Journal of Constructional Steel Research 65 (2009) 228–236

www.elsevier.com/locate/jcsr

Research on end-plate connection with non-completely penetrated welds

Yiyi Chen a,∗ , Sufang Wang b,a
a College of Civil Engineering, Tongji University, 1239 Siping Road, Shanghai, 200092, PR China
b Shanghai University of Electric Power, 2103 Pingliang Road, Shanghai, 200090, PR China

Received 25 July 2006; accepted 15 February 2008

Abstract

A series of tests were executed for H-shaped beams whose flanges were connected to end-plate by non-completely penetrated welds. The welds
were designed by equal capacity rule and the beams were mainly composed of non-compact or slender elements. The parameters introduced are
width–thickness ratios of beam flange and web, end-plate thickness, angle between flange plane and end-plate plane and so on. 24 pieces of
specimens were tested to failure under monotonic loading and 6 under cyclic loading. For comparison, a few of the specimens are those with
completely penetrated groove welds. Finite element (FE) analysis was also carried out to study the effect of parameter change on the resistance of
the welds. The parameters involved end-plate thickness, weld size, column flange thickness and bolt gauge. The results show that thinner end-plate
and greater bolt gauge will deteriorate the resistance of weld. As long as the components are properly designed, as done in tests, the connections
will not fail at the welds no matter what kind of weld form is adopted. In addition, weld form has no obvious effect on the ultimate loads and
cyclic behavior of specimens. It is confirmed that the non-completely penetrated welds connecting flanges of H-shaped beams and end-plates can
be used both under static loading and seismic loading in low-rise building structure where local buckling of the member is the predominant factor.
c 2008 Elsevier Ltd. All rights reserved.

Keywords: Portal frame; End-plate connection; Weld; Static and cyclic test; Finite element analysis

1. Introduction bolted end-plate connections figured in FEMA350 [8] where

the flange is welded to end-plate by completely penetrated
Extended end-plate connections have been widely used in groove weld. In addition, Chinese technical specification [5] for
steel frames, especially in low-rise non-residential buildings. light-weight buildings with gabled frames requests the beam
In the case that the beams or columns are composed of non- flange welded to the end-plate by full penetration weld, no
compact or slender elements (hereafter called NCSE members), matter what kind of member section is used.
there are still some issues to be discussed about the welds Full penetration weld is indeed necessary to ensure the
connecting beam flanges and the end-plate. complete development of plastic moment capacity for beams
In the region where earthquake risk should be taken into and to prevent the connection from failure prior to the members
account in structural design, there is strict requirement on if the beam section is composed of compact element as required
welding quality. The welds, which connect beam flanges to by ANSI/AISC 360-05 [3]. However, when the NCSE member
column flange, are often required to be full penetration weld. is used, the moment capacity of beams is not expected to
For example, prequalified connections are demanded to be used exceed yield moment or plastic moment, shall it still be
in special moment frames (SMF) and intermediate moment necessary? When using fully penetrated welds, steel plates
frames (IMF) by the provisions in ANSI/AISC 341-02 [2]. need to be grooved according to the thickness and more
Although there is no clear clause to specify the weld details expensive examination is also required compared with fillet or
connecting elements of beam to end-plate for the above- non-penetrated weld. Therefore, simple and economic weld is
mentioned frames in AISC specifications, the same requirement desirable if it can fulfill its structural function.
is implied in other documents. One case is the prequalified Some investigations have been made on weld details
for beam-to-column joint. Morris [13] mentioned that the
∗ Corresponding author. Tel.: +86 21 6598 2926; fax: +86 21 6598 4976. connections of portal frame should be prepared with fillet
E-mail address: yiyichen@mail.tongji.edu.cn (Y. Chen). welds, but he did not present relevant study to support the

c 2008 Elsevier Ltd. All rights reserved.

0143-974X/$ - see front matter
doi:10.1016/j.jcsr.2008.02.009
 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236 229

conclusion. Witteveen et al. [10] indicated that the failure of Table 1

the weld between the end-plate and the tension flange of the Geometrical parameters of beams
beam must be prevented prior to yielding of any other part Series ID Beam size (mm) Width–thickness ratio
of the connection. Murray [14] recommended that the full hb bfb twb tfb bfb /2tfb h wb /twb
capacity of tension flange should be developed through using A 316 200 5 8 12.5 60
either full-penetration or fillet welds even if the full bending B 616 200 5 8 12.5 120
strength of the beam is not used. Adey et al. [1] investigated C 324 200 8 12 8.3 37.5
the effect of welding process on the behavior of extended D 624 200 8 12 8.3 75
end-plate connections, but there was no significant difference
observed. Dubina et al. [7] tested 54 pieces of T-type assembly
Table 2
specimens using the following three weld types: fillet weld, Mechanical properties of beams
double bevel weld and single bevel weld. The results showed
Series ID Mechanical properties (MPa)
that double bevel weld specimens exhibited satisfying behavior,
f yf f uf f yw f uw
single bevel weld specimens had unsatisfactory behavior due to
incomplete penetration at the root of the weld, and fillet weld A 570 627 411 523
B 365 540 377 548
specimens were characterized by an intermediary behavior. C 330 522 365 540
The research program of Azuma et al. [4] mainly focused D 387 520 336 519
on the influence of weld defects in welded beam-to-column
joints. Test study and finite element analysis all indicated that
brittle fracture was unlikely to be initiated by defects in partial Table 3
penetrated groove weld reinforced by additional fillet weld. Section parameters of beams
In a word, these previous researches provided valuable data Series ID Flange Web
and opinions, but did not involve the relation between the λp λr λp λr
requirements for the weld and the capacity of NCSE members. A 7.2 15.5 84.2 127.6
The present study attempts to take one step forward. In B 9.0 16.3 87.9 133.2
this study, the bolted end-plate connection of NCSE beam to C 9.5 22.9 89.3 135.4
D 8.8 17.8 93.1 141.1
column, which is commonly used in steel frame of low-rise
building, will be examined to confirm whether non-completely
penetrated welds connecting flanges of NCSE beam to end- 2.2. Bolts and end-plate
plate satisfy the design requirements under seismic condition.
In the first phase of the study, monotonic tests and extended Since the study focuses on the welds connecting beam flange
FE analysis were carried out, by which basic conclusion can be and end-plate, the capacity of the bolts was designed to be
made. Then in the second phase, cyclic loading tests were done greater than the yield moment of beam section. All bolts were
to check the cyclic behaviors. of 10.9 grade in Chinese GB standard (ultimate strength of 1000
MPa), with a diameter of 20 mm for series A, B and 24 mm for
2. Design of test specimen series C, D. The pretension of one bolt was about 60% of its
ultimate tension capacity, and the coefficient against slippage
2.1. NCSE beam between friction surfaces was set as 0.5.
End-plate thickness, tep , was decided based on the yield
A series of cantilever beams with built-up section were line theory and the assumption that the bolt is just in its
designed. According to width–thickness ratio of the plates, the design ultimate state when yield line mechanism forms. For
beam sections were grouped into four series named A, B, C each section series, the thickness was designed into two sizes
and D. The geometrical parameters of the beams are listed in through two design criteria. Criterion I takes the plastic moment
Table 1, where h b , bfb , twb , tfb are section depth, flange width, of plate section while criterion II takes the yield moment,
web thickness and flange thickness respectively, h wb is the although a yield line mechanism is assumed. Thus the end-
depth of web. The steel plates used for all specimens are of plate designed by criterion I should be thinner. For specimens of
grade Q345 by Chinese GB standard [11], and the mechanical series A, less thickness values than designed ones were adopted
properties of beams are listed in Table 2, where f yf , f uf are for comparison. The thicknesses are summarized in Table 4.
tested yield stress and ultimate strength of the flange, and f yw , Therefore, in case that the tensile load in the bolt does not
f uw are those of the web. exceed its allowable capacity, the thicker plate shall develop
Meanwhile, the clauses in ANSI/AISC 360-05 [3] were used merely limited plastic deformation or even keep elastic. On
to classify the beam sections. The limiting width–thickness the other hand, the thinner plate which has smaller bending
ratios of beam elements, λ p and λr from the code, are listed in stiffness will develop greater deforms, and so the weld will
Table 3. Comparing the data in Table 1 and Table 3, it is known require greater deformability compared with the one connecting
that the beams of series A, B are NCSE members while the thicker plate.
beams of series C, D are compact ones with the width–thickness In some specimens, stiffeners having the same thickness as
ratios of series D close to limiting values somewhat. the beam flange were arranged on the extended part of the
230 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236

Table 4
General information of specimens

Specimen ID Weld form Weld dimension tep (mm) tep /tfb Bolt layout Beam orientation Loading type
h f (mm) S (mm) α (◦ )
A1-1EM CJP – – – 16 2.0 EPA Horizontal Monotonic
A1-2EM CJP – – – 20 2.5 EPA Horizontal Monotonic
A2-1EM-1 FW 8 – – 16 2.0 EPA Horizontal Monotonic
A2-1EM-2 FW 8 – – 16 2.0 EPA Horizontal Monotonic
A2-2EM FW 8 – – 20 2.5 EPA Horizontal Monotonic
A3-1FM PJP3 9 7.8 60 16 2.0 EPA Declining Monotonic
A1-1ER CJP – – – 16 2.0 EPA Horizontal Cyclic
A2-1ER FW 8 – – 16 2.0 EPA Horizontal Cyclic
A2-2ER FW 8 – – 20 2.5 EPA Horizontal Cyclic
B1-1EM CJP – – – 16 2.0 EPB1 Horizontal Monotonic
B1-2EM CJP – – – 20 2.5 EPB1 Horizontal Monotonic
B2-1EM FW 8 – – 16 2.0 EPB1 Horizontal Monotonic
B2-2EM FW 8 – – 20 2.5 EPB1 Horizontal Monotonic
B2-4EM FW 8 – – 16 2.0 EPB2 Horizontal Monotonic
B2-5EM FW 8 – – 20 2.5 EPB2 Horizontal Monotonic
C1-2EM CJP – – – 20 1.7 EPC1 Horizontal Monotonic
C1-3EM CJP – – – 25 2.1 EPC1 Horizontal Monotonic
C3-2EM PJP1 12 11.1 60 20 1.7 EPC1 Horizontal Monotonic
C3-3EM PJP1 12 11.1 60 25 2.1 EPC1 Horizontal Monotonic
C3-5EM PJP1 12 11.1 60 20 1.7 EPC2 Horizontal Monotonic
C3-6EM PJP1 12 11.1 60 25 2.1 EPC2 Horizontal Monotonic
D1-2EM CJP – – – 20 1.7 EPD Horizontal Monotonic
D1-3EM CJP – – – 25 2.1 EPD Horizontal Monotonic
D3-2EM PJP1 12 11.1 60 20 1.7 EPD Horizontal Monotonic
D4-2EM PJP2 12 11.1 60 20 1.7 EPD Horizontal Monotonic
D3-3EM PJP1 12 11.1 60 25 2.1 EPD Horizontal Monotonic
D3-2FM PJP3 12 12.7 60 20 1.7 EPD Declining Monotonic
D1-2ER CJP – – – 20 1.7 EPD Horizontal Cyclic
D3-2ER PJP1 12 11.1 60 20 1.7 EPD Horizontal Cyclic
D3-3ER PJP1 12 11.1 60 25 2.1 EPD Horizontal Cyclic

(a) EPA. (b) EPC1. (c) EPC2. (d) EPB1. (e) EPB2. (f) EPD.

Fig. 1. Bolt layout and geometry of end-plates (dimensions in mm).

end-plate. The detailed geometries of end-plates and bolt layout bfb tfb f ≤ 2(0.7h f bfb βf ) f fw (1)
are described in Fig. 1.
where, f and f fw
denote the design strength of beam and
fillet weld respectively, βf is the strength factor, 1.22 by
2.3. Weld connecting beam section and end-plate Chinese specification, f fw / f is approximately equal to 0.65 for
grade Q345 steel. The design strength of PJPG weld was also
In this study, both fillet weld (FW) and partial joint pene- calculated according to Eq. (1) while h f was replaced with S
tration groove weld (PJPG weld) with additional reinforcing (Fig. 2). The weld sizes finally adopted are listed in Table 4.
fillet weld on the opposite side of the flange (PJP) are called In all, the three combinations of non-completely penetrated
non-completely penetrated weld. The size of the fillet weld, h f , welds, named as FW, PJP1 and PJP2, were used to connect
was calculated by Eq. (1) based on the principle of equivalent beam flanges and end-plate as shown in Fig. 2. No welding
strength of the weld and beam flange. access hole was set.
 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236 231

(a) FW. (b) PJP1. (c) PJP2. (d) PJP3.

Fig. 2. Non-completely penetrated weld between flanges and end-plate.

located at the free end of the beam to prevent lateral torsional

buckling.
The instrumentation included oil pressure sensor, displace-
ment transducers (designated as D in Fig. 3) and strain gauges.
The tested displacements mainly involved the vertical displace-
ment at the free end of the beam and horizontal displacements
of the end-plate and linkage, as indicated in Fig. 3. Strain
gauges were set on the beam surface to monitor the local buck-
ling of compression flange.

3.2. Test procedure

In monotonic loading tests, the load was applied continu-

ously until the ultimate state of specimen was reached. In cyclic
loading tests, loading procedure was controlled by the displace-
Fig. 3. Test set-up and location of displacement transducers. ment at free end of the cantilever beam. Three levels, δe , 2δe and
3δe , were set with each repeating three cycles. δe was the ver-
To compare with non-completely penetrated weld, two or
tical displacement of beam end at which the initial yielding of
three specimens adopting complete joint penetration groove
beam occurred. After the cyclic routine of the third level fin-
weld (CJP) was designed in each series.
ished, the load was applied until the relative displacement of
Double-sided fillet weld was used for beam web to end-plate
the beam exceeded L b /25.
connection and the sizes of the weld, h f , are 6 mm for series A,
B and 8 mm for series C, D respectively. 4. Monotonic loading tests
Altogether, 30 pieces of specimen were designed. For
most specimens, the beam was perpendicular to the end-plate. 4.1. Failure mode
Considering the joint at the roof ridge and the top of side
column in portal frame, declining beam with gradient of 1:12 The failure mode is predominated by local buckling of
were designed in two specimens, as described in Fig. 2(d), and beams in all specimens even for those in series A with relatively
the welds are named as PJP3. thinner end-plate. When investigating the buckling behavior
The general information of specimens is summarized in according to the width–thickness ratio, it can be confirmed that
Table 4. For the parameters one can refer to Fig. 2. series A and B buckled locally and elastically while series C and
D buckled plastically. The appearance of the beams after tests is
3. Test setup and loading procedure shown in Fig. 4. In the weld connecting beam and end-plate, no
failure was detected. Also, the bolts were found to keep elastic
3.1. Test setup and arrangement of instrumentation during tests by the strain gauges stuck on them.

Test specimen was bolted to a linkage which acted as a 4.2. Load–displacement curves
segment of structural column. The linkage was then fixed on
the back support column. A reaction frame was used to balance Load–displacement (F1 –δ) curves are plotted in Fig. 5,
the load of oil jacks. The test set-up is shown in Fig. 3. In where δ represents the deformation at the free end of the
monotonic loading process, only the upper jack was used. For beam, excluding the settlement and decline of support column.
cyclic loading process, two oil jacks exerted load in turn. Unfortunately, the curve of specimen D1-2EM was not obtained
The loads were applied on the free end of beam at a distance, due to the sensor error.
L b , of 1500 mm (series A) or 2000 mm (series B, C, D) from the In the curves except that of specimen A2-1EM-1, there
end-plate. The load applied by the upper jack and the lower one is an arrow denoting the onset of the local buckling in the
is designated as F1 and F2 respectively. Lateral supports were compression flange, which was detected by strain gauges.
232 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236

(a) Series A. (b) Series B. (c) Series C. (d) Series D.

Fig. 4. Specimens after monotonic loading tests.

(a) Series A, tep = 16 mm. (b) Series A, tep = 20 mm. (c) Series B, tep = 16 mm.

(d) Series B, tep = 20 mm. (e) Series C, tep = 20 mm. (f) Series C, tep = 25 mm.

(g) Series D, tep = 20 mm. (h) Series D, tep = 25 mm.

Fig. 5. Load vs. displacement curves in monotonic loading tests.

 
In Fig. 5, load level Fe or F p are marked with straight lines, f yf Ib f yw Ib
Fe = min , (2)
which represents the loads on the beam end corresponding to its (h b /2) L b (h wb /2) L b
yield moment and full plastic moment respectively. The loads f yf Sfb + f yw Swb
can be calculated by Eqs. (2) and (3) where Ib is the inertia Fp = . (3)
moment of beam section, Sfb and Swb are the plastic section Lb
modulus of flanges and web respectively to the neutral axis of Since the beams of series A and B are NCSE members,
the beam section. The expected capacities of all series are listed the ultimate loads based on estimated moment capacity but
in Table 5. computed by effective area are also given in Table 5, noted as
 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236 233

Table 5
Theoretical resistances of connected beams (kN)
Theoretical resistance A B C D
Fe 159 229 137 332
Fp 218 262 156 405
FB 183 231 – –

FB . Here, the flange width, 32tfb f yf /235, and the web depth,
p

60twb f yw /235, are taken into account when computing the

p
effective area [6].
Fig. 6. Contrast of tested ultimate loads.
According to the curves in Fig. 5 and the observation on
specimens’ failure modes, the following conclusions can be
drawn: (1) Though there is unavoidable deviation, the ultimate
loads of specimens in series A and B, beams of which are
NCSE members, are all between the elastic ultimate load Fe
and FB considering the post-buckling behavior. However, the
maximum loads of specimens in series C and D are all greater
than the plastic resistance F p because of the compact beam
sections. (2) To the components of bolt and weld, the computed
moment resistances by design specification are all higher than
that of the beams, so failure did not occur in these components.
To the component of end-plate, though thinner ones adopted in
some cases, it could still bear the load in tests. Therefore, the Fig. 7. Stress vs. strain curves for finite element models.
beam determined the failure mode and resistance. (3) No matter
what type of weld is used, the connected beam can keep or were tested not only in series A but also in series D, non-
exceed the expected ability and undergo large deformation. The completely penetrated welds designed by equivalent capacity
fact reveals that under a monotonic load, the non-completely rule can be used without problem.
penetrated weld connecting the beam flanges and the end-plate
will not fail prior to the failure of the connected NCSE members 5. FE analysis for supplement of tests
or even compact ones if the weld is designed by the equivalent
capacity rule as Eq. (1) shows. 5.1. Finite element model
4.3. Influence of end-plate thickness
Finite element analysis was carried out using ANSYS
Thinner end-plate is less stiff and therefore requests the FE package. The models were constructed mainly by solid
weld to adapt larger deformation between curved end-plate and elements. In addition to beam members, the other components
beam flange. A comparison based on tested ultimate loads is were simulated to approach their feature as far as possible:
shown in Fig. 6, where F1 max is the ultimate load of individual designed pretension was applied on the bolt by pretension
specimen and F1ave is the average value of ultimate loads in element; on the interfaces between the end-plate and linkage,
each series. Notice that there is the disparity of flexural stiffness bolt head and end-plate, bolt nut and linkage, contact pairs
for end-plate in each group of contrast plates by almost two were established, so does the interface between bolt shank
times. However, it can be found that the variation of end-plate and bolt hole. For CJP welds, only the portion within flange
thickness has no obvious influence on the ultimate resistance of dimension was considered and no extra elements generated
specimen, thus no substantial influence on the weld behavior, beyond. For non-completely penetrated welds, the models were
no matter what kind of weld is used. Especially, the CJP weld constructed following their designed sizes, and the unmelted
does not show its advantage either. part between flange and end-plate was replaced by contact
pairs. For simplicity, beam web was supposed to be welded to
4.4. Influence of declining beam end-plate without unmelted part.
A kinematic hardening model based on von Mises yield
The existence of the angle between beam flange and end- criterion was adopted. The stress–strain relationship was
plate makes the welds a little different, which are composed assumed as multi-line mode described in Fig. 7. Elastic
of partial joint penetration groove weld and beveled fillet weld modulus E was taken as 2.06 × 105 MPa. The tangential
(Fig. 2(d)). However, the comparison of test results does not modulus beyond the yield point, E st , was defined as 0.1E and
show noticeable difference, both on failure mode and on test 0.02E [9] respectively for bolts and the others. Stresses at turn
curves. That is perhaps because: (1) the welds are still designed points, f y and f u , of beam flange and web were identical with
by equivalent capacity rule, (2) the gradient of the beam is not those in Table 2, and the other stress data were taken as the
so great in portal frames. Finally, since the declining beams values specified in Chinese GB code [11].
234 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236

coincide with the test results well in general, though there is

still some diversity.

5.3. Parameter analysis based on test specimen

As an extensive study, FE parametric analysis based on

specimen B2-1EM was executed by varying the values of end-
plate thickness, weld size, column flange thickness and bolt
gauge.
Fig. 8. Distribution of longitudinal residual stress.
End-plate thickness is the main parameter in test study and
then, in FE analysis, thinner end-plate was used to inspect its
effect. Weld size in tests should meet equal capacity principle
and constructional requirement, and besides, it should be
rounded up to integer in design application. Then there must
be a margin for weld size and the margin is generally taken
as 30% of the adopted size. Therefore, FE models with weld
size decreased by 30% and 20% were established and analyzed.
Column flange was the linkage flange and was regarded as the
rigid plane in the tests. In FE analysis the thickness of linkage
flange was varied to model real column flange. Bolt gauge is the
distance between the bolt rows. When the bolt gauge varies, the
bending deformation of end-plate will change accordingly and
Fig. 9. Beam deformation in finite element analysis. the effect should be investigated. The numerical models and the
analytical results are summarized in Table 6.
In order to consider the effect of residual stresses on the Among the analysed models, parameters of the bolt itself
buckling behavior of welded beam, a simplified residual stress were not varied. Therefore the following discussion will be
distribution modeled as in Fig. 8 was adopted. done under the design ultimate state of bolts.
Initial geometrical imperfection was imposed on the beam By the observation on the contour plot of von Mises stress,
plate in FE model. The initial deformed configuration was it can be found that for specimen BE2, most part of the
obtained by local buckling analysis. The maximum deformation end-plate near tension flange have amounted to yield stress
was taken as 1.5 mm, which is the upper limit value of initial or even ultimate strength under the design bending capacity
imperfection for steel plate [12]. of the bolts. With the increasing of the end-plate thickness,
the maximum stress and the area of plasticity decrease. In
5.2. Validity of FE model analysis specimen B2-1EM, there is only a little plasticity in the end-
plate whereas in B2-2EM there is none. The variation of end-
First of all, the analysis for specimens B1-1EM, B2- plate thickness has similar effect on the stress distribution
1EM and B2-2EM was carried out to calibrate the proposed of welds connecting end-plate and tension flange. Firstly the
finite element models. Local buckling also occurred in the stress distribution in welds is found not to be uniform from
models before the collapse of connection components and the the contour plot of von Mises stress. In general, stress level
deformation mode is shown in Fig. 9. The curves of F1 –δ in the weld outside the tension flange is higher than the
obtained from the finite element analysis (noted as FEA) are weld inside and for one single weld the stress in weld toe
compared with the corresponding test curves (noted as TRs) in is often higher than that in the other parts. When the end-
Fig. 10. plate becomes thinner, its bending deformation increases; and
By comparing Fig. 9 with Fig. 4(b), and by observing the the angle between tension flange and the extension part of
curves in Fig. 10, it is confirmed that both the local buckling the end-plate also increases. Thus the fillet weld outside the
patterns and load–displacement relationship from FE analysis flange will suffer greater deformation and so the plasticity in

Fig. 10. Load vs. displacement curves of test and FE analysis.

 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236 235
Table 6
General information of finite element models
Specimen ID Weld tep (mm) t f c (mm) h f (mm) Bolt gauge (mm) F1 max (kN)
B2-1EM FW 16 25 8 120 252
BE1 FW 13.3 25 8 120 247
BE2 FW 10 25 8 120 234
BE3 CJP 13.3 25 – 120 243
BE4 CJP 10 25 – 120 222
BW1 FW 16 25 5.6 120 250
BW2 FW 16 25 6.4 120 252
BP1 FW 16 16 8 120 253
BP2 FW 16 20 8 120 253
BG1 FW 16 25 8 150 248
BG2 FW 16 25 8 90 243

(a) A1-1ER. (b) A2-1ER. (c) D1-2ER. (d) D3-2ER.

Fig. 11. Load vs. displacement curves of typical specimens in cyclic loading tests.

weld develops rapidly. When the bolt is in its design ultimate Table 7
state, it appears plastic at the weld toe of specimen B2-1EM Ultimate loads in cyclic loading tests (kN)
and B2-2EM, but for specimen BE1 and BE2 many parts of Specimen ID F1 max F2 max Specimen ID F1 max F2 max
the weld have yielded. Considering the characteristics of stress A1-1ER 163 163 D1-2ER 434 449
distribution and values, it can be deduced that under ultimate A2-1ER 163 160 D3-2ER 444 472
state of bolts, end-plate and weld remain elastic in specimen A2-2ER 162 161 D3-3ER 445 461
B2-2EM, develop very limited plasticity in specimen B2-1EM F1ave 157 F1ave 453
and fail in specimen BE1 and BE2. Therefore, the decrease of
end-plate thickness will deteriorate the resistance of end-plate small as possible on condition that it meets the requirement of
and weld, but the components will be safe when the end-plate calculation and construction.
thickness is calculated by the methods discussed in Section 2.2. 6. Cyclic loading tests
In the two criteria, criterion II is recommended for safety.
To specimen BE3 and BE4 with CJP weld, there exist similar 6.1. Phenomena in cyclic process
phenomena as those mentioned above.
At the first loading level, there was a little plastic
Plastic zone in fillet weld increases with the decrease of deformation in the flange near the weld. In series A, specimens
weld size. When the size was reduced to 70% of the designed reached their ultimate capacity at the second loading level due
value, as specimen BW1, most part of the weld outside the to local buckling of compressed flange, at the third loading
tension flange becomes plastic whereas the beam and other level, the load retained almost the same value as obtained
components develop merely limited plasticity or even remain at the end of the second level and the compressed flange
elastic. Considering the results of tests and FE analysis, the began to buckle severely. In series D, almost all the loads
equal capacity design rule is reasonable and safe enough. reached ultimate value at the second loading level except F2 of
The thickness of column flange has hardly any effect specimen D3-3ER. At the third loading level the loads dropped
on the distribution and peak value of von Mises stress in obviously. No weld failure was detected in all specimens.
welds although the deformation of column flange changed
accordingly. 6.2. Test data and curves
The effect of bolt gauge is somewhat distinct. When bolt The maximum loads, F1 max and F2 max , are summarized in
gauge increases, compound bending of end-plate also increases, Table 7 as well as the corresponding value F1ave , obtained in
which makes the plastic zone develop and the stress level monotonic loading tests.
improve in the weld. Contrarily the plastic zone and stress Load–displacement curves of typical specimens are plotted
level shall decrease. As a result, bolt gauge should be as in Fig. 11.
236 Y. Chen, S. Wang / Journal of Constructional Steel Research 65 (2009) 228–236

Compared with the results of monotonic loading tests, the only if the weld is designed by equivalent capacity rule.
following conclusions can be drawn: (1) the failure modes of Furthermore, the conclusion can be applied even to the
series A and D under cyclic loading are identical with that compact element connection within thickness limitation.
under monotonic loading, i.e. the local buckling of compressed (4) The design principles for non-completely penetrated welds
flange and web in NCSE members, or the thoroughly plastic connecting end-plate and beam flange are examined by test
deformation after full plastic moment development in compact and FE analysis. However, quality examination of the weld
members. (2) None of the specimens failed due to weld should be done and be emphasized.
cracking. (3) In each series, the change of weld form brings not
more than 7% difference for F1 max and F2 max . Thus, weld form Acknowledgements
has no obvious effect on the bearing capacity of the specimen.
The specimens displayed good energy dissipating character- The research was financially supported by Bluescope Butler
istics, and weld form has no obvious effect on the cyclic behav- (Shanghai) Inc., Shanghai ABC Steel Structure Co. Ltd.
ior of the specimens. and Hangxiao Steel Structure Co. Ltd. Appreciation is also
presented to Dr. Aihui Wu for her supply of weld mechanical
7. Design rules properties.
In the steel structures with light-weight portal frames, References
non-completely penetrated weld can be used to connect the
end-plate and the beam flange. The weld should meet the [1] Adey BT, Grondin GY, Cheng JJR. Cyclic loading of end plate moment
requirement of equal capacity rule in design, namely the tension connections. Canadian Journal of Civil Engineering 2000;27:683–701.
resistance of the weld can not be less than that of the connected [2] American Institute of Steel Construction, Inc. (AISC). ANSI/AISC 341-
02. Seismic provisions for structural steel buildings. AISC, Chicago, IL,
beam flange. May 21. 2002.
Non-completely penetrated weld involves fillet weld and [3] American Institute of Steel Construction, Inc. (AISC), ANSI/AISC 360-
partial joint penetration groove weld with supplemental fillet 05. Specification for structural steel buildings. AISC, Chicago, IL, March
weld. Double-sided fillet weld can be used if tfb < 12 mm or 9. 2005.
[4] Azuma K, Kurobane Y, Makino Y. Cyclic testing of beam-to-column
the weld size will be too large. If tfb > 12 mm, partial joint
connections with weld defects and assessment of safety of numerically
penetration groove weld with supplemental fillet weld can be modeled connections from brittle fracture. Engineering Structures 2000;
used; we recommend a partial joint penetration groove weld 22(12):1596–608.
outside the flange and a fillet weld inside the flange (as shown [5] CECS 102:2002. Technical specification for steel structure of light-weight
in Fig. 2(b)). Although the target of this study is the end-plate buildings with gabled frames [in Chinese].
[6] Chen Y, Wu X, Tian H, Zhao J, Ma Y. Experiment on non-compact
connection with NCSE member, a plate thickness greater than H-shaped members and frames subjected to cyclic loads. International
20 mm could also be covered. Journal of Steel Structures 2005;6(3):215–26.
[7] Dubina D, Stratan A. Behavior of welded connections of moment
8. Conclusions resisting frames beam-to-column joints. Engineering Structures 2002;
24(11):1431–40.
[8] Federal Emergency Management Agency (FEMA). Recommended
For the weld connecting end-plate and NCSE members, the
seismic design criteria for new steel moment-frame buildings, FEMA 350,
following conclusions can be drawn. FEMA, Washington, DC; 2000.
(1) The design parameters considered in tests are typical for the [9] Bing Guo. Collapse mechanism and design criterion of steel beam-
to-column end-plate connections under cyclic loads. Ph.D. dissertation.
members and end-plate connections in light-weight portal Xi’an University of Architecture and Technology; 2002 [in Chinese].
frames. The discreteness of test results is small and the data [10] Witteveen Jelle, Stark JanWB, Bijlaard FransSK, Zoetemeijer Piet.
is credible. Welded and bolted beam-to-column connections. Journal of the Structural
(2) No test specimens failed with weld cracking. Within the Division 1982;108(ST2):433–55.
[11] Ministry of Construction. GB50017-2003, Code for design of steel
scope of the tested width–thickness ratios, the changes of structures. 2003 [in Chinese].
weld form and flange inclination have no obvious effect [12] Ministry of Construction. GB50205-2001, Code for acceptance of
on the ultimate loads, load–displacement curves and cyclic construction quality of steel structures. 2001 [in Chinese].
behavior. So does the proper change of end-plate thickness. [13] Morris LJ. A commentary on portal frame design. The Structural Engineer
(3) The non-completely penetrated weld connecting end-plate 1981;59A(12):394–404.
[14] Murray ThomasM. Recent developments for the design of moment end-
and beam flange can be used in both monotonic loading plate connections. Journal of Constructional Steel Research 1988;10:
condition and cyclic one, and so be used in seismic attack 133–62.