You are on page 1of 11

FOREWORD

On architects and engineers


Jörg Schlaich

There is still a widespread misunderstanding concerning If well formed, there are no bending but membrane
the role of architects and structural engineers: it is forces only (axial compression and tension) in a
said that architects are the designers of a building shell, permitting its thickness to be around 80mm
from concept to detail, whereas the engineers (only) for reinforced or prestressed concrete, even down
care for its stability. In fact, it is its function which to 12mm for fibre reinforced concrete (Fig. 0.1).
clearly attributes a building to either an architect or Though these concrete shells initially do not leave
an engineer only, or to both: to an architect only if much space for an architect (or even for the fantasy
it is multifunctional in a social context – typically a of an engineer), it is fortunately not unusual that the
family house where no engineer is needed – and to an two collaborate fruitfully or that an engineer himself
engineer only if it serves a singular structural purpose has the courage and imagination to go beyond strict
– typically built infrastructure such as a bridge where logic.
no architect is needed. A high-rise building typically So Pier Luigi Nervi’s Palazzetto dello Sport in Rome
needs both, an architect and an engineer. would have fulfilled its purpose at considerably lower
The more the form of a building or structure cost with a tensile ring on vertical supports instead of
develops from its flow of forces, the more it is under the inclined Y-shaped columns as built, but it would
the responsibility of the engineer. have looked like a boring and ugly tank (Fig. 0.2). Only
Especially due to the fact that most infrastructure, a creative engineer could have made such a proposal
such as towers, power plants, long-span roofs and as built!
bridges, is large and long-lasting, a responsible In case of the large hypar roof for the Hamburg
engineer will seek the advice of an architect or a Alster-Schwimmhalle, the architect insisted that the
landscape designer when deciding on the material or edge beams should be free cantilevering beyond the
the scale of their bridge or sports hall in an urban or facade. This caused vivid discussions, because according
natural environment. to the classical shell literature a hypar shell cannot
transfer shear forces from edge beams but needs direct
It is only culture that can convert our built support as evident from the wonderful Candela shells
environment into civilization. (Fig. 0.3).
But the architect insisted and thus made us find a
Shells play a special, singular role for engineers. Their revolutionary though very simple solution by super-
shape directly derives from their flow of forces, and imposing the classical saddle surface for the surface
defines their load-bearing behaviour and lightness, loads with the straight line generators’ surface for
saving material by creating local employment, their edge loads (Fig. 0.4). The result was a perfect structure
social aspect. This is especially true for thin concrete thanks to the insisting architect.
shells with their characteristic curvatures: single During the last decades concrete shells lost more
curvature (cylindrical and conical), synclastic (dome- ground to: 1) cable nets; 2) textile membranes; and
like), anticlastic (saddle-like) or free (experimental). 3) steel grids. In terms of their load bearing, these
FOREWORD: ON ARCHITECTS AND ENGINEERS IX

Figure 0.2 Palazzetto dello Sport by Pier Luigi Nervi, Rome,


1958

Figure 0.1 Bundesgartenschau Pavilion, with a 12mm fibre Figure 0.3 Hypar shell with direct edge beam supports for the
reinforced concrete shell, Stuttgart, 1977 Church of San José Obrero by Félix Candela, Monterey, 1959

can also be considered as shells. Let us discuss one they instead brought in an idea which was absolutely
example for each of them. convincing, to bring together under one continuous and
floating roof all sports facilities: the stadium, the sport
hall, the swimming hall, and all transitions connecting
Cable nets
them (Fig. 0.5). This is exactly what we engineers
In 1967, the international competition for the sport expect from our architect if structure plays a significant
fields of the 1972 Olympic Games in Munich was won role: an idea, a concept, a proposal, but leaving the
by architects Behnisch+Partners from Stuttgart, even structural solution to us. Together with Frei Otto and
though they hardly fulfilled the requirements. In fact, Fritz Auer from Behnisch+Partners, we developed a
X JÖRG SCHL AICH

Figure 0.5 The cable-net structure of the Munich Olympic


Roofs, 1972 (from left to right: Heinz Isler, Fritz Auer, Frei
Otto, Jörg Schlaich, Fritz Leonhardt, Rudolf Bergermann and
Knut Gabriel)

Steel gridshells
The courtyard of the Museum of Hamburg History,
L-shaped in plan, was to be covered with a glass roof,
Figure 0.4 The cantilevering hypar of the Alster-
Schwimmhalle, Hamburg, 1967 as light and transparent as possible. The architect
Volkwin Marg expressed with his sketch his wishes,
his ideas which stimulated our adequate structural
prestressed cable-net structure with 75cm quadran- solution: a quadrangular mesh or grid, 1.2/1.2m from
gular mesh-width, adaptable to any shape, subdivided 60mm × 40mm steel members, diagonally stiffened by
by edge cables, supported by masts, held down by thin, prestressed cables (Fig. 0.6). By change of angles,
anchors and, finally, covered with Plexiglas. This huge the grid can easily adapt to a smooth doubly curved
but nevertheless light and floating roof has been very transition between the two cylindrical shells, which
well accepted and is still very popular, thanks to an ideal themselves are stiffened by radial spokes (Fig. 0.6).
cooperation between architects and engineers, each of Thus, the fear of the client that his historic building
them playing their role in a useful manner. would suffer from the roof could be relieved.
FOREWORD: ON ARCHITECTS AND ENGINEERS XI

Figure 0.6 Sketched and completed steel gridshell for the Museum of Hamburg History, 1989

Figure 0.7 Roof for the ice skating rink of the Wolfgang Meyer Sports Centre Hamburg-Stellingen, Hamburg, 1994

Textile membranes such as those for the Wolfgang Meyer Sports Centre
Hamburg-Stellingen (Fig. 0.7).
As we know from our clothes, we can produce doubly In a fruitful cooperation of architect and engineer
curved surfaces from plane textile with the help of neither of them will impose their opinion because it
a cutting pattern. If prestressed (by reducing the is only the result that counts! To build in untouched
cutting patterns), they behave under loads as an ideal nature can only be justified by creating a responsible
membrane shell and can even permit convertible roofs, building culture.
FOREWORD

Sharing the same spirit


Shigeru Ban

Shell structures are but one of many different, inter- materials, and so on; ideals I really share with him.
esting structural systems. If a gridshell structure Like me, Frei Otto is not a form-making architect,
happens to be suitable for a project, I use it, but meaning that we always design according to a combi-
otherwise I design another appropriate system. nation of structural logics, architectural constraints
An architect shouldn’t concentrate on one type of and the programmes contained within the structure.
structure, and should take notice of all possible struc- That’s why our collaboration went very well. We share
tural systems. Without an understanding of structures, the same spirit.
we cannot design a building. If you would have some If I design just a shell, I always try to look for the
most appropriate, most structurally efficient shape
to reduce bending moments and the size of the

Figure 0.8 Japan Pavilion at Expo 2000, Hannover

preference, it would be very difficult to adjust to


different programmes.
Frei Otto is a notable exception. He has his own
specialities: cable-net structures, membrane struc-
tures and lightweight structures. He does not design
everything, yet he’s also an architect. When I got the
commission for the Japan Pavilion at the 2000 Expo
in Hannover (Fig. 0.8), I immediately contacted him,
and he agreed to collaborate with me. This had been
a dream ever since I was a student. He does not like
to make complicated connections, he always strives
for minimum effort, minimum labour, minimum
Figure 0.9 Centre Pompidou, Metz, 2010
FOREWORD: SHARING THE SAME SPIRIT XIII

Figure 0.10 Haesley Nine Bridges golf clubhouse, Jeju Island, South Korea, 2010

members. However, I might not always have total solved all the technical problems. Nowadays, for any
freedom to design the most minimal shape for the timber structure, in any country, I work with him. I
material, because the shape of the shell is also defined just send him the design and he solves all the technical
by other constraints. For the Centre Pompidou in issues. He immediately understands what I want to
Metz (Fig. 0.9), a museum, we needed to have large, do: the collaboration between us is really perfect. I
column-free spans for the galleries underneath. As a choose particular engineers for particular projects. As
result, it was designed as a large, suspended surface: an engineering student, you have to develop your own
part of it is in tension, some part in compression. At interests and understand your own abilities and limita-
first sight, the Haesley Nine Bridges golf clubhouse tions. Study under the guidance of a good professor,
has a similar system (Fig. 0.10). But, because we could because an engineer always moves into a certain type
have a repetition of the supports, it developed into a of structure as a speciality. This is not a bad thing,
set of compressive arches, creating the characteristic but merely an observation. Some engineers are an
quality of the internal space. So, the shapes came from exception and are capable of designing anything, like
the programme. Jörg Schlaich, who is also good at shell structures. But,
The structural engineer Hermann Blumer worked as I said, to me it is important that we can share the
on both of these projects, and was a key contributor to same spirit, the same priorities. Even if someone is a
the design of the Centre Pompidou. After he joined genius engineer, if we cannot share the same priorities,
our team, he went back to my original design and it doesn’t work out.
This page intentionally left blank
Introduction

In this book, leading experts from academia and comes from diverse fields – geometry, structural
practice describe design and optimization methods for mechanics, linear algebra, mathematical optimization,
the generation of efficient shell forms and topologies. computer science and reliability theory – each with
Some of these techniques are part of the long history of its own notation, and nobody is knowledgeable in
the theory of shells, others have been developed or first all these fields. Engineers may know about struc-
used in practice by the authors themselves. Recent years tural mechanics, but not be well versed in geometry;
have seen a renaissance of shells. Through advances in mathematicians may know about geometry but not be
computational techniques and power, as well as novel particularly familiar with structural mechanics. Each
fabrication and construction methods, engineers and person assembles their knowledge in a hotchpotch
architects have been imagining and creating elegant fashion since it is not possible to know all there is to
thin-shell structures. For shells to be efficient, their know about shell structures. Nonetheless, this book
shape should depend on the flow of forces and vice is intended as a good starting point and provides
versa and therefore their design requires a process of advice on further reading throughout, explaining, for
form finding. The ‘ideal’ form of a shell may need to different topics, which references will best expand on
fulfil additional architectural, mechanical or technical particular topics.
aspects necessitating some form of optimization. This
makes the design and engineering of shell structures a
Shells for architecture
highly involved process.
Shell Structures for Architecture: Form Finding and The main application of the methods described in this
Optimization offers a comprehensive overview and book are shell structures in the context of the built
hands-on textbook for the form and topology gener- environment.
ation of shells. Intended for students, researchers and
professionals, both in architecture and in engineering, Shell structures are constructed systems described by
this book presents many new and established methods, three-dimensional curved surfaces, in which one
and explains them through theory and through dimension is significantly smaller compared to the
working design examples. other two. They are form-passive and resist external
This book is written so that each chapter exists as loads predominantly through membrane stresses.
a complete entity in its own right, and can be read
as such, but also so that the chapters fit together as A ‘form-passive’ structural system does not signifi-
a whole. Some of the chapters contain mathematics cantly, actively change its shape under varying load
and for some people this aids their understanding conditions, unlike ‘form-active’ structural systems
and insight. However, it is equally true that some such as cable or membrane structures. A shell
very gifted shell designers have used physical and transfers external loads to its supports predominantly
computer models in their work and have not relied through forces acting in the plane of the shell surface,
on advanced mathematical knowledge. Thus, it is which are called membrane stresses and might be
not necessary to fully master the mathematics in compression, or a combination of compression and
one chapter before moving on to a following chapter. tension. The word ‘membrane’ might suggest a film or
However, it is recommended to have some under- fabric that can only carry tension, but the compressive
standing of linear algebra and basic calculus, or stresses in a steel, concrete or masonry shell are still
have some references concerning these topics on called membrane stresses. A ‘thin’ shell has to be
hand. The mathematics applied to shell structures sufficiently ‘thick’ to carry these compressive stresses
2 INTRODUCTION

without buckling. Shell structures can be constructed Form finding


as a continuous surface or from discrete elements
following that surface. In the latter case, we speak The process of designing form-found shapes is called
of lattice, reticulated or gridshells. These terms are form finding or shape finding, where the former term
synonymous; ‘gridshell’ will be used throughout this is used in this book. Computational models for form
book. finding may be a numerical simulation of the physical
This book distinguishes between three types of model involving hanging chains or cloth, or they
geometries for shell structures: might use imaginary properties that could not be
simulated physically. In both physical and numerical
R5 Freeform, free-curved or sculptural shells are hanging models, we note that form-active systems
generated without taking into consideration struc- are exploited to find the shape of form-passive shell
tural performance. If they are shaped digitally, then structures. We define the concept as:
they are often described by higher degree polyno-
mials (e.g. patches of Non-Uniform Rational Basis Form finding is a forward process in which param-
Splines (NURBS)). eters are explicitly/directly controlled to find an
R5 Mathematical, geometrical or analytical shells are ‘optimal’ geometry of a structure which is in static
directly described by analytical functions. These equilibrium with a design loading.
functions are often chosen for their convenience in
performing further analytical calculations and their For shells, the design loading is typically the dead
ability to describe a shell’s shape for fabrication load, most often being its self-weight. For masonry
purposes. These are often lower degree polyno- or concrete shells, or glass-clad steel gridshells this
mials (hyperboloids, elipsoids and hyperbolic or load is dominant. Timber is relatively light, but creeps
elliptic paraboloids), or trigonometric or hyper- under load and therefore the permanent dead load
bolic functions (the catenary). assumes a greater importance. The parameters that can
R5 Form-found shells include natural, hanging shapes be imposed to control the form-finding process are:
associated with the funicular structures of Antoni
Gaudí, Frei Otto and Heinz Isler, but also ‘strained’ R5 boundary conditions, supports, external loads;
gridshells that feature bending stresses. If the R5 topology of the model and;
shape is found digitally, it is initially parameterized R5 internal forces, and their relationship to the
by piecewise or higher degree polynomials. Their geometry.
final shape is the result of attaining a state of static
equilibrium. The geometry is an unknown; form finding is the
generation of geometry. However, the process may
The general topic of shell structures is introduced in require some arbitrary starting geometry. Once the
Part I of the book, ‘Shells for architecture’. Chapter final shape is found, the numerical model is updated
1 elaborates on the vast range of possible shapes for by assigning real physical material and member
shells. Chapter 2 describes how we can conceive good properties. This new updated model serves as a basis
structural forms. Chapter 3 asks what a shell is, and for structural analysis.
discusses the matter from a mathematical standpoint. Methods developed for the form finding of shell
A brief history of the use of physical models in the structures are discussed in Part II, ‘Form finding’.
design of shells is provided in Chapter 4. Chapter 5 These methods solve the problem for static equilibrium,
presents computational analogues for digital design either without requiring material properties (force
and subsequent optimization, and focuses on different density method in Chapter 6 and thrust network
strategies for geometrical parameterization of curved analysis in Chapter 7), or by incorporating (fictitious)
surfaces. This chapter serves as an introduction to the material or spring stiffness, and solving for dynamic
numerical methods discussed in Part II and Part III equilibrium (dynamic relaxation in Chapter 8 and
of the book. particle-spring simulation in Chapter 9). Part II is
INTRODUCTION 3

summarized through a comparison of these methods, optimal structures, one needs to formulate a general
applying them to a single chain in Chapter 10, before definition of the optimization problem:
concluding with Chapter 11 on steering of forms.
minimize f ( x )
Optimization
⎧ gi( x ) ≥ 0, i = 1,…,m
When additional objectives or constraints are intro- subject to ⎨ hi( x ) = 0, i = 1,…,p
duced, methods of structural optimization are needed ⎩ xi D x, i = 1,…,n
to solve for them. These methods can be applied to
structural shapes resulting from form finding, but Here, an objective function f ( x ) of the n design
also to freeform or mathematical shapes. Different variables x subject to m + p constraints is minimized.
strategies for the solution of constrained optimi- These constraints are also expressed as functions of
zation problems, in the context of the structural the design variables and they may be either equalities
design of shells, are presented in Part III, ‘Structural h( x ) – for example, that certain lengths are fixed – or
optimization’. inequalities g( x ) – for example, that the height should
not exceed some value.
Structural optimization is an inverse process in which Even though there is only one objective function,
parameters are implicitly/indirectly optimized to it can be a function of various quantities (e.g. cost,
find the geometry of a structure such that an energy consumption), which are themselves functions
objective function or fitness criterion is minimized. of the design variables. Instead of just adding up
these quantities, it is possible, for example, to
One can optimize for multiple objectives and does not minimize the sum of the squares of the cost and
have to be constrained to a single design loading. In energy consumption, in which a weighting is assigned
this case, the objectives are evaluated based on their to give a monetary value to energy consumption.
importance, either a priori using weightings or a poste- Alternatively, for multiple objective functions, a Pareto
riori by exploring the Pareto front. In multi-objective front is explored to assess the trade-offs between those,
optimization, the objectives can include goals that are often competing, objectives.
non-structural. Optimization often leads to a large Optimization problems are encountered in a vast
set of feasible shapes, called the design space. In cases number of fields, and methods to solve them can be
where such a design space is explored and we tend to roughly divided in two groups: the local and the global
more gradually develop a form, we may also encounter methods.
the term computational morphogenesis.
The ‘optimal’ shape in the design space is subject R5 Local methods reach a local optimum, and therefore
to a set of given requirements (the constraints such as offer no guarantee in finding the global one, unless
allowable deformations) and is defined with respect to the problem is convex. The most common local
one (or several) objectives, such as: methods are based on the computation (or approxi-
mation) of derivatives. Therefore, they require the
R5 the minimization of material and weight in a functions to be differentiable, and the variables to
structure, aiming for material economy; be continuous (specific methods have also been
R5 the minimization of deflections or dynamic vibra- proposed in the literature for discrete parameters).
tions of a structure, ensuring serviceability of a R5 The most popular global methods to find a
structure; global optimum are often based on an imitation
R5 the maximization of stiffness (i.e. minimization of of phenomena observed in nature. They belong
structural compliance), aiming for efficient load- to metaheuristic methods, which can be defined
bearing structures. as general trial-and-error strategies guiding the
search for optimal solutions in hard problems. In
When using mathematical optimization for finding these methods, there is no guarantee either that the
4 INTRODUCTION

optimal solution is systematically found, but they rigid frames and gridshells), the areas of the cross
are designed for the purpose of finding a global sections usually are the design variables.
optimum in complex problems and numerical
experiments have demonstrated their efficiency Chapters 12, 13, 16 and 18 each deal with shape
in certain applications. These techniques include optimization, but towards different objectives and for
simulated annealing and evolutionary algorithms. different structural typologies. The approaches and
parameterizations vary as well. Chapter 16 further
The chapters in Part III use both local and global discusses how to derive a discrete topology, or ‘struc-
methods. Chapters 12, 13 and 17 use local, gradient- tural pattern’, from the resulting shape. Chapter 21
based methods. The remaining chapters of Part III from Part IV explains shape optimization as it has
describe metaheuristic techniques. Chapter 16 uses been applied by Mutsuro Sasaki in his practice.
simulated annealing, whereas Chapters 14, 15 and 18 Chapters 14, 15 and 17 apply topology optimi-
apply genetic algorithms, which are a subset of evolu- zation. In Chapter 14, shape variables are also
tionary algorithms. These are discussed in Appendix C. included, whereas Chapter 15 includes sizing optimi-
Structural optimization is traditionally classified zation. Chapter 17 discusses continuum topology
in three categories depending on the nature of the optimization, which aims to find the optimal layout
variables involved: of material in a given region, but applies this also to
shape optimization.
R5 Shape optimization has variables acting on the
geometry of the structure, without modifying the
Precedents
topology. Practically, in discrete structures, the
node coordinates are often used directly as param- Throughout the book, examples are given of existing
eters to modify the geometry, but more advanced shell structures. Part IV, ‘Precedents’, provides more
parameterizations are available (see Section 5.4 specific insights from built examples. Chapter 19
and Appendix D). compares two seminal gridshell structures: the
R5 Topology optimization deals with the situation Mannheim Multihalle and the roof of the British
in which the topology of the structure is not Museum’s Great Court. In contrast, Chapter 20
prescribed by the designer. The structure’s compares two types of continuous concrete shells:
topology is defined by the connectivity of the the geometrical shapes by Félix Candela and the
nodes in the structure, and the existence or form-found shapes by Heinz Isler. The final chapter,
absence of elements. Chapter 21, gives an overview of Mutsuro Sasaki’s
R5 Sizing optimization has variables representing concrete shell structures.
cross-sectional dimensions or transversal thick- The book concludes with ‘The congeniality of
nesses (the geometry and the topology remaining architecture and engineering’, a short note on the
fixed). For instance, in discrete structures (trusses, future and potential of shell structures.

You might also like