Journal of Constructional Steel Research 90 (2013) 108119

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

A loaddeformation formulation for CHS X- and K-joints in

push-over analyses
Xudong Qian , Yang Zhang, Yoo Sang Choo
Centre for Offshore Research and Engineering, Department of Civil and Environmental Engineering, National University of Singapore, 1 Engineering Drive 2, 117576, Singapore

a r t i c l e

i n f o

Article history:
Received 23 February 2011
Accepted 20 July 2013
Available online 25 August 2013
Keywords:
Joint formulation
Circular hollow section
Tubular joint
Pushover analysis
Phenomenological representation

a b s t r a c t
This paper proposes a new loaddeformation formulation for circular hollow section (CHS) X- and K-joints to be
implemented in the pushover analysis of steel frames. The proposed formulation describes the loaddeformation
relationship of the CHS X- and K-joint through a simple function with the coefcients dependent on the ultimate
strength and the geometric parameters of the joint. The strength-dependent parameter follows the mean
strength equations in the latest IIW recommendations, while the geometric-dependent parameters derive
from the nite element results of the CHS X- and K-joints covering a practical geometric range. The proposed
joint formulation predicts closely the loaddeformation responses for planar CHS X- and K-joints measured in
the experiments. The non-dimensional loaddeformation formulation developed in the current study provides
a calibrated basis in the phenomenological representation of the nonlinear joint behavior in push-over analyses
of steel space frames. The experimental results from the large-scale 2-D and 3-D frame tests validate the accuracy
of the proposed formulation, which is implemented in a nonlinear pushover analysis as joint-spring elements.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
The extended service of steel offshore platforms beyond their initial
design life of 20 years has become a common practice due to economical considerations [1]. The reassessment of such platforms requires advanced nonlinear frame analyses which should include an accurate
representation of the local joint responses under overloading conditions. The rigid joint assumption in the conventional frame analysis
often underestimates the deformation and over-estimates the ultimate
resistance of the critical joint. The rigid joint hypothesis therefore may
cause strong effects on the load-distribution and the sequence of the
component failure in the structure, leading to severe deviations in the
predicted frame behavior from the real structural response. However,
frame analyses with rigid-joint assumptions do not always provide conservative estimations on the ultimate strength of the structure, since
large deformation of the joint may mobilize adjacent redundant members and leads to higher structural resistances than the prediction by
rigid-joint frame analyses. Hence, improved understandings on the effect of nonlinear joint behavior on the frame response become necessary to develop a simple and calibrated engineering representation of
the nonlinear joint characteristics in pushover analyses.
The last three decades observe substantial experimental developments in the local joint exibility (LJF) for both uni-planar and multiplaner tubular joints. Bouwkamp [2] investigates the effect of joint exibilities on the elastic response of offshore jacket structures where a
Corresponding author. Tel.: +65 6516 6827; fax: +65 6779 1635.
E-mail address: qianxudong@nus.edu.sg (X. Qian).
0143-974X/$ see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jcsr.2013.07.024

detailed shell element model describes the behavior of the joint-can

which is connected to the brace members modeled by beam elements.
Efthymiou [3] develops the elastic exibility parametric formula for
T- and TY-joints based on the FE analysis using thin-shell elements
with the weld modeled by solid elements. Fessler [4,5] improves the formula based on Araldite models for Y-, X-joint and multi-brace joints.
Chen [6] develops special elements to represent the joint exibility by
subdividing the brace end into nite strips, and derives parametric formulae for the elastic stiffness of T-, Y- and K-joint. Kohoutek [7] investigates the T-joint elastic stiffness based on the frequency measurement
by introducing a rigidity index into the stiffness matrix and calibrating
it through the natural frequency of the joint. Romeijn [8] investigates
the inuence of the joint geometry on the exibility of uni-planar and
multi-planar joints based on the nite element (FE) results. Holmas
[9,10] develops a joint shell element based on the small-displacement
theory and the load is represented by a series of concentrated loads
along the brace-to-chord intersection lines. The model yields a good
agreement with the parametric formulae for the elastic joint exibility
derived by Fessler [4,5] and Chen [6].
A natural extension of the work on the joint stiffness focuses on examining the inuence of the local joint exibility on the frame response
by incorporating the LJF into structural analyses. Sub-structural models
of the critical joint using a detailed FE mesh [2,8,9,1115] provide an accurate approach to incorporate the nonlinear joint deformation in the
frame analysis. However such methods become infeasible when applied
to a realistic steel offshore platform with multiple critical joints, which
require substantial computational resources in iterating the detailed
stress/strain/displacement elds in the 3-D local joint model.

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

Nomenclature
A
B
C
P
PE
Pu
P
Pu
d0
d1
fu
fy
g
g
k0
l0
t0
t1

E
i
u

E
u

geometric-dependent constant in the loaddeformation

formulation
geometric-dependent constant in the loaddeformation
formulation
constant in the loaddeformation formulation
applied load
load at elastic limit
peak load
non-dimensional applied load
non-dimensional peak load
outer diameter of the chord
outer diameter of the brace
material ultimate strength of the chord
material yield strength of the chord
gap between two braces
(non-dimensional) gap ratio (g/t0)
initial stiffness of the joint
length of the chord
thickness of the chord
thickness of the brace
chord length to half chord diameter ratio ( = 2l0/d0)
brace diameter to chord diameter ratio ( = d1/d0)
deformation parameter
deformation at elastic limit
limit deformation parameter in proposed joint
formulation
deformation at the peak load
non-dimensional deformation
non-dimensional deformation at elastic limit
non-dimensional deformation at the peak load
chord radius to chord wall thickness ratio ( = d0/2t0)
angle between the brace and the chord centerline
brace wall thickness over chord wall thickness ratio
( = t1/t0)
angle parameter in the proposed joint formulation

109

and brace loads in the joint response and the K-, X-, and Y-joint classications [21].
This study develops a nonlinear joint formulation which predicts
closely the loaddeformation responses for CHS X- and K-joints
subjected to brace axial compression. This proposed formulation describes the loaddeformation relationship for CHS X- and K-joints with
different geometric parameters covering a brace to chord diameter
ratio () from 0.3 to 1.0 and a chord radius to thickness ratio () ratio
from 7 to 25. The proposed loaddeformation relationship develops
from loaddeformation results computed using calibrated FE analyses.
The parametric formulation, proposed in the current joint representation,
provides a convenient approach to characterize the loaddeformation
curve and eliminates the need for the elasticplastic, large-deformation
nite element analyses on CHS X- and K-joints. The nonlinear pushover
analysis, performed in the numerical tool USFOS (an acronym for
Ultimate Strength for Framed Offshore Structures) [22], proves the validity of the proposed formulation by implementing the proposed formulation via spring elements between the chord and brace members in 2-D
and 3-D space frames.
This paper starts with an introduction including a review on the research of the joint exibility and the jointframe interaction. The next
section develops the joint phenomenological formulation to represent
the joint resistance with respect to the deformation due to the yielding
and plasticity mobilized under the remote brace loading. The study
compares the proposed joint formulation with calibrated nite element
results. The following section presents the verication of the proposed
formulation using experimental results reported on 2D and 3D frames.
The last section summarizes the main conclusions drawn from the current study.
2. Joint formulation
This section veries the accuracy of the elasticplastic, large deformation FE analysis based on the experimental results for CHS X-joints
and K-joints. A loaddeformation relationship develops subsequently
from calibrated FE analyses, covering a practical geometric range for
X- and K-joints.
2.1. Validation of nite element analysis

Alternatively, the phenomenological representation of the joint behavior through nonlinear joint springs, as shown in Fig. 1a, provides a
convenient method for practicing engineers. Ueda [16,17] develops
the elasticperfectly-plastic springs for CHS T-, Y- and K-joints. Choo
et al. [18] implement the joint spring elements described by piecewise linear loaddeformation curves in the nonlinear frame analysis.
The joint industry project, led by a UK company [19,20], develops the
MSL formulation, which includes the interaction between the chord

Table 1 lists the geometric parameters for the CHS X- and K-joint
specimens reported by van der Vegte [23] and Kurobane et al. [24], respectively. In Table 1, the ratio indicates the ratio of the brace diameter
over the chord diameter. The value stands for the chord radius to the
chord wall thickness ratio. The parameter refers to the chord length
to the chord radius ratio and denotes the ratio of the brace wall thickness over the chord wall thickness. The X-joint reported by van der Vegte
[23] experiences brace axial compression, as shown in Fig. 2a. Fig. 2b
shows the uni-axial true stress and true strain curve for the X-joint

(a)

Chord
Nodes for spring element
(Beam-column
element)

Load

Brace
(Beam-column element)

(b)

Deformation

Fig. 1. (a) Joint spring representation in the global frame analysis; and (b) loaddeformation characteristics of the joint spring.

110

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

Table 1
Geometric parameters of the X- and K-joint specimens in the reported test.
Joint

d0 (mm)

X-joint
K-joint

408.0
216.4

0.6
0.76

20
13.7

12
16

1.0
0.67

60

3.8

material. The presence of three planes of symmetry allows the use of a

one-eighth model, as shown in Fig. 2c, which illustrates an FE mesh
built from 20-node hexagonal elements in the preprocessor Patran
[25]. The elasticplastic, large-deformation analysis utilizes the generalpurpose nite element package, ABAQUS [26]. Fig. 2d demonstrates the
close agreement between the experimental loaddeformation response
of the CHS X-joint and that computed from the FE analysis, which predicts accurately the plastic deformation around the brace-to-chord intersection observed in the experiment.
Fig. 3a illustrates the boundary conditions on the gapped K-joint test.
A test frame supports the ends of the left brace and the chord by bearing
pins. The K-joint experiences axial load applied on the right brace, the
end of which is free to rotate in the plane of the K-joint. Fig. 3b shows
the uni-axial true stress and true strain relationship and Fig. 3c illustrates the typical FE mesh for the K-joint. Fig. 3d conrms the accuracy
of the FE analysis compared to the experimental loaddeformation
record.

form of the function should remain universal for different types of joints
under various loading conditions. In addition, the function should adopt
a simplest possible form for curve tting and subsequent engineering
applications.
Fig. 4 shows the typical loaddeformation curve for an X-joint under
the brace axial compression and that for a gapped K-joint under balanced brace axial loads. Except for very thick-walled chords [27], the
X-joint under brace axial compression often exhibits a peak load as
the deformation increases. The joint resistance decreases gradually
after the peak load as the plastic deformation propagates in the chord
wall. In contrast, the gapped K-joint under balanced axial loads sustains
monotonically increasing loads until the joint resistance is limited by
the ductility of the material, or extensive plastic deformations in the
chord.
Coupling the physical response of the X- and K-joints with the requirements on the expected loaddeformation function, the proposed
load deformation formula follows,



P f Pu h

where Pand P u are the non-dimensional load and ultimate load, respectively, or,
P

P sin
f y t 20

2.2. Proposed joint formulation

This research work targets at developing an accurate nonlinear relationship for the loaddeformation responses of CHS X- and K-joints, including both the elastic and the elasticplastic responses. The expected
function, which needs to describe such a relationship, shall entail the
following characteristics. The formulation should provide highly accurate estimates on the joint response for a wide range of practical geometric parameters. The function should be at least C2 continuous,
implying that the change in the joint stiffness as the load increases
should be continuous before any unstable failure occurs. The basic

Load

Saddle

(a)
t1
d1

t0

Pu

P u sin
f y t 20

where fy denotes the yield strength of the chord material, t0 refers to the
thickness of the chord member, and measures the intersection angle
between the brace and the chord. The parameter in Eq. (1) represents
the non-dimensional deformation of the joint,

d0

800

(MPa)

(b)

600

Crown
d0

400

f y = 331 MPa

200

E = 205 GPa

fu = 435 MPa

= 0.3

l0
0

Reaction

(c)

No. of nodes: 28,000

No. of elements: 5,300

15

10

20

30

(%)

40

(d)

P / f yt02

10

5
Test
FE
0
0

0.25

/ d0

0.50

0.75

Fig. 2. A CHS X-joint test: (a) test set-up; (b) uni-axial true stress and true strain curve; (c) FE mesh; and (d) comparison of the loaddeformation curve between the test and FE analysis.

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

(a)

800

111

(b)

(MPa)

Loading
600

f y = 480 MPa

400

Saddle

Pin support

f y = 532 MPa

Crown

E = 205 GPa
200
0

(c)

= 0.3

10

(%)

20

30

(d)

P sin / f yt02

30

No. of nodes: 47,000

No. of elements: 9,000
20

10

Test
FE
0

0.25

0.50

/d0

0.75

Fig. 3. A CHS K-joint test: (a) loading and boundary conditions; (b) uni-axial true stress and true strain curve; (c) FE mesh; and (d) comparison of the loaddeformation curve between the
test and FE analysis.


where d0 denotes the chord diameter. In Eq. (1), f P u is a linear func

tion of P u and h is a logarithmic function of ,

h

pi2 

P CP u 1A ln 1 B 1= A
:

In Eq. (5), CP u refers to the extreme value of the function. The derivative of the load P follows,
"

# 

1 B
dP
B
2AC ln
:
1
p
d
1 B
e A

The coefcient A determines the decreasing rate of the joint strength

beyond the peak load. A large value of A in Eq. (5) creates a sharp variation in the joint resistance as the deformation increases, while a small
value of A generates a smooth loaddeformation relationship. The value
of A, therefore, exhibits strong dependence on the geometric parameters of the joint. The coefcient B together with A determines the initial

16

(a)

P / f yt02

 p

e1= A 1
u
:
u

B
d0

The proposed joint formulation includes four independent parameters: P u , A, B and C. The current approach employs the mean strength
equations in IIW [28] forP u , which follows,


1
0:15

P u 3:16
10:7



1:6
0:3

P u 2 16
1

1
:
1:2 g=t 0 0:8

Eq. (8) denes non-dimensional ultimate strength X-joints under

brace axial compression and Eq. (9) calculates that for K-joints under

50

P sin / f yt02

(b)

40

12

30

d0 = 406 mm
= 0.6
=15

d0 = 406 mm
= 0.9
= 15
= 60o

20

4
0

stiffness of the curve. The joint displacement at the peak load derives
from Eq. (5) by setting P CP u , or by setting Eq. (6) to zero,

10

0.04

0.08

/d0

0.12

0.16

0.02

0.04

0.06

0.08

/d0

Fig. 4. Typical loaddeformation curves for: (a) an X-joint under axial brace compression; and (b) a K-joint under balanced axial brace loading.

112

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

balanced brace axial loading. The use of Eqs. (8) and (9) reduces the
number of undetermined coefcient to three: A, B and C. The value of coefcients A and B should remain in reasonable ranges to avoid a negative value of the dependent variable P, which requires a positive value


of the g function, or,
h

pi2

p2
A ln 1 B 1= A b 1 or e A N 1 B:

10

The deformation at the peak load in Eq. (7) should remain as a positive value, which requires,
A N 0 and B N 0:

11

2.3. X-joint formulation

The current study determines the value of A, B and C for CHS X-joints
through a regression analysis of the results obtained from 30 elastic
plastic, large-deformation analysis, covering a ratio from 0.3 to 1.0
and a ratio from 7 to 25.
The loaddeformation characteristics of the X-joint depend signicantly on the ratio. Based on the plastic hinge model proposed by
Togo [29], plastic hinges form at the saddle point and the mid-depth
point of the chord cross section when an X-joint reaches its peak capacity. For a joint with a small ratio, the chord wall around the brace-tochord intersection area undergoes membrane, bending and shearing actions. As the ratio increases, the two braces become closer in locations
and the load transfers from one brace to the other predominantly via
the membrane action in the chord wall material between the two
braces. To reect this change, the parametric study includes six ratios:
0.3, 0.6, 0.9, 0.93, 0.96 and 1.0 for the X-joint.
For thin-wall joints with a large value, the transverse shear across
the chord wall thickness is negligible based on the thin-shell theory. The
joint strength depends primarily on the interaction between the bending and axial stresses acting on the chord wall. Large deformations of
the chord wall create strong variations in the magnitudes of these bending stresses and axial stresses, causing a pronounced change in the resistance of the joint. This sharp change in the joint resistance with
increasing joint deformations yields a relatively large A value. As the
ratio approaches 1.0, the membrane action becomes dominant, which
leads to a much higher joint capacity than that of a joint with the
same chord size under dominant bending actions in the chord wall.
Therefore, the loaddeformation curve for X-joints with a large ratio
shows a smooth variation, corresponding to a small A value in Eq. (5).
For thick-walled joints with a small value, the transverse shear
across the chord wall contributes to the joint strength. The large bending and shear stiffness of the chord wall limits the deformation in the
chord wall and leads subsequently to a small variation in the joint resistance with increasing joint deformations, as compared to thin-walled
joints. This smooth variation in the loaddeformation relationship for
the thick-walled
leads to a relatively small value of A.

joint

The term f P u CP u in Eq. (5) characterizes a reference load level in
the loaddeformation relationship. The non-dimensional ultimate
strength P u incorporates the geometric dependence of the load resistance, while the parameter C includes the effect of joint types and loading
conditions. The C value for CHS X-joints under brace axial compression,
which often exhibits a peak in the loaddeformation curve, equals to 1.0.
The curve-tting procedure to evaluate the coefcients A and B consists of two steps. The rst step determines the values of A and B for each
discrete loaddeformation curve obtained from the nite element analysis. The geometric-dependent formulations of A and B then derive from
a nonlinear regression procedure [30] using the values of A and B for all
joints included in the parametric study. Table 2 lists the corresponding
formulation for A and B, which demonstrates a close agreement with
the discrete values obtained using the FE results, as shown by the

Table 2
Coefcient in the proposed formulation for X-joints.
Coefcient Formulation

Proposed/FE
Mean Standard deviation No. of data

A
B
C

5:2

2
2:5 7:54:3

0:27
1.00
(2333.24 40 + 820)0.6 1.01
1.00
1.00

0.08
0.09
0.02

30
30
30

small standard-deviation values. Fig. 5 compares the loaddeformation

curves predicted by the proposed joint formulation and those computed
from the FE analysis for four typical CHS X-joints. The proposed load
deformation formulation agrees well with the loaddeformation relationships computed from the large-deformation, elasticplastic FE
analysis.
This study compares the predictions of the critical joint deformation
at the peak load and the initial joint stiffness derived from the proposed
joint formulation with the reported studies [18,31] to ensure that the
proposed formulation provides reliable estimations on these important
parameters. Lu's deformation limit, which corresponds to a joint deformation equal to 3% of the chord diameter [31], has become a widely recognized deformation parameter to dene the ultimate strength of
tubular joints. The initial stiffness formulation, reported by Choo et al.
[18] based on an extensive numerical study, estimates the joint stiffness
as,
k0 P E =E cP u =E 0:8P u =E

12

where the load PE corresponds to the limit of elasticity and assumes a

value of 0.8Pu based on the FE analysis [18]. The initial joint stiffness
(k0), therefore, equals,

p

1 0:2=A
k0 0:8P u =E 0:8BP u = e
1

13

where E denotes the displacement at 0.8Pu. Table 3 shows the agreement in the u and k0 values obtained from the proposed joint formulation in comparison with Lu's deformation and the k0 results reported by
Choo et al. [18].
For X-joints under brace axial compression, a re-development of the
joint strength occurs at a large deformation level due to the direct contact of the compression braces, as observed in the BOMEL 2D and 3D
frame tests [32,33]. Fig. 6a shows the large deformation of the chord,
which leads to the contact of the two braces through the inner surface
of the chord. The direct contact of the two brace members leads to a
re-gained joint strength equal to the axial yield strength of the brace
member at a joint deformation of = 0.5d0. The initialization of the
strength re-development depends on the ratio, as shown in Fig. 6b,
which denes i to be the displacement corresponding to the initial contact of the two braces,


1
i 0:5d0 t 0 sin cos :

14

The value of i corresponds to the distance between the inner surfaces of the chord member near the saddle point, measured along a vertical axis corresponding to the mid-thickness of the brace wall, as
shown in Fig. 6c. Fig. 6d shows the close agreement between Eq. (14)
and the i values obtained from the nite element analysis, which prohibits self penetration of the chord inner surface in the contact algorithm. The proposed joint formulation includes this redevelopment of
joint strength for CHS X-joints under brace axial compression through
a bilinear model in the loaddeformation relationship, as shown by
the dashed line in Fig. 6b.

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

(a)

P / f yt02

15

113

(b)

P / f yt02

25
20

10

15

d0 = 406 mm
= 0.6
= 10

FE
Proposed
0

0.03

0.06

0.09

d0 = 406 mm
= 0.9
= 10

10

FE
Proposed

0.12

0.15

0.03

0.06

/d0

(c)

P / f yt02

15

15

/d0

0.09

0.12

0.15

(d)

P / f yt02

10

10
d0 = 406 mm
= 0.6
= 20

d0 = 406 mm
= 0.9
= 20

FE
Proposed

FE
Proposed
0

0
0

0.03

0.06

/d0

0.09

0.12

0.15

0.03

0.06

0.09

/d0

0.12

0.15

Fig. 5. Comparison between the proposed loaddeformation formulation and FE results for CHS X-joints with: (a) = 0.6, = 10; (b) = 0.9, = 10; (c) = 0.6, = 20; and
(d) = 0.9, = 20.

2.4. K-joint formulation

The loaddeformation curve for the K-joint under balanced axial
loads follows the response of the compression brace [18]. Based on
the typical loaddeformation curve for a gapped K-joint, which continues to sustain increasing loads beyond the Lu's deformation limit
[31], the K-joint formulation also follows Eq. (5).
The strength of the K-joint depends on the membrane, shear and
bending resistance of the chord wall around the brace-to-chord intersection. In addition, the gap in the chord between the two braces transfers the load from one brace to the other and experiences bending, shear
and membrane actions at large deformations. Similar to the X-joint, the
transverse shear in a thick-walled chord of a K-joint also contributes to
the joint strength. The shear contribution leads to a smooth loaddeformation curve for the thick-walled K-joint with a small ratio. Therefore,
the A value, which implies the rate of change in the load level, decreases
as decreases.
The determination of the coefcients A, B and C follows the same
procedure as that for the X-joint. The parametric FE analysis covers a
ratio from 0.3 to 1.0 and a ratio from 7 to 25. The numerical analysis
xes the gap between the two braces to be twice the wall thickness of
the chord. The boundary conditions for the K-joint follows that shown
in Fig. 7a, which provides a conservative representation of framing effect on the K-joint [34].

Table 3
Comparisons of the critical deformation at the peak load and the joint stiffness with
reference studies for X-joint.
Parameters

u/d0
k0 P d0f sin
t2
y 0

Results from Reference

0.03 (Lu's deformation limit [31])

2
1185 0:8 0:150:4 [18]

Proposed/reference
Mean

Standard
deviation

No. of
data

1.10
1.09

0.20
0.15

30
30

Table 4 illustrates the formulation for A, B and C based on the

nonlinear regression analysis, which leads to a close agreement with
the values determined from the FE analysis, as reected by the mean
and standard deviation values in the same table. Fig. 7b and c sketches
the loaddeformation curves predicted by the proposed joint formulation and those computed from the FE analysis for two typical K-joints.
The proposed joint formulation based on the IIW equation [28] provides
a close agreement with the FE results.
3. Validation of the proposed formulation in pushover analyses
The current study implements the proposed joint formulation in the
frame analysis performed using the nonlinear frame analysis tool,
USFOS [22]. The verication study utilizes experimental results from
large-scale 2D and 3D frame tests [32,33,35]. The element formulation
in USFOS employs the exact solution of the governing equation for
beam-columns subjected to end-forces, which enables the modeling
of each physical member by one element. The plastic hinges at the
mid-span and at the end of the beam-column element simulate the material nonlinearity [22].
3.1. BOMEL 2D frames
Bolt et al. [33] report an experimental study of a series of 2D largescale frame tests under the scope of a joint industry project. The frame
test consists of 6 double-bay X-frames and 4 single-bay K-frames. The
current study compares the results of three X-braced frames, namely
Frame I, Frame II and Frame III, as shown in Fig. 8. The design of
X-frames follows practical congurations representative of offshore
jacket structures. Frame I has strong joint-cans at both the top and the
bottom bays, together with a horizontal member in the middle of the
top and the bottom bays, while frame II includes a weak joint-can at
the top bay to investigate the load shedding and redistribution. Frame
III remains the same as Frame I, except that the mid-horizontal member

114

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

(a)

(b)

P
Brace yielding
capacity

i = (0.5 d0 t 0)sin

Bilinear model
= 0.5d0

Eq. (5)

(c)

i / d0

0.5

(d)

0.4

0.5d0 t0

i = (0.5d0 t0 )sin

= cos 1(

0.3
0.2

0.5d1 0.5t1
)
0.5d 0 t0

by ABAQUS

i = (0.5d0 t0 )sin

0.1
0

0.2

0.4

0.6

0.8

1.0

1.2

Fig. 6. (a) Contact of two compression braces under a large deformation level for X-joint; (b) schematic loaddeformation relationship for the X-joint with strength re-development;
(c) deformation level at the initial contact of the two compression braces; and (d) comparison of i obtained from FE analyses and Eq. (14).

is removed. Each frame connects to a test rig through pin connections at

the bottoms of the two vertical legs, with out-of-plane pin supports provided at six primary leg joints. The test arrangement applies a horizontal
load at the top of the frame until the critical joints and members deform
signicantly, causing pronounced reductions in the frame resistance.

(a)

The current study includes three types of joint formulation for each
frame analysis: 1) the rigid joint assumption, 2) MSL joint formulation
[22], and 3) the proposed joint formation. Fig. 9a compares the numerical analysis and the test results for Frame I. In Frame I, the buckling of
the top-bay compression brace (shown in Fig. 8a) dominates the

32

P sin / f yt02

(b)

24
d0 = 406 mm
= 0.6
= 15

16

FE
Proposed

8
0

0.015

0.030

0.045

0.050

/ d0
60

(c)

P sin / f yt02

45
d0 = 406 mm
= 0.9
= 20

30

15
0
0

FE
Proposed

0.02

0.04

/ d0

0.06

0.08

Fig. 7. (a) Load and boundary conditions for FE analyses of CHS K-joints; (b) the proposed loaddeformation formulation for the CHS K-joint with = 0.6, = 15; and (c) the proposed
loaddeformation formulation for the CHS K-joint with = 0.9, = 20.

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

115

Table 4
Coefcient in the proposed formulation for K-joints.
Formulation

2
0:075 1:5 26:4 e0:0017 0:0141:47 =1000
1.9
2267e
1.13

All units: (mm)

(a)

5944

1524

3.2. Kurobane's 2D frames

Kurobane and Ogawa [35] summarize the cyclic tests on 15 2D
frames, with six frame congurations investigated. The current study
veries the K-joint formulation based on three typical frames, as
shown in Fig. 10. All three frames shown in Fig. 10 experience a vertical
load at the right end of the frame, while the chord ends on the left are
xed via ange connections to a reaction wall. The test measures the rotation as the deection of the frame divided by the length of the truss.
Fig. 11 compares the numerical prediction of the frame response
based on different joint formulations with the experimental results.
Similar to the BOMEL 2D frames, the numerical study includes three
types of joint formulation in the frame analysis. Each frame analysis includes the joint formulation for all K-joints in the frame. All three analyses (the rigid joint, the MSL formulation and the proposed formulation)
predict accurately the failure mode of Frame A, which is governed by the
member buckling (Figs. 10a and 11a). The proposed formulation agrees
with the test results on both the frame stiffness and the ultimate frame
strength. The MSL formulation predicts a more exible frame response
than the test results. A detailed examination reveals that the MSL formulation predicts a much lower joint stiffness than does the proposed joint
formulation. The latter agrees with the joint stiffness obtained from a
separate FE analysis for the K-joint in Frame A.

(b)

(c)

5944

Load

168OD7.1WT

6096
356ODx12.7WT

168OD4.5WT

Buckled
brace

168OD5.1WT

Buckled
brace

168OD9.5WT
168OD6.3WT

6096

6096

168OD6.3WT

168OD7.1WT

168OD9.5WT

Buckled
brace
168OD9.5WT

1524

1524

1524
6096
6096
1524

16
16
16

Load

168OD9.5WT

All members
168OD4.5WT

0.04
0.05
0.03

168OD4.5WT

Buckled
brace

168OD6.3WT

1.00
1.00
1.00

5944

168OD4.5WT

168OD9.5WT

No. of data

prediction on the global frame response than that using the MSL formulation and that based on the rigid joint assumption.

Load
168OD 7.1WT

Standard deviation

All members
168OD4.5WT

All members
168OD4.5WT

Fig. 8. Conguration of BOMEL 2D frames: (a) Frame I; (b) Frame II; and (c) Frame III.

356ODx12.7WT

frame strength. All three analyses predict this failure mechanism. The
proposed formulation leads to a slightly better prediction on the ultimate strength of the test frame than the MSL joint formulation.
Fig. 9b compares the global response for Frame II, in which the topbay X-joint is the critical structural component. This X-joint with = 1
under brace axial compression softens gradually due to increased plastic
deformations in the chord wall beyond the peak load. The contact of the
two braces at a further deformation redevelops the joint strength sufcient to cause buckling of the compression brace in the top bay. The
global load applied on the frame thus increases until the buckling of
that compression brace occurs. The MSL formulation shows the softening of the X-joint in line with the experimental observation. However,
the MSL formulation imposes a deformation limit on the X-joint and
leads to the termination of the analysis before the unstable brace buckling takes place. The proposed joint formulation predicts both the softening of the CHS joint due to plastic deformations in the chord wall
and the re-strengthening of the X-joint at a large deformation level.
The predicted frame response using the proposed joint formulation
thus reects correctly the buckling failure of the test frame, albeit that
this brace buckling occurs at a lower load level than that observed in
the test.
Without the horizontal member, Frame III shows a similar ultimate
strength level compared to Frame I, as shown in Fig. 9c. The absence
of the horizontal member generates signicant load re-distributions beyond the top bay brace buckling. This forces the buckling of the bottom
bay brace to occur at a small global displacement level. The frame analysis based on the proposed joint formulation provides a better

Mean

1524

A
B
C

6096

Proposed/FE

356ODx12.7WT

Coefcient

116

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

1000

(a)

Global load (kN)

1200

Brace buckling

Brace buckling

800

800

600
Test
MSL
Rigid
Proposed

400
200
0

1000

(b)

Global load (kN)

Frame I
0

0.05

0.10

Test
MSL
Rigid
Proposed

Joint yielding
400
Frame II
0

0.20

0.15

0.05

Global displacement (m)

0.10

0.15

0.20

0.25

Global displacement (m)

(c)

Global load (kN)

Brace buckling (top bay)

Brace buckling (bottom bay)

800
600
400

Test
MSL
Rigid
Proposed

200
Frame III
0

0.05

0.10

0.15

0.20

Global displacement (m)

Fig. 9. Comparison of the global loaddeformation response between numerical analysis and experimental records for: (a) Frame I; (b) Frame II; and (c) Frame III.

frame exhibits a much lower strength and a much lower stiffness than
the test results. The proposed joint formulation predicts closely the softening of the joint and agrees well with the peak strength of the test frame.

Frame B fails by the out-of-plane buckling and local buckling of the

brace (Figs. 10b and 11b). The three analyses show similar strength predictions as the test, as shown in Fig. 11b. The MSL formulation provides
a lower prediction on the frame stiffness than the test frame.
Fig. 11c compares the numerical prediction and the experimental record on the global loaddisplacement response for frame T, which is
governed by the buckling of the brace shown in Fig. 10c. The rigid joint
formulation predicts a sequence of member buckling due to the stiff
joint response. The MSL formulation predicts a weak joint and the

3.3. BOMEL 3D frames

Bolt and Billington [36] report the large-scale 3D frame tests shown in
Fig. 12. The double-bay test frame consists of six vertical legs [37]. As
shown in Fig. 12a, the structure presents a hybrid of bracing conguration

(a)

(b)

3572

2418

1500

60.4OD
3.8WT

Load

Load

1000

60.5OD3.8WT

60.6OD2.2WT

60.6OD2.2WT

165.4OD5.7WT
165.5OD5.7WT
Buckled brace

Buckled brace

(c)
60.5OD
2.1WT

60.5OD
2.1WT

60.5OD
2.1WT

Load

1250

4054

139.9OD4.1WT
Buckled brace

All units: (mm)

Fig. 10. Conguration of Kurobane's 2D frames: (1) Frame A; (b) Frame B; and (c) Frame T.

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

(a)

Global load (kN)

150

200

(b)

Global load (kN)

Out-of-plane buckling

150
100

117

Brace buckling
100

Test
MSL
Rigid
Proposed

50
Frame A

Test
MSL
Rigid
Proposed

50
Frame B

0
0

0.02

0.01

0.10

Global rotation (Radian)

(c)

Global load (kN)

150

0.20

Global rotation (Radian)

Brace buckling
100

Test
MSL
Rigid
Proposed

50
Frame T
0
0.02

0.04

Global rotation (Radian)

Fig. 11. Comparison of the global loaddeformation response between the numerical analysis and experimental records for: (a) Frame A; (b) Frame B; and (c) Frame T.

typical for offshore jacket structures. The two longitudinal panels in the
horizontal plane (designated as Panel A and Panel B) are X-braced. In
Panel A (the bottom panel in Fig. 12a) the X-joints have thick jointcans. In Panel B (the top panel in Fig. 12a), the two level I X-joints do
not include joint-cans and the through chords run in opposite directions.
The transverse panels C and D are K-braced with intermediate diamond
bracing in between the two panels. In Panel C, neither of the gapped Kjoints has a joint-can. The distant transverse panel E is X-braced but

without a horizontal member in the middle height. The entire structure

is mounted in a self-reacting frame made of I- and H-sections, as
illustrated in Fig. 12a. The bottom of the self-reacting frame, which is parallel to Panel A in Fig. 12a, sits on a strong oor. The entire testing procedure includes three load cases, as shown in Fig. 12cd, in which the selfreacting frame is removed.
The testing of the 3D frame includes three load cases. In Load Case I,
the front K-braced panel along Panel C is loaded vertically upwards, as

(b)

(a)
Panel E

Load Case I

Panel D
Panel C

Panel B

Buckled
brace

Panel A

Crack

Load

Level I
Level II

Panel C
Panel E

(c)
Panel E

Load Case III

Load Case II

(d)
Crack

Crack

Panel A
Load

Buckled
brace

Load

Buckled brace
Fig. 12. Conguration of BOMEL 3D frame test: (a) test model; (b) Load Case I; (c) Load Case II; and (4) Load Case III.

118

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

shown in Fig. 12b. Fig. 13a shows the comparison of three analyses with
the experimental results. The weld-toe crack near the tension brace in
the K-joint (Fig. 12a) initiates a slight decrease in the frame strength.
However, this crack does not grow extensively under increasing loads.
Instead, the diamond brace in level I redistributes the load to Panel D
and the K-brace in Panel D buckles at the peak frame load.
None of the three types of joint formulation includes a representation on the fracture failure in tubular joints. Both the MSL and the proposed formulation predict the weakening of the joint under plastic
deformation as well as the subsequent brace buckling in the frame.
The proposed formulation provides a close prediction on the ultimate
strength of the test frame.
In Load Case II, the X-braced Panel E experiences a vertical load applied in an upward direction as shown in Fig. 12c. The weakening of
the X-joint under plastic deformation in the chord wall leads to a ductile
frame response. Similar to Frame II in the BOMEL 2D frame (see Fig. 8b),
the large deformation of the joint enables contact of the two compression braces through the inner surface of the chord. This contact leads
to the redevelopment of the joint strength and causes the buckling of
the compression brace. Both the MSL joint formulation and the proposed joint formulation predict the weakening of the joint, as shown
in Fig. 13b. Similar to Frame II (in Fig. 8b), the deformation limit in the
MSL joint formulation terminates the frame analysis at a small deformation level, insufcient to mobilize the subsequent brace buckling. The
proposed joint formulation shows a good agreement with the test results for Load Case II. The rigid joint formulation estimates a relatively
smaller frame capacity than the test by forcing the compression brace
to buckle at a very small global deformation level.
In Load Case III, a horizontal load is applied along Panel A to the bottom X-braced panel. After all the compression braces in Panel A buckles,
the horizontal braces redistribute the load to Panel B and leads to the
crack in two joints shown in Fig. 12d. The test stops after the K-brace
in Panel C buckles. Similar to Load Case I, crack initiation is not captured,
which contributes to the difference between the proposed formulation
and test results. Fig. 13c shows that the proposed formulation predicts
the frame ultimate strength accurately.

1200

(a)

Global load (kN)

4. Summary and conclusions

The current study develops a new loaddeformation formulation for
CHS X- and K-joints to describe their nonlinear loaddeformation behavior in the global pushover analysis. The current study focuses on
the loaddeformation response of X- and K-joints subjected only to
brace axial loads. The reference ultimate strength in the proposed
joint formulation follows the latest IIW recommendations [27]. The proposed joint formulation develops through regression analyses of the FE
results, which are validated against reported experimental results. The
verication study of the proposed formulation on CHS X- and K-joints
in the pushover analysis utilizes 2D BOMEL [32,33] and Kurobane
frame tests [35] as well as 3D BOMEL [36] frame experiments. The
study summarized above supports the following conclusions:
(1) The proposed joint formulation provides a convenient approach
to estimate the loaddeformation relationship for CHS X- and Kjoints. The parametric formulation based on the joint geometry
and loading conditions eliminates the need for the elasticplastic,
large-deformation nite element analyses on CHS X- and K-joints.
The verication based on the reported experimental study proves
the accuracy of the proposed formulation.
(2) The comparison between the frame analyses with various joint
formulations and the experimental data demonstrates the signicance of the nonlinear loaddeformation joint behavior in the
frame response, especially for simple 2-D frames with low redundancy. The rigid joint assumption leads to completely different failure modes in a frame with weak joints. The proposed formulation,
implemented as joint-spring elements in the frame analysis, provides close predictions on both the failure modes and the ultimate
strength for 2-D and 3-D tested frames.
(3) The proposed formulations describe the loaddeformation
behavior of the axially loaded CHS X- and K-joints without fracture
failure. The incorporation of the fracture failure as reliable phenomenological representations in the frame analysis is the focus of a separate research effort [38].

1250

(b)

Global load (kN)

1000

900

750
600

Test
MSL
Rigid
Proposed

300
Load Case I
0

0.05

0.10

250
0.15

Load Case II
0

0.10

0.20

0.30

Global displacement (m)

Global displacement (m)

3000

Test
MSL
Rigid
Proposed

500

(c)

Global load (kN)

2000
Test
MSL
Rigid
Proposed

1000
Load Case III
0
0

0.05

0.10

0.15

0.20

0.25

Global displacement (m)

Fig. 13. Comparison of the global loaddeformation response between numerical analysis and experimental records for BOMEL 3D test: (a) Load Case I; (b) Load Case II; and (c) Load
Case III.

X. Qian et al. / Journal of Constructional Steel Research 90 (2013) 108119

Acknowledgment
We acknowledge the support of Lloyd's Register Foundation towards
funding the research and development program in the Centre for Offshore Research & Engineering in National University of Singapore. The
research scholarship provided by the National University of Singapore
is also gratefully acknowledged. The authors would like to extend
their appreciation to Professor Peter Marshall for providing the very
useful suggestions on the research.
References
[1] American Petroleum Institute (API). Recommended practice for planning, designing
and constructing xed offshore platforms. 21st ed. API RP2A-WSD; 2000.
[2] Bouwkamp JG, Hollings JP, Maison BF, Row DG. Effects of joint exibility on the response of offshore towers. 12th Offshore Technology Conference; 1981.
[3] Efthymiou M. Local rotational stiffness of unstiffened tubular joints. ReportPKER.85.199;
1985.
[4] Fessler H, Webster JJ, Mockford PB. Parametric equations for the exibility matrices
of single brace tubular joints in offshore structures. Proceedings of the Institution of
Civil Engineers; 1986. p. 65973.
[5] Fessler H, Mockford PB, Webster JJ. Parametric equations for the exibility matrices
of multi-brace tubular joints in offshore structures. Proceedings of the Institution of
Civil Engineers; 1986. p. 67596.
[6] Chen B, Hu Y, Tan M. Local joint exibility of tubular joints of offshore structures. Proc.
9th International Conference on Offshore Mechanics and Arctic Engineering; 2001.
[7] Kohoutek R, Hoshyari I. Parametric formula of rigidity for semi-rigid tubular T-joints.
Proc. 1st International Conference on Offshore Mechanics and Arctic Engineering; 1991.
[8] Romeijn A, Puthli RS, Wardenier J. Flexibility of uniplanar and multiplanar joints
made of circular hollow sections. Proc. 1st International Offshore and Polar Engineering Conference; 1991.
[9] Hellan O. Nonlinear pushover and cyclic analysis in ultimate limited state design and
reassessment of tubular steel offshore structures. Doctor dissertation The Norwegian
Institute of Technology; 1995.
[10] Skallerud B, Amdahl J. Nonlinear analysis of offshore structure. Baldock, Hertfordshire
(England): Research Studies Press; 2001.
[11] Chakrabarti P, Mukkamala A, Abu-Odeh I. Effect of joint behavior on the reassessment
of xed offshore platforms in the bay of Campeche, Mexico. 24th International Conference on Offshore Mechanics and Arctic Engineering; 2005. p. 13545.
[12] Mirtaheri M, Zakeri HA, Alanjari P, Assareh MA. Effect of joint exibility on overall
behavior of jacket type offshore platforms. Am J Eng Appl Sci 2009;2:2530.
[13] Hyde TH, Leen SB. Prediction of elasticplastic displacements of tubular joints under
combined loading using an energy-based approach. J Strain Anal Eng Des 1997;32:
43554.

119

[14] Leen SB, Pan W, Hyde TH. Investigation of an iterative sub-structure method for elastic and elasticplastic framework analysis. Comput Struct 2006;84:6908.
[15] Pan W, Leen SB, Hyde TH. Static analysis of frame by localized representation of
tubular joint. Proc. 12th International Conference on Offshore Mechanics and Arctic
Engineering; 2002.
[16] Ueda Y, Rashed SMH, Nakacho K. Flexibility and yield strength of joints in analysis
of tubular offshore structures. Proc. 5th International Conference on Offshore
Mechanics and Arctic Engineering; 1986.
[17] Ueda Y, Rashed SMH, Nakacho K. Improved joint model and equations for exibility
of tubular joints. J Offshore Mech Arct Eng 1990;112:15768.
[18] Choo YS, Qian XD, Foo KS. Nonlinear analysis of tubular space frame incorporating
joint stiffness and strength. 10th International Jack-up Platform Conference; 2005.
[19] Dier AF, Hellan O. A non-linear tubular joint response model for pushover analysis. Proc.
21st International Conference on Offshore Mechanics and Arctic Engineering; 2002.
[20] Dier AF, Lalani M. Strength and stiffness of tubular joints for assessment/design
purpose. 26th Offshore Technology Conference; 1995.
[21] Zettlemoyer N. Life extension of xed platforms. Tubular structure XIII; 2010 313.
[22] USOFS. USFOS course manual. Marintek SINTEF group; 2001.
[23] Van der Vegte GJ. The static strength of uniplanar and multiplanar tubular T- and
X-joints. Doctor dissertation The Netherlands: Delft University of Technology; 1995.
[24] Kurobane Y, Ogawa K, Ochi K, Makino Y. Local buckling of braces in tubular K-joints.
Thin-Walled Struct 1986;4:2340.
[25] Patran MSC. User's manual. MSC Software Corporation; 2010.
[26] ABAQUS. ABAQUS/Standard User's Manual Version 6.10-EF. Rising Sun Mills (USA):
Hibbitt Karlsson and Sorensen Inc.; 2010.
[27] Choo YS, Qian XD, Liew JYR, Wardenier J. Static strength of thick-walled CHS X-joints
part I. New approach in strength denition. J Constr Steel Res 2003;59:120128.
[28] IIW. Static design procedure for welded hollow section joints recommendations.
3rd ed. International Institute of Welding; 2008.
[29] Togo T. Experimental study on mechanical behavior of tubular joints. Doctor dissertation Japan: Osaka University; 1967.
[30] Greenwood PE, Nikulin MS. A guide to chi-squared testing. New York: Wiley; 1996.
[31] Lu LH, GDd Winkel, Yu Y, Wardenier J. Deformation limit for the ultimate strength of
hollow section joints. Tubular structure VI; 1994 3417.
[32] Bolt HM, Billington CJ. Result from large scale ultimate strength tests of K-braced
jacket frame structures. 26th Offshore Technology Conference. BOMEL Ltd; 1995.
[33] Bolt HM, Billington CJ, Ward JK. Result from large scale ultimate load tests on tubular
jacket frame structures. 25th Offshore Technology Conference. BOMEL Ltd; 1994.
[34] Choo YS, Qian XD, Wardenier J. Effects of boundary conditions and chord stresses on
static strength of thick-walled CHS K-joints. J Constr Steel Res 2006;62:31628.
[35] Kurobane Y, Ogawa K. New criteria for ductility design of joints based on complete CHS
truss tests. 5th International Symposium on Tubular Structures; 1993. p. 57081.
[36] Bolt HM, Billington CJ. Results from ultimate load tests on 3D jacket type structures.
31th Offshore Technology Conference; 2000.
[37] BOMEL. Brief description of 3D test set up and structural conguration. BOMEL
Reference C636\06\313R. BOMEL limited; 1999.
[38] Qian X, Zhang Y, Choo YS. A load-deformation formulation with fracture representation based on the J-R curve for tubular joints. Eng Fail Anal 2013;33:34766.