Thin-Walled Structures 58 (2012) 67–78

Contents lists available at SciVerse ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

The development of fatigue design formulas for the outer brace SCFs
in offshore three-planar tubular KT-joints
Hamid Ahmadi 1, Mohammad Ali Lotfollahi-Yaghin n, Mohammad H. Aminfar 2
Faculty of Civil Engineering, University of Tabriz, 29 Bahman Boulevard, Tabriz, Iran

a r t i c l e i n f o abstract

Article history: A set of parametric stress concentration factor (SCF) analyses is carried out in the present research for
Received 30 August 2011 right-angle three-planar tubular KT-joints under three different axial loading conditions. Three-planar
Received in revised form KT-joints are one of the most important joint types in the six-legged offshore jacket-type structures
20 April 2012
since they connect the bracing members to the main legs of the jacket template. The analysis results are
Accepted 20 April 2012
Available online 25 May 2012
used to present general remarks on the effect of geometrical characteristics and loading conditions on
the SCFs at different saddle positions on the main (outer) braces. Based on the results of FE analyses
Keywords: which are verified against the experimental data, a complete set of SCF databases is developed.
Offshore jacket structure Afterwards, through nonlinear regression analysis, a new set of SCF design equations is established for
Right-angle three-planar KT-joint
the fatigue design of three-planar KT-joints under three different axial load cases. An assessment study
Fatigue
of these equations is conducted against the experimental data and the acceptance criteria proposed by
Stress concentration factor (SCF)
Parametric SCF equation the UK Department of Energy (DoE).
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction
Steel circular hollow sections (CHSs) are widely used in
offshore structures due to their good resistances against bending,
torsion and buckling, and a high strength-to-weight ratio. In a
tubular joint, the members are connected by welding the pre-
pared profiled end of the brace members onto the outer surface of
the chord member. The fatigue design of such joints constitutes a
critical factor towards safeguarding the integrity of tubular
structures. The complex joint geometry causes significant stress
concentrations at the vicinity of the welds. Under repeated
loadings, they result in the formation of cracks which can grow
to a size sufficient to cause the joint failure. The location of
maximum stress concentration is called ‘‘hot-spot’’ and the
corresponding local stress is referred to as ‘‘hot-spot stress’’ (hss).
For fatigue design purposes, the ‘‘hot-spot stress method’’ has
been quite efficient and popular. According to this method, the
nominal stress range at the joint members is multiplied by an
appropriate stress concentration factor (SCF) to provide the so-
called ‘‘geometric stress’’ S0 at a certain location. Hence, this
design method relies on the accurate prediction of SCFs for
n
Corresponding author. Tel.: þ98 914 4186490, þ98 411 3392413;
fax: þ 98 411 3363377.
E-mail addresses: h-ahmadi@tabrizu.ac.ir (H. Ahmadi),
lotfollahi@tabrizu.ac.ir (M.A. Lotfollahi-Yaghin),
aminfar@tabrizu.ac.ir (M.H. Aminfar).
1
Tel.: þ98 912 5427829.
2
Tel.: þ98 914 3112813.
tubular joints. The SCF is the ratio of the local surface stress to
the nominal direct stress in the brace. The SCF value depends on
joint geometry, loading type, weld size and type, and the location
around the weld under consideration. Geometric stresses S0 are
calculated at various locations around the welds and the max-
imum geometric stress is the hot-spot stress S. The fatigue life of
the joint is estimated through an appropriate S–N fatigue curve, N
being the number of load cycles.
Over the past thirty years, significant effort has been devoted
to the study of SCFs in various uni-planar tubular joints (i.e. joints
where the axes of the chord and the braces lay in the same plane).
As a result, many parametric design equations (formulas) in terms
of the joint geometrical parameters have been proposed, provid-
ing SCF values at certain locations adjacent to the weld for several
loading conditions. The reader is referred for example to [1–9]
(for SCF calculation at the saddle and crown positions of simple
uni-planar T-, Y-, X-, K-, and KT-joints), [10–12] (for SCF determi-
nation in uni-planar overlapped tubular joints), [13–21] (for the
study of SCF distribution along the weld toe of various uni-planar
joints), and [22–25] (for other aspects of stress concentration
analysis such as weld and crack modeling).
Multi-planar joints (Fig. 1) are an intrinsic feature of offshore
tubular structures. The multi-planar effect plays an important
role in the stress distribution at the brace-to-chord intersection
areas of the spatial tubular joints. For such multi-planar connec-
tions, the parametric stress formulas of simple uni-planar tubular
joints are not applicable in SCF prediction. Nevertheless, for
multi-planar joints which cover the majority of practical applica-
tions, very few investigations have been reported due to the

0263-8231/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tws.2012.04.011
68 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78
Fig. 1. Geometrical notation for a three-planar right-angle tubular KT-joint.
complexity and high cost involved. Following paragraph reviews
the research works currently available in the literature.
Karamanos et al. [26] proposed a set of parametric equations
to determine the SCFs for multi-planar welded CHS XX-connec-
tions. In this study, weld profile was modeled using 20-node solid
elements while 8-node shell elements were used to model the
chord and braces. This research covered the various loading
modes including reference and carry-over loadings. Chiew et al.
[27] studied the stress concentrations in DX-joints due to axial
loads. Chiew et al. [28] developed a set of design formulas to
determine the SCFs for multi-planar tubular XX-joints under
axial, IPB, and OPB loadings. Van Wingerde et al. [29] presented
a set of equations and graphs to predict the SCFs for multi-planar
KK-joints. The aim of this study was to simplify the equations for
design purposes. Karamanos et al. [30] proposed SCF equations
for multi-planar welded tubular DT-joints. In this study, the chord
member’s bending effects were included. Woghiren and Brennan
[31] developed a set of parametric formulas to predict the values
of SCF in multi-planar rack-stiffened tubular KK-joints. Lotfollahi-
Yaghin and Ahmadi [32] and Ahmadi et al. [33] proposed para-
metric formulas for prediction of SCF distribution along the weld
toe of the outer (inclined) and central (vertical) braces of right-
angle two-planar DKT-joints under axial loads. Ahmadi and
Lotfollahi-Yaghin [34] developed a set of parametric equations
to determine the chord-side saddle SCFs of central braces in
three-planar tubular KT-joints. An experimental database of SCFs
for acrylic specimens of multi-planar K- and KT-joints has been
presented in the HSE OTH 91 353 report [35] prepared by Lloyd’s
Register of Shipping. This report covers only the value of SCF at
the chord saddle position. It can be concluded from this section
that in the case of multi-planar joints, the studied connection
types and load cases are very limited. Despite the frequent use of
three-planar CHS KT-joints in the design of offshore jacket
structures, no parametric formula is available for calculation of
SCF values in such tubular joints.
The value of SCF along the weld toe of a tubular joint is mainly
determined by the joint geometry under any specific loading
condition. In order to study the behavior of tubular joints and to
relate this behavior easily to the geometrical properties of the
joint, a set of dimensionless geometrical parameters has been
defined. Fig. 1 shows a right-angle three-planar tubular KT-joint
with the geometrical parameters (b, g, t, z, a, and aB) for chord
and brace diameters D and d, and the corresponding wall
thicknesses T and t. Henceforth in the present paper, as shown
in Fig. 1, inner and outer saddle positions on the outer braces of 01
and 901 planes will be referred by abbreviated expressions. For
example, (–X)-01-OB-IS refers to the inner saddle (IS) position on
the outer brace (OB) of the 01 plane which is located on the
negative side of the X axis. Right-angle three-planar KT-joints are
one of the most important joints in the six-legged jacket-type
structures since they connect the bracing members to the main
legs of the jacket template.
In the present paper, parametric stress analysis has been
carried out for 81 steel right-angle three-planar tubular KT-joints
under three different axial loading conditions. The analysis results
are used to present general remarks on the effect of geometrical
parameters including t (brace-to-chord thickness ratio), g (chord
wall slenderness ratio), b (brace-to-chord diameter ratio), and y
(outer brace-to-chord inclination angle) on the SCF values at the
inner and outer saddle positions on the outer (main) braces.
 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78 69

Under axial loads, the hot-spot stresses generally occur at saddle
positions. Hence, the present paper focuses only on the SCFs at
saddle positions. In order to investigate the multi-planar effect,
SCFs in three-planar joints are compared with SCFs in the two-
planar DKT- and uni-planar KT-joints having the same geome-
trical properties. The effect of loading condition is also studied.
Based on the three-planar KT-joint finite element (FE) models
which are verified against both experimental results published in
HSE OTH 354 report [8] and the predictions of Lloyd’s Register
(LR) equations [8], a complete set of SCF database is constructed
for three considered axial load cases at the inner and outer saddle
positions. The FE models cover a wide range of geometrical
parameters. Afterwards, through nonlinear regression analysis, a
new set of SCF parametric equations is established for the fatigue
design of three-planar KT-joints under axial loads. An assessment
study of these equations is conducted against the experimental
data provided in the HSE OTH 91 353 report [35] and the
satisfaction of acceptance criteria proposed by the UK Depart-
ment of Energy (DoE) [36] is checked.
2. Geometrically parametric study of saddle SCFs on the outer
braces
2.1. FE modeling, analysis, and verification
The finite element method is adopted in the present research
for the numerical simulation of the stress distribution in the
considered tubular joints. To investigate the stress concentration
in three-planar tubular KT-joints, 81 models are generated and
analyzed using the multi-purpose FEM based software package,
ANSYS [37]. The aim is to study the effect of dimensionless
geometrical parameters on the saddle SCFs (on the chord side)
of the main (outer) braces. Different values assigned to each non-
dimensional parameter are as follows: b ¼0.3, 0.4, 0.5; g ¼12, 18,
24; t ¼0.3, 0.6, 0.9; y ¼301, 451, 601. These values cover the
practical range of the normalized parameters typically found in
multi-planar tubular joints of offshore structures:
0:3 r b r0:5
12 r g r 24
0:3 r t r 0:9
301r y r601 ð1Þ
The relative gap (z ¼g/D) has no considerable effect on the SCF
values in the tubular KT-joints under axial loads [20]. Hence, a
typical value of z ¼0.2 is assigned for all joints.
As shown in Fig. 2, three different axial loading conditions are
considered in the present research to study the SCFs in three-
planar KT-joints. In the 1st loading condition, all six braces
located on the 01 and 1801 planes are subjected to tensile loads
while the ones on the 901 plane are under compressive axial
loading (Fig. 2a). In the 2nd loading condition, all six braces
located on the 01 and 1801 planes are subjected to tensile loads
while the balanced axial loads are applied to the outer braces
located on the 901 plane (Fig. 2b). Under this loading condition,
the end of the central brace located on the 901 plane is assumed to
be fixed. In the 3rd loading condition, tensile loads are applied to
all nine bracing members (Fig. 2c). As shown in Fig. 2d, due to the
symmetry in geometry of the connection and loading condition,
only one fourth of the entire three-planar KT-joint is required to
be modeled under the 1st and the 3rd loading conditions.
However under the 2nd loading condition, as shown in Fig. 2e,
half of the entire joint should be modeled.

Fig. 2. (a)–(c): Three considered loading conditions; (d)–(f): geometrical models. (a) 1st loading condition: 0* and 180* planes: tensile axial loads 90* plane: compressive
axial loads; (b) 2nd loading condition: 0* and 180* planes: tensile axial loads 90* plane: balanced axial loads (c) 3rd loading condition: 0* and 180* planes: tensile axial
loads 90* plane: tensile axial loads; (d) One fourth of the entire joint required to be modeled under the 1st and 3rd loading conditions; (e) Half of the entire joint required
to be modeled under the 2nd loading condition; (f) The geometrical view along the X axis for all loading conditions.
70 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78
The welding size along the brace-to-chord intersection satis-
fies the AWS D 1.1 specifications [38]. As defined by AWS D 1.1,
the dihedral angle c is an important parameter used to determine
the weld thickness, and is the angle between the chord and brace
surface along the intersection curve as indicated in Fig. 3. The
details of weld profile according to AWS D 1.1 are shown in Fig. 4.
In this figure, L and tw refer to the length and thickness of the
weld, respectively. The dimensions of the weld quoted from AWS
D 1.1 [38] are listed in Table 1. Since it is difficult to model the
weld profile exactly due to the complicated shape of the brace-to-
chord intersecting curve, some simplifying techniques must be
used in order to make it possible to create a smooth and
adequately accurate weld profile. The techniques utilized by
Chiew et al. [27] and Lee [39] are used in the present research.
Fig. 3. Definition of the dihedral angle according to AWS D 1.1 [38].
Fig. 4. Details of the weld profile according to AWS D 1.1 [38].
It is shown at the end of this sub-section that the created weld
profile provides sufficient accuracy in predicting the SCF values
(see Table 2).
The entire tubular joint can be modeled by 3D brick ele-
ments. Using this type of element, the weld profile is simulated
as a sharp notch. This method will produce more accurate and
detailed stress distribution near the intersection in comparison
with a simple shell analysis. In the present study, ANSYS
element type SOLID95 is used to model the chord, brace and
weld profile. The element is defined by 20 nodes having three
degrees of freedom per node. These elements have compatible
displacements and are well suited to model curved boundaries.
A sub-zone mesh generation method is used during the FE
modeling, in order to guarantee the mesh quality. In this
method, the entire structure is divided into several different
zones according to the computational requirements. The mesh
of each zone is generated separately and then the mesh of the
entire structure is obtained by merging the meshes of all the
sub-zones. The mesh generated by this method is shown in
Fig. 5. The models are meshed in such a way that leads to a
compromise between the accuracy of results and the computer
analyzing time, software generated file volume, etc. To verify
the convergence of FE analysis, converging test is done and the
meshes with different densities are used in this test, before
generating the 81 models.
The chord end fixity conditions of tubular joints in offshore
structures may range from ‘‘almost fixed’’ to ‘‘almost pinned’’
with generally being closer to ‘‘almost fixed’’ [5]. In practice, value
of a in over 60% of tubular joints is in excess of 20 and is bigger
than 40 in 35% of the joints [7]. According to Morgan and Lee [13],
 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78 71

Table 1
Weld sizes specified by AWS D 1.1 [38].

Detail ID (see Figs. 3 and 4) Parameter

Root opening (R) (mm) Joint included angle (j) Complete weld

Max Min tw Lw

Detail A (C ¼ 1801  1351) 1.6–5 901 451 rt1 Z t1/sin C but r1.75t1
Detail B (C ¼1501  501) 1.6–6 901 for C r 1501 37.51; if unless use Detail C Zt1 for C Z 901 Z t1/sin C for C r 1501
Detail C (C ¼751  301) 3–13 401; if more use Detail B C/2 Zt1/sin C but r 1.75t1 Weld built up to meet
Detail D (C ¼ 401  151) 3–13 Z 2t1

Table 2
Verification of the FEA results using the experimental data and predictions of LR equations [8].

Joint typea D (mm) y a t g b Position Test LR Eqs. FEA e1b (%) e2b (%) 9e19–9e29 (%)

T 508 90 6.2 0.99 20.3 0.8 Saddle 11.4 10.54 11.26 8 1 þ7

Crown 5.4 3.92 4.6 27 15 þ 12

Y 508 45 6.2 1.05 20.3 0.8 Saddle 8.3 5.48 5.46 32 34 2

Crown 4.7 3.5 4.7 25 0 þ 25

Kc 508 45 12.6 1.0 20.3 0.5 Saddle 6.8 4.8 6.76 29.5 0.5 þ 29
Crown 4.6 4.56 4.8 1 -4 3

a
Project reference: JISSP.
b
e1¼ (Test  LR Eqs.)/Test; e2¼ (Test  FE)/Test.
c
z ¼ 0.15.

Fig. 5. Generated mesh using 3D brick elements and the sub-zone method. (a) The mesh generated using the sub-zone mesh generation method; (b) The generated mesh
near the intersection region; (c) The generated mesh for the weld profile.

changing the end restraint from fixed to pinned results in a
maximum increase of 15% in the SCF at the crown for a ¼6, and
this increase reduces to only 8% for a ¼8. In view of the fact that
the effect of chord end restraints is only significant for joints with
a o8 and high b and g values, which do not commonly occur in
practice, both chord ends are assumed to be fixed, with the
corresponding nodes restrained. Efthymiou [5] showed that
sufficiently long chord greater than six chord diameters (i.e.
a Z12) must be used to ensure that the stresses at the brace-to-
chord intersection are not affected by the end condition. Hence in
this study, a realistic value of a ¼16 was assigned for all the
models. The effect of brace length on SCF has been studied by
Chang and Dover [15]. It was concluded that there is no effect
when the ratio aB is greater than the critical value. In the present
study, in order to avoid the effect of short brace length, a realistic
value of aB ¼8 is selected for all joints.
Static numerical calculations of the linear elastic type are
appropriate to determine the SCFs in tubular joints [25]. This
type of analysis is used in the present study. Young’s modulus and
Poisson’s ratio are taken to be 207 GPa and 0.3, respectively. The
widely accepted conventional approach for fatigue strength
assessment of tubular joints is to use the geometric stresses at
the weld toe. According to IIW-XV-E [40], the peak stress is
calculated from extrapolating the geometrical stresses at two
points in a linear way to the weld toe position. The minimum and
maximum distances from the extrapolation region to the weld toe
for chord member are 0.4T and 1.4T respectively; where T is the
thickness of chord member (see Fig. 6). CIDECT 2000 [41] also
recommends the linear extrapolation but the recommended
minimum and maximum distances from the extrapolation region
to the weld toe are 0.4T and 1.0T respectively. According to
Karamanos et al. [9], both primary and principal stresses can be
72 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78

used for stress analysis of SCF calculations. In crown locations, the percentage of relative difference between the results of FE
primary and principal stresses coincide. Differences in chord model and experimental results. Hence, 9e19 9e29 indicates the
saddle locations are negligible, whereas some differences exist difference between the accuracy of LR equations and FE model.
in brace saddle and in-between locations. Generally, differences Positive sign for value of 9e19  9e29 means that the FE model
were found to be less than 10%. Although the stress component presented in this study is more accurate for predicting the values
perpendicular to the intersection is generally regarded as having a of SCF in comparison with LR equations. It can be concluded from
direct influence on fatigue crack growth, Burdekin et al. [42] and the comparison of FE results with experimental data and the
Soh and Soh [43] have demonstrated the validity of obtaining HSS values predicted by LR equations that the finite element model is
using the Maximum Absolute Principal Stress (MAPS). considered to be adequate to produce valid results.
The accuracy of the FEA predictions should be verified against
the experimental test results. As far as the authors are aware,
2.2. Results and discussion
there is no experimental database of SCFs for steel uni-planar and
multi-planar tubular KT-joints currently available in the litera-
This sub-section presents the results of numerical parametric
ture. In order to validate the finite element model, several related
study carried out to study the effect of dimensionless geometrical
geometries including T-, Y- and K-joints are modeled and the FE
parameters (b, g, t, and y) and the type of loading on the stress
results are validated against the LR equations [8] and test results
concentrations at the inner and outer saddle positions on the
published in HSE OTH 354 report [8]. The method of modeling the
outer braces of right-angle three-planar KT-joints. Multi-planar
members and the weld profile, the mesh generation procedure
effect is also investigated.
(including the selection of the element type), and the analysis
method are identical for the validating models and the considered
three-planar KT-joints. Hence, the conclusion of the verification 2.2.1. Effect of dimensionless geometrical parameters on saddle SCFs
of T-, Y-, and K-joints with the experimental test results can be Three charts are presented as an example in Fig. 7 showing the
used to validate the generated three-planar KT-joint models change of SCFs due to the change in the value of b and the
[20,21,32,33]. Verification results which are separately presented interaction of this parameter with the g. Geometrical parameters
at saddle and crown positions are summarized in Table 2. In this of the joint, the saddle position under consideration, and the
table, e1 denotes the percentage of relative difference between applied loading condition are given in the legend of each chart.
the predictions of LR equations and test results, and e2 denotes The parameter b is the ratio of brace diameter to chord diameter.
Hence, the increase of the b in models having constant value of
chord diameter leads to the increase of brace diameter.
Through investigating the effect of the b on SCFs, it can be
concluded that under the 2nd loading condition, increase of the b
leads to the decrease and increase of the SCF at the ( þX)-01-OB-IS
and (–X)-901-OB-IS positions, respectively (Fig. 7a and b); and
increase of the b leads to the decrease of the SCF at the ( þX)-01-
OB-IS position, under the 3rd loading condition (Fig. 7c). These
results are independent from values of the other dimensionless
geometrical parameters.
Nine charts are presented as an example in Fig. 8 showing the
change of SCFs due to the change in the value of the parameters t,
g, and y. Fig. 8a–c shows the change of SCFs due to the change in
the value of the t and the interaction of this parameter with the b.
The parameter t is the ratio of brace thickness to chord thickness
and the g is the ratio of radius to thickness of the chord. Hence,
the increase of the t in models having constant value of the g
leads to the increase of the brace thickness. Fig. 8d–f shows the
Fig. 6. Extrapolation procedure recommended by IIW-XV-E [40]. change of SCFs due to the change in the value of the g and the

6 6 8

7
Stress Concentration Factor

tor

5 5
Factor
Stress Concentration Fac

6
4 4
Stress Concentration

5
3 3 4

2 2 3
2
1 γ = 24 1 γ = 24 γ = 24
1
γ = 18 γ = 18 γ = 18
0 0 0
0.3 γ = 12 0.3 γ = 12 0.3 γ = 12
0.4 0.4 0.4
Value of β 0.5 Value of β 0.5 Value of β 0.5

Fig. 7. Effect of the parameter b on the SCFs at the outer saddle positions. (a) 2nd loading condition; ( þX)  01  OB-IS; t ¼ 0.3, y ¼451 (b) 2nd loading condition;
( X)  901 OB-IS; t ¼ 0.3, y ¼451 (c) 3rd loading condition; (þX)  01  OB-IS; t ¼0.3, y ¼ 601
 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78 73

6 2.5 7

5 6

tor
Stress Concentration Factor

Stress Concentration Factor

Stress Concentration Fac
4 5
1.5
4
3
1 3
2
2
0.5
1 β = 0.3 β = 0.3 β = 0.3
1
β = 0.4 β = 0.4 β = 0.4
0 0 0
0.3 β = 0.5 0.3 β = 0.5 β = 0.5
0.6 0.6 0.3
Value of 0.9 Value of 0.9 0.6
τ τ Value of 0.9
τ

25 6 35

30
tor
Stress Concentration Factor

Stress Concentration Factor

20
Stress Concentration Fac

25
4
15 20
3
15
10
2 10
5 τ = 0.9
τ = 0.9 1 τ = 0.9 5
τ = 0.6 τ = 0.6 τ = 0.6
0 0 0
τ = 0.3 τ = 0.3 12 τ = 0.3
12 12 18
18 18
Value of 24 Value of 24
Value of
γ 24 γ γ

20 9 8
18 8 7
tor
Stress Concentration Factor

16
Stress Concentration Factor

7
Stress Concentration Fac

6
14
6
12 5
5
10 4
4
8 3
6 3
2 2
4
β = 0.3 β = 0.3 1 β = 0.3
2 1
β = 0.4 β = 0.4 β = 0.4
0 0 0
30° β = 0.5 30° β = 0.5 30° β = 0.5
45° 45° 45°
Value of 60° Value of 60° Value of 60°
θ θ θ

Fig. 8. Effects of the parameters (a)–(c): t, (d)–(f): g, and (g)–(i): y on the SCFs at the outer saddle positions. (a) 3rd loading condition; ( þ X)  01 OB-IS; g ¼ 12, y ¼301(b)
3rd loading condition; (þX)  901  OB-IS; g ¼12, y ¼ 301 (c) 3rd loading condition; (þ X)  01 OB-IS; g ¼12, y ¼ 301 (d) 3rd loading condition; ( þX)  01  OB-IS; b ¼ 0.3,
y ¼451 (e) 3rd loading condition; (þ X)  901 OB-IS; b ¼ 0.3, y ¼ 451 (f) 3rd loading condition; (þX)  01  OB-OS; b ¼0.3, y ¼451 (g) 1st loading condition; (þX)  01  OB-IS;
g ¼ 24, t ¼0,3 (h) 2nd loading condition; (þ X)  01 OB-IS; g ¼ 24, t ¼0.3 (i) 3rd loading condition; (þX)  01  OB-IS; g ¼24, t ¼ 0.3

interaction of this parameter with the t. The parameter g is the parameters t and g on the saddle SCFs are considerably greater
ratio of radius to thickness of the chord. Hence, increase of the g than the effects of parameters b and y. This conclusion is not
in the models having constant value of chord diameter leads to dependent on either the applied loading condition or the saddle
decrease of chord thickness. Fig. 8g–i shows the change of SCFs at position under consideration.
( þX)-01-CB-IS position due to the change in the value of the y
(outer brace inclination angle) and the interaction of this para-
meter with the b under three considered loading conditions. 2.2.2. Comparison of SCFs: different saddle positions and loading
Through investigating the effect of the t, g, and y on SCFs, it conditions
can be concluded that under all three considered loading condi- Under the 1st loading condition, differences among the SCFs at
tions, the increase in the value of the parameter t leads to the the ( þX)-01-OB-OS, ( þX)-01-OB-IS, and ( þX)-901-OB-IS positions
increase of SCFs at all considered saddle positions. The same are small and the SCF at the ( þX)-901-OB-IS position is generally
conclusion was also drawn for the parameters g and y. Effects of the biggest one (Fig. 9a). The average and maximum differences
74 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78

35 20 18
18 16
30
Stress Concentration Factor

16

n Factor

Stress Concentration Factor

14
25 14
12
12

Stress Concentratio
20 10
10
15 8
8
10 6
6
4 4
5 τ = 0.9 τ = 0.9 τ = 0.9
2 2
τ = 0.6 τ = 0.6 τ = 0.6
0 0 0
0°-OB-OS τ = 0.3 0°-OB-OS τ = 0.3 0°-OB-OS τ = 0.3
0°-OB-IS 0°-OB-IS 0°-OB-IS
90°-OB-IS 90°-OB-IS 90°-OB-IS

5 9
4.5 8
4
Stress Concentration Factor
Factor

7
3.5
6
Stress Concentration

3
5
2.5
4
2
3
1.5
1 2
0°-OB-OS 0°-OB-OS
0.5 1
0°-OB-IS 0°-OB-IS
0 0
(–X) SCFs 90°-OB-IS 1st LC 90°-OB-IS
2nd LC
(+X) SCFs 3rd LC

Fig. 9. (a)–(c): Comparison of SCFs at different saddle positions; (d): comparison of (–X) and ( þ X) SCFs; (e): comparison of SCFs under three considered loading conditions
(LC stands for loading condition). (a) 1st loading condition; ( þ X) SCFs; b ¼ 0.5, g ¼ 18, y=451 (b) 2nd loading condition; (þX) SCFs; b ¼ 0.5, g ¼ 18, y=451 (c) 3rd loading
condition; (þ X) SCFs; b ¼ 0.5, g ¼ 18, y=451; (d) 2nd loading condition; b ¼0.4, g ¼18, y ¼ 451 (e) (þX) SCFs; b ¼ 0.4, g ¼ 18, y ¼ 0.3, y ¼ 451

among the SCFs at these three saddle positions are approximately The following relationship exists among the SCFs at a specific
6% and 15%, respectively. saddle position under three considered loading conditions
Under the 2nd loading condition, following relationships exist (Fig. 9e):
among the SCFs at six considered saddle positions (Fig. 9b): SCF1st LC 4 SCF2nd LC 4 SCF3rd LC ð4Þ
SCFð þ XÞ01OBOS 4SCFð þ XÞ01OBIS 4SCFð þ XÞ901OBIS
where LC denotes loading condition. At a specific saddle position,
SCFð2XÞ01OBOS 4SCFð2XÞ01OBIS 4 SCFð2XÞ901OBIS ð2Þ the difference between SCFs under the 2nd and 3rd loading
The SCF values at saddle positions located on the negative side of conditions is relatively small. However, under the 1st loading
the X axis [(–X) SCFs] are always bigger than the corresponding condition, the SCF at a specific saddle position is remarkably
SCFs on the positive side of the X axis [(þX) SCFs] (Fig. 9d). The bigger than the corresponding SCFs under the 2nd and 3rd
average and maximum differences between the SCF( þ X)-01-OB-OS loading conditions. Maximum and minimum differences between
and SCF(–X)-01-OB-OS are 1.7% and 30%, respectively. The average the 1st and 2nd loading conditions occur at the (þ X)-901-OB-IS
and maximum differences between the SCF( þ X)-01-OB-IS and and ( þX)-01-OB-OS positions, respectively.
SCF(–X)-01-OB-IS are 28% and 86%, respectively and the SCF(–X)-901-
OB-IS is 3.99 times bigger than the SCF( þ X)-901-OB-IS, in an average 2.2.3. Comparison of SCFs in uni-, two-, and three-planar joints
sense. Hence, it can be seen that the SCF( þ X)-901-OB-IS is remark- As can be seen in Fig. 10, highly remarkable differences exist
ably smaller than the SCFs at the other saddle positions. among the SCF values in a three-planar KT-joint and the correspond-
Under the 3rd loading condition, following relationship exists ing SCFs in the equivalent two-planar and uni-planar joints having
among the SCFs at three considered saddle positions (Fig. 9c): the same geometrical properties. It can be clearly concluded from
this observation that using the equations proposed for uni-planar
SCFð þ XÞ01OBOS 4SCFð þ XÞ01OBIS 4SCFð þ XÞ901OBIS ð3Þ
and two-planar connections to compute the SCFs in three-planar
The average difference between the SCF( þ X)-01-OB-OS and KT-joints will lead to considerably either under-predicting or over-
SCF( þ X)-01-OB-IS is about 50%. The SCF( þ X)-01-OB-OS and SCF( þ X)-01- predicting results. Hence it is necessary to develop SCF formulas
OB-IS are respectively 6 and 4.5 times bigger than the SCF( þ X)-901- specially designed for three-planar KT-joints. As shown in Fig. 10d,
OB-IS, in an average sense. Hence, it can be seen that the SCF at the under a combination of tensile and compressive axial loads, the SCF
saddle position of the 901 plane is remarkably smaller than the value at the outer saddle position on the outer brace of the 01 plane
SCFs at the saddle positions located on the 01 plane. in a three-planar KT-joint can be 60% and 150% bigger than the
 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78 75

(All 3 braces are loaded)

(All 3 braces are loaded)

90° plane

0° plane

(All 3 braces are loaded) (All 3 braces are loaded) (All 3 braces are loaded)

0°-OB-OS 0°-OB-IS 90°-OB-IS 0°-OB-OS 0°-OB-IS 90°-OB-IS

3-Planar Joint 3-Planar Joint

2-Planar Joint 2-Planar Joint

Uni-Planar Joint Uni-Planar Joint

0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
Stress Concentration Factor Stress Concentration Factor

Fig. 10. Comparison of saddle SCFs on the outer braces in three-, two-, and uni-planar KT-joints; b ¼0.4, g ¼12, t ¼0.6, y ¼301. (a) Two-planner joint: 01 plane: tensile axial
loads 901 plane: compressive axial loads (b) Two-planner joint: 01 plane: tensile axial loads 901 plane: tensile axial loads (c) Uni-planner joint: 01 plane: tensile axial loads
(d) Loading conditions: Three-planner joint: Fig. 2a Two-planner joint: Fig. 10a Uni-planner joint: Fig. 10c (e) Loading conditions: Three-planner joint: Fig. 2c Two-planner
joint Fig. 10b Uni-planner joint: Fig. 10c

corresponding SCFs in the equivalent two-planar and uni-planar KT- input data which is imported as a matrix. Each row of this matrix
joints, respectively. Under the tensile axial loads, as can be seen in involves the information about the SCF value at a specific position
Fig. 10e, the three-planar SCF at the outer saddle position of the 01 in a three-planar tubular KT-joint having specific geometrical
plane can be 2.3 and 1.4 times bigger than the corresponding SCFs in properties. The number of rows and columns of input matrix for
the equivalent two-planar and uni-planar joints, respectively. How- each equation are 81 (number of the joints) and 5 (number of
ever, the two-planar SCF at the inner saddle position of the 901 plane variables), respectively. Hence, the whole FEM SCF database is
is 80% bigger than the corresponding SCF in the equivalent three- arranged as twelve 81  5 input matrices. When the dependent
planar joint. Such observations highlight the necessity of proposing (i.e. SCF) and independent (i.e. b, g, t, and y) variables are defined,
individual parametric equations for any specific saddle position a model expression must be built with defined parameters.
under each loading condition. Parameters of the model expression are unknown coefficients
and exponents. The researcher must specify a starting value for
each parameter, preferably as close as possible to the expected
3. Developing the SCF parametric formulas for outer brace final solution. Poor starting values can result in failure to
saddles converge or in convergence on a solution that is local (rather
than global) or is physically impossible. Various model expres-
Despite the successful utilization of the finite element method sions must be built to derive a parametric equation having a high
to analyze the tubular joints, the extensive use of such a coefficient of correlation.
numerical method is not feasible in a normal day-to-day design
office operation. Instead, parametric design equations expressed
in the form of dimensionless geometrical parameters are useful 3.2. Proposed equations
and desirable for fatigue design. In the present study, 12 indivi-
dual parametric equations are proposed for the calculation of SCFs After performing a large number of nonlinear analyses, follow-
at the inner and outer saddle positions on the outer (main) braces ing parametric equations are proposed for predicting the chord-
of the 01 and 901 planes in a right-angle three-planar KT-joint side SCFs at inner and outer saddle positions on the outer (main)
under three considered axial loading conditions. braces of 01 and 901 planes in a right-angle three-planar KT-joint
under three considered axial loading conditions:
1st loading condition:
3.1. Nonlinear regression analysis
(þ X)-01-OB-IS position:

Geometrically parametric SCF design equations are derived SCF ¼ fð0:8740:834t0:755b2:495tbÞ

based on multiple nonlinear regression analyses performed by the þ ½ð2:910t0:839 þ0:557ÞUð0:788g1:512 3:953Þ
statistical software package, SPSS. Values of dependent variable
1:324 2
(i.e. SCF) and independent variables (b, g, t, and y) constitute the ð0:600b 0:254ÞUð1:737b þ 1:096b þ 0:118Þg
76 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78

0:525
ð1þ 0:971y 1:635Cos y þ 2:472 Sin yÞ; R2 ¼ 0:993: ( þX)-901-OB-IS position:
ð5Þ SCF ¼ fð0:6080:906t1:381b þ 3:513tbÞ
( þX)-901-OB-IS position: þ ½ð1:247t0:710 0:358ÞUð0:422g1:859 28:260Þ
0:155 2
SCF ¼ fð0:003 þ 3:343t0:043b1:848tbÞ ð1:282b 0:217ÞUð1:143b þ0:924b0:190Þg
1:474
þ ½ð2:152t0:781 þ 0:491ÞUð0:348g1:564 10:631Þ ð1 þ 0:882y þ 0:121 Cos y þ 0:949 Sin yÞ; R2 ¼ 0:865
0:732 2 ð15Þ
ð0:823b 0:272ÞUð1:710b þ 1:071b þ0:193Þg
0:673 2
ð1þ 0:565y 1:629 Cos y þ4:249 Sin yÞ; R ¼ 0:990 ( þX)-01-OB-OS position:
ð6Þ SCF ¼ fð0:527 þ 0:583t0:320b2:886tbÞ
( þX)-01-OB-OS position: þ ½ð1:572t0:860 þ0:304ÞUð0:556g1:568 6:721Þ
0:382 2
SCF ¼ fð0:7930:541t0:570b4:615tbÞ ð0:451b 0:143ÞUð1:337b þ 0:909b þ 0:209Þg
0:492
þ ½ð1:027t0:875 0:184ÞUð0:979g1:521 þ 0:250Þ ð1 þ 0:581y 1:521 Cos y þ 2:487 Sin yÞ; R2 ¼ 0:993
ð0:931b
1:224 2
0:117ÞUð1:100b 0:240b0:562Þg ð16Þ
0:511 2 2
ð1þ 2:052y 2:030 Cos y þ2:250 Sin yÞ; R ¼ 0:994: In Eqs. (5)–16), R denotes the coefficient of correlation and y
ð7Þ should be inserted in radians.
2nd loading condition:
( þX)-01-OB-IS position: 3.3. Assessment according to UK DoE [36] acceptance criteria

SCF ¼ fð1:2037:292t1:248b þ 7:543tbÞ The UK Department of Energy (DoE) [36] recommends the
þ ½ð1:510t0:982 0:240ÞUð0:536g0:960 þ19:977Þ following assessment criteria regarding the applicability of the
0:590 2 commonly used SCF parametric equations (P/R stands for the ratio
ð0:463b þ 0:484ÞUð0:835b þ 0:727b þ0:416Þg
1:070 of the predicted SCF from a given equation to the recorded SCF
ð1þ 11:042y 4:586 Cos y þ15:128 Sin yÞ; R2 ¼ 0:991:
from test or analysis):
ð8Þ
( þX)-901-OB-IS position:  For a given dataset, if % SCFs under-predicting r25%, i.e. [%P/
R r1.0]r25%, and if % SCFs considerably under-predicting
SCF ¼ fð0:3081:091t0:828b þ 4:044tbÞ
r5%, i.e. [%P/Rr0.8]r5%, then accept the equation. If, in
0:650 3:498
þ ½ð1:330t 0:443ÞUð0:009g 51:200Þ addition, the percentage SCFs considerably over-predicting
0:310 2
Uð0:775b þ 0:598ÞUð1:889b þ0:976b þ 0:068Þg r50%, i.e. [%P/R 41.5]Z50%, then the equation is regarded
Uð1 þ 2:829y
1:599
þ0:531 Cos y þ 2:155 Sin yÞ; R2 ¼ 0:908: as generally conservative.
 If the acceptance criteria is nearly met i.e. 25%r[%P/
ð9Þ
R r1.0]r30%, and/or 5% r[%P/Rr0.8]r7.5%, then the equa-
(–X)-01-OB-IS position: tion is regarded as borderline and engineering judgment must
0:140 1:400 1:182 1:127 be used to determine acceptance or rejection.
SCF ¼ 0:478b g t y ; R2 ¼ 0:987 ð10Þ  Otherwise reject the equation as it is too optimistic.
(–X)-901-OB-IS position:
1:141 1:103 1:155 1:122 In view of the fact that for a mean fit equation, there is always
SCF ¼ 2:310b g t y ; R2 ¼ 0:957 ð11Þ
a large percentage of under-prediction, the requirement for joint
( þX)-01-OB-OS position: under-prediction, i.e. P/Ro1.0, can be completely removed in the
SCF ¼ fð1:0106:128t þ 0:101b3:304tbÞ assessment of parametric equations [44]. Assessment results
according to the UK DoE criteria are tabulated in Table 3.
þ ½ð1:691t0:935 þ 0:256ÞUð0:287g1:241 þ 9:172Þ
As can be seen in Table 3, Eqs. (9), (11), and (15) require
0:184 2
ð0:696b þ 0:034ÞUð0:605b þ 0:854b þ0:229Þg revision to satisfy the recommended criteria. In order to revise
0:953 these equations, SCF values obtained from each of these equations
ð1þ 2:192y 2:147 Cos y þ 5:758 Sin yÞ; R2 ¼ 0:993
are multiplied by a coefficient in such a way that the resulted SCF
ð12Þ
set satisfies the UK DoE criteria. This idea can be expressed as
(–X)-01-OB-OS position: follows:
SCF ¼ fð1:99512:518t þ0:202b3:752tbÞ Design Factor ¼ SCF Design =SCF Equation ð17Þ

þ ½ ð1:214t0:952 0:187ÞUð0:586g1:302 þ 45:750Þ
where values of SCFEquation are calculated from Eqs. (9), (11), and
0:234 2
ð0:579b þ 0:787ÞUð0:256b þ 1:051b þ 0:335Þg (15) and the values of SCFDesign are expected to satisfy the UK DoE
0:821 acceptance criteria.
ð1þ 0:846y 1:957 Cos y þ6:196 Sin yÞ; R2 ¼ 0:991
Multiple comparative analyses were carried out to determine
ð13Þ
the optimum value of design factor for each equation. The results
3rd loading condition: showed that the optimum design factors for Eqs. (9), (11), and
( þX)-01-OB-IS position: (15) are 1.35, 1.09, and 1.30, respectively. Hence, the following
equations should be used for design purposes:
SCF ¼ fð0:359 þ 1:056t0:146b2:163tbÞ
þ ½ð1:150t0:817 0:268ÞUð0:911g1:556 17:056Þ SCF Design ¼ 1:35  SCF Eq: ð9Þ ð18Þ
1:313 2
ð0:651b þ 0:525ÞUð1:144b þ 0:980b þ 0:041Þg SCF Design ¼ 1:09  SCF Eq: ð11Þ ð19Þ
0:617
ð1þ 1:183y 1:959 Cos y þ 2:809 Sin yÞ; R2 ¼ 0:991
ð14Þ SCF Design ¼ 1:30  SCF Eq: ð15Þ ð20Þ
 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78 77

Table 3
Results of equation assessment according to UK DoE [36] acceptance criteria.

Load case Saddle position Equation Conditions Overall status

%P/Ro 0.8 %P/R41.5

1st Loading condition (þ X)-01-OB-IS Eq. (5) 4.9% o 5% OK 0%o 50% OK Accepted
(þ X)-901-OB-IS Eq. (6) 3.7% o 5% OK 0%o 50% OK Accepted
(þ X)-01-OB-OS Eq. (7) 2.5% o 5% OK 0%o 50% OK Accepted

2nd Loading condition (þ X)-01-OB-IS Eq. (8) 4.9% o 5% OK 0%o 50% OK Accepted
(þ X)-901-OB-IS Eq. (9) 18.5% 45% 1.2% o50% OK Requires revision
(–X)-01-OB-IS Eq. (10) 4.9% o 5% OK 0%o 50% OK Accepted
(–X)-901-OB-IS Eq. (11) 12% 45% 0%o 50% OK Requires revision
(þ X)-01-OB-OS Eq. (12) 4.9% o 5% OK 0%o 50% OK Accepted
(–X)-01-OB-OS Eq. (13) 2.5% o 5% OK 0%o 50% OK Accepted

3rd Loading condition (þ X)-01-OB-IS Eq. (14) 3.7% o 5% OK 0%o 50% OK Accepted
(þ X)-901-OB-IS Eq. (15) 23% 45% 8.6% o50% OK Requires revision
(þ X)-01-OB-OS Eq. (16) 0%4 5% OK 0%o 50% OK Accepted

Table 4
Results of validating the proposed equations at the inner and outer saddle positions using the data from a strain gauged acrylic model test.

Geometrical parameters SCF value at the saddle position

Experimental SCF Revised experimental SCFa Equation SCF Differenceb (%)

D ¼ 150 mm, y ¼45, t ¼0.6, b ¼0.5, g ¼ 12 10.90 10.35 9.13 (Eq. (7)) 13.4
D ¼ 150 mm, y ¼45, t ¼0.6, b ¼0.5, g ¼ 24 18.88 17.94 19.35 (Eq. (12)) 7.8

a
Revised experimental SCF ¼ experimental SCF  0.95 (to simulate a weld fillet; according to HSE OTH 354 [8]).
b
Difference¼ (Revised experimental SCF/Equation SCF)  1.0.

For other parametric equations, the design factor equals
to 1.00.
3.4. Verification against the experimental data
Results of validating the proposed equations against the data
from a strain gauged acrylic model test are presented in Table 4.
that the effect of parameters t and g on the outer saddle SCFs is
considerably greater than the effect of parameters b and y.
Under the 1st loading condition, differences among the SCFs at
( þX)-01-OB-OS, ( þX)-01-OB-IS, and ( þX)-901-OB-IS positions are
small and the SCF at the ( þX)-901-OB-IS position is generally the
biggest one. Under the 2nd and 3rd loading conditions, the SCF01-
OB-OS is always bigger than the SCF01-OB-IS which is itself always

The source of the experimental data is the HSE OTH 91 353 report
prepared by Lloyd’s Register of Shipping [35] in which a compre-
hensive experimental database of SCFs for acrylic complex joints
including multi-planar and overlapped K- and KT-joints has been
presented. This report covers only the value of SCF at the chord-
side saddle position. It must be noted that since the weld profile is
not included in an acrylic specimen, the SCFs extracted from
acrylic model tests should be reduced by a factor of 0.95 in order
to obtain the realistic values expected to be found in steel tubular
joints [8]. As can be seen in Table 4, there is a good agreement
between the predictions of proposed equations and the experi-
mental measurements.
4. Conclusions
Effects of dimensionless geometrical parameters on the SCF
values at different outer brace saddle positions of axially loaded
right-angle three-planar KT-joints were investigated and a new set
of SCF parametric equations was established for the fatigue design
of three-planar KT-joints under three different axial load cases.
Detailed and quantitative results of parametric study which were
extensively discussed in the text are not repeated here for the sake
of brevity and only a summary of general remarks is presented:
Under all three considered loading conditions, the increase in
the value of the parameter t leads to the increase of outer brace
SCFs at all six considered saddle positions. The same conclusion
was also drawn for the parameters g and y. It was also concluded
bigger than the SCF901-OB-IS.
Among the SCFs at a specific saddle position, the SCFs under
the 1st and 3rd loading conditions are always the biggest and
smallest ones, respectively; and the difference between SCFs
under the 2nd and 3rd loading conditions is relatively small.
However, under the 1st loading condition, the SCF at a specific
saddle position is remarkably bigger than the corresponding SCFs
under the 2nd and 3rd loading conditions.
Highly remarkable differences exist among the SCF values in a
three-planar KT-joint and the corresponding SCFs in the equivalent
two-planar and uni-planar joints having the same geometrical
properties. It can be clearly concluded from this observation that
using the equations proposed for uni-planar and two-planar con-
nections to compute the SCFs in three-planar KT-joints will lead to
considerably either under-predicting or over-predicting results.
Hence it is necessary to develop SCF formulas specially designed
for three-planar KT-joints. Good results of equation assessment
according to UK DoE acceptance criteria, high values of correlation
coefficients, and the good agreement between the predictions of
proposed equations and the experimental data guarantee the
accuracy of equations. Hence, twelve developed equations can
reliably be used for fatigue design of offshore structures.
Acknowledgments
The authors gratefully acknowledge the useful comments of
anonymous reviewers on the draft version of this paper.
78 H. Ahmadi et al. / Thin-Walled Structures 58 (2012) 67–78
References
[1] Kuang JG, Potvin AB, Leick RD. Stress concentration in tubular joints. In:
Proceedings of the offshore technology conference. Paper OTC 2205; 1975.
[2] Wordsworth AC, Smedley GP. Stress concentrations at unstiffened tubular
joints. In: Proceedings of the European offshore steels research seminar.
[3] Wordsworth AC. Stress concentration factors at K and KT tubular joint.
In: Proceedings of the conference on fatigue of offshore structural steels;
1981. p. 59–69.
[4] Efthymiou M, Durkin S. Stress concentrations in T/Y and gap/overlap K-joints.
In: Proceedings of the conference on behavior of offshore structures; 1985.
p. 429–40.
[5] Efthymiou M. Development of SCF formulae and generalized influence
functions for use in fatigue analysis. In: Proceedings of the OTJ 88; 1988.
[6] Hellier AK, Connolly MP, Dover WD. Stress concentration factors for tubular
Y- and T-joints. International Journal of Fatigue 1990;12:13–23.
[7] Smedley P, Fisher P. Stress concentration factors for simple tubular joints.
In: Proceedings of the international offshore and polar engineering con-
ference; 1991.
[8] UK Health and Safety Executive. OTH 354. Stress concentration factors for
simple tubular joints-assessment of existing and development of new
parametric formulae. Prepared by Lloyd’s Register of Shipping, UK; 1997.
[9] Karamanos SA, Romeijn A, Wardenier J. Stress concentrations in tubular gap
K-joints: mechanics and fatigue design. Engineering Structures 2000;
22:4–14.
[10] Gho WM, Gao F. Parametric equations for stress concentration factors in
completely overlapped tubular K (N)-joints. Journal of Constructional Steel
Research 2004;60:1761–82.
[11] Gao F. Stress and strain concentrations of completely overlapped tubular
joints under lap brace OPB load. Thin-Walled Structures 2006;44:861–71.
[12] Gao F, Shao YB, Gho WM. Stress and strain concentration factors of
completely overlapped tubular joints under lap brace IPB load. Journal of
Constructional Steel Research 2007;63:305–16.
[13] Morgan MR, Lee MMK. Prediction of stress concentrations and degrees of
bending in axially loaded tubular K-joints. Journal of Constructional Steel
Research 1997;45:67–97.
[14] Morgan MR, Lee MMK. Parametric equations for distributions of stress
concentration factors in tubular K-joints under out-of-plane moment load-
ing. International Journal of Fatigue 1998;20:449–61.
[15] Chang E, Dover WD. Parametric equations to predict stress distributions
along the intersection of tubular X and DT-joints. International Journal of
Fatigue 1999;21:619–35.
[16] Shao YB. Proposed equations of stress concentration factor (SCF) for gap
tubular K-joints subjected to bending load. International Journal of Space
Structures 2004;19:137–47.
[17] Ahmadi H, Lotfollahi-Yaghin MA. Geometrical effect on the stress distribu-
tion along the weld toe for tubular KT-joints under balanced axial loading. In:
Proceeding of the 8th International conference on coasts, ports and marine
structures (ICOPMAS); 2008.
[18] Shao YB, Du ZF, Lie ST. Prediction of hot spot stress distribution for tubular
K-joints under basic loadings. Journal of Constructional Steel Research 2009;
65:2011–26.
[19] Lotfollahi-Yaghin MA, Ahmadi H. Numerical parametric study of stress
concentration along the intersection of tubular KT-joints subjected to
balanced axial loading. In: Proceedings of the 19th international offshore
and polar engineering conference (ISOPE); 2009.
[20] Lotfollahi-Yaghin MA, Ahmadi H. Effect of geometrical parameters on SCF
distribution along the weld toe of tubular KT-joints under balanced axial
loads. International Journal of Fatigue 2010;32:703–19.
[21] Ahmadi H, Lotfollahi-Yaghin MA, Aminfar MH. Geometrical effect on SCF
distribution in uni-planar tubular DKT-joints under axial loads. Journal of
Constructional Steel Research 2011;67:1282–91.
[22] Lee MMK. Strength, stress and fracture analyses of offshore tubular joints
using finite elements. Journal of Constructional Steel Research 1999;51:
265–86.
[23] Lie ST, Lee CK, Wong SM. Modeling and mesh generation of weld profile in
tubular Y-joint. Journal of Constructional Steel Research 2001;57:547–67.
[24] Lie ST, Lee CK, Chiew SP, Shao YB. Mesh modelling and analysis of cracked
uni-planar tubular K-joints. Journal of Constructional Steel Research
2005;61:235–65.
[25] N’Diaye A, Hariri S, Pluvinage G, Azari Z. Stress concentration factor analysis
for notched welded tubular T-joints. International Journal of Fatigue
2007;29:1554–70.
[26] Karamanos SA, Romeijn A, Wardenier J. Stress concentrations in multi-planar
welded CHS XX-connections. Journal of Constructional Steel Research
1999;50:259–82.
[27] Chiew SP, Soh CK, Fung TC, Soh AK. Numerical study of multiplanar tubular
DX-joints subject to axial loads. Computers and Structures 1999;72:746–61.
[28] Chiew SP, Soh CK, Wu NW. General SCF design equations for steel multi-
planar tubular XX-joints. International Journal of Fatigue 2000;22:283–93.
[29] Wingerde AM, Packer JA, Wardenier J. Simplified SCF formulae and graphs for
CHS and RHS K- and KK-connections. Journal of Constructional Steel Research
2001;57:221–52.
[30] Karamanos SA, Romeijn A, Wardenier J. SCF equations in multi-planar welded
tubular DT-joints including bending effects. Marine Structures 2002;15:
157–73.
[31] Woghiren CO, Brennan FP. Weld toe stress concentrations in multi planar
stiffened tubular KK Joints. International Journal of Fatigue 2009;31:164–72.
[32] Lotfollahi-Yaghin MA, Ahmadi H. Geometric stress distribution along the
weld toe of the outer brace in two-planar tubular DKT-joints: parametric
study and deriving the SCF design equations. Marine Structures 2011;24:
239–60.
[33] Ahmadi H, Lotfollahi-Yaghin MA, Aminfar MH. Distribution of weld toe stress
concentration factors on the central brace in two-planar CHS DKT-connec-
tions of steel offshore structures. Thin-Walled Structures 2011;49:1225–36.
[34] Ahmadi H, Lotfollahi-Yaghin MA. Geometrically parametric study of central
brace SCFs in offshore three-planar tubular KT-joints. Journal of Construc-
tional Steel Research 2012;71:149–61.
[35] UK Health and Safety Executive. OTH 91 353. Stress concentration factors for
tubular complex joints. Prepared by Lloyd’s Register of Shipping, UK; 1992.
[36] London, UK: HMSO; 1983.
[37] Swanson Analysis Systems Inc. ANSYS user’s manual; 2007.
[38] American Welding Society (AWS). Structural welding code: AWS D 1.1. USA;
2002.
[39] Lee MMK. Estimation of stress concentrations in single-sided welds in
offshore tubular joints. International Journal of Fatigue 1999;21:895–908.
[40] France: International Institute of Welding; 1999.
[41] Zhao XL, Herion S, Packer JA, Puthli R, Sedlacek G, Wardenier J. Design guide
for circular and rectangular hollow section joints under fatigue loading.
CIDECT Publication, no. 8. Germany: TUV Verlag; 2000.
[42] Burdekin FM, Cheaitani MJ, Thurlbeck SD. Fatigue, fracture and plastic
collapse of offshore tubular joints using BSI PD 6493:1991. In: Proceedings
of the 14th offshore mechanics and arctic engineering conference; 1995.
p. 41–54.
[43] Soh AK, Soh CK. Stress concentrations of K tubular joints subjected to basic
and combined loadings. In: Structures and Buildings, Proceedings of the
Institution of Civil Engineers; 1996. p. 19–28.
[44] Bomel Consulting Engineers. Assessment of SCF equations using Shell/KSEPL
finite element data. C5970R02.01 REV C; 1994.