L. Vestarte, P. Vestartas, R.

Kucinskas

Corrugated Cardboard Shell

A pavilion project of an architectural workshop

Lina Vestarte1,* , Petras Vestartas2 , Romualdas Kucinskas1

1
Department of Architecture, Vilnius Academy of Arts (VAA)
01124 Maironio st. 6, Vilnius, Lithuania
2
Laboratory for Timber Construction (IBOIS) École Polytechnique Fédérale de Lausanne (EPFL)
IBOIS, ENAC, EPFL, 1015-Lausanne, Switzerland
*
Corresponding author e-mail: [email protected]

Abstract
This paper presents mesh-discretisation and assembly methods for quad-based shell-
structure. The study case is a corrugated cardboard shell 3m in height and spans
up to 6m in diameter and is supported at 5 arbitrary points. The structure was
built from-two layer 3mm corrugated cardboard sheets, consisting of 391 sets of
planar quadrilateral elements (1537 total count of unique pieces). Since the principal
fabrication constraint was a 2-axis laser cutter, the approximation of cutting angle
for finger-joints was required. The laser cutter, with the maximum bounding area
of 600 x 900 mm, also restricted the size of the panels, while the directionality of
corrugated material posed constraints on geometrical proportions of the panels and
their orientation within the cutting sheet. The shell was assembled without using
heavy scaffolding or any adhesives, thus relying on material strength and friction
provided by the connection method. The assembly sequence and insertion vector
had a crucial role in the act of such construction. The pavilion was assembled within
less than 12 hours, including the time spent for sorting the unique pieces.

Keywords: Corrugated Cardboard, Assembly sequence, Shell structures, Form-

finding, Breadth-first-search, Polygonal meshes, Finger joints

522
 Corrugated Cardboard Shell

Figure 1: The pavilion made from corrugated cardboard displayed in the art exhibition in 2019.

1 Introduction
1.1 Shell structures
Shell structures are constructed systems described by three-dimensional curved
surfaces, in which one dimension is significantly smaller compared to the other
two. They are form-passive (relatively rigid) and resist external loads predominantly
through membrane stresses (Sigrid Adriaenssens and Williams 2014, Chapter 3).
Typically made of timber (Robeller et al. 2016), concrete or masonry (Block
et al. 2017). The unique membrane performance allows us to build efficient shells
even from very cheap and/or weak (with regards to bending) materials, such as
corrugated cardboard sheets, selected for this project.

1.2 Geometry
The challenge of shell design is to find the appropriate form for a given problem and
since shell structures are considered as one of the most complex structural systems,
it is very difficult to derive structural equations and sometimes impossible to solve
them manually. Due to the fact that the structural behaviour of shells can be so
complicated, the design methods of these complex structures strongly relies on
model prototypes and testing (Sigrid Adriaenssens and Williams 2014, Chapter 4).
Therefore, exclusive attention is given to the use of digital form-finding tools (Piker
2013; Rippmann et al. 2012) and physical modelling. Form-finding methods are

523
L. Vestarte, P. Vestartas, R. Kucinskas

the ones, where the structure itself defines its own shape based on its equilibrium
under applied loads. Unlike geometric forms, which are defined mathematically,
or free-form shapes, which are generated without taking into account structural
performance, form-found shapes rely on the structure and loads themselves for
definition, thus making such shapes inherently more structurally efficient.

2 Background
The 1:1 scale pavilion (fig. 1) is a result of a two-stage architectural workshop.
The main idea for the construction method originated from a well known simple
finger-joint technique, widely used from children’s toys and furniture, to architecture.
The teaching agenda for the workshop consisted of the following parts: introducing
historical examples of form-finding, modelling with digital form-finding tools, getting
acquainted with assembly sequencing and fabrication methods.

Figure 2: Students made physical models that were used for testing the form-finding methodology
and the physicals parameters of the joint-scale, structural-thickness, panelling-topology and
foundation.

In the course of the workshop, students were divided into groups to explore various
mesh topologies (triangular, quadrilateral, hexagonal) within the realm of form-
found double-curved shapes (fig. 2). Physical models were built as a two-layer
system, from 2 mm cardboard using a laser-cutter. Each student group designed and
fabricated supports, customised to suit their designs aesthetically and structurally.
These physical prototypes served as crucial study examples, providing information
regarding the dimensions and proportions for the design of the pavilion structure.
The initial studies also exposed some important fabrication issues such as defining
assembly sequence, solving insertion vectors per panel, and indexing, which was
essential for building a larger scale structure.

524
 Corrugated Cardboard Shell

During the second stage, the corrugated cardboard pavilion was built. This shell
structure was 3m height and had a span of 6 m in diameter and was supported
at 5 arbitrary points. The structure was built from two-layer 3 mm corrugated
cardboard sheets, consisting of 391 sets of planar quadrilateral elements and a total
of 1537 unique pieces. It was created using similar surface modelling methods as
the smaller scale experiments conducted by the students, but using quadrilateral
panels.

3 Methods: design phase

This project addressed following geometry processing methods as seen in fig. 3:
mesh discretisation, digital form-finding, data-sets ordering for laser-cutting, and
CNC fabrication and assembly. A dynamic-relaxation method was used for form-
finding a discrete shell (Piker 2013; Deuss et al. 2015). Polygonal mesh processing
tools were applied to: find a 2D topology-diagram, obtain planar elements, construct
connecting elements with finger joints, and prepare the geometry for fabrication
(Vestartas et al. 2020). A nesting tool OpenNest was used for orienting, packing,
and labelling of the elements for fabrication.

Material properties, thickness, Sheet dimensions, material Assembly sequence, minimal

structural directionality burn tolerance, cutting speed scaffolding, manual assembly,
DESIGN no connecting alloy

TOOLS: MATERIAL: FABRICATION: CONSTRUCTION

VISUAL PROGRAMMING CORRUGATED CARDBOARD CNC LASERCUTTER

“Rhinoceros 3D” “Grasshopper” “Kangaroo” “NGon” “Nest”

1. Creating a free-form 2. Generating structural 3. Digital form-finding 4. Polygonal mesh 5. Sorting and labeling 6. Orienting and
plan with supports topology pattern using physics engine faces the pieces nesting the pieces

Figure 3: The workflow diagram presenting digital geometry processing tools, material constraints,
2D-Axis fabrication, and assembly.

525
L. Vestarte, P. Vestartas, R. Kucinskas

3.1 Topology diagram

The first step in the design stage was to define two-dimensional outlines and
generate a planar topology pattern within the given boundaries. Various methods
for generating topology patterns were explored, when working with students on
several smaller-scale models during the prototyping phase (fig. 2). Quad dominant
block decomposition based on the medial-axis pruning/re-branching method (Oval
et al. 2017; Oval 2019) was selected for proceeding forward to a larger-scale
structure. Based on this methodology, the planar surface, represented by the
boundary curves, was divided into a set of topologically simpler patches. Each
patch was subdivided with quad pattern and then generated mesh was smoothed.
The applied form-finding method computes discrete meshes or line networks with
a list of geometry goals (Deuss et al. 2015). As a result, the selected topology
pattern, an input to the form-finding process, is essential, since it is directly related
to the obtainable geometries (fig. 4).

Figure 4: Design Workflow. (left to right) 2D topology diagram. Digital form-finding employing
"hanging cloth reversed" model. A projection method for Planar Panels. Connecting elements for
neighbouring panels.

3.2 Form-finding
The second part was to define boundary conditions and use form-finding. The
selection and manipulation of anchor points play a major role in shaping the output
geometry, thus being another significant constituent of the design scope. Digital
form-finding was employed for the initial design phase using a stability or equilibrium
(statics) approach, by simulating gravity load in the direction of Z-Axis. Students
were acquainted with such methods first through the introduction to a “hanging
cloth reversed” (i.e. fabric forming) model proposed by Heinz Isler, which proved
to be a beneficial strategy in helping students to understand the underlying logic of
such digital methods.

3.3 Projection method for planar panels

The planar mesh-faces are obtained by projecting their outlines to the polygonal
mesh-face planes as illustrated in fig. 5. The projection method is based on a mesh
where triangle mesh faces are grouped into quads, hexagons, or n-gons, represented

526
 Corrugated Cardboard Shell

by the RhinoCommon MeshNGon class. Triangle mesh-face properties such as faces

and vertex normals and vertex-edge-face adjacencies (VEF) are used to tessellate a
polygon internally. A triangle or quad, that is referenced in the polygon list, is no
longer visualised and conceived as a single entity, which takes part in the NGon
data-structure. Consequently, the normals and nine adjacency types (VEF) are
reconstructed following the notion of a polygon. For instance, the plane of the
NGon is computed by taking an average of the boundary-vertex coordinates and
the mean of the triangle-face normals (fig. 5c). The polygonal mesh is initialised
by triangulating an array of closed polylines using an ear-clipping algorithm (Eberly
2002), then joining and welding using the R-Tree search (Guttman 1984). The
applied algorithms are detailed in the source code1 . The projection function gets
the closest-point to a plane (fig. 5e) and reconstructs the planar polygon.

f4 v10 v11
f5 f3 f3 f3
v12
p4 p3 f0 f2 v9 f0 f2 v0 f0 f2
f1
f1 f5 f5
f1
p5 p2 v9
v8 f4 v1
f2 f2 f4
f3 f9 f3 f9
p0 p1 f3
f0 f2 v7 f6 f8 f6
v2 f8
f1 f7 v3
f7
v6 v5

A B C

D E F

Figure 5: Polygonal mesh-data structure: (a) Triangulation of a polygon, (b) ngon mesh vertex
normals , (c) mesh faces vertices and ngon planes. Projection method: (d) non-planar polygons ,
(e) average polygon planes, (f) planar polygons.

A second projection method could be used by intersecting polygon planes by bisector

planes of the mesh edges, but this would only work for 3 valence meshes (Rippman
et al. 2016; Pottmann et al. 2015). The geometry representation of the mesh-graph
is no longer connected graphically, but the connectivity-graph is continued to be
used from the non-projected mesh to construct joints, offset the mesh, and index
the cardboard panels for the fabrication.

1
Ear-clipping method: TriangulateOpenPolyline in MeshUtilSimple class and R-Tree Welding
method: WeldUsingRTree in MeshUtil class (https://github.com/petrasvestartas/NGon)

527
L. Vestarte, P. Vestartas, R. Kucinskas

3.4 A two-layer structure

The preliminary tests of the cardboard structure were made using one layer hexagonal
panels and diagonal connecting elements as seen in fig. 6. After inspecting the
single-layer test model, a two-layer structure was introduced to obtain greater
stiffness and add resistance to twisting. The two-layer system was generated by
offsetting polygonal meshes using their normals by the distance of their structural
depth, taking into account the thickness of the sheet material (fig. 7).

Figure 6: One layer cardboard model.

A B C

Figure 7: Mesh offset: (a) using vertex normals, (b) two layers of the projected mesh , (c) four
projections considering the material thickness.

Several small scale models were made to check the right distance between the
layers, taking material properties into account and the ease of the assembly. For
the 1:1 pavilion structure, the distance of 9 cm between the two layers was selected,
making the full structural depth (including the height of the connecting elements) of
15 cm. The experimental models were relatively small in scale and due to material
elasticity, the assembly was possible without having defined assembly sequence and
insertion direction. However, the issue of damaging the pieces during the insertion
due to bending was exposed. Thus the larger structure needed the insertion vector
and assembly sequence sorted. The equal distance offset was taken for further
development as a straightforward solution, well suited for this particular project.

528
 Corrugated Cardboard Shell

3.5 Diagonal connections

The diagonal connection elements with simple finger-joints were chosen to connect
the pairs of panels in a double layer system, creating a closed cross-section. Several
experiments were made to test the direction of the connection element. A connection
perpendicular to the mesh edge performed well in transferring the loads but resulted
in a greater risk of damaging the panels during the construction. The insertion
direction was solved, finding one approximated vector per one set of panels (i.e.
inserting both layers of the same enumeration at the same time). The sequence
of the assembly, and thus the resulting insertion vector, could be specified either
manually drawing a path (fig. 8b) on mesh-face centres, or using graph methods
such as Breadth-First-Search as illustrated in fig. 9 and the source code2 .

A B C

D E F
Figure 8: Insertion order following (a) perpendicular to edge (b) user specified insertion sequence
(c) diagonal connections following the insertion order. (d-e-f) assembly sequence order showing
the orientation of connections following one direction per set of panels.

The modelling of the connection pieces was based on a mesh edge and face adjacency.
Each joint was placed on a planar mesh-edge, connecting the two neighbouring
panels. A mean value of the two lines was taken since the panels had shifted slightly
after the application of the projection method. The orientation of the connection
element depends on the mesh edge direction and the sum of normal vectors of the
adjacent mesh faces. In order to identify the orientation of the connection element
during the construction, a design decision to shorten the connecting part outside
2
Breadth-First-Search method for a Mesh: MeshBFS in UndirectedGraphBfsRhino class
(https://github.com/petrasvestartas/NGon)

529
L. Vestarte, P. Vestartas, R. Kucinskas

the double-layer structure was made. A series of tests were performed with smaller
prototypes as well as a cluster test conducted in the original scale and material.
This allowed defining the suitable proportions and dimensions for these elements.
For the pavilion structure, the width of the interior parts of the connection pieces
are up to 20cm, the outer part is 12cm and the total height is 15 cm. The larger
outline arches on the edges were stiffened with a double diagonal connection and
lengthened interior parts (fig. 10a).

Figure 9: Insertion order following (a) Breadth-First-Search graph method (b) finding insertion
vector (c) physical test model.

A B C

Figure 10: Details of finger joints (a) doubling of connections for the out arch (b) elongated
joinery (c) foundation part with two and three layers of panels.

3.6 Finger joints

The traditional finger joint method used for such structure results in an angular cut
through the material thickness. This is the case because of the inherent properties
of double-curved surface geometry (fig. 11). Due to the main constraints being
a simple laser-cutter and a thin sheet corrugated cardboard (3mm) material, an
approximation for a two-dimensional cut was needed. Although performing best in
compression, with well-managed friction within the joint, the traditional finger-joint
can also resist comparatively small amounts of tension. In order to exploit the
potential of friction within the joint, the offset tolerance of both the depth and
width of the finger joint had to be defined after thorough physical trials. The

530
 Corrugated Cardboard Shell

most critical aspect during such testing was a precise definition of an actual burnt
parameter for a specific material, using specific laser-cutting settings. This was
done to find a balance between ensuring fast and easy insertion, while also providing
enough friction for stabilizing the structure during the construction process.

Figure 11: View of the finger joints in the assembled pavilion model, showing one connection
element per edge for inner panels and two elongated connections for the boundary edges.

4 Methods: fabrication
4.1 Clustering
During the initial prototyping phase, the panels with the associated diagonal
connection elements were grouped into a series of clusters, to ease the fabrication,
sorting, and assembly processes. This was done due to the management issues
of large numbers of unique pieces. It also helped to divide the work among
the groups of students, when clusters were assembled in a parallel mode, and
afterwards connected. A K-Means clustering algorithm (Accord.NET) was used
to partition elements by cartesian coordinates. This technique reflects on the
cluster-seam connections because large groups were joined together using multiple
joints at once. Such rapid connection was possible because of the relatively small
scale of the physical prototypes. Within the context of a larger scale structure,
grouping strategies were used for creating sets of panels grouped with the associated
connecting elements. These small sets, consisting of only several items, were later
used for speeding the fabrication, manual sorting process, and the assembly.

4.2 Orienting and nesting

The panels and connecting elements contained geometrical information of cutting
outlines, planes, and engraving curves for indexing. The plane of the element was
used to orient the properties of the panels from 3D space to 2D XY plane. The
central cutting outline is approximated from 5-axis angles in the 3d model to 2d-axis
laser-cutting. A nesting algorithm was applied for the small scale prototypes. The

531
L. Vestarte, P. Vestartas, R. Kucinskas

packing method kept track of the transformation matrices from XY-plane to the
nested part to orient additional data such as engraving polylines. The engraving
was performed to track the data of the neighbour faces on each edge, first and
second layer index, the panel id and assembly markers. The small scale models were
cut from several sheets of cardboard with many pieces nested together. The digital
nesting method performed fast (less than a second) for a large number of elements
running only one iteration of packing. With regards to the 1:1 pavilion structure,
there was a significantly smaller number of elements per sheet to be nested, the
genetic algorithm performed less precisely, otherwise the time to compute increased
exponentially.

The material used to build the 1:1 scale pavilion, was a single wall corrugated board,
consisting of a fluted corrugated sheet and two flat liner-boards. The geometrical
proportions of panels had to be restrained due to the material properties, for
instance, elongated pieces and sharp corners had to be avoided. The directionality
of corrugation also strongly affects panel performance under stress in different
directions. After physical testing, the decision was made to mix the directionality
among the panels (fig. 12), ensuring that the same set of facets would have a
varying corrugation direction among the inner and outer layer of the structure. This
was achieved through controlled nesting of specific panels within the sheet with
the intentionally set direction of corrugation. Single corrugation direction, the best
performing in compression, was chosen for fabricating diagonal connection elements.
The fabrication constraint regarding the laser-cutter bed with the dimensions of
600 mm x 900 mm, significantly reduced the number of panels per sheet for the
pavilion structure, compared to the smaller prototype models. This resulted in
extended fabrication time but was advantageous in speeding the process of sorting
and packing the pieces.

Figure 12: The orientation of elements considering the direction of a cardboard corrugation. The
first layer panels of a two-layer system were rotated ninety degrees in relation to the second layer.

4.3 Sorting and labelling

Due to a large number of unique, but very similar pieces, the labelling was necessary
for the assembly of both the test models and the pavilion structure. The latter
consisted of 1537 pieces in total. During the construction of the pavilion, labelling

532
 Corrugated Cardboard Shell

also served for sequencing the assembly. The enumeration of the panel, according to
the assembly sequence, was engraved in the centre of the panel and the orientation of
the text followed the insertion direction (fig. 13). Each panel also had identification
for the inner or outer layer and numbers naming the neighbouring panels, with
regards to each edge. The diagonal connection element had a label marking
the neighbouring panels that it connects. The fabrication of pieces followed the
order of the enumeration. Both panels for inner and outer layers as well as
connecting elements of a consecutive order were grouped for fabrication. These
clusters, consisting of two panels and their associated connecting elements, were
packed together for safe preservation and fast access during the assembly (fig. 14).
Thorough labelling provided all the necessary information for the assembly, within
the pieces themselves, ensuring fast and precise on-site construction. For aesthetic
purposes, the labelling was hidden inside the structure, between the layers.

Figure 13: (a) A set of elements (two panels and connecting elements) and (b) the preparation
for the assembly.

Figure 14: (a) Grouping the elements two panels and their connecting elements (b) packed
pieces prepared for the on-site assembly .

533
L. Vestarte, P. Vestartas, R. Kucinskas

5 Methods: construction
5.1 Assembly sequence
The assembly sequence directly corresponded to the enumeration order of the panels
and consequently, the insertion direction. A linear path was chosen to connect
the components starting from the ground and spiralling towards the top of the
shell. The possibility to split the sequence and work parallel in several clusters was
made possible when still maintaining the sequential order: starting from a larger
enumeration number and adding the next set, marked with the successive smaller
sequence number.

5.2 Assembly process

The base and the five supports were custom designed and fabricated using a 3-axis
CNC machine from 18mm MDF panels. First, the base and the supports were
put in place. Second, following the sequential labelling order, the five stiffened
outline arches on the edges, that convey the loads to the supports, were assembled.
Then, maintaining the assembly sequence, the structure was assembled in several
clusters in parallel, merging in one assembly path towards the final pieces. The
structural balance was secured during the construction using temporary supporting
posts, but no need for a heavy scaffolding arose. This was due to the resistance to
tension provided by the friction within the joint, allowing the possibility of small
temporary cantilevers and compensating any possible instabilities occurring during
the construction. The assembly process was fast and precise, the whole structure
was built within less than 12 hours by a dozen students (fig. 15).

6 Results and reflections

The resulting pavilion project (fig. 16 and 17) provided some fruitful insights con-
sidering possible future investigations, regarding the assembly aspects in particular.
The direction of connection elements that followed one insertion vector per one
item (two panels and their connections) fostered some major challenges. The
presented study was based on extensive physical prototyping and testing fabrication
tolerances, thus it could be highly beneficial to use computational optimisation
strategies to find the best-balanced solutions taking into account the assembly
sequence, insertion vector and the distribution of loads within the shell structure.
An additional aspect worth investigating further could be the design of the assembly
sequence, employing certain clustering algorithms while taking into consideration
off-site construction and speeding up the construction process in general. This
approach would require a different type of joints for connecting multiple elements
across each cluster.

534
 Corrugated Cardboard Shell

A B

C D

E F

Figure 15: The assembly process (a) started from the base with five foundation elements (b)
then the largest boundary arch was assembled (c) which required temporary punctual supports.
The assembly (d) followed a spiralling insertion order (e) with multiple students assembling the
proposed cardboard pavilion (f) .

Figure 16: Different views of the pavilion project in the exposition space.

535
L. Vestarte, P. Vestartas, R. Kucinskas

During this study, students were introduced to digital design strategies and various
visual programming tools with more explicit emphasis on physics simulation tool for
form-finding and mesh-processing. Next to digital modelling, physical prototyping
leads to recognizing and analysing subtle nuances in precision and tolerance, which
helps to understand crucial distinctions between the digital models and the physical
ones. Although the experimentation phase was a great learning opportunity, much
appreciated by the students, having some "hands-on" building experience with a
larger structure proved to be a better lesson. Constraints of fabrication, logistics,
and material properties amplified by scale, indisputably proved the importance of
incorporated structural thinking and the act of construction, in the preliminary
design phase for such architectural structures as shells.

156 SUPPORT
VIE 0 1m 2m 05
W_ VIEW 0 1m 2m
02 _ 02
600

GROUND FLOOR PLAN TOP VIEW

6

SUPPORT
15

SUPPORT
04 05 SUPPORT
60 04
0

Ø6.00m Ø6.00m
_ 03

_ 03
VIEW

VIEW
600

0
60 SUPPORT
03
SUPPORT
15

03
6

156
60
0

SUPPORT
6
VIE

15 02
VIE

SUPPORT
W

02
_
_

01
01

SUPPORT
SUPPORT 01
01

3.20 3.20

2.75 2.75

2.30

1.85

1.65

1.25

0.85

0.58 0.60

±0.00 ±0.00

VIEW_ 01 VIEW_ 02

Figure 17: Ground floor, top view and elevation drawings.

7 Conclusion
This paper discussed design and fabrication strategies for polygonal shell-structures
using digital form-finding. It presented a study case of a quad-dominant corrugated
cardboard shell as a result of an extended workshop conducted with architecture
students. The pavilion was demonstrated in the indoor art exhibition to showcase the
studio-teaching. The structure was designed to suit the given material - corrugated
cardboard and the fabrication method - 2D-axis laser cutter, making these two
constraints as the main design scope constituents.

536
 Corrugated Cardboard Shell

Acknowledgements
We are grateful to The Vilnius Academy of Arts (VAA), Department of Architecture,
where the workshop took place in 2019, and to all the students who participated:
Agne Avulyte, Evelina Azuolaityte, Zivile Bakutyte, Vilte Bernataviciute, Raminta
Bieliauske, Agne Brazinskaite, Solveiga Buozelyte, Zivile Grigaleviciute, Rugile Jon-
aityte, Migle Karaskaite, Liucija Kuchalskyte, Jovita Laurukenaite, Kseniya Liadnik,
Karolina Masalskaite, Emilija Minikaviciute, Alisa Muchks, Simonas Pinkoraitis,
Kiril Smotkin, Ieva Sinickaite, Vytenis Stasiunas, Emilija Staskute, Justina Tautvy-
daite, Algimantas Urbonas, Rokas Vainiunas, Milda Venslaviciute, Ula Zebrauskaite,
Kasparas Ziliukas.

References
Block, P., T. Van Mele, M. Rippmann, and N. Paulson (2017, October). Beyond
Bending - Reimagining Compression Shells. Munich: Edition DETAIL.

Deuss, M., A. H. Deleuran, S. Bouaziz, B. Deng, D. Piker, and M. Pauly (2015).

ShapeOp—A Robust and Extensible Geometric Modelling Paradigm. In Modelling
Behaviour.

Eberly, D. (2002). Triangulation by ear clipping. Magic Software, Inc.

Guttman, A. (1984). R-trees: A dynamic index structure for spatial searching. In

Proceedings of the ACM SIGMOD International Conference on Management of
Data.

Oval, R. (2019). Topology Finding of Patterns for Structural Design. Ph. D. thesis,
University Paris Est, École doctorale Sciences, Ingénierie et Environnement.

Oval, R., M. Rippmann, T. V. A. N. Mele, O. Baverel, and P. Block (2017). Patterns

for Masonry Vault Design. In Proceedings of the IASS Annual Symposium 2017
- "Interfaces: architecture.engineering.science".

Piker, D. (2013). Kangaroo: Form finding with computational physics. In Architec-

tural Design. Wiley.

Pottmann, H., C. Jiang, M. Hobinger, J. Wang, P. Bompas, and J. Wallner (2015).

Cell packing structures. CAD Computer Aided Design.

Rippman, M., T. Van Mele, M. Popescu, E. Augustynowicz, T. Méndez Echenagu-

cia, C. Calvo Barentin, U. Frick, and P. Block (2016). The Armadillo Vault:
Computational design and digital fabrication of a freeform stone shell. In Advances
in Architectural Geometry.

537
L. Vestarte, P. Vestartas, R. Kucinskas

Rippmann, M., L. Lachauer, and P. Block (2012, December). Rhinovault - interac-

tive vault design. International Journal of Space Structures 27 (4), 219–230.

Robeller, C., M. Konakovic, M. Dedijer, M. Pauly, and Y. Weinand (2016). A

Double-Layered Timber Plate Shell: Computational Methods for Assembly,
Prefabrication, and Structural Design. Advances in Architectural Geometry 2016 .

Sigrid Adriaenssens, Philippe Block, D. V. and C. Williams (2014). Shell Structures

for Architecture. Routledge.

Vestartas, P., N. Rogeau, J. Gamerro, and Y. Weinand (2020). Modelling Workflow

for Segmented Timber Shells Using Wood-Wood Connections. In Impact: Design
With All Senses.

538