0% found this document useful (0 votes)
297 views69 pagesPorus Structures
Uploaded by
Michael FridayCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
297 views69 pagesPorus Structures
Uploaded by
Michael FridayCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 69
GENERATIVE ALGORITHMS
CONCEPTS and EXPERIMENTS: POROUS SHELL GENERATIVE ALGORITHMS
CONCEPTS AND EXPERIMENTS
2_PorRous SHELL
ZUBIN KHABAZI
(© 2011 Zubin M Khabazt
This book produced and published digitally for public use. No part of this book may be reproduced in any manner whatsoever,
without permission from the author, exceptin the contest of reviews,
To see the latest updates visit my website or for enquiries contact me:
www.MORPHOGENESISM.com
[email protected] To Michael Hensel
Generative Algorithms, Concepts and Experiments
Introduction
The idea behind the ‘Generative Algorithms, Concepts and Experiments’ is to explore design and
igotithmic solutions through experiments rather than pure geometrical/digarithmic subjec’s. Using curent
design problerns or ongoing researches, here the development of such subjects pusued with a poramettic
‘approach in Grasshopper/Rhino Environment.
This book is. pat of @ research that was accomplished as a master thesis in Architectural Associaton,
Emergent Technologies and Design couse under supervision of ivichael Hensel, Mike Weinstock and Achim,
‘Menges. in ths research, ful study of a natural system combined with algorithmic solutions in form-finding
techniques fo develop cn digarithmic design fool. Here just a brief summary of research pilus ifs cigorithmic
ark reflected. The research filed “Algorithmic Morphogenesis’ which generally investigated the ficld of
form generation in algorithmic/digital spaces based on natural processes of form generation.
Considering the fact that parametric design is c fast growing subject and global experiments develop
increasingly, here the cin of these series is fo open up some of these subjects with detailed discussions on
technique and methed and also in each book, trigger some of the techniques that seem fo be the point of
concentration among those who share their interests, problems and ideas in Grasshopper Forum, weblogs
or in global workshops as well. in this new book, with less description on Definitions ond a globdl overview of
the concept of algorithm progress, while all les are attached to be explored by readers, the idea is to
bring forward these concept. hone it helps you in your design career.
Zubin Khobazi Contents
1_Porous Shell System...
1_1_Introduction..
1_2_what are Micro-Organisms?...
1_3 Micro-Organism's Morphogenesis (process of form generation)
1_4 Micro-Organism's Forms: Ideal forms and Deformations.
1_5_ Geometry and Topology Analysis
2_Algorithmic Morphogenesis
2_1_Computation: General Consideration ners
2.2_Method...
2.3 Concept of the Algorithm
2.4 Algorithm Grasshopper DefiNitiON...cnuncn
Algorithm_Phase One_Center Point and Contour lines 7
Algorithm_Phase two_Triangulation...
Algorithm_Phase Three_Chamber Generation 30
3_ Fabrication.
3_1_Fabrication Techniques
3_2_Fabrication Algorithm.
Algorithm_Phase Four_Fabrication and Assembly. 39
Fabrication_Step_1_ Splitting the surface into pieces 39
Fabrication_Step_2_Joint Generation 40
Fabrication_Step_3_ Slut Generation..
Fabrication_Step_4_Nesting and Labeling
Fabrication_Step_5_Manufacturing and Assembly. 60
3.3 Outlook...
Bibliography. 65 1 Porous Shell System,
1.1Lintroduction
In a very general definition, architecture was aimed to inhabit human inside a space that covers him
from wild extreme environmental conditions. During the centuries, it developed and achieved higher
levels of complexity and advancements in technology, space organization, style, material assembly and
so on. Buildings become complex products that have so many parts and have to fulfill different tasks.
Among different building elements, building shells and envelops, as the border between interior and
exterior space, developed with different ambitions to give beauty to outside, comfort and protection to
inside, Building shells, from ancient domes, up to modern hi-tech envelopes have improved in
technology, construction techniques, size, space quality and geometry and their beauty. But it seems
that there are still more to explore.
Looking at a specific line of research in building shells
Nature always seemed to be a great source of idea for human and for architecture as well. It inspired
architects in so many different ways. In a very specific method, Frei Otto developed a technique, called
Form-Finding, through which he tried to find the state of equilibrium of predefined material systems in
predefined conditions while they become self-organized with natural forces. He used this technique to
apply as a design medium in architecture. One of his researches in the field of building shells was
studying the hard shell of some micro-organisms like Radiolaria and Diatom. In his research with
biologist 1.G. Helmcke, they carefully scrutinized how these organisms’ hard shell get the
nature and they tried to apply that process to a group of form-finding and design experiments.
form
This book is the further continuation of the methodology of Fri Otto in order to develop a shell system
based on the formation process of the micro-organisms’ hard shell. While using the methodology of
form-finding, here it is pursued via current tendencies of design and technological advancements in
building technology and the main aim is to develop the same methodology in digital space and the final
product should be able to be fabricated with new machineries like Laser-cutters or CNC machines.
It should be mentioned here that the book you are reading mostly reflects the digital algorithmic part of
a bigger research since the original text is way too much to be interesting to read as a design experiment
summary.
1.2. what are Micro-Organisms?
Micro-Organisms are small microscopic organisms which are being studied in microbiology. There is 2
specific group of micro-organisms that share the same property which is a hard skeleton as a shell,
Which remain as fossils after their death. Some of them are: Calcareous Nannofossils, Conodonts,
Diatoms, Foraminifera, Ostracods, Palynology and Radiolaria (www. radiolaria.org),
Radiolaria: Radiolaria are single cell organisms, part of marine planktons occur in all oceans. They are
non-motile structures that float in the water and their body also prepared to enhance the buoyancy of the organism for this life style. Radiolaria are interesting for biologists and other scientists because of
their silicified porous shell as one of the nature's most amazing structures. “Individual radiolarians are
normaly in the size range of hundredth to tenths of millimetres, but some reach the dimension of
milimetre or more [..]" (www.radiolaria.org).
Fig.1.1. Samples of different species of Radiolaria
1.3 Micro-Organism’s Morphogenesis (process of form generation)
The process of form generation of micro-organisms like Radiolaria or Diatoms is among interests of
different scientists up to architects and engineers. Since the result is a hard shell which has structural
capacities, functional aspects of the cell and also includes the process of self-organization inherent in
the process, itis useful to extract the concept of their form generation from the careful observation and
study for further use.
The general concept of the form generation of micro-organisms is put forward by D'Arcy Thompson,
revolutionary scientific thinker and author of the classic ‘On Growth and Form’, based on vesicular
formation. This concept is as follows:
“The organism produces closely packed foam of vesicles around its cell nuclei. Then siliceous materials
from seawater precipitate gradually in the interstices of the foam bubbles and start to shape the
organisms shell. Since this deposition happens on top of these closely packed bubbles and vesicles, it
produces a layer like an envelope over the cell (and not happens inside the cell), so that when this
formation process were completed and vesicles reabsorbed, the final skeleton would be a latticework
of holes’. Fig.1.2. Formation process of hard shell in micro-organisms
So basically, close packing of pneumatics (vesicles) and deposition of siliceous material over their
interstices is the concept of form generation of these micro-organisms which is believed by scientists,
| oe TST
2 i Dé
Fig.1.3. Close packing of Pneumatics as a concept of form generation 14 Micto-Organism’s Forms: Ideal forms and Deformations
In an ideal situation, the hard shell which is like a lattice should have a regular hexagonal pattern. This is.
because of the mathematical laws regarding identical sphere packing. But perfect geometries rarely
happen in nature. There are multiple reasons that cause the pattern not to be homogeneous at all
times. ifthe differentiation of bubbles in the net happens (which happens almost always) then the result
would be chambers like deviated polygons (not always hexagon, but it could be pentagon or heptagon
as well) or deviated circles. The shape of chambers is usually convex which is related to concept of their
formation from a pneumatic base.
Fig.1.4. Regularity/Irregularity of the pattern
_Regional organizations of chambers
Having various global forms (especially double-curved and not flattened), hard-shell of the micro
organisms could be seen as an accumulation of chambers over a global surface. So based on the form,
curvature and geometrical aspects of this global surface, general and local organization of chambers
could be an issue of observation. This might show that how different types of chambers sorted or how
they gradually change their size or pattern in order to cover the outer surface of the organism. Zonal
organization of chambers could be simplified on a sphere as follows:
JERS
Fig.1.5. Zonal organization of chambers (Frei Otto’s studies in IL) 1.5 Geometry and Topology Analysis
Geometry
The general form of Radiolaria has different varieties. There are multiple cone shapes, spheres, and their
deformations. Having this variety, in terms of geometry analysis, it is more useful to look at the lattice in
small scale.
‘As mentioned before, the general shape of the lattice has hexagonal pattern. Looking closer at each
chamber in the net, it is surrounded by six others. Chambers are not as homogeneous and regular as a
mathematical model and a closer look would reveal that each chamber by itself has a deviated and
differentiated hexagon or ellipse geometry. Especially when the net pattern converges towards the tip
point, it causes irregularity inside the net. In most cases, chamber sizes and lattice thickness are bigger
close to the edges and smaller in middle of the shell. The size differentiation is gradual across the lattice
unless in marginal areas where irregularities happen.
Fig.1.6. Geometry analysis in chambers and lattice.
Topology
While geometry includes shape, form and logic of pattern, position and other related issues, Topology
refers to the connectivity and neighbouring condition of chambers. Since pattern formation of shell
always discussed as nets which have segments and nodes, so connectivity or ‘topology’ of nodes of the
net is important. If each bubble’s (sphere) volume centroid accepted as a node in the net, then
connectivity would be defined as line segments which connect these center points. In a shell, these
points would be situated in the middle of chambers in an empty space.
There are different methods of connectivity between these nods but a triangulated net seems to be the
best representation of the connectivity pattern. In a mathematical point of view, a triangular mesh is the
best way of covering curved surfaces by meshes. 10
ee
Fig.1.7. Different methods of Chamber Connectivity (Topology)
_Summary
atom Ostoraton ee tty mnie
Zonal Organization eon eta
Diterentton in ase it ere
ener Mortoeay Gertemicctietiortomeenectcte, | teystantyentoman ncaa ce uu
2.Algotithmic Morphogenesis:
2.1_Computation: General Consideration
Studying the natural process of form generation in micro-organisms, now the aim is to start the digital
process of form generation through computational techniques but first some general notes should be
considered:
Generic form
Considering the process of form generation and looking at the shell, the general form of the radiolaria
are always respond to the pneumatic structures and in general it is mostly spherical or conical. But the
idea in digital algorithms is a bit different. Here any generic double-curved surface should be considered
as an input geometry to start the algorithm and generate the form. It is obvious that the result of the
algorithm should not just limit to the catalogue of forms of radiolaria. This process is aimed to be used
as generator of the same concept for any given surface which later on could be transferred into a
physical entity that should perform as a shell system.
Scale
One of the main issues in Biomimetic and using nature as a source of idea to generate something
artificial is scale. Some great ideas in microscopic scales are not viable to transfer and mimic in the
macro scale. That’s way any physical outcome of the biomimetic, undergoes lots of experiments and
tests to check if the idea is working on the proposed scale or not.
Here the idea of scale has a general consideration. It is possible to generate the resulting model with the
average scale of micro-organism’s chambers. So whatever is the scale of main surface, the resulting
lattice would have the same ratio as for example radiolarian has. This would be satisfying for the
algorithm in a digital space. in addition, any predefined scale, even far different from what radiolarian
has, would be possible since the form generation is still in digital realm and up to the time that this scale
is flexible, the process would be acceptable.
But if the process is aimed to end up in a physical outcome and designed for a product, then scale would
be an important factor. In this situation, there are multiple factors that should be taken into account.
What is the material? What is the scale of the project? What is the method of fabrication and assembly?
What is the degree of self-stability? What is the established performance of the system? ...
So basically shifting the process towards any physical experiment and project, means further research
that should be carried on in relation to that specific material and also criteria of performance that this
experiment designed to fulfill 2
_Ideality vs. Reality
{As studied in micro-organisms, there are different cases of irregularities and deformations in the pattern
of chambers or in the geometry of chambers by themselves. The question here is, to what extent should
the lattice be regular and where should it be allowed for irregularities. Should the algorithm even refine
the geometry in a limited degree to generate the ideal lattice or it should exactly follow the predefined
situation and use irregularities in form generation.
Answer to these questions is not an easy one. There are multiple issues that affect the process and
result and define criteria to evaluate the degree of regularity/irregularity depends on the material, scale,
fabrication method and so on. For example using only complete circles as the main shape of the
chambers might cause some parts of the lattice become thicker or thinner. This would further affect the
general weight of the structure in a big scale experiment. But deviation in chamber's shape towards
ellipses would retain the thickness of lattice almost the same with different modules,
It is also possible to refine the geometry in order to fix irregularities and generate a regular and ideal
output, yet the degree of refinement depends on the preconditions of the material experiment
(material, scale, fabrication, ..). Since these refinements would change the general geometry, in a real
project while there are connections between all elements of design, this type of geometry changes
should be precisely observed. It is assumed that it changes the position of certain points or areas inside
the general geometry that would cause disjunctions with other parts of the design.
_Final outcome
The process of computation has two main steps. The first step is the form generation process and the
result would be a digital geometry which is associated with certain external data. This is the main aim of
the algorithm which converts the natural morphogenesis of an organism to a digital one. But the final
‘outcome of such an algorithm is not something that could be directly used as a product.
The next step is to convert the generated geometry to a real material product. This could be any type of,
material with suitable fabrication technique. Based on the physical necessities and structure,
fabrication, assembly and any other consideration in the field of that specific material system, further
steps of the algorithm should be designed and developed to meet those specific conditions. The second
part of the algorithm also informs the first part in terms of size, numbers and so on. B
2.2.Melhod
Based on all general issues discussed, the method of computation should respond to the form
generation concept, studied in natural form generation phase. In this method, for any given surface, as
the final state of the shell, organized groups of chambers should be generated over the surface to
subtract and form the lattice. This lattice then should be prepared for desired method of fabrication.
In more detail, a reverse process of analysis of organism's shell, here the method of form generation in
algorithmic space is to triangulate any given generic surface to make a net and then generate chambers
over the nodes of this triangulation net. Removing the area inside each chamber from the surface would
result in a porous surface as a representative of the lattice.
The main underlying part of the algorithm is the way this input surface triangulates. It is important that
triangulation should encompass all geometrical/topological aspects of the organism's shell which is
studied before. This triangulation should be center oriented, towards the center point of the surface and
it should consider size differentiation from edges towards center point.
Fig.2.1. Analogy of the resultant net of close packing and triangulation. Topology Analysis > 3 6
2 1 € Form-Finding Logic
Fig.2.2. Process of form generation
nse timpotm ees be pots
Fig.2.3. Triangulation 2.8 Concept ofthe Algorithm
As outlined in the main research diagram above, the core of this research is the design algorithm, the
algorithmic morphogenesis tool that works based on the natural morphogenesis of the organism being
studied. This algorithm as described in the algorithm diagram has different inputs, variables, switches
and manipulations also in multiple phases. Here these phases are described as an overview:
Fig.2.4. Form-Finding Algorithm progress 16
-Algorithm Phases:
1_Phase One_Center Point and Contour lines
The organization of chambers in radiolaria is center-oriented and radial. This type of organization in any
given space needs a center point that should be found. After locating center point, several closed curves
as contour lines should be generated around it up to the edges of the surface. The general scale of the
surface and desired size of chambers define the amount and distance of the contour lines. This phase of
the algorithm corresponds to the general geometry of the input surface.
2_Phase Two. Surface Subdivision and Triangulation
After finding the center point and generating contour lines, the surface is ready for triangulation. As
studied before, this part corresponds to the logic of sphere packing. Basically the algorithm divides the
contour lines into pieces and connects them to generate the base lines of triangulation. The number of,
base lines and the distance between them are all adjustable. Manipulation of them depends on the
chamber sizes and local surface curvature. The result of this phase is the virtual position of all spheres
(bubbles) as if they packed together. This phase corresponds to the physical behavior of the natural
morphogenesis of the organism.
3.Phase Three_ Chamber Generation
In this phase all preconditions for chamber generation is met and process should focus on the single task
of extracting data and drawing chambers. Each node of the net is a center point for a chamber to be
drawn around it, using parameters of all net segments connected to that point. After all chambers being
generated, these curves and their inside area should be subtracted from the main surface which makes
allattice shape surface which is porous shell
4Phase Four_Fabrication
Although did not mention before, all issues regarding size and number of chambers, thickness of lattice,
etc. should be defined by the values coming from the product, design, material and fabrication
necessities. While the design phase has been informed by such information, now the porous shell is
ready to go to the process of fabrication and assembly. It should be decided which fabrication technique
going to be used and what is the material and how the final material product would be assembled 7
2.4 Algorithm. Grasshopper Definition
Algorithm_Phase One_Center Point and Contour lines
As mentioned, this part of the alge m designed in order to find the
center of organization to generate closed curves as contour lines from.
KET J emer toners edes tom sme ses toworé ge ones) hich
KL A | eu te hve es fr ttgatn hese cvs sed fo
my edges towards the center point).
9
Fig.2.5. Algorithms input; A generic NURBS surface. 18
Fig.2.6. The main surface introduced to canvas by . Components of this surface as a BRep
d to get its edges. These edges are d in order to generate start points for radial
subdivision lines. At the same time, mid-point of the main surface also d as the second point
for radial subdivision lines, Since start points and end points of edge curves are overlap and produce two
points at each corner, end point of all division point lists omitted using with the index number
equal to the number of divisions.
Fig.2.7. Radial subdivision lines. Although these lines are not lay on the surface, for this level of
algorithm they would be fine.
To generate contours, these radial subdivision lines should be divided in order to provide points for
curve generation (as contours). The next step is to prepare parameters to evaluate subdivision lines. 19
Fig.2.8. Non evenly distributed values between 0 and 1 generated and first and last numbers are also
excluded from the list
Fig.2.9. Evaluation of subdivision lines. Here by using for
subdivision lines and for evaluation parameters (t), the way that
data (points) branches, reversed so it could be used directly to generate
separate closed contour lines. 20
Fig.2.10. Using all d points as vertices to as contour lines.
AB
Fig.2.11. A. Radial subdivision lines with evaluated points on them. 8. Contour lines
Based on the way Contour lines generated, they are generally affected by the geometry of the edge
curves of the main surface which shifts the irregularity of the boundary of surface to the central part of
it. But Organization of chambers in the organism is a bit different and more regular at the central part.
So it seems that a level of modification is needed, in order to achieve more regular geometries in central
part of the surface. ae 6
DN
Fig.2.12. In the above diagram, the first one from left shows how curves are generated under the effect
of boundary condition of the surface, and the middle one shows if they wanted to be circular around the
center point. In the third one, curves started with the circular shape but gradually changed to adapt to
the shape of edge curves. This manipulation is the aim of the next part of the algorithm.
Fig.2.13. In order to achieve smooth curves, here they wit
would be circle-like curves.
less control points. The result
Fig.2.14. looking at figures above, in the first one from left, non-modified contours seem to be affected
by surface edges too much. In the third one, rebuild curves seem to be affected by the smoothness of
the central-circular shape of curves too much. Looking at the middle figure and comparison between
two sets of curves, it seems that a solution between these two sets of curves could be the desired oni 2
The technique is to use rebuild curves as the final state which non-modified curves tend to
metamorphosis towards them with the maximum amount in center and the minimum amount close to
the edges.
Fig.2.15. In order to literally ‘morph’ contour lines towards rebuild curves, sufficient amount of sample
points across both of them are needed. These sample points would be used to pull points of original
curve towards the rebuild one. That's why both curves are d and series of vectors are generated
from sample points of contour lines towards rebuild ones. U part of the is set to True.
aa
Fig.2.16. While the idea is to move sample points of contour lines, using these vectors, the power of
these vectors should be modified. So at the end, vectors close to the center would have maximum
power and those close to the edges would be minimum. 23
Fig.2.17. All modification vectors are visualized.
Fig.2.18. If these d sample points d again, the result would be modified contour
lines. 24
Fig.2.19. Selected curves in green are modified contour lines.
Fig.2.20. Now if these modified contour lines being projected onto the main surface, the result would be
‘the desired contour lines on the main surface. Here edge curves and center point are again brought
forward to the end part to be ‘Baked’ for the next step of the algorithm. 25
Fig.2.21. Baked contour lines, Edge curves and surface Mid-point.
Algorithm_Phase two_Triangulation
This is the second part of the algorithm with a new file to proceed.
This part is designed in order to triangulate the surface in the space
between contour lines. The idea is to divide contour lines and get
some points, and in the space between any two curves, connect
these points, one from each curve with a zig-zag pattern to generate
triangles in between. But the strategy to design the algorithm should
be clarified first.
It is possible to divide edge and contour curves and in each pair of
adjacent curves, connect every other point to get a triangulated net.
But let's look at the diagram below to check what might happen in
this technique:
If edge curves assumed as one closed curve and then this closed
curve and contour curves are divided and connected, it would be
impossible to control the relation of triangles with the corners of the
surface. So when triangulated, position of triangles would be
irrelevant to the boundary conditions and final position of chambers
would be unorganized in the surface. Here in order to avoid this
situation, division of points should be made upon contours which are
splitted by diagonal subdivision lines of surface (which were
visualized in the next figure). 26
Fig.2.23. In order to generate a zig-zag pattern of connectivity, two sets of points are needed. That's why
here two lists of curves are prepared to split and divide for the next steps. These two lists a
sorted contour lines without the last one. 2. list of sorted contour lines without the First one.
list of 7
Fig.2.24. The next step is to prepare number of divisions for splitted curves. Here it should be noticed
that number of divisions increase one by one, so when the first curve divides into two parts the next one
divides into three, the next four, .
Fig.2.25. Checking the resulted geometry, it is clear
now that curves are divided in an incremental
pattern and there is always one division point in
each corner of the surface and its associate division
line up to the center of the surface. 28
Fig.2.26. While division points are all available, it is now possible to generate the list of points to
generate zig-zag lines. Just a small bit is that again here last part of each zig-zag line would overlap and
lay on the first part of the adjacent zig-zag line and that’s why here by using a component, all
first points of second lists are omitted. The result is a component with data prepared for line
generation.
Fig.227. Woven data list used to feed a
component to generate polylines as triangle edges. 29
Fig.2.28. Again as triangle edges, as nodes of net and a new which connects sides of
triangles across contour lines, are brought forward to be ‘Baked’ for the next stage.
Fig.2.29. Final Triangulated surface with baked net lines and nodes. 30
Algorithm_Phase Three_Chamber Generation
In terms of form-finding, this is the last step of the progress. Here a
closed curve as a chamber should be generated on any node of the
network of triangulation
‘The strategy is to check each node and list all lines which are
connected to that node. Then a closed curve would be generated
‘through the midpoint of these selected lines.
Fig.2.30. Polycurves are exploded in Rhino
=m) and introduced to. the canes by
. In the first step marginal
nodes (those which lay on the edge
curves) should be removed from the list.
That's because in this specific case it is not
ed to generate any chamber at the
edges.
Fig.2.31. In this part of the algorithm all lines of the net which associate with each node should sort in
cone separate data branch. So nodes are ed and the of each node is calculated from
Start and End point of net lines to check whether the start/end point is close to (almost lay on) the node
or not. Fig.2.32. Sorted net lines are being d in their midpoint at t= 0.5 (it is possible to use other
parameters to evaluate curves but then all curves starting from each node should have the same
direction. This would happen later on this book).
Fig.2.33. The above diagrams show how chambers are going to be generated and how the final shell’s,
porosity could be changed according to the chambers’ sizes. As lines evaluated in their mid point at the
previous step, this would result in almost maximum size of chambers and thinnest lattice size for shell. Fig.2.34. Now if all evaluated points feed a component with True value for its ‘Periodic Curve’
option, the resultant geometry would be something like this figure. It is now clear that all points are
listed in groups of six around each node but they are not sorted.
Fig.2.35. In order to sort points in their lists, the best possible option is to use Convex Hull of any set of
points. The Convex component sorts points and provides ‘Index’ numbers of them as well, so
points could be selected in this fashion. 33,
Fig.2.36. Feeding with sorted points. The ‘Periodic curve’ option is set to True.
Fig.2.37. The main surface with its chambers created on it in a radial organization, with desired size
differentiation form center to edges. 34
Fig.2.38. The final step. All chamber curves are ‘Baked’ and the main surface splitted by these curves in
Rhino. The main surface is also splitted by the smallest contour line in its center. Surface pieces inside
chambers are deleted. The result isa lattice, as a form found geometry of the ‘Porous Shell’. 35
3 Fabrication
3.1 Fabrication Techniques
The Algorithm that developed so far is a general form-finding algorithm with several variables. In order
to set these variables it is important to subject the form finding process to a material outcome. In such
case it is possible to set different values like scale, size of chambers, thickness of lattice. There are
different possible avenues in front of this research to materialize the final outcome. Here three possible
methods are discussed.
1. Porous Cast
Fig.3.1. Frei Otto's experiments on porous shell systems.
After Frei Otto and Helmcke developed the concept of deposition of siliceous materials over a cluster of
pneumatics as form generation process of hard shells, they worked on the ‘Porous Cast’ as a form-
finding technique to materialize the form of these shells in experiments. This method was quite the
same as the natural form generation process.
2. Component based - Geodesic Dome
Geodesic Domes invented by Richard Buckminster Fuller in the 1940 and the aim was to cover large
spaces. Fuller developed different techniques and geometrical configurations for geodesic domes. Since
making a complete sphere was geometrically complex, he tried to find methods in which triangulation of.
a sphere could be organized in a way that all parts become the same size. One of his projects was Fly Eye
dome which he developed for affordable housing. It was a modular based system with a porous shell but
covered with transparent pieces. 36
Fig.3.2. Buckminster Fuller’s ‘Fly Eye’ dome.
3. Sheet Material /Cutting based Fabrication
Fig.3.3. Flat-Piece Modulated geometry for sheet material fabrication (Mock up).
The third option of the materialization is to use current technological equipment and machineries for
fabrication, like Laser-Cutter or CNC-Machines while using flat sheet materials. In this technique,
geometry of the porous shell should be splitted into pieces to be cut and then assembled together, to
form the desired geometry. 37
Fig.3.4. Mock up
Fig.3.5. Comparisons and Evaluations 38
Selection of the Technique
Since the idea is to use algorithmic tools and digital machineries, the third technique suits the project in
different ways and sounds more viable for this stage.
Fig.3.6. One piece of splitted porous shell as a mock up, cut out of flat sheet.
3
2 Fabrication Algorithm
Fabrication necessities
The geometry should be splitted into pieces to project onto a flat sheet to cut.
A system of joints should be developed to attach and assemble pieces.
All pieces should be labeled in order to address them in assembly phase.
A proper material should be selected for fabrication. Properties of this material would affect the
algorithm in certain ways.
5. Assembly instructions and plans might be needed and should be provided by the algorithm
beforehand.
6. Any post-production action, if affects the final piece or assembly, should be considered
beforehand.
7. Preparation for fabrication usually needs data management, file exchange between softwares,
technical manipulation of drawings to match the requirements of machineries and so on. It
contains manual jobs and non-parametric! computation as well.
BeN 39
Algorithm_Phase Four_Fabrication and Assembly
brication Preparation Process
Based on the sheet material fabrication the rest of the algorithm is designed.
In order to prepare the porous shell/surface to fabricate by flat sheet material with the bridle joints to
assemble, there are five main steps that should be taken:
1. The porous surface should split into pieces (components) which are able to be cut through sheet
material
2+ Joints should be generated
3+ Sluts for bridle joints should be cut from all pieces and joints
4- Pieces and joints should be labeled and nested into sheets.
5S- All prepared files should be transferred for Manufacture and Assembly
Fabrication_Step_1_Splitting the surface into pieces
1m order to fabricate the porous shell, the first step is to split the form-found geometry into pieces.
Generally for fabrication purposes for any project, there are multiple steps, some of them might be in
Grasshopper, some of them in Rhino, and even in other softwares.
To generate all pieces, the porous surface of the previous step has been splitted in Rhino, using
‘Triangulation’ net lines as cutting objects. As visualized in following figures, the surface divided into
pieces, each one is inside one triangle of the triangul:
Fig.3.7. Using all lines of triangulation (exclude edge lines) as ‘cutting objects’, the porous shell (surface)
splitted into pieces in Rhino. As clear in the second image, now pieces can be selected individually. 40
Fabrication_Step_2_Joint Generation
2-1- Sluts/Joints Baselines
Since the surface of the porous shell divided into pieces, it needs joints to connect these pieces
together. These joints are made from the same sheet material, in the shape of small cylinders with two
sluts on them in the direction of a diameter, that any of the surface pieces should slide into these sluts
‘gently and by three connections for each piece, they would remain fix in their position. So the next step
would be joint generation for all pieces in their division line, adjacent to the next piece.
In the first stage of joint generation, position of these joints on the surface pieces should be extracted
from the model. These positions are on the connectivity lines of surface-pieces mid-points. So if all
adjacent surface-pieces mid-points connect together, the baselines for joints would be there. Lets look
at Grasshopper definition:
Fig.3.8. Overall view of the Grasshopper Definition: “4_Slut_Baseline_Generation’.
Working with component based systems, it might encompass hundreds, sometimes thousands of
relatively small pieces and it is crucial for any fabrication algorithm to sort them efficiently, usually
based on the assembly logic. So for instance if pieces should assemble row by row on top of each other,
‘then the sorting mechanism should be in the same fashion.
Fig.3.9. The first part of the algorithm sorts all pieces in separate rows and then connects their centers
to each other to make the first level of connectivity lines. There are three inputs for this part
, , and , all baked geometries from previous parts. 41
Fig.3.10. The main challenge in this part is sorting all pieces which is being described in more detail. In
the very first step, all components are introduced to canvas by and all contour lines are
also, by . Contour lines are sorted by their and their list reversed to have a list
of all contour lines from longest one to the shortest.
In order to sort all pieces, first they should sort in rows from bottom to top of the shell. That's because
assembly of them would be in the same order, row by row, from bottom to top. In this case, the helping
geometry to sort these pieces is contour lines. Pieces between any two adjacent contour line should be
sorted as one row.
The next step would be sorting pieces in each row one by one in a circular fashion. So at the end there
would be separated groups of pieces sorted in rows and in each row, sorted pieces by their connectivity. 42
The way that these pieces are sorted is as follows:
‘triangle has been used to check whether the
plece is inside the area between two adjacent
contour lines or not. First, the closest point of
the piece’s mid-point has been found for both
contour lines. So there would be two lines that
pty bpp xy Artonteat
* Lupa < tee
connect piece’s mid-point to these new
founded points (Lp1 and Lp2), and there
another line that connects these two points to
each other (Lpp}. To check ifthe piece is inside
the area between two contour lines, this third
line (Lop) should be always longer than other
two (Lp1 and Lp2), otherwise the piece would
be outside and should be omitted from the list.
Fig.3.11. In the first step, pieces should be sorted between two adjacent contours, and here two lists of
contour lines needed. The first one is a list of contour lines (without the last one), and the next one
would be shifted list (contour lines without the first one). 43
Fig.3.12. Now all surface mid-points are calculated by , and the closest point to both lists of
contour lines are also founded (P1 and P2). The same component gives the distance between any mid-
point and its associate closest point on curve as well (Lp1 and Lp2).
Closest points of the first points list are also calculated to get the distance between two contour lines at
that position.
Fig.3.13. Since the gives the distance as well, here the length of Lp1 and Lp? are used to make
comparisons with the distance between contour lines Lpp. By a gate , it is now possible to find
those distances wl
are both smaller than the distance of contour lines (Lp1 both lists of components and their
‘mid-points. Components and mid-points are sorted row by row in different data branches.
is time to sort them in each row in a circular fashion.
lines again to sort
Now all pieces are sorted between contour lines, it
‘Although it would be better to use polar coordinates, here the idea is to use contout
pieces one after each.
Fig.3.15. All sorted mid-points are used to find their
closest points on . Since
‘the components gives the parameter (t) of that
closest point, this parameter as a numerical value
used to sort all components and mid-points across
contour lines. Fig.3.16. First set of connectivity lines in horizontal direction. 46
Fig.3.17. In the next part of the algorithm, all adjacent surface-piece’s mid-points should be connected
to each other in vertical direction. The main problem in here again is the way that points should be
sorted to connect via a component.
Fig.3.18. The result of the line connectivity would be a group of lines that generally converts the slut
baselines into a pattern of deformed hexagons. |
Fig.3.19. In the last step, all connectivity lines are being sorted in groups, each group consists of lines
attached to a surface mid-point. This has been achieved by controlling the distance of their end-
points to the surface-mid-point. For each point, those lines whose start or end points are almost has the
0 distance from the mid-point are sorted into a group.
Fig.3.20. In order to select part of each connectivity line as the position of sluts, all lines whose end
points are close to the surface-mid-point are ed to have the same direction as others, so at the
end, all lines have their start point, close to the surface-mid-point.
In the last bit, as the position of joints (and sluts), just a portion of each line is selected.
Fig.3.21. Baselines for Sluts/Joints. 48,
2:2- Joints
Fig.3.22. Since all baselines are ready, it is time to generate couple of planes at each surface edge and
perpendicular to the surface and generate basic circles of the joints on them. The size of the circles is
relative to the size of the surface piece and two different formulas could be used, one for generation of
joints, and one for sluts which are smaller than joint a bit.
.23. basic circles for joint generation. 49
Fig.3.24. Circles are being d for this next step to become cylinders in both sides. The amount
of extrusion should be extracted from the material thickness that aimed to be cut from.
Fig.3.25. Joints. 50
Fig.3.26. In this small bit, all surface pieces are also extruded a bit to perform Boolean subtraction on
‘them later on.
Fig.3.27. Surface pieces and their associate joints. All these pieces and joints should be baked in order to
continue with later stages. Here the (x/2) as the size of sluts is used. s1
Fabrication_Step_3_Slut Generation
Fig.3.28. All cylinders (as representative of sluts which are smaller than actual joints) are being
subtracted from the surface pieces in Rhino. The result is surface pieces with sluts on them. 52
Fabrication step_4.Nesting and Labeling
-1_Nesting Pieces
All components are being generated, now it is time to label them and nest them into sheets to cut out
from desired material
Fig.3.29. The very first step is to convert pieces from a volume to a surface to project only a surface into
nesting sheet. Here by a single algorithm all surface volumes are exploded and sorted by their size and
‘then their biggest faces has been selected. Since these files are potentially having loads of calculations
they are separated into multiple files in order to prevent memory/computer crashes.
Fig.3.30. Surface pieces should be baked for the next step of the algorithm.
Fig.3.31. The algorithm for nesting has four main parts. The very first part is exactly the same as the
sorting part of the previous section when all pieces were sorted. Here the new pieces with sluts are used
as . 53
Fig.3.32. After sorting all pieces again, surface midpoints are evaluated to find their associate surface
Normals, used to pieces onto a sheet. The sheet is a rectangular surface, divided to have
bunch of frames across it. Since there are more frames on this surface than needed, these frames are
splitted by the number of pieces. All surface pieces are ed, then d and their edges are
into a surface the same as the previous sheet but a bit bigger. These projected lines are
<t again to give some closed shapes, nested in the sheet, ready to cut.
Fig.3.33. Nested pieces. sa
Fig.3.34. In order to label pieces, a list of text tags should be provided. These text tags should show in
which row each piece is posited and the number of each piece in the row. With the same fashion as
nested pieces, all pleces of the porous shell in their original positions are also labeled which help in later
assembly stages to check the correct positions of pieces with the digital model.
HELP
In order to label pieces, a string should be generated that shows the position of the piece in rows and
its index number in that row. So for example ‘Row_2_#_1' will show that the piece is in the row 2 of
‘the shell and in that row, it is the first piece.
So here two different numbering systems needed. To generate row numbers, the algorithm should
know how many data branches the list of components has. Here using a and a
- component calculates the number of branches. This would be the number of values that 2