You are on page 1of 8

CHAPTER ONE

Introduction

1.1 INITIAL CONSIDERATIONS


The current dynamics of scientific progress and technological innova-
tions, in addition to the markets globalisation, imposed changes in the engi-
neers training. On the other hand, an innovation of processes and integrated
computer systems was necessary, in order to enable the construction industry
to compete at international level. This change of attitude can also be seen in
the adoption of more efficient and economical structural systems like the
ones that use steel and composite solutions aiming to become cost-effective
and viable alternatives.
One of the main objectives of this book is to enable and boost the use of
steel and composite structures in buildings. For that, it is intended to
improve the training of a new generation of engineers who are familiar with
its behaviour to widespread its use in Brazil. The structural design develop-
ment associated with new constructive techniques turns out to be a direct
consequence of such ideas, but this will only be possible with a better under-
standing of the behaviour of the structural elements that form the global
structure.
This understanding is based on the complete development of experi-
mental and numerical models contemplating the behaviour of elements
and structural systems. This strategy enables a better understanding of
phenomena such as strength, structural stability, and stiffness; fabrication
processes effects; erection aspects; and the steel and composite structural
systems’ dynamic response.
The structural behaviour requires the understanding of a series of phy-
sical phenomena related to the occurrence of ultimate limit states like flange
local buckling, web local buckling, lateral torsional buckling, plastic hinge
formation, or even, the crack distributions shown in Figs 1.1–1.5.
Naturally, over the last few years there has been a considerable evolution
of procedures and models used in the steel and composite structures design.
This evolution was the result of an increased knowledge level of structural

Modelling Steel and Composite Structures Copyright © 2017 Elsevier Editora Ltda.
http://dx.doi.org/10.1016/B978-0-12-813526-6.00001-5 Published by Elsevier Inc. All rights reserved. 1
2 Modelling Steel and Composite Structures

(A) (B)
Fig. 1.1 Flange local buckling.

(A) (B)
Fig. 1.2 Web local buckling.

Fig. 1.3 Lateral torsional buckling of a cantilever beam [8].


Introduction 3

(A) (B)
Fig. 1.4 Plastic hinge formation.

(A) (B)
Fig. 1.5 Crack formation.

behaviour obtained with the use of new numerical and experimental


techniques. The consolidation of such knowledge has been made through
structural design scientific publications present in journals and conferences.
These publications generated discussions and ideas that were later consoli-
dated by the technical committees of structural design standards, and scien-
tific societies such as ECCS, Eurocodes, and ABNT.
However, the incorporation of the outcomes of experimental and
numerical investigations in the design standards is not a simple process,
giving rise to numerous discussions and to the development of more accu-
rate models to be inserted in their design recommendations. These structural
behaviour models can be associated with various natures, complexities, and
formats where the most frequently used will be detailed in the next section of
this chapter.
4 Modelling Steel and Composite Structures

1.2 STRUCTURAL BEHAVIOUR MODELS


1.2.1 Analytical, Mathematical, and Hybrid Models
One of the simplest ways to understand these types of models comes from
examples such as the structural joints. The joints classification in terms
of moment versus rotation curves can be divided into three types: the analy-
tical, mathematical, and combined models. In the analytical models, the
moment versus rotation curve is based on their physical characteristics.
In mathematical models, on the other hand, this curve is expressed by a
mathematical function in which the parameters are determined by a curve
adjusted to experimental results. Finally, the combined models use both
analytical and mathematical models.
The analytical models can be used to predict the joint stiffness based on
their geometric properties and components arrangement. The joint
mechanical behaviour can be predicted by numerical methods such as finite
elements based on the connection components deformation mechanism
hypothesis. With this in hand, the components deformation and the connec-
tion moment capacity can be determined as well as their associated moment
versus rotation curve.
Generally, parametric studies are conducted considering the effects of
several geometric variables related to joint components. Practical values
of these variables are then analysed to produce data for the analysis. How-
ever, the cost and time involved are usually unsatisfactory for practical appli-
cations, because each type of joint or joint component configuration
requires a new formulation for obtaining the moment versus rotation curve,
Chan [1]. In addition, the joint uncertainties can significantly affect their
model predicted stiffness. There is still the fact that additional data-handling
procedures are necessary to incorporate analytical results in the semirigid
frame analysis. Another classic model for evaluating joints is characterised
by the components method present in the Eurocode 3 pt. 1.8 [19], which
is also known as a classical mechanical model. Further details on this model
will be made in the examples present in Chapters 2–4.
Another method used to determine the joints moment versus rotation
curve consists in approximating a curve to match experimental data by using
simple expressions. These expressions are called mathematical models,
which directly relate the connections moment and rotation by mathematical
functions, using curve-fitting procedures. When these procedures are deter-
mined by experimental data adjustment, the moment versus rotation curve
can be explicitly expressed and directly used in structural analysis.
Introduction 5

The mathematical models are simpler than the analytical models previ-
ously mentioned. Examples of mathematical models include the model of
Richard and Abbott [2], Lui and Chen [3,4], and Al-Bermani et al. [5]. Since
numerous tests on various types of connections have been carried out in
recent decades, many data, for various types of connections, are accessible
for determining the parameters or constants required for the mathematical
models.
Composite models are combinations of analytical and mathematical
models. In the composite models formulation, the moment versus rotation
curves are expressed in terms of both adjustment constants as well as geomet-
ric parameters. With the goal of developing a general expression for all
connections with a similar component layout, usually these functions are,
for convenience, standardised. The curves adjustment constants are deter-
mined by curves approximation techniques whilst the geometric parameters
are based on the connection components geometry.
Composite models require few parameters when compared with the
mathematical models and, in addition, similarly to analytical models, they
keep the geometric parameters that establish a correlation with the joint
physical characteristics, a feature that is not found in the curves adjustment
mathematical models. In general, the hybrid models can be used to evaluate
the initial stiffness of particular types of connections and to predict their
nonlinear response. The polynomial model [6] and the Ramberg–Osgood
[7] are the examples of hybrid models.
Another example of a mechanical/analytical model is shown in Fig. 1.6.
In this figure, the web local buckling phenomenon is illustrated in the region
where a direction change in the beam bottom flange occurs. This ultimate
limit state is particularly significant in tapered beams. Chapters 2 and 3 will
depict experimental and numerical models related to this phenomenon.
These results enabled the development of a simple column model where
the constants K1 and K2 were calibrated to resemble an effective width
and a buckling length compatible with the already mentioned experimental
and numerical results.

1.2.2 Experimental Models


The steel and composite structural models may also involve experiments
carried out in the laboratory. These experiments are crucial to the under-
standing and calibration of the investigated structural behaviour and will
be described in detail, as well as the arrangements, instrumentation, and
techniques usually adopted in these tests. This chapter will also address
6 Modelling Steel and Composite Structures

h
√K2
F2
TF
TF K1h
F2

Fig. 1.6 Tapered beam’s local web buckling analytical model.

some aspects and general questions to elucidate their key planning


points in order to ensure the success of the demanded experimental
programme.

1.2.3 Numerical Models


Structural behaviour models can also involve computer simulations. Many
types of simulations are possible involving finite differences, finite strips,
finite elements, computational intelligence techniques, amongst others.
Chapters 3 and 4 of this book will focus on the latter two types of steel
and composite structures computational modelling. First, the numerical
models developed based on the finite element method will be discussed
and presented. Several simulations showing nonlinear behaviours will be
presented, as well as their main details and restrictions in terms of boundary
conditions, major difficulties, solution strategies, and methods to overcome
convergence difficulties. This will continue with the presentation of a series
of examples of computational intelligence techniques application in the sim-
ulation of steel and composite structures response. Thus, neural networks,
neuro-fuzzy networks, and genetic algorithms nebulae will be described
and applied focusing on their main advantages, scope, and limitations.

1.3 STATISTICAL ANALYSIS OF INPUT DATA


AND RESULTS
The correct planning of the investigated models must be carried out
very carefully in order to consider the effects of the main variables that influ-
ence the studied structural behaviour. One of the classic forms of planning
Introduction 7

uses a number of techniques and procedures based on statistics and design of


experiments.
The design of experiments aims at determining enough information with
the least possible number of tests seeking to optimise the number of tests
necessary to encompass all the possible variables. For that, it is necessary
to follow some suggestions for a better planning: recognise, establish, and
clearly delimit the problem; identify the possible factors that can affect
the investigated problem; check which factors can be kept fixed, therefore,
will not have their effects assessed in the experiments; identify, for each
factor, the range and levels that will be studied; choose a suitable design
of experiments (i.e. how the levels of the factors can be combined in order
to solve the proposed problem with the lowest number of tests); and choose
which variables properly measures the process results (quality, performance,
etc.) for the tests data analysis planning.
This can be done through an orthogonal arrangement, in which the
sequence and the number of experiments to be carried out can be deter-
mined. This type of arrangement includes the combination of variables
and their adopted levels. Several methods were developed to represent
the design of experiments such as ‘Latin Square’, Taguchi, ‘D-Optimal’,
and others. The Taguchi method, for instance, works with equal levels for
all variables of the project. Alternatively, the factorial method ‘D-Optimal’
allows the use of different levels for each variable through a quadratic opti-
misation of the complete factorial arrangement minimising the amount of
experiments required to encompass the entire set.
The correct interpretation of the results, curve adjustment, and the main
variables significance assessment that influence the problem can also be done
with statistics-based methods such as those that use tables like the new
Taguchi processes. These procedures, when used properly, can simplify
the process of generating structural design equations. They can identify
and quantify the terms that should be present and which terms may be dis-
carded without an accuracy loss. Further details of these processes can be
found in Lochner and Matar [9].

1.4 BOOK SCOPE


This chapter presented a description of the main models used in the
simulation of the of steel and composite structures behaviour. Chapter 2
contains a detailed description of the experimental models, as well as the
8 Modelling Steel and Composite Structures

main requirements and precautions that should be taken for a steel or com-
posite structure test in the laboratory.
Chapter 3 introduces a series of numerical models, developed with the
aid of the finite element method. Various simulations depicting nonlinear
responses are presented as well as their main details and restrictions in terms
of boundary conditions, major difficulties, solution strategies, and methods
to overcome convergence difficulties.
Chapter 4 summarises the use of computational intelligence techniques
in the simulation steel and composite structures response. Neural networks,
neuro-fuzzy networks, and genetic algorithms are described and applied
focusing on their main advantages, scope, and limitations.
Finally, Chapter 5 presents a synthesis of what was presented in this book
and establishes the main contributions, finalising with some considerations
and thoughts about the present publication.

You might also like