You are on page 1of 11

Ocean Engineering 186 (2019) 106074

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Notch stress analysis and fatigue strength assessment of tube-flange welded


joints under torsion loading
Yaoyu Hu a, b, Renjun Yan a, b, Wei Shen a, b, Kang Liu a, b, *
a
Key Laboratory of High Performance Ship Technology (Wuhan University of Technology), Ministry of Education, Wuhan, 430063, China
b
School of Transportation, Wuhan University of Technology, 430063, Wuhan, China

A R T I C L E I N F O A B S T R A C T

Keywords: A simple method named singularity length method (SLM) with a key parameter singularity length ‘as’ is proposed
Notch stress to estimate the shear stress distribution at the V-notch corner. For tube-flange welded joints (TFWJs) with 135�
Singularity length (as) weld toe notches, the singularity length ‘as’ is observed to be determined by different geometries and is sum­
Tube-flange welded joints
marized into concise formulas. A rapid calculation method of the mode III notch stress intensity factor (NSIF K3)
Notch stress intensity factor
Fatigue strength assessment
without any FEA process is presented based on the SLM. Good agreement is showed between the fine meshed FE,
the coarse meshed FE for the strain energy density (SED) and the proposed method. Six fatigue tests data of
TFWJs is collected to conduct fatigue strength assessments based on the nominal shear stress and the rapidly-
calculated NSIF K3. The results shows the NSIF K3 can be well applied as a fatigue parameter.

1. Introduction Hasebe (1997).


The clear analytical frame of NSIFs parameters at weld toe V-notches
Full-penetration steel tube-flange welded joints (TFWJs) are widely was established by Lazzarin and Tovo (1998), who had quantified the
used to connect tubes and plates. As is well recognized, geometrical contributions of in-plane tensile and in-plane shear stress modes.
discontinuities at weld toe corners always lead to localized stress con­ Approximate expressions of the stress distributions around the
centrations, which drive fatigue cracks initiating and propagating with V-notches were summarized by finite element analysis (FEA), which
alternating loads from sea waves (Cheng et al., 2018; Dong and Guedes considered the effect of geometric sizes of welded joints and could be
Soares, 2019). This phenomenon has brought about the study of stress further applied for fatigue strength assessment. However, the quantita­
distribution around the corners, which is necessarily required for an tive formulas were rather complex. Xu et al. (2013) simplified the
accurate fatigue design of welded structures (Lazzarin and Tovo, 1996). quantitative formulas applicable to the NSIF K1 (for in-plane tensile
The weld toe profiles is generally treated as a special case of sharp V- stress, mode I) and neglected the contribution of the NSIF K2 (for
notches (2α ffi 135o ), having a notch tip radius ρ ¼ 0 with hypothesis of in-plane shear stress, mode II). A more comprehensive formula consid­
conventional arc-welding technologies and un-dressed welds (Lazzarin ering the effect of both NSIF K1 and NSIF K2 were then proposed by Shen
and Tovo, 1998; Radaj et al., 2006), see left side of Fig. 1. The earliest et al. (2015). Xu et al. (2013) and Shen et al. (2015) focused on the
approach to evaluate V-notches stress fields was the eigenvalue expan­ planar V-notches which exist only in-plane tensile and shear stress. As
sion of such stress fields according to the asymptotic solution by Wil­ actual welded structures such as TFWJs are three-dimensional and suffer
liams (1952). Williams presented that, in linear elastic stress field, the from complicated loads, it is essential that further researches on shear
asymptotic stress near V-notch tip is singular and the singularity degree stress field and NSIF K3 (mode III) of 3D V-notches subject to
is depended on the notch angle. To accurately describe the singularity of out-of-plane loading should be implemented.
V-notch stress fields, Gross and Mendelson (1972) defined the Notch Fatigue strength assessment is a critical issue for offshore structures
Stress Intensity Factors (NSIFs) for in-plane tensile and in-plane shear (Ma et al., 2018). According to the welding standards EC3 and IIW
stress fields at V-notch tips, by extending the well-known Stress Intensity (Eurocode 3; 2005, Hobbacher, 2016), the nominal stress approach is
Factors (SIFs) concept for crack tips or slit tips. Definition of the NSIF the most direct method to assess the fatigue strength of steel welded
subjected to out-of-plane shear loading was later performed by Qian and joints. However, it does not consider the local geometrical

* Corresponding author. Key Laboratory of High Performance Ship Technology (Wuhan University of Technology), Ministry of Education, Wuhan, 430063, China.
E-mail address: cpmk12345@gmail.com (K. Liu).

https://doi.org/10.1016/j.oceaneng.2019.05.056
Received 26 March 2019; Received in revised form 24 May 2019; Accepted 27 May 2019
Available online 24 June 2019
0029-8018/© 2019 Elsevier Ltd. All rights reserved.
Y. Hu et al. Ocean Engineering 186 (2019) 106074

processing. On the other hand, few attentions have been focused on the
fatigue strength assessment of welded joints under out-of-plane shear
loading.
In this paper, a simple method (singularity length method, SLM) is
proposed to estimate the shear stress distribution near V-notches under
mode III loading. Subsequently, this method is applied to predict shear
stress around the weld toe of full-penetration TFWJs subjected to torsion
loading. The predicted stress distributions are compared with the FE
results to confirm the validity of the SLM. The analytical solution is
further extended to perform rapid NSIF calculation. The results of the
fine meshed FE, coarse meshed FE for the strain energy density (SED)
and the proposed rapid NSIF calculation method are in good agreement.
Fig. 1. Three-dimensional polar reference coordinates system centered at the Finally, a number of torsion fatigue experimental data of TFWJs is
V-notch of the weld toe of a typical full-penetration TFWJ. summarized. Statistical analysis of these data expressed in nominal
shear stress range and NSIF K3 range is performed to assess fatigue
discontinuities at the weld toe, which cause the assessment results strength of TFWJs.
inaccurate. It is widely recognized that better accuracy can be obtained
from local approaches (Radaj et al., 2006), such as hot spot stress 2. Stress field analysis of V-notch subject to mode III loading
method (Fricke and Kahl, 2005; Liu et al., 2014), NSIF method (Lazzarin
and Livieri, 2001; Lazzarin et al., 2004; Radaj, 2014), strain energy 2.1. Williams’ solution
density (SED) method (Lazzarin et al., 2008, 2010; Radaj, 2015), peak
stress method (PSM) (Meneghetti and Lazzarin, 2007; Meneghetti et al., Williams (1952) pointed out that the classical expressions, which
2017). These methods effectively avoid the influence of the stress sin­ describes the singular stresses at crack tips under elastic material hy­
gularity at the weld toe notch, but it still requires the process of estab­ pothesis, can be transferred to the shape of V-notches (e.g. weld toe
lishing finite element models, especially for some typical welded joints. notches). However, the degree of the stress singularity is reduced and
It drives the demand for a more convenient estimation method for the depends on the notch angle of V-notches, see Eq. (1). In a polar reference
fatigue characteristic values. coordinate system (r;θ), the stress field around V-notch is the summation
The NSIF method was proposed to estimate not only the crack of mode I and mode II stress fields.
initiation life but also the total life of welded joints (Radaj, 2014), 8 9 8 9 8 9
> > > > > >
because the vast majority of fatigue life is consumed during the crack >
< σθθ >
= >
< f1;θθ ðθÞ >
= >
< f2;θθ ðθÞ >
=
initiation stage. However, the fatigue parameter NSIF cannot be used for σrr ¼ λ1 rλ1 1 a1 f1;rr ðθÞ þ λ2 rλ2 1 a2 f2;rr ðθÞ (1)
> > > > > >
comparison between various V-notch angles due to the unit differences >
: τrθ >
; >
: f1;rθ ðθÞ >
; >
: f2;rθ ðθÞ >
;
according to Williams’ theory (Williams, 1952). Another disadvantage is
that the direct calculation of the NSIF needs very refined FE meshes to where r is the distance to the V-notch corner,λi is the eigenvalue defined
obtain the stress distribution on the bisector route (Fischer et al., 2016). in Eqs. (2) and (3),ai is the complex constant, fi ðθÞ is the stress functions.
To overcome the above-mentioned drawbacks of the NSIF, Lazzarin and Parameters are defined in Fig. 1 in the case of plane polar coordinate
co-authors (Lazzarin et al., 2008a,b) proposed the SED method based on system (r; θ).
the NSIF concept. The SED can be calculated over a material-related
control volume radius Rc surrounding the weld toe or the weld root sinðλ1 qπ Þ þ λ1 sinðqπÞ ¼ 0 (2)
(Lazzarin and Zambardi, 2001). The radius values Rc ¼ 0.28 mm for steel
sinðλ2 qπ Þ λ2 sinðqπÞ ¼ 0 (3)
welded joints and Rc ¼ 0.14 mm for aluminum welded joints were rec­
ommended for engineering fatigue assessment (Livieri and Lazzarin,
where q is related to the notch opening angle2α, see Eq. (4).
2005). More than 900 experiment data containing weld toe and weld
root fatigue failure form various welded joints was synthesized with the 2π 2α
2α ¼ 2π qπ; q ¼ (4)
SED, which was unified into aΔWscatter band (Berto and Lazzarin, π
2014). The SED method has been also verified for mixed loading brittle
fracture for the notch componets (Pook et al., 2016; Campagnolo et al., 2.2. Qian & Hasebe’s method
2016). It has been confirmed that the SED is mesh insensitive and a
coarse mesh is available for the SED method to assess fatigue strength Performing similar steps as Williams’ solution for mode I and mode II
(Lazzarin et al., 2008a,b). Rapid calculations of the NSIF of welded joints loading, Qian and Hasebe (1997) determined the local stress distribu­
can be realized by the SED from coarse meshes (Lazzarin et al., 2010). tions subject to out of plane shear loading (mode III) around the
Another NSIF-based method PSM was proposed to speed up the nu­ V-notches, see Eq. (5). Symbols are defined as shown in Fig. 1
merical estimation of NSIF parameters with special coarse meshes 8 9 8 9
< = < =
(Meneghetti et al., 2017). The advantage of the PSM is that only the τzθ cosðλ3 θÞ
¼ λ3 r λ3 1
a3 (5)
linear elastic peak stresses evaluated at the V-notch tip are sufficient, : τzr ; : sinðλ3 θÞ ;
instead of a number of numerical stress results in specific route or vol­
ume. The PSM has been verified for uniaxial or multiaxial fatigue where
strength assessment of welded joints (Meneghetti et al., 2017, 2019).
As mentioned above, the analysis of the local stress distribution and sinλ3 ð2π 2αÞ ¼ 0 (6)
the NSIF needs extremely refined FE meshes due to stress singularity The first non-zero eigenvalue is:
near the weld V-notch, which has limited the development of the NSIF
π
method. The SED method can improve the calculation of NSIF by rela­ λ3 ¼ (7)
2π 2α
tive coarse meshes, but the process of FE modeling is still necessary. It
would be more convenient if an engineering method can determine the Specifically, for2α ¼ 135o weld toe notch, the eigenvalueλ3 is equal to
stress distribution and the NSIF of a weld toe notch without FE 0.8.
By extending the definition of Notch Stress Intensity Factors (NSIFs)

2
Y. Hu et al. Ocean Engineering 186 (2019) 106074

proposed by Gross and Mendelson (1972) for mode I and mode II stress Therefore, according to above equations, the shear stress distribution
state, the mode III NSIF that quantifies the intensity of the out-of-plane of mode III around the weld toe V-notch can be described as
shear stress is defined as follows: h � i 1 λ3
τ 1
pffiffiffiffiffi τzθ ¼ pffi0ffiffiffiffi � 1 λ3 � fhtLf ðh; t; LÞ1=1 λ3 θ θÞ1=1 λ3 t (12)
K3 ¼ 2πlimðτzθ Þ r1 λ3 (8) 2π x
r→0 θ¼0

Applying the definition (8) to Eq. (5), Qian & Hasebe’s formula can where x is the distance to the notch corner, see Fig. 2, f1 ðh; t; LÞis a
be presented as a function of the NSIF K3 as follows: function of h,t, L, λ3 can be computed by Eq. (7).
8 9 8 9 For a given V-notch angle 2α at a given direction θ, parameter λ3 and
< =
τzθ K3 λ3 1 < cosðλ3 θÞ = cosðλ3 θÞ can be seen as constants and π is a fixed value. Taking all the
¼ pffiffiffiffiffir (9) constants intof1 ðh; t; LÞ, a new function can be defined as
: τzr ; 2π : sinðλ3 θÞ ;
ðh; t; LÞ ¼ f1 ðh; t; LÞðh; t; LÞ1=1 λ3
cosðλ3 θÞθÞ1=1 λ3
tπ 1=2ð1 λ3 Þ
(13)
It should be noted that mode III is always coupled with mode II,
especially near the crack (2α ¼ 0) tips (Pook et al., 2015, 2016). To simplify the calculation, a new equivalent parameter ‘as’ (as ¼
However, when the notch opening angle 2α is large than 102.6� , the f2 ðh; t; LÞ) is employed, then Eq. (12) can be simplified as follows:
mode II stress distribution is no longer singular (Radaj, 2014). The fa­
τ �as�1 λ3
tigue failure of TFWJs can be associated only with the mode I and mode τzθ ¼ p0ffiffi (14)
2 x
III stress distribution, neglecting the effect of mode II stress (Lazzarin
et al., 2004). So the torsion loading on TFWJs is considered to cause only Obviously, the equivalent parameter ‘as’ directly affects the singu­
mode III stress in this study. larity degree of the shear stress distribution. Meanwhile, ‘as’ has a unit
of length (its unit should be the same with the distance x). So ‘as’ can be
2.3. An improved approach: singularity length method regarded as a characteristic length of the singularity and is named
‘singularity length’.
Lazzarin and Tovo (1998) researched the scale effect of NSIF values, In next section, an approximate formula of ‘as’ for TFWJs under
found out that NIFS can be expressed in terms of non-dimensional torsion loading is carried out. The shear stress distribution can be
quantities as follows: conveniently estimated by Eq. (14) when the geometric parameters are
known to determine ‘as’ value. Similarly, this simple stress estimation
Ki;ref 1
Ki ¼ D λi
(10) method is named as ‘singularity length method (SLM)’.
D1 λi ref
3. Numerical analysis
whereKi;ref ði ¼ 1; 2; 3Þare the NIFs determined for a reference geometry
with a reference geometrical dimension equalling to Dref (for instance, 3.1. FE modeling
Dref could be the thickness of the principal plate in TFWJs), whereλi ði ¼
1; 2; 3Þare the eigenvalues depending on the notch angle2α. Linear elastic FE analysis is performed by the finite element software
The Ki;ref =D1 λi ref in Eq. (10) is dimensionally a stress, it can be ANSYS. Among the ANSYS element library, a special four-node, two-
expressed as a reference stressσ 0 . So that, when the considered welded dimensional and harmonic axisymmetric element PLANE 25 is adopted.
attachment is subjected to uniaxial loading, Eq. (10) can be rewritten as This employed element type is capable of simulating axisymmetric
Eq. (11) structures with external loads (not necessarily axisymmetric), which
K3 ¼ k3 τ0 t1 λ3
(11) must be defined as a series of harmonic functions (Fourier series).
Therefore, PLANE 25 is chosen for modeling three-dimensional
where τ0 is the remotely applied nominal torsion stress referring to the axisymmetric components subjected to torsion loading, keeping the
maximum shear stress at the outer surface of the tube (for TFWJs, it is advantage of building two-dimensional FE models. Fig. 3 shows an
defined as Eq. (15)), t is the thickness of the main plate of welded joints example of 2D FE model in terms of a TFWJ. The model is axisymmetric
(for instance, the tube thickness of TFWJs),k3 are non-dimensional pa­
rameters which depend on overall welded detail geometry and loading
types. When it comes to a typical TFWJ under torsion loading, k3 is a
function of three principal geometric parameters, namely h (the weld
size), t (thickness of the tube) and L (thickness of the flange). Geometric
parameters are shown in Fig. 2b. Note that g/R ¼ 1 means the flange is
solid.

(a) (b)

Fig. 2. Geometry details of a typical full-penetration TFWJ: (a) General view


(quarter model) (b) Parameters definition and stress distribution under tor­
sion loading. Fig. 3. Two-dimensional FE model with refined mesh using element PLANE 25.

3
Y. Hu et al. Ocean Engineering 186 (2019) 106074

around the Y-axis. As the flange is usually bolted in fatigue test and engineering range (0 < L/t � 3.0 and 0 < h/t � 4), the value of L is
engineering use, fixed constraints are assumed at the back face of the changed from 0.5 cm to 3 cm, the value of h is changed from 0.5 cm to
flange for FEA (see left end of the flange in Fig. 3). For steel, the Pois­ 4 cm and the value of t is changed from 1 cm to 4 cm. The following
son’s ratio v is 0.3 and the Young modulus E is 206 GPa. The nominal conclusions can be revealed:
torsion stress applied at the right end of tube is set as 1 MPa, and the
nominal torsion stress of tube is defined by the following classical (1) The tube thickness t plays a dominant role on the notch shear
expression: stress. With the increase of t, the stress distribution is obviously
enhanced.
Mt Mt
τ0 ¼ ¼ 4
(15) (2) The influence of h depends on the value of L and t. The stress
Wt 4
distribution has almost no change as h increases when t is small,
π ð2Rþ2tÞ ð2Ri Þ
16 2Rþ2t
while the increase of h enlarges the stress distribution in condi­
where R is the inner radius and t is the tube thickness, see Fig. 2b. tion of large t. Comparing Fig. 5a (L ¼ 0.5 cm), Fig. 5b (L ¼ 2 cm)
Since stress concentration at the V-notch corner is very serious, so and Fig. 5c (L ¼ 3 cm), it can be concluded that smaller L makes
that an extremely fine mesh is required nearby the weld toe while the effect of h more remarkable on the shear stress distribution.
coarser mesh is chosen at the other parts. The mesh convergence study is
conducted to ensure the accuracy of the FE results. The degree of 3.3. Approximate formulas of ‘as’
refinement close to the corner is determined by the spacing ratio set in
the ANSYS software. The spacing ratio is the ratio of the maximum mesh According to the analysis in section 3.2, three main parameters t, h,
size to the minimum one on a path or line where the mesh sizes are and L are considered as the major controllable variables to determine the
changed in gradient. The stress distributions along the x line (see equivalent parameter ‘as’. Varying t, h, and L among the engineering
Fig. 2b) adopting different mesh sizes are presented in Fig. 4. It can be parameters ranges, 168 FE models are conducted to obtain corre­
seen that the shear stress distributions are essentially unchanged when sponding shear stress distribution in the direction of tube thickness (see
the minimum mesh size is 5 � 10 3 mm or smaller. Considering the x line in Fig. 2b). Partial results have been presented in Fig. 5. The
balance of simulation accuracy and acceptable computational cost, the detailed dimensions of these FE models are summarized in Table 1.
minimum mesh size at the notch corner is set to 10 4 mm in the FE results provide discrete data of the stress valueτzθ and the distance
following FE analysis (see enlarged scale in Fig. 3). from the corner x. The eigenvalueλ3 is 0.8 for 135� weld toe notch.
Therefore the parameter ‘as’ can be well determined by processing such
3.2. Parametric study data in MATLAB software with Eq. (14) using nonlinear least squares
fitting method. The fitting results (t ¼ 2 m) are shown in Table 2. It can
The geometrical parameters of a typical TFWJ are defined in Fig. 2b. be seen there is a function relation between L, h and t and the corre­
A series of finite element analysis are employed by using ANSYS para­ sponding fitting value of ‘as’.
metric design language (APDL), aiming to determine the contribution of As demonstrated in Fig. 6, two cases of g/R ¼ 0 (hollow flange) and
each parameter to the shear stress distribution in the vertical direction of g/R ¼ 1 (solid flange), are respectively considered when fitting ‘as’ for­
the tube (see x line in Fig. 2b).The ranges of geometrical parameters mula with parameter L, h and t. When the value of h/L is small, the in­
values correspond to the actual structure sizes. The length of the tube l is fluence of h is invisible and can be neglected. The ‘as’ formula can be
set as a relatively large value to ensure providing uniform stress field obtained according to the other two parameters L and t. However, with a
near the notch corner. large value of h/L, a big size of the weld size h enhances the local stress
From engineering view, the flange thickness L, the weld size h, and concentration at the corner, the formula needs to contain the obvious
the tube thickness t would affect the stress concentration at the weld toe, effect of h.
while other parameters (the outer flange length H and the tube inner The analysis of Fig. 6 reveals that ‘as’ is determined by two
radius R) have little influence on the stress distribution. Therefore, the approximate formulas. The smaller of calculated values of these two
influences of parameters L (thickness of the flange), h (the weld size), formulas is closer to the fitting ‘as’ from FEA. Therefore, estimation
and t (thickness of the tube) on shear stress distribution are discussed by formulas of ‘as’ for TFWJs under torsion loading can be deduced as
keeping other parameters as constant, see Fig. 5. In view of the following simple form:
� � t �0:9 �
asg=R¼0 ¼ min 0:09L þ 0:9t; ð0:2L þ h 0:1tÞ (16a)
h
� � t �0:9 �
asg=R¼1 ¼ min 0:9 � ð0:09L þ 0:9tÞ; 0:9 � ð0:2L þ h 0:1tÞ
h
(16b)
For convenience and sufficinet accuracy, the approximate formula of
g/R ¼ 0 is set to 0.9 times the formula of g/R ¼ 1. Fig. 7 presents the
errors of ‘as’ values between the estimation formulas and fitting results
of FEA. Most errors are less than 10%, and only case holds the largest
error of 21%. It can be concluded that the approximate formulas Eq. (16)
can estimate ‘as’ values with satisfactory accuracy.
For practical aspects in engineering, it is useful to note that the
similarity of stresses is contained in ‘as’ formulas. If two similar geom­
etries are scaled by factor ‘d’, i.e. the absolute dimension of case b is
d times as large as that of case a. The parameter ‘as’ can be seen as a
function linearly related to the whole geometry size. When the boundary
conditions on external forces and displacements are the same, the
0 following relations can be derived from Eq. (14):

Fig. 4. Shear stress distributions along the x line adopting different mesh sizes.

4
Y. Hu et al. Ocean Engineering 186 (2019) 106074

Fig. 5. Shear stress distribution curves under unit torsion loading: R/t ¼ 3, H/h ¼ 5, g/R ¼ 1 and (a) L ¼ 0.5 cm; (b) L ¼ 2 cm; (c) L ¼ 3 cm.

� � � �
τ �asb �1 λ3 τ0 d � asa 1 λ3 τ0 asa 1 λ3
τzθ; b ðxÞ ¼ p0ffiffi ¼ pffiffi ¼ pffiffi
Table 1 2 x 2 x 2 x=d
Dimensions of FE models. ¼ τzθ;a ðx=dÞ⇒τzθ;a ðxÞ ¼ τzθ;b ðdxÞ (17)
t(cm) h(cm) L(cm) g/R R/t H/h

1–4 0.5–4 0.5–3 0,1 3 5 4. Verification of singularity length method

When the geometric dimensions of a TFWJ are known, the ‘as’ value
can be rapidly computed by Eq. (16). Then substituting ‘as’ into Eq. (14),

Table 2
The fitting results of ‘as’ (t ¼ 2 cm) (the unit of ‘as’ is cm).
L/cm h ¼ 0.5 cm h ¼ 1 cm h ¼ 2 cm h ¼ 4 cm

asg=R¼0 asg=R¼1 asg=R¼0 asg=R¼1 asg=R¼0 asg=R¼1 asg=R¼0 asg=R¼1

0.5 1.2106 1.1455 1.5198 1.4709 1.8362 1.8187 1.9972 1.9954


0.75 1.4292 1.2976 1.6433 1.5546 1.8803 1.8502 2.0109 1.9992
1 1.8878 1.5761 1.8887 1.7066 1.9619 1.9036 2.0253 2.0185
1.5 1.6156 1.4152 1.7463 1.6198 1.9158 1.8742 2.0175 2.0129
2 2.1331 1.7285 2.0412 1.8156 2.0188 1.9473 2.0277 2.0189
2.5 2.0495 1.6737 1.9812 1.7687 1.9921 1.9247 2.0274 2.0193
3 2.1764 1.7617 2.0786 1.8502 2.0320 1.9599 2.0285 2.0194

Fig. 6. The prediction of ‘as’: (a) g/R ¼ 0; (b) g/R ¼ 1.

5
Y. Hu et al. Ocean Engineering 186 (2019) 106074

or smaller in other cases. It can be drawn that the SLM has good
performance on shear stress estimation near the weld toe corner.
(2) The shear stress decreases rapidly away from the corner.
Specially, when the distance from the corner equals to as/200,
the shear stress reduces to about twice the nominal shear stress
(2τnom ). This phenomenon can be explained by substituting
x ¼ as/200 into Eq. (14), the shear stress is
τ �as�1 λ3 τ0
τzθ ¼ p0ffiffi ¼ pffiffi ð200Þ0:2 ¼ 2:04τ0 � 2τ0 (18)
2 x 2
Similarly, it can be inferred that when the distance x ¼ as/5.5, the
shear stress τzθ is about equal to the nominal shear stressτnom .

5. Rapid NSIF calculation

It is widely recognized that NSIFs can be employed as fatigue


parameter for life prediction of cyclically loaded welded joints with V-
notch corners (Lazzarin and Tovo, 1998; Fischer et al., 2016). However,
the computation of exact NSIF requires extremely refined finite element
Fig. 7. The error of ‘as’ values between the estimation formulas Eq. (16) and
the fitting results of FEA. mesh, where a row of finite element nodes must be along the notch
bisector line and the mesh size around the notch tip need to be adjusted
around 10 4 mm (Meneghetti, 2013). As the concise formulas of ‘as’ for
the shear stresses along the x line (see Fig. 2b) can be obtained. This is
THWJs has been established, it is feasible to rapidly calculate the mode
the process how singularity length method (SLM) determines the shear
III NSIF K3 without any FE process by transforming the formulas in the
stress distribution. To verify the capability of the method, shear stress
SLM. Based on Eq. (9) and Eq. (14), the mode III NSIF can be expressed
distributions near the toe corner (0–100 mm) derived from the SLM and
as:
the FE results are compared via four examples, as shown in Fig. 8 (g/
pffiffiffiffiffi pffiffiffi
R ¼ 0) and Fig. 9 (g/R ¼ 1), the following can be concluded: 2πτθz πτ0 as1 λ3
K3 ¼ λ 1 ¼ (19)
r 3 cosðλ3 θÞ cosðλ3 θÞ
(1) Figs. 8a and 9a show the cases with the largest ‘as’ error between
the formula Eq. (16) and the fitting value. While even in these two For a TFWJ with a 135� notch opening angle (λ3 ¼ 0:8θ ¼ :22:50 ),
worst cases, the shear stresses errors between SLM and FEA re­ the NSIF K3 becomes:
sults are less than 6%. Moreover, the stress errors are within 5%

Fig. 8. Stress comparison and error analysis when g/R ¼ 0. (a) Example 1: L/t ¼ 0.125, h/t ¼ 0.25. (b) Example 2: L/t ¼ 0.125, h/t ¼ 1. (c) Example 3: L/t ¼ 3, h/
t ¼ 0.5. (d) Example 4: L/t ¼ 3, h/t ¼ 4.

6
Y. Hu et al. Ocean Engineering 186 (2019) 106074

Fig. 9. Stress comparison and error analysis when g/R ¼ 1. (a) Example 1: L/t ¼ 0.125, h/t ¼ 0.125. (b) Example 2: L/t ¼ 0.125, h/t ¼ 1. (c) Example 3: L/t ¼ 3, h/
t ¼ 0.5. (d) Example 4: L/t ¼ 3, h/t ¼ 4.

pffiffiffi
π modeling process. According to the NSIF K3 definition Eq. (8), any shear
K3 ¼ as0:2 τ0 ¼ 1:864as0:2 � τ0 (20)
cosð0:1πÞ stress value along the bisector, but adequately close to the weld toe
corner, can be used to determine the NSIF. The shear stresses and the
where ‘as’ can be directly determined by Eq. (16). corresponding distance rj of the nodes along the bisector line are taken
Lazzarin and Tovo (1998) employed fine meshed FE models to into account. At each node j, the mode III NSIF can be computed as
directly evaluate the Mode I and II components of NSIF. According to follows:
their method, the NSIF K3 can be obtained in the same way. Fig. 10 pffiffiffiffiffi
shows a typical FE model of a TFWJ (L ¼ 1 cm, h ¼ 0.5 cm, t ¼ 1 cm, K3;j ¼ 2πτzθ;j r1j λ3 (21)
g/R ¼ 1) to compute NSIF K3 and the mesh detail of the notch tip (order All these nodes provide the NSIF values by means of Eq. (21), as
of magnitude of the smallest mesh size equals to 10 4 mm). shown by the red line in Fig. 10. The shear stress distribution of the
Two-dimensional element PLANE 25 is employed to simplify the nodes near the corner is nearly linear in a log-log plot. There is a large
stability zone ranging from about 10 4 r/t to 10 1 r/t, where the analysis
gives a constant and well-defined value for NSIF K3. However, it is very
complicated to conduct the process of fine meshed FE method. When
applying the Eq. (20), it becomes much easier to gain the value of NSIF
K3.
Based on the closed-form expressions correlating NSIFs and local
strain energy density (SED) of sharp notches, the NSIFs can be rapidly
calculated from the SED on a control volume around the weld toe, since
coarse meshes can provide sufficiently accurate SED (Lazzarin et al.,
2010). Under pure mode III loading, the link between local SED and K3
can be expressed as follow (Lazzarin et al., 2008a,b).
" #2
e 3 K3
W¼ (22)
E R1C λ3

It gives:
sffiffiffiffiffiffiffiffiffiffiffiffiffi
E�W pffiffiffiffi
K 3 ¼ RC1 λ3
¼ 691:91 W (23)
e3

where E is the Young’s modulus, Parameter e3 depends on the notch


angle 2α and the Poisson’s ratio v. For 2α ¼ 135o and v ¼ 0.3, e3 is found
Fig. 10. Shear stresses along the bisector and mode III NSIF calculation by to be 0.2586 (Lazzarin et al., 2008a,b). Rc is the radius of the control
refined meshed FE model.

7
Y. Hu et al. Ocean Engineering 186 (2019) 106074

6. Fatigue strength assessment

The most common way to assess and to design the fatigue strength of
welded joints is using the reference S–N curves based on the nominal
stress range, which are divided into different fatigue classes (FAT, ex­
press in MPa, of the fatigue strength at 2 million cycles) curves for
various structural details (Hobbacher, 2016). These fatigue strength
curves are linear in the double logarithm coordinate, which can be
defined as following expressions:
logðNÞ ¼ log C þ dσ k logðSÞ (24)

where N is the predicted fatigue life (cycles) under the stress range S; C is
a constant whose value is specific to each structural detail; σ is the
standard deviation of log (N); d is the number of the standard deviation,
and d equals to 2 (below the mean curve) for safe fatigue design with
97.7% survival probability; k is the inverse slope of S–N curve.
The IIW (International Institute of Welding) and EC3 (Eurocode 3;
2005) standards provide several FAT S–N curves for various welded
structural details under direct or normal stress ranges (difference be­
tween maximum and minimum value). However, for shear stress ranges,
the fatigue phenomena are less concerned and there is no well-defined
Fig. 11. Coarse meshed FE model for local SED. constant amplitude fatigue limit. The standards just provide a unique
S–N curve (FAT ¼ 100, k ¼ 5) for toe fatigue failure of steel welded joints
on the basis of the shear stress range.
volume, it turns out to be 0.28 mm for welded joints in mild steels
The approach based on notch stress intensity factor (NSIF) naturally
(Livieri and Lazzarin, 2005). The coarse meshes used to obtain the local
takes the influence of weld toe detail and the scale effect into account,
SED of TFWJs are referring to the literature (Lazzarin et al., 2010).
and has been considered to be more accurate than the nominal stress
Fig. 11 shows overall meshes of an FE example. At the weld toe, the
approach (Fischer et al., 2016). Livieri and Lazzarin (2005) systemati­
control volume is a circle of radius 0.28 mm, embedded in a square with
cally analyzed large amount of fatigue tests data under tension or
4 mm sides.
bending loading using the NSIF approach, proposed a design mean curve
To compare aforementioned three NSIF calculation methods (fine
and corresponding scatter bound of mode I NSIF for welded joints.
meshed FE, coarse meshed FE for SED, and Eq. (20)), six typical TFWJ
However, fatigue strength assessment under shear stress ranges, which is
geometry sizes are investigated here. The external load is assumed as
related to the mode III NSIF, is also ambiguous.
torsion loading with nominal torsion stress 1 MPa for convenience. The
In order to assess the fatigue strength of welded joints in alternating
comparison results are summarized in Table 3. The fine meshed FE re­
shear stress state, and to validate the application of rapid NSIF calcu­
sults are taken as the reference. The errors varies from 4.21% to 5.16%
lation for fatigue strength assessment at the same time, six series of fa­
for coarse meshed FE, and are within 2% for Eq. (20). Both the coarse
tigue test data concerning TFWJ specimens subjected to pure torsion
meshed FE method and the proposed analytical solution Eq. (20) have
have been summarized from the literature. All specimens are made of
good performance on rapid NSIF calculation. It demonstrates that the
structural steels. Table 4 and Fig. 12 report the main geometry sizes and
NSIF K3 of TFWJs can be rapidly calculated though the proposed
fatigue loading conditions of the specimens. Note that the round or
analytical solution, having an advantage of very simple expressions (Eq.
square flange would make no difference according to the parametric
(20) and Eq. (16)) without any FE processing.
study in section 3.2. For all specimens, a weld toe angle equalling to
135� is assumed the same as that in the theory and numerical analysis in

Table 3
Comparison of the mode III NSIF calculations between Eq. (20), fine meshed FE and coarse meshed FE for SED.
L (cm) h (cm) T (cm) g/R Fine mesh K3 (Mpa cm0.2) Coarse mesh for SED Rapid calculation by Eq. (20)
0.2
W (N mm/mm ) 3 K3 (Mpa cm ) Δ% as(cm) K3 (Mpa cm0.2) Δ%

5
3 0.5 1 1 1.815 1.91 � 10 1.908 5.16 0.891 1.821 0.36
5
3 2 2 1 2.132 2.59 � 10 2.223 4.28 1.863 2.111 0.98
5
3 4 4 1 2.440 3.43 � 10 2.557 4.83 3.483 2.392 1.95
5
1 0.5 1 0 1.858 1.97 � 10 1.936 4.21 0.990 1.860 0.12
5
3 2 2 0 2.148 2.64 � 10 2.244 4.48 2.070 2.156 0.37
5
3 4 4 0 2.472 3.50 � 10 2.581 4.41 3.870 2.443 1.16

Table 4
Geometry sizes and fatigue loading conditions of the tube-flange welded specimens.
Series References specimen geometry (Fig. 12) L (mm) T (mm) h (mm) g/R as (mm) Number of specimens

1 Sonsino, 1995, 2001 a&R 25 10 9 1 10.1 14(R ¼ 1)


2 Yousefi et al. (2001) a&R 25 8 10 1 8.5 9(R ¼ 0); 8(R ¼ 1)
3 Amstutz et al. (2001) a&R 25 7.7 9 1 8.3 4(R ¼ 0); 3(R ¼ 1)
4 Seeger, 1987 b&S 20 8 6.3 0 9.0 5(R ¼ 1)
5 Siljander et al. (1992) c&S 9.3 9.5 8 0 9.4 2(R ¼ 0); 4(R ¼ 1)
6 Razmjoo (2000) c&R 12 7 8 0 7.4 2(R ¼ 0)

8
Y. Hu et al. Ocean Engineering 186 (2019) 106074

Fig. 12. Specimen geometry. (R or S means the flange is round or square).

the previous sections. Moreover, considering the worst case of un- strength Δτ with survival probability of 97.7% is 138 MPa, which is
dressed welding in all fatigue specimens, the weld toe cross section is much higher than the recommended value 100 MPa (FAT 100). The
a pointed V-notch with zero radius. All selected specimens are under inverse slope of the fatigues curves is 6.6, and the scatter index defined
pure torsion loading (mode III loading) and their failure location are at by Eq. (25) is 1.84.
the weld toe.
ΔτðNA ; Ps ¼ 2:3%Þ
Fig. 13 shows all summarized fatigue tests results expressed in Tσ ¼ (25)
ΔτðNA ; Ps ¼ 97:7%Þ
nominal shear stress range. According to the IIW and EC3 standards,
FAT 100 curve is recommended for these data. It is obvious that the A rapid calculation method for NSIF K3 has been established in the
design curve of standards is very safe but too conservative, especially in section 5. Fig. 14 shows the fatigue data expressed in rapidly-calculated
high cycle region. NSIF K3 range. Statistical analysis of these NSIF values has been carried
To be more accurate, statistical analysis has been conducted with the out with K–N curves. K–N curve can be defined as Eq. (26) referring to
same tests data. S–N curves with Ps (survival probability) ¼ 2.3% and the basic S–N curve expression. The mean value of ΔK3 at
97.7% (i.e. mean curve � two standard deviations) are demonstrated in NA ¼ 2⋅106 cycles is 136 MPa m0.2.The scatter index TK is 1.83 and the
Fig. 13. At the reference 2 million cycles (NA ¼ 2⋅106 cycles), the fatigue inverse slope of K–N curves is 6.7, both values are close to the analysis

Fig. 13. Fatigue assessment of torsion fatigue tests results: statistical analysis of nominal shear stress range and the design curve in standards (FAT 100, k ¼ 5).

9
Y. Hu et al. Ocean Engineering 186 (2019) 106074

The mean fatigue strength at NA is 120 MPa m0.2. It can be seen that the
proposed design K–N curves are quite suitable for the tests data with
satisfactory accuracy. The K–N curve with survival probability of 97.7%
can be employed for safe fatigue design of steel TFWJs by the fatigue
parameterΔK3 .Based on Eq. (26), this curve can be expressed as
following simple form:
logðNÞ ¼ 5 logðK3 Þ þ 16 (27)

7. Conclusions

A brief method named singularity length method (SLM) is introduced


to determine the shear stress distribution at the V-notch corner subject
to mode III loading. The key of SLM is the equivalent parameter ‘as’
named singularity length, which directly relates to the stress singularity
and physically has a unit of length. For tube-flange welded joints
(TFWJs) under torsion loading, ‘as’ is expressed as formulas of three
main geometric parameters though a series of finite element analysis.
Thus, the shear stress distribution near the weld toe corner can be easily
estimated by Eq. (14) for known dimensions of a TFWJ. Following this,
Fig. 14. Fatigue assessment of torsion fatigue tests results: statistical analysis of
the rapidly-calculated formula Eq. (20) of the mode III NSIF is derived
rapidly-calculated mode III NSIF (ΔK3 ) range.
based on the SLM. Six series of fatigue tests data of TFWJs under pure
torsion are employed for fatigue strength assessments based on the
results of the nominal shear stress range. One possible reason may nominal shear stress range and the rapidly-calculated mode III NSIF
explain the phenomenon. The available experiment data provides a range. The following conclusions can be drawn:
small change of tube thicknesses (7 mm-10mm) and weld sizes (6.3 mm-
10mm), which means each test specimen have similar size effect. It 1. The shear stress distribution around weld toe corner predicted by the
shows that mode III NSIF can be applied for fatigue strength assessment SLM is in good agreement with the FE results for TFWJs under tor­
of TFWJs under torsion loading. sion loading. Among all the geometric parameters of TFWJs, thick­
logðNÞ ¼ logCK þ dσ k logðKÞ (26) ness of the flange L, the weld size h, and thickness of the tube t play
the dominant role on the shear stress distribution.
Furthermore, modified K–N curves with a uniform scatter index and 2. Comparison of NSIF K3 calculations shows excellent agreement be­
curve slope would be more practical in fatigue design. Fig. 15 shows the tween the fine meshed FE, the coarse meshed FE for SED and pro­
proposed design K–N curves. The proposed inverse slope k is 5.00 ac­ posed analytical solution Eq. (20).
cording to the inverse slope of S–N curve for shear stress ranges in 3. Fatigue assessment of the employed test data indicates that the
standards. The proposed scatter index with survival probabilities design S–N curve in the standards is too conservative, and the mode
2.3–97.7% is 1.90, which is consistent with the scatter index T ¼ 1.50 III NSIF is well applied for fatigue strength assessment of TFWJs. For
with survival probabilities of 10–90% determined by Haibach (1989).

Fig. 15. Fatigue assessment of torsion fatigue tests results: proposed design S–N curves of mode III NSIF (ΔK3 ) range.

10
Y. Hu et al. Ocean Engineering 186 (2019) 106074

practical fatigue design, modified K–N curves with a uniform scatter Lazzarin, P., Tovo, R., 1998. A notch intensity factor Approach to the stress analysis of
welds. Fatigue Fract. Eng. Mater. Struct. 21 (9), 1089–1103.
index and curve slope are further proposed, and the lower curve used
Lazzarin, P., Zambardi, R., 2001. A finite-volume-energy based approach to predict the
for safe fatigue design has a concise expression. static and fatigue behavior of components with sharp V-shaped notches. Int. J. Fract.
112 (3), 275–298.
Acknowledgements Liu, G., Liu, Y., Huang, Y., 2014. A novel structural stress approach for multiaxial fatigue
strength assessment of welded joints. Int. J. Fatigue 63, 171–182.
Livieri, P., Lazzarin, P., 2005. Fatigue strength of steel and aluminium welded joints
The research project is supported by the National Natural Science based on generalised stress intensity factors and local strain energy values. Int. J.
Foundation of China (Grant No. 51609185). Fract. 133 (3), 247–276.
Ma, Y., Han, C., Qu, X., 2018. Fatigue assessment method of marine structures subjected
to two Gaussian random loads. Ocean. Eng. 165, 107–122.
References Meneghetti, G., 2013. The peak stress method for fatigue strength assessment of tube-to-
flange welded joints under torsion loading. Weld. World 57 (2), 265–275.
Amstutz, H., St Rzel, K., Seeger, T., 2001. Fatigue crack growth of a welded tube–flange Meneghetti, G., Campagnolo, A., Babini, V., Riboli, M., Spagnoli, A., 2019. Multiaxial
connection under bending and torsional loading. Fatigue Fract. Eng. Mater. Struct. fatigue assessment of welded steel details according to the peak stress method:
24 (5), 357–368. industrial case studies. Int. J. Fatigue 125, 362–380.
Berto, F., Lazzarin, P., 2014. Recent developments in brittle and quasi-brittle failure Meneghetti, G., Campagnolo, A., Rigon, D., 2017. Multiaxial fatigue strength assessment
assessment of engineering materials by means of local approaches. Mater. Sci. Eng. R of welded joints using the peak stress method – Part I: approach and application to
Rep. 75, 1–48. aluminium joints. Int. J. Fatigue 101, 328–342.
Campagnolo, A., Meneghetti, G., Berto, F., 2016. Rapid finite element evaluation of the Meneghetti, G., Lazzarin, P., 2007. Significance of the elastic peak stress evaluated by Fe
averaged strain energy density of mixed-mode (I þ Ii) crack tip fields including the analyses at the point of singularity of sharp V-notched components. Fatigue Fract.
T-stress contribution. Fatigue Fract. Eng. Mater. Struct. 39 (8), 982–998. Eng. Mater. Struct. 30 (2), 95–106.
Cheng, B., Li, C., Lou, Y., Zhao, X., 2018. Parametric Fe modeling to predict hot spot Pook, L.P., Campagnolo, A., Berto, F., 2016. Coupled fracture modes of discs and plates
stress concentrations of bird-beak shs joints in offshore structures. Ocean. Eng. 160, under anti-plane loading and a disc under in-plane shear loading. Fatigue Fract. Eng.
54–67. Mater. Struct. 39 (8), 924–938.
Dong, Y., Guedes Soares, C., 2019. Stress distribution and fatigue crack propagation Pook, L.P., Campagnolo, A., Berto, F., Lazzarin, P., 2015. Coupled fracture mode of a
analyses in welded joints. Fatigue Fract. Eng. Mater. Struct. 42 (1), 69–83. cracked plate under anti-plane loading. Eng. Fract. Mech. 134, 391–403.
Eurocode 3, 2005. Design of Steel Structures – Part 1–9: Fatigue. CEN. Qian, J., Hasebe, N., 1997. Property of eigenvalues and eigenfunctions for an interface V-
Fischer, C., Fricke, W., Rizzo, C.M., 2016. Review of the fatigue strength of welded joints notch in antiplane elasticity. Eng. Fract. Mech. 56 (6), 729–734.
based on the notch stress intensity factor and sed approaches. Int. J. Fatigue 84, Radaj, D., 2014. State-of-the-Art review on extended stress intensity factor concepts.
59–66. Fatigue Fract. Eng. Mater. Struct. 37 (1), 1–28.
Fricke, W., Kahl, A., 2005. Comparison of different structural stress approaches for Radaj, D., 2015. State-of-the-Art review on the local strain energy density concept and its
fatigue assessment of welded ship structures. Mar. Struct. 18 (7–8), 473–488. relation to thej -integral and peak stress method. Fatigue Fract. Eng. Mater. Struct.
Gross, B., Mendelson, A., 1972. Plane elastostatic analysis of V-notched plates. Int. J. 38 (1), 2–28.
Fract. Mech. 8 (3), 267–276. Radaj, D., Sonsino, C.M., Fricke, W., 2006. Fatigue Assessment of Welded Joints by Local
Haibach, E., 1989. Service Fatigue-Strength-Methods and Data for Structural Analysis. Approaches, second ed. Woodhead Publishing, Cambridge.
Dusseldorf, VDI. Razmjoo, G., 2000. In: Abington (Ed.), Fatigue of Load-Carrying Fillet Welded Joints
Hobbacher, A., 2016. Recommendations for Fatigue Design of Welded Joints and under Multiaxial Loadings. TWI REF. 7309.02/96/909, Cambridge, UK.
Components. IIW Collection. Springer International Publishing. Seeger, T.,O.R., 1987. Tolerable and allowable shear stresses at fatigue loaded welded
Lazzarin, P., Berto, F., Gomez, F., Zappalorto, M., 2008. Some advantages derived from joints. Stahlbau 56 (8), 231–238.
the use of the strain energy density over a control volume in fatigue strength Shen, W., Barltrop, N., Yan, R., Liu, E., Qin, K., Song, L., 2015. Stress field and fatigue
assessments of welded joints. Int. J. Fatigue 30 (8), 1345–1357. strength analysis of 135-degree sharp corners based on notch stress strength theory.
Lazzarin, P., Berto, F., Zappalorto, M., 2010. Rapid calculations of notch stress intensity Ocean. Eng. 107, 32–44.
factors based on averaged strain energy density from coarse meshes: theoretical Siljander, A., Kurath, P., Lawrence, F.V., 1992. Nonproportional fatigue of welded
bases and applications. Int. J. Fatigue 32 (10), 1559–1567. structures. In: Mitchel, M.R., Landgraf, R. (Eds.), Adv. Fatigue Lifetime Predict. Tech.
Lazzarin, P., Livieri, P., 2001. Notch stress intensity factors and fatigue strength of ASTM STP 1122. ASTM, Philadelphia, PA, pp. 319–338, 1992.
aluminium and steel welded joints. Int. J. Fatigue 23 (3), 225–232. Sonsino, C., 1995. Multiaxial fatigue of welded joints under in-phase and out-of-phase
Lazzarin, P., Livieri, P., Berto, F., Zappalorto, M., 2008. Local strain energy density and local strains and stresses. Int. J. Fatigue 17 (1), 55–70.
fatigue strength of welded joints under uniaxial and multiaxial loading. Eng. Fract. Sonsino, C., 2001. Influence of load and deformation-controlled multiaxial tests on
Mech. 75 (7), 1875–1889. fatigue life to crack initiation. Int. J. Fatigue 23 (2), 159–167.
Lazzarin, P., Sonsino, C.M., Zambardi, R., 2004. A notch stress intensity approach to Williams, M.L., 1952. Stress singularities resulting from various boundary conditions in
assess the multiaxial fatigue strength of welded tube-to-flange joints subjected to angular corners of plates in extension. J. Appl. Mech. 19 (4), 526–528.
combined loadings. Fatigue Fract. Eng. Mater. Struct. 27 (2), 127–140. Xu, L., Gao, S., Nbarltrop, 2013. Simple method of fatigue assessment for structure with
Lazzarin, P., Tovo, R., 1996. A unified approach to the evaluation of linear elastic stress singular point. Ship Eng. 76 (6), 1105–1112.
fields in the neighborhood of cracks and notches. Int. J. Fract. 78 (1), 3–19. Yousefi, F., Witt, M., Zenner, H., 2001. Fatigue strength of welded joints under multiaxial
loading: experiments and calculations. Fatigue Fract. Eng. Mater. 24 (5), 339–355.

11

You might also like