Professional Documents
Culture Documents
www.elsevier.com/locate/jcsr
Abstract
This paper presents parametric equations to predict the stress concentration factors (SCF)
of completely overlapped tubular K(N)-joints under lap brace axial compression. 192 finite
element (FE) models of the joint have been created for the parametric study using a com-
mercial FE package, MARC. In the analysis, both 8-node thick shell and 20-node solid ele-
ments are found suitable for modelling the joint. The comparison of strain concentration
factors (SNCF) between the FE models and the tested specimen shows that the difference is
less than 0.42. However, in view of the computational effort, the FE model with 8-node
thick shell element is used for the parametric analysis. The study reveals that the effect of
through brace to chord wall thickness ratio on SCF of the joint is most crucial. The
maximum SCF occurs on the through brace saddle near the lap brace. The gap size on the
through brace surface between the outer surfaces of the chord and the lap brace significantly
affects the SCF. Based on the FE data, a set of parametric equations is derived to predict the
SCF of the joint. The reliability of the equations has been verified against the requirement of
the Fatigue Guidance Review Panel.
# 2004 Elsevier Ltd. All rights reserved.
Keywords: Completely overlapped tubular joint; Stress concentration factor; Strain concentration factor
Abbreviations: FE, Finite element; SCF, Stress concentration factor; SNCF, Strain concentration factor.
Corresponding author. Tel.: +65 67906391; fax: +65 67921650
E-mail address: CWMGho@ntu.edu.sg (W.-M. Gho).
0143-974X/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2004.05.003
1762 W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782
Nomenclature
L Chord length
lT Through brace length
lL Lap brace length
D Chord diameter
dT Through brace diameter
dL Lap brace diameter
T Chord wall thickness
tT Through brace wall thickness
tL Lap brace wall thickness
gT Gap size
aC Chord length parameter ð2L=DÞ
aT Through brace length parameter (2lT =dT )
bCT Through brace to chord diameter ratio (dT =D)
bTL Lap brace to through brace diameter ratio (dL =dT )
cC Chord radius to wall thickness ratio (D=2T)
cT Through brace radius to wall thickness ratio (dT =2tT )
sCT Through brace to chord wall thickness ratio (tT =T)
sTL Lap brace to through brace wall thickness ratio (tL =tT )
n Gap size to chord diameter ratio (gT =D)
E Young’s modulus
m Poisson’s ratio
1. Introduction
lapped joints [1]. Since the loads are predominantly transferred between the braces,
the punching shear stresses on the chord wall can be reduced. Additionally, the
joint configuration can also reduce the cost significantly without the use of tempor-
ary members to support the braces during fabrication of offshore steel jackets [2].
The recent experimental study of a full-scale completely overlapped tubular K(N)-
joint at Nanyang Technological University revealed that the joint could have
strength equivalent to a simple gap K-joint [3]. A good ductility performance of
the joint under lap brace axial compression was obtained. On cyclic behaviour, the
failure location of the joint shifted from the chord to the through brace [4] which
indicated that the use of completely overlapped joints might be able to prevent the
structure from complete collapse under extreme load conditions.
Owing to the complexity of geometry and stress distributions, the understanding
of the behaviour of completely overlapped joints is very limited. No parametric
equations are currently available for design of the joint. In API RP2A [5], a mini-
mum stress concentration factor (SCF) of six is recommended for brace members
of overlapped joints. In CIDECT [6], the SCF of simple gap joints and partially
overlapped joints are specified, but not of completely overlapped joints.
In the present work, the SCF of the completely overlapped K(N)-joint under lap
brace axial compression is studied. The load condition is unbalanced as the axial
1764 W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782
load of the lap brace is distributed among members at the joint according to their
stiffness. Under this load condition, the lap brace is subjected to axial compression
while the through brace is in tension. Here, the definition of load condition is dif-
ferent from that of simple gap K-joints where a balanced load case is considered if
the loads of the two braces are of the same magnitudes and the vertical load com-
ponents are in opposite directions. On the other hand, an unbalanced load con-
dition is referred to those joints with both braces subjected to loads in compression
or tension [7]. The geometrical properties of the completely overlapped joint can be
seen in Fig. 1. In order to have a more reliable analysis of the joint in the current
study, the through brace to chord diameter ratio is limited to 0.8. The reason is
that the joints with through brace to chord diameter ratio above 0.8 have shown
different behaviour to those with low and medium diameter ratios [8]. Further-
more, the only available experimental result for the verification of the finite
element (FE) models has a through brace to chord diameter ratio of 0.6 [9].
In the FE analysis, both 8-node thick shell and 20-node solid elements are exam-
ined in detail to determine the suitable element for modelling the joint. A commer-
cial FE package MARC [10] is used to perform the analysis. Besides the
comparison of SCF between the two FE models, the results from the FE analysis
are also compared to those from the test. For the parametric study, 192 FE models
of the joint subjected to lap brace axial compression have been created with geo-
metrical parameters within the range of practical limits. Based on the FE data, a
set of parametric equations is derived to predict the SCF of the joint at the crown
toe, saddle and crown heel of members. The reliability of these equations has been
verified against the requirement of the Fatigue Guidance Review Panel.
2. Literature review
be considered as a chord member in the analysis. But this has been proven to be
inappropriate in the recent study on the behaviour of K(N)-joint by Gho et al. [9].
Their analysis results showed that only at a large gap could the through brace of
the joint be taken as the chord member. On the failure behaviour, Dexter and Lee
[12] highlighted from their numerical study that the gap region of the through
brace of the joint could be subjected to high shear stress with failure by cracking at
low load level. This was otherwise shown in the experimental investigation of the
joint under cyclic loading by Soh et al. [4]. In their experiment, compression buck-
ling occurred on the through brace surface with cracking initiated at the saddle of
the through brace near the lap brace. The difference of the failure behaviours could
be due to the geometric properties of braces and the load conditions.
Apart from the cyclic behaviour, a full-scale completely overlapped tubular
K(N)-joint specimen has also been tested to failure under static incremental load-
ing by Fung et al. [13]. Their experimental result revealed that the capacity of the
joint was slightly higher than that of simple gap K-joints. Furthermore, the joint
showed a good ductility performance with a high deformation limit of about four
times the lap brace yield deformation. It was noted that the ultimate capacity of
completely overlapped K(N)-joint at large gap size was equivalent to that of simple
T/Y-joints [12].
The completely overlapped tubular joint under lap brace axial loading showed
less local flexibility in comparison to simple T/Y-joints. The study by Gho et al.
[14] revealed that the local flexibility of the completely overlapped joint increased
with its gap size. The stress interaction between the chord and through brace
showed an insignificant effect on the local flexibility of the joint with large gap size.
However, the local flexibility of the completely overlapped joint was equivalent to
that of simple T/Y-joints with gap size of about three times the through brace
diameter.
In the study of SCF of completely overlapped tubular K(N)-joints by Gho et al.
[9], 8-node thick shell element was used for modelling the joint in the FE analysis.
Their analytical results revealed that the SCF of the joint at a small gap size was
lower than that of simple T/Y-joints. However, as a solid element is highly recom-
mended for the FE analysis of the joint [15], a comparison of SNCF between the
experiment conducted by Gho et al. and the FE models with 8-node thick shell and
20-node solid element is performed in the current study. The results of the com-
parison indicate that both 8-node thick shell and 20-node solid elements are suit-
able for modelling the joint.
3. Numerical investigation
models. The SCF and SNCF at the crown toe, saddle and crown heel of members
are considered for comparison. In this study, the SCF and SNCF of the joint are
defined as the ratios of hot spot stress and strain at the weld toe to nominal stress
and strain of the lap brace, respectively. The load condition of the joint is unba-
lanced with the compressive load applied at the end of the lap brace (Fig. 2). In the
present work, the joint is first modelled according to the geometric and material
properties of the tested specimen for verification. A parametric study is subsequently
conducted for the development of equations to predict the SCF of the joint.
In the current research, weld elements are included in both 8-node thick shell
and 20-node solid FE models for accurate stress analysis of the joint. The method
of welding and the thickness of weld elements are determined based on the
dihedral angle (W) along the joint perimeter as specified in AWS [16]. For the two
1768 W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782
Table 1
Geometric parameters of tested specimen by Gho et al. [9]
v
FE models, the range of the dihedral angles is within 30–180 (Fig. 5). The com-
parison of weld sizes between the two FE models is presented in Table 2.
In the experimental investigation of the joint by Gho et al. [9], a specific type of
strain gauge was used for stress concentration measurement. This strain gauge
comprised five single gauges of 1 mm long, each bonded on a common backing
which enabled the nonlinear distribution of stresses to be captured. All the strain
gauges were placed perpendicular to the weld toe at locations on the crown toe,
saddle and crown heel of members (Fig. 6). Therefore, in the current FE analysis, a
W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782 1769
Fig. 5. Sizes of weld elements. (a) Weld elements for 8-node thick shell FE model; (b) weld elements for
20-node solid FE model. Note: (W) dihedral angle, (a,b) FE sizes, ðtw Þ weld thickness.
combination of linear and parabolic curves is adopted to achieve the curve fitting
of all data points in and around the extrapolation region so that the hot spot strain
at the weld toe can be determined. These data points are obtained from the node
points of the finite elements near the joint intersections of the FE model.
The hot spot strains at the weld toe and the nominal strain of the lap brace are
obtained from the FE models. The nominal strain is calculated based on the
Young’s modulus and the cross sectional area of the lap brace, while the hot spot
strain is determined through the extrapolation of strain values at node points near
the joint intersections. The comparison of SNCF between the tested specimen and
the FE models is summarised in Table 3. From the table, the differences of SNCF
in some cases are not presentable in terms of percentage. Although the difference
of crown heel SNCF of the lap brace between the tested specimen and the 8-node
thick shell FE model is 0.22, the difference in percentage is as high as 46.8%. It
1770 W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782
Table 2
Sizes of weld elements (mm)
Location Element Toe Saddle Heel
(a) Chord and through brace
Chord Thick shell 12.8 10.0 31.2
Solid 1.5 1.7 2.8
Through brace Thick shell 19.2 18 33.4
Solid 14.0 14.0 14.6
Weld thickness Thick shell 5.3 18.0 17.6
Solid 15.2 15.6 14.9
(b) Through brace and lap brace
Through brace Thick shell 9.5 7.7 25.2
Solid 1.1 1.3 2.2
Lap brace Thick shell 16.8 15.6 27.2
Solid 10.7 10.7 11.2
Weld thickness Thick shell 4.3 4.0 14.8
Solid 11.6 11.9 11.4
should be noted that the recommended minimum SCF of tubular joints for design
is 1.5 [17,18]. If the SCF is to be taken as 1.1 times the SNCF as recommended by
UEG [8], then the joints with SNCF less than 1.36 will be considered less significant.
Table 3
Comparison of SNCF between tested specimen and FE models
Member Location Tested FE model Difference % difference
specimen
Thick Solid
shell
From Table 3, the maximum difference of SNCF between the tested specimen
and the 8-node thick shell FE model is 0.36 which occurs on the through brace
crown heel near the lap brace. However, the maximum SNCF of the joint occurs at
the through brace saddle near the lap brace. At this location, the FE model yields
a higher SNCF than the tested specimen with a difference of 0.09. Although the
maximum difference of the lap brace saddle SNCF is 16.4%, the difference in terms
of numeric is only 0.32. Thus, in consideration of this small difference, the SNCF
obtained from the 8-node thick shell FE model is concluded to be in good agree-
ment with the test values.
7.2. Tested specimen and 20-node solid FE model
The comparison of SNCF between the tested specimen and the 20-node solid FE
model shows that the maximum value occurs on the through brace saddle near the
lap brace (Table 3). The difference of SNCF between the tested specimen and the
FE model at this location is 0.24. However, the maximum difference of SNCF (=
0.42) occurs at the through brace crown heel near the lap brace. The average differ-
ence of SNCF of the FE model in comparison with the tested specimen is 9.6%.
This is slightly higher than the average value of 8.4% obtained from the compari-
son of SNCF between the tested specimen and the 8-node thick shell FE model.
Although the maximum difference of SNCF at the chord saddle is 14.6%, the
difference in terms of numeric is only 0.27. It could therefore be concluded that the
SNCF of the 20-node solid FE model also compared reasonably well with those of
the tested specimen.
1772 W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782
Figs. 7 and 8 illustrate the SCF distributions at the joint intersections (chord and
through brace, through brace and lap brace) of the 8-node thick shell and the 20-
node solid FE models. In the figures, / is the angle measured from the crown toe
to saddle and then to crown heel. It is noted that the maximum SCF of the joint is
located at the through brace saddle near the lap brace. At this location, the
maximum SNCF for the 8-node thick shell and the 20-node solid FE models are
2.93 and 3.08, respectively, with a difference of 0.15 (Table 3). Both the models are
found to have SNCF greater than the experimental value. For the SCF, the differ-
ence between the two FE models is only 0.36, which is less than the recommended
minimum SCF of 1.5 [17,18] for design. From the comparison, both the 8-node
thick shell and the 20-node solid elements are suitable for modelling completely
overlapped tubular K(N)-joints. However, in view of the computational effort, the
FE model with 8-node thick shell element is used throughout the parametric study.
1774 W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782
8. Parametric study
192 FE models of the completely overlapped tubular K(N)-joint have been cre-
ated for the parametric study. The load condition is unbalanced with the joint sub-
jected to lap brace axial compression. The boundary conditions of the chord and
the through brace are assumed fixed. The length parameters for the chord
(ac ¼ 2L=D) and the through brace (aT ¼ 2lT =dT ) are taken as 12. As the load is
predominantly transferred between the braces of the joint, the boundary condition
of fixed or pinned as well as the length parameters of 12 have shown insignificant
effect on SCF. In the current analysis, the Young’s modulus (E) and the Poisson’s
ratio (m) of steel material are assumed 200 kN/mm2 and 0:3, respectively. The ran-
ges of geometric parameters of the joints are limited as follow.
0:2 bCT 0:8 ð0:2; 0:4; 0:6 and 0:8Þ
0:2 bTL 0:8 ð0:2; 0:4; 0:6 and 0:8Þ
0:2 sCT 1:0 ð0:2; 0:4; 0:6; 0:8 and 1:0Þ
ð1Þ
0:2 sTL 1:0 ð0:2; 0:4; 0:6; 0:8 and 1:0Þ
8 cC 32 ð8; 16; 24 and 32Þ
0:15 n 1:5 ð0:15; 0:25; 0:5; 1:0 and 1:5Þ
Several geometric parameters of the completely overlapped K(N)-joint are inves-
tigated. For the effect of bCT (= dT/D), the SCF at large cC (= D/2T) is crucial.
The maximum SCF (= 5.42) occurs on the chord saddle at large bCT (Fig. 9). But
for small and medium cC, the maximum SCF occurs somewhere around
bCT ¼ 0:55. For the effect of bTL (= dL/dT), the maximum SCF (= 7.08) is at the
through brace saddle near the lap brace of the joint. As shown in Fig. 10, the SCF
increases with cT (= dT/2tT) and the maximum SCF occurs in the range of
bTL ¼ 0:35 0:45.
For the effect of sCT (= tT/T), the maximum SCF (= 11.09) occurs on the
through brace saddle near the lap brace at small sCT ¼ 0:2 (Fig. 11). The SCF
increases with cC but reduces with sCT. On the contrary, the SCF increases almost
linearly with cT and sTL (Fig. 12). For the effect of sTL (= tL/tT), the maximum
SCF (= 7.29) occurs on the chord saddle at sTL ¼ 1:0.
In summary, the effect of through brace to chord wall thickness ratio (sCT) at
large cC is most crucial as it yields maximum SCF of the joint. For the effect of
through brace to chord and lap brace to through brace diameter ratios (bCT and
bTL), the impact on SCF is found to be less significant. Note that the SCF increa-
ses with chord and through brace diameter to wall thickness ratios (cC and cT)
which indicates that the wall thickness of members is the most important geometric
parameter to minimise the SCF of the joint.
The gap size (gT) of the completely overlapped K(N)-joint on the through brace
surface between the outer surfaces of the chord and the lap brace is an important
parameter. In this study, only the results of SCF due to sCT and sTL are presented
as they are shown to have significant impact on SCF. A non-dimensional para-
meter n ¼ gT =D) is introduced to examine the effect of gap sizes on SCF of the
Fig. 13. Effect of gap size (n) on SCF of the joint—through brace saddle (near lap brace).
joints. For the effect of sCT, the maximum SCF (= 10.27) occurs on the through
brace saddle near the lap brace at large n (= 1.5) (Fig. 13). The difference of SCF
between the maximum and the minimum gaps is 6.28. The SCF at large sCT is
almost constant throughout the gap sizes. A drastic change of curves around
n ¼ 0:2 as shown in Fig. 13 could be due to the restraint of weld elements at the joint
intersection. For the effect of sTL, the behaviour is opposite in comparison to that
of sCT as the SCF reduces with the increment of gap sizes (Fig. 14). The maximum
SCF (= 3.33) occurs on the through brace saddle near the chord at n ¼ 0:2.
A set of parametric equations is proposed to predict the SCF at the joint inter-
sections of completely overlapped tubular K(N)-joints under lap brace axial com-
pression (Appendix A). These equations are derived based on data obtained from
the FE analysis of 192 models. A non-linear regression curve fitting program,
DataFit [19], is adopted for the development of the equations. Each parametric
equation consists of three parameters namely F1, F2 and Q. The parameters F1 and
F2 take account the effect of stress interactions between the chord and through
brace as well as the through brace and lap brace, respectively. The parameter Q
considers the effect of gap sizes on SCF of the joint. The functions of the three
1778 W.-M. Gho, F. Gao / Journal of Constructional Steel Research 60 (2004) 1761–1782
Fig. 14. Effect of gap size (n) on SCF of the joint—through brace saddle (near chord).
Table 4
Assessment of parametric equations with FE database
Note: SCF of 1.5 has been assigned at chord toe (Y1), chord heel (Y3), through brace heel near chord
(Y6) and through brace heel near lap brace (Y9).
Table 5
Assessment of parametric equations with experimental results
Location SCF(P) SCF(R) P/R
Proposed Tested
equations specimen
Y2 Chord saddle 2.19 2.02 1.08
Y4 Through brace toe near chord 1.70 0.86 1.97
Y5 Through brace saddle near chord 2.16 2.49 0.87
Y7 Through brace toe near lap brace 1.61 1.71 0.94
Y8 Through brace saddle near 3.25 3.12 1.04
lap brace
Y10 Lap brace toe 1.69 1.95 0.87
Y11 Lap brace saddle 2.35 2.15 1.09
Y12 Lap brace heel 0.63 0.52 1.20
11. Conclusions
In the FE analysis, both 8-node thick shell and 20-node solid elements are found
suitable for modelling completely overlapped K(N)-joints. The comparison of
SNCF between the tested specimen and the two FE models shows that the differ-
ence is less than 0.42. The difference of SCF between the two FE models is only
0.36. However, in view of the computational effort, the FE model with 8-node
thick shell FE element is used throughout the analysis.
For the parametric study, the effect of through brace to chord wall thickness
ratio is most crucial. A high SCF is obtained at small through brace to chord wall
thickness ratio but at large gap size. Based on data obtained from the FE analysis,
a set of parametric equations is proposed to predict the SCF of the joint at the
crown toe, saddle and crown heel of members. However, owing to small SCF at
the chord toe, chord heel and through brace heel of the joint, it is recommended
that SCF of 1.5 should be assigned at these locations for design. The reliability of
the proposed parametric equations has been verified against the requirement of the
Fatigue Guidance Review Panel.
References
[1] Wardenier J, Kurobane Y, Packer J, Dutta D, Yeomans N. Design guide for circular hollow sec-
tion Joints under predominantly static loading. Koln, Germany: Verlag TUV Rheinland GmbH;
1991.
[2] Cheung LY, Gho WM, Fung TC, Soh CK. Design economics of offshore structures—eccentric
jacket. Journal of the Institution of Engineers, Singapore 1997;38(3):42–9.
[3] Fung TC, Soh CK, Gho WM, Qin F. Ultimate capacity of completely overlapped tubular joints I:
an experimental investigation. Journal of Constructional Steel Research 2001;57(8):855–80.
[4] Soh CK, Fung TC, Qin F, Gho WM. Behaviour of completely overlap tubular joints under cyclic
loading. Journal of Structural Engineering, ASCE 2000;127(2):122–8.
[5] American Petroleum Institute, API RP2A, Recommended practice for planning, designing and con-
structing fixed offshore platforms-working stress design, 21st edition, 2000.
[6] Zhao XL, Herion S, Packer JA, Puthli R, Sedlacek G, Wardenier J, Weynand K, Wingerde A,
Yeomans N. Design guide for circular and rectangular hollow section joints under fatigue loading,
CIDECT. IIW Doc. XV-EE-98-236, 1998.
[7] Efthymiou M, Durkin S. Stress concentrations in T/Y and gap/overlap K-joints, Behaviour of off-
shore structures, Elsevier, Netherlands, 1985:429–40.
[8] Underwater Engineering Group, Design of tubular joints for offshore structures, UEG/CIRIA,
London, UK, 1985.
[9] Gho WM, Fung TC, Soh CK. Stress and strain concentration factors of completely overlapped
tubular K(N)-joints. Journal of Structural Engineering, ASCE 2003;129(1):21–9.
[10] MARC. User manual and primer. California, USA: MARC Analysis and Research Corporation;
2001.
[11] Gho WM. An eccentric braced system for offshore jacket structures, 7th International Conference
on Steel and Space Structures, Singapore, 2002:83–90.
[12] Dexter EM, Lee MMK. Static strength of axially loaded tubular K-joints I: Behavior. Journal of
Structural Engineering, ASCE 1999;125(2):194–201.
[13] Fung TC, Soh CK, Gho WM. Ultimate capacity of completely overlap tubular joints II: beha-
vioural study. Journal of Constructional Steel Research 2001;57(8):881–906.
[14] Gho WM, Fung TC, Soh CK. The local joint flexibility of completely overlapped tubular joints
under axial loading, 7th International Conference on Steel and Space Structures, Singapore,
pp. 311–18, 2002.
[15] Romeijn A, Puthli RS, Wardenier J. Guidelines on the numerical determination of stress concen-
tration factor of tubular joints, Proceedings of 5th International Symposium on Tubular Structures,
Nottingham, UK, 1993:625–39.
[16] American Welding Society, Structural welding code—steel. ANSI/AWS, D1.1-2000, Miami, USA,
2000.
[17] Smedley P, Fisher P. Stress concentration factors for simple tubular joints, Proceedings of 1st Inter-
national Offshore and Polar Engineering Conference, ISOPE, Edinburgh, UK, 1991:475–83.
[18] Chang E, Dover WD. Parametric equations to predict stress distribution along the intersection of
tubular Y and T-joints. International Journal of Fatigue 1999;21(1999):361–81.
[19] DataFit, Version 8.0, Oakdale Engineering, USA, 2002
[20] Fatigue Guidance Review Panel, MaTSU, Fatigue backgroud guidance document, an offshore tech-
nology report. By HSE Books as an Offshore Technology Report, 1996.