Proceedings of The Fifteenth (2005) International Offshore and Polar Engineering Conference

Seoul, Korea, June 1924, 2005

Copyright 2005 by The International Society of Offshore and Polar Engineers
ISBN 1-880653-64-8 (Set); ISSN 1098-6189 (Set)

Ultimate Strength Formulation for Axially Loaded CHS Uniplanar T-Joints

1

G.J. van der Vegte 1,2 and Y. Makino 1

Department of Architecture and Civil Engineering, Kumamoto University

Kumamoto, Japan
2
Faculty of Civil Engineering and Geosciences, Delft University of Technology
Delft, The Netherlands

moments (i.e. chord stress) on the overall strength of the T-joints. As a

result, for T-joints, the most governing parameters (chord stress and
chord length) are not included satisfactory in the current design codes.

ABSTRACT
As presented by Van der Vegte in 1995, the strength of axially loaded
uniplanar T-joints is governed by a combination of (local) joint failure
and failure due to (global) chord in-plane bending moments. The current
study, which is more rigorous compared to the research in 1995, presents
the effect of chord length, boundary conditions and chord end conditions
on the strength of uniplanar T-joints. Comparisons are made with the new
T-joint strength formulations developed by API (2003) and ISO (2004).
Finally, a simple ultimate strength equation is proposed for uniplanar Tjoints with compensating chord end moments. An interaction contour is
established to describe the effect of chord in-plane bending moments on
the strength of uniplanar T-joints.

In 1995, Van der Vegte proposed the following approach, using finite
element (FE) analyses, to separate and evaluate the two aforementioned
failure modes. A T-joint under brace axial load is numerically analyzed
twice : once excluding and once including the effect of chord in-plane
bending moments. In the first analysis, without the effect of chord inplane bending moments, compensating (negative) in-plane bending
moments are applied to the chord ends. By forcing the chord end
moments to be equal to those caused by to the brace load, the bending
moments at the cross sections at the crown points remain zero throughout
the loading history. The second analysis simply considers an ordinary
T-joint without chord end moments.

KEY WORDS: Circular hollow section; tubular; T-joint; steel;

numerical simulations; ultimate strength.

In the present study, further FE analyses have been conducted on

uniplanar T-joints under brace compression. Sixteen combinations of the
brace-to-chord diameter ratio and chord diameter-to-thickness ratio 2
have been numerically analyzed for five values of the chord length
parameter . In comparison to the research programme in 1995, the
current study considers a much wider array of geometric parameters ,
and 2. Material- and geometric non-linear FE analyses have been
conducted with the general purpose package ABAQUS/Standard (2003).

INTRODUCTION
In recent years, a number of references were published regarding the
effect of chord stress on the strength of tubular circular hollow section
(CHS) joints. Numerical studies carried out by Pecknold et al. (2000),
Van der Vegte et al. (2001) and Choo et al. (2003) identified the
influence of tensile or compressive chord pre-stress on the strength of
either thin- or thick walled X-joints for various types of brace load.
Other research presented by Pecknold et al. (2001) and Van der Vegte
et al. (2002) focused on the chord stress effect of uniplanar gap K-joints.
For the third type of the basic uniplanar joint configurations, the T-joint,
the evaluation of the chord stress effect is not as straightforward as for
X- and K-joints. Unlike the X-joint configuration, where axial brace
loads do not cause equilibrium induced chord loads, for uniplanar Tjoints, axial brace loads will lead to chord in-plane bending moments,
which will affect the joint strength.

The database derived from the FE analyses enables answers to the

following important questions. Is it indeed possible to exclude the effect
of the chord length parameter for uniplanar T-joints under brace load
by applying compensating chord end moments ? In other words : will the
strength of a T-joint become independent of ? A second issue addressed
in this study involves chord end plate effects. Unlike X-joints, for which
the interaction between chord length and end plate effects can be
determined rather straightforward, such an assessment is complicated for
axially loaded T-joints, because of the strong, but, at the same time,
unknown relation between local joint failure and failure due to chord
bending.

For uniplanar T-joints under axial brace load, several failure modes can
be identified. Depending on the chord length, failure may occur by
chord face plastification (local failure) or by a combination of overall
chord in-plane bending and shear (chord member failure). In most cases,
these two failure modes are strongly related. Experiments alone may
not provide the data necessary to quantify the effect of in-plane bending

Because of the increasing number of data available, especially numerical

data, improved strength formulations for tubular joints were proposed
recently.

279

- In the framework of the project to update the API RP2A (1993),

extensive series of FE analyses were conducted by Pecknold et al. (2001,
2002) as mentioned above. A relatively small part of this programme was
devoted to axially loaded uniplanar T-joints. The strength formulations
developed for T-joints, adopting the moment-free baseline approach
described above (API, 2003), will be included in the 22nd edition of API
RP2A.
- Based on a strictly screened experimental database, Dier and Lalani
(1998) presented a new set of ultimate strength equations for various
types of CHS joints to be incorporated in the new ISO 19902 (2004).

FE MODELLING
The numerical analyses were carried out with the finite element
package ABAQUS/Standard (2003). Due to symmetry in geometry and
loading, only one quarter of each joint has been modelled, whereas the
appropriate boundary conditions have been applied to the nodes in the
various planes of symmetry.
The joints are modelled using eight-noded thick shell elements
employing reduced integration (ABAQUS element S8R). Seven
integration points through the shell thickness are applied.

For both ISO and API, the strength equations developed for T-joints and
chord stress functions are briefly described and evaluated in section
Comparison of T-joint data with existing or proposed design rules.

For all joints except = 0.98, the geometry of the welds at the bracechord intersection has been modelled using shell elements. Previous
research (Van der Vegte et al., 1991) revealed that the use of shell
elements to simulate the welds, provides accurate predictions of the
load-displacement response of axially loaded uniplanar and multiplanar
T- and X-joints. The dimensions of the welds in the numerical model
are in accordance with the specifications recommended by the AWS
(1992).

Based on the complex expression derived from the analytical ring

model approach, Van der Vegte (1995) proposed an ultimate strength
equation for uniplanar T-joints including compensating chord end
moments. In the present study, a much simpler ultimate strength
equation is established. Furthermore, an interaction contour is presented
to describe the effect of the chord length and chord bending moments
on the strength of uniplanar T-joints.

Since the incorporation of material- and geometric non-linearity in

ABAQUS requires the use of true stress-true strain relationships, the
engineering stress-strain curve is modelled as a multi-linear relationship
and subsequently converted into a true stress-true strain relationship.
The hardening rule proposed by Ramberg-Osgood has been used to
describe the true stress-true strain behaviour. Both the engineering
stress-strain curve and the true stress-true strain relationship of S355
(the steel grade used in this study) are shown in Fig. 2.

RESEARCH PROGRAMME
The configuration of uniplanar T-joints and the definition of the
geometric parameters are illustrated in Fig. 1. The geometric parameters
of the T-joints considered are summarized in Table 1. For all joints, the
chord diameter d0 = 406.4 mm. Each of the 16 combinations of and 2
has been analyzed for the following five values of the chord length
parameter (= 2l0/d0) : 12, 16, 20, 24 and 28.

True stress - true strain curve

Engineering stress-strain curve

800

t1

Stress [N/m m ]

=2l0/d0
= d1/d0
2 = d0/t0
= t1/t0

N1

d1

l1 = 4d1

d0

600
S355

400

200

t0
l0
0

Fig. 1. Configuration of a CHS uniplanar T-joint.

10

20

30

40

Strain [% ]

Table 1. Geometric parameters analyzed for uniplanar T-joints

0.25
0.48
0.73
T1
T2
T3
2 = 25.4
T5
T6
T7
2 = 36.9
T9
T10
T11
2 = 50.8
T13
T14
T15
2 = 63.5
- all joints are analyzed for = 12, 16, 20, 24 and 28
- d0 = 406.4 mm

Fig. 2. Engineering and true stress - true strain curves for S355

Chord End Conditions

0.98
T4
T8
T12
T16

Various approaches, illustrated in Fig. 3, can be identified to restrain

the chord and apply chord end bending moments.
In Fig. 3a, the compensating in-plane bending moments are a result of
equal but opposite point loads at the top- and bottom nodes of the chord
end. End-plates are modelled to react the brace load. In the analyses by
Van der Vegte (1995), 40 mm thick, shell elements were used for the
end-plates. Although such end-plates are very stiff in-plane, out-ofplane deformations are still possible.
An alternative approach is shown in Fig. 3b, where rigid beams are

The steel grade used for the tubular members is S355 with fy = 355
N/mm2 and fu = 510 N/mm2.

280

attached to the chord end. For clarity, the last row of chord elements is
not displayed. Chord bending is modelled through a pure bending
moment applied to the reference node located in the centre of the chord
end cross section. Boundary conditions are applied to the reference
node to react the brace load. Because of the rigid behaviour of the
beams, the chord-end cross section cannot deform, but is able to rotate.
As this set of boundary conditions is expected to give a more consistent
description of the bending behaviour at the chord end as compared to
the approach shown in Fig. 3a, these boundary conditions are adopted
in the present study.

N1

M0,chord end

Point loads

Shell elements, t = 40 mm

=
Fig. 4. Loading scheme and moment distribution of T-joints

(a) Deformable end-plates

Numerical Results and Observations

For T-joints with compensating chord end moments, the primary failure
mode is chord face plastification. Because the chord axial stresses due to
in-plane bending moments have been eliminated, severe plasticity is
observed around the chord-brace intersection of each T-joint.

Rigid beam elements

Bending moment
Reference node

For each T-joint analyzed, a load-ovalization curve has been derived

whereas ovalization is defined as the vertical displacement of the crown
point minus the vertical displacement of the bottom surface of the chord
at mid-span. Fig. 5 illustrates the set of five load-ovalization curves
obtained for the T-joints with = 0.73 and 2 = 63.5. Ultimate load is
defined as the brace load corresponding to the peak in the loadindentation curve. The ovalisation at ultimate load may exceed the
deformation limit proposed by Lu et al. (1994).

(b) Rigid beams (last row of chord elements not shown)

Fig. 3. Various chord end restraints

ANALYSES ON T-JOINTS WITH COMPENSATING CHORD

END MOMENTS

From Fig. 5 it becomes clear that for this particular T-joint, the loadovalization curves are identical for values equal to or larger than 20.
Similar observations can be made from Fig. 6, depicting the ultimate
strength of T-joints ( = 0.73) with compensating moments against the
chord length parameter . The strength of the T-joints is independent of
if 20. This is not only noticed for = 0.73, but also for the other
values. It may therefore be concluded that a chord length of 10d0 is
sufficiently long to exclude any influence on the ultimate strength of Tjoints caused by chord end conditions.

Strategy
In order to exclude the effect of equilibrium induced chord in-plane
bending moments on the strength of uniplanar T-joints, compensating
(negative) in-plane bending moments are applied to the chord ends,
illustrated in Fig. 4. By forcing the chord end moments M0,chord end to be
proportional to the brace load N1 according to Eq. 1, the bending
moments at the cross sections beside the intersection of chord and brace
(i.e. crown points) remain zero throughout the loading history, depicted in
the moment distribution diagrams.
M 0,chord end = 0.25 N1 (l 0 - d1 )

For values less than 20, the effect of depends on the values of and
2. For the four values considered, the chord length parameter is not
seen to have any effect on the strength of the (thick walled) T-joints with
2 = 25.4, illustrated in Fig. 6 for = 0.73. Although not presented, for
the four joints with = 0.25, the joint strength is again independent of .

(1)

Because the moment M0,chord end is directly related to the brace load, the
chord end moments may lead to premature chord failure depending on
and 2. To avoid member failure at the chord ends taking place prior to
joint failure, the yield strength of the elements close to the chord ends is
increased. Only those elements for which the in-plane bending moments
at ultimate load would cause overall yielding, are reinforced. The zone of
high-strength elements is kept as small as possible to prevent possible
influence of the reinforced zone on the joint strength.

On the other hand, for T-joints with relatively short chords ( 16) in
combination with values exceeding 0.25 or 2 values larger than 25.4,
shorter chords cause an increase of ultimate strength due to the stronger
effect of chord end conditions. This influence becomes more significant
for increasing values of and 2, as chord ovalisation not only becomes
more severe (for larger values) but also dampens out more slowly along
the chord length (for increasing 2 values). Hence, the largest strength

281

enhancing effect is found for the T-joints with = 0.98 and 2 = 63.5.
Using the strength of the T-joint with = 28 as reference, a strength
enhancement of 10 % is found for the T-joint with = 12.

The influence of the chord length is demonstrated in Fig. 7, plotting the

ultimate strength of the T-joints with = 0.73 as a function of . Because
of the presence of chord stresses caused by the in-plane bending moments,
the chord length parameter has a strong effect on the strength of the Tjoints. Shorter chords lead to a rise in ultimate strength, not only caused
by a reduction in chord stresses, but also due to a more pronounced
influence of chord end conditions.

CHS T-joints with

chord end moments
= 0.73
2 = 63.5

24

32

2 = 63.5
2 = 50.8
2 = 36.9
2 = 25.4

CHS T-joints without

chord end moments
= 0.73

16

N1,u / fy0 t02

N1 / fy0 t02

32

= 28
= 24
= 20
= 16
= 12

may take place beside the crown points.

0
0.00

0.02

0.04

0.06

0.08

24

16

0.10

/ d0
8

Fig. 5. Load-ovalisation curves for T-joints ( = 0.73 and 2 = 63.5) with

compensating chord end moments

N1,u / fy0 t02

32

2 = 63.5
2 = 50.8
2 = 36.9
2 = 25.4

CHS T-joints with

chord end moments

= 0.73

20

25

30

Fig. 7. Ultimate strength of T-joints ( = 0.73) without compensating

chord end moments

Interaction Contour
Fig. 8 displays the interaction contour derived from the FE results
obtained for uniplanar T-joints. Fig. 8a presents all data points, while
Fig. 8b shows the data for 2 = 63.5. Interaction contours make it
possible to quantify the contribution of local joint failure and overall
chord bending failure. For the T-joints with compensating moments,
the joints with = 20 are taken as reference. Hence, on the vertical axis
of Fig. 8, Qf depicts the strength N1,u of each T-joint divided by the
strength N1,u,loc of the corresponding T-joint (with = 20) with
compensating chord end moments. As the in-plane bending chord
moments cause compressive stresses at the brace-chord intersection, the
moments on the horizontal axis are assumed as negative.

16

10

15

24

10

15

20

25

30

Fig. 6. Ultimate strength of T-joints ( = 0.73) with compensating chord

end moments

The following observations can be made :

- for increasing values of , overall chord bending failure becomes
more dominant. This is visualized in Fig. 8b where, for each value, a
line connects the data points obtained for different chord lengths. The
data point at the right refers to the T-joint with the smallest value of
(= 12), while the other end of each line marks the data for the joint with
the longest chord ( = 28).
- for increasing values, chord in-plane bending moments at failure
become larger. Hence, for large joints, the data points are located at
the left side of the interaction contours as shown in Fig. 8b.
- smaller values of 2 lead to relatively larger strength ratios Qf =
N1,u/N1,u,loc. A similar trend was also found for axially loaded uniplanar
X- (Van der Vegte et al., 2001) and K-joints subjected to axial chord
loading (Van der Vegte et al., 2002).
- for thick walled T-joints, it is observed that the moment ratio
M0,ipb/Mpl,0 can exceed -1.0, caused by strain hardening.
- another point of interest is the effect of the chord end restraints on the

ANALYSES ON T-JOINTS WITHOUT COMPENSATING

CHORD END MOMENTS
Numerical Results and Observations
For T-joints with equilibrium-induced chord bending moments, failure is
caused by a combination of (local) chord face plastification and (global)
chord in-plane bending failure. The contribution of each failure mode
depends on the geometric parameters , and 2. For thin walled Tjoints with a small value, chord face plastification clearly dominates
failure. The brace load causing local joint failure is relatively small and
not even sufficient to reach the yield moment in the chord member. On
the other hand, for thick-walled T-joints with large values for and , the
joints fail as a result of chord bending accompanied with pronounced
strain hardening. For thin-walled, large T-joints, chord wall buckling

282

interaction contours. As mentioned before, the strength ratios depicted

on the vertical axis are based on the ultimate strengths of the T-joints
with = 20, which is considered sufficiently long to exclude chord end
conditions. However, for T-joints including the effect of chord bending
moments, chord end effects cannot be excluded. This is especially clear
for the T-joints with large 2 values which are most sensitive for short
chord effects, thus explaining why these data points slightly move up
for relatively short chords ( = 12) as illustrated in Fig. 8b.

Nu =

f y0 t 02
sin

with the strength factor Qu for T-joints under brace compression given
as:
Q u = 2.8 + (20 + 0.8 ) 1.6 but 2.8 + 36 1.6

Q f = 1.0 + k1[

0.8
ISO

N y,0

]- k2 [

M 0,ipb
M pl,0

] k 3 [(

N 0,p 2
M0 2
) +(
) ]
N y,0
M pl,0

Qf

2 = 5 0 .8
2 = 3 6 .9

0.2
A PI

In the framework of a joint industry project, an extensive database of

available test data was compiled and strictly screened by MSL, after
which existing design equations were re-examined and new strength
formulations for uniplanar CHS joints were developed (Dier and Lalani,
1998). These formulations are to become part of the new ISO 19902
(2004). Based on the results included in the screened database, Dier and
Lalani developed the following functions to assess the characteristic
strength of uniplanar T-joints under brace axial compression :

0.0
-0.8

-0.6

-0.4

-0.2

(5)

ISO (2004)

2 = 2 5 .4

-1.0

(4)

The constants k1, k2 and k3 have different values for each brace load
condition and joint configuration. For T-joints under brace compression,
k1 = 0.3, k2 = 0.0 and k3 = 0.8. A factor of 1.18 should be applied to this
set of equations to obtain the mean strength values.

2 = 6 3 .5

0.0

M0,ipb/Mpl,0
(a) All data
1.0

0.8

Nu =

IS O
0.6

f y0 t 02
sin

(6)

Q u Qf

Qf

where the basic joint strength parameter Qu is given as :

Q u = (1.9 + 19 ) Q 0.5

= 0 .9 8

0.4

= 0 .7 3

API

Q = 1.0

for 0.6

= 0 .2 5

Q =

0.0
-1.0

-0.8

-0.6

-0.4

-0.2

(7)

with :

= 0 .4 8

0.2

-1.2

N 0,p

2
2
where M 0 = M 0,
ipb + M 0,opb

0.6

0.4

(3)

Eq. 2 shows the general format Qf to account for the effect of chord
stress on the ultimate strength :

1.0

-1.2

(2)

QuQf

0.0

M0,ipb/Mpl,0

0.3
(1 - 0.833 )

(8)
for > 0.6

Eq. 9 describes the factor Qf to account for chord axial load and chord
bending moments :

(b) Data for 2 = 63.5

Q f = 1.0 U 2

Fig. 8. Interaction contour for axially loaded uniplanar T-joints

(9)

where = 0.030 for brace axial load

= 0.045 for brace in-plane bending
= 0.021 for brace out-of-plane bending

COMPARISON OF T-JOINT DATA WITH EXISTING OR

PROPOSED DESIGN RULES
Proposed Update API RP2A (2003)

U 2 = k1 (

In 1991, the API Task Group for Tubular Connection Research initiated
a ten-year plan to upgrade the API RP2A. Extensive FE research by
Pecknold et al. (2000, 2001) was conducted to extend the available
database. As part of this project, a moment-free baseline for uniplanar
T-joints is proposed (API, 2003). After removing the factor of safety (=
1.60), the general format of the capacity equation is :

M 0, ipb 2
M 0,opb 2
N0 2
) + k2(
) + k2(
)
N y,0
M pl,0
M pl,0

(10)

The constants k1 and k2 are dependent on the joint configuration and

brace load conditions. For uniplanar T-joints under brace compression,
k1 = 25 and k2 = 11. A factor of 1.336 should be applied to this set of
equations to obtain the mean strength values.

283

not presented in Fig. 10, these T-joints have a large value ( = 0.73 or
0.98). As mentioned before, especially for these T-joints, the chord end
conditions are found to have a pronounced strength enhancing
influence.
- For many cases, the ISO equations overpredict the strength of the Tjoints analyzed, as can be concluded from the mean value of the FE/
prediction ratios given in Table 2. It may be concluded that for chord
bending moments, the reduction of the chord stress factor Qf is not
covered adequately, also suggested by Fig. 8a. As a result, when using
the ISO strength equations, a large number (24) of the T-joints are
predicted to fail by member failure (M0,ipb/Mpl,0 < -1). For these 24
joints, failure load is determined by assuming M0,ipb,u = Mpl,0 in the
chord cross sections beside the brace. Would the data for which ISO
predicts chord member failure have been removed from the database,
the comparison with the FE data would have been slightly worse, as
shown in the third column of Table 2.
- The performance of the API strength predictions on the FE database is
better than that of ISO. Not only is the average of FE/prediction ratios
for API larger than 1.0, but also the scatter is less, resulting in a smaller
value for the coefficient of variation in comparison with the ISO
predictions. In general, the strength predictions derived with the API
formulations are lower than the values obtained with ISO primarily
caused by the smaller values (i.e. stronger reductions) of APIs chord
stress factor Qf.

Illustration of API and ISO Strength Equations

For both the API and ISO, the strength of a uniplanar T-joint under
axial brace load is obtained by multiplying the basic joint strength
parameter Qu with the chord stress factor Qf. For both codes, the
formulation for Qu (including the mean bias factors) is illustrated in Fig.
9. Because the expression for Qu proposed by API, is dependent on 2,
a value of 2 = 36.9 is selected as example. The chord stress factors Qf
recommended for axially loaded T-joints without axial chord pre-load
are displayed in Fig. 8.
Fig. 9 shows that the API and ISO codes give different predictions for
the strength factor Qu. For large values, API provides larger values
for Qu than ISO. On the other hand, API proposes a much stronger
strength reduction for chord in-plane bending moments. For M0,ipb/Mpl,0
= -1.0, the ratio between the two chord stress functions even exceeds 3.
As a result, for the overall T-joint strength derived from the API code,
the higher values found for the strength factor Qu are partly
compensated by the lower values obtained for the chord stress function
Qf.
50

40

1.4

API

1.2

Qu

30

FE / ISO

1.0

20
ISO
10

0.8
0.6

2 = 63 .5
2 = 50 .8

0.4

0
0.0

0.2

0.4

0.6

0.8

2 = 36 .9
2 = 25 .4

0.2

1.0

2 = 14 .5

0.0
10

Fig. 9. Basic joint strength parameter Qu for 2 = 36.9

15

20

25

30

20

25

30

(a) ISO

Comparison with FE Data

1.4

In order to evaluate the strength predictions of API and ISO,

comparisions are made with the available FE data. Not only the 80 FE
data derived in this study are included, but also the 15 FE data (with
12 and 0.6) generated by Van der Vegte (1995). The strength
equations described by Eqs. 2-10 are used to calculate the predicted
strength for each of the 95 FE data points. In order to enable the
comparisons on the basis of mean values, mean bias factors of 1.18
and 1.336 are adopted for the API and ISO equations respectively.

1.2

FE / API

1.0
0.8
0.6

2 = 6 3 .5
2 = 5 0 .8

0.4

For both sets of equations, Fig. 10 plots the ratios between FE- and
predicted strength as a function of the chord length parameter , while
Table 2 summarizes the most relevant parameters obtained from the
comparisons. FE/prediction ratios < 1.0 imply that a code provides (unconservative) strength predictions, exceeding the values found in the
FE analyses.

2 = 3 6 .9
2 = 2 5 .4

0.2

2 = 1 4 .5

0.0
10

15

(b) API

The following observations can be made from Fig. 10 and Table 2:

- It is observed that for both API and ISO, the most conservative
predictions are obtained for the T-joints with relatively short chords (
= 12) in combination with large 2 values (2 = 50.8 or 63.5). Although

Fig. 10. FE data versus API or ISO strength predictions

284

Table 2. Comparisons between FE strength and predictions by API or

ISO
Mean

CoV

Maximum Minimum
ratio
ratio

95

1.067

0.077

1.31

0.94

95

0.941

0.104

1.30

0.80

71

0.921

0.110

1.30

0.80

60

40

30

NEW STRENGTH FORMULATION

20

T-joints with Compensating Chord End Moments

10

Based on the complex expression derived from the analytical ring

model approach, Van der Vegte (1995) proposed an ultimate strength
equation for uniplanar T-joints with compensating chord end moments
with the following format :
N1,u
f y0 t 02

2
A
D E
B C

0
0.0

(11)

0.6

0.8

1.0

T-joints without Compensating Chord End Moments

2
3.9
0.85 0.55
1 0.8

In general, various formulations such as linear equations or parabolic

functions of n (= M0,ipb/Mpl,0) can be used to describe the influence of
chord stress. In order to derive functions that are applicable to multiple
joint configurations of circular and rectangular hollow sections and
load types, the following equation is selected as basis for the regression
analysis:
f(n) = (1.0 - n A ) (B+ C + D)

The database consists of the 80 FE data points obtained in the present

study for the T-joints without compensating chord end moments as well
as 15 data points (with 12 and 0.6) analyzed by Van der Vegte
(1995). However, the following data are not included in the regression
analyses :
- Eq. 13 is invalid for M0,ipb/Mpl,0 < -1. In addition, the curves

(12)

representing Eq. 13 become very steep in the vicinity of M0,ipb/Mpl,0 =

-1. For data points located in this area, deviations between actual- and
predicted strength values will not only become large in percentage
terms but also in absolute sense, having a significant effect on the
final regression results. Hence, the data points with M0,ipb/Mpl,0 < 0.95 are excluded from the regression analysis.
- As mentioned before, for thin-walled T-joints with large values
and = 12 or 16, chord end conditions may enhance the strength of
the joints. As this effect is limited to these configurations, these data
are removed from the regression analyses. The final strength equation
is expected to underpredict the strength of these joints.

Table 3. Results of the regression analyses of T-joints with compensating

chord end moments (Eq. 12)
Mean
1.000

(13)

The same format was used by Liu et al. (2004) to describe the chord
stress effect of RHS T- and X-joints under axial brace load. Regression
analyses are conducted to determine the values of the regression constants
A, B, C and D.

For 2 = 14.5, Eq. 12 is displayed for 0.6 only. For large values (
> 0.9), Eq. 12 is expected to overestimate the strength for T-joints with 2
= 14.5. Evidence on axially loaded thick walled X-joints (Choo et al.,
2003) suggests that the strength becomes proportional to t0 instead of t02
observed for thin-walled joints. Furthermore, FE analyses on thick walled,
large T-joints with compensating chord end moments may be
influenced by chord end conditions as the zone of high-strength chord
elements modelled to prevent premature chord failure spreads over a
significant part of the chord member.

No. of data
33

0.4

Fig. 11. Ultimate strength of axially loaded T-joints with compensating

chord end moments

The 16 data points obtained in the present study for the T-joints with =
20 are used as input for the regression analyses. In addition, 13 data
points analyzed by Van der Vegte (1995) for which the chord is
considered sufficiently long to exclude the influence of chord end
restraints as well as four data on T-joints with = 0.91 [unpublished
results] are included in the regression analyses. The final ultimate
strength equation is given in Eq. 12, while the results of the regression
analyses are summarized in Table 3. Fig. 11 displays all data points as a
function of as well as the predictions obtained from Eq. 12.
f y0 t 02

0.2

In this equation, the terms A, B and C represent complicated functions of

the geometric parameters and 2, while D and E are simple regression
coefficients. In the current study, Eq. 11 is again used to serve as a basis
for the ultimate strength formulation. However, the terms A, B and C are
now simplified to scalars. Furthermore, in analogy with X-joints, B is
assumed as 1.

N1,u

CHS T-joints with

chord end moments

50

N1,u / fy0t02

API all data

ISO all data
ISO M0,ipb/Mpl,0 -1.0

Number
of data

2 = 63.5
2 = 50.8
2 = 36.9
2 = 25.4
2 = 14.5

CoV.
0.044

The final equation is given in Eq. 14, while the results of the regression
analyses are summarized in Table 4.
f(n) = (1.0 - n )(0.1 + 0.008 )

285

(14)

- In general, the ultimate strength predictions derived with the ISO

formulae are higher than the values obtained with API. Because of the
relatively large values of the chord stress factor Qf (i.e. the reduction is
not severe enough), ISO may lead to un-conservative strength predictions.
- Eq. 12 is proposed to describe the ultimate strength of uniplanar T-joints
with compensating chord end moments, while Eq. 14 represents the
interaction contour between local joint failure and (global) member
failure. The combined formulation of Eqs 12 and 14 provides accurate
strength predictions for uniplanar T-joints under axial brace load.

with n = M0,ipb/Mpl,0.
Table 4. Results of the regression analyses for the interaction contour
(Eq. 14)
No. of data
69

Mean
0.994

CoV.
0.055

In order to assess the performance of the new strength equations, the

combined formulation of Eqs. 12 and 14 is used to predict the strength of
the 95 T-joints in the FE database. The results are summarized in Table 5
and displayed in Fig. 12. A very good correlation between the predicted
and actual strength values is observed. As expected, the strength
equations give the largest underpredictions for the T-joints affected by
chord end conditions.

REFERENCES
ABAQUS/Standard (2003). Version 6.4, Hibbitt, Karlsson and
Sorensen, Inc, USA.
American Petroleum Institute (1993). Recommended Practice for
Planning Designing and Constructing Fixed Offshore Platforms Working Stress Design, API RP2A-WSD, 20th Edition.
American Petroleum Institute (2003). Proposed Updates to Tubular Joint
Static Strength Provisions in API RP2A 21st Edition.
American Welding Society (1992). Structural Welding Code, AWS
D1.1-92.
Choo, YS, Qian, XD, Liew, JYR, and Wardenier, J (2003). Static
Strength of Thick-Walled CHS X-Joints Part II. Effect of Chord
Stresses, J of Constructional Steel Research. Vol 59, No 10 pp
1229-1250.
Dier, AF, and Lalani, M (1998). New Code Formulations for Tubular
Joint Static Strength, Proc 8th Int Symp on Tubular Structures,
Singapore, pp 107-116.
International Organization for Standardization (2004). Petroleum and
Natural Gas Industries - Fixed Steel Offshore Structures, ISO DIS19902.
Liu, DK, Wardenier, J, and Vegte, GJ van der (2004). New Chord Stress
Functions for Rectangular Hollow Section Joints, Proc 14th Int
Offshore and Polar Engineering Conf, Toulon, France, Vol IV, pp
178-185.
Lu, LH, Winkel, GD de, Yu, Y, and Wardenier, J (1994). Deformation
Limit for the Ultimate Strength of Hollow Section Joints, Proc 6th Int
Symp on Tubular Structures, Melbourne, Australia, pp 341-347.
Pecknold, DA, Ha, CC, and Mohr, WC (2000). Ultimate Strength of DT
Tubular Joints with Chord Preloads, Proc 19th Int Conf on Offshore
Mechanics and Arctic Engineering, New Orleans, USA.
Pecknold, DA, Park, JB, and Koppenhoefer, KC (2001). Ultimate
Strength of Gap K Tubular Joints with Chord Preloads, Proc 20th Int
Conf on Offshore Mechanics and Arctic Engineering, Rio de Janeiro,
Brazil.
Vegte, GJ van der, Koning, CHM de, Puthli, RS, and Wardenier, J (1991).
Numerical Simulation of Experiments on Multiplanar Tubular Steel
X-Joints, Int J Offshore and Polar Engineering, ISOPE, Vol 1, No 3,
pp 200-207.
Vegte, GJ van der (1995). The Static Strength of Uniplanar and
Multiplanar Tubular T- and X-Joints, Doctoral Dissertation, Delft
University of Technology, Delft, the Netherlands, Delft University
Press, ISBN 90-407-1081-3.
Vegte, GJ van der, Makino, Y, Choo, YS, and Wardenier, J (2001).
"The Influence of Chord Stress on the Ultimate Strength of Axially
Loaded Uniplanar X-Joints," Proc 9th Int Symp on Tubular
Structures, Dsseldorf, Germany, pp 165-174.
Vegte, GJ van der, Makino, Y, and Wardenier J. (2002). The Effect of
Chord Pre-Load on the Static Strength of Uniplanar Tubular KJoints, Proc 12th Int Offshore and Polar Engineering Conf,
Kitakyushu, Japan, Vol IV, pp 1-10.

Table 5. Results of comparisons between FE strength and predictions by

Eqs. 12 and 14

Eqs. 12 & 14

Number
of data

Mean

CoV

95

1.000

0.053

Minimum Maximum
ratio
ratio
0.91

1.18

1.4

FE / Eqs.12 &14

1.2
1.0
0.8
0.6

2 = 63 .5
2 = 50 .8

0.4

2 = 36 .9
2 = 25 .4

0.2

2 = 14 .5

0.0
10

15

20

25

30

Fig. 12. FE data versus strength predictions by Eqs 12 and 14.

SUMMARY AND CONCLUSIONS
FE analyses are conducted on uniplanar T-joints under axial brace load.
Sixteen combinations of the geometric parameters and 2 are
analyzed for five values of the chord length parameter . Each T-joint
is analyzed with and without the presence of compensating chord end
moments. The ultimate strength of the T-joints without chord end
moments is compared with the strength predictions of the API (2003)
and ISO (2004). Based on the results, the following conclusions can be
drawn :
- The approach of applying compensating moments to the chord end
with the purpose to eliminate equilibrium induced chord bending
moments is confirmed to provide ultimate strength values for T-joints
not affected by the chord length parameter .
- For axially loaded uniplanar T-joints with compensating moments, a
chord length of 10d0 ( = 20) is sufficiently long to exclude the influence
on the ultimate strength caused by chord end conditions. For shorter
chords, chord end restraints may affect (i.e. enhance) the joint strength,
depending on the values of and 2.
- For both API and ISO, the ultimate strength of T-joints under axial
brace load is obtained by the combined formulation of the basic strength
factor Qu and the chord stress function Qf. However, API and ISO may
recommend substantially different values for the components Qu and Qf.

286