JID: ADES

ARTICLE IN PRESS [m5G;February 22, 2016;14:45]

Advances in Engineering Software 0 0 0 (2016) 111

Contents lists available at ScienceDirect

Advances in Engineering Software

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

Numerical fatigue assessment of welded and HFMI-treated joints by

notch stress/strain and fracture mechanical approaches
Martin Leitner, David Simunek, Syed Faheem Shah, Michael Stoschka
Chair of Mechanical Engineering, Montanuniversitt Leoben, Austria

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

Article history: Post-treatment methods enhance the fatigue strength of welded high-strength steel joints signicantly. In
Available online xxx industrial applications, one commonly applied technique is the high frequency mechanical impact (HFMI)
treatment. Attained local benets increasing fatigue strength are the compressive residual stress state,
Keywords:
Structural weld simulation the reduced notch effect at the weld toe, and additionally the local work hardening of the material. This
Local fatigue assessment paper presents the set-up of a closed simulation loop including structural weld simulation, numerical
High frequency mechanical impact (HFMI) computation of the HFMI-process, and a numerical evaluation of the local fatigue life.
Noch stress/strain approach A thermo-mechanical coupled weld process simulation using Sysweld is built-up to assess the course
Fracture mechanics of residual stress and strain due to welding. The resulting local material behavior is transferred as me-
Fatigue of welded joints chanical cards to the solver Abaqus for the subsequent numerical simulation of the HFMI-process. Hence,
major inuences such as transient change of material parameters during welding and subsequent cool-
down, effect of process dependent clamping conditions and changing contact characteristics are consid-
ered. Finally, a numerical evaluation of the local fatigue behavior by the local stress/strain approach and
by crack propagation is performed by the aid of the software packages Femfat, nCode and Franc2D. A
comparison of the simulated fatigue life with experimental test results proofs their basic applicability;
but also numerical limitations of the presented simulation tools are determined. The major benet of
the established simulation chain is the opportunity to study different weld process and HFMI-treatment
parameters in regard to fatigue strength without the need of comparably expensive fatigue tests.
2016 Civil-Comp Ltd. and Elsevier Ltd. All rights reserved.

1. Introduction further major enhancement is provided by post-treatment meth-

In general, welding leads to signicant amounts of distortion
and residual stresses in a component or structure. Acting external
loads and internal weld residual stresses are superimposed which
may affect the local stress condition. For high-cycle fatigue appli-
cations, the notch topography, microstructure in the heat-affected
zone (HAZ), and the residual stress state have a signicant inu-
ence on the fatigue lifetime.
This paper is based upon [1] and additionally provides en-
hanced information about the specimen manufacturing, simulation
work including process denitions, material data and representa-
tion of the results, and a fatigue assessment by a linear-elastic frac-
ture mechanical approach. According to the IIW-recommendation
[2], the fatigue strength of welded steel joints is independent of
the steel grade. One opportunity to utilize the potential of high-
strength steels offers an optimization of weld process parameters,
such as ller metal and shielding gas, as investigated in [3,4]. A
ods. One widespread technique in industrial applications repre-
sents the high frequency mechanical impact (HFMI) treatment. This
post-treatment method is statistically proven to increase fatigue
life of welded joints signicantly, especially for high-strength steels
[5]. Thereby, a hardened metal pin is hammering onto the weld
toe surface with high frequency and causes local plastic deforma-
tion in this area. The weld toe geometry is round-out; this re-
duces thereby the notch stress concentration, compressive residual
stresses are induced, the top surface layer is hardened and weld-
ing induced small cracks on the surface are closed. The change in
local geometry and material properties due to welding and HFMI-
treatment changes the geometric and metallurgical notch at the
most-stressed weld toe of full-penetrated joints. In this work, nu-
merical investigations based on experimental data are made to
build-up a closed three-dimensional simulation loop and to deter-
mine the change in local lifetime of a butt joint specimen.
2. Objectives

Corresponding author. Tel.: +43 38424021463. The aim of the work is to built-up a simulation loop including
E-mail address: martin.leitner@unileoben.ac.at (M. Leitner). structural thermo-mechanical coupled weld simulation, transfer of

http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
0965-9978/ 2016 Civil-Comp Ltd. and Elsevier Ltd. All rights reserved.

Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
JID: ADES
ARTICLE IN PRESS [m5G;February 22, 2016;14:45]

2 M. Leitner et al. / Advances in Engineering Software 000 (2016) 111

Table 1
Nominal mechanical properties of base and ller material.

Type Yield Ultimate Elongation

strength [MPa] strength [MPa] [%]

Base material 355 575 22

Filler material 440 530 30

Table 2
Nominal mechanical properties of base and ller material.
Fig. 1. Steps of numerical simulation loop.
Current Voltage Traverse Wire feed Heat input
Current [V] speed [cm/min] rate [m/min] [kJ/cm]

the transient results into a structural nite element code, numeri-
cal investigation of the HFMI-treatment and nally a computation
of the local fatigue strength or lifetime. Fig. 1 depicts a owchart
of the characteristic process steps and the involved software pack-
ages.
3. Experimental work
The benet of fatigue strength due to high frequency mechani-
cal impact (HFMI) treatment of welded joints depends on a reduc-
tion of residual tensile stresses, shallower geometric notch, and a
local hardening of the hammered surface layer. In the following,
detailed insight in the manufacturing process of the investigated
S355 butt joint specimens and fatigue testing is provided.
3.1. Manufacturing of specimens
In Fig. 2, the design and manufacturing set-up of the inves-
tigated butt joint is illustrated. The weld process is dened as
GMAW (MAG metal active gas welding) and the specimens are
continuously welded in arrays in order to assure similar weld seam
qualities, stiffness and residual stress states. The lateral run-on and
off parts of the base plate are to stabilize the arc during ignition
and further on to avoid end crater cracks in the testing section. In a
subsequent manufacturing step after welding and HFMI-treatment,
these lateral ends are truncated from the fatigue testing specimen.
Reproducible clamping conditions for every specimen are ensured
by a systematically positioning of the tacking points at the array.
Furthermore, additional tackings at the joint clearance between the
specimens minimize the magnetic arc blow and lead to a more
consistent welding process.
The nominal mechanical properties of the applied base and
ller material are summarized in Table 1. For the GMAW welding
process, a solid wire G3Si1 in combination with a standard weld
process gas M21 (82% Ar, 18% CO2 ) is utilized.
290.0 27.5 60.0 10.0 8.0
Table 2 provides an overview of the welding process parame-
ters which are applied for the GMAW-process in spray arc mode
without pulsed current pattern.
Subsequently to the welding process and nal cool-down, half
of each specimen lot is mechanically post-treated by HFMI using
the Pneumatic Impact Technology (PIT). For the pneumatic actua-
tion energy, a common industrial air pressure of 0.6 MPa (87 psi)
is applied. The angle of the pin-axis with respect to the plate sur-
face is set between 80 and 100, the angle of the pin-axis with
respect to the direction of travel about 90, and the working speed
between 1 and 3 mm/s. The radius of the hardened pin used in this
study is R = 2 mm, which ts to the weld seam size of the inves-
tigated specimens. To ensure a high-quality post-treated weld toe,
the HFMI-treatment is performed in accordance to the proposed
procedures and quality assurance guidelines depicted in [6]. The
post-treatment is applied without additional static pre-stressing
or former cyclic damage. Due to the HFMI-treatment, the pulsed
transaction of the hardened pin rounds-out the curvature of the
weld toe region and thereby reduces the geometric stress concen-
tration factor at the weld toe. In addition, compressive stresses are
induced which counteract the superposition of residual weld and
external load applied tensile stresses.
3.2. Fatigue tests
Fatigue tests at a tumescent stress ratio of R = 0.1 involving
butt joint specimens are carried out to investigate the inuence
of the post-weld treatment technique on the lifetime experimen-
tally. The results are part of an extensive test program for HFMI-
treated joints focusing on the effect of different high-strength steel
grades and joint geometries, see [7]. As the weld toe notch factor
of the investigated thin-walled butt joint is comparably low, the

Fig. 2. Specimen geometry and set-up of welding array [4].

Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
JID: ADES
ARTICLE IN PRESS [m5G;February 22, 2016;14:45]

M. Leitner et al. / Advances in Engineering Software 000 (2016) 111 3

Table 3
Summary of fatigue test results.

Condition  n (2e6) k Nk 1/TS

. 6
Base material 250 7.6 1.5 10 1.07
As-welded 185 5.3 2.0. 106 1.17
HFMI-treated 220 6.3 9.0. 105 1.15

Table 4
Chemical composition in weight % of S355 mild steel.

C Mn Si S P

0.18 1.60 0.55 0.035 0.035 Fig. 3. Half-symmetrical model for structural welding simulation.

benet in improvement due to the HFMI-treatment is quite small

but both are at the upmost fatigue performance. Due to the post-
treatment, ninety percent of the base material fatigue strength is
reached, which conrms the benecial effects of this technique.
Table 3 summarizes the fatigue test results and provides a com-
parison of the evaluated nominal stress range  n , slope k in the
nite life regime, transition knee point Nk , and scatter index TS for
the investigated conditions, base material, as-welded, and HFMI-
treated.

4. Manufacturing process simulation Fig. 4. Double-ellipsoidal heat source model [9].

4.1. Structural weld simulation

The thermo-mechanical coupled structural weld simulation is
performed by the aid of the commercial software package Sysweld.
Firstly, the transient temperature distribution including the time-
temperature dependent phase transformations are calculated ac-
cording to Eq. (1).
 2  ter nal-cooling. The heat-input q(x,y,z,t) is dened as a double-
T K T 2T 2T 1 Q
= + + +
t c x2 y2 z 2 c t
In Eq. (1), T represents the temperature, t the time, K the ther-
mal conductivity, c the specic heat capacity, the density and
Q the internal heat generation rate. After the thermo-metallurgical
run, the mechanical computation of the local strain state is oper-
ated according to Eq. (2).
tot = el + th + p + v p + pt
In Eq. (2), tot represents the total strain, el
th the thermal proportion, pl and vp the plastic- and visco-
plastic part and pt the phase-transformation induced strain. The
unclamping occurs after nal cool-down of the welded plate and
not-yet-deposited material is treated as quiet elements with signif-
icantly reduced mechanical properties. The chemical composition
of the mild steel for the numerical analysis is shown in Table 4
and basically equal to the base material used in testing.
A half symmetrical 3D-model of the butt joint consisting of
about 135,0 0 0 hexahedral elements and 150,0 0 0 nodes is built-
up according to the specimen geometry of the fatigue test sam-
ples, see Fig. 3. Thereby, also the two run-out areas are added
along the weld line, whereat the inhomogeneous start and stop
regions of the weld seam are not located within the test speci-
men cross section. The subsequent HFMI-treatment is applied onto
the whole weld toe length and the overlapping lateral surfaces are
cut-off after the treatment, leading to an identical replication of
the real manufacturing procedure. The cross shape of the joint is
based on micrographs of metallographic sections and the applied
weld toe radius possesses a value of 0.3 mm, which was gained
by non-destructive laser-confocal measurement of the weld toe re-
gion. The input parameters of the structural weld simulation, en-
ergy input per unit length and welding speed, the boundary and
clamping conditions are dened as applied during welding of the
specimens.
The weld toe area requires a ner mesh for the plastic deforma-
tions of the subsequent HFMI post-treatment simulation. Accord-
ingly to the manufacturing process, the specimens are clamped
with tacking points at the end of the plates and unclamped af-
(1) ellipsoid model [8] and adjusted to the real weld process with an
constant ellipse shape of a = 7.50 mm, b = 4.00 mm, cf = 3.33 mm,
and cr = 6.67 mm. A normalized power maximum power source
intensity at the front area of qf = 1.00 and at the rear side of
qr = 0.83 is used to split up the total heat input per unit length
of 8.0 kJ/cm, see Fig. 4. Finally, a general welding heat eciency of
= 0.85 for the GMAW is considered in the process denition.
To ensure comparable results to the experimental investiga-
(2)
tions, the knowledge of thermal and mechanical properties of the
the elastic part, material is essential. Thermal properties like specic heat, conduc-
tivity, emissivity, and latent heat depend on the current temper-
ature, whereas mechanical properties like Youngs modulus, Pois-
sons ratio, and the material plasticity are in addition inuenced by
the work-hardening behavior controlled by the effective strain rate.
Numerical investigations in [10,11] exhibit that only a few proper-
ties are technically important; therefore, numerous extensive and
costly experiments may be avoided. Moreover, the default database
in Sysweld provides numerous temperature-dependent thermal and
mechanical properties for the applied carbon mild steel. Fig. 5 de-
picts two major thermal properties, as specic heat c and thermal
conductivity K, illustrating a distinctive phase-dependent differ-
ence up to approximately Ac1 -temperature. Additionally, the den-
sity is pre-dened exhibiting no signicant phase-dependency and
an almost linear decrease from 7.9 kg/m for 20 C to 7.29 kg/m for
1500 C.
Regarding the material properties not only temperature-
dependent quasi-static values like yield stress, ultimate stress,
Youngs modulus, Poissons ratio, and density, but also cooling
rate-dependent phase transformation have to be considered. To
evaluate the local microstructure and hardness, time-temperature-
transformation (TTT) diagrams are utilized as a basis, see Fig. 6.
It is shown that especially for higher cooling rates a primary

Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
JID: ADES
ARTICLE IN PRESS
4 M. Leitner et al. / Advances in Engineering Software 000 (2016) 111
Fig. 5. Thermal phase-dependent properties.
Fig. 6. TTT-diagram for mild steel S355 [12].
bainitic-martensitic microstructure is achieved, leading to an in-
creased hardness in the HAZ and primary brittle fracture mecha-
nisms. Therefore, it is most important to follow the recommended
cooling time for the base material, which is veried by thermo-
couple measurements during welding of the specimens leading to
a predominantly ferrite-bainite microstructure in the HAZ.
Fig. 7. Volume change due to phase transformation [12].
Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
[m5G;February 22, 2016;14:45]
Especially for carbon steels a consideration of the volume
change due to austenitic-martensitic phase transformation is es-
sential to compute the local residual stress state properly. The de-
fault data-base incorporates this information based on dilatometer
tests for the applied base material, see Fig. 7.
After the material assignment, the weld process parameters
have to be dened. Preliminary investigations [13] showed that
by application of isotropic hardening model, the simulated resid-
ual stress conditions are in good accordance to measured val-
ues. In contrast, the kinematic hardening model underrates phase-
dependent hardening effects. Hence, the isotropic hardening model
is applied within the structural weld simulation. Fig. 8 shows the
nodal temperature distribution during welding and compares the
molten area with a micrograph. The three main areas - base metal
(BM), heat-affected-zone (HAZ) and molten ller metal (FM) are
labeled accordingly, exhibiting that the shape of the numerical
weld seam is in good accordance to the real weld process.
After completing the thermo-metallurgical computation, the
mechanical step including a calculation of the residual stress and
distortion condition is conducted. In regard to the fatigue life of
the investigated specimen, the residual stress state in transverse
direction to the weld seam is of great importance, which is dis-
played in Fig. 9. The maximum tensile values are located in the
center of the weld toe and exhibit stresses of about 290 MPa. Dur-
ing cyclic loading, these tensile residual stresses are superimposed
with the tumescent external load and fatigue crack initiation is
mainly caused at this region.
Considering the preceding thermo-mechanical rolling process,
the evaluated residual stress values at this region agree well to X-
ray measurements, up to a maximum deviation of fteen percent.
4.2. Simulation of HFMI-treatment
After structural welding simulation in Sysweld, the results of
the butt joint including residual stress state, plastic strains and
phase-dependent material properties are transferred as mechan-
ical cards into Abaqus. Although the amount of data gained by
three-dimensional weld simulation is excessive, the simulation of
the HFMI-treatment is applied onto the whole model; thus avoid-
ing errors due to two-dimensional plane strain simplications as
discussed in [14]. On contrary to the structural weld simulation,
whereat an isotropic hardening model is applied, it is herein ut-
most important to incorporate the cyclic hardening behavior prop-
erly. To study the effect of the hardening law on the stress-strain
JID: ADES
ARTICLE IN PRESS [m5G;February 22, 2016;14:45]

M. Leitner et al. / Advances in Engineering Software 000 (2016) 111 5

Fig. 8. Temperature distribution during welding and comparison of molten zone.

Fig. 9. Residual stress condition after welding process.

Fig. 10. Isotropic and combined isotropic/kinematic model [after [15]].

relationship, a single element was stressed by a reversal load cy- force value denes the displacement-controlled impact step in the
cle. Fig. 10 compares the results of isotropic hardening law with a subsequent HFMI simulation.
mixed isotropic-kinematic hardening model. In addition, high-speed camera recordings are operated to in-
The combined hardening model includes the backstress after a vestigate the dynamic movement of the pin during service. An
change of the loading direction and matches suciently for the
mild construction steel S355. In [16], the combined Chaboche-
model [17], implemented in Abaqus, is applied for the simulation
of the HFMI-treatment of austenitic and duplex stainless steels and
shows a good cyclic applicability. Hence, for the subsequent struc-
tural elastic-plastic simulation of the HFMI-treatment in this work,
the mixed hardening model and a single non-step-wise indenta-
tion of the pin is applied.
As the impact power of the hardened pin is essential for the
plastic deformations at the HFMI-treated area, strain gauge mea-
surements at the surface of the pin are carried out. The evaluation
of the dynamic impact force matches well to the transient move-
ment of the pin. Each compressive peak intends an impact of the
pin onto the specimen surface. The results of such strain gauge
measurements, either with a clamped device or operated manu-
ally, are shown in Fig. 11. Due to a high logging frequency of at
least 1.2 kHz the necessary time steps for evaluation of the tran-
sient impact force are maintained. It is shown that both techniques
are almost equal, exhibiting an impact peak force up to 2.5 kN. This
Fig. 11. Results of impact force measurements.

Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
JID: ADES
ARTICLE IN PRESS [m5G;February 22, 2016;14:45]

6 M. Leitner et al. / Advances in Engineering Software 000 (2016) 111

Fig. 12. Numerical simulation of HFMI-process.

Fig. 13. Residual stress condition after HFMI-treatment.

evaluation of the pin stroke and velocity distribution by the point
tracking method acts as basis to estimate the local strain rate dur-
ing penetration. Strain rates up to 500 s1 are determined, which
are basically in good agreement to recent studies [18] with the
same post-treatment device. In general, it is preferred to perform
numerical analyses of the HFMI treatment on the basis of an ex-
plicit simulation to consider the dynamic process properly. In this
case, one great drawback is the comparably high numerical insta-
bility and computational effort. In [18], it is shown that an explicit
simulation with a combined isotropic-kinematic and strain rate de-
pendent hardening behavior leads to the best agreement with the
measured residual stress values.
However, in this work an implicit displacement-controlled com-
putation is executed, showing a more stable computing behavior
featuring the calculation of complex structures as well. As shown
later on, the numerical results after HFMI-treatment are in a good
accordance to X-ray measurement values, proving that this method
is an ecient and well applicable technique.
The pin is modeled as rigid body with a tip radius of 2 mm
and the specimen is clamped to the padding support. The applied
axial travelling distance of the pin equals 0.1 mm, relating to the
measured indentation of the micrographs, and leading to force val-
ues almost equal to the measurements. After each indentation the
specimen is moved 0.4 mm along the weld path to achieve a su-
cient overlapping.
Fig. 12 illustrates the HFMI-treatment process, clearly depicting
the already post-treated area at the weld toe by change in hydro-
static pressure. As expected, tensile residual stresses are reduced
and the amount of compressive residual stresses in the weld toe
region increases due to the cyclic mechanical impacts.
After completing the HFMI-treatment process, the run-out
plates are numerically cut off by deactivating the lateral elements
and performing an additional equilibrium step. This causes minor
stress redistribution at the sideways edges, but no signicant effect
Table 5
Comparison of residual stress values at weld toe region.
Residual stress S11 [MPa] Simulation Measurement
z = 0 m (at surface) -190 210
z = 200 m (in depth) -310 290
on the weld toe zone. The nal residual stresses state is shown in
Fig. 13.
A comparison of the residual stress condition at the surface
layer of the weld toe in mid-position is provided in Table 5. The
resulting values agree well with differences up to ten percent be-
tween the measurement and simulation, showing that a numerical
computation by an implicit formulation is well applicable.
5. Local fatigue assessment
The local fatigue behavior of welded and HFMI-treated joints
can be preferably evaluated by strain-based approaches due to
the quite excessive elastic-plastic material condition in the treated
weld toe region [19,20]. In this work, a stress and strain-based, and
a fracture mechanical approach is applied to assess the fatigue life.
5.1. Safety factor based on local stress approach
Firstly, an assessment of the local safety factor based on a stress
approach is performed in Femfat. Thereby, the local stress concept
[21] takes into account the stress gradient to dene the local stress
fatigue strength limit, the number of cycles at the fatigue limit and
the slope in the nite lifetime region.
To consider the inuence of mean stresses various methods
have been developed, e.g. by Gerber [22], Goodman [23], Soderberg
[24] and Morrow [25]. In general, tensile mean stresses are detri-
mental in regard to fatigue life, while compressive mean stresses

Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
JID: ADES
ARTICLE IN PRESS
M. Leitner et al. / Advances in Engineering Software 000 (2016) 111
Fig. 14. Local safety factor at R = 1 for as-welded (left) and HFMI-treated (right) condition calculated by stress approach in Femfat.
Table 6
Local safety factor at different R-ratios for as-welded and HFMI-
treated condition calculated by Femfat.
Local safety factor at N = 1.5e5 As-welded HFMI-treated
R = 0.1 0.82 0.93
R = 1 1.31 1.53
are benecial. Most of these existing mean stress models are appli-
cable within conventional stress based methods for correcting the
fatigue limit or fatigue strength in the high cycle fatigue regime of
106 up to 108 load-cycles [26]. For the present investigations the
mean stress inuence according to the German FKM-guideline [27]
is applied. Thereby, the simulated manufacturing induced resid-
ual stresses for the as-welded and HFMI-treated condition are de-
ned as mean stresses which are superimposed with external load
stresses. For the analysis of the local safety factor a stress ampli-
tude of 150 MPa at 1.5e5 load-cycles and an evaluation by con-
stant stress ratio with a survival probability of PS = 97.5% is chosen.
Based on the experimental fatigue test results depicted in Fig. 2
this would lead to a safety factor of about one in the most-
damaged weld toe region of the butt-joint. In Fig. 14, the results
for the as-welded and HFMI-treated condition at a stress ratio of
R = 1 calculated by Femfat are pictured.
For each node a local S/N-curve is calculated based on nodal
characteristics taking fatigue support, size effect, mean stress, sta-
tistical inuence and surface roughness as parameters into account.
In Table 6 the results by Femfat for the two investigated manu-
facturing conditions and stress ratios are shown. The experimental
fatigue tests are performed at a tumescent stress ratio of R = 0.1
whereat the as-welded and HFMI-treated condition show almost
equal load-cycles until burst fracture. For comparison, the exper-
imental values exhibit a local safety factor of one. In case of the
numerical results the calculation for the as-welded condition is
slightly conservative but the HFMI-treated specimen shows a good
accordance to the experiments. For the stress ratio of R = 1 the
safety factor increases by about 60% for both conditions caused by
the reduced mean stress state.
For comparison purposes a numerical calculation of the local
safety factor is also performed with the aid of nCode. To ensure
comparable results, the same material behavior and assessment at-
titudes are dened. In the as-welded condition, the local safety
factor at R = 1 is determined as 1.21 and at R = 0.1 as 1.09. These
values show a good compliance to the results calculated by Femfat.
However, for the HFMI-treated condition the results show a signif-
icant overestimation of the fatigue behavior. The local safety factor
at R = 1 for the as-welded condition calculated by the stress ap-
proach in nCode is pictured in Fig. 15.
Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
[m5G;February 22, 2016;14:45]
7
Summarized, the stress approach to calculate the local safety
factor shows a good accordance to the experimental fatigue tests
for the examined load level. The inuence of the applied stress ra-
tio is well considered and the calculated safety factors agree well
to the experiments. However, due to a high notch effect and over-
loads plasticization at the weld toe may occur which is basically
not well described by stress-based methods. Therefore, a further
fatigue assessment based on the local strain approach is executed
to consider the elastic-plastic behavior at the notch more accurate.
5.2. Damage based on local strain approach
Due to the superposition of the residual and load based stresses
the material yield limit is exceeded locally and plastic defor-
mations occur. This plastic rearrangement deduces a subsequent
change in the stress condition which is more clearly feasible by
the notch strain approach. For the evaluation of the local notch
stress and strain relation, the stabilized cyclic stress-strain curve
by RambergOsgood [28] is used:
na   n1
a,t = a,e + a,p = +
na
(3)
E K
where a,t is the total strain amplitude, a,e and a,p the elastic
and plastic terms, na the nominal stress amplitude, E the Youngs
modulus, K the strength coecient and n the strain hardening ex-
ponent. The calculation of the local notch stress condition is per-
formed based on Neubers rule [29]:
(Kt na )2
 e  e = (4)
E
where  e is the notch stress range,  e the notch strain range
and Kt the stress concentration factor. The hysteresis loops are de-
rived from stress-strain curves by applying Masings rule [30,31].
Further on, the relationship between local strain and number of
load-cycles until crack initiation is dened by Manson [32], Con
[33] and Basquin [34]:
f  b  c
a,t = a,e + a,p = 2N f + f 2N f (5)
E
where f is the fatigue strength coecient, Nf the fatigue life, b
the fatigue strength exponent, f the fatigue ductility coecient
and c the fatigue ductility exponent [35]. For the current inves-
tigations, the damage parameter according to Smith, Watson and
Topper [36] is applied do consider the mean stress effect within
the strain approach:

PSW T = max a E (6)
where max is the maximum local stress amplitude and a the lo-
cal strain amplitude. For a dened lifetime and for different com-
binations of the strain amplitude and maximum stress the model
assumes that the damage parameter keeps constant [26,31].
JID: ADES
ARTICLE IN PRESS [m5G;February 22, 2016;14:45]

8 M. Leitner et al. / Advances in Engineering Software 000 (2016) 111

Fig. 15. Local safety factor at R = 1 for as-welded condition calculated by stress approach in nCode.

Fig. 16. Local damage at R = 0.5 for as-welded and HFMI-treated condition calculated by strain approach in nCode.

Table 7 Table 8
Local damage for as-welded and HFMI-treated condition calcu- Local damage at different R-ratios by strain approach for as-
lated by stress and strain approach nCode. welded and HFMI-treated condition calculated by nCode.

Local damage at N = 1e5 As-welded HFMI-treated Local damage at N = 1e5 As-welded HFMI-treated

Stress approach 0.41 < 0.01 R = 1 0.96 < 0.01

Strain approach 0.96 < 0.01 R = 0.1 1.37 < 0.01
R = 0.5 1.86 0.24

An implementation of the SWT damage parameter in the strain-

life equation shows that the product of the maximum tensile stress
max and the strain amplitude a considers the inuence of the
mean stress:
 2
f  2 b  b+c
max a = 2N f + f f 2N f
E
Based on this approach, a strain-based local damage calculation
is performed in nCode. For this study, the cyclic material data of
a mild-strength steel (UTS500) is taken from the default database.
The load amplitude is set to 100 MPa and the linear damage ac-
cumulation is done for a dened number of 1e5 load-cycles. For
comparison, the experimental values exhibit a local damage of one.
A tabular comparison of the local damage between the stress and
strain approach for the as-welded and HFMI-treated condition at a
survival probability of PS = 97.5% is shown in Table 7.
The results for the as-welded condition exhibit that the local
damage calculated by the two approaches are roughly in the same
range. Minor deviations might be caused by the different method
of implementations of the mean stress effects caused by the high
residual stress conditions after the welding process. In case of the
HFMI-treated specimen, both concepts calculate a minor damage
sum by reason of the high compressive residual stress state in the
weld toe area which counteracts to the external loadings. To study
the effect of higher mean stresses further calculations based in the
strain approach for different R-ratios are performed. The results for
(7) R = 0.5 are shown in Fig. 16.
A summary of the results is given in Table 8. Thereby signicant
differences between the investigated stress ratios are recognizable.
The local damage sum for R = 0.5 is almost two times higher com-
pared to R = 1 in the as-welded condition. In addition it is shown
that even for high external mean stresses the calculated damage
for the HFMI-treated condition is quite low in contrast to the as-
welded state.
5.3. Fatigue life based on fracture mechanical approach
To evaluate not only the crack initiation stage by notch
stress or strain-based approaches but also the subsequent crack
propagation, fracture mechanical concepts provide suitable and
proven tools. Especially for welded joints containing cracks, aws
or crack-like imperfections, the estimation of crack propagation is

Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
JID: ADES
ARTICLE IN PRESS [m5G;February 22, 2016;14:45]

M. Leitner et al. / Advances in Engineering Software 000 (2016) 111 9

Fig. 17. Results of numerical crack propagation analysis in Franc2D at a nominal stress range of  n = 300 MPa.

Fig. 18. Results of analytical crack propagation analysis in nCode at a nominal stress range of  n = 300 MPa.

fundamental to assess the service lifetime and inspection periods
properly.
In general, the fatigue life evaluation by crack-growth models is
based on the stress intensity factor K as driving force for the crack
propagation. Nowadays, several methods are available to calculate
the factor K, such as closed form solutions for simple structural
details [3739], weight function approaches [4042], and numeri-
cal methods for complex welded structures [4345]. In case of the
investigated butt joint parametric formulas are provided in [4648]
at which the non-linear stress peak at the weld toe is considered
by an additional function Mk . Therefore, the range of the stress in-
tensity factor K is dened as shown in Eq. (8) incorporating the
applied stress range  , the actual crack length a, the geometry
function Y(a), which considers the crack geometry, and the factor
Mk (a):
K =  a Y ( a ) M k ( a )
Further on, the crack growth rate da/dN and subsequent the to-
tal lifetime N can be calculated following the equation by [49] for
actual stress intensity values K above the threshold value Kth :
da
= C0 K m for K > Kth
dN
For welded joints the material parameters C0 and m are derived
from experimental data and conservative values are given in [50].
For steel structures the coecient C0 = 5.0E-13 [N. mm3/2 ] and the
exponent m = 3 are applicable which are used for the analytical
and numerical calculations in this paper.
For the application of fracture mechanical approaches an as-
sumption of an initial crack depth ai is required. In [50] it is
stated that a specication of ai = 0.15 mm is recommended for a
conservative assessment, but in case of mechanical engineering ap-
plications ai can be dened as 0.05 mm, which is also selected for
the investigations in this paper.
Welding as thermo-metallurgical process causes complex resid-
ual stress states, global and local distortions, and changes the
local microstructure and hardness within the heat-affected-zone
(HAZ). Among all these effects, the residual stress condition has
been identied as the most inuential factor on fatigue [51]. An
evaluation of the local residual stress distribution can be either
performed by experimental measurements [52], by numerical anal-
ysis using structural weld simulation [53] or the inherent strain
method [54]. To incorporate residual stresses in fracture mechan-
ical calculations a superposition of the external applied stress in-
tensity factor range Kapp and the stress intensity factory due to
residual stresses Kres is proposed by [55,56]. Hence, the range of
effective stress intensity factor Keff is dened as follows, which is
not affected by a superposition of Kres :
 max   min 
Ke f f = Kapp + Kres Kapp max
+ Kres = Kapp min
Kapp = Kapp
(8)
(10)
In contrary, the local effective R-ratio Reff at the crack tip is af-
fected by the superimposed stress intensity factory due to residual
stresses Kres and continuously changes during crack propagation:
(9)  min 
Kemin Kapp + Kres
=  
ff
Re f f = (11)
Kemax
ff Kapp + Kres
max
To calculate the crack growth rate da/dN incorporating the pres-
ence of residual stresses several models exist [5759]. Within this
paper, rstly, a numerical crack propagation analysis by applica-
tion of the software package Franc2D [60] is conducted for the
butt joint in as-welded condition. Thereby, the inuence of residual
stresses is considered by a crack closure approach, but no redistri-
bution of the residual stress eld during the crack propagation is
taken into account [61]. The results of the crack growth simulation
in Franc2D are shown in Fig. 17.

Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
JID: ADES
ARTICLE IN PRESS
10 M. Leitner et al. / Advances in Engineering Software 000 (2016) 111
Table 9
Comparison of fatigue life for as-welded condition calculated by frac-
ture mechanical approaches.
Fatigue life at  n = 300 MPa Number of load-cycles
IIW-recommendation (nominal S/N-curve) 24,000
Franc2D (a/c = 0) 28,000
nCode (a/c = 0.1) 46,000
Fig. 19. Notch dependency of endurance limit  E assessment [62,63].
In addition, an analytical crack propagation assessment in nCode
by a preceding calculation of the stress intensity factor based on
given parametric formulas in [37] is executed, see Fig. 18.
A comparison of the results to the IIW-recommended lifetime
by the nominal S/N-approach is summarized in Table 9. As ex-
pected, the fatigue life based on the nominal stress approach is
most conservative. The difference between Franc2D and nCode can
be explained by the assumption of slightly different a/c-ratios,
but in general a good compliance for the as-welded condition is
achieved.
In case of the HFMI-treated condition the application of crack
propagation approaches seems to be not utile. As shown in Fig. 19,
the validity of applying stress or strain-based and fracture mechan-
ical approaches in dependency of the stress concentration factor Kt
in the vicinity of the notch is presented.
Due to a signicant reduction of the stress concentration at the
weld toe by the HFMI-treatment, only a mild notch is expectable
and the application of stress or strain-based approaches to assess
the fatigue limit is preferable. Furthermore, a local hardening and
the presence of compressive residual stresses lead to an extension
of the crack initiation phase leading to a reduced amount of crack
propagation in relation to the total lifetime. Strain gauge measure-
ments with accompanying optical crack detection of the weld toe
revealed that for HFMI-treated joints a decrease of the measured
stress only appears closer to the end of the fatigue test, and there-
fore, the crack initiation phase seems to be dominant in this case.
6. Conclusion
A three-dimensional simulation loop is built-up incorporating
a structural elastic-plastic weld simulation, a numerical computa-
tion of the HFMI-process; and nally, local fatigue assessment by
stress/strain-based and fracture mechanical approaches. The weld-
ing simulation takes phase-dependent material properties into ac-
count and both the transient stress-strain results and the locally
change in mechanical properties of the weld process are mapped
to a structural elastic-plastic mechanical simulation. For the im-
plicit, displacement controlled HFMI-treatment simulation, the dif-
ferent elemental mechanical properties of the weld simulation
are grouped into material cards and the elastic-plastic material
behavior is considered by a combined hardening model. A com-
parison of the obtained residual stress condition to radiographic
Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
[m5G;February 22, 2016;14:45]
measurements shows a good agreement with differences up to f-
teen percent for the as-welded and up to ten percent for the HFMI-
treated state.
Finally, a local fatigue assessment incorporates the utilized sim-
ulation tools within the manufacturing steps for the investigated
thin-walled butt joint. The assessed manufacturing and load pa-
rameters include also the effect of different load-ratios. The simu-
lated lifetimes are in good accordance to the fatigue tests, which
proofs the basic applicability of the presented procedure. Linear-
elastic fracture mechanical calculation of the crack propagation is
well applicable for the as-welded condition, but on contrary, for
the HFMI-treated state further investigation is needed to incorpo-
rate the enhanced crack initiation phase properly.
Acknowledgments
Financial support by the Austrian Federal Government (in
particular from Bundesministerium fr Verkehr, Innovation und
Technologie and Bundesministerium fr Wirtschaft, Familie und
Jugend) represented by sterreichische Forschungsfrderungsge-
sellschaft mbH and the Styrian and the Tyrolean Provincial Govern-
ment, represented by Steirische Wirtschaftsfrderungsgesellschaft
mbH and Standortagentur Tirol, within the framework of the
COMET Funding Programme is gratefully acknowledged.
References
[1] Leitner M, Simunek D, Stoschka M. Local fatigue assessment of welded and
high frequency mechanical impact-treated joints based on manufacturing pro-
cess simulation. In: Ivnyi P, Topping BHV, editors. Proceedings of the ninth in-
ternational conference on engineering computational technology. Stirlingshire,
United Kingdom: Civil-Comp Press; 2014 paper 41.
[2] Hobbacher A. IIW recommendations for fatigue design of welded joints and
components. WRC bulletin, 520. New York: The Welding Research Council;
2009.
[3] Stoschka M, Leitner M, Fssl T, Posch G. Effect of high-strength ller metals on
fatigue. Weld World 2012;56:209.
[4] Stoschka M, Leitner M, Posch G, Eichlseder W. Effect of high-strength ller
metals on the fatigue behaviour of butt joints. Weld World 2013;57:8596.
[5] Yildirim H, Marquis G. Fatigue strength improvement factors for high strength
steel welded joints treated by high frequency mechanical impact. Int J Fatigue
2012;44:16876.
[6] Marquis G, Barsoum Z. Fatigue strength improvement of steel structures by
high-frequency mechanical impact: proposed procedures and quality assurance
guidelines. Weld World 2014;58:1928.
[7] Leitner M, Stoschka M, Eichlseder W. Fatigue enhancement of thin-walled
high-strength steel joints by high frequency mechanical impact treatment.
Weld World 2014;58(1):2939.
[8] Goldak J, Chakravarti A, Bibby M. A new nite element model for welding heat
sources. Metall Trans B 1984;15(2):299305.
[9] Wohlfahrt H. Simulation der vorgnge im schmelzbad beim laser-
strahlschweien zur voraussage der nahtausbildung, gefge, verzug und
schweieigenspannungen. Technische Universitt Braunschweig; 2001. (in
German).
[10] Asserin O, Loredo A, Petelet M, Iooss B. Global sensitivity analysis in welding
simulations what are the material data you really need? Finite Element Anal
Des 2011;47:100416.
[11] Zhu X, Chao Y. Effects of temperature-dependent material properties on weld-
ing simulation. Comput Struct 2002;80:96776.
[12] ESI Group. SYSWELD analysis reference manual. ESI Group; 2009.
[13] Sakkiettibutra J, Loose T, Wohlfahrt H, Vollertsen F. Zur wahl des verfes-
tigungsmodells bei der berechnung von schweieigenspannungen. Univer-
sittsverlag Weimar; 2009. (in German) p. 2133.
[14] Rahmann M, Enzinger N, Sommitsch C. Deformation and distortion induced by
pneumatic impact treatment (PIT) in a welded joint for HS steel. In: Proceed-
ings of the international workshop on thermal forming and welding distortion;
2011. p. 20313.
[15] Rsler J, Harders H, M Bker. Mechanisches verhalten der werkstoffe. Springer-
Verlag; 2012.
[16] Baptista R, Infante V, Branco C. Fully dynamic numerical simulation of
the hammer peening fatigue life improvement technique. Procedia Eng
2011;10:19438.
[17] Chaboche J. A review of some plasticity and viscoplastic constitutive theories.
Int J Plast 2008;24:164293.
[18] J. Foehrenbach, V. Hardenacke, M. Farajian, High frequency mechanical Impact
treatment (HFMI) of welded joints: numerical and experimental investigations
to describe the condition in the surface layer, IIW Document XIII-2587-15,
2015.
JID: ADES
ARTICLE IN PRESS
M. Leitner et al. / Advances in Engineering Software 000 (2016) 111
[19] Le Quilliec G, Lieurade H-P, Bousseau M, Drissi-Habti M, Inglebert G,
Macquet G, Jubin J. Fatigue behaviour of welded joints treated by high fre-
quency hammer peening: part II numerical study, Paris: International Insti-
tute of Welding; 2011. IIW Document XIII-2395-11.
[20] Schaumann P, Keindorf C. Experiments and simulation of welded joints with
post weld treatment. In: Proceedings of the international conference on
welded structures, Miskolc/Hungary; 2008.
[21] Eichlseder W. Fatigue analysis by local stress concept based on nite element
results. Comput Struct 2002;80:210913.
[22] Gerber W. Bestimmung der zulssigen spannungen in eisenkonstruktionen. Z
Bayer Arch Ing Ver 1874;6:10110.
[23] Goodman J. Mechanics applied to engineering, vol. 1. 9th ed. London: Long-
mans Green and Co.; 1930.
[24] Soderberg CR. ASME Trans 1930;52:1328 (APM-52-2).
[25] J. Morrow, Fatigue properties of metals, section 3.2, Fatigue design handbook,
Pub. No. AE-4. SAE International, Warrendale, PA, 1968.
[26] Ince A, Glinka G. A modication of Morrow and SmithWatsonTopper mean
stress correction models. Fatigue Fract Eng Mater Struct 2011;34:85467.
[27] Hnel B, Haibach E, Seeger T, Wirthen G, Zenner H. Rechnerischer festigkeit-
snachweis fr maschinenbauteile (Numerical proof of strength of mechanical
parts). Frankfurt, VDMA Verlag: FKM-Richtlinie; 2002.
[28] W. Ramberg, W.R. Osgood, Description of stressstrain curves by three param-
eters, Technical Note No. 902, National Advisory Committee for Aeronautics,
Washington DC, 1943.
[29] Neuber H. ber bercksichtigung der spannungskonzentration bei festigkeits-
berechnungen. Konstruktion 1968;20:24551 (in German).
[30] Masing G. Stretching and hardening for brass. In: Proceedings 2nd interna-
tional congress on applied mechanics; 1926. p. 3325.
[31] Karakas . Consideration of mean-stress effects on fatigue life of welded mag-
nesium joints by the application of the SmithWatsonTopper and reference
radius concepts. Int J Fatigue 2013;49:117.
[32] Manson SS. Fatigue: a complex subject some simple approximation. Exp
Mech 1965;5:193226.
[33] Con LF. A study of the effect of cyclic thermal stresses on a ductile metal.
Trans ASME 1954;76:93150.
[34] Basquin OH. The exponential law of endurance tests. Am Soc Test Mater Proc
1910;10:62530.
[35] Niesony A, Dsoki C, Kaufmann H, Krug P. New method for evaluation of the
MansonConBasquin and RambergOsgood equations with respect to com-
patibility. Int J Fatigue 2008;30:196777.
[36] Smith KN, Watson P, Topper TH. A stressstrain function for the fatigue of
materials. J Mater 1970;5:76778.
[37] A. Hobbacher, Update of the fracture mechanics chapters of the iiw fatigue
design recommendation, IIW Document XIII-2370r1-11/XV-1376r1-11, 2011.
[38] Brennan FP, Dover WD, Kar RF, Hellier AK. Parametric equations for T-butt
weld toe stress intensity factors. Int J Fatigue 1999;21:105162.
[39] Wang T, Yang JG, Liu XS, Dong, ZB, Fang HY. Stress intensity factor expression
for center-cracked butt joint considering the effect of joint shape. Mater Des
2012;35:729.
[40] Shen G, Plumtree A, Glinka G. Weight function for the surface point of
semi elliptical surfece crack in a nite thickness plate. Eng Fract Mech
1991;40(1):16776.
[41] Moftakhar AA, Glinka G. Calculation of stress intensity factors by ecient in-
tegration of weight functions. Eng Fract Mech 1992;43(5):74956.
[42] Lindroth P, Marquis G, Glinka G. Fatigue crack growth of arbitrary planar
cracks in welded components. Weld World 2013;57:42535.
Please cite this article as: M. Leitner et al., Numerical fatigue assessment of welded and HFMI-treated joints by notch stress/strain and
fracture mechanical approaches, Advances in Engineering Software (2016), http://dx.doi.org/10.1016/j.advengsoft.2016.01.022
[m5G;February 22, 2016;14:45]
11
[43] Meneghetti G, Guzzella C. The peak stress method to estimate the mode I
notch stress intensity factor in welded joints using three-dimensional nite
element models. Eng Fract Mech 2014;115:15471.
[44] Leung AYT, Zhou Z, Xu X. Determination of stress intensity factors by the nite
element discretized symplectic method. Int J Solids Struct 2014;51:111522.
[45] Lee CH, Chang KH. Finite element computation of fatigue growth rates
for mode I cracks subjected to welding residual stresses. Eng Fract Mech
2011;78:250520.
[46] Hobbacher A. Stress intensity factors of welded joints. Eng Fract Mech
1993;46(2):17382 vol. 49 (2), pp. 323, 1994.
[47] Bowness D, Lee MMK. Prediction of weld toe magnication factors for semi-
elliptical cracks in T-butt joints. Int J Fatigue 20 0 0;22:36987.
[48] Bowness D, Lee MMK. Weld toe magnication factors for semi-elliptical
cracks in Tbutt joints - comparison with existing solutions. Int J Fatigue
20 0 0;22:38996.
[49] Paris P, Erdogan F. A critical analysis of crack propagation laws. J Basic Eng
1963;85(4):52834.
[50] Hobbacher A. The new IIW recommendations for fatigue assessment of welded
joints and components a comprehensive code recently updated. Int J Fatigue
2009;31:508.
[51] Servetti G, Zhang X. Predicting fatigue crack growth rate in a welded butt
joint: the role of effective R ratio in accounting for residual stress effect. Eng
Fract Mech 2009;76:1589602.
[52] Rossini NS, Dassisti M, Benyounis KY, Olabi AG. Methods of measuring residual
stresses in components. Mater Des 2012;35:57288.
[53] Stoschka M, Ottersbck M, Leitner M. Integration of phase-dependent work-
hardening into transient weld simulation. In: Proceedings of the ninth interna-
tional conference on engineering computational technology, Naples/Italy; 2014.
[54] Deng D, Murakawa H. Prediction of welding distortion and residual stress in a
thin plate butt-welded joint. Comput Mater Sci 2008;43:35365.
[55] Glinka G. Effect of residual stresses on fatigue crack growth in steel weldments
under constant and variable amplitude load. American Society for Testing and
Materials; 1979. p. 198214.
[56] Parker AP. Stress intensity factors, crack proles, and fatigue crack growth
rates in residual stress elds. American Society for Testing and Materials; 1982.
p. 1331.
[57] Forman RG, Kearney VE, Engle RM. Numerical analysis of crack propagation in
cyclic-loaded structures. Trans ASME J Basic Eng 1967;89:45964.
[58] Walker K. The effect of the stress ratio during crack propagation and fatigue
for 2024-T3 and 7075-T6 aluminium. ASTM STP 462, effects of environment
and complex load history on fatigue life, 462. ASTM International; 1970. p. 1
14.
[59] Forman RG, Mettu SR. Behavior of surface and corner cracks subjected to ten-
sile and bending loads in Ti6Al4 V alloy. In: Proceedings of 22nd symposium
on fracture mechanics, 1, ASTM STP 1131. Philadelphia: American Society for
Testing and Materials; 1992. p. 51946.
[60] FRANC2D, Version 3.2, http://www.cfg.cornell.edu/.
[61] Barsoum Z, Barsoum I. Residual stress effects on fatigue life of welded struc-
tures using LEFM. Eng Fail Anal 2009;16:44967.
[62] Frost NE, Marsh KJ, Pook LP. Metal fatigue. Oxford: Oxford University Press;
1974.
[63] Smith RA, Miller KJ. Prediction of fatigue regimes in notched components. Int
J Mech Sci 1978;20:2016.