Journal of Constructional Steel Research 63 (2007) 1241–1253

[Link]/locate/jcsr

The semi-rigid behaviour of three-dimensional steel beam-to-column joints

subjected to proportional loading. Part I. Experimental evaluation
J.M. Cabrero ∗ , E. Bayo
Department of Structural Analysis and Design, School of Architecture, University of Navarra, 31080 Pamplona, Spain

Received 29 June 2006; accepted 8 November 2006

Abstract

An experimental investigation of statically loaded extended end-plate connections in both major and minor column axes was undertaken at the
University of Navarra, Spain. The aim was to provide insight into the behaviour of these joints when a proportional load is applied to both axes
(three-dimensional loading). The rotational stiffness of the joints increases with this type of three-dimensional loading. The findings also show
that an increase in the end-plate thickness results in an increase in the connection’s flexural strength and stiffness, and a decrease in its rotation
capacity.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: End plate connections; Major axis connections; Minor axis connections; Three-dimensional behaviour; Experimental testing; Resistance; Stiffness;
Rotation capacity; Steel connections

1. Introduction corresponding analytical model [14,15]; Costa Neves carried

out an experimental programme on extended end-plate joints
It is currently recognised that a great majority of joints bolted to the column’s web and proposed an analytical model
exhibit a rather different behaviour than that idealised as for both stiffness and resistance[16].
rigid or pinned. Extensive investigation of the behaviour of Plane joints subjected to in-plane loading have constituted
such semi-rigid joints in the major axis has been carried the most studied configuration. Few references are available
out. Works by Zoetemeijer [1], Yee and Melchers [2] and accounting for three-dimensional out-of-plane behaviour with
Jaspart [3] constitute the basis for the current component joints in both axes. Gibbons, Kirby and Nethercot tested several
method proposed by Eurocode 3 [4] for the analysis of joints. type of joints in three-dimensional frames [17]. Janss, Jaspart
Faella, Pilusso and Rizzano proposed several modifications and Maquoi also reported tests for three-dimensional and mi-
to this method [5]. The works headed by Chen [6–9] are nor axis joints [18]. Recently, Vértes and Iványi [19] ran a test
also a major source regarding experimental results of semi- program on three-dimensional joints and proposed an analytical
rigid joints and reference advanced analysis methods for semi- model based on the plate formula developed in Eurocode 3 [20].
rigid frames. Works by Aggarwal and Krishnamurthy [10, This paper presents the experimental programme carried out
11] are also a major reference for experimental results and at the University of Navarra, Spain, on a new proposal for a
analytical models for extended end-plate joints [12]. Another three-dimensional semi-rigid steel joint. Proportional loading
source for experimental results is the database compiled by was applied in two different arrangements: to the joint axis
Cruz, Simões da Silva, Rodrigues and Simões [13]. Less (bi-dimensional or in-plane loading) or to both axes (three-
effort has been dedicated to analysing the behaviour of minor- dimensional loading).
axis joints and to proposing adequate analytical models. The proposed three-dimensional joint configuration is
Lima and Andrade tested several angle joints and proposed a presented in Fig. 1. It consists of extended end-plate joints for
both major and minor axes. The minor-axis joints are not bolted
∗ Corresponding author. Tel.: +34 948425600x2704; fax: +34 948425629. to the column web, but to an additional plate welded to the
E-mail addresses: jcabrero@[Link] (J.M. Cabrero), ebayo@[Link] column flanges (see Fig. 2). By virtue of this additional plate,
(E. Bayo). the corresponding minor-axis beams do not get into the column

0143-974X/$ - see front matter c 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/[Link].2006.11.004
1242 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253

Nomenclature S j.h initial joint rotational stiffness measured accord-

ing to the beam vertical displacement
Greek letters S [Link] initial joint rotational stiffness
S j.l initial joint rotational stiffness measured accord-
δb beam elastic displacement ing to the end-plate deformation
δc displacement corresponding to joint rotation S j. p−l post-limit joint rotational stiffness
δv beam elastic shear deformation
δh.i vertical displacement measured at point i Lower cases
δl.c horizontal deformation of the end-plate at the
bb beam section width
compression zone
bc column section width
δl.t horizontal deformation of the end-plate at the
bep end-plate width
tension zone
dinf K −R rotation comprised between the lower bound
u ultimate strain
of the knee-range of the joint moment–rotation
y strain at the yield point
curve and the rotation capacity
rem remaining strain
dsup K −R rotation comprised between the upper bound
φ rotational deformation of the joint
of the knee-range of the joint moment–rotation
φCd rotation capacity
curve and the rotation capacity
φinf K −R rotation at the lower resistance bound of the knee-
e edge distance
range of the joint moment–rotation curve
ex lateral edge distance
φsup K −R rotation at the upper resistance bound of the
fu ultimate or tensile stress
knee-range of the joint moment–rotation curve
fy yield stress
Ψj joint ductility index
f cd tensile stress at failure
Ψinf K −R ductility index evaluated at the lower bound
hb beam section height
of the knee-range of the joint moment–rotation
hc column section height
curve
hr distance from the bolt-row r to the centre of
Ψsup K −R ductility index evaluated at the upper bound
compression
of the knee-range of the joint moment–rotation
h ep end-plate height
curve
ki stiffness coefficient for component i
ρy yield ratio
p pitch of the tension bolts
θi joint rotation
p2−3 distance between the tension bolt-row 2 and the
Upper cases compression bolt-row 3
tep end-plate thickness
[Link] effective tension resistance of the end-plate of tfb beam flange thickness
bolt-row r tfc column flange thickness
[Link] clearance of the column below the end-plates twb beam web thickness
[Link] clearance of the column between the minor-axis w gauge of the bolts
end-plates z lever arm
[Link] clearance of the column above the end-plate
Li distance between the point i and the end-plate
face and the joints do not hinder each other, so they may be designed
L beam length of the beam individually. This plate also acts as a stiffener for the joint in the
L c−t distance between the transducers located in the major axis.
end-plate at the tension and compression zones The end-plate of the proposed and tested minor-axis joints
L load distance between the load application point and consists of two partial end-plates. The remaining gap is required
the face of the end-plate for practical reasons, to access and tighten the bolts properly.
M bending moment These end-plates could be executed as a single end-plate
[Link] flexural resistance of the end-plate extended on both sides if the column size were big enough to
Minf K −R lower resistance bound of the knee-range of the allow for those operations from the upper or the lower side of
joint moment–rotation curve the joint.
M [Link] joint flexural resistance
Mmax maximum bending moment 2. Description of the experimental program
Msup K −R upper resistance bound of the knee-range of the
joint moment–rotation curve 2.1. Test details
Q total load applied
Ri reaction force at support i The experimental programme essentially comprised two
test details on the joint configuration shown in Fig. 1. The
 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253 1243

Fig. 1. Proposed three-dimensional semi-rigid joint.

(a) Major-axis joint. (b) Minor-axis joint.

Fig. 2. Additional plate welded to the column flanges.

(c) Complete test arrangement.

specimens were fabricated from a set of two major- and minor-
axis joints, as shown in detail in Table 1 and Fig. 3. One main Fig. 3. General test detail.
parameter was varied: the end-plate thickness, tep —A series
corresponds to a thick plate (tep = 16 mm), while B series to a Table 1
thin plate (tep = 10 mm)—. The chosen steel grade for plates Test details
and sections was S275. Series Column Minor-axis beam Major-axis beam Bolt tep
Hand tightened full-threaded grade 8.8 M20 bolts in 22 mm
A HEB 160 IPE 240 IPE 330 TR 20 16
drilled holes were used in all the sets. The column had a section B HEB 160 IPE 240 IPE 330 TR 20 10
profile HEB160. In addition, due to the loading procedure, the
clearances above ([Link] ) and below ([Link] ) of the end-plate
were close to 150 mm. All specimens were designed to comply 2.2. Geometrical properties
with Eurocode 3 requirements [4], so that the component end-
plate and bolts in the tension zone were the determining factors The actual geometry of the connection elements was
for the failure. recorded before starting the test. The actual dimensions of
1244 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253

Table 2
Actual geometry of the connection (dimensions in mm, referred to those indicated in Fig. 3)

Column profile

Test ID hc bc tfc [Link] [Link] [Link]

A 160.80 159.00 12.30 153.50 76.50 146.50
B 158.20 159.50 12.30 150.00 78.50 150.00

Beam profile

Test ID hb bb tfb twb L beam L load

Major axis A-M 329.63 161.84 11.44 7.52 1247.75 948.00

B-M 329.75 161.90 11.41 7.38 1248.75 949.50
Minor axis A-n 241.60 119.30 9.51 6.08 1248.00 947.00
B-n 241.04 120.21 9.71 6.23 1249.25 949.50

End-plate and connection geometry

Test ID h ep bep tep e w ex p p2−3

Major axis A-M 409.75 161.50 16.55 29.65 99.04 31.79 89.66 228.41
B-M 408.50 159.50 10.10 28.83 97.98 31.81 89.21 227.81
Minor axis A-n 150.38 142.00 16.35 30.64 79.33 31.83 89.81 –
B-n 150.67 141.50 10.10 31.03 77.95 32.00 89.55 –

Table 3 Table 4
Average characteristic values for the structural steel Average characteristic values for the bolts

Specimen End-plate Beam Steel grade 8.8

Steel grade S275 S275
f y (MPa) 793.75
f y (MPa) 383.12 357.82 f u (MPa) 879.60
f u (MPa) 521.00 499.10 f cd (MPa) 438.08
f cd (MPa) 340.20 248.57
y 25.99
y 5.32 4.00 max 34.43
max 23.28 46.76 cd 43.75
u 33.35 62.56 rem 20.74
rem 18.00 36.29
ρ 0.902
ρ 0.735 0.717

in accordance with UNE-EN 10002-1 [22]. The average

the profiles, plates and connection geometry are summarised
in Table 2. These values correspond to the average of the properties are shown in Table 4.
different measurements for each series. There was not a large
scattering for geometrical measurements. Actual dimensions 2.4. Test arrangement and instrumentation
met the requirements according to the standard UNE-EN
10034:1994 [21]. The actual connections were rotated downwards for practical
reasons. The load was applied to the upper-side of the column
2.3. Mechanical properties by a 400 kN testing machine (hydraulic jack with maximum
2.3.1. Tension tests on the structural steel piston stroke of 200 mm). The beams were pinned supported at
a distance of 950 mm to the end-plate face. In order to prevent
The test programme included one steel grade for the end-
plates and sections: S275. lateral torsional buckling of the beam while loading, a beam
The coupon tension test on the structural steel material was guidance device near the support had to be provided.
performed according to the appropriate UNE procedures [22]. The instrumentation plan is described in Figs. 4 and
The average characteristics are set out in Table 3. The Table 5. The primary requirements of the instrumentation were
gives the values for the static yield and tensile stresses, f y measuring the applied load and the relevant displacements of
and f u , the stress at failure, f cd , the yield ratio, ρ y = f y / f u , the connection (e.g. the vertical displacement of the beam, the
the strain at the yield point,  y , the maximum strain, max , the horizontal displacement of the end-plate). The displacements
ultimate strain, u , and the remaining strain after the test, rem . were measured by means of linear variable displacement
transducers (LVDTs, shown as δl in Fig. 4, which accounted
2.3.2. Tension tests on the bolts for total deformation of the end-plate and the column flange),
Two machined bolts were tested under tension in order and linear position transducers (δh ) located as indicated in
to determine the mechanical properties of the bolt material, Fig. 4 (which recorded the vertical displacements of the beams).
 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253 1245

Fig. 5. Joints instrumentation plan.

followed by the major-axis joints. In both cases, the testing

Fig. 4. General instrumentation plan adopted for each beam. procedure was the same:

Strain gauges (maximum strain ±2%) were added to the end- • Joints were loaded (after initial preloading allowing for
plate (front side) and to the column as shown in Fig. 5 to provide settlement) up to 2/3M [Link] (which corresponds to the
insight into the strain distribution. theoretical elastic limit according to Eurocode 3 [4]), where
Total applied loads (Q) were obtained directly from the M [Link] is the full plastic moment resistance of the joint. After
hydraulic jack. Reactions (Ri ) were measured by load-cells unloading, joints were loaded up again to this same load
located at the beam supports. degree twice. This first part of the tests was carried out under
force control. The maximum load was kept for 30 s to record
quasi-static forces in the joint.
2.5. Testing procedure
• In a second phase, the joints were tested until failure
occurred. The test control was changed to displacement
The dimensions of the plates were recorded before installing control.
the specimens into the testing rig. The bolts were hand-
tightened up to a torque value of 165 N m (approximately 3. Test results
30% of the total torque value for preloaded bolts) to simulate
the snug-tight condition. The specimens were then placed and The results presented in the following sections relate
aligned. mostly to the second phase of the tests, involving controlled
The measurement devices and strain gauges were then displacement. The first, elastic phase, only gives the elastic
connected. Electronic recording started and all the equipment rotational stiffness. The plotted graphs refer to the applied
was verified. The specimens were subjected to monotonic force, load, displacement and strain direct readings and to the
which was applied to the column as explained previously (load corresponding bending moments and deformations.
is indicated as Q in Fig. 4).
The joints were tested with load applied either along one 3.1. Preliminary concepts
or along both axes. The aim of the tests was to characterise
the joints’ behaviour, and to verify whether or not there were 3.1.1. Bending moment
coupling effects between the joints in the minor and major axes The bending moment, M, acting on the connection
when submitted to proportional loading. corresponds to the reaction force, Ri , multiplied by the distance
First, the assembly was tested with three-dimensional between the load application point and the face of the end plate,
proportional loading (load applied to both axes). The four L load (Fig. 4):
beams were supported at the same distance, so loads were
M = Ri × L load . (1)
applied to both axes in a symmetric pattern. This test is
referred to from now on in this paper as three-dimensional 3.1.2. Joint rotation
loading. The test was only carried out in the elastic range The rotational deformation of the joint, φ, is quantified
behaviour of the joints, under force control. Initially, joints in three different ways, according to the measurements taken
were loaded up to a small degree and partially unloaded, to (Fig. 4):
suppress slipping and to allow the different parts to settle and
adjust. Subsequently, joints were loaded up to the start of their • By beam and column vertical displacements, shown as δh in
non-elastic behaviour and unloaded afterwards. This complete Fig. 4. The total displacement at a point i, δi , is the sum of
loading cycle (comprising both loading and unloading) was the displacement corresponding to the joint rotation δc , the
repeated twice. beam elastic, δb , and shear, δv , deformations at that point:
After the three-dimensional test, different tests were δi = δc,i + δb,i + δv,i . (2)
conducted for both individual axes submitted to in-plane
loading (bi-dimensional tests). The non-tested joints remained The displacement related to the joint rotation is
fastened to the column. The minor-axis joints were loaded first, δc,i = δi − δb,i − δv,i , (3)
1246 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253

The stiffness values are obtained by means of linear

regression analysis of the quasi-linear branches before and
after the knee-range. In particular, the corresponding final
unloading portion of the M–φ curve (final step for the second
complete loading cycle of the elastic-range tests) measured by
the transducer located 500 mm from the end-plate face is used
to determine S [Link] (as will be explained in Section 3.3).

3.1.4. Joint ductility

The ductility of a joint is a property that reflects the length
of the yield plateau of the M–φ response. This property can be
quantified by means of an index, Ψ j , that relates the maximum
rotation of the joint, φCd , to the rotation value corresponding to
the joint’s plastic resistance, φ M Rd [25]:
Fig. 6. Moment–rotation curve characteristics.
φCd
and joint rotation is finally obtained as Ψj = . (7)
φ M Rd
δc,i
θ1 = arctan . (4) Other authors [26,27] propose to define ductility as the
Li difference between the rotation value corresponding to the joint
• By the same vertical displacement measurements (Fig. 4): plastic resistance, φ M Rd , and the total rotation capacity, φCd
the joint rotation may be also obtained by means of (Fig. 6).
the resulting displacement difference between two of Both indexes are evaluated in this paper. Since the
the measurement points, after suppressing both beam experimental value of the joint’s plastic resistance is not easily
deformations: assessed [25], two different resistance levels are taken into
δi − δb,i − δv,i − (δ j − δb, j − δv, j )

account: the lower and upper boundaries for the knee-range
θ2 = arctan . (5) of the M–φ curve, corresponding to φinf K −R and φsup K −R ,
Li − L j respectively.
• By the LVDTs located in the end-plate (δl , Fig. 4): Eurocode 3 [4] gives quantitative rules to predict joint
flexural plastic resistance and initial rotational stiffness. These
δl.t − δl.c
θ3 = arctan . (6) structural properties are evaluated below using the geometric
L c−t and mechanical nominal properties. An alternative stiffness
Rotations θ1 and θ2 led to similar results, and were finally formulation presented in [5] will also be evaluated.
adopted, as will be shown in Section 3.3. The reliability of this
method has been already proved in [23]. In the same section, 3.2. Moment–rotation curves
it will be shown how the actual location of the sensors for θ3
leads to unreliable results. As explained above, the M–φ curves for the several
connections are obtained from the beam or end-plate
3.1.3. Joint moment–rotation characteristics displacement readings and the applied load. In the case of
The M–φ curve of the connection may be characterised the beam readings, beam deformations are excluded through
by using the aforementioned relationships. The main features application of Eqs. (3) and (5). Results for the three vertical
of this curve are: moment resistance, rotational stiffness and displacement sensors, two of which are located in the beam
rotation capacity. In particular, the following characteristics are (250 and 500 mm away from the end-plate face respectively)
assessed for the different tests [24], as drawn in Fig. 6: and one in the column, led to similar results.
Fig. 7 shows the complete rotational behaviour obtained for
• the initial stiffness, S [Link] ; each of the tested joints. Both stiffness and resistance decrease
• the post-limit stiffness, S j. p−l ; when the end-plate thickness is reduced. The thickness of the
• the plastic flexural resistance, M [Link] , which corresponds end-plate also affects the joint’s ductility: the thin end-plate
to the intersection point of the previous two regression joints exhibit a greater rotation capacity.
lines obtained for the initial (S [Link] ) and post-limit (S j. p−l ) Fig. 8 depicts the yielding sequence detected by the strain
stiffnesses; gauges for each of the individual tests. In the case of thin
• the maximum bending moment, Mmax , and its corresponding plates, the bending components (mainly the end-plate in the
rotation, φmax ; tension zone) are the weak components. It may be observed
• the knee-range of the M–φ curve, which is defined as the in Fig. 6 how in the case of the A-M test (major-axis thick-
transition zone between the initial and post-limit stiffnesses, plate), its stiffness decreases in three sudden steps. These are
with its lower boundary at Minf K −R and rotation φinf K −R , and due to nut stripping. Bolt cracking caused the final collapse of
its upper limit at Msup K −R and rotation φsup K −R ; the joint. Both phenomena involve brittle failure. Nut stripping
• and the rotation capacity, φCd . was reported for most of the bolts on the tension side, as shown
 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253 1247

Table 5
In-plane loading

Test ID S [Link] S j. p−l

S j.h S j.l
A-M 50 782.13 71 478.79 (+10 000) 3023.38
B-M 31 511.77 27 539.31 (+45 000) 1055.69
A-n 29 867.85 27 797.87 (+32 000) 875.45
B-n 15 157.78 17 457.77 (+1000) 713.69
Initial stiffness (values in kN m/rad).

3.3. Initial stiffness

Tables 5 and 6 present the resulting rotational stiffness

Fig. 7. Resulting moment–rotation behaviour for the tested joints (bi- values for each tested joint. As previously indicated, they
dimensional loading).
correspond to the last unloading path in the elastic analysis
in Fig. 10. These phenomena were not observed in the thick for each of the joints. Results are shown for both measuring
plate tests for the minor-axis joints. methods, S j.h and S j.l .
S j.h corresponds to the previously explained rotations θ1
3.2.1. Discussion on the moment–rotation curves (4) and θ2 (5), according to the transducers that measure the
Fig. 9 shows the application of the Frye and Morris beam and column vertical displacements. Both lead to similar
polynomial model [28–30] for the tested joints. The model results. The rotational stiffness values for each joint relate to
predicts less stiffness and resistance than experimentally the measure of the beam’s vertical displacements (in this paper,
observed. The differences seem reasonable, since tested the given values correspond to those of the transducer located
connections are three-dimensional, and are thus additionally 500 mm away from the end-plate face). The same Table 5 also
stiffened by means of the joint in the other axis. describes post-limit stiffness values.

(a) Major-axis thick-plate (A-M). (b) Major-axis thin-plate (B-M).

(c) Minor-axis thick-plate (A-n). (d) Minor-axis thin-plate (B-n).

Fig. 8. Yielding sequence for each of the joints tested. The strain gauge numbering refers to Fig. 5. Corresponding component and bolt-row are given between
brackets.
1248 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253

Table 6
Three-dimensional loading

Test ID S [Link]
S j.h S j.l
A-M 59 408.68 109 534.95
B-M 35 521.46 64 301.99
A-n 32 294.94 64 752.02
B-n 18 674.62 24 308.70
Initial stiffness (values in kN m/rad).

beam and column displacements gave similar values for both

joints. The LVDTs were located as close to the beam flange as
permitted by the weldings. As shown in Fig. 13, the resulting
(a) Major-axis tests.
deformations in the actual locations (slightly above and below
of the flanges of the beam) are less than those occurring in
the beam flanges. This fact explains in part the differences in
stiffness measured by these sensors. Also, the variation on the
stiffness values between the two symmetrical tested joints may
be due to differences in the location of the sensors because of
the weldings, and also to their possible slipping on the plates
during loading.
As shown in Table 6, both axis joints exhibit an increase
of initial stiffness when they are subjected to three-dimensional
proportional loading. Increased stiffness due to coupling effects
between the joints located in both axes is thus observed. The
stiffness increase for the major-axis joints is about 15%. In the
case of the minor-axis joints, this ranges from 8% for joint A-n
(thick end-plate) up to 23% for B-n (thin end-plate).
(b) Minor-axis tests.

Fig. 9. Application of the Frye–Morris model [28–30] to the tested joints

3.3.1. Discussion of the stiffness results.
subjected to in-plane loading. The initial rotational stiffness of the end-plate bolted joints is
given by a mechanical model [4] composed of several springs,
each one representing a component of the joint, assembled as
follows:
z
S [Link] = P 1 . (8)
ki
i

The lever arm z is taken to be equal to the distance from

the centre of compression to a point midway between the two
bolt rows in tension and ki is the stiffness coefficient for joint
component i.
For the double-sided major-axis beam-to-column joints, the
stiffness coefficients to be taken into account are those related
to the column web in shear (kcws ) and in transverse compression
(kcwc ), and an equivalent stiffness coefficient keq that represents
the basic components related to the bolt rows in tension. The
latter coefficient is based on the stiffness coefficients for the
column web in tension, the column flange in bending, the
Fig. 10. Major-axis joints. Bolts after joint failure. end-plate in bending and the bolts in tension. Although all
these components are given in Eurocode 3 [4], some minor
S j.l corresponds to the rotation θ3 (6), measured by the modifications must be made to include the additional stiffening
LVDTs located in the end-plate. In regard to these latter provided by the additional plate welded to the column flanges.
values, the difference in stiffness between the two tested joints Currently available analytical methods should consequently
(measured by the LVDTs located in the end plate) is given as an underestimate the major-axis response.
additional value between brackets. These variations are of great Table 7 sets up the predicted values for the major-axis
importance in most cases, while the results according to the joints’ initial stiffness and compares them with the experiments.
 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253 1249

Table 7 3.4.1. Discussion on resistance results

In-plane loading Eurocode [4] lacks components to obtain the resistance of
Test ID A-M B-M the tested joints. But as the collapse was mainly due to the
Experimental 50 782.13 31 511.77 end-plate failure in the tension zone, the analytical end-plate
Eurocode [4] 67 041.22 42 065.04 resistance can be compared to the experimental joint resistance.
Error 32.02% 33.49% End-plate flexural resistance is evaluated as follows:
Faella et al. [5] 40 573.10 26 701.52
2
Error −20.10% −15.26% X
[Link] = h r [Link] (9)
Comparison between the experimental and theoretical stiffness characteristics r =1
(values in kN m/rad).
where h r is the distance of the r -th bolt-row from the centre
Table 8 of compression, and [Link] is the effective tension resistance
In-plane loading of the end-plate of bolt row r . The end-plate model relies
on the T-stub idealisation, which can fail according to three
Test ID Knee-range M [Link] Mmax
Minf K −R Msup K −R possible plastic collapse mechanisms: Type-1 is characterised
by complete flange yielding, Type-2 corresponds to bolt failure
A-M 140.37 179.22 167.77 223.08
with flange yielding, and Type-3 involves only bolt failure.
B-M 70.97 96.49 93.24 171.50
Applying this procedure provides the results presented in
A-n 62.20 89.65 78.49 109.77
B-n 41.65 60.66 55.23 95.78
Table 9. When comparing the theoretical predictions with the
experiments, they show values below the experimental values.
Experimental resistance characteristics (values in kN m). Most of them are close to the lower limit of the knee-range.
The same Table 9 also presents the results provided by the
Eurocode [4] predicts stiffer joints. This overestimation has also
method proposed by Murray et al. [33,34]. This model predicts
been reported in [25] when testing joints with stiff columns.
around 30% less resistance for the thin plate tests. Slightly
The alternative stiffness formulation proposed in [5] (which
better agreement is observed for the thick plate connections.
mainly changes the T-stub model for the end-plate and column
Both methods also provide predictions for the type of failure
flange in bending) underestimates the stiffness. However, this
of the end-plate. They lead to similar results, and show a good
theoretical value seems to be more appropriate, as the analytical
agreement with the experimental behaviour. Thin plate joints
components still do not take into account the additional
fail due to end-plate yielding. Thick-plate joints fail according
stiffening provided by the additional plate welded to both
to a mixed failure involving both end-plate yielding and bolt
column flanges.
fracture.
In the case of the minor-axis beam-to-column joints, a
similar spring model can be obtained. While some of the 3.5. Rotation capacity
components have already been developed and presented in
Eurocode [4], others are not yet available. Even though some The experimental values for the rotation capacity and
proposals have been made [15,16,31,32] for minor-axis joints, the corresponding ductility indices for the various tests are
they only apply when joints are directly bolted to the column presented in Tables 10 and 11. Tests with thin plates present
web, so none of them may be applied in this case. No prediction higher ductility. The different rotation capacity is due to the
for the stiffness can then be obtained for the minor-axis joints, different failure mechanism type produced for the end-plate:
due to the lack of required components. in the case of thin plates, failure follows a Type-1 mechanism,
A new model for both major and minor axis joints, which shows very high ductility. The tests with thick plates fail
also accounting for three-dimensional loading effects, will be according to Type-2, which includes the brittle bolt failure.
proposed in the second part of this paper.
4. Behaviour of the plates in bending
3.4. Plastic flexural resistance
4.1. End-plate deformation behaviour
Tests with three-dimensional loading (in both axes) were
performed just in the elastic range, so the experimental values One of the most significant characteristics describing the
presented herein only correspond to the joints subjected to in- overall behaviour of the extended end-plate joints is related to
plane loading (bi-dimensional loading). the end-plate deformation. Different end-plate behaviours are
The experimental value for the moment resistance of the observed when thin and thick end-plate joints are compared.
joint, M [Link] , is obtained as the intersection point between the Accordingly, their different failure mechanisms are shown in
two regression lines obtained for the initial, S [Link] , and post- Figs. 11 and 12. The resulting deformation for the thin plates
limit, S j. p−l , stiffnesses. Two additional resistance values are clearly relates to the theoretical yield line pattern for Type-1
also indicated, those corresponding to the beginning (Minf K R ) mechanism.
and end (Msup K ) of the knee-range behaviour (Fig. 6). All Deformations are greater for the tension side. Little
R
these values are given in Table 8. deformation is noticed in the compression side, as shown
1250 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253

Table 9
In-plane loading

Eurocode [4]

Test ID A-M B-M A-n B-n

Experimental 167.77 93.24 78.49 55.23
Eurocode [4] 126.00 65.79 83.97 43.56
Type of failure epb (m1)-epb (m2) epb (m1)-epb (m1) epb (m1)-epb (m2) epb (m1)-epb (m1)
Error M [Link] −24.90% −29.44% 6.98% −21.13%
Error Minf K −R −10.23% −7.30% 35.00% 4.59%
Error Msup K −R −29.69% −31.82% −6.33% −28.19%

Murray et al. [33,34]

Experimental 167.77 93.24 78.49 55.23

Murray et al. [33,34] 169.47 62.94 99.54 39.70
Type of failure C A B A
Error M [Link] 1.01% −32.50% 26.82% −28.12%
Error Minf K −R 20.73% −11.31% 60.03% −4.68%
Error Msup K −R −5.44% −34.77% 11.04% −34.55%

Legend for the types of failure (Eurocode):

epb (m1) End plate in bending, type 1 (end plate yielding).
epb (m2) End plate in bending, type 2 (mixed failure with plate yielding and fracture of the bolts).
The two predictions indicated for each test account for the first and the second bolt-rows in the following way: pred. for 1st bolt row-pred. for 2nd bolt row.
Legend for the types of failure (Murray et al.):
A Thin plate behaviour controlled by end plate yielding (Eurocode’s Type 1).
B Thin plate behaviour controlled by bolt rupture (w/prying action) (Eurocode’s Type 2).
C Thick plate with behaviour controlled by bolt rupture (no prying action) (Eurocode’s Type 3).
Comparison between the experimental and theoretical resistance characteristics (values in kN m).

Table 10 to “follow” the end-plate deformation, rather than restraining it

In-plane loading (the restriction due to this second bolt row is hardly appreciated
Test ID φinf K −R φsup K −R φ Mmax φCd in the end-plate deformation). This is consistent with the Type-2
(m rad) (m rad) (m rad) (m rad) mechanism reported for the thick plates.
A-M 3.71 8.83 25.70 41.58
In both series, the lever arm seems to comprise the distance
B-M 3.01 8.30 96.30 96.30 between the axis of the beam flange under tension (maximum
A-n 3.90 11.04 31.36 33.85
deformation) and the lower bolt-row, since no deformation
B-n 3.24 11.06 68.00 68.00 originates below this point.
Experimental rotations. 4.2. Additional plate deformation behaviour

Table 11 The additional plate’s deformation behaviour in the tension

In-plane loading zone was measured using a linear transducer in the minor-
Joint Ψinf K −R Ψsup K −R dinf K −R dsup K −R axis and three-dimensional tests. As pointed out in Fig. 14, its
(m rad) (m rad) stiffness increases when both axes are loaded. This increase
results in stiffer behaviour in the minor axis joints, as shown
A-M 6.93 2.91 37.87 32.75
B-M 31.97 11.60 93.28 87.99 in Section 3.3.
The additional plates in the tension and compression zones
A-n 8.04 2.84 29.95 22.81
B-n 20.97 6.15 64.76 56.94 behave in different ways, as demonstrated by their resulting
deformation (Fig. 15(a)). The plate in the tension zone behaves
Ductility assessment.
as a laterally pinned plate, as proved by the single curvature
in Fig. 13 for major-axis joints. The deformability of the of its resulting deformation (Fig. 15(b)). The restraining effect
end-plate increases for smaller values of tep . This effect was of the welds may be assumed to be pinned. By contrast, in the
already noticed in the moment–rotation behaviour, since the compression zone, an additional restriction is provided by the
components’ end-plate and bolts are the main sources of column flanges: the plate may be assumed to act as though it
was laterally clamped.
connection deformability.
The restraining effect of the second bolt row is clear in 5. Conclusions
the tension side of the thin plate (Fig. 13(a)). In the thick
plates (Fig. 13(b)), bolts are submitted to greater tension forces, This paper presented an experimental programme on
resulting in greater elongations. As a consequence, bolts seem extended end-plate beam-to-column symmetrically loaded
 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253 1251

(a) Thin plate specimen. (b) Thick plate specimen.

(a) Thin plate specimen. Tension (b) Thick plate specimen. General General view. General view.
end-plate and additional plate view.
deformation.

(c) Thin plate specimen. (d) Thick plate specimen.

End-plate deformation in the Tension zone (notice the lower
tension zone. bolt cracking and the upper bolt
(c) Thin plate specimen. Tension nut stripping).
end-plate deformation.
Fig. 12. Major-axis joints failure.

results. The alternative proposal in [5] seems to obtain better

agreement.
3. Eurocode 3 [4] proposals give safe approaches for predicting
joint resistance. The predictions underestimate the resistance
in these particular cases.
4. Special attention should be given to the nut stripping
behaviour observed in the major axis thick plate tests, as this
is a brittle failure.
5. The initial rotational stiffness results for all tests indicate that
three-dimensional proportional loading (load in both axes)
(d) Thick plate. Deformation of the additional plate in the significantly affects the joint structural behaviour. The initial
tension zone. stiffness increases for this loading arrangement. In the case
Fig. 11. Minor-axis joints’ failure.
of the minor axis joints, additional stiffening may be related
to the additional plate’s reported stiffer behaviour.
6. Analytical models for these joints are required. Additional
joints in both major and minor axes under bi- and three- components to those presented in Eurocode [4] are also
dimensional proportional loading carried out at the University needed. Models should provide an effective way to include
of Navarra, Spain. The main conclusions of this investigation the three-dimensional reported effects.
are:
The experimental investigation presented in this paper
1. Both the moment resistance and the rotational stiffness of provides a first insight into the semi-rigid behaviour of three-
the joint increase when the thickness of the end-plate is also dimensional joints subjected to proportional loading. More
increased. experimental investigation is required to fully assess this
2. Eurocode 3 [4] overestimates the rotational stiffness of the behaviour. The second part of this paper presents appropriate
tested major axis joints when compared with experimental analytical models that have been developed for this purpose.
1252 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253

(a) Thin plate (B). (a) General view. Column is shown

downwards, so the tension zone is
located below, and the compression
zone above.

(b) Thick plate (A).

(b) Detail of the additional plate deformation in the
Fig. 13. Major axis joints. Remaining deformation on end-plate.
tension zone.

Fig. 15. Additional plate deformation.

References

[1] Zoetemeijer P. Summary of the research on bolted beam-to-column

connections (period 1978–1983). Tech. rep. 6-85-M. Delft: Steven
Laboratory; 1983.
[2] Yee YL, Melchers RE. Moment–rotation curves for bolted connections.
Journal of Structural Engineering ASCE 1986;112(3):615–35.
[3] Jaspart JP. Etude de la Semi-Rigidite des Noeuds Poutre-Colonne et son
Influence sur la Resistance et la Stabilite des Ossatures in Acier. Ph.D.
thesis. University of Liege; 1991.
[4] CEN, Eurocode 3: Design of steel structures. Part 1.8: Design of Joints
(EN 1993-1-8:2005). 2005.
[5] Faella C, Piluso V, Rizzano G. Structural steel semirigid connections:
Theory, design and software. In: New directions in civil engineering. Boca
Fig. 14. Additional plate force–deformation behaviour. Raton (Florida): CRC Publishers; 2000.
[6] Chen W. Practical analysis for semi-rigid frame design. Singapore: World
Acknowledgements Scientific; 2000.
[7] Chen W, Goto Y, Liew JR. Stability design of semi-rigid frames. John
Wiley & sons, Inc.; 1996.
The support given to this work by the Arcelor Chair [8] Chen W, Toma S. Advanced analysis of steel frames: Theory, software
of the University of Navarra is greatly acknowledged. The and applications. In: New directions in civil engineering. CRC Press, Inc.;
Alumni Navarrenses association and the Spanish Science 1994.
and Education Ministry are also gratefully acknowledged for [9] Chen W, Lorens R, Kato B. Semi-rigid connections in steel frames.
providing sponsorship to the first author. McGraw-Hill, Inc.; 1993.
[10] Krishnamurthy N. Fresh look at bolted end-plate behaviour and design.
The assistance provided by the staff at the University of Engineering Journal AISC 1978;2(15):39–49.
Navarra School of Architecture Laboratory in making the test [11] Krishnamurthy N, Huang H, Jeffrey P, Avery L. Analytical
specimens available and their support in conducting the tests is moment–rotation curves for end-plate connections. Journal of the
most appreciated. Structural Division ASCE 1979;105:133–45.
 J.M. Cabrero, E. Bayo / Journal of Constructional Steel Research 63 (2007) 1241–1253 1253

[12] Aggarwal AK. Comparative tests on endplate beam-to-column connec- [23] Aribert J-M, Braham M, Lachal A. Testing of simple joints and their
tions. Journal of Constructional Steel Research 1994;30(2):151–75. characterisation for structural analysis. Journal of Constructional Steel
[13] Cruz PJS, Simões da Silva LAP, Rodrigues DS, Simões RAD. Database Research 2004;60:659–81.
for the semi-rigid behaviour of beam-to-column connections in seismic [24] Girão Coelho AM, Simões da Silva L, Bijlaard FSK. Ductility analysis
regions. Journal of Constructional Steel Research 1998;46(1–3):233–4. of bolted extended end-plate beam-to-column connections. In: The third
[14] de Lima LRO, de Andrade SAL, da S Vellasco PCG. Bolted semi- international conference on steel & composite structures. 2004.
rigid connections in the column’s minor axis. In: Studnicka J, Wald F, [25] Girão Coelho AM, Bijlaard FSK, Simões da Silva L. Experimental
Machacek J, editors. Proceedings of the Eurosteel’99 conference. Prague: assesment of the ductility of extended end plate connections. Engineering
Czech Technical University in Prague; 1999. p. 473–6. Structures 2004;26:1185–206.
[15] de Lima LRO, de Andrade SAL, da S Vellasco PCG, da Silva LS. [26] Schleich JB, Chantrain P, Chabrolin B, Galéa Y, Bureau A, Anza J et al.
Experimental and mechanical model for predicting the behaviour of Promotion of plastic design for steel and composite cross sections: New
required conditions in Eurocodes 3 and 4, practical tools for designers.
minor axis beam-to-column semi-rigid joints. International Journal of
European commission; 1998.
Mechanical Sciences 2002;44:1047–65.
[27] Gil B, Cabrero JM, Goñi R, Bayo E. An assessment of the rotation
[16] da Costa Neves LF. Comportamento monotónico e cı́clico de ligaçoes de
capacity required by structural hollow sections for plastic analysis.
exixo fraco e tubulares em estruturas metálicas e mistas aço-betão. Ph.D.
In: Jaurrieta MA, Alonso A, Chica JA, editors. Tubular structures X. Lisse
thesis. Civil Engineering Department, Science and Technology Faculty,
(Holland): A.A. Balkema Publishers; 2003. p. 277–92.
University of Coimbra; 2004. [28] Frye J, Morris G. Analysis of flexibly connected steel frames. Canadian
[17] Gibbons C, Kirby P, Nethercot DA. Experimental behaviour of 3-D Journal of Civil Engineering 1975;(2):280–91.
column subassemblages with semi-rigid joints. Journal of Constructional [29] Kishi N. PR connection database. In: Practical analysis for semi-rigid
Research 1991;19:235–46. frame design. Singapour: World Scientific; 2000. p. 261–323.
[18] Janss J, Jaspart J-P, Maquoi R. Strength and behaviour of in plane weak [30] Kishi N. Semi-rigid connections. In: New directions in civil engineering.
axis joints and 3-D joints. In: Connections in steel structures: Behaviour, Advanced analysis of steel frames, CRC Press; 1994. p. 91–138.
strength and design. Elsevier; 1988. p. 60–8. [31] Costa Neves L, Simões da Silva L, da S Vellasco PCG. A model for
[19] Vértes K, Iványi M. Investigation of minor axis and 3d bolted end-plate predicting the stiffness of beam to concrete filled column and minor
connections – experimental and numerical analysis – load tests. Periodica axis joints under static monotonic loading. In: Hoffmeister B, Hechler O,
Polytechnica Ser. Mechanical Engineering 2005;49(1):47–58. editors. Eurosteel 2005: 4th European conference on steel and composite
[20] CEN, Eurocode 3: Design of Steel Structures. Part 1.7: Strength and structures, vol. C. Maastricht: Verlag Mainz; 2005. p. 4.10-131–4.10-138.
stability of planar plated structures to out of plane loading (prEN 1993-1- [32] Silva LAP, Neves LFN, Gomes F. Rotational stiffness of rectangular
7). 2005. hollow sections composite joints. Journal of Structural Engineering ASCE
[21] UNE, UNE-EN 10034:1994. Perfiles I y H de acero estructural. 2003;129(4):487–94.
Tolerancias dimensionales y de forma. (Versión oficial EN 10034:1993). [33] Murray T, Shoemaker WL. Flush and extended multiple row: Moment
AENOR; 1994. end-plate connections. Steel design guide series, Chicago (USA):
[22] UNE-EN 10002-1: Materiales metálicos. Ensayos de tracción. Parte 1: American Institute of Steel Construction; 2002.
Método de ensayo a temperatura ambiente. july 2002 Edition, AENOR; [34] Murray T, Sumner E. Extended end-plate moment connections. 2nd ed.
2002. Steel design guide, vol. 4. American Institute of Steel Construction; 2004.