International Journal of Fatigue

International Journal of Fatigue 146 (2021) 106144
Contents lists available at ScienceDirect
International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue
An overview and comparative assessment of approaches to multi-axial

fatigue of welded components in codes and standards
Jifa Mei a, *, Pingsha Dong b, *, Shizhu Xing a, Anoop Vasu a, Alain Ganamet a, Jerry Chung a,
Yogesh Mehta a
a
American Axle & Manufacturing, Inc, Detroit, MI, United States
b
University of Michigan, Ann Arbor, MI, United States
A R T I C L E I N F O A B S T R A C T
Keywords: Three types of commonly used local stress approaches recognized by Codes and Standards for fatigue of welded
Multi-axial fatigue joints are reviewed and evaluated with respect to a comprehensive collection of multi-axial fatigue test data
Welded joints available in literature. These local stress approaches are surface-extrapolated hot spot stress, effective notch
Non-proportional loading
stress and traction structural stress. The hot spot stress and notch stress approaches are based on those stipulated
Traction structural stress
in Eurocode 3 (2005) and IIW Recommendations (2016). Traction structural stress is based mainly on ASME VIII
Hot spot stress
Effective notch stress Div 2 Code (2007) and its latest development. The results show traction structural stress method offers a
Residual stress relief significantly improved data correlation over hot spot stress and notch stress approaches.
1. Introduction (a) The majority of these test results are presented in terms of nom
inal stress definitions and a clear or consistent way of combining
The multi-axial fatigue of welded joints can be introduced either by normal and shear stress is typically lacking.
the multiple sources of cyclic loadings or by a complex load transfer of a (b) Consistent with key findings in test data for non-welded compo
dominant external cyclic load applied at a remote location. When nents, most of the test data for welded components shows that
several independent load inputs are applied asynchronously, stress non-proportional loading can cause significant additional fatigue
components at a potential failure location can vary independently. As damage, as much as up to a factor of 10 [9] when compared with
such, it is often referred to as non-proportional multi-axial fatigue proportional loading. However, some tests [14] also demonstrate
loading. In contrast, proportional multi-axial fatigue loading refers to the insensitivity of welded joint to the phase difference between
the situation when stress components vary over time in proportion with tensile and shear loading applied at the same time.
each other. (c) When dealing the asynchronous loading, i.e. the different loading
The presence of a multi-axial stress state at welded joints is almost frequencies between tensile loading and shear loading, some
inevitable in engineering structures subjected to a variety of loading unresolved issues include how to define a fatigue cycle for the
scenarios. However, the study of multi-axial fatigue of welded joints has loading event and how to combine tensile and shear stresses
not been carried out as extensively as that of non-welded components [7,13].
[1–3]. There are only limited amount of experimental studies [4–14] (d) A large percentage of tested components were subjected to post-
carried out thus far. Most of them clearly demonstrated a need in weld heat treatment for relieving residual stress before multi-
developing improved multi-axial fatigue assessment procedures for axial fatigue testing. It is still not clear how much benefit the
taking into account of both normal and shear stress effects properly process can bring in terms of fatigue life as compared with as-
when more than one dominant stress component are involved. Some welded components.
general observations on the challenges of treating multi-axial fatigue of
welded joints can be made by examining these available tests as a whole: To provide guidance to the engineering community, these issues
need to be clarified in terms of how they are treated by each of the
codified procedures and the resulting effectiveness in correlating the
* Corresponding authors.
E-mail addresses: meijifa@umich.edu (J. Mei), dongp@umich.edu (P. Dong).
https://doi.org/10.1016/j.ijfatigue.2021.106144
Received 3 November 2020; Received in revised form 29 December 2020; Accepted 4 January 2021
Available online 6 January 2021
0142-1123/© 2021 Elsevier Ltd. All rights reserved.
J. Mei et al. International Journal of Fatigue 146 (2021) 106144
multi-axial fatigue test data available to date. capabilities in treating arbitrary multi-axial loading conditions through
One possible reason for not being addressing some of the above is the path-dependent maximum range (PDMR) cycle counting method
sues effectively in the past is that the current methodologies in the most [25–27] and moment of load path (MLP) based multi-axial fatigue
of the current Codes and Standards [15–17] for fatigue design and damage parameter [28–30].
assessment of welded joints were originally developed by focusing on Given the major differences among these three local stress ap
only one nominal stress component, i.e. the normal stress at a weld toe proaches for treatment of multi-axial fatigue of welded joints, there is a
location. This is exemplified by the fact that a large portion of current growing interest in a comparative assessment of their relative effec
codes and standards [15–17] is dedicated to dealing with welded joints tiveness by evaluating available component test data from various
under normal loading directions, as illustrated by different classes of sources. Along this pathway, Bäckström and Marquis [31] analyzed
weld details together with applicable loading conditions. Guidelines multi-axial test data available by using the extrapolation based hot spot
given for treating multi-axial fatigue are rather limited, e.g., through the stress method and presented a comparative analysis of available multi-
use of an interaction equation representing linear damage accumulation axial test results using various existing interaction equations. More
caused by tensile and shear stress components [16,17]. Therefore, no recently, Pedersen [32] also carried out an evaluation of existing test
direct interactions are actually considered between tensile and shear data by adopting the effective notch stress approach. At the same time,
stress in this type of interaction equations. As far as non-proportional Mei and Dong [28] examined how traction structural stress can be used
loading is concerned, its effect on fatigue damage is either not taken to correlate the same well-documented multi-axial test data. In their
into account or accounted for empirically by imposing a smaller study, both a path-dependent maximum range (PDMR) cycle counting
allowable damage. More details on this topic will be reviewed in a later procedure and the moment of load path (MLP) based damage parameter
section. were used and were shown effective for treating both synchronous and
In addition to the conventional nominal stress approach used in asynchronous sinusoidal bending and shear stress histories. As a result,
combination with joint classification, different definitions of local they showed that a large amount of multi-axial fatigue test data
stresses have been adopted in codes and standards [15–17] and can be collected from welded components can be effectively collapsed into a
used for multi-axial fatigue assessment. Among them, hot spot stress narrow scatter band.
(geometric stress), effective notch stress and traction-based structural However, a comprehensive comparative study remains to be done for
stress are the most commonly practiced methods. Hot spot stress is a demonstrating relative effectiveness of the three local stress based
stress definition aiming to include all stress-raising effects of a structure methods. To do so, the first part (Section 2) of this paper starts with a
detail but excluding the local stress caused by the weld profile itself critical assessment of the basic assumptions, the possible physical ex
[17,18]. Such a hot spot stress is calculated by extrapolation of surface planations underlying the three different local stress approaches and
stresses from predefined reference points to the hot spot location (where their relative applicability for consistently treating multi-axial fatigue
fatigue failure is expected to occur). Depending on the types of joints, problems. They are generally termed as local stress approaches here
finite element mesh sizes and element types, different extrapolation because of the local concept involved for the fatigue assessment, in
procedures can be used as recommended in IIW Recommendations [17]. contrast to the classical nominal far field stress concept. Among the
Due to its dependency on element size and type, the robustness of the three types of local stress approaches evaluated, traction structural
method is still subjected to debate since mechanics-based interpretation stress is examined in greater detail since it is found to be more effective
of reference points for surface extrapolation is lacking. Furthermore, for in data correlation, as shown in the later part of the report. Also note that
the hot spot stress based fatigue evaluation, several reference joint de various other approaches, e.g., strain energy density [33] will not be
tails and multiple S-N curves are still needed. covered in this assessment since these methods have not been adopted in
The effective notch stress approach aims to take into account of both dominant Codes and Standards. Then, the three local stress based
structural and local stress riser effect, including the one caused by weld methods stipulated in current codes and standards are assessed by
toe profile [19]. This notch stress is the total and maximum stress at the focusing on the following specific areas in Section 3: (a) how multi-axial
root of a notch. However, considering the variation of a weld profile and stress state is considered in recommended interaction equation; (b) how
transition radius, only the worst case with an ideal sharp notch is non-proportional loading effect is captured; (c) how as-welded and re
considered and it is further represented by an effective notch with a sidual stress relieved conditions are treated. With the above aspects
specified notch radius based on the concept of micro-support theory clarified, the three local stress approaches are then applied for analyzing
[19]. For structural steels and aluminum alloys, an effective radius of a comprehensive collection of multi-axial fatigue test data of welded
1.0 mm is suggested. For engineering applications, modeling such a components of various structural forms and multi-axial loading condi
small notch radius will require very refined meshes, making it almost tions in Section 4. A special attention is given to the relative effective
impractical for complex welded structures. Furthermore, the cross sec ness of each method for correlating the test data. Finally, the most
tion of weld toe and weld root profile along a single weld line varies effective method identified is further investigated in Section 5 and
significantly [20] even for a well-controlled welding procedure. As a Section 6 for elucidating the effects of some special cases on fatigue
result, the actual local stress can deviate significantly from the one damage, e.g., pure shear loading, in-phase and out-of-phase loading
calculated using a fixed fictitious notch radius. It is also worth noting between normal and shear stresses as well as residual stress relief.
that in addition to the stress state at notch tip, the opening stress dis
tribution along fatigue crack growth path must also be considered when 2. Overview of local stress approaches to fatigue analysis of
dealing with welded joints, according to fracture mechanics. welded joints
To capture both stress concentration and its effects over an antici
pated crack path, a traction-based structural stress definition was The fatigue design of weld structures based on nominal stress and
introduced by imposing equilibrium conditions [21,22] through the use joint classification assumes the welded joint to be examined has a good
of nodal forces/moments available from FE results. In addition to its match with those available in weld joint classification [15–17] in terms
demonstrated mesh-insensitivity, the method has led to the develop of joint type, loading conditions as well as other essential parameters.
ment of the Master S-N curve adopted by ASME Div 2 [23] since 2007. The nominal stress approach is increasingly regarded as a disadvantage
One advantage of this method over the other methods discussed above is since such a search for a match requires both experience and expertise
that the shear stresses can be calculated in the same manner, enabling an considering the vast amount of joint types. Furthermore, welded struc
effective structural stress parameter for assessing multi-axial fatigue tures are usually so complicated that the definition of a nominal stress
damage regardless of out-of-phase angle under synchronous sinusoidal itself is challenging or impossible. This issue also arise in FEA based
loading conditions [24]. Recent developments further extend its design and assessment nowadays. Most importantly, the fatigue damage
2
of a welded joint is recognized as a localized process [34] and it can’t be has been successfully used for tubular joints for many years [37,39]. It is
well-described by a global or nominal stress. As a result, several local based reference nodes defined in terms of tube characteristic length
√̅̅̅̅
approaches, including fracture mechanics based analysis, emerge and parameter rt in which r and t stand for tube radius and wall thickness,
gradually gain popularity as an indispensable part of methodology for respectively. In thin shell theory, the parameter serves as a measurement
welded joint design. These approaches typically require processing of of tube or pipe deformation wave length, while in plate weldments, plate
various types of stresses defined locally at the location of interest in thickness t cannot be argued to have a similar physical meaning.
contrast to the nominal stress. As a result, they are generally termed as Consequently, the extrapolated hot spot stress is better defined in
local stress approaches here. Nevertheless, some of these stress defini tubular joints than those used in plate joints.
tions do capture the overall or structural level of stress distribution to Therefore, the “extrapolation region” in tubular joints, including
different extent. A brief overview of the fundamentals of the three hollow section joints is well established [37]. As along as a consistent
popular local stress approaches i.e., extrapolation-based hot spot stress, extrapolation strategy is strictly followed, the hot spot stress is then
effective notch stress and traction-based structural stress that are supposed to serve as an indicator for local stress state at the weld toe.
adopted by current codes, standards and recommendations [16–19] are However, both the linear and quadratic extrapolation procedures
presented. permitted by Codes suggest that there is no general agreement on how to
separate the local effect caused by local weld profile and those influ
2.1. Surface-extrapolated hot spot stress approach enced by overall joint geometry or structural discontinuity. As a
demonstration, three types of extrapolated hot spot stresses [36] are
The hot spot stress approach is also known as structural hot spot shown in Fig. 1b, calculated from different extrapolation positions. With
stress (σhs ) or geometric stress approach. It is intended to take into ac FEA being widely used today, the hot spot stress approach has been
count of the stress-raising effects caused by an overall structural detail or adopted by major Codes and Standards for applications in plate and shell
the macro geometric stress concentration effects (Fig. 1a) of a welded structures [17,18]. In order to avoid large scatter of extrapolated
joint configuration. The very local concentration effect caused by weld stresses, there is a need for stipulating both FE modeling and specific
toe geometry is excluded [18,35,36]. It also assumes that the variation extrapolation details [34]. Current codes and standards [17,18] provide
of a local weld toe profile and related secondary stress concentration detailed and somewhat complicated instructions for performing ex
effects are included in the scatter band of hot spot stress S-N design curve trapolations, depending on the types of hot spot, the element size,
(s). The key assumption for determining hot spot stress is that localized element type, etc. Especially, different reference points are suggested for
notch stress, which is often singular in nature, can be separated from the both linear and nonlinear extrapolations.
hot spot stress which can be related to the overall joint geometry. As a final comment on the hot spot stress approach, the method can
The implementation of such a concept in practice can be challenging. only be applied to welded joint failure originating from a weld toe
For the widely used hot spot stress methods considered in this study, i.e., location. It is mainly used for cases when dominant loading is applied
those adopted in Eurocode 3 and IIW [16,17], the mechanics basis for normal to the weld line, even though a few studies [31], have been re
achieving the separation between local notch and hot spot stresses has ported on its application for pure shear or torsional dominated loading
not been explicitly offered. Instead, only an intuitive argument was conditions. As to be demonstrated in a later section, shear dominated
given by stating that the notch stress effect introduced by weld toe ge loading in plate weldments tend to cause rather localized stress con
ometry is limited to a localized region near the weld toe [35]. As a result, centration within a distance far less than 0.4 t or 0.5 t from a weld toe
the nodes of linear extrapolation should be positioned at some distance, location, making extrapolations more difficult to apply.
e.g., 0.4 t and 1.0 t in terms of plate/tube thickness t, away from a weld
toe position of interest in order to avoid the influence of local weld toe
2.2. Effective notch stress approach
geometry. Similarly, in ISO Design Standard [37], it is argued that
geometric stress increases linearly or quadratically as the weld toe is
Notch stress is the stress calculated at the root of a notch typically
approached and it is dependent on the stiffness of the plate or tube to be
with linear elastic material behavior assumed. Notch stress and notch
joined [36].
fatigue are classical fatigue problems in non-welded components and
Such a hot spot stress definition first originates from strain gage
numerous models such as Neuber’s equation [40], Peterson’s equation
measurements at certain distances away from a weld toe location. It was
[41] were proposed to deal with high cycle or fatigue strength of ma
first used as a reference strain for fatigue assessment of welded joint
terials. It has only recently been used for the analysis of welded joint.
[38]. The hot spot stress method based on surface strain measurements
When it comes to the welded joint, the notch stress is the highest
elastic stress calculated, among all different local stress approaches, at
the weld toe or root position. It includes the local stress (Fig. 1a) caused
Fig. 1a. Illustration of notch stress (σnotch ) and hot spot stress (σhs ) at weld toe
location as toe radius of (ρ). Fig. 1b. three different definitions of hot spot stress examined by Maddox [36]
3
by weldment profile itself. This method had originally been used for 2.3. Traction structural stress approach
high cycle fatigue and it was later on extended for medium and low cycle
fatigue by IIW Recommendations [17]. An effective radius (Fig. 2) is 2.3.1. Traction-based structural stress calculation from nodal force output
typically assumed at either weld toe or weld root location in order to of FEA model
simplify and avoid the variation of the weld transition radius profile. A traction-based structural stress method developed by Prof. Dong
The basic logic of an effective notch radius can be connected to the [21] has been shown quite effective in correlating a huge amount of test
micro-support concept proposed by Neuber [40] for high cycle or fatigue data obtained from different joint geometries, plate thicknesses, and
limit evaluation of non-welded notches. It is typically found that the loading conditions. This local stress approach is built upon a solid un
surface stress at the root of a notch under-predicts fatigue life, derstanding of the fatigue damage process of welded joints as a fatigue
depending on the materials examined. Neuber argued that it is more crack growth dominated issue. Fracture mechanics is employed to
appropriate to use the average stress calculated over a length scale formulate such a local stress framework and consequently, a unified
named as micro-structural length. Considering the worst case of a weld Master S-N curve is found to be sufficient to correlate test data from
toe with a sharp corner, i.e., an actual radius of zero [42], this would different types of weld joints. The traction-based structural stress
indicate that the effective notch radius is directly related with the micro- approach is emerging as a very popular and dominant local stress
structural length. Bearing in mind that such an extension of the concept approach and is currently adopted by several codes and standards
to welded joint analysis implies that this effective notch stress is more [23,44].
appropriate for the high cycle fatigue or crack initiation dominated The structural stress is calculated from nodal forces/moments of FEA
analysis of weld joint. In the specific application of the approach, an output with respect to a hypothetical cut plane and is found to be mesh-
effective radius of r = 1.0mm is typically suggested for both steels and insensitive [22]. The method has also been shown capable of extracting
aluminum alloys for plate thickness larger than 5.0 mm according to the two shear stress components on the same cut plane, which can be used to
IIW [17]. As plate thickness is getting thinner, a much smaller radius of describe multi-axial fatigue behavior under both proportional and non-
0.05 mm is advised [19]. proportional loading conditions [24,28]. Such a traction-based struc
In reality, the local profile of weld toe varies significantly even for tural stress relevant to weld toe cracking in a tube-flange joint is illus
different sections of one weldline and the weld toe profile can hardly be trated in Fig. 4(a), in which the component is subjected to both remote
modeled in an accurate manner. Fig. 3 shows typical measurements of cyclic bending (My ) and torsion (Mz ). The hypothetical cut into tube
radius at different weld toe locations of a welded cruciform joint rep thickness simulating weld toe cracking (Fig. 4(b)) exposes three traction
resenting current level of manufacture practice [20]. It shows that local stress components that are responsible for three different modes of fa
weld toe radius can be quite random and can hardly be fully considered tigue crack growth behavior in the context of fracture mechanics. These
in real analysis practice. As a result, it is a significant extrapolation components are referred to as normal, in-plane shear, and transverse
between a physical weld toe profile and the one with a defined effective shear stress components i.e., σ s (t), τs (t) and τz (t) defined with respect to
notch radius, especially considering that the approach is aimed at taking the cut plane (x’ − z’ plane) along weld toe, as illustrated in Fig. 4(b). As
into account of the notch effect caused by weld toe profile. described in detail in [22], these traction stress components with respect
To conclude, the advantage of notch stress approach is that it allows to the local coordinate system(x’ − y’ − z’) can be directly related to line
for the local weld geometry to be included so the effects of different force and line moments obtained from nodal forces and nodal moments
geometric configurations of weldments can be evaluated for detailed at each nodal position in shell or plate finite element model by:
design purpose [19,43]. The implementation of such an approach can be
fy’ 6mx’
laborious since very fine meshes are needed for notches with 1.0 mm σs = σm + σb = − (1)
radius, making it less attractive for analyzing complex structures. On the t t2
other side, since it is a much localized stress parameter and it is origi fx’ 6my’
nally proposed for fatigue initiation or micro-structurally small crack τs = τm + τb = + 2 (2)
t t
growth dominated fatigue life prediction, the extension of its application
to low cycle or medium cycle fatigue requires further comprehensive fz’
τz = (3)
studies. t
In Eqs. (1) through(3), fx’ , fy’ , fz’ represents line forces andmx’ , my’
represents line moments with respect to local coordinate system
x’ − y’ − z’ , respectively, as shown in Fig. 4(b). These lines forces and line
moments can be solved by a system of simultaneous equations using
nodal forces and line moments obtained from finite element analysis
[22,45].
1.0 mm For most applications of welded joints, such as the cases to be eval
radius uated in this paper, transverse shear component (τz ) is usually negligible
and only normal and in-plane shear structural stress components are
dominant when dealing with multi-axial fatigue of welded joints. These
two traction-based structural stress components will be used to formu
late an equivalent structural stress as elaborated in the next subsection.
2.3.2. Moment of load path (MLP) based equivalent structural stress for
multi-axial loading
2.3.2.1. Load path representation. For the multi-axial fatigue loading

caused by independent loading inputs such as asynchronous bending
and torsion, it is typically convenient to introduce a stress plane with the
horizontal axial representing the normal stress component and vertical
axis for shear stress component to characterize its load path. As a result,
the multi-axial loading histories can be fully captured. For example, if
Fig. 2. Sketch of FEA model for notch stress calculation [17]
4
Fig. 3. Measurements of toe radius along weld a toe line: (a) 3D surface of weld, (b) results showing small and large variations in the measurements of the weld toe
radius [20]
Fig. 4. Illustration of traction-based structural stress definition: (a) Tube-to-flange specimen subjected to both bending and torsion; (b) hypothetical free body cut
along weld toe and traction structural stress components exposed.
the nominal stress is used to characterize the load of a welded joint, the present load path in terms of local stress terms. As far as the traction-
load paths for cases of pure tension, pure torsion, in-phase loading be based structural stress in concerned, any given load path depicted on
√̅̅̅
tween tension and torsion, 90◦ out-of-phase loading are illustrated in a nominal stress based plane (i.e. σn − βτn plane) can be mapped onto a
√̅̅̅
Fig. 5(a). traction structural stress based plane (i.e., σ s − βτs plane). The coeffi
√̅̅̅
As mentioned in the introduction, the fatigue damage is a local cient β can be interpreted as a fatigue equivalency parameter between
stress/strain dominated process. It is therefore more appropriate to S-N data obtained under pure bending and pure torsion, typically taking
Fig. 5. Comparison of load paths between nominal and traction structural stress plane based representations: (a) nominal stress plane; (b) structural stress plane.
5
√̅̅̅
on a value of 3. In this process, the abscissa and ordinate representing range is found, projection to the next point needs to be made (R*
√̅̅̅
the nominal normal stress and shear stress on σn − βτn plane is point in Fig. 6(a)) to secure the stress range increase is always
stretched by normal and shear structural stress concentration factors, monotonic. The monotonically increasing load path identified in
respectively, to form a local structural stress load path. As an illustra this case is thenORR* T shown in Fig. 6(b) marked in red. This
tion, Fig. 5(b) shows the traction-based structural stress load paths after segment of load path information as a half cycle is then recorded.
√̅̅̅
mapping the path from Fig. 5(a). The load path on the σs − βτs plane is (iii) Repeat the steps in (i)-(ii) for the remaining load path. In this case
subject to a shape and size change since stress concentrations for normal the remaining load path is RSR* . Further maximum range
stress component is typically much higher than that of shear stress. In searching will then break it into RS and SR* segments, each with
the following, it is based upon the load path on the structural stress plane different stress ranges.
that an equivalent structural stress parameter is defined and validated
for multi-axial fatigue of welded joints. With the above multi-axial cycle counting procedure implemented,
the load paths information and counted cycles are available. The next
2.3.2.2. Path dependent maximum range (PDMR) multi-axial cycle step is to come up with a multi-axial fatigue damage parameter. In this
counting procedure. Multi-axial fatigue assessment of welded joints respect, path length as a fatigue parameter [25–27] was found to be
invariably involves the issue of cycle counting to break loading histories effective in collapsing a good amount of test data. In the following, as a
of different channels into simple segments for damage accumulation. further development, a structural stress range based parameter together
While this issue is a much simpler one for uniaxial loading for which with a non-proportionality parameter is used to formulate a fatigue
rainflow cycle counting is the dominant procedure, it is much more damage parameter.
challenging when multiple channels of loading histories are considered.
Also, different cycle counting procedures may lead to significant dis 2.3.2.3. Load path non-proportionality and moment of load path (MLP)
crepancies in predicted results [28]. Among the several multi-axial fa based equivalent structural stress. As presented in detail in recent papers
tigue related cycle counting procedures [46], a path dependent [28,30] for a non-proportional load path from A to B, i.e., AB, ̃ on
maximum range cycle counting procedure proposed (see more detailed
discussions by Dong et. al [26] and Wei and Dong [26,27] in the context
of PDMR cycle counting method) defined on the structural stress plane is
found to be attractive. The procedure essentially aims to capture
maximum possible change, i.e. the maximum range of stress or strain out
of available loading paths in a successive manner such that each segment
is counted exactly once. Additionally, this maximum range searching
algorithm is a further extension and generalization of the conventional
rainflow cycle counting method for which maximum range is also the
most important parameter to be identified.
The PDMR for breaking down a load path into simple segments is an
integral part of analyzing multi-axial fatigue of welded joints by traction
structural stress approach. Therefore, a simple demonstration of its
application to break a non-proportional multi-axial load path is shown
in Fig. 6(a). The procedure for the maximum range search is as follows:
(i) Search for the maximum range between any two points of the
whole loading path (ORSR* T) in stress plane. For this simple case,
it is clear that the starting point O and ending point T lead to the
maximum range. Note that starting point and ending points don’t
have to be at the start and the end of a whole path in most cases.
(ii) For the maximum range identified, we can take the starting point
O as reference and monotonically increase its stress range with
respect to O, until the maximum stress range is reached. If a Fig. 7. Non-proportional load pathAB,
̃ non-proportional circular load path AB
̂
turning point (R point in this case), leading to a decrease of stress and reference load path.AB
Fig. 6. PDMR cycle counting for loading path (a) a simple schematic load pathORSR* T on structural stress plane; (b) the identified load path segmentORR* T with
maximum stress range and the remaining segment RSR* to be further counted by PDMR.
6
√̅̅̅
σ s − βτs plane illustrated in Fig. 7, it constitutes one half cycle load path parameter takes a similar form as:
according to PDMR cycle counting procedure. A multi-axial fatigue
Δσ e
damage parameter Dfor any given non-proportional load path AB ̃ is ΔSe = 2− m 1 (9)
t* 2m I(re )m
assumed to be consisted of two parts:
where Δσe is effective stress range as shown in Fig. 7 and re is defined
D = DP + DNP (4)
as:
in which DP represents damage caused by the reference loading event ⃒
⃒Δσb,e ⃒
⃒
from A to B (i.e., AB), which can be directly related to the distance from re = ⃒⃒ ⃒ ⃒ ⃒ (10)
Δσm,e ⃒ + ⃒Δσb,e ⃒
A to B, or the effective stress range Δσ e . DNP represents non-
proportionality caused fatigue damage due to any excursion of load in which
path AB
̃ deviating from the reference load path (AB). Therefore, one √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
possible and intuitive way of representing a load path non- Δσb,e = Δσ2b + βΔτ2b (11)
proportionality in terms of a dimensionless form can be expressed as:
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∫ ∫
Δσm,e = Δσ2m + βΔτ2m (12)
DNP ̃ r’|sin(θ) |ds’ ̃ r’|sin(θ) |ds’
gNP = = ∫ AB = AB (5)
DMax ̂ R|sin(θ) |ds’ 2R2 Note that m in Eq. (9) is assumed to be the same as the one developed
AB
for Mode I crack growth crack growth, i.e.,m = 3.6 for the time being. As
where DMax represents the non-proportional fatigue damage caused
will be demonstrated later, Eq.(9)-Eq.(12) typically yield satisfactory
by the semi-circular load path represented by dashed lines in Fig. 7. gNP
prediction.
in Eq. (5) can be referred as a normalized load-path non-proportionality
With load path dependency considered by Eq. (6), the final form of
damage factor with respect to the maximum possible non-proportional
MLP based equivalent stress parameter that takes into account of load
damage DMax . With a dimensionless form of load path non-
path non-proportionality, thickness and bending ratio effect for non-
proportionality, a stress range parameter taking into account of non-
proportional multi-axial fatigue loading is:
proportional load path induced damage can be written as:
ΔσNP
ΔσNP = Δσe (1 + α∙gNP ) (6) ΔSNP = 2− m 1 (13)
t* 2m I(re )m
where Δσe is shown in Fig. 7 and it is the effective stress range of
reference path or the maximum range of the segment of the load path In the following section, the effectiveness of Eq. (13) in correlating
based on PDMR. Note that a material sensitivity parameter α is inserted well-documented multi-axial fatigue test data will be examined and
in Eq. (6) to accommodate the fact that some materials are more sen compared with both hot spot stress approach and notch stress approach.
sitive to non-proportional multi-axial fatigue loading than others, as
observed in [47]. For various structural steels investigated [30], it is 3. Assessment of procedures for multi-axial stress state, non-
generally found that α ≈ 1 gives a reasonable correlation with very few proportional loading, residual stress in codes and standards
exceptions based on the study of available historical multi-axial fatigue
test data of welded joints. This aspect will be elaborated in detail in data A quick summary of recommended procedures for the multi-axial
correlation section. For aluminum alloys ranging from 2000 to 7000 fatigue assessment of welded joint from Eurocode 3 and IIW recom
series, they are observed to be less sensitive to non-proportional loading mendations are presented here. Since no clear distinction is made
than structural steel and their material sensitivity parameters α vary regarding the type of stress definitions needed for the application of
from 0.35 to 0.5 depending on the ductility of aluminum alloys [29]. The these recommendations, these procedures are implemented in terms of
specific procedures for material sensitivity determination can be found both hot spot and notch stress approaches. On the other side, the traction
in [29]. structural stress approach is applied by following the integrated pro
cedures outlined in Section 2.3 with the final form of equation shown in
2.3.2.4. Thickness and bending ratio effects. Based on a two-stage crack Eq. (13).
growth model [48], an equivalent stress parameter (ΔSσ ) taking into
account of both plate thicknesses and bending ratio effect is given by
3.1. Multi-axial stress state and non-proportional multi-axial loading
Dong et al. [22] under normal traction stress dominated loading
conditions:
The treatment of multi-axial fatigue of welded joint is still in its early
Δσs development and limited information is available in most codes and
ΔSσ = (7)
standards, especially in Eurocode 3. Specifically, when multi-axial stress
2− m 1
t* 2m I(rσ )m
state in terms of normal tensile stress and shear stress need to be
In Eq. (7), Δσ s is traction structural stress range defined in Eq. (1); t* considered, the following interaction equation in Eurocode 3 should be
is a relative thickness with respect to a reference thickness (tref ), i.e., t * = implemented [16]:
t/tref , which enables the transferability of the resulting S-N curve for ( )3 ( )5
using a different reference thickness value such as those used in BS 7608 Δσ
+
Δτ
≤ DEC (14)
[15] etc. For the present discussion, tref = 1mm used here is consistent Δσ f Δτ f
with the 2007 ASME Code [23]. Also note that m = 3.6 was obtained by where Δσ and Δτ are normal and shear stress ranges calculated based
correlating both short and long crack growth data [48] under Mode I on their respective nominal or local stress definitions; Δσf is referred to
loading conditions. The integral I(rσ ) is a dimensionless function of as fatigue strength under pure axial loading and Δτf is the fatigue
bending ratio rσ , defined as:
strength under pure shear loading. These fatigue strength parameters
|Δσb | are defined by the specific types of stresses used for analysis. The term
rσ = (8)
|Δσm | + |Δσ b | DEC refers to the total accumulated damage and it is equal to unity for
both proportional and non-proportional loading conditions. Note that
The specific form of I(rσ ) can be found in [22], developed under the left side of Eq. (14) represents a simple linear summation of the
dominantly uniaxial fatigue loading conditions. For treatment of multi- damage caused by each of the two stress components carrying a negative
axial fatigue, it is postulated that an equivalent structural stress inverse slope of 3 and 5 in their S-N curves on a log–log scale,
7
respectively. In other words, no interaction is assumed between normal stress ratio (R). The increase of fatigue resistance is essentially equiva
stress and shear stress in Eq. (14) based on current Euorcode 3. In the lent to the scale of stress range by a factor of 1/f(R). Three different
process of the Eurocode 3 implementation, Eq. (14) can be rewritten as cases are distinguished: (a) Unwelded base material, residual stress
an effective stress range for facilitating data presentation: relieved welded components without constraints in assembly; (b) Small-
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ scale thin-walled structures containing small welds without assembly
ΔσEC = (Δσ)3 + kEC (Δτ)5 (15) constraints; (c) Complex two or three dimensional welded, thick-walled
3
components with global residual stresses. The specific definition of f(R)

/
kEC = Δσf 3 Δτf 5 for each case is defined as a piece-wise function and can be found in
[17]. Compared with Eurocode 3, the IIW Recommendations factors in
The IIW Recommendations [17] is more specific for combining the effect of assembly constraint and is more specific in the correction
normal and shear stress when multi-axial stress state needs to be introduced by residual stress relief. As for the multi-axial stress state, the
considered. The several cases to be distinguished are: correction of shear stress component is also beyond the scope of current
recommendation. Therefore, same correction is also done for shear
(a) If the shear stress range is less than 15% of normal stress, the stress part in Eq. (17) when residual stress relieved test data are evalu
effect of shear stress may be neglected. ated by IIW Recommendations.
(b) If the normal and shear stress are in phase or the plane of
maximum principal stress does not change significantly (less 4. Multi-axial fatigue test data analysis
than20◦ ), the maximum principal stress range should be used.
Also, the maximum principal stress shall be used, provided its 4.1. Data collection and assessment
direction is within ± 60◦ from the normal to the weld for a
constant amplitude loading. Otherwise, interaction equation is It is crucial that the three major local stress approaches to multi-axial
recommended. fatigue assessment reviewed in the last section are to be validated with a
(c) For the cases of both proportional and non-proportional loading, large amount of test data from different joint geometries, different multi-
regardless of constant amplitude or variable amplitude loading, axial loading conditions, and multiple sources. In doing so, some
the following interaction equation is recommended: essential requirements must be imposed to ensure that test data
(
Δσ
)2 ( )2
Δτ considered in such a validation effort were documented with sufficient
+ ≤ DIIW (16) details. This includes the parameters involved in calculating local
Δσ f Δτ f
stresses, consistent failure modes and failure criteria for comparison
Note that the total damage DIIW varies between proportional loading purposes. Specifically, the requirements and considerations for
and non-proportional loading in order to consider non-proportional including a particular set of existing test data are as follows:
damage to some extent. For the proportional loading, the total dam
age is still assumed to be DIIW = 1 whereas DIIW = 0.5 is recommended (a) Specimen geometry and testing conditions: To facilitate the
for non-proportional loading. Interestingly, the interaction equation Eq. computation of local stresses and their corresponding SCFs (see
(16) cannot be interpreted as damage accumulation from each stress Sec. 4.2), detailed specimen geometry and dimensions must be
component anymore since the exponent of both terms are not equal to available for creating a sufficiently detailed finite element model,
the slopes of their respective fatigue resistance curves. Similar to Eq.(15) including specimen mounting and loading conditions
, an equivalent form of Eq. (16) can be written as: (b) Weld toe failure mode: Only test data that exhibit weld toe failure
1
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ are considered in this study since weld root or throat failures
ΔσIIW = √̅̅̅̅̅̅̅̅̅ (Δσ)2 + kIIW (Δτ)2 (17) would require more detailed documented information on weld
DIIW
size and penetration in each specimen, as discussed recently by
As a final note on the above interaction equations in current pro Xing et al. [49].
cedures for multi-axial fatigue of welded joints, no specific recommen (c) As-welded versus stress-relieved conditions: Test data to be
dation is provided to deal with multi-axial cycle counting. It is therefore considered in this paper should have a clear indication on if
ambiguous or challenging to process non-proportional loading, espe welded components had gone through post-weld heat treatment
cially for the case of asynchronous loading between different load in (PWHT) based stress-relief or not prior to testing. A cursory re
puts. In the current evaluation of hot spot stress and effective notch view of existing multi-axial test data indicates that there exist
stress approaches by following these procedures, the cycles defined in noticeable differences in S-N data between as-welded and stress-
original documents are adopted. relieved conditions. This is also the main reason for the test data
to be grouped into as-welded and residual stress relieved joints in
3.2. Residual stress latter comparison and discussions.
(d) Through-thickness failure dominated failure criteria: The ma
It is typically assumed that the design S-N curve for each joint clas jority of test data for which tests were typically stopped as along
sification incorporates the effect of residual stress implicitly. Further as through-thickness failure was observed. This is consistent with
modification is needed only if reliable information is available to sup the formulation of the equivalent structural stress parameter in
port the beneficial effect of residual stress relief. A very simple way of Eq. (7) and Eq. (13), recalling through-thickness hypothetical cut
considering the stress-relieved welded joint is adopted in Eurocode 3. in Fig. 4. Some exceptions include the test data from R. Archer
That is, when part of or all of the cyclic stress is compressive, the [14] for which tests were stopped when a 20 mm crack was
effective stress range can be calculated by adding the tensile stress and observed. Its abnormal behavior may be related to this failure
60% of the compressive portion of the stress. When it comes to multi- criterion and will be looked into in detail.
axial stress cases, no further details are available for the shear stress
component. As a result, the same correction as normal stress is adopted After a comprehensive literature survey and detailed critical assess
for shear stress component before they are combined into the effective ments, various component types, loading conditions, test sources [4–14]
stress definition in Eq. (15) following the Eurocode 3. As for the as- to be considered for evaluation of different local stress approaches in
welded joint, no correction is made. this study are illustrated in Fig. 8 under as-welded and Fig. 9 under
IIW Recommendations introduces a fatigue enhancement factor residual stress relieved conditions, respectively. Additional test details
(f(R)) for fatigue strength of classified weld joint details as a function of regarding multi-axial loading types, plate thickness, loading ratios, etc.
8
Fig. 8. As-welded multi-axial fatigue test specimens and loading conditions.
are given in Table 1. captured. Traction structural stress method clearly has the advantage of
being applied to large scale structures for which a large amount of weld
joints need to be evaluated.
4.2. Calculation of local stresses for different welded joints For a consistent processing of the different types of local stresses, all
the stresses are presented in terms of normal and shear stress concen
A good amount of effort is made to create models for extracting the tration factors as shown in Table 2 since nominal stress can be easily
three different types of local stresses. Based on the available information calculated for all joints studied. For example, the normal structural and
for the tests that fit the criteria given in the previous section, all the shear structural stress concentration factors are calculated as follows:
essential details such as boundary constraints, geometrical parameters, σs
etc. that might affect the local stress distribution are captured to our best Kσ s = (18)
σn
interpretation. Fortunately, systematic researches [31,32] into a good
amount of these test data in Table 1 by surface-extrapolated hot spot τs
Kτ s = (19)
stress approach and effective notch stress are already available. These τn
local stresses in terms of local stress concentration factors (SCF), i.e. hot
where normal structural stress and shear structural stress are calcu
spot stress SCFs and notch stress SCFs are adopted [31,32] in our com
lated from Eq. (1) and Eq.(2) and are composed of membrane and
parison. Otherwise, FEA models are created by following the guideline
bending parts.
of IIW Recommendations [17]. As far as structural stress approach is
With local stress components calculated, they are then combined into
concerned, all FEA models are built and structural stresses are calcu
an equivalent local stress according to the current codes and standards.
lated. Fig. 10 is a comparison of tube-flange FEA models used for
The equivalent stress equations in Eq. (15) and Eq. (17) are applied
extraction of hot spot stress, notch stress and structural stress. As
according to Eurocode 3 and IIW in terms of hot spot stress and notch
reviewed in the Section 2, hot spot stress model in Fig. 10(a) requires the
stress. As regard to residual stress relieved data, corrections according to
precise positioning of elemental nodes near weld toe, depending on the
each procedure as reviewed in Section 3.2 are implemented.
extrapolation technique as well as the tube and plate thickness involved.
The implementation of traction structural stress, more specifically,
Notch stress model in Fig. 10(b) has very fine meshes at weld toe and a
MLP based equivalent structural stress incorporates the integral process
small transition radius, i.e. 1.0 mm is needed to capture the local notch
of PDMR cycle counting (Section 2.3.2.2), load path non-proportionality
stress. Structural stress model in Fig. 10(c) has very coarse meshes since
characterization (Section 2.3.2.3) as long as non-proportional multi-
it is mesh-insensitive and it can be further simplified to shell-element
axial loading is involved. More detailed information and explanation
based model as long as the through-thickness stress distribution is
9
Fig. 9. Residual stress relieved multi-axial fatigue test specimens.
Table 1
A summary of multi-axial fatigue tests and test details examined in this paper.
Multi-axial fatigue tests Joint types Steels types Thickness(mm) Load combinations Normal stress Shear stress Residual stress state
R ratios R ratios
Sonsino and Kueppers [4] tube-to-flange StE460 10.0 Bending, torsion − 1 − 1 relieved
Sonsino and Łagoda [5] tube-to-tube StE460 6.0 Tension, torsion − 1 − 1 relieved
Siljander et al. [6] tube-to-plate ASTM A519 7.95 Bending, torsion − 1,0 − 1,0 relieved
Yousefi et al. [7] tube-to-flange P 460 8.0 Bending, torsion − 1 − 1 relieved
Witt and Zenner [8] tube-to-flange StE 460 M 8.0 Bending, torsion − 1 − 1 relieved
Razmjoo [9] tube-to-flange Grade 50E 3.2 Tension, torsion 0 0 as-welded
Yung and Lawrence [10] tube-to-plate ASTM A519 7.95 Bending, torsion − 1 − 1 as-welded
Takahashi et al. [11,12] fillet welded gusset JIS SM400B 6.0 Tension, tension 0 0 as-welded
Dahle et al. [13] butt welded box Domex 350 10.0 Bending, torsion − 1 0 as-welded
R. Archer [14] fillet welded gusset Grade 43C 6.0 Bending, torsion 0,-1 − 1 as-welded
Fig. 10. A comparison of models used in three local stress approaches: (a) surface-extrapolated hot spot stress approach FEA model; (b) notch stress FEA model [32]
and (d) traction-based structural stress FEA model.
can be found in [28]. Additionally, the thickness and bending ratio

correction are carried out as shown in the final form of equivalent
structural stress in Eq. (13).
10
Table 2
Structural stress based SCFs, hot spot stress SCFs and notch stress SCFs calculated for test components described in Fig. 8 and Fig. 9.
Multi-axial fatigue Structural stress SCF of Structural stress SCF of Hot spot SCF of Hot spot SCF of Notch stress SCF of Notch stress SCF of
tests normal stress (σm + σb )/σn shear stress (τm + τb )/τn normal stress shear stress normal stress shear stress
Membraneσm /σn Bending Membrane Bendingτb /τn

σb /σn τm /τn
Sonsino and 0.925 0.79 0.97 0.13 2.2 1.1 3.31 1.74
Kuepper [4]
Sonsino and 1.0 0.0 1.0 0.0 1.0 1.0 1.65 1.36
Łagoda [5]
Siljander et al. [6] 0.834 0.585 0.95 0.15 1.25 1.1 2.34 1.63
Yousefi et al. [7] 0.80 0.89 0.97 0.13 1.37 1.1 2.88 1.62
Witt and Zenner 0.80 0.89 0.97 0.13 1.37 1.1 2.88 1.62
[8]
Razmjoo [9] 0.94 0.96 0.96 0.14 1.4 1.1 4.06 1.46
Yung and 0.834 0.585 0.95 0.15 1.25 1.1 2.8 1.57
Lawrence [10]
Takahashi et. al 1.41 0.54 0.0 0.0 1.73 0.0 2.62 (Uniaxial) 0.0
[11,12] 3.15(bi-axial) 0.0
Dahle et al. [13] 1 0.0 1.0 0.0 1.0 1.0 1.79 1.55
R. Archer [14] 1.26 0.6 1.0 0.1 1.8 1.0 2.0 1.57
5. Comparison of three different local stress approaches by data involved. Among the various types of weld failures of box structure in
correlation Dahle et al. [13], only weld toe failure along transverse weld is
considered. The gusset joint in Archer’s research [14] involves the
The potential of the three local stress approaches in collapsing multi- asynchronous loading cases and the structure typically still function
axial fatigue test data will be examined in this section. To prevent the even if a small crack is observed at the weld toe. This is part of the reason
pollution of data, leading to a larger scatter band of test data, the as- a 20 mm crack was observed before tests were stopped.
welded and residual stress relieved test data are presented separately The Fig. 11 presents hot spot stress based plot following the Eurocode
in the following section. The possible beneficial effect of residual stress 3 and IIW Recommendations. First of all, it can be found that the test
relief in comparison to as-welded test results will be discussed in a data from Archer [14] is on the upper side of the mean S-N curve,
separate section. indicating that test stop criteria of the gusset weld joint on top of the
box-section structure is more of a criteria for the structure failure than
the joint itself. Also, the test data from Dahle et al. [13] is highly scat
5.1. Data correlation by three local stress approaches tered in nature. Although the Eurocode 3 in Fig. 11(a) yields slightly
better correlation than that of IIW based Fig. 11(b), it can be seen that
5.1.1. As-welded weld joint data correlation surface-extrapolation based hot spot stress does not offers satisfactory
Among the several types of welded joints in Fig. 8 in as-welded correlation regardless of the procedures implemented.
condition, tube-flange welded joint under combined tension and tor It is impressive to notice that the effective notch stress approach
sion is the most common joints examined by researchers. The gusset improved the data correlation, as shown in Fig. 12. The plot shows a
joints studied by Takahashi et al. [11,12] are bi-axially loaded joints and much better correlation with two procedures implemented when
it is essentially close to uniaxial loaded joints since shear stress is not
Fig. 11a. Hot spot stress based fatigue analysis for as-welded joints following EC3 procedure.
11
Fig. 11b. Hot spot stress based fatigue analysis for as-welded joints following IIW procedure.
compared with hot spot stress method. The main reason for such an Fig. 13 together with the Master S-N curve adopted by ASME Div 2 [23]
improved correlation is again related to the test data from Archer’s test and API 579 [44]. One can observe that almost all the multi-axial loaded
[14]. In such a plot, these test results are not significantly shifted away test data falls into the scatter band defined by two times of standard
from other tested joints in Fig. 12 (a) in contrast to those shown in deviation of the Master S-N curve. This correlation is encouraging and
Fig. 11 (a). Among the different factors, it is believed that the normal exciting in view of the fact the S-N curve was derived based on test data
stress SCFs calculated by notch stress approach play a significant role. As from welded joints under uniaxial loading. Overall, a much narrower
can be seen in the five rows from the bottom in Table 2 for as-welded scatter band is found for structural stress based correlation (Fig. 13) than
joints, there is a significant increase of normal stress SCF values when hot spot stress (Fig. 11) and notch stress plots (Fig. 12). More specially,
notch stress is used in comparison with hot spot stress, except for the the test results from Takahashi et al. [11,12] loaded by bi-axial tension
Archer’s welded joints. In this way, the equivalent notch stress in Fig. 12 (Fig. 8 (c)) fall right on the top of the mean Master S-N curve. This is
(a) calculated for Archer’s welded joints is thus smaller when compared consistent with our previous comment that the tests are essentially
with others. For the current data examined, the IIW based correlation in equivalent to uniaxial fatigue tests, as confirmed from the zero shear
Fig. 12(b) is not performing better than Eurocode 3 in Fig. 12(a), structural stress concentration factors in Table 2. Special attention
although it is a more detailed procedure, as reviewed in Section 3.1. should be paid to the test data distribution from Archer [14]. First of all,
Traction-based equivalent structural stress correlation is shown in the data is found to be on upper shelf among all and it further confirms
Fig. 12a. Notch stress based fatigue analysis for as-welded joints following EC3 procedure.
12
Fig. 12b. Notch stress based fatigue analysis for as-welded joints following IIW procedure.
Fig.13. MLP based equivalent structural stress based fatigue analysis for as-welded joints.
that the longer life is related to the test failure criteria with a 20 mm simple joint types and through-thickness failure criterion compared with
crack found on the complex structure with a large cross section (Fig. 8). some of the complex structures in as-welded joints in Fig. 8, these tests
Secondly, in contrast with what is shown in Fig. 11 and Fig. 12, the test are believed to be better candidates to further examine the effectiveness
data is not lumped together anymore, because a clear cycle counting of the three local stress approaches as well as the current codes and
process by PDMR is carried out for asynchronous loadings involved. standards.
Finally, the non-proportional loading damage is taken into account in a The hot spot stress based plot of residual stress relieved test data are
quantitative manner. More specific discussion on this special case shown in Fig. 14(a) and Fig. 14(b) for Eurocode 3 and IIW based multi-
regarding the material sensitivity can be found in Section 5.2. axial stress interaction respectively. It is observed that the IIW recom
mendation performs much better than Eurocode 3. The large amount of
5.1.1.1. Residual stress relieved welded joint data correlation. It is inter test data indicates the better correlation achieved by IIW procedure is
esting that there is much more test data available from residual stress more convincing than the as-welded data in Fig. 11. There is no denying
relieved welded joints than as-welded joints. Also, most of these test data that that the current level of scatter in Fig. 14(b) is still unsatisfactory as
is from tube-flange and tube-tube joints as shown in Fig. 9. In view of the indicated by the standard deviation and correlation coefficients.
13
Fig. 14a. Hot spot stress based fatigue analysis for stress-relieved welded joints following EC3 procedure.
Fig. 14b. Hot spot stress based fatigue analysis for stress-relieved welded joints following IIW procedure.
When the notch stress approach is adopted, the correlation in Fig. 15 relieved test data fall on the upper side of the Master S-N curve
(a) and Fig. 15(b) shows improvement over hot spot stress approach compared with the as-welded test data in Fig. 13. This shift of the large
(Fig. 14(a) and Fig. 14(b)). Furthermore, the IIW procedure out- amount of test data in relation to the Master S-N curve reveals that there
performs Eurocode 3 is further confirmed when notch stress is used is indeed a noticeable beneficial effect from residual stress relief treat
for stress relieved test data. Collectively, the notch stress and IIW pro ment. The data in Fig. 16 offers us a valuable reference on how much
cedure combined lead to the best collapse of data so far among the four beneficial effect will be introduced with respect to Master S-N curve for
alternatives when these two local stress approaches and two different which as-welded joints is applicable. As a final note on such an effective
interactions are combined. data collapse by the structural stress approach, the essential factors such
The traction structural stress plot in Fig. 16 is found to be performing as PDMR cycle counting, load path non-proportionality, bending ratio
much better than the other two local stress approaches regardless of and thickness correction are all integrated consistently, to achieve such
Eurocode 3 or IIW procedure they follow. The much smaller scatter band a correlation. More detailed examination on some of these aspects such
indicated by the standard deviation is in sharp contrast with those as non-proportional damage effect will be discussed in the next section.
shown Fig. 14 and Fig. 15. Noticeably, most of the residual stress
14
Fig. 15a. Notch stress based fatigue analysis for stress-relieved welded joints following EC3 procedure.
Fig. 15b. Notch stress based fatigue analysis for stress-relieved welded joints following IIW procedure.
5.2. Discussions the Master S-N curve was aimed for mode I type of weld fracture. Till
now, very limit data was analyzed [50] for welded joints by structural
With sufficient as-welded and residual stress relieved data to stress for pure torsional loading. Therefore, it is revealing to fill this gap
demonstrate that structural stress approach is a robust and reliable local by looking into test results under pure torsional loading.
stress approach among all, some of the test data will be explored in terms The as-welded and residual stress relieved test data under pure tor
of structural stress to have a deeper understanding of the behavior of sion is presented in Fig. 17(a) and Fig. 17(b), respectively. The Master S-
welded joints under different circumstances. This includes: (a) the fa N curve is also shown, characterizing the scatter band for a wide variety
tigue of welded joints subjected to pure torsion; (b) the non-proportional of as-weld joint under mode I tensile loading. It is a pity that very limited
loading effect by comparing simple in-phase and out-of-phase loading; amount of test data is available for as-welded joints and most of the data
(c) the effect of residual stress relief treatment. points are form Archer [14] tested on a complex structure. Still, most of
these test data fall into the Master S-N curve scatter band indicating that
5.2.0.1. Fatigue design curve for weld joints under pure shear loading the Master S-N curve is still applicable for pure torsional welded joint
It is not surprising that very limited test data is currently available for analysis, except for the high cycle fatigue life regime beyond 2 million
welded joints under pure torsional loading. This is especially the case cycles for which the Master S-N curve tends to be conservative. As
when structural stress is applied as a local stress approach considering reiterated before, the fatigue life in Archer’s [14] tests in Fig. 17(a) tends
15
Fig. 16. MLP based equivalent structural stress analysis for stress-relieved welded joints.
Fig. 17a. MLP based equivalent structural stress analysis for as-welded joints under pure torsional loading.
to be longer since a much longer crack size, instead of through-thickness the residual stress relief effect, as it is typically believed the heat treat
failure was used to stop the tests. ment improves fatigue resistance more significantly in high cycle than in
The stress relieved test data in Fig. 17(b) for pure torsion is much low cycle regime. This argument is kind of reinforced by the as-welded
more abundant than as-welded tests and all of the tests were done on data in Fig. 17(a) for which no clear deviation of slope from Master S-N
tube-flange joints. Two interesting observations can be made from the curve is found, admitting that only limited test data is currently avail
plot. To start with, the test results are mostly located beyond the upper able. On the other side, as currently adopted in most codes, the nominal
side the S-N curve with two times of standard deviation, in comparison shear stress S-N curve has a slope of 5, much shallower than that of
with Fig. 17(a), indicating the clear benefit from residual stress relief normal stress S-N curve. Considering the shear structural stress con
even for pure torsional loading. It is clear that the deviation is much centration factors (Table 2) are typically very small and are around 1.1
larger in the high cycle regime while the beneficial effect is almost non- for all cases, the slope in Fig. 17(b) might be partly caused by inherent
existent in the low cycle regime, probably caused by the plastic defor nature of the shear stress loading. Further investigations and tests on as-
mation in low cycle fatigue. Second of all, the slope of the pure torsional welded joint are needed for a better understanding of this detail.
test data seems shallower and it is deviating from the slope of Master S-N
curve for which m = 3.6 is used. On the one side, this might be caused by
16
Fig. 17b. MLP based equivalent structural stress for residual stress relieved welded joints under pure torsional loading.
5.2.0.2. The effect of in-phase and out-of-phase loadings on welded joint sufficient for most steels and this value is adopted for all cases except the
fatigue one from Archer’s test [14]. This special case will be explored here and it
The simplest multi-axial loading is combined sinusoidal tension and is believed to be related with the long crack failure criterion.
sinusoidal torsion with the same frequency as demonstrated in Fig. 5. Fig. 18(a) and Fig. 19 (a) show the in-phase and out-of-phase test
When there is not a phase difference, it is call in-phase loading. Other data without considering the non-proportional damage effect, i.e the
wise it is called out-of-phase loading and it is a special case of non- second term in Eq. (6). The empty symbols represent the out-of-phase
proportional loading. A large amount test data currently studied are test data and the solid ones are in-phase test results. It clearly shows
from these two types of loadings. These sets of test data are good can that the out-of-phase test life is much shorter than the in-phase. Sur
didates for looking into the non-proportional damage effect, which is prisingly, almost one order of magnitude lower fatigue life is observed
currently taken into account in a qualitative manner (Section 3.1) in IIW for out-of-phase tests in comparison with in-phase in tests led by Sonsino
Recommendations. [4,5]. The extent of deviation between in-phase and out-of-phase de
MLP based equivalent structural stress in Eq. (13) offers a quantita pends on both the stress amplitudes and phase angles involved.
tive treatment of non-proportional damage by Eq. (6). The simple form When the non-proportional damaging effect is considered by Eq. (6),
of equation in Eq. (6) captures the non-proportional damage effect by the non-proportional test results merge with in-phase data and a good
two parameters: the load path non-proprotionality (gNP ) and a material collaspe of data is achieved for both as-weld joints and stress relieved
senstivity parameter (α). The first parameter is explained in detail joints in Fig. 18(b) and Fig. 19(b), respectively. This quantatitive
(Section 2.3.2) in a phenomenological way and the determiantion of the treatment of non-proportional damage is one essential factor for
latter one usually involves in-phase and out-of-phase test results. Most of achieving the excellent correlation in multi-axial fatigue analysis by
the detailed information on these two parameters can be found in structural stress approach. As mentioned before, α = 1 is used for steels
[28–30]. Based on a previous study [28], a specific value of α = 1 is consistently for all cases except for Archer’s test [14] in the plot of
Fig. 18a. In-phase and out-of-phase test results for as-welded joints without non-proportional damage effect considered.
17
Fig. 18b. In-phase and out-of-phase test results for as-welded joints with non-proportional damage effect considered.
Fig. 19a. In-phase and out-of-phase test results for residual stress relieved joints without non-proportional damage effect considered.
Fig. 18(b). It is found that α = 0.3 is a best choice found for collasping all The as-weld data is mostly located within the Master S-N curves while
the test data within this specific group. One possible explanation is that the same curves are conservative for residual stress relieved joints,
the non-proportional damage effect might be different or less evident as indicating the beneficial effect of residual stress relief. As expected, the
long as a marco-crack is formed. In the Archer’s case, the tests were not beneficial effect of stress relief increases significantly as high cycle
stopped until cracks with about 20 mm crack size were found. Consid regime is approached, especially when the design life goes beyond a
ering the structural behavior of the test specimen in Fig. 8(e), it is likely million cycles. To put it in a different way, it seems that there is a slight
that a good amount of fatigue life is caused by macro-crack growth after difference in terms of the slope for stress relived test data since stress
a crack penetrates the plate. This extra macro-crack growth life is relief is less effective in low cycle regime.
confirmed by the fact that this group of test data is always shown on the
upper side of mean Master S-N curve in Fig. 13 and Fig. 18(b).Fig. 19a. 6. A unified fatigue design curve for as-welded and stress-
relieved joint
5.2.0.3. The effect of residual stress relief on welded joint fatigue
A noticeable difference will be observed when the as-welded test As noted in the above section, there exist noticeable differences be
data from Fig. 13 and stress relieved data from Fig. 16 is shown in the tween as-welded and stress relieved joints in terms of their fatigue
same chart in Fig. 20. By comparing with as-welded test data, stress- performance. One alternative of fatigue design can be done by adopting
relieved test data tends to show a noticeable upward shift in S-N plots. two separate S-N curves, one is Master S-N curve for as-welded joints
18
Fig. 19b. In-phase and out-of-phase test results for residual stress relieved joints with non-proportional damage effect considered.
Fig. 20. MLP based equivalent structural stress plot of as-welded and residual stress relived test data.
and the other is the S-N curve for residual stress relieved welded joints. parameter in Eq. (13) becomes:
Working is ongoing to collect more stress-relieved data to present such a
ΔσNP
curve with more confidence. Currently, it would be great if the data from R
ΔSNP = 2 2− m 1 (20)
both types of joints can be put into one S-N curve with some kind of (1 − R)m t* 2m I(re )m
correction for stress relieved joints. In order to achieve such a goal, Currently, no mean stress correction is needed for as-welded joints
further look into of the stress relieved tests, especially the test data from and the corresponding Master S-N curve has the mean stress effect
Siljander et al.[6] with different load ratios is done. Two important 2 1
included. The term (1 − R)m is valid for negative R, (1 − R)m for positive
observations are: (a) the normal stress load ratio (R in Table1) does have
R when dealing with stress-relieved test data. Noting that the stress ratio
influence on stress relieved welded joint with higher load ratio being
used here is only for normal structural stress ratio since shear stress ratio
more damaging; (b) the shear stress load ratio has no clear effect on
has little effect on fatigue life. Applying Eq. (20) to stress relieved test
fatigue damage. These critical findings indicate that normal stress load
data leads to the plot shown in Fig. 21, in which Master S-N curve scatter
ratio can be used to take into account of mean stress effect.
band (i.e., mean ± 2 STD) is also shown. The result suggests a rather
The 2007 ASME Code [23] provides an applied stress ratio (R)
reasonable comparison between all multi-axial test data analyzed in this
correction scheme for an added conservatism, which is derived based on
paper and those (over 1000 tests) under uniaxial loading conditions used
a two-stage crack growth model [48] under ideal stress-relieved condi
by 2007 ASME Code. Even though a somewhat poor correlation occurs
tions. As such, it can be shown that the equivalent structural stress
19
Fig. 21. Equivalent structural stress range based correlation of all test data including stress ratio correction for stress-relieved test data using Eq. (20).
at high cycle regime, say above 106, the use of the ASME design S-N depending on the joint stiffness involved. On the other side,
curve yields conservative life estimate. effective notch stress aims to capture the local stress as a whole
As a final comment, it should be pointed out that only set of the without distinguishing its primary or secondary components in
multi-axial test data from Yousefi et al.[7] not considered in Fig. 21 is leading to failure. It originates from classical high cycle fatigue of
the one obtained under load ratio of R = 0. This set of data is behaving notched non-welded components for which fatigue initiation is
inconsistently with all the rest of stress-relieved data for which a load assumed to play a dominant role. Aside from the complexity
ratio effect is typically observed and therefore mean stress correction is involved for fine mesh control near the effective notch radius, its
needed as demonstrated in Eq. (20). During the re-examination of the application to low or medium cycle regime requires further
data, we notice that one plot in a previous publication by the first two investigation.
authors [28] of this paper had an inadvertent calculation error, resulting (2) The traction structural stress definition extracted from a hypo
a reasonable correlation of the data between R = -1 and R = 0. The thetical cut plane into base material thickness at weld toe loca
authors would like to use this opportunity to alert readers about the tion represents the dominant controlling parameter involved in
detail in the previous work. the physical fatigue crack growth process as far as fracture me
chanics is concerned. Its calculation based on nodal forces and
7. Conclusions nodal moments from FEA output is shown be to mesh-insensitive
and as a result, coarse mesh is sufficient for implementation.
This paper establishes an evaluation and a comparison of the three Furthermore, its recent developments such as multi-axial cycle
widely used local stress approaches, i.e. the surface-extrapolated hot counting by PDMR and treatment of non-proportional damaging
stress, effective notch stress and traction structural stress when they are effect bring further potential for multi-axial fatigue analysis of
applied to multi-axial fatigue assessment of welded joints. A compre welded joints.
hensive critical review of these stress definitions are made regarding the (3) Both Eurocode 3 and IIW Recommendations provide limited in
fundamental logic behind these local stress approaches. Furthermore, formation on multi-axial fatigue of welded joints. They aim to
the details pertinent to multi-axial fatigue of welded joints from Euro address problem by introducing simple interaction equations,
code 3 and IIW Recommendations are also overviewed as they are going either based on damage accumulation or empirical equation.
to be implemented for hot spot stress and notch stress methods. Major Eurocode 3 currently does not take non-proportional loading ef
emphasis is put on the structural stress approach and its recent devel fect into account while IIW Recommendation provides a quali
opment for multi-axial fatigue analysis in view of the popularity it is tative assessment approach. Further improvements for dealing
gaining in recent years. Extensive amount of multi-axial fatigue test data with issues such as the cycle counting of asynchronous loading
is collected and analyzed by the three stress definitions in order to events are needed.
establish a convincing comparison among them. Some of the major (4) A large amount of multi-axial test data currently available is
findings are: investigated by three local stress approaches. It is found that for
both as-welded joints and residual stress relieved joints, Large
(1) Currently available local stress methods provide an essential scatter bands are found for both hot spot stress and notch stress
pathway to fatigue assessment of welded joints by acknowledging approaches, regardless the code or standard followed. It seems
that fatigue damage is a localized damage process. The surface- notch stress is performing slightly better when IIW Recommen
extrapolated hot spot stress approach attempts to separate the dations is followed. The best collapse of test data is achieved by
local stress contributed by the macro-geometric detail from the MLP based equivalent structural stress for which the integral
part introduced by local weld profile. Its actual implementation is processes of PDMR cycle counting, local path non-proportionality
challenging since this delineation depends on the macro- calculation, etc. are carried out consistently as long as complex
geometric-affected zone from where extrapolation takes place. non-proportional loadings (e.g. asynchronous loading) are
In reality, this extrapolation zone varies among different joints involved.
20
(5) It is interesting to observe that the currently available pure [19] Fricke W. IIW recommendations for the fatigue assessment of welded structures by
notch stress analysis. Woodhead Publ 2012.
torsional test data also falls into the Master S-N curve for as-
[20] Barsoum Z, Jonsson B. Influence of weld quality on the fatigue strength in seam
welded joints. On the other side, a slightly different slope is welds. Eng Fail Anal 2011;18(3):971–9.
found for pure torsional fatigue loading of residual stress relieved [21] Dong P. A structural stress definition and numerical implementation for fatigue
joints. The in-phase and out-of-phase test results indicate that analysis of welded joints. Int J Fatigue 2001;23:865–76.
[22] Dong P, Hong JK, Osage DA, Dewees DJ, Prager M. “The Master S-N Curve Method:
non-proportional loading reduces fatigue life significantly and it An Implementation in 2007 ASME Div 2 International Codes for Boilers and
should be considered quantitatively. The equivalent structural Pressure Vessels,” Welding Research Council Bulletin, No. 523, 2010.
stress approach addresses this issue by incorporating one [23] 2007 ASME Boiler and Pressure Vessel Code, Sec VIII,Div. 2 (ASME BPVC-VIII-2-
2007), American Society of Mechanical Engineers, New York, 2007.
parameter to characterize load path non-proportionality and the [24] Dong P, Hong J. A Robust Structural Stress Parameter for Evaluation of Multiaxial
other for material sensitivity. Finally, residual stress relief brings Fatigue of Weldments. J ASTM Int 2006;3(7):206–22.
a clear beneficial effect in improving fatigue life of welded joint [25] Dong P, Wei Z, Hong JK. A path-dependent cycle counting method for variable-
amplitude multi-axial loading. Int J Fatigue 2010;32:720–34.
under multi-axial loading and it can be significant in high cycle [26] Wei Z, Dong P. A generalized cycle counting criterion for arbitrary multi-axial
regime. Both as-welded and stress-relived joint data can be fatigue loading conditions. J Strain Anal Eng Des 2014;49:325–41.
collapsed into the currently available Master S-N curve scatter [27] Wei Z, Dong P. Multiaxial fatigue life assessment of welded structures. Eng Fract
Mech 2010;77:3011–21.
band, as long as mean stress correction is made for stress-relieved [28] Mei J, Dong P. An equivalent stress parameter for multi-axial fatigue evaluation of
joints. welded components including non-proportional loading effects. Int J Fatigue 2017;
101:297–311.
[29] Mei J, Dong P. Modeling of path-dependent multi-axial fatigue damage in
Declaration of Competing Interest
aluminum alloys. Int J Fatigue 2017;95:252–63.
[30] Mei J, Dong P. A new path-dependent fatigue damage model for non-proportional
The authors declare that they have no known competing financial multi-axial loading. Int J Fatigue 2016;90:210–21.
interests or personal relationships that could have appeared to influence [31] Bäckström M, Marquis G. A review of multiaxial fatigue of weldments:
Experimental results, design code and critical plane approaches. Fatigue Fract Eng
the work reported in this paper. Mater Struct 2001;24:279–91.
[32] Pedersen MM. Multiaxial fatigue assessment of welded joints using the notch stress
References approach. Int J Fatigue 2016;83:269–79.
[33] Lazzarin P, Lassen T, Livieri P. A notch stress intensity approach applied to fatigue
life predictions of welded joints with different local toe geometry. Fatigue Fract
[1] Socie DF, Marquis G. Multiaxial fatigue. Warrendale, PA: Society of Automotive Eng Mater Struct 2003;26(1):49–58.
Engineers; 1999. [34] Radaj D, Sonsino CM, Fricke W. Recent developments in local concepts of fatigue
[2] Susmel L. Multiaxial Notch Fatigue. Elsevier; 2009. assessment of welded joints. Int J Fatigue 2009;31:2–11.
[3] Fatemi A, Shamsaei N. Multiaxial fatigue: An overview and some approximation [35] Fricke W. Recommended hot-spot analysis procedure for structural details of ships
models for life estimation. Int J Fatigue 2011;33:948–58. and FPSOs based on round-robin FE analyses. Int J Offshore Polar Eng 2002;12:
[4] Sonsino CM, Kueppers M. Multiaxial fatigue of welded joints under constant and 40–7.
variable amplitude loadings. Fatigue Fract Eng Mater Struct 2001;24:309–27. [36] Maddox SJ. Hot-spot stress design curves for fatigue assessment of welded
[5] Sonsino CM, Łagoda T. Assessment of multiaxial fatigue behaviour of welded joints structures. Int J Offshore Polar Eng 2002;12:134–41.
under combined bending and torsion by application of a fictitious notch radius. Int [37] Zhao XL, Packer JA. Fatigue design procedure for welded hollow section joints:
J Fatigue 2004;26:265–79. Recommendations of IIW Subcommission XV-E. Woodhead Publ 2000.
[6] Siljander A, Kurath P, Lawrence FV. Nonproportional fatigue of welded structures. [38] Haibach E. Betriebsfestigkeit – Verfahren und Daten zur Bauteilberechnung. 2nd
In Advances in fatigue lifetime predictive techniques. ASTM Int. 1992:319–38. ed. Berlin: Springer-Verlag; 2002.
[7] Yousefi F, Witt M, Zenner H. Fatigue strength of welded joints under multiaxial [39] Radaj D. Review of fatigue strength assessment of nonwelded and welded
loading: experiments and calculations. Fatigue Fract Eng Mater Struct 2001;24: structures based on local parameters. Int J Fatigue 1996;8:153–70. https://doi.
339–55. org/10.1016/0142-1123(95)00117-4.
[8] Witt M, Zenner H. Multiaxial fatigue behaviour of welded flange-tube connections [40] Neuber H. Kerbspannungslehre, Springer-Verlag, Vienna (1987) [Reprinted in
under combined loading. In: Experiments and lifetime prediction. 5th International translation as Theory of notch stresses. Berlin: Springer Publishers; 1958].
conference on biaxial/multiaxial fatigue and fracture; 1997. p. 421–34. [41] Peterson RE. Notch sensitivity. Sines G, Waisman JL (Eds).Metal fatigue. MacGraw-
[9] Razmjoo GR. Fatigue of load-carrying fillet welded joints under multiaxial loading. Hill; 1959: 293–307.
Fatigue – core research from TWI. Woodhead, UK, 63-99. [42] Radaj D. Design and analysis of fatigue resistant welded structures, Cambridge.
[10] Yung JY, Lawrence FV. Predicting the fatigue life of welded joints under combined England: Abington Publishing; 1990.
bending and torsion. Biaxial and multiaxial fatigue, EGF3. London 1989:53–69. [43] Schijve J. Fatigue predictions of welded joints and the effective notch stress
[11] Takahashi I, Ushijima M, Takada A, Akiyama S, Maenaka H. Fatigue behaviour of a concept. Int J Fatigue 2012;45:31–8. https://doi.org/10.1016/j.
box-welded joint under biaxial cyclic loads. Fatigue Fract Eng Mater Struct 1999; ijfatigue.2012.06.016.
22:869–77. [44] API. 579 RP-1/ASME FFS-1, Fitness-For-Service. American Petroleum Institute,
[12] Takahashi I, Takada A, Ushijima M, Akiyama S. Fatigue behaviour of a box-welded Houston 2007.
joint under biaxial cyclic loading: Effects of biaxial load range ratio and cyclic [45] Dong P. A robust structural stress method for fatigue analysis of offshore/marine
compressive loads in the lateral direction. Fatigue Fract Eng Mater Struct 2003;26: structures. J. Offshore Mech. Arct. Eng 2005;127(1):68–74.
439–48. [46] Mei J. Modeling of Multi-Axial Fatigue Damage Under Non-Proportional Variable-
[13] Dahle T, Olsson KE, Jonsson B. High strength welded box beams subjected to Amplitude Loading Conditions. PhD Thesis 2017.
torsion and bending – test results and proposed design criteria for torsion/bending [47] Sonsino CM. Influence of material’s ductility and local deformation mode on
interaction. In: First North European Engineering and Science Conference; 1997. multiaxial fatigue response. Int J Fatigue 2011;33:930–47.
p. 143–61. [48] Dong P. Stresses and stress intensities at notches: ‘anomalous crack growth’
[14] Archer R. Fatigue of a welded steel attachment under combined direct stress and revisited. Int J Fatigue 2003;25:811–25.
shear stress. International Conference of Fatigue of Welded Constructions, [49] Xing S, Dong P, Wang P. A quantitative weld sizing criterion for fatigue design of
Brighton, England; 1987;63-72. load-carrying fillet-welded connections. Int J Fatigue 2017;101:447–58. https://
[15] BS 7608: 2014 Guide To Fatigue Design And Assessment Of Steel Products, British doi.org/10.1016/j.ijfatigue.2017.01.003.
Standards Institution; 2014. [50] Vasu A, Xing S, Mei J, Chung CJ, Ravi D. A Computational Methodology for Multi-
[16] Fatigue CEN;2005.. Objective Fatigue Life Optimization of Welded Brake Flange on Full Beam Axles.
[17] Hobbacher AF. IIW recommendations for fatigue design of welded joints and SAE Technical Paper 2018-01-1019.
components. Springer; 2016.
[18] Niemi E, Fricke W, Maddox S. Structural hot-spot stress approach to fatigue
analysis of welded components - designer’s guide. Springer; 2016, iSBN: 978–981-
10-4458-3.
21

International Journal of Fatigue

Uploaded by

Copyright:

Available Formats

International Journal of Fatigue

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

International Journal of Fatigue

Uploaded by

Copyright:

Available Formats

International Journal of Fatigue 146 (2021) 106144

Contents lists available at ScienceDirect

International Journal of Fatigue

An overview and comparative assessment of approaches to multi-axial

2.3.2.1. Load path representation. For the multi-axial fatigue loading

components with global residual stresses. The specific definition of f(R)

Fig. 8. As-welded multi-axial fatigue test specimens and loading conditions.

Fig. 9. Residual stress relieved multi-axial fatigue test specimens.

can be found in [28]. Additionally, the thickness and bending ratio

Membraneσm /σn Bending Membrane Bendingτb /τn

You might also like