Journal of Constructional Steel Research 55 (2000) 69–89

www.elsevier.com/locate/jcsr

General report: session on connections

*
J.P. Jaspart
Department MSM, University of Liège, Quai Banning 6, B-4000 Liège, Belgium

Abstract

As a result of a world-wide intensive research activity, a new design concept for structural
moment resistant joints has been suggested and implemented in design codes. In the present
paper, its background is first briefly described and the main contributions to its development
are presented. As a matter of fact, because of the high number of past and ongoing researches,
reports and papers devoted to this topic, an exhaustive list of all these works could not be
established in a limited number of pages.
The possibilities of extension of the new design concepts, which have been first developed
for steel beam-to-column joints under static loading, are particularly pointed out. This aspect
appears quite important as in the near future it should provide designers with a unified design
approach for structural joints whatever their loading, their configurations and the nature of
their constitutive material(s).  2000 Elsevier Science Ltd. All rights reserved.

Keywords: Structural joints; Building frames; Economy; Characterisation; Modelling; Classification;

Idealisation; Experimentation; Fitting; Mechanical model; Analytical model; Numerical model; Practical
design tool; Static loading; Earthquake; Fire

1. Introduction

In the last fifteen years, much research work has been devoted world-wide to
structural moment resistant joints and their influence on the resistance and the sta-
bility of building frames. All these researches have progressively led to a deep under-
standing of the joint behaviour, to the development of design tools for the evaluation
of the mechanical properties of the joints and to guidelines and recommendations
for the design of building frames including the joint behaviour. The motivation of
these works is of a scientific nature, but it is also clearly linked to the economical

* Tel.: +32-4-3669247; fax: +32-4-3669192.

E-mail address: [email protected] (J.P. Jaspart).

0143-974X/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 9 9 ) 0 0 0 7 8 - 4
70 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

benefit which results from a better consideration of the importance of the joint design
as far as fabrication and erection costs are concerned.
The new concept for joint design has been first developed for beam-to-column
steel joint configurations between H or I profiles and subject to static bending and
shear forces. More recent works have allowed to extend the concept to composite
joints but also to other types of configurations such as beam-to-beam ones, beam
splices and column bases. Further actions are now in progress to apply this new
design approach to other types of profiles (tubular sections, built-up sections, …),
to other loading conditions (fire, earthquakes, …) and to other materials (timber,
pre-cast concrete, …).
In this article, the new design concept is described and the significant scientific
contributions to its development are commented on.

2. A consistent approach for structural joints

The rotational behaviour of actual joints is well recognised as being often inter-
mediate between the two extreme situations, i.e. rigid or pinned.
Consider now the bending moments and the related rotations at a joint (Fig. 1):
When all the different parts in the joint are sufficiently stiff (i.e. ideally infinitely
stiff), the joint is rigid, and there is no difference between the respective rotations
at the end of the members connected at this joint (Fig. 1a). The joint experiences a
single global rigid-body rotation which is the nodal rotation in the commonly used
analysis methods for framed structures.
Should the joint be without any stiffness, then the beam will behave just as simply
supported whatever the behaviour of the other connected member(s) (Fig. 1b). This
is a pinned joint.
For intermediate cases (non zero and non infinite stiffness), the transmitted
moment will result in a difference f between the absolute rotations of the two con-
nected members (Fig. 1c). The joint is semi-rigid in these cases.

Fig. 1. Classification of joints according to rotational stiffness.

 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 71

The simplest means for representing the concept is a rotational (spiral) spring
between the ends of the two connected members. The rotational stiffness S of this
spring is the parameter that links the transmitted moment Mj to the relative rotation
f. It is defined as the slope of the so-called moment-rotation curve Mj⫺f which
depends on the joint properties.
At the global analysis stage, the effect of having semi-rigid joints instead of rigid
or pinned joints is to modify not only the displacements, but also the distribution
and magnitude of the internal forces throughout the structure.
As an example, the bending moment diagrams in a fixed-base simple portal frame
subjected to a uniformly distributed load are given in Fig. 2 for two situations, where
the beam-to-column joints are respectively either pinned or semi-rigid. The same
kind of consideration holds for deflections.
The concept of semi-rigid joints is introduced in some different codes, and in
particular in Eurocode 3 Part 1.1 [1] for the design of steel structures under static
loading.

3. The merits of the consistent approach for structural joints

Both the Eurocode 3 requirements and the desire to model the behaviour of the
structure in a more realistic way leads to the consideration of the semi-rigid beha-
viour when necessary.
Many designers would stop at that basic interpretation of the code and hence
would be reluctant to confront the implied additional computational effort involved.
Obviously a crude way to deal with this new burden will be for them to design
joints that will actually continue to be classified as being either pinned or fully rigid.

Fig. 2. Elastic distribution of bending moments in a simple portal frame.

72 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

However such properties will have to be proven at the end of the design process
and, in addition, such joints will certainly be found to be uneconomical in a number
of situations.
It should be noted that the concept of rigid and pinned joints still exists in Euro-
code 3. It is accepted that a joint which is almost rigid, or almost pinned, may still
be considered as being truly rigid or truly pinned in the design process. How to
judge whether a joint can be considered as rigid, semi-rigid or pinned depends on
the comparison between the joint stiffness and the frame stiffness.
The designer is strongly encouraged to go beyond this “all or nothing” attitude.
Actually it is important to consider the benefits to be gained from the semi-rigid
behaviour of joints. Those benefits can be brought in two ways:

1. The designer decides to continue with the practice of assuming — sometimes

erroneously — that joints are either pinned or fully rigid. However, proper con-
sideration has to be given to the influence that the actual behaviour of the joints
has on the global behaviour of the structure, i.e. on the precision with which the
distribution of forces and moments and the displacements have been determined.
This may not prove to be easy when the joints are designed at a late stage in the
design process since some iterations between global analysis and design checking
may be required. Nevertheless, the following situations can be foreseen:
앫 So that a joint can be assumed to be rigid, it is common practice to introduce
web stiffeners in the column. Eurocode 3 now provides the means to check
whether such stiffeners are really necessary for the joint to be both rigid and
have sufficient resistance. There are practical cases where they are not needed,
thus permitting the adoption of a more economical joint design.
앫 When joints assumed to be pinned are later found to have fairly significant
stiffness (i.e. to be semi-rigid), the designer may be in a position to reduce
beam sizes. This is simply because the moments carried by the joints reduce
the span moments in the beams.
2. The designer decides to give consideration, at the preliminary design stage, not
only to the properties of the members but also to those of the joints. It may be
shown [2] that this new approach is not at all incompatible with the sometimes
customary separation of the design tasks between those who have the responsi-
bility for conceiving the structure and carrying out the global analysis and those
who have the responsibility for designing the joints. Indeed, both tasks are very
often performed by different people, or indeed, by different companies, depending
on national or local industrial habits. Adopting this novel early consideration of
joints in the design process requires a good understanding of the balance between,
on the one hand, the costs and the complexity of joints and, on the other hand,
the optimisation of the structural behaviour and performance through the more
accurate consideration of joint behaviour for the design as a whole. Two examples
are given to illustrate this:
앫 It was mentioned previously that it is possible in some situations to eliminate
column web stiffeners and therefore to reduce costs. Despite the reduction in
its stiffness and, possibly, in its strength, the joint can still be considered to
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 73

be rigid and be found to have sufficient strength. This is shown to be possible

for industrial portal frames with rafter-to-column haunch joints in particular,
but other cases can be envisaged.
앫 In a more general way, it is worthwhile to consider the effect of adjusting the
joint stiffness so as to strike the best balance between the cost of the joints
and the cost of the beams and the columns. For instance, for braced frames,
the use of semi-rigid joints, which are not necessarily more costly than the
pinned joints, leads to reducing the beam sizes. For unbraced frames, the use
of less costly semi-rigid joints, instead of the rigid joints, may lead to increased
beam sizes and possibly column sizes.

Of course the task may seem a difficult one, and this is why a design handbook
devoted to this new concept has recently been published [3]. The whole philosophy
could be termed as “Because you must do it, take advantage of it”. The designer
has therefore the choice between a traditionalist attitude, where something may often
be gained, and an innovative attitude, where the most economical result [4] may
best be sought.
It is important to stress the high level of similarity that exists between the member
and joint responses. This topic is addressed in the next section.

4. A parallel between member sections and joints

Member cross-section behaviour may be considered through an M-f curve for a

simply supported beam loaded at mid-span (M: bending moment at mid-span; f:
sum of rotations at the span ends). Joint behaviour will be considered through a
similar relationship, but with M = Mj being the bending moment transmitted by
the joint and f being the relative rotation between the connected members. Those
relationships have similar shapes as illustrated in Fig. 3. To flexural stiffness EI/L
and the design resistance Mb.Rd of the member correspond the initial stiffness Sj,ini
and the design resistance Mj.Rd of the joint.
According to Eurocode 3 member cross-sections are divided into four classes
according to their varying ability to resist local instability, when partially or totally
subject to compression, and the consequences this may have on the possibility for
plastic redistribution. Therefore their resistance ranges from the full plastic resistance
(class 1 and 2) to the elastic resistance (class 3) or a reduced elastic resistance
(class 4).
The allocation of a cross-section to a specific class is governed by the assump-
tions on:

앫 The behaviour to be idealised for global analysis (i.e. class 1 will allow the forma-
tion of a plastic hinge and permit the redistribution of internal forces in the frame
as loads are increased up to or beyond the design loads);
앫 The behaviour to be taken into account for local design checks (i.e. class 4 will
74 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

Fig. 3. M-f characteristics for member cross-section and joint.

imply that the resistance of the cross-section is based on the properties of a rel-
evant effective cross-section rather than of the gross cross-section).

In Eurocode 3, the classification of a cross-section is based on the width-to-thick-

ness ratio of the component walls of the section. Ductility is directly related to the
amount of rotation during which the design bending resistance will be sustained. For
joints, the rotation capacity concept is equivalent to the ductility concept for sections.
In a manner similar to that for member cross-sections, joints are classified in terms
of ductility or rotation capacity. This classification is a measure of their ability to
resist premature local instability and, even more likely, premature brittle failure
(especially due to bolt failure) with due consequences on the type of global analysis
allowed. The practical interest of such a classification for joints is to check whether
an elasto-plastic global analysis may be conducted up to the formation of a plastic
collapse mechanism in the structure, which implies such hinges in at least some of
the joints.
Joints may therefore be classified according to both their stiffness and their duc-
tility. Moreover, joints may be classified according to their strength.
In terms of their strength, joints are classified as full-strength or partial-strength
according to their resistance compared to the resistance of the connected members.
For elastic design, the use of partial-strength joints is well understood. When plastic
design is used, the main use of this classification is to foresee the possible need to
allow a plastic hinge to form in the joint during the global analysis. In order to
permit a further increase of loads beyond that corresponding to the formation of the
hinge, a partial-strength joint may be required to act as a hinge from the moment
when its plastic bending resistance is reached. In that case, the joint must also have
sufficient ductility.
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 75

A parallel between member sections and partial-strength joints may also be ident-
ified in the case of cyclic loading under seismic action.

5. The concept of joint representation

For many years the research activity in the field of joints mainly concentrated on
two aspects:

앫 the evaluation of the mechanical properties of the joints in terms of rotational

stiffness, moment resistance and rotation capacity;
앫 the analysis and design procedures for frames including joint behaviour.

However, progressively, it has been understood that there were intermediate steps
to consider in order to integrate in a consistent way the actual joint response into
the frame analysis; this is known as the joint representation.
The joint representation includes four successive steps respectively named:

앫 the joint characterisation: i.e. the evaluation through appropriate means of the
stiffness, resistance and ductility properties of the joints (full M-j curves or
key values);
앫 the joint modelling: i.e. the way on how the joint is physically represented in
view of the frame analysis;
앫 the joint classification: i.e. the tool providing boundary conditions for the use of
conventional types of joint modelling (e.g. rigid or pinned);
앫 the joint idealisation: i.e. the derivation of a simplified moment-rotation curve so
as to fit with specific analysis approaches (e.g. linear idealisation for an elastic
analysis).

The state of development and knowledge in these four domains is discussed in

the following sections.

6. Joint characterisation

6.1. Generals

This paragraph gives an overview of the existing characterisation procedures by

particularly highlighting their advantages and drawbacks.
The more accurate, but also expensive, way to characterise the deformability and
the resistance of joints is the experimentation in laboratory. The use of this tech-
nique — which requires much money and time — is basically limited to research
activities and can consequently not be recommended for daily practice.
The existence of numerous test results for a large variety of joint configurations
and connection types led progressively some researchers to develop computerised
76 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

databanks [5,6]. The low probability for the designer to find information for the
specific joint he is studying and the risk to misinterpretate the results listed in the
databank — no standardised procedure for the testing of joint exists at present —
limit considerably the practical interest of these tools. On the other hand, it now
appears clearly that the databanks have to be considered as quite valuable tools for
the validation of mathematical models aimed at predicting the joint response on the
basis of the geometrical and mechanical joint properties.
Most of the well-known mathematical models available in the literature are
described by Nethercot and Zandonini [7]. They can be classified in four main categ-
ories:

앫 curve fitting;
앫 simplified analytical models;
앫 mechanical models;
앫 finite element analysis.

These categories are briefly presented hereafter.

6.2. Curve fitting

Definition:

앫 Attempt to fit a mathematical representation to characteristic M-f curves obtained

by means of:
experimental tests in laboratory;
numerical simulations.
앫 Attempt to link the coefficients of the mathematical representation with physical
parameters of the joint.

General advantages and/or disadvantages:

앫 Capacity of representing with extreme accuracy any shape of M-f curve;

앫 Purely empirical → range of application limited to joints, the geometrical and
mechanical properties which are similar to those considered when calibrating
the formula.
앫 Inability to recognize that, depending on the relative actual values of the geometri-
cal and mechanical parameters, the type of connection behaviour as well as the
contribution of each component to the overall joint response may change signifi-
cantly.

6.3. Simplified analytical models

Definition:

앫 Simplified analytical methods to predict the main characteristic values of the M-

 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 77

f curves (initial stiffness, design moment resistance,...); this step requires the
knowledge of the mechanical and geometrical properties of the joints.
앫 Verification of these analytical methods by comparison with test data or results
of numerical simulations.
앫 Description of the M-f behaviour by curve fitting using the calculated initial stiff-
ness, plastic and/or ultimate moment,... in suitable mathematical expressions.

General advantages and/or disadvantages:

앫 Allow approximation of the form of the M-f curves without resorting to testing;
앫 Still require empirical curve fitting to generate the full curve (but limited!);
앫 Special remark: Eurocode 3 Revised Annex J [8] method refers basically to this
section; it provides an original way to generate the M-f curves.

6.4. Mechanical models

Definition:

앫 Set of rigid and deformable elements each representing the behaviour of specific
parts of the joint.
앫 Non-linearity of the joint response is then accounted for by inelastic constitutive
laws adopted for the deformable elements.
앫 These constitutive laws are obtained from test data, numerical simulations or ana-
lytical models.

General advantages and/or disadvantages:

앫 Really suitable for modelling provided that a knowledge of the load deformation
curve of the key components is available.
앫 May be easily extended to different types of joint configurations and of connec-
tions provided the knowledge of the key components is available.
앫 Often require the use of computer programs to generate the M-f curves.

6.5. Finite element analysis

Definition:

앫 Prediction of M-f curves by means of a finite element analysis

General advantages and/or disadvantages:

앫 Suitable to predict the response of welded joints.

78 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

앫 Sufficient ability to model the non-linear 3-D response of joints with bolted con-
nections and in particular:
the actual bolt action;
the contact phenomena;
the slips;
not always fully attained.

An updated list, with references, of existing models for each category may be found
in [9].
From this description, it may be concluded that:

앫 The finite element analysis is not yet likely to be used to predict the semi-rigid
response for any type of connection detailing and its use is reserved to research
activities, even if significant progress has been achieved in the last years [10].
앫 The simplified analytical models and the mechanical models are characterised by
a wider field of application than curve fitting; this results from the theoretical
background of these two kinds of models. However, they have to be extended to
other types of connections in order to cover the main needs of the designers.
앫 The simplified analytical models are the only ones — with the formulae resulting
from curve fitting — suitable for hand calculations.

6.6. The component method

Plenty of analytical models are available in the literature for different types of
joint configurations and connection types. But progressively one of these models,
because of the advantages it offers in comparison to the others, slowly became the
reference and is now considered as such by most of the researchers. In particular it
has been followed in Eurocode 3 Revised Annex J [8]. It is known as the compo-
nent method.
Roughly speaking the component method may be presented as the application of
the well-known finite element method to the calculation of structural joints.
In the characterisation procedures, a joint is generally considered as a whole and
is studied accordingly; the originality of the component method is to consider any
joint as a set of “individual basic components”. In the particular case of Fig. 4 (joint
with an extended end-plate connection subject to bending), the relevant components
are the following:

앫 compression zone:
column web in compression;
beam flange and web in compression;
앫 tension zone:
column web in tension;
column flange in bending;
bolts in tension;
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 79

Fig. 4. Joint with end-plate in bending.

end-plate in bending;
beam web in tension;
앫 in shear zone:
column web panel in shear.

Each of these basic components possesses its own level of strength and stiffness
in tension, compression or shear. The coexistence of several components within the
same joint element — for instance, the column web which is simultaneously sub-
jected to compression (or tension) and shear — can obviously lead to stress interac-
tions that are likely to decrease the strength and the stiffness of each individual basic
component [11]; this interaction affects the shape of the deformability curve of the
related components but does not call the principles of the component method into
question again.
The application of the component method requires the following steps:

1. identification of the active components for the studied joint;

2. evaluation of the mechanical characteristics of each individual basic component
(specific characteristics — initial stiffness, design strength,... - or the whole
deformability curve);
3. “assembly” of the components in view of the evaluation of the mechanical charac-
teristics of the whole joint (specific characteristics — initial stiffness, design resist-
ance,.... — or the whole deformability M-j curve).

These three steps are schematically illustrated in Fig. 5 in the particular and simple
case of a beam-to-column steel joint with a welded connection.
As specified above, the parallelism with the finite element method is obvious.
To “component” and “joint” may then be substituted the words “finite element”
and “structure”.
The assembly is based on a distribution of the internal forces within the joint. As
a matter of fact, the external loads applied to the joint distribute, at each loading
step, between the individual components according to the instantaneous stiffness and
80 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

Fig. 5. Application of the component method to a welded steel joint (simplified bi-linear component
and joint deformability curves).

resistance of each component. Distributions of internal forces may be obtained

through different ways as discussed in [9].
The application of the component method requires a sufficient knowledge of the
behaviour of the basic components. To review all the publications devoted to this
subject would be quite long and therefore references are made here to recent sum-
mary reports, in particular those prepared within the COST C1 European Action on
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 81

“Control of the Semi-Rigid Behaviour of Civil Engineering Connections” (1992–

1996).
The components active in the traditional steel joints have been deeply studied and
recommendations for their characterisation are given in the Revised Annex J of Euro-
code 3 [8]. The combination of these components allows a wide range of joint con-
figurations to be covered, which should largely be sufficient to satisfy the needs of
practitioners as far as beam-to-column joints and beam splices in bending are con-
cerned. The application is however limited to joints between H or I hot-rolled profiles
or built-up ones with similar dimensions.
Some new fields of application have recently been investigated:

앫 Weak axis joints where the beam is connected to the web of an H or I column
profile are characterised by an out-of-plane deformability of the column web under
the tension and compression forces carried over by the beam. For this “column
web in transverse compression and tension” component, stiffness and strength
characterisation tools are now available [12]. Knowledge about this component
allows the extension of the component method to joints where tubular columns
with rectangular hollow sections are used, as shown by Vandegans [13]. As a
matter of fact, similar phenomena occur in the face of the hollow section where
the beam is connected.
앫 For joints subject to bending moment (and shear) and axial compression or tension
forces, a specific assembly procedure has to be considered, that corresponds to a
particular distribution of the internal forces for stiffness and strength calculation,
the component properties remaining unchanged whatever the type of loading. Sig-
nificant progress in this field has been recently achieved [9].
앫 Column bases are subjected to combinations of bending moments and axial forces
and possess specific components for which a knowledge had to be provided.
For instance:
concrete block in compression;
base plates with specific geometry;
anchorages in tension;
contact between soil and foundation;
...
Such information has recently been published by an international working group
[14]. In the report, the validity of the proposed models is shown through compari-
sons with results of experimental tests and numerical simulations.
앫 An improvement of the mechanical properties of the joints may be achieved
through the use of beam haunches, end-plate stiffeners or high strength steels.
These aspects are covered in [9].
앫 In pitch-roof portal frames, connected beams and columns form an angle higher
than 90°. This requires specific amendments to be made to the existing characteris-
ation procedures, see [9].
앫 When columns with rather slender webs are used, the usual rules for “column
web panels in shear” are no more valid because of the apparition of shear buckling
and post-critical phenomena. Ad-hoc recommendations are available in [9].
82 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

In [15,16], details and rules for the extension of the component method to composite
joints are given. The new components to consider are as follows: reinforcement bars
in tension; stiffening and strengthening of the column web panel in shear and of the
column web in compression by means of encased concrete.
Preliminary works also indicate that the component method also seems suitable
for the characterisation of joints subjected to extreme loading conditions such as
earthquakes or fire. Besides that, first attempts have been made in the COST C1
project to apply the component method to joints in pre-cast and timber construction.
So it may reasonably be thought nowadays that a unified characterisation procedure
for all structural joints is now developing and will be the common basis for the
future design codes whatever the material or the combination of materials used. This
is likely to lighten the work of the designers, in particular when composite construc-
tion is of concern and the composite action between the materials is not effective
according to the erection stages.

6.7. Levels of refinement of the component method

The framework of the component method is sufficiently general to allow the use
of various techniques of component characterisation and joint assembly. In particular,
the stiffness and strength characteristics of the components may result from exper-
imentations in laboratory, numerical simulations by means of finite element programs
or analytical models based on theory. Often experimentation and numerical simula-
tions are used as references when developing and validating analytical models. The
latter may be developed with different levels of sophistication:

앫 expressions as those presented in PhD. theses cover the influence of all the para-
meters which significantly affect the component behaviour (strain hardening, bolt
head and nut dimensions, bolt prestressing,...) from the beginning of the loading
to collapse (plasticity, instability, …);
앫 rules such as those which have been introduced in the Revised Annex J of Euroc-
ode 3 [8] are more simple and therefore more suitable for hand calculations;
앫 as an ultimate step in the simplification process, simplified calculation procedures
and design tables for standardised joints or components have been produced
[3,17]; they allow a quick and nevertheless accurate prediction of the main joint
properties. Besides that, design software for joints are also available on the market.

Similar levels of sophistication also exist for the joint assembly.

7. Joint modelling

7.1. Generals

Joint behaviour affects the structural frame response and shall therefore be mod-
elled, just as for beams and columns, for the frame analysis and design. Traditionally,
the following types of joint modelling are considered:
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 83

For rotational stiffness:

앫 rigid
앫 pinned

For resistance:

앫 full-strength
앫 partial-strength
앫 pinned

When the joint rotational stiffness is of concern, the wording rigid means that no
relative rotation occurs between the connected members whatever the applied
moment may be. The wording pinned postulates the existence of a perfect (i.e.
frictionless) hinge between the members. In fact these definitions may be relaxed.
Indeed rather flexible but not fully pinned joints and rather stiff but not fully rigid
joints may be considered as effectively pinned and perfectly rigid, respectively. The
stiffness boundaries allowing one to classify joints as rigid or pinned are discussed
in the next section.
For joint resistance, a full-strength joint is stronger than the weaker of the connec-
ted members, which is in contrast to a partial-strength joint. In the everyday practice,
partial-strength joints are used whenever the joints are designed to transfer the
internal forces but not to resist the full capacity of the connected members. A pinned
joint is considered to transfer only a limited moment. Related classification criteria
are expressed in the next section.
Consideration of rotational stiffness and joint resistance properties leads tradition-
ally to three significant joint models: rigid/full-strength, rigid/partial-strength and
pinned.
However, as far as the joint rotational stiffness is considered, joints designed for
economy may be neither rigid nor pinned but semi-rigid. There are thus new possi-
bilities for joint modelling: semi-rigid/full-strength and semi-rigid/partial-strength.
With a view to simplification, Eurocode 3 accounts for these possibilities by intro-
ducing three joint models (Table 1):

Table 1
Types of joint modelling

Stiffness Resistance

Full-strength Partial-strength Pinned

a
Rigid Continuous Semi-continuous
a
Semi-rigid Semi-continuous Semi-continuous
a a
Pinned Simple

a
Without meaning.
84 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

앫 continuous: covering the rigid/full-strength case only;

앫 semi-continuous: covering the rigid/partial-strength, the semi-rigid/full-strength
and the semi-rigid/partial-strength cases;
앫 simple: covering the pinned case only.

The following meanings are given to these terms:

앫 continuous: the joint ensures a full rotational continuity between the connected
members;
앫 semi-continuous: the joint ensures only a partial rotational continuity between the
connected members;
앫 simple: the joint prevents from any rotational continuity between the connected
members;

The interpretation to be given to these wordings depends on the type of frame

analysis to be performed. In the case of an elastic global frame analysis, only the
stiffness properties of the joint are relevant for the joint modelling. In the case of a
rigid-plastic analysis, the main joint feature is the resistance. In all the other cases,
both the stiffness and resistance properties govern the manner in which the joints
should be modelled. These possibilities are illustrated in Table 2.

7.2. Exact and simplified joint modelling

The difference between the loading of the connection (bending) and that of the
column web (shear) in a beam-to-column joint requires, from a theoretical point of
view, that account be taken separately of both deformability sources when designing
a building frame. However, doing so is only feasible when the frame is analysed by
means of a sophisticated computer program which enables a separate modelling of

Table 2
Joint modelling and frame analysis

Modelling Type of frame analysis

Elastic analysis Rigid-plastic analysis Elastic-perfectly plastic

and elastoplastic
analysis

Continuous Rigid Full-strength Elastic-perfectly plastic

and elastoplastic
analysis
Semi-continuous Semi-rigid Partial-strength Rigid/partial-strength;
semi-rigid/full-strength;
semi-rigid/partial-
strength
Simple Pinned Pinned Pinned
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 85

both deformability sources [18]. However for most of the available software, such
a sophisticated modelling is not possible and a simplified one (Table 3) has to be
adopted by concentrating the sources of deformability into a single rotational spring
located at the intersection of the axes of the connected members.
For the other joint configurations (beam splices, column bases, …), this “separ-
ation problem” is not of concern and the simplified modelling is used as it closely
represents the reality.

Table 3
Simplified modelling for joints

Joint modelling Beam-to-column joints (major axis Beam splices Column bases
bending)

Simple

Semi-continuous

Continuous

8. Joint classification

8.1. Stiffness classification

The stiffness classification into rigid, semi-rigid and pinned joints is performed
by comparing simply the design joint stiffness to two stiffness boundaries (Fig. 6).
For sake of simplicity, the stiffness boundaries are usually derived so as to allow a
direct comparison with the initial joint stiffness, whatever the type of joint idealis-
ation that is used afterwards in the analysis.
Different stiffness boundaries have been suggested by some authors. They mainly
differ by the criteria used as a basis for classification and the level of the sophisti-
cation in their expression. In [19], these proposals are discussed, together with that
included in Eurocode 3 Revised Annex J [8].
86 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

Fig. 6. Stiffness classification boundaries.

8.2. Strength classification

The strength classification simply consists of comparing the joint design moment
resistance to “full-strength” and “pinned” boundaries (Fig. 7). The boundaries
adopted in Eurocode 3 Revised Annex J [8] seem to be well accepted at the inter-
national level and no further discussion on these took place in the last few years.

Fig. 7. Strength classification boundaries.

8.3. Ductility classes

Experience and proper detailing results in so-called pinned joints which exhibit a
sufficient rotation capacity to sustain the rotations imposed on them. For moment
resisting joints the concept of ductility classes is introduced to deal with the question
of rotation capacity.
Few studies have been devoted to the evaluation of the rotation capacity of joints.
This is clearly illustrated in Eurocode 3 Revised Annex J [8] where only a limited
amount of information is given. Criteria should therefore be established to distinguish
between “ductile”, “semi-ductile” and “brittle” joints. Ductile joints are suitable for
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 87

plastic frame analysis while brittle ones do not allow any redistribution of internal
forces. The use of semi-ductile joints in a plastically designed frame can only result
from a preliminary comparison between the available and required rotation
capacities. All this matter is extensively discussed in [19].
Under cyclic loading, the dissipation capacity of the joints is a major parameter
to be investigated before allowing the use of partial-strength joints under seismic
loading.

9. Joint idealisation

The non-linear behaviour of the isolated flexural spring which characterises the
actual joint response does not lend itself towards everyday design practice. However
the moment-rotation characteristic curve may be idealised without significant loss
of accuracy. One of the most simple idealisations possible is the elastic-perfectly
plastic relationship (Fig. 8). This modelling has the advantage of being quite similar
to that used for the modelling of member cross-sections subject to bending.
The moment Mj,Rd that corresponds to the yield plateau is termed the design
moment resistance in Eurocode 3. It may be considered as the pseudo-plastic moment
resistance of the joint. Strain-hardening effects and possible membrane effects are
henceforth neglected, which explains the difference in Fig. 8 between the actual M-
f characteristic and the yield plateau of the idealisation.
The value of the joint constant stiffness Sj.ini/η is discussed [9] and practical values
are given in Eurocode 3 Revised Annex J [8]. This coefficient results from the high
non-linearity of the joint M-j curves in comparison to those of the members.
In fact there are different possible ways to idealise a joint M-f characteristic. The

Fig. 8. Bi-linearisation of moment-rotation curves.

88 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89

choice of one of them is dependent upon the type of frame analysis which is contem-
plated, see Eurocode 3:

앫 Elastic idealisation for an elastic analysis;

앫 Rigid-plastic idealisation for a rigid-plastic analysis;
앫 Non-linear idealisation for an elastic-plastic analysis.

10. Conclusions

Much research progress has been achieved in the last years and practical guidelines
are now available for most of the steel building frames made of I or I profiles and
subjected to static loading. The extension of the design procedures to other types of
profiles and to other loading situations is now in progress and results are expected
to be obtained through further research projects. One of the main aspects which
would focus the attention in the future will be the transfer of these new concepts to
practice through appropriate actions (e.g. seminars) and education efforts.

References

[1] Eurocode 3. Design of Steel Structures. Part 1.1: General Rules and Rules for Buildings, European
Prestandard — ENV 1993-1-1, 1992.
[2] SG/TC10a. Momentverbindingen, Staalbouwkundig Genootschap, Branchecentrum Staaltechnologie,
Rotterdam, The Netherlands, 1999.
[3] Maquoi R, Chabrolin B. Frame design including joint behaviour, ECSC Report 18563. Luxembourg:
Office for Official Publications of the European Communities, 1998.
[4] Weynand K. Sicherheits- und Witschaftlichkeits-untersuchungen zur Anwendung nachgiebiger
Anschlüsse im Stahlbau, Heft 35. Aachen, Germany: Shaker Verlag, 1997.
[5] Kishi N, Chen WF. Steel beam-to-column connections — A review of test data, CE-STR-86-26.
West Lafayette, Indiana, USA: School of Engineering, Purdue University, 1986.
[6] Weynand K, Huter M, Kirby PA, Simoes da Silva LAP, Cruz PJS. Sericon — A databank on semi-
rigid joints. Proceedings of the COST C1 International Conference in Liège, September 17–19, 1998,
pp. 217–227, Report EUR 18854 EN, Office for Official Publications of the European Communities,
Luxembourg, 1999.
[7] Nethercot D, Zandonini R. Methods of prediction of joint behaviour. In: Narayanan R, editor. Struc-
tural Connections — Stability and Strength. Elsevier Applied Science Publisher, 1989:23–62.
[8] Revised Annex J of Eurocode 3. Joints in building frames, European Prestandard ENV 1993-1-
1:1992/A2. Bruxelles, Belgium: CEN, Bruxelles, 1998.
[9] Jaspart JP. Recent advances in the field of steel joints — Column bases and further configurations
for beam-to-column joints and beam splices, Professorship Thesis, Department MSM, University of
Liège, Belgium, 1997.
[10] COST C1.c. Numerical simulation of semi-rigid connections by the finite element method. In: COST
C1 report edited by K. Virdi, European Commission, Bruxelles, Luxembourg, 1999.
[11] Guisse S, Jaspart JP. Influence of structural frame behaviour on joint design. In: Proceedings of the
Workshop on Connections in Steel Structures II held in Trento, Italy, on May 29–31, 1995, pp.
321–330, Oxford, U.K.: Elsevier Science, 1996
[12] Steenhuis M, Jaspart JP, Gomes F, Leino F. Application of the component method to steel joints.
 J.P. Jaspart / Journal of Constructional Steel Research 55 (2000) 69–89 89

In: Proceedings of the COST C1 International Conference in Liège, September 17–19, 1998, pp.
125–143, Report EUR 18854 EN. Luxembourg: Office for Official Publications of the European
Communities, 1999.
[13] Vandegans D. Use of the threaded studs in joints between I-beams and RHS columns. In: Proceedings
of the Istanbul Colloquium on Semi-Rigid Connections held on September 25–27, pp. 53–62. Zürich,
Switzerland: IABSE, 1996.
[14] COST C1.a. Column bases in steel building frames. In: COST C1 report edited by Weynand K.
Bruxelles, Luxembourg: European Commission, 1999.
[15] COST C1. Composite steel-concrete joints in braced frames for buildings. In: COST C1 report edited
by Anderson D. Bruxelles, Luxembourg: European Commission, 1997.
[16] COST C1.b. Composite steel-concrete joints in frames for buildings: design provisions. COST C1
report edited by Anderson D. Bruxelles, Luxembourg: European Commission, 1999.
[17] BSCA and SCI. Joints in Steel Construction — Moment Connections, Publication 207/95. Ascot,
U.K.: The Steel Construction Institute, 1995
[18] Maquoi R, Jaspart JP. Modelling of beam-to-column joints for the design of steel building frames.
In: Proceedings of the first World Conference on Constructional Steel Design held in Acapulco,
Mexico on December 6–9, pp. 196–205. Barking, Essex, UK: Elsevier Applied Science, 1992
[19] COST C1.d. Recent advances in the field of structural steel joints and their representation in the
building frame analysis and design process. In: COST C1 report edited by Jaspart JP. Bruxelles,
Luxembourg: European Commission, 1999.