Вы находитесь на странице: 1из 11

PARAMETRIC & GENERATIVE:

The road to shape design

Domenico D’Uva

The wide spread of digital parametric tools creates a hectic series of possible ways to approach
design tasks, thus producing a deeper way to analyze these tools. The analysis is aimed to the
comprehension of which tools to use for specific needs and more generally if it is possible to
define character of a tool, which may solve the widest available range of problems in
architectural design.
The needs solved by digital tools are basically of two kinds; the first is reaching a good grade of
efficiency in design and fabrication processes, the second is the searching for innovative,
effective and complex shapes for architecture.
The reach for efficiency is the main field of research in parametric, thus creating the family of
modeling software commonly known as BIM. The main theme of this paper is instead the task of
finding and fabricating optimized shapes for architecture through generative digital tools.
The case study in the application of such task is the Parametric CAD 3d drawing course held in
Politecnico di Milano. The student’s task is the comprehension - through experimentation of
digital generative tools – of how architectural complex surfaces have been created. The
spontaneous student approach to this problem is most of the time based on CAD software,
which gives multiple constraints. These are mainly related to a general lack of surface complexity
editing, typical of common CAD solution. This is due to a better grade of knowledge of this family
of tools by students. A small part of class went through their research with advanced modeling
tools aided with generative digital solutions. The result achieved by this part of the students were
much better in quality than the rest of students, because of personal contribution given to the
model that does not remain static, but it acquires new status of shape-equilibrium as a reaction
to parameters modification.
The research taken through this paper analyses the results got from the course in Architecture
faculty in Politecnico di Milano called Parametric 3d Cad Drawing. The 3d representation level of
students is basic, so the approach proposed tries to give them digital tools to perceive
architecture behaviour.
The teaching is provided between a theoretical approach and a practical application on a
building chosen by student.
The theoretical approach, developed by professor Andrea Rolando, is given through a method to
analyse shape-complex buildings, creating a model composed of key-elements only, through a
representation of the students’ interpretation of the built shape.
Key-elements are shape driver components, most of the times hidden under the building skin,
which render up the complexity. Shape-creator elements should be found analysing the structure
through geometry, because it is essential for structure to answer to precise spatial dimensions,
whose only way to accomplish is geometry. Therefore analysing building structure is useful to
spoil the complexity of the visible skin.
The analysis approach is paralleled with a widening in studies of geometrical representation,
because it is discovered a general ignorance in basic elements also, like ellipse, hyperbola, or a
catenary. This curves are perceived by student like metaphysical entities, not connected with
real life architecture; this leak of knowledge leads to a misuse of complex shapes and a discard
of solution relying on these forms. It leads, however, to a stagnation of visual thinking ability
because awesome unknown shapes are mantled with an aura of fascination that hides the real
connection between perception of shape and method of creation of the shape itself.
The practical part is therefore accomplished by drawing the key elements of a building chosen by
each students through digital tools taught during lessons. At first, they are provided a general
knowledge of digital modelling, making students understand the basis of computer aided design
and then they are taught with the basics of Digital parametric and generative design solutions.
It is important, indeed, to provide a method of solving basic geometrical problems with digital
tools and the basis for choosing the right tool for the right problem, because it is frequent, in
students’ behaviour to solve all of their problems with the only tool they know, which is frequently
AutoCAD. Digital modelling knowledge in students is frequently provided with a series of
practical AutoCAD lessons, or with self-teaching. That is why it is essential to give students a
roadmap of the existing tools. The roadmap provides a classification in families of digital tool,
tracing the way from common Cad solution to contemporary parametric tools.
In parallel with digital model generation it is given a general vision of advanced model
representation. This is based on general enlightenment theory; from basic light sources to the
most recent render solutions, as unbiased and GPU-based engines.
The core of the course is focused on the concept of parameter, both as a group of tools, both as
building key components, therefore one of student’s aim is finding a smart way to represent
building key elements (parameters) through parametric digital tools.

Digital Tools: Parametric and generative

Parametric tools are subdivided in two software families, general parametric and generative.
General parametric software answers to a pragmatic demand for efficiency with a focus quite
shifted from the efficient form finding and understanding aim of the course. This becomes clear
studying the workflow and software interface oriented to minimise editing time, not to create new
approach to building itself.
Generative software, instead, gives importance to the pattern recognition as a principle of shape-


  
 

 
creation, because it works as a computer, which executes the code written, following instruction
after instruction until the end program. Coding these lines is a common activity for programmers,
but it is very difficult for shape creators like designer and architects.
Since a couple of decade, all repetitive activities, or automated procedures not coded in the
original software packages, were written directly by advanced users through small programs
called scripts.
Today these problems have been (partially) overcome by using graphical interface, which has
become common with software like Grasshopper (GH), or Generative Component. The graphical
interface, though, covers only a part of the problems that can be solved with software of such
kind; for all the rest of needs there is still the manual coding solution.

Key-elements: Pattern and singularity

Because buildings are composed of a great number of different components, there is the need to
make up a way of categorize the group of similar pieces in hierarchy of importance in the
process of form understanding.
Therefore it is essential to hierarchize elements in patterns and singularity, both in understanding
building shape process both in choosing digital representation tools.
To understand the general composition scheme of a building it is needed to decompose it in
small parts, linked each other with a scheme of geometrical rules. The system of rules that
geometrically links parts may be called pattern.
Singularities, instead, are the exceptions in pattern scheme; in building an example of pattern
may be the Cartesian array of columns, the stairwell will act as singularity.
As patterns may be laid up in buildings, it is possible to analyse and understand living being as
examples complex and efficient element compositions. It is clear that natural process are
improved in efficiency and aesthetics through evolutionary process, so connecting them with
building construction brings to a general improvement in costs and quality.
To achieve a smart composition of pattern and singularity it is useful to give answers to
questions like how can we evolve our architecture ideas in a parsimonious way, or how can we
avoid (as much as possible) blind alleys in evolutionary process. It is clearly impossible to
reproduce the natural evolutionary process, but a theoretical evolutionary model is feasible by
establishing some basic evolution rules and making the system goes through a series of steps
(algorithm) in an accelerated time flux. It is therefore important to fix a set of rules to make
embryo evolve into an efficient but predictable result. This principle seems to work for most sets
of rules but, as Wolfram’s rule 30 showed1, it is impossible to always predict results, even with a
very simple set of rules. The
unpredictability isn’t the only
problem with this kind of
processes because most of
the chosen evolutionary
processes bring to efficient
solutions, which are
completely un-useful for
coherent building purpose.
Even if we add rules for
coherence, as the flatness

    
and horizontality of pavement
       
new complications come
through as cost constraints connected to component feasibility.
Surfaces grown through evolutionary processes are frequently characterized by a high degree of
complexity, which is generated by the number of steps that increase curvature and face
complications. Therefore it is needed, after the generation process, a discretization work to make
component constructible.
Another problem with generative algorithm is the randomness, because even with the process
above described it is not sure whether the solutions will converge to an optimised result or will
deviate toward an incoherent structure.
Due to this series of difficult connected with generative components it was quite tough for
students to create new series of patterns, which could answer to all the issue above-mentioned.
So the path proposed is an interpretation study of architectural specific examples of parametric
and generative created building, making them aware of the creation process, which is not a
spark of creativity, but a long series of semi-predictable steps, which evolves into the final shape
of the building. The general analysis scheme, followed by most of the students, starts from


Stephen Wolfram, A New Kind of Science, Wolfram Media, Champaign, IL, 2002
understanding and represent the presence of pattern in plan, after the elevation, then generating
the whole building characters only as a final step of the work.
Students’ work examples presentation is divided in two main groups. The first is about pattern
recognition, the second’s about generative design.
Students’ works – Pattern recognition
The first work proposed is Land Formation One by Zaha Hadid, which is a building where fluid
shapes have been driven into a precise modelling scheme and analysed in different steps.
The pattern analysis starts in plan with the intersection between lines from two different

 

      

   

directions, the first group of lines starts from a point within a finite distance, the second are
parallel lines (infinite distance intersection). Chosen intersection of lines becomes tangent lines
of Bezier curves, which are geometrical entities, whose tangent in end points become control
polygon of the curve. It is important to underline this issue, because the control of the curve
forms is given by the direction of lines, keeping it easy to work on. Plan surfaces, which
characterize this famous piece of architecture, are created by a couple of Bezier curves starting
from the same point, joint at the opposite side with a perpendicular edge. The pattern works with
all the concrete flattish surface of the building, but it is the pond beneath the building that makes
up the singularity, brokening the scheme. The flat surface of pond is made up by the connection
of two curves that crosses all the others underneath. The pattern is applied in elevation: a centre
pole around which each surface rotates generates volumes. These volumes interfere with the
ground with different section plan that becomes shifted edges for the volumes.
The analysis goes further in deepening how volumes are composed in detail, creating an abacus

    


  
  
of Boolean operation, with a series of further editing processes.
The second example analyses the pattern and singularities of Meiso no Mori Crematorium by
Toyo Ito. Due to the high complexity of the building it has been necessary to limit the analysis to
surface roof. It has been conceived as a Digital Terrain Model (DTM) ruled by contour lines and
a reference plane called “zero plane” to which refers all the height of the representation.
As a DTM, the roof is
composed of peak
(maximum) and nadir
(minimum). The
intersection between
vertical plans connecting
these points and the
DTM forms curves. The
pattern formed by these
curves is the changing
point in second
derivative that lies on the
zero plane. Therefore if
the height of zero plane
         varies, then the curve
   

inclination varies though


leaving maximum and
minimum as constraint.
The areas where there is
higher occurrence of
intersection between
maximum-minimum-
maximum curves are areas
of high complexity, and
higher aesthetical interest.
Because all the building is
substantially without

       



    
properly intended connection to the ground, the bearing wall on the west side becomes a
singularity. The bearing walls, together with the lack of columns creates a space, designated to
crematorium with an ample vault, which interrupts the pattern which worked for the rest of the
building.

Student’ work – Generative design


Students found quite
tough to cope with
generative design, so
only a brief
application taken from
a group exercise has
been considered in
this paper. It is about
London City Hall by
Lord Norman Foster.
Analysis has been 
      
focused on the façade
related to the shape
of levels. The façade toward Thames River is profiled from a portion of a sphere, but the section
of the whole building has a silhouette, which can be resembled to a revolutionary rotational

        
  
         
ellipsoid. After a first analysis has been accomplished it was clear that an ellipsoid composed of
an ellipsis revolving around and axis with a circular orbit cannot give an accurate shape
representation. Therefore it was necessary to modify the generation scheme of the ellipsoid into
an elliptical orbit ellipsoid. Once external shell shape has been defined it was necessary to focus
on the relation between floors and the ellipsoid.
The floors have all a circular shape but if we make a section of the revolutionary ellipsoid with
horizontal plane we obtain elliptical planes. The question is, if we rotate the ellipsoid around an
axis does it exist an angle that generates section of the ellipsoid which shapes is a perfect
circle? The answer to this question was possible applying to the surface a generative design tool
as Grasshopper and produce a number of solutions from where we can pick the right one.
It was necessary to apply a geometrical concept to produce the sample circle section to which
refers all other sections. In fact if we section a general ellipsoid we have an ellipse, but this
ellipse has two axis of different length. We can pick up the smaller axe as a ray of the sample
circle to which refer all other sections. Then while manually rotating the whole ellipsoid, GH
produces sections in real time, until we find the correct angle, which gives us circular sections.

           


   
   

Conclusion
As analysis of students’ work has demonstrated the approach to the generative and generally
parametric tools proved to exchange their way to think architecture. The achievement possible
with this kind of teaching is the enhancing of students visual thinking, their ability to read
complexity and re-elaborate constructions in small controllable pieces, instead of a single
complex architectonical organism.
The process of learning parametric tools must be paralleled with geometrical knowledge
because tools need perceptive cognition to work in a meaningful way.
General parametric tools study finds students more enthusiastic because practical results are
immediately ready, but the creativity is reduced and the general approach to fast building
solution seems to prevail over thinking of innovative shapes.
Generative tools, on the other side find more resistance to become popular because of the
complexity and the youth of tools, which makes them less intuitive to most of users.
At present day it seems difficult to find a single tools that can solve pragmatic approach to shape
production and creative approach to form-finding problems.
The only tool that can try and solve part of the problem seems Rhinoceros, which tries to go
beyond its limit of pure modeller tool toward generative and building information modelling plug-
in.

List of sources of illustrations


1. Author
2. Stephen Wolfram, A New Kind of Science, Wolfram Media, Champaign, IL, 2002
3. Egle Costantinopoli
4. Egle Costantinopoli
5. Egle Costantinopoli
6. Alessandro Menini, Gerardo Semprebon
7. Alessandro Menini, Gerardo Semprebon
8. Copyright for betteraccess.org and thinkofthefuture.com @ Seedgen Ltd.
9. Author
10. Author

Вам также может понравиться