Professional Documents
Culture Documents
Roberts 4259769 ResearchArticle
Roberts 4259769 ResearchArticle
Contact Information
Name: Sarah Jean Roberts
Address: Oostplein 1, 2628 EE Delft
Telephone No. +31 (0) 618083641
Email: roberts.sarahjean@gmail.com
University Information
TU Delft Student No. 4259769
Faculty: TU Delft Faculty of Architecture
Department: Architectural Engineering
Studio: MSc 3/4 Graduate Studio 11
Advisors: Moritz Fleischmann, Tjalling Homans, Martijn Stellingwerff
Designing Form-Active Shelters | MSc. Arch. Thesis 2
Abstract
1 Background........................................................................................................................ 6
1.1 Introduction + Relevance........................................................................................................................6
1.1.1 Temporary Architecture...........................................................................................................6
1.1.2 Portable Shelters........................................................................................................................7
1.1.3 Form-Active Systems................................................................................................................7
1.2 Problem Statement..................................................................................................................................8
1.3 Research Questions..................................................................................................................................9
1.4 Research Objectives.................................................................................................................................9
2 Research Methods and Results........................................................................................ 10
2.1 Research Framework.............................................................................................................................10
2.1.1 Overview...................................................................................................................................10
2.1.2 Case Study: Tent......................................................................................................................10
2.1.3 Research Scope + Limitations...............................................................................................10
2.2 Methods: Form-Finding........................................................................................................................11
2.2.1 Introduction.............................................................................................................................11
2.2.2 Form-Finding Methods..........................................................................................................12
2.3 Methods: Fabrication............................................................................................................................20
2.3.1 Introduction.............................................................................................................................20
2.3.2 Possible Methods: Material Selection...................................................................................20
2.3.3 Possible Methods: Cutting Patterns......................................................................................24
3 Conclusions...................................................................................................................... 26
3.1 Main Conclusions.................................................................................................................................26
3.1.1 Parametric Tools......................................................................................................................26
3.1.2 Form-Finding...........................................................................................................................26
3.1.3 Materiality, Cutting Patterns, and Nesting...........................................................................27
3.2 Reflection..............................................................................................................................................27
3.3 Further Research...................................................................................................................................29
4 List of Figures.................................................................................................................. 32
5 List of References............................................................................................................ 34
6 Acknowledgements.......................................................................................................... 37
Introduction
Temporary Architecture
Figure 1.1.1. (2010) Altare della Patria. Monumento Nazionale a Vittorio Emanuele II.
Figure 1.1.2. Asal, Berk. (2010) Spacebuster. A mobile inflatable public space designed for New York City.
Figure 1.1.3. Renzo Piano Building Workshop. Examples of Traditional Portable Shelters: North American Tipi, Bedouin Nomadic Tent, Asian Yurt.
1.1.2 Portable Shelters They are also used to flex- Bending-Active Systems
ibly handle modern-day threats,
Portable shelters are one including climatic fluctuations and A class of form-active sys-
type of temporary architecture built extreme environments (Fig. 1.1.5). tem whose geometry is based on
since the start of humankind. They The common tent is a ubiquitous the elastic deformation of initially
are typically lightweight and deploy- type of portable shelter; it provides planar elements can be classified as
able with a relatively short lifespan. a key case study for this research. a bending-active system [7]. Bending-ac-
They are also easily assembled, dis- tive structures rely on pre-stressed
assembled, and transported. Com- singly or doubly curved units for
mon vernacular portable shelters 1.1.3 Form-Active Systems stability and strength. (Pre-stress in-
include tents, tipis, and yurts (Fig. volves the intentional introduction
1.1.3). Many portable shelters qual- of stress into a structure to improve
These vernacular shelters ify as form-active systems, or systems its performance [8]). Bending-active
resurface in contemporary architec- in which the envelope has nota- structures (including tents) have the
tural practice, especially when envi- ble structural capacities. As stated power to integrate their geometry,
ronmental, social, and economical by Moritz Fleischmann and Julian structural system, and envelope in a
circumstances demand alternatives Lienhard, form-active structures single functioning entity (Fig. 1.1.7).
for human survival (Fig. 1.1.4) [5]. enable a high level of integration
in long-spanning and lightweight
Portable shelters also de- structures [1]. Many portable shel-
part from their vernacular roots, ap- ters made of stretched cloth, cords,
pearing in high-tech modern forms. and poles – such as tents – fall un-
Contemporary deployable models der the form-active class. Such shel-
are often flat packable and conserve ters are the focus of this research
material through prefabrication [6]. (Fig. 1.1.6).
These makeshift structures are of-
ten used for recreational purposes.
Problem Statement
1.2.2.
Primary research questions: The goal of this research was More generally, this article aims
to explore some possible techniques to promote the creation of more
1. How can computational para- for the design of small-scale equi- structurally efficient, aesthetically
metric tools be used in the design librium structures. This comprised beautiful, and financially economic
process for small-scale form-active the implementation and evaluation shelters. The computational meth-
structures? of certain digital tools (namely Kan- ods explored would ideally be paired
2. More specifically, what are the garoo for Grasshopper). Further, it in- with traditional analog techniques to
strengths and weaknesses of Kan- volved the identification of key fac- create customized shelter typologies
garoo for Grasshopper, a plug-in for tors that should be considered in the for a wide range of scenarios. This
Rhinoceros, in the form-finding design process, including the phases might include solutions serving the
process? of form-finding, selecting materials, requirements of different programs,
and generating cutting patterns. The scales, and environmental contexts.
output of this research would ideal- Ultimately, these methods can sup-
Secondary research questions:
ly be useful for designers and indus- port built solutions to flexibly han-
try partners alike, to help guide the dle modern-day threats, including
1. Once form-found, what are some
design and manufacture process of climate fluctuations, environmental
possible methods for the fabrication
portable structures such as tents. extremes, and social upheaval.
of small-scale form-active struc-
tures?
2. More specifically, how might Kan-
garoo and Rhinoceros aid in the pro-
cesses of selecting materials and
generating cutting patterns? Which
alternative methods are recom-
mended?
Figure 1.2.2. Tomlow, J. (1989) Antoni Gaudis Hangemodell und seine Rekonstruktion.
Research Framework
Overview
Case Study: Tent
Research Scope and Limitations
2.1.1.
Figure 2.1.1. Roberts, Sarah. (2011) Sierra Designs Tent. Mom and Aunt Sarah camping at Beehive Lake, Idaho, USA.
2.2.1
Methods: Form-Finding
Step 1: Construct Base Geometries
2.2.2.
Kangaroo is just one of Kangaroo can simulate var- 2.2.2 Form-Finding Methods
several possible approaches using a ious forces (i.e. gravity and spring
spring-based particle system for simula- forces) affecting the particles in the
tion. (Processing, for example, can system. Forces can originate from 2.2.2.1 Step 1: Construct
achieve identical results.) A particle many different sources designated Base Geometries
system can be thought of as a net- by the designer, such as geomet-
work of independent objects, usu- ric constraints or material elasticity In this step, the basic start-
ally depicted as simple dots [13]. In [12]. All forces are represented with ing members of the case study were
Kangaroo, particles can be connect- idealized force vectors, allowing for modeled prior to the simulation
ed with damped springs to approx- live interaction and negotiation be- of forces. This included modeling
imate real physical performance of tween them. Spring force vectors, bending-active members, ten-
non-rigid structures. More detailed for example, are based on Hooke’s sion-active members, and anchor
information about spring-based law of linear elasticity [15, 16]. For points to be included in the spring-
particle systems can be found in this research, a Velocity Verlay solv- based simulation. This initial phase
the publications of Axel Kilian and er was used. (The solver denotes the started to define the basic function-
John Ochsendorf [14]. integration method used by Kanga- al features and formal appearance
roo to calculate new positions of of the structure. The choices made
particles [12]). during this phase also inform possi-
ble manufacture techniques.
Figure 2.2.2. Left: Geometry of Rods: Before and After Bending Force. Right: Variations in Parameter: Spring Rest Length.
2.2.2.1.1 Base Geometries – Here, the arcs that resem- Base geometries can be
Digital Model ble the tent poles were subdivided modeled in Rhinoceros or in Grass-
into smaller line segments. Each hopper. As a parametric design tool,
1. Bending-Active Rods segment acted as one spring (Fig. Grasshopper offers the possibility
2.2.2). It should be noted that in to quickly alter the definition. Of
reality the rod’s mass would have course, Rhinoceros also has its own
Two bending-active rods
been distributed throughout its advantages for modeling, especial-
composed the support struc-
length, instead of at the endpoints ly for those not yet proficient in
ture for the tensile membrane. In
of each line segment [12]. While Grasshopper.
Kangaroo, a spring is represent-
this configuration was not accu-
ed as a line segment connecting
rate in terms of continuum me-
two points (particles). Therefore,
chanics, it still provided a formi-
in order to approximate the be-
dable estimation of real behavior.
havior of a flexible member, the
overall shape must be broken into
smaller line segments and mod-
eled as a series of separate springs.
Figure 2.2.3. Left: Planar Mesh Before Tension Force. Right: Variations in Parameter: Spring Rest Length.
Methods: Form-Finding
Step 1: Construct Base Geometries
2.2.4
Figure 2.2.4. Left: Anchor Points, Pre-Simulation. Right: Anchor Points, Post-Simulation
Figure 2.2.5. Analog (Physical) Model of the Tent Case Study (Scale 1:10)
Methods: Form-Finding
Step 2: Implement Forces
2.2.6
Figure 2.2.7. Spring Force: Rod-Membrane Connections. Left: Base Geometry Pre-Simulation: Flat Membrane and Rods.
Right: Variations in Parameter: Spring Stiffness.
Methods: Form-Finding
Step 2: Implement Forces
While simulating the bend- Spring Force: Mesh Relaxation Spring Force: Rod-Membrane
ing force in Kangaroo, variations in Connections
certain parameters were tested in re- Here, springs were used to
al-time. (Fig. 2.2.2) shows variations model the textile membrane of the Springs were also used to
in the spring rest length parameter, tent. The starting mesh of intercon- model the connections between
which triggered changes in tent nected springs and particles was “re- the textile membrane and the two
height: (A) Rest Length = 400mm, laxed” to form-find its initial shape, rods (Fig. 2.2.7). The end points of
(B) Rest Length = 500mm, (C) Rest given the assigned anchor points. each line segment subdivided from
Length = 580mm. Further infor- While the simulation ran, particles each rod were connected to corre-
mation about the bending force in moved relative to each other. Thus sponding points on the mesh. These
Kangaroo can be found in the work the sum of forces acting upon each springs were crucial; they acted on
of S.M.L. Adriaenssens and M.R. particle was in constant flux [13]. both rods and membrane. They
Barnes [18]. maintained tension in the mem-
Tensile membranes carry
brane by providing the required lev-
loads only through tension force.
el of pre-stress.
2.2.2.2.1.2 Spring Force Pre-stress in the membrane elimi-
nates compressive forces. This type The stiffness parameter for
Introduction: Spring Force of system can be extremely light- each spring is one of many which
weight and achieve a significant de- can be adjusted to the preferences
gree of transparency, if desired [8]. of the designer. (Fig. 2.2.7) shows
As previously explained, Pre-stress also gives the structure variations in spring stiffness that
Kangaroo uses springs to simulate stability given external loads such as changed membrane stability and
the behavior of non-rigid struc- wind and snow. tent height: (A) Stiffness = 10, (B)
tures. This is possible because all Stiffness = 500, (C) Stiffness =
In Kangaroo, springs can be
materials, even relatively stiff ones, 1000. If the connection points are
ascribed certain rudimentary param-
can stretch and compress in re- not satisfactory, their positions can
eters, including rest length and stiffness.
action to the forces exerted upon be tweaked until membrane behav-
These parameters were changed in
them [12]. In Kangaroo, springs be- ior is in-line with the designer’s in-
real-time to allow for optimiza-
have in tension or compression ac- tentions.
tion. (Fig. 2.2.3) shows variations
cording to Hooke’s law F = kx. This in spring rest length, which induced
law states that the force F needed changes in tent surface area, elas-
to extend or compress a spring by ticity, and transparency: (A) Rest
a distance x is proportional to that Length = 0mm, (B) Rest Length =
distance. Various basic spring forc- 40mm, (C) Rest Length = 80mm. In
es applied to the tent in the digi- variation (A), the area of the mesh
tal simulation are explained below. was minimized, much like the be-
havior of a soap film.
Methods: Fabrication
Introduction
Possible Methods: Material Selection
2.2.8
2.3.2.1 Introduction
2.3.1 Introduction
Figure 2.2.8. Spring Force: Edge Cables. Left: Base Geometry Pre-Simulation: Flat Membrane and Rods. Right: Variations in Parameter: Spring Stiffness.
Figure 2.2.9. Left: Digital Tent Model. Right: Analog Tent Model (Scale 1:10).
Methods: Fabrication
Possible Methods: Material Selection
2.3.1. 2.3.2.
Figure 2.3.1. SL-Rasch, Sefar Architecture, Buro Happold. (2010) Courtyard of the Al-Masjid al-Nabawi, Medina, Saudi Arabia.
250 deployable sun umbrellas (each 306m2) composed of steel structures with PTFE membranes.
Figure 2.3.2. (2014) Load-Elongating Curves for Select Fibers. Stress/Strain Curves: Comparison of tensile properties of man-made and natural fibers.
This proposal can be at- But such software can also GSA Fabric, developed by
tempted using Grasshopper’s ex- be used for the design of small- Oasys, provides analysis of non-lin-
isting math and list functions. The scale shelters such as tents. While ear fabrics in dynamic relaxation.
Python code PyFEM could also different physical systems and ma- Maya is comparable to Kangaroo as
be implemented, as it can handle terials are at play, the same approach it uses a spring-based particle system
non-linear material behavior. As still holds valid at different scales. for simulation. But it also provides
stated by Clemens Preisinger, de- the option to select from a handful
Multimedia Engineering
veloper of the Karamba plug-in, cur- of common membrane materials.
Pte. Ltd. produces WinFabric, soft-
rently Karamba cannot handle ma- ware that integrates the properties
terial non-linearity. of commonly used fabrics obtained
from biaxial testing (Fig. 2.3.2).
2.3.2.3 Possible Methods: WinFabric also facilitates the tasks
Alternative Software of form-finding, nonlinear finite el-
ement load analysis, patterning, con-
It may be more efficient nection detailing, fabrication, and
(and effective) to address material- assembly. German-based company
ity not within Kangaroo, but using Sofistik AG produces software that
alternative software. Large-scale allows users to define stress-strain
form-active structures are already curves for materials and spring ele-
largely designed with specialized ments (Fig. 2.3.3).
software incorporating specific
material properties. Of course, dif-
ferent scales require different to-
pologies, materials, and structural
detailing.
Figure 2.3.3. Multimedia Engineering Pte. Ltd. (2014) WinFabric Software. Setting Fabric Properties.
Figure 2.3.4. Sofistik. (2014) Sofistik Software: Measuring Maximum Principal Tension Stress.
Methods: Fabrication
Possible Methods: Cutting Patterns
2.3.5. 2.3.6.
2.3.3 Possible Methods: in turn affects cutting patterns in the for cutting. It also allows the design-
Cutting Patterns manufacture process. Traditionally, er to assign priorities to objects and
cutting patterns for different mate- identification tags (Fig. 2.3.4).
2.3.3.1 Introduction rials were determined by manually
disassembling physical prototypes.
The process of generating The following overview provides The Shrinking Method
cutting patterns for tensile mem- some possible digital techniques to One approach to generating
branes requires translation of the approach the patterning process. cutting patterns in Rhinoceros is to
form-found 3D geometry into 2D
deliberately shrink the fabric mem-
shapes. These shapes can then be
brane by intentionally cutting its
nested, meaning their position and 2.3.3.2 Possible Methods: parts too small [21]. First, the mem-
orientation in relation to each oth- Grasshopper and Rhinoceros brane seam layout is defined by its
er can be optimized to minimize
mesh subdivision pattern. Optimal
fabric waste. The final shapes can Generating cutting patterns subdivision can be attained by eval-
then be cut from the appropriate can be challenging because tensile uating stress in the membrane for
flat material by the manufacturer structures assume their equilibrium adequate distribution. As a general
and fabricated together. This pro- shape under a certain degree of pre- rule, seams should trace the zones
cess can be difficult because the stress. While a rigid panel can be un- of greatest stress. And regardless
necessary pre-stress in the mem- rolled in Rhinoceros, a doubly curved of scale, designs of the same shape,
brane must be accounted for. membrane is a different animal. material, and pre-stress typically re-
Rhinoceros and Grasshop- quire the same number of panels.
In the design of form-active
structures, different materials (i.e. a per indeed offer some features and Once the final form-found
very stretchy nylon versus a more plug-ins that facilitate patterning for geometry is baked out of Kanga-
rigid polyester) affect structural membrane structures. For example, roo into Rhinoceros geometry, the
behavior. This behavior affects the the plug-in Rhino Nest (Rafael del anti-clastic shape can be converted
structure’s equilibrium shape, which Molino) optimizes the position and to flat panels derived from the mesh
orientation of parts on a flat surface
faces. These panels can be down- 2.3.3.3 Possible Method 3: tern with dimensions 1-2% less than
sized according to their material Alternative Software those of the final surface. K3-Tent
type. Thus the necessary pre-stress can also define appropriate mar-
is present when the membrane For cutting patterns, MPan- gins for cutting and implement al-
pieces are welded together along el-R can be used as a plug-in for Rhi- gorithms for nesting. These cutting
its seams, ensuring stability and noceros to facilitate basic form-find- lines become prospective seams for
strength in the assembled structure. ing, seam allowance, and nesting. the final design.
This method was used by Anish Ka- Alternatively, the final form-found ExactFlat can convert a 3D
poor and Arup for the construction geometry can be exported out of model into a 2D pattern, including
of Marsyas, an immense elliptical Kangaroo and Rhinoceros and into necessary compensation for mate-
membrane structure installed in the different software to address issues rial stretching. It can also nest pat-
Turbine Hall of the Tate Modern in of patterning and nesting. Compad terns on a cutting template to opti-
2002 (Fig. 2.3.5). by Tentnology and Sofistik AG’s mize yield. Further, ExactFlat allows
software (introduced in Possible
Since non-linear materi- the designer to make adjustments to
Methods: Material Selection) can output
al behavior depends on the quality material type and dimensions while
cutting patterns (Fig. 2.3.6).
of applied stresses, compensation simultaneously providing cost in-
values are often determined from German company Technet formation. Lectra and Gerber can
experience (and physical material GmbH produces EasyNT, soft- also aid in patterning and nesting
tests) instead of calculation. Values ware that aids in the comprehensive processes (Fig. 2.3.7).
around 1% warp (the direction of membrane structure design process,
the fabric’s long threads, generally including form-finding, non-linear
stronger) and weft (the direction of load analysis, cutting pattern gener-
the cross-threads, generally weaker) ation, and nesting. K3-Tent by GeoS
are typical for common membrane can subdivide doubly curved surfac-
fabrics. es and approximate their flat state
by mapping them onto a plane. As
a rule, the mapping results in a pat-
Main Conclusions
Parametric Tools
Form-Finding
3.2.1.
3 Conclusions per and Kangaroo is visually intui- on Java and Processing) can all achieve
tive, allowing direct manipulation of nearly identical results. RhinoMem-
the design without a high level of brane, a plug-in for Rhinoceros, uses
3.1 Main Conclusions specialized knowledge or technical FEM-based algorithms and can
skill. estimate specific pre-stress levels.
3.1.1 Parametric Tools The ixForten software can support
3.1.2 Form-Finding a wide range of processes beyond
In retrospect, parametric form-finding, including engineering
computational tools proved to be Kangaroo proved to be an and production phases (i.e. struc-
useful for aiding certain aspects of effective tool for approximating re- tural non-linear analysis and pattern
the design process for small-scale al-life physics in the form-finding layouts).
equilibrium structures, particularly process. But Kangaroo is just one of
for form-finding. The primary ad- many spring-based particle systems
vantage of parametric modeling lies which can facilitate form-finding.
in its ability to test several iterations Like its comparable tools, Kanga-
in a short time period, demanding roo has its own strengths and weak-
minimal effort from the designer. nesses. Processing, RhinoVault, and
Further, the interface of Grasshop Karsten Schmidt’s Toxiclibs (based
Figure 3.2.1. Roberts, Sarah. (2014) Tent Typology 1. Scale: 4 m2, Capacity: 4 people, System: bending-active + tension-active. Form-found in Kangaroo.
3.2.2.
Figure 3.2.2. Roberts, Sarah. (2014) Tent Typology 2. Scale: 3 m2, Capacity: 1 person, System: bending-active + tension-active. Form-found in Kangaroo.
Reflection
3.2.3.
While the case study se- Please note that the digital
lected for this research was a sim- methods deemed effective in this re-
ple small-scale tent, many other search are still most effective when
form-active typologies can be cre- paired with physical tests. Simula-
ated with similar techniques. For tions are abstractions that provide
example, canopies, hammocks, and a basic understanding of real-world
kites all support different functions physics. An idealized computation-
but share some related geometric al model, no matter how precise,
and physical systems. Kangaroo is cannot account for all the complex
flexible enough to roughly model conditions that exist in reality.
these typologies and output custom-
ized models for specific situations.
(Fig. 3.2.1, 3.2.2, 3.2.3, 3.2.4, 3.2.5,
3.2.6) show some example typolo-
gies all form-found in Kangaroo.
Figure 3.2.3. Roberts, Sarah. (2014) Tent Typology 3. Scale: 40 m2, Capacity: 5 people, System: tension-active. Form-found in Kangaroo.
3.2.4.
Figure 3.2.4. Roberts, Sarah. (2014) Canopy Typology. Scale: 200 m2, Capacity: 50-100 people, System: tension-active. Form-found in Kangaroo.
Further Research
3.2.5.
3.2.6.
Figure 3.2.5. Roberts, Sarah. (2014) Installation Typology. Scale: 200 m2, Capacity: N/A, System: tension-active. Form-found in Kangaroo.
Figure 3.2.6. Ibid.
This article was written in partial fulfillment of the requirements for a Master of
Science Degree in Architecture at the Delft University of Technology in the Netherlands. I
would like to thank all the advisors, contributors, technical experts, inspirations, and friends
who have supported me during this project. Further information can be requested from the
author by email: roberts.sarahjean@gmail.com.
Andreas Brun
Andrew Barrett
Bruce Roberts
Daniel Piker
Debbie Roberts
Elisa van Dooren
Gabi Schillig
Henry Shires
Jia-Rey Chang
Jim Keysor
Joe Heywood
Judith Reitz
Leon Spikker
Lieneke van Hoek
Lucas Bolte
Marcel Bilow
Mårten Nettelbladt
Martijn Stellingwerff
Matthew Tanti
Michael Probyn
Molly Roberts
Moritz Fleischmann
Paul Soethout
Peter Eigenraam
Rasmus Holst
Rogier Houtman
Romain Thijsen
Thibault Clar
Thijs Asselbergs
Thomas Billet
Tjalling Homans
Virginia Clasen
Yuxiao He