Journal of Constructional Steel Research 84 (2013) 1–16

Contents lists available at SciVerse ScienceDirect

Journal of Constructional Steel Research

Bending behavior and design model of bolted flange-plate connection

Y.Q. Wang ⁎, L. Zong, Y.J. Shi
Key Laboratory of Civil Engineering Safety and Durability of China Education Ministry, Department of Civil Engineering, Tsinghua University, Beijing 100084, China

a r t i c l e i n f o a b s t r a c t

Article history: This paper focuses on the bending behavior of flange-plate connections under pure-bending and aims for
Received 5 April 2012 putting forward a practical design model. Four basic types of bolted flange-plate connections are tested
Accepted 25 January 2013 and related finite element analysis is implemented. The finite element model is verified by experimental
Available online xxxx
results and proved to be precise and reliable. Based on the finite element analysis, the distribution of von-
Mises strain and contact pressure at end plates of the connections is revealed. The valuable information
Keywords:
Bending behavior
can be directly used in the theoretical model to present a relatively clear yield line mechanism and defi-
Bolted flange-plate connection nite pressure center. The bending capacity determined by flange-plates is derived with the virtual work
Design model principle. It is proved that the theoretical model can give a good prediction for the yield bending capacity
Experimental study of the connections. Meanwhile, traditional T-stub analogy is introduced to obtain the bending capacity
FEA determined by bolts. Combining with the two different design models and assuming that the end plates
should fail before high strength bolts, a practical design procedure is put forward. The connections
designed according to this procedure will meet the demand of safety and economy. Furthermore, the de-
sign model herein can provide useful reference for practical design of other kinds of bolted flange-plate
connections.
© 2013 Elsevier Ltd. All rights reserved.

1. Introduction Each type of connection could have stiffener or not [3]. Thus, four
basic types are specified and shown in Fig. 2. It should be noticed
With long span buildings and high-rise buildings constantly emerg- that the rectangular hollow section in Fig. 2 is a square hollow section
ing, steel tubular structures with unique advantages are increasingly (SHS) particularly, which is commonly used in tubular structures.
widely used. Bolted flange-plate connections become more and more Many full-scale experiments were taken on different types of the
popular as an important solution of splicing steel tubular struc- connections. Kato et al. took a series of experimental investigations
tures. Fig. 1 shows a small but typical flange-plate connection in steel on unstiffened connections in axial loading, including 63 specimens
tubular structure. The connections provide easy and fast installation, for CHS [4] and 36 specimens for RHS [5]. The key parameters such
nice shape, simple constitution without the need for field welding [1]. as diameter and the thickness of end plates were paid high attention
Typical loading situations for the connections include axial tension by comparing different specimens' behavior. Igarashi focused on the
and bending moment. Axial loading often occurs in steel tubular trusses, connections for CHS and took experimental study on 15 unstiffened
while bending moment often appears when tower structures bear later- specimens [6] and 7 stiffened specimens [3]. Afterwards, J.A. Packer
al load such as wind or earthquake [2]. The loading capacity under axial [7] and X.W. Gou [8] tested more specimens and assembled different
tension has been studied by many scholars [1]. However, study on experimental data to verify related design codes for flange-plate
flexural capacity of flange-plate connections is limited. This situation connections.
restricts the generalization of the nice connections. Thus, it is important Other approaches, including theoretical study and numerical simu-
to do further study on the behavior of flange-plate connections under lation had been applied to the study on the tensile capacity of flange-
bending moment. plate connections besides experiments. Typical theoretical methods
Fortunately, lots of completed work on flange-plate connections were T-stub analogy and yield line theory. T-stub model was put for-
under axial load has laid a good foundation for further study. Now- ward by J.H.A. Struik and J.de Back [9] and is often used to consider
adays, the connection can be generally classified into two types — the prying action of the connections. Many scholars used this model
flange-plate connections splicing circular hollow sections (CHS) and directly to calculate the prying force. As for yield line theory, many
flange-plate connections splicing rectangular hollow sections (RHS). researchers such as Kato [5], Igarashi [3,6], Packer [10,11] and J.J.Cao
[12] used different yield line mechanisms to develop suitable design
⁎ Corresponding author at: Department of Civil Engineering, Tsinghua University,
models. Some models were ushered into related design codes such as
Beijing 100084, China. Tel.: +86 10 6278 8623; fax: +86 10 6278 8623. AISC Hollow structural sections connections manual [13], and provided
E-mail address: [email protected] (Y.Q. Wang). a good prediction of the connections' bearing capacity under axial load.

0143-974X/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jcsr.2013.01.012
2 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

to take FEA on flange-plate connection is to choose suitable element

types and simulate the possible contact faces precisely. Usually, 3-D
solid element and contact pair should be used by some powerful gen-
eral FEA software [14].
Thanks to the solid work done by experimental, numerical and
theoretical methods, the prediction of the tensile capacity for flange-
plate connection was complete and reliable. As for study on the bending
capacity, Wheeler et al. [15,16] had done significant work which
focused on the connections for RHS with four bolts arranged at corners.
However, the behavior of the connections could vary enormously with
the change of the connections' geometrical parameters. The research
on four-bolt connections was not enough for the practical demand.
This paper focuses on the bending behavior of flange-plate con-
nections with eight bolts under pure bending (Fig. 2). Four specimens
have been carefully tested and each one is corresponding to each
basic type shown in Fig. 2. Afterwards, related FEA is done and the
results obtained from experiments and FEA are compared with each
other. It is proved that FEA was precise and reliable. With the help
of accurate FE models, the bending behavior of the connections can
Fig. 1. Picture of a small but typical flange-plate connection. be revealed better and directly used in corresponding design model.
Combining with the T-stub analogy and yield line theory, the design
procedure for the connections under bending moment is proposed.
Some others were very precise but too complicated to be applied in As it is impossible to do research on the features of all flange-
practical design [1]. plate connections with different geometrical parameters, the speci-
Finite element analysis (FEA) was also widely used in the related fied study aims to present a general method to predict the con-
study. Many important features which could not be measured in nections' bending capacity and provide valuable reference for the
experiments conveniently might be obtained from FEA. The key point design of other kinds of flange-plate connections.

Do Do
D e1 e 2 D e1 e2
Di Di

End
Tube End
Plate Tube Plate

Stiffener

(a) Unstiffened flange-plate connections (b) Stiffened flange-plate connections

splicing CHS splicing CHS

ew ef Stiffener ew ef
End Plate End Plate
hl ew ef

ew ef
ho

ho
c

c
hi

c
hi

Tube Tube
hl

hi hi

c c
h h
ho ho

(c) Unstiffened flange-plate connections (d) Stiffened flange-plate connections

splicing SHS splicing SHS
Fig. 2. Four basic types of flange-plate connections with 8 bolts.
 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16 3

Notation εu The steel ultimate strain;

φ Be equal to π/n;
B Secant modulus of the measured Moment-Midspan dis- θAB, etc. Rotation of line LAB, etc.;
placement curves; γ Tensile capacity index of one plate zone.
Btube The bending stiffness of a pure steel tube;
c Distance between adjacent bolts parallel to hollow section;
3. Experimental study
D External diameter of circular tube, D/2 is denoted by r;
Do External diameter of end plate for CHS;
3.1. Specimen preparation
Di Inner diameter of end plate for CHS;
di Arm force of the ith plate zone;
Four specimens have been tested in total. As an exploratory study,
d0 Diameter of bolt hole;
the main aim of the experiments is to observe the connections' bend-
e1 Distance from bolt line to circular hollow section face;
ing behavior and check the reliability of the related FEA and design
e2 Distance from bolt line to end plates' edge of circular
model. Therefore, the number of specimens is not so much. However,
section;
the specimens are carefully selected according to practical demand.
es Distance from bolt line to stiffener face;
The dimension of tube and end plates are very common in steel tubu-
ew Distance from bolt line to square hollow section face;
lar structures, and their details are shown in Table 1. The section was
et Distance from bolt line to end plates' edge of square
designed to avoid local buckling. Full penetration welds are applied
section;
between the tube and end plates. For stiffened connections, the thick-
f Design value of the steel yield stress;
ness of stiffener is 6 mm and the welds between stiffener and tube
fy Measured steel yield stress;
are fillet welds with 6 mm leg size.
fu Measured steel ultimate stress;
High strength bolts with grade 8.8S are used in the connection,
h External height of square tube for SHS;
whose nominal yield stress is 640 Mpa. According to the testing report
ho External height of end plate for SHS;
provided by the bolt factory, the actual yield stress of the bolt is
hi Inner height of end plate for SHS;
660 Mpa and the ultimate stress is 800 Mpa. The preload of the bolts
hl Be equal to (h − c)/2;
is 125 kN and the distance from the edge of the plate to the center of
lAB, etc. Length of line LAB, etc.;
the bolt is set to be 36 mm according to Ref. [17]. All holes are standard
mp Full plastic moment of end plate per unit length;
holes of 26 mm for M24 bolts [18]. The material of tube and end plates
mp′ Moment at first yield of end plate per unit length;
are different among the four specimens so that the influence of material
n Total number of bolts;
could be considered, and all the material property is tested according to
t0 Thickness of the steel tube;
the procedure mentioned in Ref [19]. The material property of the tubes
t Thickness of end plates;
and end plates (t = 18 mm) are tested by the authors and the material
tc Thickness of end plate when no prying action occurs;
of end plates (t = 10 mm) and end plates (t = 16 mm) are tested by the
td Thickness of end plate when the plate is in double
factory. The material details are listed in Table 2.
curvature;
M Given bending moment which the connection should
transfer; 3.2. Experimental program
Mtube The bending capacity of the tube;
My, My,F, My,E Yield bending capacity of the connections obtained As shown in Fig. 3, four beams which were spliced by the flange-
from theoretical model, FEA and experiment; plate connections were tested under four-point loading condition. The
Mu,E Ultimate bending capacity of the connections obtained length of the beam was 2 m. The two jack loading points were right at
from experiment; the quarter of the beam and the connection was located at the middle
N Tension force which a bolt should transfer; of the beam, resulting in that the connection was subjected to pure-
Nmax Maximum tensile force of bolts; bending. Two jacks with 60 t loading capacity could work well in syn-
Ntb Design value of the bolt tensile capacity; chronism with each other as their source of power was from the same
Q Prying force; channel. To investigate the variation of bolt force, four gauged bolts
T Tensile capacity of plate zone; were also introduced into each connection.

Table 1
Dimensions and materials of specimens and test results.

Flange-plate connections for CHS

Specimen Tube Plate dimension Bolt parameter Material of tube Material of My,E (kN · m) Mu,E (kN ·m) Mu,E/My,E Failure mode
dimension flange-plate

D t0 Do Di t e1 e2 c

J1(type a) 168 10 312 118 16 36 36 160.3 Q345B Q345B 79.2 96.05 1.21 Weld fracture
J2(type b) 168 10 312 118 10 36 36 160.3 Q345B Q235B 89.9 113.03 1.26 Excessive
deformation

Flange-plate connections for SHS

Specimen Tube Plate dimension Bolt parameter Material of tube Material of My,E (kN · m) Mu,E (kN ·m) Mu,E/My,E Failure mode
dimension flange-plate

h t0 ho hi t ew ef c

J3(type c) 200 5 344 150 18 36 36 75 Q235B Q345B 105.1 122.1 1.16 Weld fracture
J4(type d) 200 5 344 150 10 36 36 75 Q235B Q235B 97.8 120.2 1.23 Excessive
deformation
4 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

Table 2 strain of the end plates in J1 and J3 which is shown in Fig. 6 is over
Measured material property of all the tubes and end plates in the test. 2200 με before the welds fracture. It means the end plates have
E/N/mm2 fy/N/mm2 fu/N/mm2 εu already reached yield status and so the experimental results are still
valuable for connection design.
CHS 2.12e5 380 524 0.137
SHS 1.96e5 306 443.1 0.137 As for the stiffened connections, failure occurred when the longitu-
End plate(t = 10 mm) 2.06e5 290 420 0.29 dinal deformation of the end plates were deemed excessive. Residual
End plate(t = 16 mm) 2.06e5 390 530 0.29 deformation of end plates was about 10.0 mm (Fig. 7 (a)) in J2 and
End plate(t = 18 mm) 2.04e5 410 561 0.29
10.3 mm in J4 (Fig. 8 (a)). The connections showed good ductility.
High strength bolts 2.01e5 660 800 –
For the four specimens, the measured strain of gauged bolt which
transferred the maximum tension force was settled and converted
3.3. Experimental results and observations into bolt load. The bolt load curve is presented in Fig. 9. Two impor-
tant points could be obtained directly from the figure as follows:
With the incremental load, the load–displacement curve for each
a. Yield strength of the bolt is 289 kN based on that its yield strength
specimen is obtained and presented in Fig. 4. It should be noticed
of material is 640 Mpa and its diameter is 24 mm. All the bolts of
that Jack load is converted into bending moment and the midspan
the specimens haven't reached the yield status. It means that the
displacement is average value of LVDT-1 and LVDT-2. The point on
yield capacity of all the connections is determined by the end
the curve shows the step of the loading program. The bi-linear yield
plates and the result can be used to verify the design model in
load approximation method [20] is shown in the figure. The peak
Section 4.2.
point of the curve is regarded as ultimate moment point. The values
b. It is proved that the prying action exists in all the specimens. Ob-
of yield moment My,E are recorded in Table 1. As shown in this
viously, prying action exists in stiffened connections, but its value
table, failure of the unstiffened connections for CHS and SHS occurs
is quite less than the unstiffened connections. The conclusion
when the full penetration welds between the tube and end plates
should be considered in the related design model.
fracture (see Fig. 5 (a) and (b)). The failure mode is unexpected
because the ultimate load of the specimen is not limited to the con- As recorded in Table 2, the ratio of Mu,E over My,E is about 1.2. This
nection failure completely. Fortunately, strain gage is distributed on value can represent the safety margin of the connections to some
the tensioned side of each end plate of unstiffened connection. The di- extent. What's more, the value of the stiffened connections is larger
rection which the strains is measured is along the longer edge of the than that of the unstiffened connections. It shows that the safety mar-
gauge, which is perpendicular to the radial direction of CHS. The mea- gin of the stiffened connections is better. One possible reason is that
sured strain of the gage can represent the deformation of end plates the stiffener and related fillet welds can share the load of penetration
to some extent. Fig. 6 shows the location of the gage and load–strain welds. It should be noticed that the unstiffened connections should be
curve. The trend of the curve is similar for FEA and test. The measured more carefully designed and given more safety margin.
Furthermore, the stiffness of the specimens spliced by connections
should be focused on after discussion on the strength of the connections.
The qualitative property of the stiffness of the specimens can be illus-
trated by Fig. 4. Secant modulus of the measured Moment-Midspan
displacement curves noted by B can represent the bending stiffness of
the whole specimens. Meanwhile, the bending stiffness of a pure steel
tube noted by Btube can be easily obtained according to structural
60 ton Jack Gauged Bolt
mechanics. The contribution of the connections can be revealed by
Simple comparing B and Btube. It is shown that the stiffness of the specimens
Support Loading
decreases as the load increases. However, this change is not obvious be-
Support fore the load gets to yield moment. After the load exceeds the yield load,
the stiffness of the specimens decreases rapidly. When the load gets to
LVDT-2 LVDT-1 ultimate moment, the connections can be regarded as hinge approxi-
mately. So, as for the connections, what is the design demand for its stiff-
ness? An easy way to judge whether the stiffness of the connections
meets the design demand is proposed here, that is the connections
(a) Testing Set for CHS should make positive contribution on the stiffness of the specimen
before the load gets to My. It's required that the secant modulus of the
Moment-Midspan displacement curve before My is larger than Btube.
All the specimens in this paper don't meet the demand as we focus on
is the capacity of the connection instead of members. We design the
specimens with weaker connection on purpose to reveal the collapse
of the connections sooner and better, in case the tube's collapse or stiff-
60 ton Jack ness softening has effect on the behavior of the connections. The exper-
Gauged Bolt
imental results prove that the design on the specimens is reasonable.
Simple
Support 4. Numerical simulation

4.1. Finite element models

LVDT-2

LVDT-1 Numerical simulation was implemented using finite element

method. Finite element model was established by ANSYS software.
(b) Testing Set for SHS The stress–strain relationship for the end plates and tubes was
taken as multi-linear model with yield plateau (Fig. 10). The strain
Fig. 3. Testing set. at the yield plateau εst was taken as 2.5% for Q235 [21] and 2.26%
 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16 5

100 Mu,E 120

Mu,E
Mu,F Mu,F
My,E 100
75 Experiment
My,F FEA My,E Experiment

Moment/kN·m
Pure tube 80
Moment/kN·m

Arbitrary secant line My,F FEA

Btube Mu,E,My,E Pure tube
50 B Mu,F,My,F 60 Arbitrary secant line
Btube B Mu,E,My,E
Mu,F,My,F
40
Btube =96EI/11L2
25 Btube =96EI/11L2
20

0 0
0 10 20 30 40 50 0 10 20 30 40 50
Midspan Displacement/mm Midspan Displacement/mm
(a) J1 (b) J2

125 Mu,E 125 Mu,F

Mu,F Mu,E
My,E My,F Experiment
100
100 My,F Experiment My,E FEA
FEA Pure tube
Moment/kN·m

Pure tube Arbitrary secant line

Moment/kN·m

75 Arbitrary secant line 75 Mu,E,My,E

Btube
Mu,E,My,E Mu,F,My,F
Mu,F,My,F
B 50
B
50
Btube
Btube =96EI/11L2 Btube =96EI/11L2
25 25

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30 35
Midspan Displacement/mm Midspan Displacement/mm
(c) J3 (d) J4
Fig. 4. Comparison between the load–displacement curve obtained from test and FEA.

for Q345 [22]. For high strength bolts, strain hardening initiated at the welds fracture can't be simulated and the ultimate moment of the
the end of the elastic limit. The ultimate strain εu was taken as unstiffened connections calculated by FEA should not be compared
0.136 [14] and the other data needed could be found in Table 2. with that obtained in tests. Thus, only yield moment is taken as a stan-
In the modeling herein, all elements of the bolts were meshed by dard index to do the comparison as shown in Table 3. The bi-linear yield
SOLID 95 and all elements of end plates and stiffeners were meshed load approximation method is also used to obtain My,F.
by SOLID 92. To simulate the preload of bolts, PREST179 was intro- Besides the comparison of yield bending capacity, comparison
duced. The model was exactly designed based on the experimental on the deformation of end plates in test and FEA was implemented.
specimens and boundary conditions. To reduce computing time, the Figs. 7 and 8 show the comparison of the ultimate deformation of
model for SHS was based on half of the entire geometric model for end plates in connections J2 and J4. The deformation obtained from
simplification applying symmetry. However, as two bolts of connec- test and FEA is found to be very similar with each other.
tions for CHS were located in the symmetry axis, the model had to For unstiffened connections, the ultimate deformation can't be
be based on an entire geometric model (Fig. 11). simulated as the weld is weaker than end plates. Although the end
The key to simulate the behavior of the connections was to consid- plates have reached yield status, their deformation is not obvious
er the contact problem appropriately. Contact pairs with 3-D target and hard to be measured. So comparison on the deformation of end
surface elements TARGE 170 and 3-D contact elements CONTA 174 plates in J1 and J3 has to be given up.
were created at all possible contact interfaces as shown in Fig. 12.
4.3. Discussion on distribution of contact stress
4.2. Comparison of results by FEA and experiments
The aforesaid comparison items prove that the agreement between
Comparison of results by FEA and experimental load–displacement numerical simulation and experiment is very close. What's more, the
curves is shown in Fig. 4. As welds are not considered in the FE model, numerical predictions of the bending capacity from FEA are slightly
6 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

200

Welds
150
Fracture

Jack Load/kN
100

50

0
p1-1 p1-3
(a) Failure mode of J1 0 500 1000 1500 2000 2500 3000
Average Strain of p1-1 and p1-3/10-6
(a) Load-strain curve of J1
250
Welds
Fracture 200

Jack Load/kN 150

100

50
(b) Failure mode of J3
0
Fig. 5. Failure mode of unstiffened connections. p3-1
0 500 1000 1500 2000 2500
conservative. It can be concluded that the FE model herein can simulate Strain of p3-1/10-6
the connection behavior well. The slight discrepancy may be caused by
several reasons, such as the fabrication error of the specimen, neglect of (b) Load-strain curve of J3
welds and inaccurate preload of the bolts. It can be found that the initial
Fig. 6. Load–strain curve for unstiffened connections.
preload of the bolt is not exactly equal to the standard preload 125 kN
from Fig. 9. With the help of FEA, some interesting features of bending
behavior can be inspected. For connections bearing bending moment,
it is important to find its pressure center. In the pressed region of here considers the effect of both flexural yielding of end plates and
flange-plate connection, the pressure is transferred depending on con- yielding of high-strength bolts. Firstly, the bending capacity controlled
tact of end plates rather than high-strength bolts. Thus, the distribution by end plates and bolts is derived separately, which is presented in
of contact stress between end plates is the key to find the pressure center. Sections 4.2 and 4.3. Secondly, a design procedure is shown in Section
With ANSYS, the distribution of contact press at different load step can be 4.4. The procedure here is based on four basic assumptions below:
obtained. Figs. 13 and 14 show that the distribution of contact press for J1
and J2 bearing 10% yield moment, 50% yield moment and 100% yield mo- a. As for the whole specimen, the flange-plate connection should be
ment. It can be obviously seen that the distribution of contact stress have stronger than the HSS members which it splices. Only the bending
little change, which means that the pressure center can be considered to capacity of the connection will be focused on here, while, it is nec-
be constant. Thus, the pressure center of flange-plate connection can be essary to check whether the capacity of HSS members is less than
easily obtained at any load step using FEA. This information is valuable that of the connection.
for the related design model for connections. The feature is also fit for b. The failure of end plates is always ductile and foreseeable, but the
connections of SHS, which is not shown here for paper space saving. failure of bolts is brittle and more dangerous. Thus, the bending
Besides pressure center, FEA can provide other valuable information capacity of the connection is assumed to be controlled by the
such as the distribution of von-Mises strain of end plates, which can end plates instead of high strength bolts. It means that the end
describe the outline of yield line for the design model applying yield plates will fail before the bolts. What's more, it's assumed that
line theory. The details will be presented in the next section. end plates below the geometrical neutral axis become fully plastic
when end plates are regarded as failure.
5. Design model c. The effect of welds is not considered here. It is required that full
penetration welds should be used between end plates and tubes
5.1. Basic assumption and the strength of the welds should not be lower than the
strength of base metal of steel.
Both the experimental and numerical study on flange-plate con- d. Enough safety margins should be prepared in the design model of all
nections are aimed to propose an effective design model. The model kinds of connections, especially for the unstiffened connections.
 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16 7

Expand
10.3mm

(a) Test (a) Test

(b) FEA (b) FEA

Fig. 7. Comparison of ultimate deformation of connection J2. Fig. 8. Comparison of ultimate deformation of connection J4.

5.2. Bending capacity determined by end plates where T is the tensile capacity of each zone and δ is virtual displacement
generated at the junction of tube and end plates. It is assumed that no
Yield line theory, a traditional kinematic method, is used here to displacement is generated at the location of bolts. mp is the plastic
derive the bending capacity determined by end plates. The key to moment of the flange plate per unit length and it's given by
apply the yield line theory is to determine appropriate yield line
mechanism and the pressure center of end plates under bending mo- 1 2
mp ¼ f t ð2Þ
ment. It's relatively easy to be done with the help of precise FEA re- 4 y
sults when the load of the connection gets to My,F. Usually, the end
Thus, T can be written as
plate is divided into several plate zones to simplify the calculation for-
mula. There are 3 fundamental steps in the whole derivation of bend- ∑li θi
1 2
ing capacity. Firstly, derive the tensile capacity of plate zone based on T¼ f t γ; where γ ¼ ð3Þ
4 y δ
yield theory. Secondly, determine the pressure center and the arm of
force of each zone. Lastly, multiply the tension capacity of each plate γ is a dimensionless index. Here, the yield line mechanism has been
zone by the arm of force to obtain the bending capacity. The detailed calculated [8] and its tensile capacity index γ is given by
derivation will be illustrated by taking the unstiffened flange-plate

connection of CHS as example. It will be shown in next chapter. r þ e1 π
γ¼8 tan ð4Þ
e1 2n
5.2.1. Unstiffened flange-plate connections for CHS
The FEA results when the load of the connection gets to My,F To derive the equation above, analytic geometry method is used in
and design model for unstiffened flange-plate connections of CHS is literature [8]. As is shown in Fig. 15 (c), an individual plate zone is
shown in Fig. 15. Fig. 15 (a) presents the distribution of von-Mises strain taken out for study and a user-defined coordinate system is built.
for end plates. The yield line mechanism in Fig. 15 (c) can be obtained. It The coordinates of all key points in the coordinate system can be
can be found that there are 3 whole plate zones and 2 half plate zones. easily obtained. A(r,0,δ); B(rcosφ, − rsinφ, δ); C(rcosφ, rsinφ, δ);
The dashed area in Fig. 15 (c) is an individual zone for study. An D(r + e1,0,0); E(xcosφ, − xsinφ, 0); F(xcosφ, xsinφ, 0). Then, the nor-
enlarged figure of the dashed area is provided with key points and mal vector of each surface can be obtained and given by
yield lines, which is note by dashed line. The following equation can
→
be obtained according to the virtual work principle: NABC ¼ ð0; 0; 1Þ ð5Þ


→ 1− cosφ
NACD ¼ δ; δ; e1 ð6Þ
Tδ ¼ mp ∑li θi ð1Þ sinφ
8 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

240 240

200 200
Measured Bolt Load/kN

Measured Bolt Load/kN

Q
160 160
Q
120 120

80 80

40 40
o o
45 45
0 0
0 50 100 150 200 250/kN 0 50 100 150 200 250/kN
Tension force the bolt should transfer /kN Tension force the bolt should transfer /kN
(a) J1 (b) J2

240 240

200 200
Measured Bolt Load/kN

Measured Bolt Load/kN

160 Q 160 Q
120 120

80 80

40 40
45o 45
o

0 0
0 50 100 150 200 250/kN 0 50 100 150 200 250/kN
Tension force the bolt should transfer /kN Tension force the bolt should transfer /kN
(c) J3 (d) J4
Fig. 9. Variation curve of the maximum bolt load.


→ 1− cosφ Thus, γ can be derived according to Eq. (3). As the kinematic method
N ABD ¼ δ; − δ; e1 ð7Þ
sinφ is based on the upper bound theory, a minimum γ should be used.
When x equals to r + e1, γ gets γmin, which is shown in Eq. (4).


→ r þ e1 −x cosφ ðr þ e1 Þðr−xÞ The first step of the whole derivation, which is to derive the tensile
N CDF ¼ δ; − δ; ð8Þ
x sinφ x capacity of plate zone, is already finished. The next step is to determine
the pressure center and the arm of force of each zone. As for the pres-
With these normal vectors, the rotation angle of each line, which sure center, it's set to be the top of the tube according to the distribution
is the included angle of two surfaces containing the line, can be calcu- of contact pressure presented in Fig. 15 (b), which is a little lower than
lated. As a result, the rotation of all yield lines is given by the average pressure center and more conservative for the sake of safe-
ty. Thus, the arm of force for each plate zone can be shown in Fig. 15(d).
δ Multiply the tension capacity of each plate zone by the arm of force. The
jθAB j ¼ jθAC j ¼ ð9Þ bending capacity denoted as My is given by
e1 cosðφ=2Þ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi My ¼ T ðd1 þ 2d2 þ d3 Þ ð14Þ

jx−r−e1 j ðr þ e1 Þ2 þ r 2 −2ðr þ e1 Þr cosφ
jθCD j ¼ jθBD j ¼ δ ð10Þ
e1 ðr þ e1 Þðx−rÞ sinφ 5.2.2. Stiffened flange-plate connections for CHS
Just as the unstiffened connections for CHS, all needed information
2 tanðφ=2Þ for calculation is obtained by FEA. The yield line mechanism shown in
jθAD j ¼ δ ð11Þ
e1 Fig. 16(c) is more simplified than that descripted in Ref. [3]. It should
be noticed that line LCF is parallel with line LDQ and LBE is parallel with
2jx−ðr þ e1 Þ cosφj line LDP. The coordinates of key points are given by
jθCF j ¼ jθBE j ¼ δ ð12Þ
ðr þ e1 Þðx−r Þ sinφ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Aðr; 0; δÞ; Bðrcosφ; −rsinφ; δÞ; C ðrcosφ; rsinφ; δÞ; Dðr þ e1 ; 0; 0Þ;
x2 þ ðr þ e1 Þ2 −2ðr þ e1 Þx cosφ Eððr þ e1 þ e2 Þcosφ; −ðr þ e1 þ e2 Þsinφ; δÞ;
jθED j ¼ jθFD j ¼ δ ð13Þ
ðr þ e1 Þðx−r Þ sinφ F ððr þ e1 þ e2 Þcosφ; ðr þ e1 þ e2 Þsinφ; δÞ
 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16 9

5.2.3. Unstiffened flange-plate connections for SHS

For unstiffened flange-plate connections of SHS, the distribution
Stress/MPa
fu
of von-Mises strain is not so clear that the yield line mechanism
can be easily determined. Many mechanisms put forward by former
fy researchers are very complicated. Herein a simplified model is proposed
as shown in Fig. 17 (c). The yield line mechanism conforms roughly
with the distribution of von-Mises strain. It should be paid attention
E that line LDE, LAC and LBC are not yield lines, which are noted by solid
line. What's more, the displacement of point C in Fig. 17 (c) is assumed
to be zero. Actually, the plates will contact with each other very tightly
at the corner so the assumption is reasonable. Thus, the coordinates of
all key points can be given by

0 Aðh=2; h=2; δÞ; Bðh=2 þ ew ; c=2; δÞ; C ðho =2; ho =2; 0Þ; Dðh=2; 0; δÞ;
fy/E εst εu Eðh=2 þ ew ; 0; δÞ:
Strain/10-6
(a) Constitutive model of flange-plates and tubes The length and rotation of each yield line can be derived as
follows:

lBE ¼ c=2; θBE ¼ δ=ew ð20Þ

Stress/MPa

fu
lAD ¼ h=2; θAD ¼ δ=ew ð21Þ
fy qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 δef ⋅lAB
lAB ¼ e2w þ hl ; θAB ¼   ð22Þ
ew ⋅ðew þ hl Þ⋅ ew þ ef
E
Then the tensile capacity of each plate zone is given by
 
1 2 hþc ef ⋅ ew 2 þ hl 2
T ¼ f t γ; where γ ¼ þ   ð23Þ
4 2ew ew ⋅ðew þ hl Þ⋅ ew þ ef

0
fy/E εu Take the top of tube as the pressure center just like the unstiffened
Strain/10-6 connections for CHS. The average pressure center is a little above the
top of the tube and the design model will be conservative. The bend-
(b) Constitutive model of high-strength bolts ing capacity of the connections can be given by

Fig. 10. Constitutive model.

My ¼ T ð2d1 þ 2d2 Þ ð24Þ

The coordinates of P and Q are not necessary for the derivation.

The length and rotation of each yield line can be derived as follows: 5.2.4. Stiffened flange-plate connections for SHS
As unstiffened connections for SHS, the distribution of von-Mises
lAB ¼ lAC ¼ rφ; θAB ¼ θAC ¼ δ=e1 ð15Þ strain of stiffened connections is not so clear that the yield line mech-
anism can't be easily obtained (Fig. 18). The mechanism used here is
also roughly corresponding to the results of FEA. The coordinates of
lBE ¼ lCF ¼ e1 þ e2 ; θBE ¼ θCF ¼ δ=½ðr þ e1 Þ sinφ ð16Þ all key points in Fig. 18(c) are given by

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Aðh=2; 0; δÞ; Bðh=2; h=2; δÞ; C ðho =2; 0; δÞ; Dðho =2; h=2; δÞ;
lBD ¼ lCD ¼ ðr þ e1 −r cosφÞ2 þ ðr sinφÞ2 ; Eðho =2; c=2; 0Þ; Oðh=2 þ ew ; c=2; 0Þ:
θCD θBD 2φ e1 rð2− cosφÞ ð17Þ
¼ ¼ 8 cotφ sin þ þ
δ δ 2 ðr þ e1 Þ sinφ e1 sinφ The length and rotation of each yield line can be derived as follows:

lPD ¼ lQD ¼ e2 secφ; θPD ¼ θQD ¼ δ=½ðr þ e1 Þ sinφ ð18Þ lAB ¼ h=2; θAB ¼ δ=ew ð25Þ

According to Eq. (3), the tensile capacity of each plate zone is lBD ¼ ew þ ef ; θBD ¼ 2δ=hl ð26Þ
given by
lAC ¼ ew þ ef ; θAC ¼ 4δ=c ð27Þ
1 2 ∑li θi lAB 2l þ 2lPD 2l θ
T ¼ f y t γ; where γ ¼ ¼ þ BE þ CD CD ð19Þ
4 δ e1 ðr þ e1 Þ sinφ δ lOE ¼ ef ; θOE ¼ 2δ=c þ δ=hl ð28Þ

Taking the nearest bolt line above the top of tube as the pressure qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
center, the bending capacity can be given based on Eq. (14). lBO ¼ ew 2 þ hl 2 ; θBO ¼ δ ew 2 þ hl 2 =ðew hl Þ ð29Þ
10 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

(a) J1 (b) J2

(c) J3 (d) J4
Fig. 11. FE models.

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 5.2.5. Modeling verification

lAO ¼ ew 2 þ c2 =4; θAO ¼ 2δ ew 2 þ c2 =4=ðew cÞ ð30Þ
As specified above, the yield capacity of the specimens in the test
is determined by end plates and could be used to verify the yield line
Then the tensile capacity of each plate zone is given by mechanism in the design model. Comparison of results from theoret-
    ical model and experimental load–displacement curves is shown in
1 2 6 ew þ ef 3 ew þ ef h Table 4, and all the yield moment data points obtained by tests, FEA
T ¼ f y t γ; where γ ¼ þ þ ð31Þ and the theoretical model are presented in Fig. 19. It shows that the
4 c hl ew
design model can give a desired safe prediction of bending capacity.
The probable reason for that the theoretical results are lower than ex-
Taking the nearest bolt line above the top of tube as the pressure perimental results is the assumed yield line mechanism according to
center, the bending capacity can be given based on formula (24). FEA is not so accurate. The distribution of von-Mises strain for all con-
nections shows that yield region occurs except for yield lines. The
actual bending capacity should be larger than the calculated value
which only considers the boundary line of the yield mechanism.
Contact between nut
Contact between
and end plates
screw and hole wall

Table 3
Comparison of yield moment obtained from FEA and experiments.

Contact Specimen Test (kN ·m) FEA (kN · m) FEA/test

between end J1 79.2 75.1 0.95

J2 89.9 80.1 0.89
plates
J3 105.1 103.3 0.98
J4 97.8 100.1 1.02
Fig. 12. Possible contact interfaces.
 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16 11

(a) 10% My,F (b) 50% My,F (c) 100% My,F

Fig. 13. Development of distribution of contact pressure between flange-plates for J1.

(a) 10% My,F (b) 50% My,F (c) 100% My,F

Fig. 14. Development of distribution of contact pressure between flange-plates for J2.

Average Pressure
Center

Top of Tube

(a) FEA results – Distribution of von-Mises (b) FEA results – Distribution of

Strain Contact Pressure

O
y
= /n
Geometrical neutral axis
r

x
d3

O
d2

O
d1

B C
A T/2 T/2
B C
T T
e1

A
T
E F E F
D D

(c) Yield line mechanism (d) Calculation schematic diagram

Fig. 15. Design model for unstiffened connections of CHS.

12 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

Bolt
line

Yield
Region
(a) FEA results – Distribution of (b) FEA results – Distribution of
von-Mises Strain Contact Pressure

O
y
= /n

r
Geometrical neutral axis
x

d3
B C

d2
O

d1
y A
x T/2
e1
T/2
B C
A
D T T
T
e2
D E F
P Q
E F
P Q

(c) Yield line mechanism (d) Calculation schematic diagram

Fig. 16. Design model for stiffened connections of CHS.

Average
Pressure Center

Top of Tube

(a) FEA results – Distribution of (b) FEA results – Distribution of

von-Mises Strain Contact Pressure

O
y
h/2

Geometrical neutral axis x

h/2
d2

O
d1

D A
y
x T T
ew

D A
E B T T
c/2 hl
ef

B
E
C C

(c) Yield line mechanism (d) Calculation schematic diagram

Fig. 17. Design model for unstiffened connections of SHS.
 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16 13

Bolt line

(a) FEA results – Distribution of von-Mises (b) FEA results – Distribution of

Strain Contact Pressure
O
y
h/2
x
h/2

c/2 hl
Geometrical neutral axis A B

d2
d1
O
y
ew

x T T
O
T T
ef

C E D
(c) Yield line mechanism (d) Calculation schematic diagram
Fig. 18. Design model for stiffened connections of SHS.

5.3. Bending capacity determined by high-strength bolts where b is the average width of the plate zone (see Fig. 21) and β is
the net sectional coefficient. The index α is the ratio of the sagging mo-
Bending capacity determined by high-strength bolts is seriously ment at the bolt line (M2) to the hogging moment at the stem of the
affected by prying action. As specified above, both the unstiffened tee (M1). To calculate Q, the value of α is the key. M1 can be given by
connections and the stiffened connections have prying action. For
unstiffened connections, the prying action is more obvious and
should be paid more attention. Thus, the design model of unstiffened 1 2
M1 ¼ ft b ð34Þ
connections will be focused on firstly. 4
As for the first step of the model, prying action is not considered
and distribution of the load of bolts group is triangular according to
related codes [17]. The load which should be transferred by a bolt
can be calculated. Then a half T-stub model is applied according to 105
Ref. [9]. The model is shown in Fig. 20. The method proposed by FEA results
100 J4
Results obtained by FEA and

J. H. A. Struik and J. de Back [9] is applied to derive the prying action. Theoretical results
theoretical method /kN·m

The following basic equations have been given by the two scholars:
95
ð1 þ αβÞM 1 ¼ Ne1 ð32Þ J3
90
αβ e1
Q¼ ð33Þ 85
1 þ αβ e2

80
Table 4
J2
Comparison between yield moments obtained by tests and theoretical calculation. 75
Specimen Theoretical model My/kN ·m Test My,E/kN · m My/My,E J1
70
J1 73.5 79.2 0.93 70 75 80 85 90 95 100 105
J2 80.3 89.9 0.89
J3 100.8 105.1 0.96 Test results /kN·m
J4 88.8 97.8 0.91
Fig. 19. Yield moment data.
14 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

Q The practical thickness of end plate is usually between tc and td.

e2(ef)
Then choose an appropriate t and calculate the index α by combining
Eqs. (35) and (36):
B M2

e1(ew)
"
 #
M1 1 tc 2 N
α¼ −1 ð38Þ
N β t Nbt

Fig. 20. Mechanical model of bolts. Then put α obtained by Eq. (38) into Eq. (33). The prying force can
be calculated.
For SHS, ew and ef need to be used instead of e1 and e2. The design
Combining Eqs. (32) and (34), t can be given as model for unstiffened connections has been derived until now. As
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi for the stiffened connections, the prying action is not so obvious.
4Ne1 The model should be simple. Taking the following two points into
t¼ ð35Þ
bf ð1 þ αβÞ consideration:
a. The prying action is not obvious due to the stiffeners as they share
Usually, the value of α is between 0 and 1. One extreme case is part of tensile load.
that α is 0. It represents that the prying action won't occur. If the ten- b. Each yield line in a yield line mechanism shares part of tensile load.
sion of bolt gets up to Ntb and α is set to be 0, the maximum thickness
of end plate tc can be obtained; Actually, T-stub analogy describes a yield line mechanism con-
taining two yield lines which is shown in Fig. 21(a) and (c). An indi-
sffiffiffiffiffiffiffiffiffiffiffiffiffi
vidual plate zone noted by dashed area can be used to explain the
4N bt e1
tc ¼ ð36Þ following method. The two yield lines, L1 and L2, share the tensile
bf
load N, which is the load causing prying action. However, yield lines
will occur along the stiffener, which can also share the tensile load
Another extreme case α is 1 when the prying forces is really sub-
for the stiffened connections. So the tensile load which causes prying
stantial and the end plate is in double curvature. The related thickness
action will decrease. The modified models for stiffened connections
td can also be derived:
are shown in Fig. 21(b) and (d). The two yield lines, L1 and L2, share
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi N s in the modified model, which can be given by:
4Nbt e1
td ¼ ð37Þ
bf ð1 þ βÞ l1 θ1 þ l2 θ2
Ns ¼ N
l1 θ1 þ l2 θ2 þ l3 θ3 þ l4 θ4
l1 ðδ=e1 Þ þ l2 ðδ=e1 Þ bðl1 þ l2 Þ
For CHS; Ns ¼ N¼ N
l1 ðδ=e1 Þ þ l2 ðδ=e1 Þ þ 2l3 ð2δ=bÞ bðl1 þ l2 Þ þ 4e1 2
2
2bðδ=ew Þ 2b
For SHS; N s ¼ N¼ 2 N
2bðδ=ew Þ þ 2ew ð2δ=bÞ 2b þ 4ew 2
ð39Þ

Putting Ns into formula (38) instead of N, the prying force can be

b=(L1+L2)/2 b=(L1+L2)/2 calculated and used to check the capacity of the bolts.

L1 L1 5.4. Design procedure

L3 L4
L2 L2 Based on the above two design models determined by end plates
and high-strength bolts separately, a practical design procedure is pro-
posed here (Fig. 22). It should be noticed that the procedure is based on
(a) T-stub model used for (b) Modified model used for the basic assumptions descripted in Section 4.1. First of all, the proce-
dure can ensure that the connection is stronger than the tube because
stiffened connection of CHS stiffened connection of CHS
the thickness of end plates is calculated according to the bending capac-
ity of the tube. The procedure can also ensure that the end plates will fail
before the bolts by taking full account of prying action. What's more, the
moment at first yield of end plate per unit length denoted as mp′ is used
instead of mp as shown in formula (2). mp′ is given by

1 2
0 ft 1 2
mp ¼ 4 ¼ ft ð22Þ
1:5 6
b b Thus, the tensile capacity of each plate zone will be given as
L1 L1
L3 L4
L2 L2
1
f t2 γ 1 2
T¼4 ¼ ft γ ð23Þ
1:5 6

(c) T-stub model used for (d) Modified model used for where f is the design value of yield strength based on related design
codes. Obviously, the actual yield moment of the connections will
stiffened connection of SHS stiffened connection of SHS
be 1.5 times the design value. It can meet the demand of safety and
Fig. 21. Calculation schematic diagram for Ns. economy appropriately.
 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16 15

Fig. 22. Design procedure.

6. Conclusions As known to all, the behavior of the connections might vary

enormously with the change of geometrical parameters. How-
ever, it should be noticed that the theoretical model in this
(1) A practical design procedure is proposed in this paper. The pro-
paper can provide useful reference for other kinds of bolted
cedure combines two design models which are based on yield
flange-plate connections.
line theory and T-stub analogy separately. The model based on
yield line theory can give an accurate prediction of yield bend-
ing capacity except for stiffened connections of CHS. The model Acknowledgements
based on T-stub analogy can derive the value of prying forces.
What's more, the model can be slightly revised and simplified The writers gratefully acknowledge the support for this work,
to calculate prying forces of stiffened connections. With the de- which was funded by the Natural Science Foundation of China (No.
sign procedure herein, the bolted flange-plate connections can 51038006).
meet the demand of safety and economy in practical engineer-
ing. References
(2) It is found the ratio of Mu,E over My,E is about 1.2 and the safety [1] Willibald S, Packer JA, Puthli RS. Experimental study of bolted HSS flange-plate
margin of the stiffened connections is better. It should be no- connections in axial tension. J Struct Eng 2002;128(3):328–36.
ticed that the unstiffened connections should be more carefully [2] GB50135-2006. Code for design of high-rising structures. BeiJing: China Planning
Press; 2006 [In Chinese].
designed and given more safety margin.
[3] Igarashi S, Wakiyama K, Inoue K, Matsumoto T, Murase Y. Limit design of high
(3) Finite element analysis is an effective method to simulate the strength bolted tube flange joints Part 2. Joint with rib-plates and ring-stiffeners.
behavior of the connections. It is proved that the results Proceedings of A.I.JTokyo: Architectural Institute of Japan; 1985. p. 71–82
[In Japanese].
obtained from FEA in this paper are identical with that
[4] Kato B, Hirose R. Bolted tension flanges joining circular hollow section members.
obtained from tests in both capacity and deformation. With J Constr Steel Res 1985;5(2):79101.
the help of FEA, the behavior of the connections is better re- [5] Kato B, Mukai A. Bolted tension flanges joining square hollow section members.
vealed. The pressure center and a relatively clear yield line J Constr Steel Res 1985;5(3):163–77.
[6] Igarashi S, Wakiyama K, Inoue K, Matsumoto T, Murase Y. Limit design of high
mechanism are found and they give a strong support for the strength bolted tube flange joints Part 1. Joint without rib-plates and
theoretical model. ring-stiffeners. , Proceedings of A.I.JTokyo: Architectural Institute of Japan; 1985.
(4) The model in this paper is limited to bolted flange-plate con- p. 52–66 [In Japanese].
[7] Willibald S, Packer JA, Puthli RS. Experimental study of bolted flange-plate con-
nection with eight bolts. What's more, the layout of the bolts nections for square hollow section tension members. Proceedings of ISOPE;
should meet the Chinese Code for design of steel structures. 1999. p. 46–53.
16 Y.Q. Wang et al. / Journal of Constructional Steel Research 84 (2013) 1–16

[8] Gou XW. Research on bolted flange-plate connections of tubular structures. Research [16] Wheeler AT, Clarke MJ, Hancock GJ. FE modeling of four-bolt, tubular moment
Institute of Building and Construction; MCC Group Beijing; 2010 [In Chinese]. endplate connections. J Struct Eng 2000;126(7):816–22.
[9] Struik JHA, de Back J. Tests on Bolted T-stubs with respect to a Bolted Beam-to-Column [17] GB 50017–2003. Code for design of steel structures. Beijing: China Planning Press;
Connection. Stevin Laboratory Report 6-69-13. Delft, The Netherlands: Delft Univ. of 2003.
Technology; 1969. [18] GB/T 1228–2006. High strength bolts with large hexagon head for steel struc-
[10] Packer JA, Bruno L, Birkemoe PC. Limit analysis of bolted RHS flange plate joints. tures. Beijing: China Standard Press; 2006 [In Chinese].
J Struct Eng 1989;115(9):2226–42. [19] GB/T 228–2002. Metallic materials tensile testing at ambient temperature. Beijing:
[11] Willibald S, Packer JA, Puthli RS. Design recommendations for bolted rectangular China Standard Press; 2002 [In Chinese].
HSS flange-plate connections in axial tension. Eng J 2003;40(1):15–24. [20] Kosteski N, Packer JA, Puthli RS. A finite element method based yieldload determi-
[12] Cao JJ, Packer JA, Yang GJ. Yield line analysis of RHS connections with axial loads. nation procedure for hollow structural section connections. J Constr Steel Res
J Constr Steel Res 1998;48(1):1–25. 2003;59(4):453–71.
[13] American Institute of Steel Construction (AISC). Hollow structural sections con- [21] Wang GZ. Theory and Design of Steel Structures [M]. Beijing: Tsinghua University
nections manual; 1997 [Chicago]. Press; 1993 [In Chinese].
[14] Shi G, Shi YJ, Wang YQ. Numerical simulation of steel pretensioned bolted end-plate [22] Shi WL. Experimental and theoretical study on semi- rigid beam-to-column com-
connections of different types and details. Eng Struct 2008;30(10):2677–86. posite joints with flush end-plate connection [D]. Shanghai: Tongji University;
[15] Wheeler AT, Clarke MJ, Hancock GJ, Murray TM. Design model for bolted moment end 2006 [in Chinese].
plate connections joining rectangular hollow sections. J Struct Eng 1998;124(2):
164–173s.