See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/260948331

A Structural Strain Method for Fatigue

Evaluation of Welded Components

Article in International Journal of Pressure Vessels and Piping · July 2014

DOI: 10.1016/j.ijpvp.2014.03.003

CITATIONS READS

8 460

4 authors, including:

Pengli Dong Xianjun Pei

University of Science and Technology Beijing University of Michigan
118 PUBLICATIONS 1,728 CITATIONS 11 PUBLICATIONS 66 CITATIONS

SEE PROFILE SEE PROFILE

Shizhu Xing
University of Michigan
6 PUBLICATIONS 21 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Residual Stress Profile Estimation Scheme View project

Process Window Prediction for Friction Stir Welding View project

All content following this page was uploaded by Xianjun Pei on 18 November 2017.

The user has requested enhancement of the downloaded file.

 International Journal of Pressure Vessels and Piping 119 (2014) 39e51

Contents lists available at ScienceDirect

International Journal of Pressure Vessels and Piping

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

A structural strain method for low-cycle fatigue evaluation of welded

components
P. Dong a, *, X. Pei b, S. Xing b, M.H. Kim c
a
Department of Naval Architecture and Marine Engineering, University of Michigan, Ann Arbor, MI 48105, USA
b
University of Michigan, USA
c
Pusan National University, Pusan, South Korea

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

Article history: In this paper, a new structural strain method is presented to extend the early structural stress based
Received 29 September 2013 master SeN curve method to low cycle fatigue regime in which plastic deformation can be significant
Received in revised form while an elastic core is still present. The method is formulated by taking advantage of elastically
26 February 2014
calculated mesh-insensitive structural stresses based on nodal forces available from finite element so-
Accepted 4 March 2014
lutions. The structural strain definition is consistent with classical plate and shell theory in which a linear
through-thickness deformation field is assumed a priori in both elastic or elasticeplastic regimes. With
considerations of both yield and equilibrium conditions, the resulting structural strains are analytically
solved if assuming elastic and perfectly plastic material behavior. The formulation can be readily
extended to strain-hardening materials for which structural strains can be numerically calculated with
ease. The method is shown effective in correlating low-cycle fatigue test data of various sources docu-
mented in the literature into a single narrow scatter band which is remarkable consistent with the
scatter band of the existing master SeN curve adopted ASME B&PV Code since 2007.
With this new method, some of the inconsistencies of the pseudo-elastic structural stress procedure in
2007 ASME Div 2 Code can now be eliminated, such as its use of Neuber’s rule in approximating
structural strain beyond yield. More importantly, both low cycle and high cycle fatigue behaviors can
now be treated in a unified manner. The earlier mesh-insensitive structural stress based master SeN
curve method can now be viewed as an application of the structural strain method in high cycle regime,
in which structural strains are linearly related to traction-based structural stresses according to Hooke’s
law. In low-cycle regime, the structural strain method characterizes fatigue damage directly in terms of
structural strains that satisfy linear through-thickness deformation gradient assumption, material
nonlinear behavior, and equilibrium conditions. The use of a pseudo-elastic structural stress definition is
not fundamental, but merely a means to put low-cycle and high-cycle fatigue test data in a conventional
stress-based SeN data representation which is typically preferred in engineering practice, than a strain-
based representation.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction structural stress calculation method as a post-processing procedure

Since 2007, ASME Section VIII Division 2 Code has adopted an
alternative fatigue evaluation method for welded joints, i.e., the
mesh-insensitive structural stress based master SeN curve method
[1]. A comprehensive discussion on its theoretical basis, analysis
procedures, and validations using fatigue test data on this method
can be found in a recent publication [2]. The master SeN curve
method consists two basic elements: (a) a novel nodal force-based
* Corresponding author.
E-mail address: dongp@umich.edu (P. Dong).
to finite element structural solutions; (b) an equivalent structural
stress parameter that captures a combined effect of stress con-
centration, plate thickness, and loading mode on fatigue behavior
of welded joints [2e5]. The effectiveness of the master SeN curve
method has also been validated for applications in offshore struc-
tures independently, e.g., by Healy [7] and in automotive structures
by Kyuba [8].
The design master SeN curve in the 2007 ASME Div 2 [1,2] was
developed by introducing an equivalent structural stress range
parameter that collapses a large number of fatigue test data (about
1000 tests of large scale and full scale specimens) into a narrow

http://dx.doi.org/10.1016/j.ijpvp.2014.03.003
0308-0161/Ó 2014 Elsevier Ltd. All rights reserved.
40 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51
scatter band. These tests span a wide range of joint geometries,
plate or pipe wall thicknesses, and loading modes. Through a
standard statistical analysis of the data in the form of equivalent
structural stress range versus cycle to failure, the design master Se
N curve is defined as the mean master SeN curve minus three
standard derivations before an environmental effect factor is
considered. For applications not governed by ASME pressure vessel
codes, a design master SeN curve based on mean minus two
standard derivations is typically used [6e8].
The test data on which the design master SeN curve was based
have fatigue lives as low as a few hundreds of cycles to as long as
nearly 108 cycles to failure. It should be noted that in low cycle
regime, i.e., typically lower than 104 or 105, pseudo-elastic struc-
tural stress ranges [2,9] based on reported pseudo-elastic loads in
displacement controlled low-cycle fatigue tests, such as those in
Refs. [10,11]. In supporting the 2007 ASME Div 2 Code development
effort [1], Dong et al. [9] proposed a preliminary low-cycle fatigue
(LCF) treatment procedure for adapting the master SeN curve
method which was mainly focused upon high-cycle fatigue to low-
cycle fatigue applications in pressure vessel applications mostly
subjected to load-controlled conditions. The procedure involves
converting elastically calculated structural stresses under a given
loading condition into a through-thickness linearly distributed
structural strain according to Hooke’s law, then searching for a
structural strain definition that both satisfy yield conditions and
through-thickness linear deformation conditions. Unsuccessful at
the time, they assumed that Neuber’s rule can be used to calculate
approximate structural strains beyond yield using the elastically
calculated structural stresses. The resulting structural strains
parameter is then used to obtain pseudo-elastic structural stresses
by applying Hooke’s law or multiplying Young’s modulus if
assuming uniaxial stress state prevails. Although showing an
improved fatigue life estimation over purely elastic-based assess-
ment procedure (i.e., without any plastic deformation consider-
ations) for some available low-cycle fatigue test data [8], there exist
a number of inconsistencies or weaknesses in that approach:
(a) Although a structural strain concept was first introduced in
Ref. [7], its implementation in elasticeplastic deformation
regime was largely incomplete in view of the fact that a local
strain definition had to be used in order to approximate
structural strain according Neuber’s rule that has been
typically used for notch stress and strain characterization
beyond elastic regime. As a result, the very structural strain
definition intended to characterize linear through-thickness
deformation no longer possess its original meaning;
(b) It is preferable that any low cycle fatigue correction proce-
dure should provide an indication on extent of plastic
deformation, e.g., elastic core size. The presence of an elastic
core is important since it help justify that an approximate
proportionality in fatigue damage accumulation so that an
elastic FE stress analysis can still be used in fatigue design. To
the authors’ best knowledge, existing low-cycle fatigue
procedures [9,11,12] for welded structures are not capable of
providing any information regarding if an elastic core is still
present, nor its size at a location of interest;
(c) Lastly, one advantage of the nodal force based structural
stress definition is its statically equivalent decomposition of a
through-thickness traction stress state in terms of mem-
brane and bending. Once Neuber’s rule is applied for esti-
mating structural strains in elasticeplastic deformation
regime, the previous method [9] is no long capable of
tracking membrane and bending composition or bending
ratio after calculating the pseudo elastic structural stress. As
a result, elastic bending ratio must be used for calculating the
equivalent pseudo-elastic structural stress range in order to
use the design master SeN curve. Fortunately, under strictly
load controlled conditions, fatigue lives are only weakly
dependent upon bending ratio [2]. However under-
displacement controlled conditions, a much stronger de-
pendency has been shown in Ref. [2]. Therefore, an improved
treatment of low-cycle fatigue is needed.
The purpose of this paper is to present a structural strain pro-
cedure that is consistent both with the mesh-insensitive structural
stress method [2,3] and the original intent expressed in Ref. [9]
when the terminology of structural strain was perhaps first intro-
duced for the treatment of low cycle fatigue. The plan of this paper
is as follows. We start with a brief discussion of some relevant
definitions and procedures associated with the mesh-insensitive
structural stress based master SeN curve method. Emphasis will
be placed upon how displacement-controlled low cycle fatigue data
had been interpreted in the development of the master SeN curve
covering both low-cycle and high-cycle fatigue regimes [2]. The
needs to develop a more general and consistent low cycle fatigue
evaluation procedure are then discussed. A structural strain defi-
nition valid for both elastic and elasticeplastic deformation re-
gimes is then presented. Analytical solutions of structural strains
and resulting elastic core size are then presented by assuming
elastic perfectly-plastic material response. Some existing low cycle
fatigue tests both under load- and displacement-controlled con-
ditions are analyzed to validate the effectiveness of the new
structural strain method. Finally, its implementation in supporting
the effective use of the master SeN method in 2007 ASME Code [1]
is discussed in light of the present development.
2. Master SeN curve and LCF data
The master SeN curve method [1e3,6] embodies two key
technical advances in finite element analysis based fatigue evalu-
ation: (1) a robust stress concentration calculation procedure based
on a novel nodal force method that is mesh insensitive at weld toe
or weld root; (2) a single master SeN curve representation of a
large amount of SeN data regardless of joint geometries, loading
modes, thicknesses, etc. A brief description of the master SeN curve
is provided here both for completeness and contrasting the differ-
ences between the present developments in this paper and the
early master SeN curve method. Detailed discussions on the master
SeN curve method, basic principles, numerical procedures as well
as validation examples can be found in numerous publications,
including [2e4].
2.1. The traction-based structural stress definition
As discussed in Ref. [2], a structural stress parameter can be
directly formulated by representing the traction conditions on a
hypothetical crack plane in the form of their respective membrane
and bending components. Consider a stress state at a weld toe on
the chord wall (e.g., a tubular T joint) along a through-thickness
hypothetical cut shown in Fig. 1a, the corresponding stress com-
ponents characterizing the traction conditions along cut plane AeA
are sx, sy , and sz under general loading conditions. Transverse shear
sz is often negligible. Within the context of structural mechanics,
these stress components are presented in the form of membrane
and bending which can be directly related to their corresponding
line force and line moments if shell or plate element models are
used. Such a characterization can be directly generalized to a 3D
geometry, such as along the entire curved weld line shown in
Fig. 1b. As such, a local coordinate system is used in Fig. 1b so that
 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51 41

sx0 is normal to the cut surface and sy0 is parallel, while transverse
shear is the local z0 direction.
The traction stresses on the curvilinear hypothetical cut in
Fig. 1b can be extracted using nodal forces (or internal forces) and
nodal moments (or internal moments) in shell/plate finite element
models or solid models. In view of the displacement-based finite
element theory used by all major commercial FE packages, these
nodal forces always satisfy equilibration conditions both with
respect to a node, nodes contained within each element or a group
of elements. Stresses however often do not satisfy equilibrium
conditions across a node or an element boundary, particularly with
the presence of geometric discontinuities such as at weld toe. Such
a traction-based structural stress definition is also consistent with
fracture mechanics principles in view of the fact that stress in-
tensity factors for a crack situated along the hypothetic plane in
Fig. 1 are fully described by the three traction components. Spe-
cifically, Mode I stress intensity factor is solely contributed by the
normal structural stress ss, Mode II by the in-plane shear structural
stress ss and Mode III by the transverse shear sz.
nodal forces in y0 direction while the lower case f1,f2,...,fn1 the line
forces in the same direction. In the matrix on the right hand side of
Eq. (1), li (i ¼ 1, 2, ., n1) represents the edge length of ith element
situated on the weld line. The corresponding line moments can be
calculated in an identical manner by replacing nodal forces
0
F1,F2,...,Fn1 in local y direction with nodal moments M1,M2,...,Mn1
0
with respect to x in Eq (1) above.
Note that nodal force Fi in Eq. (1) represents the summation of
the nodal forces at node i from the adjoining weld toe elements
0
situated on the positive side of y axis, regardless of the number of
elements involved, as shown in Fig. 1c. The linear system of equa-
tions described in Eq. (2) can be solved simultaneously to obtain the
line forces for all nodes along the weld line connecting all weld toe
nodes. Then, the normal structural stress in terms of membrane
and bending (or structural stress) can be calculated at each node as
(depicted in Fig 1):
fy0 6mx0
ss ¼ sm þ sb ¼  2 (2)
t t

2.2. Calculation procedure Similarly, through the use of the Eq. (1) with the nodal forces in
the x0 direction and nodal moments with respect to the y0 direction,
If a shell or plate element model is used (e.g., with respect to and the corresponding in-plane shear structural stress can be
plate mid-thickness surface in Fig. 1b as shown in Fig. 1c), the calculated as:
transformation of nodal forces and moments to work-equivalent
[2e4,6] line forces and moments along a weld line simply be-
fx0 6my0
comes inverting a matrix equation as described below. As an ss ¼ sm þ sb ¼ þ 2 (3)
t t
example, for weld toe cracking into the chord wall (Fig. 1b), the
nodal forces and moments along the curve but with respect to the and the transverse shear being given as, the same as that from
elements on the chord are first rotated into the local x0 y0 z0 co- strength of materials:
ordinate system from the global xyz system used in a FEA
analysis. For instance, the line forces in local y0 direction contrib-
fz0
uting to the membrane stresses normal to the weld line in Fig. 1c or sz0 ¼ : (4)
the curvilinear plane (A-A-C-C) in Fig. 1b. The line moments with t
respect to x0 contribute to the bending stresses in the same y0 di- If a 3D solid element model is used for modeling the same
rection. Imposing the work-equivalent argument for all nodes tubular T joint, a pre-processing step of translating nodal forces
describing the weld line in Fig. 2, one can obtain the following collected from the curvilinear hypothetical cut to the mid-surface
matrix form of a system of linear equations relating the nodal forces to obtain relevant nodal forces and moments that is required Eq.
in y0 direction to the nodal line forces as: (1).

2 3
ðl1 þ ln1 Þ l1 ln1
6 0 0 / 7
6 3 6 6 7
6 7
8 9 6 l1 ðl1 þ l2 Þ l2 78 9
6 0 / 0 7
>
> F1 >
> 6 6 3 6 7> f1 >
>
> > 6 7>> >
> f2 >
>
> F >> 6 7> >
>
< 2 > = 6 l2 ðl2 þ l3 Þ l3
« 7>< >
=
F3 6 0 0 7 f3
¼ 6 6 3 6 7 (1)
>
> « >
> 6 7> « >
>
> > 6 1 1 1 7>> >
>
>
> « >
>
> 6 0 0 0 7>>
> « >
>
>
: ; 6 7: ;
Fn1 y’ 6 ln3 ðln3 þ ln2 Þ ln2 7 fn1 y’
6 « 1 0 7
6 6 3 6 7
6 7
4 5
ln1 ln2 ðln2 þ ln1 Þ
/ 0 0
6 6 3

In the above equation, the use of linear shell or plate elements is
assumed. For parabolic elements, the corresponding equation is
given in Ref. [2]. For a closed weld line such as a typical tubular
joint, the weld line forms a closed curve around the brace and chord
intersection, i.e., there exist a total of n1 independent equations in
Eq. (1), where n is the total number of nodes describing the weld
line in Fig. 1c. The upper case F1,F2,...,Fn1 in Eq. (1) represent the
2.3. A calculation example
As an example, a curved lap fillet weld loaded under tension is
shown in Fig. 2. The representative finite element models with
different mesh refinements used in this study are also shown in
Fig. 2, in which mesh sizes in the weld area varying from 0.5t  0.5t
to 2t  2t, with t is the thickness of the base plate. For this joint
42 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51

Fig. 1. Traction-based structural stress definition with respect to chord wall in a tubular joint. (a) 2D equilibrium-equivalent stress definition with respect to Line AeA; (b) 3D
Equilibrium-equivalent stress definition on a hypothetical cut surface (local x0 ez0 plane) encompassing Line AeA.

type, the stress gradient along the weld line is relatively high due to
the curvature of the fillet weld line.
Fig. 3a shows the structural stress calculated along the weld toe
plotted as a function of the curvilinear distance along the weld toe
from the location exhibiting the highest stress concentration. The
3D solid element size varies from 0.25t to 2t (note that the mesh
corresponding 0.25t  0.25t is not shown in Fig. 2), with various
element types, e.g., linear elements with reduced and full integra-
tion schemes, and parabolic elements with reduced and full
integration schemes. The structural stress distributions for all these
cases show little difference from one to another, demonstrating a
remarkable mesh-size and element type sensitivity in the stress
concentration characterization at the weld toe.
For comparison purpose, a stress-based linearization procedure
[2] similar to ASME stress classification based linearization method
[14] was applied at the peak stress location (see Fig. 2) for all cases.
Results are summarized in Fig. 3b. The effectiveness of the mesh-
insensitive structural stress method (“Nodal Force-Based”) is

Fig. 2. A doubling plate fillet weld specimen and representative 3D solid models with different mesh sizes (quarter symmetry).
 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51 43

fully described by one normal traction stress and two shear stress
components [2]. All three must satisfy equilibrium conditions by
definition. This is why the method is mesh-insensitive as shown in
Fig. 3b, labeled as “Nodal Force-Based”.

2.4. Fatigue data correlation

Once a good mesh insensitivity of the traction-based structural

stress parameter has been demonstrated for various joint config-
urations given in Ref. [2], it can then be argued that the structural
stress so calculated at a fatigue prone location only depends on
joint geometry and loading modes without typical artifacts caused
by stress singularity at weld toe. Indeed, with fracture mechanics
considerations, an equivalent structural stress parameter

Dss
DS ¼ 2m 1
(5)
t * 2m $IðrÞm

was formulated and proven to effective in collapsing a large num-

ber of fatigue test results into a single narrow band, termed as a
master SeN curve [2e4]. In Eq. (5), t* is a relative thickness, with
respect to a unit thickness of, say, 1 mm (i.e. t* ¼ t/1 mm) as used in
Refs. [2e6] and m ¼ 3.6 derived based a two stage crack growth
model in Ref. [2]. The equivalent DS retains a stress unit. The life
integral I(r) is a dimensionless function of bending ratio r defined
as:

jsb j
r ¼ (6)
jsm j þ jsb j
Under load-controlled conditions, I(r) after numerical integra-
tion and curve fitting can be expressed as a dimensionless poly-
nomial function of the bending ratio r:
1
IðrÞm ¼ 0:0011$r 6 þ 0:0767$r 5  0:0988$r 4 þ 0:0946$r 3
þ 0:0221$r 2 þ 0:014$r þ 1:2223 (7)

The use of Eq. (5) resulted in a single narrow band representa-

tion of a large number of fatigue tests (close to 1000) in logelog
scale with a rather small standard derivation of about 0.25 as shown
in Fig. 4 [2]. The master SeN curve [1,2] is given in the form of,

Fig. 3. Comparison of the structural stress based SCF calculated at weld toe in a DS ¼ C$Nh (8)
doubling plate fillet weld specimen loaded in tension (meshes shown in Fig. 2).
where C and h are constants documented which the 2007 ASME Div
2 Code [1], obtained through statistical analysis of the fatigue tests
evident. Note that both linear solid element (8-noded) with full shown Fig. 4.
integration scheme (“8f”) and parabolic solid elements (20-noded)
with reduced integration scheme (“20r”) are used with respect to
each of the three mesh designs in Fig. 2. The good mesh-
insensitivity of the structural stress method is clearly shown in
Fig. 3b, whereas, the stress concentration factors (SCFs) calculated
using the stress linearization procedure [9] show a significant
variation when coarse meshes were used as shown in Fig. 3b,
labeled as “Stress-Based”.
It should also be noted that the stress-based linearization pro-
cedure discussed above in fact is not the same as the ASME line-
arization procedure [14], which requires linearization of all six
stress components at a given stress classification line. As such, the
equilibrium conditions can no longer be enforced with respect to all
the six stress components since three of out of the six components
never need to satisfy the equilibrium conditions. In contrast, the
mesh-insensitive structural stress definition in Fig. 1b only refers to
traction conditions by introducing a hypothetical cut at a stress Fig. 4. Master SeN curve definition and supporting test data developed for 2001 ASME
classification line [14]. The tractions along the hypothetical cut are Div 2 [1].
44 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51

Fig. 5. Typical ASME pipe weld fatigue test data interpretation procedures and derivation of pseudo elastic load (Fa) at a given displacement amplitude (da): (a) cantilever bending;
(b) four-point bending; (c) pseudo elastic stress determination using the extrapolated linear elastic port of the measured stiffness curve.

2.5. Treatment of LCF data displacement in a measured load-displacement curve can be

justified for comparing deformation behavior on the basis of a
It is important to note that some of the data are clearly situated stress unit with elastic deformation behavior as far as fatigue is
in low cycle regime (i.e., with cycles to failure being less than concerned. Furthermore, the presence of high stress concentration
104 w 105) and form a consistent narrow band with respect to high at welds will be accompanied by localized plastic deformation in
cycle fatigue data, as shown in Fig. 4. Most of these low cycle test low cycle fatigue. The presence of an elastic core ensures the val-
data are from full scale welded pipe tests reported in Refs. [10,11]. It idity of the assumption of local strain-controlled conditions at
should be pointed out that the structural stress range Dss in Eq. (5) stress riser locations, which are determined by global deformation
for these low cycle fatigue tests were calculated directly using or its corresponding pseudo-elastic load. This was one main argu-
linear elastic analysis by applying pseudo-elastic loads provided in ment (not yet quantitatively proven) given in Refs. [9e11] in
Refs. [9,10], which were based on extrapolation of linear portion of explaining why the low cycle fatigue data correlate so well with
load-displacement curves obtained prior to fatigue tests. Therefore, high-cycle data in Fig. 4, which extends to a very high cycle regime
the elastic structural stress calculated corresponding a given at about 108.
pseudo-elastic load range is actually a pseudo-elastic structural From a fatigue design evaluation point of view, it is highly
stress, without needing any further treatment. The use of a pseudo- desirable to rely on linear elastic finite element analysis for stress
elastic stress can be traced back to as early as Markl’s work [17] in calculations in low-cycle regime. Additionally, a low-cycle fatigue
the 1950s. As shown in Fig. 5 for either cantilever or four-point evaluation procedure should also be able of dealing with load-
bending cyclic fatigue tests, a cyclic loading was accomplished by controlled conditions. This is because a component can be of a
imposing a constant displacement amplitude (da). The corre- much more complex geometry (e.g., nozzle to shell connection in
sponding actual load amplitude the component experienced pressure vessel or a side-shell connection in ship structure) than a
should be Fm or measured from a load cell reading. For the given simple pipe under bending fatigue problems discussed above. In
displacement amplitude, a pseudo elastic load (Fa) was obtained by the former, often only applied loads (or pressure) are available for
extrapolating the linear elastic portion of the stabilized cyclic
stiffness curve up to a specified displacement amplitude (da), as
shown in Fig. 5c. The fatigue test data were then reported in the
form of a pseudo-elastic nominal stress amplitude calculated a
simple elastic beam bending formula under the pseudo-elastic
load. For the case of four-point bending (Fig. 5b), strain measure-
ments can be more reliably measured within the constant moment
regime. The pseudo stress was then simply defined as E  ε, i.e.,
Young’s modulus by the measured longitudinal strain. If a compo-
nent is tested in the linear elastic regime, the pseudo elastic stress
becomes the actual stress experienced by the component.
For simple beam bending problems such as pipes, a load-
displacement curve provide a direct indication of beam stress and
deformation as a components goes through cyclic loading condi-
tions beyond linear elastic limit. Even if plastic deformation may be
somewhat significant, the presence of an elastic core of a sufficient
size is all that is needed to justify some of the basic assumptions on
deformation behavior in plate and shell theories [19,20]. It then
follows that a pseudo-elastic stress based on an actual Fig. 6. Structural strain definition in elastic and plastic deformation regime.
 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51 45

performing fatigue evaluation. Test-based load-displacement Then, a pseudo elastic structural stresses analogous to the
curves are often not available, nor could such curves be even pseudo-elastic nominal stresses used in Refs. [11,12,16] in low-cycle
meaningful at all for inferring relevant pseudo-elastic loads for fatigue testing can then be obtained by multiplying Young’s
stress calculation purposes. For this reason, a new structural strain Modulus of material as:
method is presented below, which is based on an analytical
formulation for estimating structural strains from traction stresses E sm t
s0m ¼
calculated using linear elastic finite element analysis using Eqs. (1) RSY
(14)
through (4). Et
s0b ¼
2R
3. Structural strain method

if only is the normal traction stress operative.

Consistent with classical plate theory, a structural member
going through elastic or elasticeplastic deformation is assumed to
3.2. Membrane-dominated loading
deform in such a way that a through-thickness normal to plate mid-
surface before deformation remains straight after deformation. For
Refer to Fig. 6, if sm is relatively high, plastic deformation may
simplicity but without losing generality, the material response can
only occur near the outer surface, i.e., s0i  SY , by a similar derivation
be further assumed to be isotropic and elastic perfectly plastic.
process to the one shown in the previous case, the elastic core size c
Additionally, it is assumed that only the normal traction stress is
can be expressed as:
operative. Two cases are considered here: one is bending-
dominated loading that may generate plastic deformation at both  
plate surfaces; the other is membrane-dominated loading that may t 3SY  3sm  sb
c ¼ (15)
develop plastic deformation at one plate surface. 2 SY  sm

3.1. Bending-dominated loading

resulting in a curvature of bending as,
For a through-thickness hypothetical cut, an elastically calcu-
1 8ðSY  sm Þ3
lated normal traction stress is depicted as sm and sb in Fig. 6. The ¼ (16)
combined normal traction stress in this case exceeds material yield
R tEð3SY  3sm  sb Þ2
strength SY. Therefore, the linear-elastically calculated normal
traction stress distribution in terms of sm and sb must be re- Then, structural strains at both outer and inner surfaces have the
distributed to satisfy equilibrium conditions (i.e., force and following expressions:
moment balances) and yield criterion (here it is assumed von Mises
criterion is applicable). The resulting stress distribution (see SY ðt  cÞ
ε0 ¼ þ
Appendix for derivation details) is shown as the thick lines in Fig 2. E R
(17)
Parameters c and e represent the corresponding elastic core size SY c
and shift of the neutral axis of bending, respectively, εi ¼ 
E R
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u "  2 # The pseudo-elastic structural stresses become:
u sm 2s
c ¼ t t3 1   b (9)
SY 3SY E
s0m ¼ SY þ ðt  2cÞ
2R
(18)
sm t tE
e ¼ (10) s0b ¼
2SY 2R

Eq. (9) yields the well-known classical limit stress state of It is not surprising that the pseudo-elastic bending stress
sb ¼ 3SY/2 if sm ¼ 0 and c ¼ 0. A non-zero elastic core condition can component s0b in Eq. (18) has an identical expression to the one in
be stated as: Eq. (14), since bending stress by definition is only related to cur-
c t vature 1/R and Young’s Modulus.
< e (11) The above developments can be readily extended to plane strain
2 2
conditions (i.e., εx0 ¼ 0 in Fig. 1b) by simply replacing material yield
Assuming the presence of elastic core dominates the through- strength SY in the above with an effective yield stress,
thickness deformation behavior, one can show the resulting cur-
vature of bending becomes: SY
S0Y ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1  v þ v2
1 2SY
¼ (12)
R cE if von Mises yield criterion is used and replacing Young’s Modulus E
in Eqs. (14) and (18) by
where R represents the bending radius or radius of curvature. Then,
the structural strains at both outer and inner fibers become: E
E0 ¼
1  v2
 
1 t
εo ¼ eþ where v is Poisson ratio. More realistic stressestrain curves with
R 2
  (13) strain hardening effects can be considered, but, numerical tech-
1 t niques would have to be resorted to solve the resulting structural
εi ¼ e
R 2 strains.
46 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51

Fig. 7. Flat heat vessel geometry and fatigue failure location tested under cyclic pressure conditions [15].

3.3. Validation by LCF test data welding processes, and as-build thicknesses are listed in Table 1
taken from Ref. [15]. Detailed test conditions can be found [15].
Available low cycle fatigue tests from literature are analyzed Among the five vessel tests, Vessel #4 (referred to as Joint #4) had a
here to validate the structural strain method discussed in the pre- different weld type and its failure origin was somewhat subjected
vious section. to interpretation as discussed in [54]. Therefore, Vessel #4 will not
be included here. A linear elastic finite element analysis was per-
formed in this investigation based on nominal vessel dimensions
3.3.1. Flat heat vessel tests
and weld geometry and the resulting traction-based structural
Fatigue test results (using cyclic pressure) of five full scale flat
stress of sm ¼ 12.7 MPa and sb ¼ 481.5 MPa using the post-
vessels were recently reported in Ref. [15], among which four are
processing procedure described in Eqs. (1) and (2). The calculated
carbon steel vessels (SA-516-70 type with a nominal yield of
traction stress (membrane plus bending) exceeds material yield
350 MPa) and one is stainless steel vessel (SA-240 304 type with a
strengths for both the carbon steel and stainless steel vessels. The
nominal yield of 310 MPa). The vessel geometry is shown in Fig. 7,
corresponding structural strains are estimated through Eq. (13) and
including weld geometry details at failure location. Materials,
pseudo-elastic structural stress through Eq. (14). The latter is only
intended to compare the vessel test results with the data forming
Table 1 the master SeN curve in Fig. 4 through Eq. (5).
Master SeN curve parameters. The results are plotted in Fig. 8 against the 2007 ASME Div 2
Statistical basis C h master SeN curve in terms of its mean, mean 2s and mean 3s,
where s represents the standard derivation defined with respect to
Mean 19930.2 0.32
þ2s 28626.5 cycle to failure in log scale. In Fig. 8, the four vessel test data are
2s 13875.8 shown as solid symbols if the elastically calculated structural stress
þ3s 31796.1 range is directly used (i.e., without applying the low-cycle fatigue
3s 12492.6 correction scheme based on new structural strain method) and as
DSs ¼ C  Nh. empty symbols if the new structural strain method is used by
 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51
Fig. 8. Comparison of flat head vessel test data with master SeN curve scatter band
(Lines: mean master SeN curve  2s and 3s; solid symbols: Neuber’s rule based
correction [10]; empty symbols: pseudo-elastic structural stress based the structural
strain method).
means of Eqs. (13) and (14). Without low cycle fatigue correction,
most of the four vessel data are situated near the mean 2s line.
With the new low cycle fatigue treatment procedure presented
here, the improvement in correlation between the vessel test data
and the existing master SeN curve can be clearly seen.
3.3.2. Longitudinal gusset specimens
A large number of double-sided longitudinal gusset joint spec-
imens were tested in both low cycle and intermediate cycle fatigue
regimes, as reported in Ref. [18] with an attempt to develop a
plastic strain range based approach, which was shown not that
promising. The specimen geometry is shown in Fig. 9a. These tests
were performed under load-controlled conditions from zero to a
specified peak load and continued until failure (defined as com-
plete separation) [18]. For this type of joints, a 3D finite element
Fig. 9. Test specimen, finite element model, and traction-based structural stress calculation results.
47
model is required to calculate the traction-based structural stress
concentration factor (SCF) under linear elastic conditions. The
normalized structural stress values or SCF are: sm/sn ¼ 1.35 and sb/
sn ¼ 0.58., where sn is applied nominal stress used in tests. The test
specimens were made of three base materials of significantly
different yield strengths [18].
After completing the elastic structural stress calculations using
the nodal force method described in the earlier section (Eqs (1)e
(2)), it becomes apparent that these tests are all in the category
of all membrane stress dominated conditions as discussed in a
previous section. In addition, in some of the low cycle tests,
membrane stresses (sm) calculated can be near and even beyond
material yield strengths. Eq. (12) indicates that as long as the
membrane stress becomes close to yield, any through-thickness
stress concentration would deem the disappearance of elastic
core. Furthermore, in some of these low cycle fatigue tests, applied
nominal stress amplitude had already exceeded material yield
strength, sometime by a significant margin. Therefore, analysis of
these test data requires some special considerations beyond the
structural strain method described through Eqs. (14)e(18) in which
the presence of an elastic core is a prerequisite. It should be pointed
out that in practice, cases with membrane stresses or applied
nominal stresses exceeding material yield strength are unlikely to
occur as a low cycle fatigue problem. This is because static strength
design criteria would have screened out these cases before fatigue
evaluation becomes warranted. Nevertheless, from a methodology
demonstration point of view, these tests should be worthy of
analyzing, particularly in view of the unique specimen geometry
used by the authors [16], which involve very high stress
concentration.
As illustrated in Fig. 10, for a given applied nominal stress
amplitude (sn) that’s beyond yield strength SY, a nominal strain
amplitude (based on the distance between grips) can be deter-
mined as εn. Again, uniaxial stress/strain conditions are assumed
here for simplicity. Similar to the pseudo-elastic load determina-
tion procedure shown in Fig. 5, the pseudo-elastic nominal stress
48 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51
Fig. 10. Application of structural strain method in cases where elastic core is no longer
present.
can be found as sp,n at Point A0 whereas the actual nominal stress is
at Point A. The traction structural stress based SCF calculated from
the finite element model (Fig. 9) can then be used to obtain the
pseudo-elastic nominal stress to obtain the structural stress
amplitude at weld toe ( i.e. SCF  sp,n,) in which the highest SCF
occurs at the weld end on the symmetry plane (Fig. 9c), as expected.
After attaining its peak, the applied nominal stress is reduced to
zero in an linear elastic manner in completing the first cycle with a
pseudo-elastic nominal stress arrange of sp,n. Upon reloading to the
same nominal stress level of sn, no further plastic strain is gener-
ated. Therefore, the corresponding pseudo-elastic nominal stress
becomes the same as the actual nominal stress applied for the rest
of loading cycles until failure. The corresponding pseudo-elastic
structural stress becomes SCF  sn for the remaining loading cy-
cles after the first.
To plot this set of test data which involve two blocks of cyclic
loadings against the master SeN curve, a Miner’s rule based
equivalent stress corresponding to the total number of cycles can be
used as below:
 h
1 X
DSeq ¼ ni DSi (19)
ne
P
where, ne ¼ ni, ni and DSi are the number of cycles contained in
ith loading block and corresponding equivalent structural stress
range according to Eq. (5), respectively. The results are shown in
Fig. 11. The empty symbols represent a direct use of Eq. (5) with
considering low-cycle fatigue or plastic deformation effects. The
solid symbols of the same shapes represent the use of the pseudo
elastic structural stress definition shown in Fig. 10 and Eq. (19).
Fig. 12. Cruciform joints used by DSME [14] and 2D finite element model used in this study for traction structural stress calculation.
Fig. 11. Comparison of low-cycle fatigue test data with the master SeN curve scatter
band: with and without structural strain based corrections: (Lines: mean master SeN
curve  2s and 3s; empty symbols: without structural strain based correction [10];
solid symbols: with structural strain based correction).
When elastically calculated structural stresses are near material
yield strengths, the structural strain method gives essentially the
same results as before, as shown in Fig. 10 for cycle to failure
beyond 2000 cycles. Below 1000 cycles to failure, the use of
pseudo-elastic structural stress parameter is essential in correlating
these data point with the master SeN curve scatter band, further
proving the validity of the structural strain method.
3.3.3. Cruciform joint tests
Heo et al. [13] recently performed a large number of cruciform
joint low cycle fatigue tests under displacement controlled condi-
tions (see Fig. 12). Similar to the longitudinal gusset specimens
discussed in the previous section, the test data reported by Heo
et al. [13] all involved significant plastic deformation since the
associated membrane traction components all exceeded reported
yield strengths of the materials. By following the same procedure
discussed in the previous section, but starting with applied nominal
strain ranges given in Refs. [13], elastic finite element analysis using
a quarter-symmetry model (Fig. 12b) was used to calculate the
traction-based structural stresses SCF under a given uniform
displacement range consistent with a gauge length of 100 mm
which ws used in performing these fatigue tests in Ref. [13].. Then,
this SCF is used to scale all load cases corresponding to nominal
stress ranges corresponding applied strain ranges. The test results
are shown in Fig. 13 in terms of equivalent pseudo-elastic stress
range versus cycle to failure against the master SeN curve scatter
band. All test data fall within narrow band defined by the mean 2s
lines of the master SeN curve.
4. Discussions
4.1. Structural strain versus structural stress
The mesh-insensitive structural stress method was first pro-
posed [5] for achieving stress calculation consistency at a notch
location such as at weld toe where local stresses are ill-defined due
 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51
Fig. 13. Analysis of DSME’s low cycle fatigue data using structural strain approach and
comparison with master SeN curve scatter band.
to the presence of singularity. This was accomplished through an
equilibrium-equivalent transformation of nodal forces/moments
available from standard finite element solutions (e.g., Eq. (1)) into
line forces/moments. A structural stress definition was then intro-
duced [2e5] based on classical structural mechanics considerations
in terms of membrane and bending components [19]. This is one
main reason that the method has been referred to as a structural
stress method to emphasize its structural mechanics root in rep-
resenting a through-thickness stress state. However, there appears
to be some confusion over the differences between the traction-
based structural stress method and surface stress extrapolation
based hot spot stress methods, since the letter have also been
referred to as a structural stress method in some recent publica-
tions, e.g., in Ref. [19]. It must be emphasized here that the former
calculates the structural stress by enforcing equilibrium conditions
both with respect to plate thickness and along a weld line through
Eq. (1), which is an analogous to an arbitrary free-body cut that
exposes up to three traction stress components as shown in Fig. 1,
while the latter doesn’t.
Along this line, it is more appropriate to refer the mesh-
insensitive structural stress method as a traction-based structural
stress method [2,7]. More importantly, the linear traction stress
description can be directly inferred from the structural strain
definition introduced in this paper in linear elastic deformation
regime within the context of classical plate and shell deformation
theory [19]. Beyond elastic deformation limit, the structural strain
definition remains valid as long as an elastic core remains dominant
[20]. Furthermore, as demonstrated in this work, the structural
strain method can still yield rather reasonable results in fully plastic
deformation regime for fatigue data correlation purposes, as
demonstrated in this study.
The introduction of the pseudo-elastic structural stress defini-
tion (e.g., in Eqs. (14) and (18)) is not necessary in concept. Its only
intent is to convert actual structural strain to a fictitious stress
definition so that the low-cycle fatigue data can be presented in the
same stress based framework as high cycle fatigue data shown in
Fig. 4 for comparison purposes. Obviously, for data that involves
negligible plastic deformation, the pseudo-elastic structural
stresses are the same as actual elastic structural stresses according
to the structural strain method discussed in this paper. In fact, it is
structural strain that serves as a fundamental parameter for char-
acterizing fatigue damage of welded joints in both low-cycle and
high-cycle regime, as demonstrated in this paper. By tradition, such
as those stated in ASME Codes and Standards (see Ref. [1]), engi-
neers tend to relate to stresses better than strains.
49
The structural strain method presented here should not be
confused with traditional strain life methods. The key difference
between the two is that structural strain method derives strain at a
hypothetical crack plane, which satisfies a plane-remaining-as-
plane condition, while strain life method uses local strain
definition.
4.2. Elastic core
Applications of the structural strain method in low-cycle fatigue
evaluation fall into two major categories: (1) load-controlled con-
ditions; (2) displacement-controlled conditions. Under load-
controlled conditions, if material yield limit is not exceeded, elas-
tically calculated traction structural stresses are directly used in Eq.
(5). Otherwise, if non-zero ( i.e., c>0) elastic core conditions (Eqs.
(9) and (15)) are met, structural strains can be analytically calcu-
lated through Eqs. (13) and (17). For applications involving full
thickness yielding (i.e., c¼0), although unlikely to occurring as a
low-cycle fatigue design problem in practice, a reference strain
corresponding to remote loading conditions is required to apply the
method, as shown in Fig. 10. Under displacement-controlled con-
ditions, the traction stresses calculated through Eqs. (1)e(4) using
linear elastic FE models directly provide the structural strain based
pseudo-elastic structural stresses for fatigue design and test data
correlation purpose through Eq. (5). No further treatment is
needed.
There are a number of areas warranting further investigations
that are currently on-going. One of them is that by definition, cyclic
yield strength SY should be used in calculating the structural strains
for a given application. However, this type of material data may not
be readily available. This is the case for all low-cycle test data of
welded joints analyzed in this investigation in which nominal
monotonic yield strengths have been used due to the lack of in-
formation on cyclic yield strengths for the materials involved. This
undoubtedly introduces errors in addition to the assumption of
elastic perfectly plastic hardening behavior (e.g., see Fig. 6). How-
ever the errors tend to be less significant if elastic core size (c) is
sufficiently large. For illustration purposes, Fig. 14 presents the
pseudo-elastic structural stresses calculated using Eq. (11) as a
function of structural stress component sb. Effects of membrane
structural stress sm are also shown. All stress parameters are
normalized by material yield strength SY. The corresponding elastic
core size (c/t) for each case is given as dashed lines. Under bending
stress dominated conditions without any strain hardening and at a
given sm/SY, the pseudo-elastic structural stress initially increases
in an approximately linear manner with an increasing bending
stress sb/SY until elastic core c/t reduces to a critical value. Then, the
Fig. 14. Relationship among normalized pseudo-elastic stress, elastic structural
bending stress, elastic membrane stress, and elastic core size.
50 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51

pseudo-elastic structural stress increases rapidly as c/t approaches (b) The fact that both low-cycle and high cycle fatigue test data
zero. This trend becomes much more pronounced as membrane follow into the same scatter bands as long as the structural
structural stress (sm/SY) increases. The turning point at which c/t strain parameter is used regardless of joint geometries and
decreases from a linear to a nonlinear dependency on sb/SY can be loading types, etc implies that structural strain (rather notch
defined as the critical elastic core size for each sm/SY ratio. Above strain or notch stress) is a fundamental parameter for char-
the critical elastic core size, it may be argued that plastic defor- acterizing fatigue of welded joints. The conversion of struc-
mation is contained and that the error in using nominal yield tural strain as introduced in this paper to a pseudo-elastic
strength should be limited in view of the typical scatter in fatigue structural stress is not essential, but only as a way to present
data (see Fig. 4). If below the critical elastic core size, the use of the both low-cycle and high cycle fatigue data in a stress-based
structural strain method, at least Eq. (11), may become question- SeN curve form that is typically used in Codes and Stan-
able. More research is needed to further clarify the limits of the dards [e.g., [1,2, 19]]. Note that in high cycle regime, traction
present approach, particularly as more relevant test data become based structural stresses are directly proportional to struc-
available. tural strains.
(c) A statically equivalent membrane and bending based stress
4.3. Membrane/bending decomposition decomposition is essential in arriving at consistent solutions
of structural strains that are valid in both elastic and elastice
The importance of decomposing a through-thickness stress plastic deformation regimes. As a result, elastic core condi-
state into membrane and bending components in a statically tions can be quantitatively determined, leading to a more
equivalent manner cannot be overstated in view of membrane clearly defined application limits in order to maintain an
stress effects shown in Fig. 14 in low-cycle fatigue regime. In high- adequate elastic core size implied by some of structural
cycle regime, I(r) in Eq. (5) measures bending ratio (r) effects on design Codes and Standards such ASME Div 2 [1,2].
fatigue, which shows a less dependency on r under load controlled (d) For applications in which elastic core is no longer present at a
conditions (see Eq. (6)) than displacement controlled conditions location of concern, the same structural strain method can
[2]. It is worth pointing out that surface extrapolation based hot still be applied for test data analysis purposes, as demon-
spot stress method [12,13,19] cannot separate bending stress strated in this paper. Howzever, it should be noted that in
component from membrane, at least not in its present form [18]. An such situations static strength design criteria might have
earlier version of the structural strain method reported by the first already predicted structural deficiencies before invoking the
author and his former co-workers [9] suffers the same problem, in needs for low-cycle fatigue evaluation.
which a Neuber’s rule was introduced to convert elastically calcu-
lated structural stresses to local strains that were then used as
approximated structural strains. In doing so, the bending content of Acknowledgments
their pseudo-elastic structural stress can’t be correctly determined
and was approximated by using elastic structural stresses [9]. A The work reported here was supported by the National Research
consistent separation of membrane and bending stress compo- Foundation of Korea (NRF) Grant funded by the Korea government
nents, which is inherent in the traction based structural stress (MEST) through GCRC-SOP at University of New Orleans under
method [2e5], proves to be essential in arriving at the final struc- Project 2-1: Reliability and Strength Assessment of Core Parts and
tural strain and elastic core solutions presented in this paper. As a Material System.
result, application limits of the new low-cycle fatigue evaluation
procedure based on the new structural strain method can now be
Appendix A
better defined and areas of further improvement can be identified
for further investigation.
Derivation of structural strain
5. Concluding remarks
Refer to Fig. 6, for simplicity, elastic perfectly plastic material is
assumed with yield strength of material designated as SY. The
In this paper, a new structural strain method is presented as a
resulting stress distribution (thick lines) must satisfy equilibrium
post-processing calculation procedure after traction-based struc-
conditions:
tural stresses become available through linear elastic finite element
analysis. The structural strain method is formulated based on the   
t t
basic assumption of classical plate and shell theory in which linear sm t ¼ SY  ðc  eÞ  SY ce
2 2
through-thickness deformation gradient is assumed a priori. Under (A1)
the assumption of elastic and perfectly plastic materials, a set of t 2 sb
¼ m1 þ m2 þm3
analytical solutions of structural strain and elastic core size have 6
been developed and validated using available test data. The method
can be readily extended to applications to materials that straining where,
hardening effects need to be considered, for which numerical so-   
t 1 t
lution techniques can be used for obtaining final structural strain m 1 ¼ SY  ðc  eÞ  þ ðc  eÞ
solutions. Major findings resulted from this study can be summa- 2 2 2
rized as follows: 1
m2 ¼ ð2cÞ2 (A2)
6
(a) The proposed structural strain method has been shown not   
t 1 t
only effective for treating low cycle fatigue problems, but m3 ¼ SY  ðc þ eÞ  þ ðc þ eÞ
2 2 2
also fully encompassing the earlier mesh-insensitive struc-
tural stress based master SeN curve approach that was Eq. (A1) states balance of forces and Eq. (A2) balance of mo-
developed mainly for dealing with high-cycle fatigue prob- ments. Since c > 0 represents non-zero elastic core, Eq. (A1) leads
lems associated with welded joints. to:
 P. Dong et al. / International Journal of Pressure Vessels and Piping 119 (2014) 39e51 51

Conference on Offshore Mechanics and Arctic Engineering; June, 2004. Paper

sm t No. OMAE2004e51228, Vancouver, British Columbia, Canada.
e ¼ (A3) [8] Kyuba H, Dong P. Equilibrium-equivalent structural stress approach to fatigue
2SY
analysis of a rectangular hollow section joint. Int J Fatigue 2005;27:85e94.
[9] Dong P, Cao Z, Hong JK. Low-Cycle Fatigue Evaluation Using the Weld Master
Substitution of Eq. (A3) into the second equation of Eq. (A2)
S-N Curve. In: Proceedings of ASME PVP 2005 Conference (PVP2006-ICPVT11-
yields elastic core size: 93607); July, 2005. Vancouver, Canada.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi [10] Scavuzzo RJ, Srivatsan TS, Lam PC. Fatigue of butt-welded pipe. Report 1 in
u "  2 # fatigue of Butt-welded pipe and effect of testing methods. Welding Research
u sm 2s
c ¼ t t3 1   b [11]
Council Bulletin; July 1998. 433.
Wais E, Rodabaugh EC. Investigation of torsional stress intensity factors and
SY 3SY
stress indices for girth welds in straight pipes. EPRI Report, #1006905, Palo
Alto, CA; April, 2002.
The above derivation can be readily applied for linear hardening [12] Wang X, Kang JK, Kim Y, Wirsching P. Low cycle fatigue analysis of marine
or power-law hardening materials. However, numerical solutions structures. Paper No. OMAE 2006-92268. In: Proc OMAE 2006, 25th Int. Conf.
on Offshore Mechanics and Arctic Engineering; June 4e9, 2006. Hamburg,
would be required for computing both the neutral axis shift e and
Germany.
elastic core size c. [13] Heo J-H, Kang JK, Kim Y, Yoo YS, Kim KS, Urm HS. A study on the design
guidance for low cycle fatigue in ship structures. In: Proceedings of the 9th
Symposium of Practical Design of Ships and Other Floating Structures; 2004.
References Luebeck-Travemunde, Germany.
[14] Hinnant C. Fatigue testing and life estimates of welded flat head pressure
vessel joints. In: Proceedings of PVP2006-ICPVT-11 2006 ASME Pressure
[1] 2007 ASME Boiler and Pressure Vessel Code, Sec VIII, Div. 2 (ASME BPVC-VIII- Vessel & Piping Conference; July 23e27, 2006 (Paper No. PVP2006-ICPVT11-
2e2007). New York, NY: American Society of Mechanical Engineers; July 93967), Vancouver, BC, Canada.
2007, ISBN 0791830691. [15] Korenke WC. Classification of finite element stresses according to ASME
[2] Dong P, Hong JK, Osage DA, Dewees DJ, Prager M. The Master S-N Curve section III stress categories. In: Proceedings of ASME Pressure Vessel and
Method: An Implementation in 2007 ASME Div 2 International Codes for Piping Technology Conference: A Decade of Progress. New York, New York:
Boilers and Pressure Vessels. Welding Research Council Bulletin, No. 523, ASME; 1985. pp. 107e40.
December, 2010, New York, New York. [16] Markl ARC. Fatigue tests of piping components. Trans ASME 1952;74:287e
[3] Dong P, Hong JK. Fatigue Tubul Joints: Hot Spot Stress Method Revisited. 303.
ASME Trans J Offshore Mechanics and Arctic Engineering August 2012;134. [17] Harrison JD. Fatigue performance of welded high strength steels, a compen-
031602e1. dium of reports from sponsored research programme. Abington Hall, Arbin-
[4] Dong P. A structural stress definition and numerical implementation for fa- ton, Cambridge CBI 6AL, England: The Welding Institute; 1974.
tigue evaluation of welded joints. Int J Fatigue 2001;23(10):865e76. [18] Hobbacher A. Recommendations for fatigue design of welded joints and
[5] Nie C, Dong P. A traction stress based shear strength definition for fillet welds. components. IIW Document, XIII-2151e07/XV-1254-07. International Insti-
Int J Strain Anal 2012;47(8):562e75. tute of Welding (IIW); May 2007.
[6] Dong P. A robust structural stress method for fatigue analysis of offshore/ [19] Timoshenko S, Woinowsky-Krieger. Theory of plates and shells. 2nd ed.
marine structures. ASME Trans J Offshore Mech Arct Eng 2005;127:68e74. McGraw-Hill Book Company; 1959.
February. [20] Hill R. The mathematical theory of plasticity. New York: Oxford University
[7] Healy B. A case study comparison of surface extrapolation and Battelle Press, New York; 1956.
structural stress methodologies. In: Proceedings of 23rd International

View publication stats