) For shallow and narrow cracks, the notch effects are very pronounced. Large
iv) D/d effect is relatively more pronounced than r/d as far as notch sensitivity is
a. From Fig. 19-Fig. 20, it is observed that r/d effect is very pronounced only for
v) The shallow and wide cracks display similar small spread of f s at both DPE and
CPE.
vi) Irrespective of crack group, f s at Point B (CPE) is more notch sensitive compared
to f s at Point A (DPE).
a. This is supported by the evidence of shape change for all four groups of crack
in Fig. 18, indicating that when notch severity increase, the f s at CPE
b. Also, refering to Fig. 19-Fig. 20, when f s at DPE and f s at CPE are compared
for any given crack, f s at CPE is relatively higher to the value of f s of the
smooth round bar and than that of the f s at DPE to the value of f s of the
smooth bar .
vii) Fig. 17 provides for, at a quick glance, the distinct spread and behavior of f s at
viii) Table 7 provides for a comparison on the quantitative spread of f s and measures
of the shape change due the MN and LN shaft for a typical crack of all crack
groups. The shape change is described in the sense of the difference and ratio of
values at the DPE and CPE. The ratio of f s MN / f s LN gives the spread of f s as notch
32
In addition to the primary findings, supplementary findings are as follows:
i) For narrow (a/c > 1.0) and deep (a/d > 1.0) cracks, the increase in notch severity
results in large increase in Point B resulting in steep curves where f s,A MN will dip
below the magnitude of f s,A LN (see Fig. 18(d) and Fig. 21). Hence, for this specific
crack domain, that f s,A MN and f s,A LN will represent the minimum and maximum f s
respectively.
ii) For very wide crack (a/c equal to 0.33), the f s MN value is distinctly higher than the
f s LN value for all crack size (a/d) and shaft notch domain. Hence, it can be
concluded that for very wide crack (a/c < 0.33), the maximum f s value is at Point
iii) For a/c between 0.5 to 1.0, throughout crack size (a/d) domain, there is a gradual
shape change from f s maximum value being at DPE (f s,A is maximum) for the
(f s,B is maximum) for the MN shaft specimen. This is evidenced from Fig. 18(a)
and (b) that for a/c = 0.5, f s,A is maximum for the LN shaft specimen whilst f s,B is
iv) Fig. 18(a) and (b) are referred. The concavity of the crack front curve shape for
crack aspect ratio of a/c = 0.5 is seen to be changing as notch severity varies.
Table 6 is an auxiliary addendum to track this change for all crack size (a/d)
v) Using a moderately notched shaft shaft of D/d = 1.5 and r/d = 0.15, and for a
typical crack geometry for each group, Table 5 provides for the magnification
factor of f s and its crack front shape in the sense of the absolute difference at the
DPE and CPE. This would supplement the qualitative description in the main
findings.
33
Fig. 17. f s at DPE vs CPE for all crack groups
34
SHALLOW (a/d <1.0) DEEP (a/d >1.0)
Low Crack ratio, a/c <1.0
WIDE
High Crack ratio, a/c >1.0
NARROW
Fig. 18. Graphical summary of shaft notch severity effect on shape of plot curves (MN
denotes most severely notched shaft specimen, LN denotes least notched shaft specimen)
35
Fig. 19. f s vs D/d for deep (high a/d) crack
36
Fig. 20. f s vs D/d for shallow (low a/d) crack
37
Table 5
Comparison of f s for moderately notched shaft (D/d=1.5, r/d=0.15) and smooth bar
Point A (DPE) Point B (CPE)
Crack group Shaft Smooth Diff. Ratio Shaft Smooth Diff. Ratio
(i) (ii) (i)-(ii) (i)/(ii) (iii) (iv) (iii)-(iv) (iii)/(iv)
Deep & Narrow
(a/d=0.1,a/c=2.0) 1.00 1.00 0.00 1.00 2.13 1.61 0.52 1.33
Deep & Wide
(a/d=0.1,a/c=0.33) 1.03 1.00 0.03 1.03 0.91 0.71 0.20 1.29
Shallow & Narrow
(a/d=0.025,a/c=2.0) 1.42 1.00 0.42 1.42 2.59 1.57 1.02 1.65
Shallow & Wide
(a/d=0.025,a/c=0.33) 1.39 1.00 0.39 1.39 1.04 0.66 0.38 1.57
Table 6
Summary of f s curve shape for a/c = 0.50 as notch varies, as addendum to Fig. 18
r/d
D/d a/d
0.05 0.10 0.15 0.20 0.25
0.025 A A A A A
0.05 B A A A A
1.1
0.075 B A A A A
0.1 B A A A A
0.025 A A A A A
0.05 B A A A A
1.2
0.075 B A A A A
0.1 B A A A A
0.025 B A A A A
0.05 B B A A A
1.5
0.075 B B B A A
0.1 B B B B A
0.025 B A A A A
0.05 B B A A A
2.0
0.075 B B B A A
0.1 B B B B A
A denotes maximum f s at Point A (DPE)
B denotes maximum f s at Point B (CPE)
38
Table 7
Comparison of f s values for most notched (MN) & least notched (LN) shaft specimen
3.5 Crack front Shape Factor (f s ) Behavior with Respect to Crack Geometry Variation
Next, we also study the f s behavior with respect to crack geometry (a/c and a/d). Thus, Fig.
21-Fig. 22 show the f s spread with relation to crack geometry for Point A (DPE) and Point B
i) The change in the f s at CPE is more pronounced than at DPE when we change
a. f s at CPE (f s,B ) is more sensitive and increases at a larger rate than f s at DPE
(denoted by the larger f s,B spread shown in Fig. 22 than the more uniform f s,A
b. Table 7 also shows that the f s at CPE is more sensitive than DPE as
1.69 for CPE and comparatively lower range of 1.01 to 1.36 for DPE.
39
ii) The f s at DPE (f s,A ) is predominantly affected by a/d (refer Fig. 21(a)), but is more
or less uniform with respect to a/c, with the exception of very wide crack, a/c
iii) The f s at CPE (f s,B ), is more pronouncedly affected by the change in a/c than by
iv) From Fig. 22, the slope of f s MN plots are steeper than the slope of the f s LN plots for
In summary, the plots established in Fig. 18-Fig. 22 represent SIF spread corresponding to
stepped shaft geometry variation for the large range of crack geometry of this work.
Qualitative description of the SIF behaviour has been narrated in details with interspersed
notch varies and as crack varies. Within the domain and common shaft geometry
configurations of the range of D/d = 1.1-2.0 and r/d = 0.05-0.25 that is drawn from Petersen’s
Handbook [46], we can conclude that this work is useful to provide a reasonable estimate
and an appreciation of the behavior of mode I SIF at DPE and CPE for any surface crack
geometry solely through obtaining the SIF mode I value of DPE for the smooth specimen; as
the notch shape factor f s is the same as the SIF modulo the SIF at the DPE for the
corresponding smooth shaft. The literature resources of [9, 18-34] provides for the SIF of
smooth shafts. The aforementioned estimate SIF spread is accomplished through referring
to Fig. 18-Fig. 22, to obtain the estimate f s spread for the given crack and stepped shaft
sample within this domain. Hence, the findings of this work may potentially be used to
provide further insight to failure analyses and the study of crack in general.
40
fs
Fig. 21. Point A (DPE) f s for most notched (MN) & least notched (LN) shaft specimen vs
crack geometry
41
fs
3.6 Comparison with Thompson and Sheppard [8] and Hojfeldt and Ostervig [7]
As mentioned in the Introduction, Thompson and Sheppard [8] and Hojfeldt and Ostervig [7]
are the only rare find on SIF of stepped shafts in the literature. Both presented simulated and
experimental results; with Thompson and Sheppard [8] on axial and torsional loading, and
Hojfeldt and Ostervig [7] on pure bending. In Thompson and Sheppard [8], an initial semi-
circular surface of approximately 0.03 in depth was introduced on a stepped of D/d = 1.396
and r/d = 0.20, and subsequent fatigue growth was studied; however, the simulation of
growth presents partly circular cracks that deviates from experimentally observed shapes of
the beach marks. Results of simulation was provided up to the depth of 0.6, and graphical
comparsion of crack shapes were illustrated for several selected depths. In Hojfeldt and
Ostervig [7], a slightly oblique semi-elliptical surface crack of depth a/d = 0.068 and width
about a/c = 0.5 was introduced on a stepped shaft of D/d = 1.35 and r/d = 0.176, and
bending moment was cyclically applied to study crack growth; simulation presents results up
to the depth of 0.6, and the crack shape was observed to grow wider in both experiment and
42
simulation. In addition, both of these papers presented a polynomial fitted SIF that can be
used to reasonably extrapolate to slightly lower crack depth. In Fig 23, we put together our
results of SIF at DPE (Point A) and that of Thompson and Sheppard [8] for axial loading, and
that of Hojfeldt and Ostervig for flexural loading [7]. Given the slight differences in the notch
geometry, results are observed to be well in agreement. We remark here that for the
particular stepped shaft shown, the SIF of notched blends into that for smooth at a non-
Fig. 23. Comparison of SIF at Point A with Thompson and Sheppard [8] and Hojfeldt and
Ostervig [7].
43
4. Conclusions
The stress intensity factor for semi-elliptical surface crack on stepped shaft of various
geometry were evaluated and analysed using the dual boundary element package of
BEASY. There is good agreement with published results of Newman and Raju [28] on similar
smooth specimen and crack domain, validates BEASY’s DBEM package used. This study
also attained enhanced understanding on the sensitivity of the SIF as notch varies and as
crack varies via the use of an effective notch shape factor of f s . The findings of this work
that adopts the DPE value of smooth shaft as the common denominator.
• Deep crack shows less sensitivity to notch effects than shallow crack.
• The f s tends to be nearly notch invariant at the smallest depth for deep and wide
crack.
• With the exception of relatively sharp fillet radius, the effect of diameter ratio on f s is
predominant.
• The notch effect is very pronounced for CPE of shallow and narrow cracks, showing
large change of f s at CPE. Notch effect caused by fillet radius change is also very
• For shallow and wide crack, f s at CPE and DPE display similar spread as notch
varies.
• The effect of crack geometry on f s at CPE is more pronounced than at DPE for a
given notch.
• For a given notch, the f s at DPE is predominantly affected by crack depth, but almost
• The f s at CPE is more pronouncedly affected by the crack width than the depth for a
given notch.
44
Acknowledgement
The authors gratefully acknowledge the partial support from UM-BKP Grant
(UM.TNC2/IPPP/PPGP/638/BK007-2015).
References
[1] da Fonte M, Reis L, de Freitas M. The effect of steady torsion on fatigue crack growth
under rotating bending loading on aluminium alloy 7075-T6. Fracture and Structural
[2] Fonte M, de Freitas M. Marine main engine crankshaft failure analysis: a case study,
[3] Kumar D, Kiln support roller shaft crack propagation rate estimation, MS Thesis,
[4] Urquizaa G, Garcíaa JC, Gonzáleza JG, Castroa L, Rodrígueza JA, Basurto-Pensadoa
MA, Mendozab OF. Failure analysis of a hydraulic Kaplan turbine shaft. Eng Fail Anal
2014;41:108-17.
[5] Momcilovic´ Dejan, Odanovic´ Zoran, Mitrovic´ Radivoje, Atanasovska Ivana, Vuherer
Tomazˇ. Failure analysis of hydraulic turbine shaft. Eng Fail Anal 2012;20:54–66.
[6] Bachschmid N, Pennacchi P, Tanzi E. Cracked rotors – a survey on static and dynamic
[7] Hojfeldt E, Ostervig CB. Fatigue crack propagation in shafts with shoulder fillets. Eng
[8] Thompson KD, Sheppard SD. Stress intensity factors in shafts subjected to torsion and
[9] Carpinteri A, Brighenti R, Vantadori S. Surface cracks in notched round bars under
[10] Guo W, Shen H, Li H. Stress intensity factors for elliptical surface cracks in round bars
45
[11] Lefort P. Stress intensity factors for a circumferential crack emanating from a notch in a
[12] Toribio J, Matos JC, Gonzalez B. Notch effect on the stress intensity factor in tension-
[13] Lin XB, Smith RA. Fatigue growth simulation for cracks in notched and unnotched round
[14] Lin XB, Smith RA. Shape evolution of surface cracks in fatigued round bars with a
[15] Nord KJ, Chung TJ. Fracture and surface flaws in smooth and threaded round bars. Int
J Fract 1986;30:47-55.
[16] Toribio J, Sanchez-galvez V, Astiz MA, Campos JM. Stress intensity factor solutions for
a cracked bolt under tension, bending and residual stress loading. Eng Fract Mech
1991;39:359-71.
[17] Toribio J. Stress intensity factor solutions for a cracked bolt loaded by a nut. Int J Fract
1992;53:367-85.
[18] James LA, Mills WJ. Review and synthesis of stress intensity factor solutions applicable
[19] Lin XB, Smith RA. Shape growth simulation of surface cracks in tension fatigued round
[20] Couroneau N, Royer J. Simplified model for the fatigue growth analysis of surface
[21] Daoud OEK, Cartwright DJ, Carney M. Strain-energy release rate for a single-edge-
[22] Salah el din AS, Lovegrove JM. Stress intensity factors for fatigue cracking of round
1981;17(6):553–66.
46
[24] Wilhem D, Fitzgerald J, Carter J, Dittmer D. An empirical approach to determining K for
[25] Nisitani H, and Chen DH. Stress intensity factor for a semi-elliptic surface crack in a
[26] Lorentzen T, Kjaer NE, Henriksen TK. The application of fracture mechanics to surface
[27] Forman RG, Shivakumar V. Growth behavior of surface cracks and fatigue crack
propagation behavior in hollow cylinders. Fract Mech ASTM Spec Tech Publ
1986;905:59-74.
[28] Raju IS, Newman JC. Stress-intensity factors for circumferential surface cracks in pipes
[29] Astiz MA. An incompatible singular elastic element for two- and three-dimensional crack
[30] Carpinteri A. Elliptical-arc surface cracks in round bars. Fatigue Fract Eng Mater Struct
1992;15(11):1141–53.
[31] Levan A, Royer J. Part-circular surface cracks in round bars under tension, bending and
[32] Carpinteri A. Shape change of surface cracks in round bars under cyclic axial loading.
[33] Shin CS, Cai CQ. Experimental and finite element analyses on stress intensity factors of
elliptical surface crack in a circular shaft under tension and bending. Int J Fract
2004;129:239–64.
[34] Toribio J, Álvarez N, Gonzalez B, Matos JC. A critical review of stress intensity factor
solutions for surface cracks in round bars subjected to tension loading. Eng Fail Anal
2009;16:794–809.
[35] BEASY Release 10.0r14, BEASY Ashurst Lodge, Ashurst, Southampton SO40 7AA,
47
[36] Chandra D, Purbolaksono J, Nukman Y, Liew HL, Ramesh S, Hamdi M. Fatigue crack
growth of a corner crack in a square prismatic bar under combined cyclic torsion-tension
[37] Chandra D, Purbolaksono J, Nukman Y, Liew HL, Ramesh S, Hassan MA. Fatigue
growth of a surface crack in a V-shaped notched round bar under cyclic tension. J.
[38] Chandra D, Putra IS, Ariffin AK, Nukman Y, Mardi NA, Purbolaksono J. Fatigue growth
analysis of a surface crack in a solid cylinder under combined cyclic axial-torsion loading.
[39] Joseph RP, Purbolaksono J, Liew HL, Ramesh S, Hamdi M. Stress intensity factors of a
corner crack emanating from a pinhole of a solid cylinder. Eng Fract Mech 2014;128:1–
7.
[40] Zhou D., Liew HL, Purbolaksono J., Andriyana A, Chong WT. Stress intensity factors for
embedded cracks within torsionally loaded square prismatic bars. Adv Mech Eng 2019;
11(4):1-11.
[41] Rice JR. A path independent integral and the approximate analysis of strain
[42] Cherepanov GP. Crack propagation in continuous media. J Appl Math Mech 1967;
31(3):503-512.
[43] Hutchinson JW. Singular behavior at the end of a tensile crack in a hardening material,
[44] Cruse TA. Boundary-integral equation formulation and solution. Bound Elem Anal Comp
[45] Mi Y, Aliabadi MH. Dual boundary element method for three-dimensional fracture
[46] Pilkey WD, Pilkey DF. Peterson's stress concentration factors. 3e, Hoboken, NJ: John
48
[47] Corn DL. A study of cracking techniques for obtaining partial thickness cracks of pre-
[49] Tipton SM, Sorem JR, Rolovic RD. Updated stress concentration factors for filleted
49
Dear Professor,
On the behalf of all authors, I declare that we have no financial and personal
interest on the manuscript
“Stress intensity factors for elliptical surface cracks in tensile stepped shaft”
Sincerely,
Haw-Ling Liew