Journal of Constructional Steel Research: Hugo Augusto, Luís Simões Da Silva, Carlos Rebelo, José Miguel Castro

Journal of Constructional Steel Research 116 (2016) 271–293
Contents lists available at ScienceDirect
Journal of Constructional Steel Research
Characterization of web panel components in double-extended bolted

end-plate steel joints
Hugo Augusto a,⁎, Luís Simões da Silva a, Carlos Rebelo a, José Miguel Castro b
a
Institute for Sustainability and Innovation in Structural Engineering (ISISE), Department of Civil Engineering, University of Coimbra, Polo II, Rua Luís Reis Santos, 3030-788 Coimbra, Portugal
b
Department of Civil Engineering, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias, s/n, 4200-465 Porto, Portugal
a r t i c l e i n f o a b s t r a c t
Article history: The prediction of the behaviour of joints relies on experimental and numerical tests that provide accurate infor-
Received 15 January 2015 mation for the characterization of the various joint components. This paper presents the calibration and valida-
Received in revised form 7 August 2015 tion of a parametric FE model of a beam-to-column double extended end-plate steel joint, using test results, and
Accepted 8 August 2015
characterizes the web panel components using the validated FE models. The model implemented in the ABAQUS
FE package which takes into account the non-linear geometrical and material behaviour, non-linear contact and
Keywords:
Steel joints
slip. The calibration/validation of the model is based on results from an experimental research programme on
Web panel three double-extended end-plate partial-strength beam-to-column joints. The behaviour of the joints is charac-
Double-extended terized both globally and in terms of the critical components. The experimental and numerical results are com-
End-plate pared, revealing good agreement, allowing their further use in the more detailed components assessment.
Components Thus, based on the validated stress and deformation fields of FE results, a procedure is proposed to extract the
Finite elements (FEA) force-deformation behaviour of the column web components. The results are compared with those obtained
from Eurocode 3 and from the literature.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction characterization is possible, in the second case the required number of

specimens to have acceptable statistical significance is prohibitively
Predicting the behaviour of beam-to-column steel joints remains expensive. Therefore, numerical simulations are an effective way to over-
an important task, as properly detailed steel joints are crucial to come this difficulty, given that the developed FE models reliably repro-
achieve safety and economy of a structure. Double extended end- duce the behaviour of the real joints [2].
plate beam-to-column connections have the potential to offer a solu- This paper aims to contribute to the mechanical characterization of
tion with moderate to low costs, particularly for cases that do not re- components in double-extended beam-to-column joints using a de-
quire full strength/rigid joints. Given that the behaviour of bolted tailed parametric numerical model developed in ABAQUS [3] and taking
joints is generally much more complex than the connected elements advantage of the Python programing language to develop a scripting in-
and influence the overall structural response, simple and reliable terface for ABAQUS. The model applies to end-plate beam-to-column
models are essential that can be incorporated in advanced analyses joints and considers a three dimensional detailed representation of
of the structures. the various connection components taking into account the several
Eurocode 3 part 1–8 (EC3-1-8) [1] establishes unified procedures phenomena involved in the connection behaviour, namely the nonline-
concerning the modelling of steel joints as the assembly of basic compo- arities related to the geometry, contact, slip and material properties.
nents. The underlying component method (CM) uses a suitable assem- Since further developments of the study will also deal with cyclic load-
blage of non-linear springs and rigid links to determine the resistance, ing, a combined isotropic and kinematic material-hardening model is
stiffness and rotation capacity of steel joints. also included in order to characterize the joint behaviour under load
The mechanical characteristics of the components are obtained from reversal.
experimental tests using specimens similar to the component to be char- The calibration of the numerical model is based on the results of an
acterized, e.g. T-stub, or using complete full-scale connections. While experimental research programme and it is comprehensively described
in the first case acceptable experimental parametric and statistical in this paper.
In the following sections, the monotonic behaviour of the joints is
⁎ Corresponding author. characterized, both globally and also in terms of the critical compo-
E-mail addresses: hugo.augusto@dec.uc.pt (H. Augusto), luisss@dec.uc.pt (L. Simões da
nents, comparing the results of the experimental, numerical and an-
Silva), crebelo@dec.uc.pt (C. Rebelo), miguel.castro@fe.up.pt (J.M. Castro).
URLs: http://www.uc.pt/fctuc (H. Augusto), http://www.uc.pt/fctuc alytical models. Using the validated FE models a detailed procedure
(L. Simões da Silva), http://www.uc.pt/fctuc (C. Rebelo), http://www.fe.up.pt (J.M. Castro). is described to isolate column web components, namely the column
http://dx.doi.org/10.1016/j.jcsr.2015.08.022
0143-974X/© 2015 Elsevier Ltd. All rights reserved.
272 H. Augusto et al. / Journal of Constructional Steel Research 116 (2016) 271–293
web panel in shear and the column web in transverse compression or

tension, and to identify their mechanical behaviour analysing the
stress and deformation fields in the FE models. The results are com-
pared with those of EC3-1-8 [1] and with other models available in
the literature.
2. Literature review
2.1. Introduction
Beam-to-column joints modelling must deal with a variety of phe-

nomena, mainly in bolted connections that make the prediction of its
mechanical behaviour a complex task [4]. As summarized by Faella
et al. [5], the methods for predicting the joint behaviour can be divided
in five different categories: empirical models, analytical models, me-
chanical models, finite element models and physical models (experi-
mental testing).
Mechanical models have gained wide acceptance because they
achieve a good balance between accuracy (accounting for resistance,
deformability and ductility) and ease of use. Joints are decomposed in
several parts, called components, that represent a specific part of a
joint that, depending on the type of loading, make an identified contri-
bution to one or more of its structural properties [6]. The constitutive re-
lations of the components and the way they are assembled determine
the joint behaviour. The relation between the components and the
joint's mechanical properties is determined through equilibrium and Fig. 1. Behaviour of the unstiffened column web panel in an external node end-plate
connections.
compatibility relations. In the framework of EC3-1-8 [1], the implemen-
tation of the component method allows for the determination of the
resistance, stiffness and rotation capacity of a variety of steel joint formulation is summarized in Table A.2, using the notation adopted
configurations. by Faella et al. [5].
Finite element models provide the most sophisticated and realistic Following Zoetemeijer's work [10] on the definition of the rele-
representation of joint behaviour but their use in design is very time vant stress interactions when the load-introduction effect due to
consuming and cumbersome, presents convergence and calibration dif- the beam forces is considered, a research programme was conducted
ficulties and is prohibitively expensive if combined with physical at the University of Innsbruck to separately characterize the compo-
models [7]. nents due to shear and load-introduction effects in the column web
This section presents a brief state of the art on bolted end-plate panel using experimental tests on welded joints. The insights gained
beam-to-column joint modelling and behaviour focussed on the use of in the characterization of the web panel behaviour were exploited by
mechanical models and finite element models. Jaspart [8,11] to develop multi-linear analytical models (AJM) for the
two identified deformation modes in the web panel, namely the
2.2. Characterization of the column web panel components load-introduction, due to the beam moment binary forces, and the
distortion due to the pure shear installed in the panel. The induced
A bolted beam-to-column steel joint may be divided in two main forces in the web panel produce a complex stress state with three
parts [8]: column web panel and connection(s) (left and/or right stress components, as depicted in Fig. 1: τ is the shear stress associ-
beams). In many situations, the column web panel controls the behav- ated with the shear forces on the contour of the column web panel;
iour of the joint. Especially under cyclic loading conditions, the column σi is the horizontal normal stress associated with the load-
web panel provides large energy dissipation. In this paper, special atten- introduction and σ n is vertical normal stress due to the axial load
tion is paid to deriving the components associated with the column web and bending of the column. The analytical model determines the
in double extended bolted end-plate joints. strength of the column web panel by analysing its stress state. In
Fig. 1 illustrates the contributions to the deformation of the column the application of the Von Mises yield criterion, some simplifications
web panel resulting from the combined effect of tensile and compres- were proposed by Zoetemeijer [10] and Jaspart [11] for the stress in-
sive forces generated by the bending moment in the beam. The load- teractions: (i) due to the localized effect of the σi stresses in the as-
introduction and shear effects are identified in Fig. 1; they require sessment of the shear resistance only the τ–σn interaction needs to
appropriate characterization of the components behaviour which are be considered; and (ii) for the assessment of the load-introduction
associated to the column web in tension, column web in compression resistance only the τ–σi interaction needs to be considered, because
and the column web in shear. Annex A summarizes the related design σn has a low influence on the load-introduction resistance for axial
models, figures, equations and notations. load in the column lower than 0.7Npl,Rd.
The characterization of the column web panel was the subject of Subsequently, analytical models were developed for welded
relevant research work conducted by Krawinkler and his co- joints [11] and further adjusted for bolted joints in the case of the
workers [9] who studied the column web behaviour, mainly sub- load-introduction, summarized in Annex A, Fig. A.3, Tables A.3 and
jected to shear, using experimental tests on welded joints stiffened A.4. In the extrapolation from welded joints to bolted end-plate
with continuity plates. They concluded about the significant post- joints there are relevant differences that must be taken into account
yielding resistance, due to the strain hardening and the bending [11]. For the load-introduction effect, based on a beam on elastic
contribution of the boundary elements composed by the column foundation model, the deformation in welded joints is similar in
flanges and the transverse stiffeners, and proposed a trilinear the tension and compression zones. However, for bolted end-plate
model for the shear force versus distortion of the panel zone. The joints, because of the different ways of applying the tension and
relevant features of the model are illustrated in Fig. A.2 and the compression loads by the beam (see Fig. 1) the corresponding Fb–Δ
H. Augusto et al. / Journal of Constructional Steel Research 116 (2016) 271–293 273
curves are distinct, as shown in Table A.3. Differences in the bolt column joints with pretension bolts. Trying to overcome some ex-
rows change the deformed configuration and the position of the neu- perimental test limitations, the authors used FE results to provide
tral axis. Therefore, the rotation needs to be computed considering additional valuable data for the joint's behaviour, such as the distri-
the deformation contributions of each component: bolts, end-plate, bution of pressure caused by bolt pretension, the friction between
column flange and column web. Also for the load-introduction effect, the end-plate and the column flange and the principal stress flow
the infinitely stiff condition for the column flange provided by the in the connection.
welding of the beam web to the column flange cannot be considered Despite the extensive numerical work available in the literature,
valid in bolted end-plate joints. In the case of the shear model, there are still no established procedures to extract the individual com-
the stresses may be considered uniformly distributed due to the in- ponent behaviour from complex 3D FE models.
fluence of the column flanges. It is also important to note that the
presence of transverse stiffeners aligned with the beam flanges influ- 3. Experimental tests on double-extended end-plate joints
ences the resistance of the joint.
Recent studies on the column web panel extended its scope of ap- A set of experimental tests [21] carried out at the University of Coim-
plication by proposing new and more sophisticated models. Castro bra were used to calibrate the numerical model presented in this paper.
et al. [12] proposed a model for steel and composite joints suitable The tested joints were designed aiming to study the behaviour of several
for use within frame analysis procedures. The model accounts for external bolted beam-to-column partial-strength joint typologies with
the effect of different boundary conditions, as well as shear and flex- double extended end plates. The relevant features of these tests are de-
ural deformation modes, in evaluating the elastic and inelastic re- scribed in the following sections.
sponse. Jordão et al. [13] based their studies on the work done by
Zoetemeijer [10] and Jaspart [8] and proposed new behaviour 3.1. General description
models for internal joints with beams of unequal depth, based on ex-
perimental and numerical work. Due to the unequal beam depths, The experimental programme comprised four external double-
the column web panel is divided into two sub-panels and three extended bolted end-plate joints, J1 to J4, whereby the column section
levels are considered for the load-introduction effect. An appropriate size and/or the beam size were varied, as well as the presence of axial
mechanical model was then proposed to determine the overall re- force in the column in test J2.1, as can be observed in Table 1. The spec-
sponse of the joint [13]. imens reproduced joints of a moment-resisting frame system and in-
cluded transverse stiffeners in the column aligned with the beam
2.3. Finite element modelling flanges. The joint's details are illustrated in Figs. 2 to 4. The configura-
tions of the joints differ mainly in the column and beam cross-
Although good approximations achieved using mechanical models sections. For all cases, 18 mm thick end-plates were chosen, connected
are in most cases sufficient for the design of joints, it is increasingly re- to the beam-ends by full strength 45° continuous fillet welds, shop
quired to obtain very accurate approximations for the joint behaviour, ei- welded in a down and up position. A manual metal arc welding pro-
ther to validate the mechanical models or to develop new and more cedure was used, with Autal Gold 70S electrodes. Hand tightened,
accurate tools for the design of joints or to incorporate in the global full-threaded M24, grade 10.9, in 26 mm diameter drilled holes
frame analysis. This is a relatively difficult task due to the number of com- were employed in all joints.
ponents involved in the joint, the material non-linearity (plasticity, iso- Test J2.1 will not be addressed in this paper as it is similar to test J1.1
tropic and kinematic strain-hardening), the non-linear contact and slip, except for the presence of axial force in the column.
the geometrical non-linearity, the residual stress conditions and the com-
plicated geometrical configurations. Furthermore, a limited number of 3.2. Test detail, instrumentation and procedures
experimental tests are required to characterize in detail the mechanical
behaviour of the joint and to validate the modelling assumptions. The loading was applied monotonically in a displacement-controlled
The significant development of the computer technology, allowing procedure.
increasing complexity of numerical tools, makes it affordable to perform To extract the relevant test information, the sub-assemblages were
parametric studies on experimentally calibrated and validated finite instrumented according to the scheme shown in Fig. 5. All the instru-
element (FE) models [14]. Historically, the first attempts, in end-plate mentation and load cells were connected to a data logger and measure-
FE modelling and behaviour simulation, are attributed to Krishnamurthy ments were taken every 10 s. The displacements were measured by
and Graddy [15]. With the available computer resources at the time, means of TML displacements transducers. Strain gauges TML PFL-10-
the authors attempted to correlate the results from an elastic three- 11-1L, FLK-1-11 and FRA-6-11 (general use, with 2%, 3% and 3% maxi-
dimensional FE analysis to those of an elastic two-dimensional FE analy- mum strain, respectively) were used to measure the surface deforma-
sis. A similar approach was used by Kukreti et al. [16] focusing on the tion. The bolt deformations were measured with special TML BTM 1C
prediction of maximum end-plate separation through parametric analy- strain gauges embedded in the bolt shaft, as shown in Fig. 6.
ses covering the various geometric and force related variables found in The columns were 3.0 m high and the beam was approximately
practical ranges. Several works followed using 3D FE models, overcoming 1.2 m long. The loading was applied in the vertical direction, at the
some limitations of 2D analysis [17]. Sherbourne and Bahaari [18] used 3D end of the cantilever beam by means of a 100 ton hydraulic actuator,
shell models to deal with thin plates in which yielding occurs due to as shown in Fig. 7a). Fig. 7b) shows the test setup for joint J-3.3 that
biaxial bending, although, for thick plates the results were also satis- comprised a reaction wall on the left side, a loading steel beam at the
factory. They demonstrated the need to use 3D models in these prob- top and a steel frame on the right side anchored to the floor. The
lems and concluded that the lack of stiffeners changes the behaviour
of the connections in the tensile and compression zones of the col-
Table 1
umn flange. Maggi et al. [19] performed some parametric analyses
Bolted beam-to-column double extended end-plate joints test programme.
on the behaviour of bolted extended end-plate connections studying
the interaction between the end-plate and bolts. They concluded Joint Beam Column type Bending Axial
that the T-stub failure mode type 2 presented levels of interaction Test J-1.1 IPE360 HEA320 Monotonic M− –
between the end-plate and bolts that are difficult to predict with ac- Test J-2.1 IPE360 HEA320 Monotonic M− N−
curacy. Shi et al. [20] presented the development of a FE model to Test J-3.1 IPE360 HEB320 Monotonic M− –
Test J-4.1 HEA280 HEA320 Monotonic M− –
simulate the mechanical behaviour of different types of beam-to-
Fig. 2. Detail of the joint for Groups 1 and 2.
Fig. 3. Detail of the joint for Group 3.
Fig. 4. Detail of the joint for Group 4.
specimens were supported by a concrete block with a pinned joint at Eqs. (B.1) and (B.5) for the lesser instrumented J3.1 test and the as-
the bottom. At the top of the model the connection to the steel beam sociated failure modes are illustrated in Fig. 8 for each joint. The
had the vertical displacement free. analysis of the results of the J1.1 test allows concluding that the
governing component was the column web panel in shear, reaching
3.3. Material properties an ultimate moment resistance of 419 kNm and a corresponding ro-
tation of approximately 70 mrad. This test was further loaded until
The material properties were obtained from steel coupons extracted rupture developed on the extended part in tension of the end plate
from the web and flanges of the beams and columns and also from the (see Fig. 8(b), not represented in Fig. 8(a)), for a rotation larger
end-plates and were subjected to uniaxial tensile tests. Some bolts were than 100 mrad. In the case of the J3.1 test, the joint failed by the col-
also subjected to tensile tests until rupture occurred. umn web panel reaching an ultimate moment resistance of 476 kNm
Tables 2 and 3 reproduce the measured values of the steel mechan- and a corresponding rotation of approximately 47 mrad, see
ical properties for the steel components and the bolts. An average Fig. 8(a) and (c). This joint was not taken to rupture. For the J4.1
overstrength in the range of 10% to 30% is noted for the yield stress of test, the joint rupture was due to failure of one bolt from the first
the steel profiles. The real dimensions of the prototypes were also mea- bolt row at a measured rotation in excess of 100 mrad; however,
sured, without significant deviations. from 80 mrad onwards, the membrane effects in the extended ten-
sion part of the end plate governed the response, see Fig. 8(d).
3.4. Global behaviour Table 4 summarizes the experimental results. The maximum rotation
and the experimental failure modes achieved for each test are included in
The experimental moment–rotation curves, obtained according the table, where EP stands for end-plate and BT for bolts in tension. In the
to Annex B, using Eqs. (B.1) and (B.4) for the J1.1 and J4.1 tests and case of the J3.1 test, first yielding occurred in the column web panel, but
and the column are established according to the experimental tests setup
used in the validation of the FE models.
The models are composed of solid (mainly C3D8RH) finite elements
in the joint region. Beam elements (B31) are used in the adjacent
regions of the joint, i.e., in the beam and column segments, aiming at
reducing the computational time. The technique adopted is similar to
the one used in [22], the kinematic relations being transferred from
the solid to the beam elements by multi-point constraint equations.
The model and its parts are illustrated in Fig. 9, with the corresponding
adopted mesh, namely the wire columns, the wire beam, the solid col-
umn with or without stiffeners and with the bolt holes in the flange,
the solid beam with the corresponding fillet welds, the end plate with
the bolt holes and the bolts composed by the shank, head and nut. The
various parts interact with each other by constraints or interactions, to
form the beam-to-column end-plate joint mechanism.
4.2. Finite elements options
The solid elements used are mainly quadrilateral and hexahedra that
are the standard volume elements available in ABAQUS. For complex
nonlinear analyses involving contact, plasticity and large deformations,
and taking into account the recommendations of Bursi and Jaspart [17,
23], three finite elements available in ABAQUS [3] were considered:
C3D8H, a 8-node linear or first-order brick element (nodes only at cor-
Fig. 5. Instrumentation apparatus.
ners) with full integration (8 Gauss points), hybrid formulation and fea-
turing constant pressure; C3D8RH, also a 8-node linear brick element
no rupture was achieved during the test. Fig. 8 also shows that the joints but with reduced integration (normally using a scheme one order less
exhibited ductile behaviour and significant rotation capacity. than the full scheme to integrate the element's internal forces and stiff-
ness with only 1 Gauss point) [3], hourglass control using the artificial
stiffness method [3] given in [24] and hybrid formulation; and
4. Numerical modelling C3D8IH, a first-order full integrated (8 Gauss points) element enhanced
by incompatible modes to improve its bending behaviour, allowing to
4.1. Modelling overview overcome the overly stiff condition due to the parasitic shear stresses.
Due to the 13 additional degrees of freedom, these elements are more
The numerical model developed in ABAQUS [3] consists of sub- computationally demanding than the regular first-order displacement
assemblages of a column and a beam connected to each other by means elements, but less demanding than second-order elements [3]. Follow-
of an end-plate welded to the beam and bolted to the column flange ing the recommendations of the benchmark proposed in [17], a sensitiv-
and are representative of an external node of a moment-resisting framed ity study was conducted to analyse the response of the 3 solid brick
structure with double-extended end plate joints. The lengths of the beam elements, which, according to the previous statement, are the ones
Fig. 6. Strain gauges disposition.

Table 3
Average mechanical properties of the tested bolts.
Bolt Young's modulus Ultimate stress Ultimate

Em (GPa) ƒu (MPa) strain εu
M24 (10.9) 213 1170 0.030
that might best solve the specific problem. Using the J1.1 and the J3.1
models, for which the geometric properties are shown in Fig. 2 and in
Fig. 3, Fig. 10 shows a comparison of the numerical response using the
3 different element types with the corresponding experimental re-
sponse. It was concluded that the element that best predicts the real be-
haviour for large rotations is C3D8RH. In the linear elastic branch all the
elements presented similar and satisfactory results. The results lead to a
logical choice for the solid element to use in the FE analyses, C3D8RH,
because of its better efficiency.
The preferential use of quadrilaterals and hexahedral elements was
due to the better convergence rate of these elements in comparison
with the triangles and tetrahedral elements, providing equivalent accu-
racy with less computational effort, for regular meshes, which is the
case in most of the model zones. As first-order (linear) triangles and
tetrahedral are usually overly stiff, these elements can only provide ac-
curate results with very refined meshes. With the intent of saving com-
putational time, reduced integration elements were adopted using a
lower-order integration to form the element stiffness. Although in this
case the numerical problem concerning shear locking is overcome, the
hourglass can be a real problem for the linear reduced-integration ele-
ments. As the elements only have one integration point, it is possible
to have distortional deformation modes in such a way that their stiffness
is severely reduced. In problems governed by bending deformations this
effect can have a direct influence on the accuracy of the results. To avoid
this problem, at least 3 layers of elements were considered across the
thickness, see Fig. 11(a), and also the hourglass control formulation was
activated for these elements. In their convergence study, Bursi and Jaspart
[17] concluded, using 1, 2, 3, 5 and 8 layers of elements in clamped and
simply supported beam models subjected to bending, that at least 3 layers
should be considered in bending-dominated problems. The elements
chosen have also hybrid formulation, normally used for fully incompress-
ible material behaviour or, as in this case, if severe plastic deformation is
expected, because the rate of total deformation becomes incompressible
as the plastic deformation starts to dominate the response [3]. For the
Table 4
Bolted beam-to-column double extended end-plate joints test programme.
Model Mmax (kNm) Sji (kNm/rad) Rot.max (mrad) Failure mode
J1.1 419 73,500 +70 Cracking EP

J3.1 N476 115,000 N47 –
J4.1 391 34,600 100 BT
Fig. 7. Test's setup.
Table 2
Average mechanical properties of steel tested.
Section size Component Yield stress Nominal yield Δ (%) Young's Modulus Nominal value Δ (%) Ultimate stress Nominal value Δ (%)
ƒym (MPa) ƒy (MPa) Em (GPa) E (GPa) ƒum (MPa) ƒu (MPa)
IPE360 Flanges 430.0 355 +21.1 206.0 210 −1.9 554.2 490 +13.1
Web 448.2 355 +26.3 213.6 210 +1.7 552.9 490 +12.8
HEB320 Flanges 393.9 355 +11.0 208.8 210 −0.6 520.7 490 +6.3
Web 398.8 355 +12.3 216.1 210 +2.9 521.1 490 +6.3
HEA320 Flanges 414.8 355 +16.8 204.9 210 −2.5 531.4 490 +8.4
Web 449.6 355 +26.7 207.4 210 −1.3 553.4 490 +12.9
HEA280 Flanges 439.7 355 +23.9 209.4 210 −0.3 547.7 490 +11.8
Web 461.7 355 +30.1 210.2 210 +0.1 575.9 490 +17.5
End-plate 220 × 18 405.1 355 +14.1 210.3 210 +0.1 534.0 490 +9.0
End-plate 300 × 18 392.9 355 +10.7 208.4 210 −0.8 523.0 490 +6.7
Stiffeners 286.4 235 +21.9 205.9 210 −2.0 451.8 360 +25.5
500
Bending Moment (kNm)

400
300
200
100 Experimental curve - J1.1

Experimental curve - J3.1
Experimental curve - J4.1
0
0 20 40 60 80 100
Rotation (mrad)
(a) Moment rotation curves. (b) Test J-1-1 failure mode.
(c) Test J-3-1 failure mode. (d) Test J-4-1 failure mode.
Fig. 8. Moment–rotation curves and failure modes.
non-solid column and beam parts, the B31 element available in ABAQUS with cyclic loading cases. Experimental evidence has shown the differ-
is used, i.e., a three-dimensional first-order linear beam element with 2 ence between the material properties obtained from the monotonic
nodes, using a beam section integrated during the analysis to define the and cyclic specimens, and therefore simple elasto-plastic isotropic or ki-
section behaviour, with 13 wall points in total, because it is a spatial nematic hardening models are not adequate to deal with load reversal.
model. It is not expected that the beam elements develop large plasticity This constitutive model [25] uses the Von Mises yield criterion and an
due to their location, hence the standard number of points in the section associative flow rule is assumed. The monotonic coupon test results
seems to be reasonable in these circumstances. are used to obtain key coordinates in the stress–strain relationship
As previously mentioned, the various parts of the model interact (see Tables 2 and 3). In all parts of the model the elasto-plastic constitu-
with each other through constraints or interactions. The beam elements tive material laws were used, non-linear in the solid elements and tri-
are constrained to the solid column and beam parts by a multi-point linear in the beam elements.
constraint beam formulation that uses the concept of slave and master
nodes to define the same degrees of freedom between both. Between
4.3. Finite elements analyses
the end-plate and the solid beam a tie constraint is imposed, using the
same master and slave philosophy, the degrees of freedom of the de-
The analyses were performed using the ABAQUS/Standard solver
pendent nodes are eliminated; the two surfaces will have the same
that iteratively solves a system of equilibrium equations implicitly for
values of their degrees of freedom. The interactions between the end-
each solution increment [3].
plate and the column flange and the interactions between the bolts
Second-order effects were considered in all the analyses to account
and the end-plate or column flange are imposed by the general contact
for large displacements effects. Due to the several nonlinearities present
algorithm, that uses “hard contact” formulation, using the penalty
in the models, namely material nonlinearity, geometric nonlinearity
method to approximate the hard pressure-overclosure behaviour, that
and contact nonlinearity, general analysis steps were adopted with au-
acts in the normal direction to resist penetration, the contact force
tomatic control for the increments, i.e. the increment size is determined
being proportional to the penetration distance, and tangential behav-
according to the response of the model.
iour to take into account the friction between surfaces. It is important
to note that Lagrange multipliers will be automatically added to the
solution if the comparison of the contact stiffness to the underlying ele- 5. Validation of the numerical models
ment stiffness is overcome in 1000 times, at the expense of solution
cost, but avoiding numerical errors related to ill-conditioning of the 5.1. Overview
stiffness matrix.
The bolts were pre-loaded using the adjust length option and deter- In this section a detailed analysis of the joints will be performed both
mining the length magnitude by the elastic elongation needed to pro- for the global and components behaviours. The results from numerical
duce the required amount of force in the bolts, normally a percentage and analytical models are compared with the experimental test results
of the ultimate strength, see Fig. 11(b). described in Section 3. The analytical models proposed in EC3-1-8 [1]
For the material properties, the combined isotropic/kinematic hard- are used in the comparisons, using the real material properties of the
ening model available in ABAQUS was adopted. Although this paper steel obtained in the coupon tests and adopting partial safety factors
only deals with monotonic loading, the model is prepared to deal also equal to unity.
Fig. 9. Meshed parts of the FE model.
500 500

400 400
300 300
200 200
J1.1 - Experimental results J3.1 - Experimental results
C3D8H - Full integration (FI) C3D8H - Full integration (FI)
100 C3D8RH - Reduced integration 100 C3D8RH - Reduced integration
C3D8IH - FI with Incompatible modes C3D8IH - FI with Incompatible modes
0 0
0 20 40 60 80 0 20 40 60 80
Rotation (mrad) Rotation (mrad)
Fig. 10. Sensitivity study on the ABAQUS solid elements.
In order to compare experimental and numerical results, forces, the J1 and J3 models. In the J4 model only 1% of the ultimate bolt
bending moments, displacements and rotations were determined ac- strength was applied because the bolts were not prestressed in the
cording to the procedure described in Annex B. For the experimental experimental test, whereas in the model a residual value is needed
tests, the load cells and the displacement transducers (DT) indicated to initiate the contact between the parts. The elasto-plastic material
in Fig. 5 were used. In the same way, for the FE models, the reactions properties were based on the mean values of the coupon tests, listed
and the displacement of some predefined nodes were selected, as in Tables 2 and 3.
shown in Fig. B.1. The strain gauges were used to determine the column Fig. 13(a), (b) and (c), compares the experimental, numerical and
web strains, in the experimental tests, and the logarithmic strain were analytical moment–rotation curves for the 3 joints, J1.1, J3.1 and J4.1,
extracted from the FE models. The equations used to obtain the M–θ re- respectively.
lationships in this section can be found in Table B.1, complemented by The joints were analysed according to the component-based proce-
Fig. B.1. dure given in EC3-1-8, using the web and flanges measured material
properties, from Tables 2 and 3. The main results are shown in
5.2. Global behaviour Table 6, highlighting the critical design components. It is noted that
the failure mode found among all the components in tension was al-
A set of models were developed according to the tests set up, using ways related to the T-Stub components, mode 2.
the same geometry, see Fig. 12 and Table 5, respecting the same bound- For the experimental results, the bending moment was obtained
ary conditions and the applied loading. Following the test procedure, using Eq. (B.1). The joint rotation was obtained by Eq. (B.6). Similarly,
the bolts were preloaded with 20% of the ultimate bolt strength for for the FE models, Eq. (B.2) was used, where dDT20 is the distance
Fig. 11. Detail for the elements in the thickness and bolts pre-load plane.
L1
Lc1
d
Column
Beam
ht Lc Lc2 disp.
Lb1 Lb2
Lb L2
Lc3 2(Y)
3(Z)
1(X)
J1.1 J3.1 J4.1 Geometric properties
Fig. 12. FE models used in the validation.
between the point where the imposed displacement occurs (DT20) and studied joints the continuity transverse web stiffeners limit the column
the column flange, and R2DT20 is the reaction force obtained at the same flanges deformation, so the large contribution is provided by the other
point. The total rotation of the joint is obtained from Eq. (B.7), the sum of two components. The column web crossheads were used to determine
the column web contribution with the other components contributions the rotations of these components, so Eqs. (B.9) and (B.10) are used to
(end-plate, column flanges and bolts). determine the joints column web rotation, for the experimental and nu-
Fig. 13(d) depicts the adopted criterion for extracting the initial merical models, respectively. Eqs. (B.11) and (B.12) are used to deter-
stiffness and moment resistance from the experimental and numer- mine the end plate contribution for the experimental and numerical
ical M–θ curves. The initial stiffness is easier to identify as it may be models, respectively.
estimated by adjusting a straight line between the point of zero load- Fig. 14 illustrates the relative dissipation of energy for these compo-
ing and the elastic moment of the joint (Me, first yield in the joint). It nents and compares the moment–rotation curves for each joint and the
is noted that the experimental procedure should always contemplate two critical components, for the experimental and numerical results. For
elastic unloading during this stage and the elastic stiffness should be all tests, the column web panel in shear exhibits the largest contribution
fitted to the unloading branch. Assessing the plastic resistance from towards the deformability of the joint (66% to 83%), while the end-plate
experimental or FE numerical results is a more complex task, as in bending represents 17% to 34% of the overall energy dissipated. The re-
yielding of the joint is a gradual process, typically occurring between maining components have little contribution as they remain elastic.
points A and B in Fig. 13(d). Furthermore, in many cases, the behav-
iour of the joint is not only driven by plasticity and a plastic plateau is 5.3.2. End-plate and column flange in bending and bolts in tension
less obvious, either because of membrane effects and/or strain harden- Table 8 summarizes the experimental initial stiffness and resistance
ing. According to the literature [26], it is common to define the plastic of the critical component end-plate in bending, comparing the experi-
moment resistance at the intersection of the two straight lines corre- mental tests results with the analytical model. The stiffness values of
sponding to the initial stiffness of the joint and a post-limit stiffness. the components, obtained according to EC3-1-8, are computed using
The latter can only be properly established on a case-by-case basis, di- only the equivalent stiffness coefficient keq, for all the bolt-rows in
rectly from the full M–θ curve. In this work it is defined as the tangent tension.
to the M–θ curve with a slope given by Sj,ini/h, where h is adjusted for Fig. 15 represents the critical component end-plate in bending for
each case. Finally, it is noted that this procedure is very sensitive to the J1.1, J3.1 and J4.1 joints, also representing the moments achieved
the adjustment of the tangent point and the slope, in comparison to by each bolt-row, using for the lever arm the distance between each
analytical procedures. row to the middle of the compressed beam flange. It is possible to con-
Table 7 compares the moment resistance and initial stiffness (equal clude that J4.1 is more ductile than J1.1, roughly twice and that the
for hogging and sagging because of the symmetry of the connection), bending moments are well estimated by EN 1993-1-8.
using the linearized responses of the experimental and numerical
results.
5.3.3. Column web panel in shear
Table 9 summarizes the initial stiffness and resistance of the compo-
5.3. Components behaviour nent column web panel in shear, comparing the experimental results,
using the same procedure illustrated in Fig. 13(d), with the ones obtain-
5.3.1. Numerical versus experimental results ed with the Atamaz-Jaspart [11] model (AJM) and the Krawinkler [9]
In a ductile extended end-plate joint it is expected that the most dis- model.
sipative components are the end-plate and/or column flanges in bend- Fig. 16 shows the comparison of the column web rotation between
ing and the column web panel in shear. In the particular case of the the experimental test and the model proposed by Krawinkler as well
Table 5
Geometrical properties of the FE models.
Column Beam ht (mm) Lc (mm) Lc1 (mm) Lc2 (mm) L1 (mm) L2 (mm) Lb1 (mm) d (mm)
J1.1 HEA320 IPE360 3260 3000 1375 740 1322 1385 468 1166
J3.1 HEB320 IPE360 3260 3040 1395 740 1340 1385 468 1180
J4.1 HEA320 HEA280 3260 2991 1298 690 1316 1483 358 1159

500 500
400 400
300 300
200 J1.1 - Joint rot. - Exp. 200 J3.1 - Joint rot. - Exp.
Joint rot. - FE Joint rot. - FE
Linearization Exp. Linearization Exp.
100 100
Linearization FE Linearization EF
EN1993-1-8 EN1993-1-8
0 0
0 20 40 60 80 100 0 20 40 60 80 100
(a) (b)
500 Mj
400 Kst=Sj,ini/h
300 Mpl B
200 J4.1 - Joint rot. - Exp. Mel A

Joint rot. - FE Mel ≤ 23 Mpl
Linearization Exp.
100 Linearization FE
EN1993-1-8
Sj,ini θ
0
0 20 40 60 80 100
Rotation (mrad)
(c) (d)
Fig. 13. Global experimental, numerical and analytical results.
as the Atamaz-Jaspart model. It is noted that the Krawinkler model con- stiffeners, and the logarithmic strain from the corresponding
siders that the post-yielding stiffness is developed up to the yielding of nodes in the FE models, revealing a good agreement.
the column flanges assumed equal to four times the yielding rotation
(distortion). After the yielding of the flanges the hardening of the steel 6. Characterization of the column web components using FE models
is 2% of the initial stiffness. The reference points for the AJM are:
(i) first yielding of the web; (ii) strain-hardening in the web, γst, and 6.1. Component's identification methodology
plastic shear stress uniformly distributed in the whole panel, Vny,
(iii) ultimate shear force, Vnu, followed by a plateau until the joint's rup- In this section a methodology to extract the column web panel
ture, see Fig. A.3(b) and Table A.4 from Annex A for more details. components behaviour, due to the load-introduction effect and the
From the figures it is possible to observe that for the J1.1 and J3.1 the shear effect, is presented. It is fully based on the results provided
initial stiffness obtained with the Atamaz-Jaspart model has a better by the FE models and allows the derivation of force-displacement
adjustment to the experimental curve, while for the J4.1, the Krawinkler and shear-distortion curves of the components. This methodology
model provides a better adjustment. In terms of resistance, both was developed using numerical integration of the stress fields
methods achieved good results, although the Krawinkler method obtained from FE models in order to characterize the component
seems to be more conservative than the Atamaz-Jaspart model. In column web panel in shear and the component column web in
Fig. 16 it is also depicted a comparison between the top and bottom transverse compression/tension. The methodology is based on con-
strain gauges, positioned in the column web at 60 mm from the siderations presented in [13] for welded joints, it is summarized in
Table 6
Main design properties of the joints according to EN1993-1-8.
Test Global prop. Component prop.
Shear Compr. Tension comp.

comp. comp.
Sj,ini Mpl,Rd Vwp,Rd Fc,fb,Rd 1st Row 2nd Row 3rd Row
(kNm/rad) (kNm) (kN) (kN)
Ft,wc,Rd Ft,fc,Rd Ft,ep,Rd Ft,wb,Rd Ft,Rd Ft,wc,Rd Ft,fc,Rd Ft,ep,Rd Ft,wb,Rd Ft,Rd Ft,wc,Rd Ft,fc,Rd Ft,ep,Rd Ft,wb,Rd Ft,Rd
(kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN)
J1.1 60,246 355 1025 1288 846 601 534 – 534 835 598 604 988 491 835 598 604 986 0
J3.1 72,950 389 1166 1288 929 727 534 – 534 918 722 604 966 604 918 722 604 986 27
J4.1 37,865 271 1050 1776 1110 568 574 – 568 1080 560 591 1680 483 1080 560 591 1680 0
Bold data indicates significance the driving components.

Underline data indicates significance the components that cannot developed full strength because the compression component has reached its full strength.
Table 7
Comparison of strength and stiffness between the experimental, numerical and analytical results.
Experimental values Numerical values Analytical values
Mpl (kNm) Sj,ini (kNm/rad) Mrd (kNm) Error (%) Sj,ini (kNm/rad) Error (%) Mrd (kNm) Error (%) Sj,ini (kNm/rad) Error (%)
J1.1 335 71,340 339 +1.2 60,660 −15.0 354.5 +5.8 60,246 −15.6
J3.1 395 95,057 391 −1.0 77,480 −18.5 389.3 −1.4 72,950 −23.3
J4.1 304 38,495 312 +2.6 34,125 −11.4 270.5 −11.0 37,865 −1.6
0 1
ð2t fc =3þr 0h 1
Table B.2 and illustrated in Figs. 17 to 19 for the load-introduction Z Þback Zwc
B C
effect and for the shear effect, respectively. The forces are obtained B
Vn ¼ @ C
τ23 dzA t 1 þ @ τ 23 dzA t wc
by integration of the stress field according to Eqs. (B.21) and
(B.23) due to the load-introduction effect and from Eq. (B.16) due 0 1
ð2t fc =3þr
Z Þfront
to the shear effect. In these equations, σ33 (or σzz) is the normal hor- B C
izontal stress in the column web along path P1, integrated over the þB
@ τ23 dzC
A t1 : ð2Þ
lengths hc for compression and ht for tension (Eq. (1)), τ23 (or τyz) is
the shear stress in the column web along path P3, integrated over the
corresponding lengths (Eq. (2)), tfc and twc are the column flange and
web thicknesses, respectively, and t1 is the thickness according to the
shear area defined in Fig. 19(b). The location of path P1, along which the normal stress diagrams are
evaluated to compute the compression and tension forces, coincide
0h 1 0 1 with the inner nodes of the first line of elements in the web beyond
Zc Zht;i
B C the root radius, as identified in Fig. 18(b). This allows the identification
F c ¼ @ σ 33 dyA t wc or F t;i ¼ @ σ 33 dyA t wc : ð1Þ
of the forces introduced by the beam into the web and avoids the stress
perturbation observed in the corner.
500 kNm.mrad
35000
400 30000
25000 22992 22753
18683 18859 FE (J1.1)
300 20000 EXP. (J1.1)
15000
200 10000
End-plate + c. flanges rot. - Exp. 4310 3973
5000
End-plate + c. flanges rot. - FE
100 0
Column web panel rot. - Exp. (J1.1)
100% 81% 19%
Column web panel rot. - FE
0 100% 83% 17%
0 20 40 60 80 100 Total Energy Column web panel End-plate and c.
Rotation (mrad) Dissip. flanges
500
kNm.mrad
35000
400 30000
FE (J3.1)
25000 18990 19369
20000 EXP. (J3.1)
300 14628
12943
15000
10000 6539
200 4362
End-plate + c. flanges Rot. - Exp. 5000
End-plate + c. flanges rot. - FE 0
100 100% 77% 23%
Column web panel rot. - Exp. (J3.1)
Column web panel rot. - FE 100% 66% 34%
0
500
kNm.mrad
35000 31716 31360
400 30000
23807 23318
25000 FE (J4.1)
300 20000 EXP (J4.1)
15000
7909 8052
200 10000
End-plate + c. flanges rot. - Exp. 5000
End-plate + c. flanges rot. - FE 0
100
Column web panel rot. - Exp. (J4.1) 100% 75% 25%
Column web panel rot. - FE
0 100% 74% 26%
Fig. 14. Comparison of the moment–rotation responses and the energy dissipated.
Table 8
Strength and initial stiffness of the critical components.
Test End plate in bending
Experimental EN1993-1-8
Kij Me Mpl Kij M1st M1st + 2nd M1st + 2nd + 3rd

(kNm/rad) (kNm) (kNm) (kNm/rad) (kNm) (kNm) (kNm)
J1.1 155,000 233 350 133,416 210.4 354.5 354.5

J3.1 208,000 267 400 160,579 210.4 387.8 389.3
J4.1 175,000 200 300 68,003 172.3 270.5 270.5

To determine the shear force, the stress field is evaluated at a properties adopted in the analyses the nominal values in Eurocode 3
cross section in the middle of the column length, where the stresses part 1–1 [27] were used (E = 2.10E5 N/mm2; fy = 355 N/mm2; fu =
reach their maximum values, also using the middle elements of the 490 N/mm2) even for the 40 mm plate.
web to extract the shear stresses, see Fig. 19(b). The stresses in The main design properties, according to EC3-1-8, are shown in
the column web panel may be assumed to be uniformly distributed, Table 11. Notice that C1_1ppv2_tp40 was designed so that the weakest
due to the action of the column flanges, as proposed by Jaspart [11]. components are the column web in tension and compression. The
Therefore, the shear stresses are only extracted along path P3. 40 mm thick end plate was adopted to be sufficiently stiff to allow the
The elongation and rotation corresponding to the load-introduction column web to be the main source of plastic deformation.
effect and shear effect, respectively, are obtained from the displacement
fields. In the case of the load-introduction effect, the displacements are
extracted from paths P1 and P2, and the elongation is determined as the

difference between them, see Fig. 18(a), Eqs. (B.22) and (B.24). In the 500
case of the shear rotation, the displacements are extracted from the
predefined nodes indicated in Fig. B.1 using Eq. (B.17). 400
To account for the additional shear strength provided by the frame
formed by the column flanges and the transverse stiffeners aligned 300
with the beam flanges, a plastic mechanism is idealised considering
that the plastic hinges will occur in the column flanges due to the higher 200
Exp. End-plate rot. (J1.1)
stiffness of the stiffeners. The stresses are extracted at a section near the
100 EN1993-1-8 - 1st row
flanges maxima normal stresses, σ22, as depicted in Fig. 20. Subsequent- EN1993-1-8 - 1st + 2nd rows
ly, Eq. (3) is used to determine the bending moment in the flanges, and Linearization
0
Eq. (4) is used to obtain the additional shear strength Vc. 0 10 20 30 40 50
Rotation (mrad)
Zt fc
Mfc ¼ zðσ 22 dAÞ: ð3Þ 500
MfcðP4Þ þ M fcðP5Þ þ M fcðP6Þ þ MfcðP7Þ 400

Vc ¼ : ð4Þ
hb −t s
300
For simplification it is considered that the stresses are uniform along
the width of the column flanges, so the integration of the stress along 200
Exp. End-plate rot . (J3.1)
each path results in a single force at that thickness level. The bending mo- EN1993-1-8 - 1st row
ment is obtained by multiplying the force at each layer by the distance to 100
EN1993-1-8 - 1st + 2nd rows
the neutral axis. Taking into account the three elements and four lines of EN 1993-1-8 - 1st+2nd+3rd rows
0
nodes in the flange thickness, the bending moment can be computed by
0 10 20 30 40 50
the forces determined in the two outer paths (P4.1 and P4.2). However, Rotation (mrad)
and because it is necessary to determine the neutral axis position, a
third path (P4.3) is required to determine the location where the stresses 500
are zero. It is also important to note that the third path crosses the radius
and web elements and hence the integrated forces and the neutral axis 400
evaluation need to take that into account.
300
6.2. Application to the joints
200
Exp. End-plate rot . (J4.1)
The methodology described above was applied to the calibrated FE Linearization
100
models for test J1.1. The force–deformation (F–Δ) and shear–distortion EN1993-1-8 - 1st row
(V–γ) relationships of the column web panel components were deter- EN1993-1-8 - 1st + 2nd rows
0
mined for use in a mechanical model capable of characterizing the contri- 0 10 20 30 40 50
bution of these components to the joint rotation. Rot.ation (mrad)
Some variants of the original test specimen were considered in order
to assess the sensitivity of the methodology to the presence of trans- Fig. 15. End-plate and column flanges M–θ relationships for the J1.1, J3.1 and J4.1,
verse column web stiffeners, see Table 10. For the steel mechanical respectively.
Table 9
Strength and stiffness of the critical component column web in shear.
Test Distortion of the column web panel
Experimental Krawinkler model Atamaz-Jaspart model
Kij Me Mpl Kij Me Mpl Kij Me Mpl

(kNm/rad) (kNm) (kNm) (kNm/rad) (kNm) (kNm) (kNm/rad) (kNm) (kNm)
J1.1 126,000 220 330 80,120 260.72 318.39 113,954 209.2 351.4
J3.1 190,500 260 390 110,111 305.03 392.54 153,122 250.4 388.9
J4.1 50,500 193 290 60,901 200.80 260.02 84,325 158.6 267.1
6.2.1. Load-introduction effect results due to the fact that the introduction of the loads in the web is not
The stress and deformation fields are represented in Fig. 21. completely horizontal due to the shear in the beam. It is also important
By analysing the stress fields in the joint, it can be concluded that the to realise that the maximum web elongations occur aligned with the
maximum stresses evolve from a more or less symmetric configuration, second bolt row under the beam flange because of the higher stiffness
in the two tension bolt rows in the elastic range with identical stress of the second bolt row in tension.
values for the external and internal row, to a configuration in the plastic By applying the methodology described before it is possible to com-
range where a single maximum stress occurs for the two tension bolt pare the numerical F–Δ results with the Atamaz-Jaspart model [8,11]
rows between the external line of bolts and the beam flange. This is and the EC3-1-8 [1] design procedure, see Fig. 22. The computed F–Δ
600

600
500 500
400 400
300 300
200 Exp. column web panel rot. (J1.1) 200 Exp. column web panel rot . (J3.1)
Linearization Linearization
100 Krawinkler model 100 Krawinkler model
A-J model A-J model
0 0
0 20 40 60 80 100 0 20 40 60 80 100
600
600
Exp. sup J3.1
500 500 FE sup J3.1
Exp. inf J3.1
400 400 FE inf J3.1
300 300
200 Exp. column web panel rot. (J4.1) 200

Linearization
100 Krawinkler model 100
A-J model
0 0
0 20 40 60 80 100 -1500 -500 500 1500
Rotation (mrad) Strain (μ)
Fig. 16. Column web panel M–θ relationships for the J1.1, J3.1 and J4.1, respectively. And comparison of the strains obtained in the column web panel of J3.1.
800 1000
700 Ft(1st) 900 Ft(1st)
h t(1st) h t(1st) 800
600
500
Ft(2nd) 700
h t(2nd) 600 Ft(2nd)
h (mm)
h (mm)
400 h t(2nd) 500

400
hc(3rd)
300 Fc(3rd)
300
200
200 hc(4th)
Fc 100 hc Fc(4th) 100
0 0
-800 -600 -400 -200 0 200 400 600 800 -800 -600 -400 -200 0 200 400 600 800
σ33 (N/mm2) σ33 (N/mm2)
(a) (b)
Fig. 17. Procedure to obtain the normal force of the column web in transverse tension and compression components: (a) without stiffeners, (b) with stiffeners.
Fig. 18. (a) Procedure to obtain the web elongation of the column web in transverse tension and compression components; (b) location of the paths P1 and P2.
Fig. 19. Procedure to obtain the shear force of the column web panel in shear component.
relationship was, in general, slightly stiffer than the analytical predic- relationship, as in the AJM a multi-linear approach is used and in the
tions, although the initial stiffness is well adjusted to the AJM for the iso- FE model a nonlinear relationship is used. In the case of the isolated
lated behaviour of the two tension bolt rows. In terms of resistance, for rows behaviour it is possible to see that the F–Δ relationship obtained
the column web in compression there is a good agreement between the for the first bolt row is different from the second bolt row, mainly for
AJM and the numerical F–Δ relationship, although for the model with the model with tp = 40 mm. This difference can be explained by
tp = 18 mm the ultimate force obtained in the numerical simulation is analysing the stress fields depicted in Fig. 23.
slightly larger than the analytical prediction. For the tension compo-
nent, grouping the first and second bolt rows, the AJM predicts well
the first yielding resistance and the ultimate resistance of the compo- 6.3. Shear results
nent, although the Δu prediction in the AJM model occurs for an elonga-
tion larger than in the F–Δ relationship. This can be related to the By applying the methodology described before it is possible to
stiffness of the hardening branch that depends on the steel uniaxial obtain the V–γ relationship. The stress fields for the joints with
Fig. 20. Determination of the bending moment in the flanges by the stress integration.
Table 10
Joints analysed.
Joints based on J1.1 ts (mm) tp = 18 mm tp = 40 mm

ppv = 100 mm ppv = 120 mm
mx = 40 mm mx = 50 mm
hp = 540 mm hp = 560 mm
M24 (10.9) M30 (10.9)
Without stiffeners C1_1ppv1_tp18_ts0 C1_1ppv2_tp40_ts0
20 C1_1ppv1_tp18_ts20 C1_1ppv2_tp40_ts20
Table 11
Main design properties of the joints according to EN1993-1-8.
Model Global prop. Component prop.
Shear comp. Compr. comp. Tension comp.
Sj,ini Mpl,Rd Vwp,Rd Fc,wc,Rd Fc,fb,Rd 1st row 2nd row
Ft,wc,Rd Ft,fc,Rd Ft,ep,Rd Ft,wb,Rd Ft,Rd Ft,wc,Rd Ft,fc,Rd Ft,ep,Rd Ft,wb,Rd Ft,Rd
(kNm/rad) (kNm) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN) (kN)
C11ppv1_tp18ts0 45,394 212 759 566 1042 646 507 459 – 459 646 507 520 781 107
C11ppv1_tp18ts20 61,274 290 832 – 1042 671 515 459 – 459 662 512 520 781 373
C11ppv2_tp40ts0 48,676 228 759 581 1042 528 536 910 – 528 528 536 1010 763 52
C11ppv2_tp40ts20 64,003 300 832 – 1042 528 536 910 – 528 528 536 1010 763 304

and without column web stiffeners are quite similar, as shown in with or without stiffeners. From the numerical F–Δ curve a good
Fig. 24. agreement can be seen for the additional shear strength, due to the
Fig. 25 compares the numerical results with the AJM [8] and the EC3- stiffeners, between the proposed procedure and the EC3-1-8 and
1-8 model. It is possible to conclude that there is a good agreement be- the AJM. The differences found can be justified by the partial hinges
tween the numerical V–γ relationship and the AJM, and consequently observed in the column flanges of the FE models, as can be seen in
with the EC3-1-8 model. The prediction of the bare web shear contribu- Fig. 28.
tion, in the AJM, (dash line) achieves a slightly better agreement, than The influence of the transverse stiffeners in the joints strength and de-
the prediction of the sum of the web and stiffeners contribution together formation can be neglected in the elastic range and increases with the
(solid line). yielding of the column flanges, as shown in Fig. 26. Fig. 26(b) represents
The additional shear resistance provided by the stiffeners is the relative reduction of the rotation due to the stiffeners, obtained by
achieved by the frame action developed by the column flanges and the difference in the rotation achieved for the same level of strength,
the stiffeners, as the bare column web shear resistance is similar at each load increment, taking as reference the maximum rotation
Fig. 21. Stress fields (left), and web elongation fields (right), for increasing levels of bending moment.
1200 0
-8 -6 -4 -2 0
1000 -200 FE (tp18) FE (tp40)
Ft row 1+2 (kN) JM (tp18) JM (tp40)
800 EC3 (tp18) EC3 (tp40)
-400
1 yield (tp18) 1 yield (tp40)
Fc (kN)
600
-600
400 FE (tp18) FE (tp40)
JM (tp18) JM (tp40) -800
200 EC3 (tp18) EC3 (tp40)
1 yield (tp18) 1 yield (tp40) -1000
0
0 2 4 6 8 -1200
Web elongation (mm) Web elongation (mm)
(a) (b)
1200 1200
JM (tp18) JM (tp40)
1000 1000 Ft,wc,Rd(tp18) Ft,wc,Rd(tp40)
FE (tp18) FE (tp40)
Ft row1 (kN)
Ft row2 (kN)
800 800 1 yield (tp18) 1 yield (tp40)
600 600
400 FE (tp18) FE (tp40) 400

JM (tp18) JM (tp40)
200 Ft,wc,Rd(tp18) Ft,wc,Rd(tp40) 200
1 yield (tp18) 1 yield (tp40)
0 0
0 2 4 6 8 0 2 4 6 8
Web elongation (mm) Web elongation (mm)
(c) (d)
Fig. 22. F–Δ behaviour for the components due to the load-introduction effect without column web stiffeners: (a) for the tension zone bolt rows 1 and 2 together, (b) for the compression
zone, (c) and (d) for the tension zone bolt rows 1 and 2, respectively.
imposed in the joints, leading to a reduction of the rotation of 55% in ABAQUS uses solid (or continuum) elements in the connection zone.
for C1_1ppv1_tp18 and 41% for C1_1ppv2_tp40. For computational efficiency beam elements were used in the column
The evolution of the flanges bending moment (Mfc) vs the web rota- and beam regions located away from the connection. The model
tion is depicted in Fig. 27(a) and (b) for the two joints with transverse considers non-linear material and geometrical behaviour, non-linear con-
stiffeners. Note that the stress distribution of Mfc near the bolt row in tacts, re-contacts and slip. A material model combining both isotropic and
tension is affected by the presence of the holes in the flange. The stress kinematic hardening was employed.
field is also affected by the bolt forces transferred to the column flange, The model was calibrated against experimental tests carried out at
as can be seen in Fig. 28, resulting in a shift of the position of the plastic the University of Coimbra on double extended end-plate beam-to-
hinge towards the bolt row alignment. column joints. The experimental tests were extensively detailed, and
the results were analysed, globally and at the component level, and
7. Conclusions they were compared with the analytical values obtained using the
EN1993-1-8 and other methodologies, namely the Krawinkler model
This paper described the development of a finite element model for the column web distortion. The excellent agreement between the
capable of representing the behaviour of extended end-plate joints clas- experimental data and the numerical results allowed to conclude that
sified as partial-strength according to Eurocode 3. The model developed the model is capable of simulating with accuracy the behaviour of
Fig. 23. Detailed stress fields for the column web in tension.
Fig. 24. Shear stress fields for increasing levels of bending moment (a) for the unstiffened column web and (b) for the stiffened web.
1200 1200
C1_1ppv1_tp18 C1_1ppv2_tp40
1000 1000
800 800
V (kN)
V (kN)
600 600
FE (ts0) FE (ts0)
FE(ts20) FE(ts20)
400 AJM (ts0) 400 AJM (ts0)
AJM (ts20) AJM (ts20)
200 EC3 (ts0) 200 EC3 (ts0)
EC3 (ts20) EC3 (ts20)
0 0
0 25 50 75 100 125 0 25 50 75 100 125
γ (mrad) γ (mrad)
(a) (b)
Fig. 25. V–γ behaviour for the shear load.
600 600
500 491 500

459
400 370 400
300 336 300
200 C1_1ppv1_tp18_ts0 200

C1_1ppv1_tp18_ts20
100 C1_1ppv1_tp18
100 C1_1ppv2_tp40_ts0
35 C1_1ppv2_tp40_ts20 C1_1ppv2_tp40
55
0 0
0 50 100 150 200 0% 10% 20% 30% 40% 50% 60%
Rotation (mrad) Rotation reduction due to stiffeners (%)
(a) (b)
Fig. 26. Influence of the column web stiffeners in the joints strength and deformation.
8 8
6 6
4 4
Mpf (kNm)
Mpf (kNm) 2
C1_1ppv1_tp18_ts20
2
C1_1ppv2_tp40_ts20
0 0
-2 -2
P4 -4 P4
-4
P5 P5
-6 P6 -6 P6
P7 P7
-8 -8
0 50 100 150 0 50 100 150
γ (mrad) γ (mrad)
(a) (b)
Fig. 27. Evolution of the bending moments in the column flanges.
end-plate beam-to-column joints, and that it can be used in detailed the beam to the column. The model deformability component responses
analyses of the components behaviour. can be implemented in a spring's model assembled according to the
A practical and efficient methodology for the extraction of the joints geometry, or used directly in frame analysis and design.
behaviour of the individual components in the column web panel The double extended end plate steel joints analysed in this paper
from experimental tests and numerical simulations was presented, ap- proved to be an excellent solution to be used in moment-resisting
plicable to bolted end plate joints. It uses the integration of stress fields frame structures, due their ductility, reversal loading ability and energy
from the FE models of the joints, namely the column web panel in shear dissipation capacity, particularly when continuity stiffeners were
and the column web in transverse compression/tension components. It employed which contributed to increase the stiffness and strength of
is worth noting that the accuracy of the methodology depends of the the web column region.
column web element size and the number of stress fields analysed
from the available load increments. It is also important to refer that Acknowledgment
this methodology can be time consuming, but if well applied, the
resulting force–displacement relationships can be easily used in me- Financial support from the Portuguese Ministry for Education and
chanical models. Science (Ministério da Educação e Ciência) (through the agency
Double-extended end plate joints were analysed and used to assess Fundação para a Ciência e a Tecnologia (FCT)) under contract grant
the proposed methodology. The numerical results were compared with PTDC/ECM/116904/2010 (3D Joints), and also within QREN–
the EC3-1-8 results, the Atamaz-Jaspart and the Krawinkler models. The POPH — Typology 4.1 — Advanced Education, reimbursed by the
numerical results and the proposed methodology generally confirm the European Social Fund and Portuguese National Funds MEC, under
accuracy of the Atamaz-Jaspart model and the EC3-1-8 design model, contract grants SFRH/BD/91167/2012, for Hugo Augusto, is grate-
further allowing to explain in a detailed way the load transfer from fully acknowledged.
Fig. 28. σyy stress distribution comparison for the C1_1ppv1_tp18_ts20 on the left and C1_1ppv2_tp40_ts20 on the right.
Annex A. Design models for column web panel components
A.1. Geometrical definitions and global joint properties
Fig. A.1. End plate joint geometrical properties.
Table A.1
Eurocode 3 [1] rotational stiffness and moment resistance main equations.
Stiffness (S) Resistance (M)
Sj ¼ Ez2 M j;Rd ¼ ∑ hr F tr;Rd If Mj,Ed ≤ 2/3Mj,Rd ⇒ μ = 1

1
μ∑
r
If 2/3Mj,Rd b Mj,Ed ≤ Mj,Rd ⇒ μ = (1.5Mj,Ed/Mj,Rd)ψ
i ki
Where
z lever arm
μ stiffness ratio
ψ coefficient that depends on the type of connection
ki stiffness coefficient for basic joint component i
hr distance from bolt-row r to the centre of compression
Ftr,Rd effective design tension resistance of bolt-row r
r bolt-row number
E modulus of elasticity
A.2. Krawinkler model
The Krawinkler model [9] is an analytical representation of the shear behaviour of the column web panel in shear, stiffened by continuity
stiffeners, using a tri-linear model (see Fig. A.2 and Table A.2).
M
Kcws,h
Vp ht
Kcws,p h2t
Vy ht
ht
Kcws h2t
γy 4γ y γ
Fig. A.2. Krawinkler model [9].
Table A.2
Krawinkler formulation [9].
Stiffness (K) Resistance (M) Rotation (γ)
Yielding of the panel zone 2 Gt wc hc hb f f

K cws ht ¼ V y ht ¼ pyffiffi t wc hc hb γ y ¼ pffiffi3yG
β 3 β
Yielding of the column flanges in bending 2
K cws;p ht ¼
24EIfc 1 Vpht = Vyht + 3Kcws,ph2t γy γp = 4γy
5t fc β
Strain hardening of the panel zone 2
K cws;h ht ¼ Eh 2
K cws ht – –
E
Where
ht lever arm
Eh is the hardening modulus of the material;
Ifc is the inertia of the column flanges including the web transition radius (Ifc = bc·t3fc / 12)
β is a factor to account for the beneficial effect of column shear above and below the joint, it can be taken equal to 1
E modulus of elasticity
G shear modulus
A.3. Atamaz-Jaspart model
The multi-linear Atamaz-Jaspart model [11] covers the load-introduction effect Fig. A.3(a) and Table A.3 and the shear effect, Fig. A.3(b) and
Table A.4 for bolted end plate joints with and without continuity stiffeners.
(a) (b)
Fig. A.3. Multi-linear model [11] for: (a) load-introduction effect and (b) shear effect.
Table A.3
Multi-linear model values due to the load-introduction effect for joints with extended end-plate.
Stiffness (K) Resistance (F) Displacement (Δ) Effective length
a) 2t wc F be Compression:
K be ¼ 2kcw
λ μ F be ¼ λ μσ ciy Δe ¼ K bi pffiffiffi
with: beff ;wc ¼ t f b þ 2a f 2 þ 2t p
with: with:
kcw = Etwc/hw Tension:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi μ ¼ 1−e−ξξ ;cosξ τ ce ¼ F be −V ce
Avc beff,wc = tfb + 2wp
λ ¼ 4 kcw =4EIfc qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðσ ciy Þ2 þ 3ðτ cy Þ2 ¼ f ywc
ξ = beff,wcλ/2
b) K bfst ¼ kbe EEst Fbppl = twcbeff,wcσciy Compression:

pffiffiffi
Δst = εsthwc/2 beff ;wc ¼ t f b þ 2a f 2 þ 2t p þ 5ðt fc þ r c Þ
Adopted material uniaxial response
c) Minimum of Fbuy and Fbb: F Tension:

Δu ¼ K buy
Fbuy = twclpσciu bfst beff,wc = beff,t,wc,EC3
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi with:
F bb ¼ μ F be F bcr qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Compression:
pffiffiffi
with: ðσ ciu Þ2 þ 3ðτ cu Þ2 ¼ f uwc beff ;wc ¼ t f b þ 2a f 2 þ 2t p
Fbcr = (hc − 2tfc) ⋅ twc ⋅ k ⋅ D F bu −V cu
2
τ cu ¼ Avc
π2 E t wc
D ¼ 12ð1−ν 2 Þ ðh −2t Þ
c wc
a) First yielding of the of the web under the combined stresses τ–σi.
b) Pseudo-plastic force in the web, Fbppl, followed by strain hardening in the web, Δst.
c) Ultimate force, Fbu, carried over from the beam to the column.
Table A.4
Multi-linear model values due to the shear effect of steel joints.
Stiffness (K) Resistance (V) Rotation (γ) Stiffeners
Vne = min(τyc(tens); τyc(comp))Avc

γ y ¼ VKne
a) Kn = GAvc
with: with: n
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
G = E/2(1 + v) ðσ ciy Þ2 þ 3ðτ cy Þ2 ¼ f ywc
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
λ ¼ 4 kcw =4EIfc F be −V ce
τcy ¼ Avc
ξ = beff,wcλ/2
b) Adopted material uniaxial response Vny = τcyAvc γy = Vy/Kn 24EIfc

pffiffiffi Kf ¼
with: γ st ¼ 0:5 3ðεst −εy Þ þ γy d2b
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðσ cny Þ2 þ 3ðτ cy Þ2 ¼ f ywc 4Mpl;fc
with: V cy ¼ db
F by −V cy εy = fy/E
τcy ¼ Avc
Mc
σ ny ¼ W
c
cy
þ NAcc
pffiffiffi
c) Knst = GstAvc Vnu = τcuAvc γu ¼ 3ðεu −εst Þ þ γ st
with: with:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Gst = Est/3
Est ≅ E/50
ðσ cnu Þ2 þ 3ðτcu Þ2 ¼ f uwc
F bu −V cu
τcu ¼ Avc
σ cnu ¼ WMcuc þ NAcc
a) First yielding of the of the web under the combined stresses τ–σi.
b) Plastic shear stress uniformly distributed in the whole panel, using the interaction of stresses τ–σn, followed by the apparition of the strain hardening in the web, γst.
c) Ultimate shear force, Vnu.
Table A.5
Notation
kcw — is a parameter used in the calculation of the load-introduction stiffness;

λ, ξ and μ — are parameters related with the assessment of the column web elastic foundation properties, for the calculation of column flanges bending resistance in the
load-introduction effect;
Avc — is the column web shear area;
Ifc — is the inertia of the column flanges plus the web transition radius (for the Krawinkler model Ifc = bc·t3fc / 12);
beff,wc — is the effective width of column web in compression or in tension, for tension the EC3-1-8 effective width recommendations should be used;
Kbe and kbfst — is the elastic and post-elastic stiffness for the load-introduction model, respectively;
Kn and knst — is the elastic and post-elastic stiffness for the shear model, respectively;
Fbe — is the value of Fb when the web panel reaches the first yielding (associated to the biaxial stress state σciy vs τcy) for the load-introduction model;
Fbppl — is the value of Fb when the pseudo-plastic moment Mbppl of the joint is reached, for the load-introduction model;
Fbuy — is the value of Fb when the web panel reaches the ultimate plastic resistance (associated to the biaxial stress state σciu vs τcu) for the load-introduction model;
Fbb — is the value of Fb when the web panel reaches the instability under compression effects, for the load-introduction model;
Fbcr — is the value of Fb for the elastic linear instability load of the web, for the load-introduction model;
k — is a coefficient that accounts for the type of loading, and web support conditions (1 for cross nodes symmetrically loaded and 2 for external nodes);
Vce and Vcu — shear forces entering in the web from the global system, corresponding to the first yielding and the ultimate resistance of the web;
fywc and fuwc — are the nominal values of the yield strength and the ultimate tensile strength of the steel grade adopted for the column web;
σciy and τcy — are the horizontal normal stress and shear stress, respectively, in the Von Mises criterion that induces the first yielding in the column web panel;
σciu and τcu — are the horizontal normal stress and shear stress, respectively, in the Von Mises criterion that induces the ultimate plastic stress state in the column web panel;
Δe, Δst and Δu — are the column web relative elongation, between the axis of the web and the introduction-load zone, at the first yielding in the web, the apparition of the
strain-hardening in the web and at the ultimate web resistance;
Vne, Vny and Vnu — are the column web shear resistance at the first yielding, at the complete yielding of the whole panel and at the ultimate plastic resistance, respectively, for
the shear model;
γe, γst and γu — are the column web shear rotation at the first yielding in the web, the apparition of the strain-hardening in the web and at the ultimate web resistance, for the
shear model, respectively, for the shear model;
σcny and τcy — are the vertical normal stress and shear stress, respectively, in the Von Mises criterion that induces the complete yielding of the whole panel;
σcnu and τcu — are the vertical normal stress and shear stress, respectively, in the Von Mises criterion that induces the ultimate plastic stress state in the column web panel;
Mc and Nc — are the bending moment and axial forces entering in the web from the global system;
Vcy — is the additional plastic web capacity, due to the presence of the transverse stiffeners aligned with the beam flanges, that forms with the column flanges a frame with
additional shear resistance;
kf — is the rate of increase of the additional plastic web capacity resistance Vcy.
Annex B. Extraction procedures from experimental tests and numerical models
(a) (b)
Fig. B.1. Points of interest: (a) in the experimental tests and (b) in the FE models.
Table B.1 shows the equations used in the extraction of the M–θ relationships, using typically available instrumentation of the experimental tests.
The left column of the table lists the equations for the experimental tests and in the right column the ones for the FE models.
Table B.1
Equations for the extraction of the M–θ curves from the experimental and numerical results.
Experimental FE
MEd = Fload_cell ⋅ d (B.1) MEd = RDT20 ⋅ dDT20 (B.2)

θtotal = θcolumn_web + θend − plate − θelast_column − θelast_beam − θblock (B.3)
θtotal ¼ arctanðDT11−DT12
d5
Þ−θelast beam −θelast column −θblock or
(B.4) θtotal ¼ arctanðDT1U3 −DT2U3 Þ þ θend−plate −θelast column
(B.7)
DT1DT2
θtotal ¼ L1 Þ−θelast beam −θelast column −θblock or
arctanðDT20 (B.5) or (B.8)
θtotal ¼ arctanðDT3−DT4 Þ þ arctanðDT1−DT2Þ−θelast column −θblock (B.6) θtotal ¼ arctanðDT11U3 −DT12U3 Þ−θelast coluna
d4 d3 DT11DT12
θcolum web ¼ arctanðDT3−DT4
d4
Þ−θelast column −θblock
(B.9) θcolumn web ¼ arctanðDT1U3 −DT2U3 Þ−θelast column
(B.10)
DT1DT2
θend−plate ¼ arctanðDT1−DT2Þ−θelast beam
DT1DT2
(B.11) θend−plate ¼ arctanðDT13U3 −DT14U3 Þ−θcolumn web (B.12)
DT13DT14
F load cell dL22 2 F load cell dL (B.13)
þ − F load cell dL2
θelast column ¼ L 6
EIyðColumnÞ
x2 (B.14)
F load cell dx− F load cell 2
θelast beam ¼ EI yðBeamÞ
θblock ¼ arctanðDT5−DT6Þ (B.15) –
d1
See also Fig. 5.

θblock parasitic displacements (global rotation of the setup due to the gaps)
θelast_beam elastic deformation of the beam
θelast-column elastic deformation of the column
θcolumn_web column web panel rotation
θend-plate end plate rotation
There are several ways to measure the global rotations, for the experimental tests, the one that has the best results is given by Eq. (B.4), because
can capture all the deformability sources and is sufficiently near to the joint to avoid bigger contributions of the beam elastic deformation. Due to the
lesser instrumented J3.1 it was necessary to use Eq. (B.5) to obtain the total rotation. Because in the numerical models DT11 and DT12 couldn't be
Table B.2
Equations for the extraction of the F–Δ curves from the numerical results.
Component Resistance (F) Deformation (Δ)

! !
Column web panel in shear ð2t fc =3þrÞback hwc (B.16) γ ¼ arctanðDT1−DT2
h
Þ þ arctanðDT3−DT4
hc
Þ (B.17)
Vn ¼ ∫ τ23dz t1 þ ∫ τ23dz t wc þ b
!
ð2t fc =3þrÞ
∫ front
τ23dz t1
Column web panel in shear with stiffeners t fc (B.18)

Mfc ¼ ∫ zðσ 22 dAÞ
MfcðP4Þ þMfcðP5Þ þMfcðP6Þ þMfcðP7Þ (B.19)
Vc ¼ db
V = Vn + Vc (B.20)
!
Column web In transverse compression hc (B.21) Δc = (ΔP1 − ΔP2)compressed_beam_flange (B.22)
Fc ¼ ∫ σ 33 dy t wc
!
Column web in transverse tension ht;i (B.23) Δt,i = (ΔP1 − ΔP2)row_i (B.24)
F t;i ¼ ∫ σ 33 dy t wc
Integration and deformation limits according to Figs. 17 to 19.
The deformations in the compression, Eq. (B.22), and in the tension, Eq. (B.24), zones are determined from the deformation of the integration
paths P1 and P2. It is important to refer the following correspondence between the directions of the global axes 1 = x, 2 = y and 3 = z.
References [6] K. Weynand, J.P. Jaspart, M. Steenhuis, The stiffness model of revised Annex J of
Eurocode 3, in: R. Bjorhovde, A. Colson, R. Zandonini (Eds.), Connections in steel
structures III, Proceedings of the 3rd International Workshop on Connections.
[1] CEN, EN 1993-1-8. Eurocode 3: Design of Steel Structures — Part 1–8: Design of Trento, Italy 1995, pp. 441–452.
Joints, European Committee for Standardization, Brussels, May 2005. [7] F. Gentili, R. Costa, L. Simões da Silva, Definition of a simplified model for connec-
[2] H. Gervásio, L. Simões da Silva, L. Borges, Reliability assessment of the post- tions in steel structures, Proceedings of Congresso Nacional de Mecânica Experi-
limit stiffness and ductility of steel joints, J. Constr. Steel Res. 60 (4) (2004) mental, 15−17 October, Aveiro, Portugal, 2014.
635–648. [8] J.P. Jaspart, Étude de la Semi-Rigidité des Noeuds Poutre-Colonne et son Influence
[3] ABAQUS User's, Theory and Scripting Manuals (2012), Version 6.12. ABAQUS, Inc., sur la Resistence et la Stabilité des Ossatures en Acier(PhD Thesis) Faculty of Applied
Dassault Systèmes Simulia Corp., Providence, USA, 2012. Sciences, University of Liège, Belgium, 1991 (in French).
[4] L. Simões da Silva, Towards a consistent design approach for steel joints under gen- [9] H. Krawinkler, V.V. Bertero, E.P. Popov, Shear behaviour of steel frame joints, J.
eralized loading, J. Constr. Steel Res. 64 (2008) 1059–1075. Struct. Division ASCE 101 (ST11) (1975) 2317–2336.
[5] C. Faella, V. Piluso, G. Rizzano, Structural Steel Semirigid Connections. Theory, De- [10] P. Zoetemeijer, The influence of normal, bending and shear stresses on the ul-
sign and Software, CRC Press LLC, 2000. timate compression force exerted laterally to European rolled sections, The
Netherlands, Stevin laboratory: Steel Structures, Report 6-75-18, Delft Univer- [20] G. Shi, Y. Shi, Y. Wang, M.A. Bradford, Numerical simulation of steel pretensioned
sity of Technology Department of Civil Engineering, February 1975. bolted end-plate connections of different types and details, Eng. Struct. 30 (2008)
[11] J.P. Jaspart, Shear and load-introduction deformability and strength of column web 2677–2686.
panels in strong axis beam-to-column joints, Internal Report MSM, n° 202, Univer- [21] P. Nogueiro, Cyclic Behaviour of Steel Connections(PhD Thesis) Civil Engineering
sity of Liége, April 1990. Department, University of Coimbra, Portugal, 2009 (in Portuguese).
[12] J.M. Castro, A.Y. Elghazouli, B.A. Izzuddin, Modelling of the panel zone in steel and [22] Y.I. Maggi, R.M. Gonçalves, L. Calado, Numerical and experimental behavior of
composite moment frames, J. Eng. Struct. 27 (2005) 129–144. beam-to-column extended end plate connections, in: F.M. Mazzolani (Ed.), Behav-
[13] S. Jordão, L. Simões da Silva, R. Simões, Behaviour of welded beam-to-column iour of Steel Structures in Seismic Areas, STESSA 2003, Lisse 2003, pp. 343–349.
joints with beams of unequal depth, J. Constr. Steel Res. 91 (2013) 42–59. [23] O.S. Bursi, J.P. Jaspart, Calibration of a finite element model for isolated bolted end-
[14] S. Ádány, L. Dunai, Finite element simulation of the cyclic behaviour of end-plate plate steel connections, J. Constr. Steel Res. 44 (3) (1997) 225–262.
joints, Comput. Struct. 82 (2004) 2131–2143. [24] D.P. Flanagan, A. Belytschko, A uniform strain hexahedron and quadrilateral with or-
[15] N. Krishnamurthy, D.E. Graddy, Correlation between 2- and 3-dimensional finite el- thogonal hourglass control, Int. J. Numer. Methods Eng. 17 (1981) 679–706.
ement analysis of steel bolted end-plate connections, Comput. Struct. 6 (4–5) [25] J.L. Chaboche, Time-independent constitutive theories for cyclic plasticity, Int. J.
(1976) 381–389. Plast. 2 (2) (1986) 149–188.
[16] A.R. Kukreti, J.M. Murray, A. Abolmaali, End-plate connections moment-rotation re- [26] ECCS, Recommended testing procedure for assessing the behaviour of structural steel
lationship, J. Constr. Steel Res. 8 (1987) 137–157. elements under cyclic loads, Technical Committee 1, TWG 1.3 — Seismic Design, Publ.
[17] O.S. Bursi, J.P. Jaspart, Benchmarks for finite element modelling of bolted steel con- n.° 45, European Convention for Constructional Steelwork, Brussels, Belgium, 1986.
nections, J. Constr. Steel Res. 43 (1–3) (1997) 17–42. [27] CEN, EN 1993-1-1. Eurocode 3: Design of Steel Structures — Part 1–1: General Rules
[18] A.N. Sherbourne, M.R. Bahaari, 3D simulation of end-plate bolted connections, and Rules for Buildings, European Committee for Standardization, Brussels, May 2005.
J. Struct. Eng. ASCE 120 (ST11) (1994) 3122–3136.
[19] Y.I. Maggi, R.M. Gonçalves, R.T. Leon, L.F.L. Ribeiro, Parametric analysis of steel bolted
end plate connections using finite element modeling, J. Constr. Steel Res. 61 (2005)
689–708.

Journal of Constructional Steel Research: Hugo Augusto, Luís Simões Da Silva, Carlos Rebelo, José Miguel Castro

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Constructional Steel Research: Hugo Augusto, Luís Simões Da Silva, Carlos Rebelo, José Miguel Castro

Uploaded by

Copyright:

Available Formats

Journal of Constructional Steel Research 116 (2016) 271–293

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Characterization of web panel components in double-extended bolted

1. Introduction characterization is possible, in the second case the required number of

web panel in shear and the column web in transverse compression or

Beam-to-column joints modelling must deal with a variety of phe-

Fig. 2. Detail of the joint for Groups 1 and 2.

Fig. 3. Detail of the joint for Group 3.

Fig. 4. Detail of the joint for Group 4.

4.2. Finite elements options

Fig. 6. Strain gauges disposition.

Bolt Young's modulus Ultimate stress Ultimate

M24 (10.9) 213 1170 0.030

Model Mmax (kNm) Sji (kNm/rad) Rot.max (mrad) Failure mode

J1.1 419 73,500 +70 Cracking EP

Bending Moment (kNm)

100 Experimental curve - J1.1

(a) Moment rotation curves. (b) Test J-1-1 failure mode.

Fig. 8. Moment–rotation curves and failure modes.

Fig. 9. Meshed parts of the FE model.

Bending Moment (kNm)

Fig. 10. Sensitivity study on the ABAQUS solid elements.

J1.1 J3.1 J4.1 Geometric properties

Fig. 12. FE models used in the validation.

Bending Moment (kNm)

Bending Moment (kNm)

200 J4.1 - Joint rot. - Exp. Mel A

Fig. 13. Global experimental, numerical and analytical results.

Test Global prop. Component prop.

Shear Compr. Tension comp.

Bold data indicates signiﬁcance the driving components.

Experimental values Numerical values Analytical values

Test End plate in bending

Kij Me Mpl Kij M1st M1st + 2nd M1st + 2nd + 3rd

J1.1 155,000 233 350 133,416 210.4 354.5 354.5

Bold data indicates signiﬁcance the driving components.

Bending Moment (kNm)

Mfc ¼ zðσ 22 dAÞ: ð3Þ 500

MfcðP4Þ þ M fcðP5Þ þ M fcðP6Þ þ MfcðP7Þ 400

Test Distortion of the column web panel

Experimental Krawinkler model Atamaz-Jaspart model

Kij Me Mpl Kij Me Mpl Kij Me Mpl

Bending Moment (kNm)

200 Exp. column web panel rot. (J4.1) 200

400 h t(2nd) 500

Joints based on J1.1 ts (mm) tp = 18 mm tp = 40 mm

Model Global prop. Component prop.

Shear comp. Compr. comp. Tension comp.

Sj,ini Mpl,Rd Vwp,Rd Fc,wc,Rd Fc,fb,Rd 1st row 2nd row

Bold data indicates signiﬁcance the driving components.

400 FE (tp18) FE (tp40) 400

Bending Moment (kNm)

500 491 500

300 336 300

200 C1_1ppv1_tp18_ts0 200

Fig. 27. Evolution of the bending moments in the column ﬂanges.

Annex A. Design models for column web panel components

A.1. Geometrical deﬁnitions and global joint properties

Fig. A.1. End plate joint geometrical properties.

Stiffness (S) Resistance (M)

Sj ¼ Ez2 M j;Rd ¼ ∑ hr F tr;Rd If Mj,Ed ≤ 2/3Mj,Rd ⇒ μ = 1

A.2. Krawinkler model