Determination of the main characteristics of

semi-rigid beam-to-column connections

through numerical and experimental method
K. Vértes & M. Iványi
Budapest University of Technology and Economics,
Department of Structural Engineering

Abstract
Connections play an essential role in steel structures which is why it is necessary
to know the behaviour of these structural elements. This study provides a FEM
model for the determination of the main characteristics of a semi-rigid joint.
Although this joint form is very economical, it still not widely used in practice.
The main reason for this is that its characteristics are hard to determine. In this
study an overview is provided about the handling of such connections by
introducing Eurocode 3 [1] and some numerical and experimental methods and
results.

1 Introduction
The main characteristics of a joint are its strength, stiffness and rotation capacity.
The most often appearing problems of using semi-rigid connections hide firstly
in the modelling of the connection, then in the determination of their stiffness
and at last in the handling of the softening part of the moment-rotation curve [2].
Three ways are known for the determination of the joint characteristics.
These are the following:
• Experimental method – accurate but expensive and inconvenient
• FE analysis – its accuracy depends on the model
• Analytical method – EC 3 offering
This study discusses the finite element modelling of steel semi-rigid
connections by demonstrating the finite element model of welded and bolted
beam-to-column connections.

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5
678 High Performance Structures and Materials II

2 Basic information on steel beam-to column connections

2.1 Joint classification

There are several classification systems. Here we just mention the one according
to Eurocode 3.
Based on overall studies, EC3 has introduced its classification system. The
new idea of this classification is to consider the joint as a part of a structure. It
means that not only the joint-type influences the structure behaviour, but the
structure-type (braced, or unbraced) has influence on the joint behaviour as well.
According to this phenomenon the following joint classes can be distinguished:
classification by rigidity for elastic analysis (nominally pinned connections; rigid
connections; semi-rigid connections) and classification by strength for plastic
analysis (nominally pinned connections; full-strength connections; partial
strength connections) [4].

2.2 Main parameters influencing a joint’s characteristics

Joints usually have quite a complex behaviour. It is because a lot of parameters

influence it. The flexibility sources of a joint have a major effect on the joint
characteristics. These flexibility sources of a joint are the following:
• Connection flexibility, which consists of the deformation of its
components (plates, cleats, T-stubs, etc.), the slip between its components (if
bolted) and its fastener deformation (bolt and rivet elongation).
• Localised deformation of the column (plastification, flanges
deformation, web buckling, shear deformation of the panel zone and
elongation and shortening in the tension and compression regions of the
panel zone).
• Localised deformation of the beam. (Less important than the two
previous ones).
The parameters that can affect these flexibility sources and consequently
influence the joint behaviour can be the connection types (web cleats, flange
cleats end plates etc.) and details and the environment of the connection (the
column axis to that the connection is fastened, the connection location, the type
of the fasteners, the joint loading type, the material) [3].

2.3 The moment-rotation curve of a joint

The moment-rotation curve describes the behaviour of a joint. It is a non-linear

curve. However this non-linearity is hard-to-handle in everyday practice. In most
cases simplifications of the curve can be done without causing relevant
inaccuracy. Such simplifications can be bi-linear law multi-linear low or the
non-linear low can be used for an elastic-plastic calculation too.
According to Eurocode 3, the main parameters of a bended joint are the
moment resistance and the initial elastic stiffness of the joint. These two
parameters provide the whole moment-rotation curve (fig. 1)

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5
 High Performance Structures and Materials II 679

Figure 1: Non-linear M-Φ curve according to EC3 Annex.

If the rotation capacity (ΦCd) is not limited, this curve has three parts. Up to
66% of the moment resistance (MRd), the curve is linear, and the stiffness of this
part is the so-called initial stiffness. From 2/3 Mj,Rd up to MRd the curve is
non-linear and after the moment have reached the yielding limit, only the
rotation increases. The non-linear curve between 2/3 Mj,Rd and Mj,Rd:
S j ,ini
Sj = Ψ (1)
 1.5M Sd 
 
 M Rd 
where Ψ=2,7 for end-plate and welded joints and 3,1 for flange cleat joints.

2.4 Eurocode 3’s component method [4]

For determining a joint’s characteristics, EC3 provides the component method.

Its main idea is to handle a joint as a set of individual basic components, where
each of these possesses its own strength. During the calculation process the
building components of the joint have to be identified, then the stiffness and
resistance of the basic components have to be calculated and finally these
components have to be put together in a statically appropriate way. In case of a
general bolted end-plate connection in bending, the following components can be
distinguished:
Compression zone:
• Column web in compression
• Beam flange and web in compression
Tension zone:
• Column web in tension
• Column flange in bending
• Bolts in tension
• End-plate in bending
• Beam web in tension

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5
680 High Performance Structures and Materials II

Shear zone:
• Column web panel in shear

2.5 Numerical modelling of joints

Nowadays numerical models are developing because of the development of the

computer capabilities. Some often-used models will be introduced here:
• Spring element, which incorporates the joint stiffness as the element
rigidity. It consists of a rotational spring an sometimes it also contains
translational springs incorporating axial and shear semi-rigidity
• Equivalent beam elements with a reduced stiffness assuring the M-Φ
characteristics to be respected
• Truss model, which consists of a particular arrangement of truss
elements representing the several kinds of joint flexibility [3].

3 Results of numerical and analytical calculations compared

to experiments
In the following section the FEM (Ansys 5.7) [5] model of two different
connection types, (a welded and an end-plate bolted connection) will be shown
and compared to analytical (EC3) solutions and experimental results. The stress
distributions and stress levels and the moment-rotation curves have been
determined for both cases by the FEM. In order to compare the results between
the Eurocode’s component method and our results, the moment-rotation curves
have been determined by the component method as well.

3.1 Welded joint

The numerical modelling of a welded (continuous) connection is relatively easy

compared to a bolted (non-continuous) one. Here a welded connection built up
by IPE 100 sections has been analyzed. A monotonously increasing vertical
loading was applied at the end of the beam, so the connection was bent. For the
analysis a non-linear inelastic material model was used.
The FE model was built up by 8node non-linear brick elements. The mesh
was refined at the highly stressed areas according to the joint’s basic components
(such as the column web and flange and the beam flanges and web at the
connection part).
The stress results and the deformed shape are shown in figure 2. It is clearly
visible that the stress distribution in the connection meets the assumed
components according to EC3 (the tensioned and compression zones appear in
the flanges and the column web in shear is visible too). The moment rotation
curve also agrees with the EC assumption it is linear up to 2/3 MRd and then
non-linear, finally the yield plateau appears (Figure 3).

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5
 High Performance Structures and Materials II 681

Figure 2: Welded joint stress distributions (vertical normal stress left above,
horizontal normal stress right above, shear stress underneath).

30
25
moment (kNm)

20 FE result
15 EC3
10 Experiment
5
0
0 0.02 0.04 0.06 0.08
rotation

Figure 3: Moment-rotation curves.

3.2 Bolted end-plate joint

Bolted connections are not continuous so this discontinuity has to be taken into
account when modelling such joints. This makes their modelling more
complicated. At present example two IPE 100 sections were connected to each
other with four bolts of M12. The thickness of the end-plate was taken to 10 mm.

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5
682 High Performance Structures and Materials II

No stiffeners were built in into the joint. The FE model was built up by 8node
non-linear brick elements and the mesh was extremely fine at the bolts and also
fine at the assumed components of the connection. Again, the monotonously
increasing vertical load was applied at the beam end. The stress levels at each
load step and the moment-rotation curve has been determined by FEM. Also in
this case the moment-rotation curve has been determined by the Eurocode
offering too.
The main problem of the modelling of a bolted joint is to model that the
bolts play the force transmission role in the connection. First of all the bolts have
to be built up from solid elements and in order to simulate the complex stress
state in it a quite fine mesh is required, not to mention the contact problems.
For the modelling of this connection 8node non linear brick elements were
used. Contact surfaces were defined between the end plate and the column
flange, between the shank of the bolt and the surface of the bolthole and between
the bolthead and the endplate/column flange. The material model was identical to
the previously mentioned one. The results are shown in figures 4, 5, and 6.

Figure 4: Bolted joint, vertical normalstress distribution.

Figure 5: Bolted joint, shear stress distribution.

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5
 High Performance Structures and Materials II 683

Figure 6: Bolted joint, horizontal normalstress distribution.

3
2,5
moment (kNm)

2 FE result
1,5 EC3
1 Experiment
0,5
0
0 0,05 0,1 0,15
rotation

Figure 7: Moment-rotation curve for bolted joint.

The stress distributions here also show the assumed components of this
connection. The column web panel in shear, the column web in tension, the bolts
in tension, the end-plate and column flange, the column web in compression, the
beam flange and web in compression appear in full view. Also the
moment-rotation curve is near to the EC3 curve.

4 Frames with semi-rigid connections

Investigations were carried out on multi-storey, multi-bay steel frames in order to
study the effect of semi-rigidity on the load bearing capacity and ultimate
behaviour of these structures.
After the numerical analysis the results showed that the presence of
semi-rigid connections significantly influences the load bearing capacity and
deformations of frames. However, this influence was not clearly proportional

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5
684 High Performance Structures and Materials II

having frames of different numbers of bays and storeys. Primary the sequence of
occurrence of plastic hinges governed the magnitude of ultimate
displacements [6].

5 Conclusions
The results of the FE calculation were close to the ones calculated according to
EC3 offering and the experimental results. In both cases the connection
behaviour was semi-rigid.
For the future spatial semi-rigid connections will be modelled and analyzed.
Then a parametrical study will follow, by which the characteristics of these
spatial semi-rigid joints can be determined.

References
[1] Eurocode 3. Design of steel structures
[2] K. Vértes Overall behaviour of steel frames with semi-rigid connections,
[Link] Int. Ph.D. Symposium in Civil Engineering, ed. Peter Schiessl,
München, 2002, pp. Vol2., 391
[3] Wajd Atamaz Sibai (1991), Semi-rigid joint modelling for non-linear
analysis of flexibly connected frames, These No. 967 Lausanne, EPFL, 1991
[4] Jaspart J.P. (1999) Semi-Rigidity in Connections of Structural Steelworks:
Theory, Analysis and Design: Characterisation and Idealisation of Moment
Resisting Joints, Udine, 1999.
[5] Ansys Release 5.7 SAS IP Inc., 1997.
[6] Miklós Iványi (1995) PECO COST C1 Technical Report, Column Base
Strength and Rigidity Modelling, Budapest, 1995.

High Performance Structures and Materials II, C.A. Brebbia & W.P. De Wilde (Editors)
© 2004 WIT Press, [Link], ISBN 1-85312-717-5