Full Terms & Conditions of access and use can be found at
Architectural Engineering and Design Management
ISSN: 1745-2007 (Print) 1752-7589 (Online) Journal homepage: https://www.tandfonline.com/loi/taem20
Computational knit – design and fabrication
systems for textile structures with customised and
graded CNC knitted fabrics
Martin Tamke, Yuliya Sinke Baranovskaya, Filipa Monteiro, Julian Lienhard,
Riccardo La Magna & Mette Ramsgaard Thomsen
To cite this article: Martin Tamke, Yuliya Sinke Baranovskaya, Filipa Monteiro, Julian Lienhard,
Riccardo La Magna & Mette Ramsgaard Thomsen (2020): Computational knit – design and
fabrication systems for textile structures with customised and graded CNC knitted fabrics,
Architectural Engineering and Design Management, DOI: 10.1080/17452007.2020.1747386
To link to this article: https://doi.org/10.1080/17452007.2020.1747386
Published online: 04 May 2020.
Submit your article to this journal
View related articles
View Crossmark data
Computational knit –design and fabrication systems for textile
structures with customised and graded CNC knitted fabrics
, Yuliya Sinke Baranovskaya
, Filipa Monteiro
, Julian Lienhard
Riccardo La Magna
and Mette Ramsgaard Thomsen
CITA, Copenhagen, Denmark;
A. Ferreira & Filhos SA, Caldas de Vizela, Portugal;
Structure GmbH, Berlin, Germany
In this paper, we are presenting a design to fabrication system, which
allows to produce eﬃciently and highly automated customised knitted
textile elements for architectural application on industrial computer-
controlled knitting machines (Computer Numerical Control (CNC)
knitting machines). These textile elements can, in this way, be individual
in both geometry, detailing and material behaviour. This work extends
recent work on CNC knitted tensile members and presents a set of
innovations in design and manufacturing, which together allow to build
structural systems, in which highly individualised membrane members
allow a structure to take on multiple structural states. Underlying these
innovations is a shift from the focus on geometry and homogeneity in
material and behaviour, expressed in current state-of-the-art membrane
structures and materials. Instead our research lays the foundation for a
new class of membrane materials with varying bespoke local material
properties. In this paper we present the underlying digital tools and
processes for design, analysis and manufacturing of these hyper
speciﬁed textile membranes. We showcase and evaluate the potentials
of Computational Knit for novel structural membrane systems through
the large-scale installation Isoropia designed and built for the Danish
Pavilion in the 2018 Venice Architectural Biennale.
Received 13 November 2019
Accepted 20 March 2020
Architecture; digital design;
bending active textile
membrane hybrids; digital
Chain –integration of design;
simulation and fabrication;
CNC knit; computational knit
The current state-of-the-art materials in the membrane architecture aim to be homogenous, stiﬀand
show little stretch. They are usually based on weave and produced in large-scale industrial production
using specialised looms. The research presented in this paper investigates a new material paradigm,
where a consistent use of computation in all steps from design to production, provides the ability to
specify and produce geometry and material properties on a scale smaller than the membrane
pattern. We use the CNC knit as a means of production for these membranes, as it allows for the spe-
ciﬁcation of material properties through the variation of knit structure and ﬁbre on a loop scale. The
resulting membrane material is stretchable though the use of high-performance ﬁbres provides it
with a relatively high strength (Tamke et al., 2016).
Our research is a simultaneous inquiry into the design and analysis tools and processes for the
conception, speciﬁcation and production of membrane structures made from the CNC knit, as well
as a speculative inquiry into the new spatial and structural possibilities that come with the new
material. The inquiry takes place through built demonstrators, namely Hybrid Tower (Thomsen et
al., 2015), Tower Guimaraes (Tamke et al., 2016) and Isoropia for the Danish Pavilion at the
© 2020 Informa UK Limited, trading as Taylor & Francis Group
CONTACT Martin Tamke firstname.lastname@example.org CITA, Copenhagen 1435, Denmark
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT
Architectural Biennale 2018 (Thomsen et al., 2019). A special structural, as well as aesthetic and spatial
focus, has been set on the strategic use and control of the stretch of the material. While Tower
(Figure 1) uses the stretch to brace the structure and create a tantalising interior of the Structure
(Figure 1), Isoropia develops an integral system of bending-active elements with purely tensile
elements, where the necessary shape of the membranes is solely achieved through the stretch of
the knitted material (Figure 2).
In Isoropia we achieved furthermore, for the ﬁrst time, a full mass customisation and automatisa-
tion of the design pipeline of all knitted elements which are all diﬀerentiated in shape, knit structure
and detailing. With the developed process we achieve a suﬃcient precision, structural stability and
longevity (Tamke et al., 2019).
The use of CNC knit as architectural material is still in its infancy and many questions, regarding the
design and simulation of graded CNC knitted membranes as well as modes for the analysis of the
material, (Thomsen & Tamke, 2016) are still open. We want to give, in this paper, an account of
the state of our research on the basis of Isoropia.
We will at ﬁrst position the research in the perspective of computation in architecture and
the related state of the art in architectural CNC knit and thereafter –with a focus on the
digital process and technologies, introduce our developments in terms of structure and analysis,
design-, material and fabrication system. We show how this approach steps beyond the state of
the art and how it creates avenues for new spatial experiences in textile architecture, share how
strategic prototyping solved the still existing challenge, that the behaviour of knitted fabrics can
not be currently reliably measured and described for simulation, and identify further research
Contextualisation of the research
Tensile membrane structures on an architectural scale are per se customised, as the shape of the
speciﬁc membrane can be determined only through the process of form ﬁnding. As a result, building
scale tensile membrane structures is usually one-oﬀ, they are designed and engineered speciﬁcally to
the context, they are situated with equally speciﬁc membrane patterns and manufactured in predo-
minantly manual processes (Knippers, Cremers, Gabler, & Lienhard, 2011). Designers, engineers and
manufacturers try to increase the eﬃciency of the production through rationalisation of the mem-
brane panels, e.g. through a repeat of the same or similar elements.
In contrast, in the architectural ﬁelds outside of membrane structures the use of non-standard
design and fabrication methods is today common. The success of these designs is based on the
implementation of computational techniques in design and analysis and on the integration of
digital design and digital fabrication on element and as well material level (Tamke & Thomsen, 2009).
Figure 1. Interior of Hybrid Tower Guimaraes (left image). Isoropia –a bespoke textile installation for the Danish Pavilion at the
Architectural Biennale 2018 (right image).
2M. TAMKE ET AL.
These digital design methods and workﬂows have enabled new spatial and tectonic solutions
across all scales (Figure 3) in which elements change and morph in order to change the atmosphere
and expression for the humans within. The underlying computational approaches provide further-
more the base for future highly material-eﬃcient structures, which rely on the total integration of
design and robotic manufacturing (Nicholas, Zwierzycki, Nørgaard Clausen, Hutchinson, &
Thomsen, 2017; Solly, Frueh, Saﬀarian, Prado, & Menges, 2018).
In the ﬁeld of tensile architecture the use of computational systems for non-standard approaches
has been tested early on in small-scale academic prototypes, such as the 2007 AA Component Mem-
brane installation and related work (Hensel & Menges, 2008). However, similar site and human scale-
speciﬁc installations, based on the morph of the size and shape of membranes across the structure,
have, until now, not seen wide implementation in the ﬁeld. Instead, discrete and varying membrane
fabric elements of same size are repeated, as in the case of the King Fahad National Library ﬁnished
building in 2014 (Dupont, 2014), or an intended morphing expression is created through a combi-
nation of a customised steel structure and standard membrane elements, as in the Nizhny Novgorod
Stadium ﬁnished building in 2018 (Bernert, 2018).
While the authors of the latter structure do not reveal the reason why their design shifted from a
non-standard approach towards the pattern cut of the membrane elements to a standardised one,
accounts from small-scale structures with non-standard approaches to geometry and detailing
(Hensel & Menges, 2008), point at the labour intensity in design, simulation, fabrication and assembly
Figure 2. Basic structural principle of Isoropia, demonstrated in a single element. A –bending compression element (GFRP rods), B
–Tension element (Rope), C –Circular interface between the tension rope and the textile to keep the structure in a shape.
Figure 3. Interior Panels in Oslo Opera Hall (OlafurEliasson), It’s A small World (CITA), Arch Union Architects (Dongdaemun), Design
Plaza (ZahaHadid architects), Aqua Tower (Studio Gang) –all photos CC.
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 3
as a challenge. The inherent stiﬀness of the traditional laminated membranes limits furthermore the
range of scales in which it can be used cost eﬃciently, as smaller sized membranes still need a similar
amount of seams (to assemble patches and to achieve double curvature) and detailing, which
increases relative costs.
New opportunities for eﬃcient methods to create non-standard logics for new spatial and tactile
experiences emerge through the integration of design, analysis and fabrication combined with the
shift of the textile system from weave to CNC knit (Ahlquist & Menges, 2013; Popescu et al., 2018;
Sabin, 2013; Thomsen et al., 2015). The inherent ﬂexibility of knit, the ability to integrate shaping
and detailing in the textile fabrication process open opportunities especially for designing mem-
branes at the smaller scale of state-of-the-art membrane architecture, as in Ron Herrons Imagination
Building (Lyall, 1992). So far applications of CNC knitting in the ﬁeld have focused on small scale pro-
totypes. They did however not include building scale objects or devised knitted textiles a role as
structural member. This is however necessary, when constructing Form Active Hybrid Structures of
Active Bend and Tensile members (Thomsen et al., 2015).
Isoropia –Outset and framework
Isoropia, which in Greek means balance, equilibrium and stability, is a 35 m long structure made from
41 custom CNC knitted patches of up to 7 m length for the 2018 Venice Biennale Installation. The
textile membrane is set in structural equilibrium with bend glass ﬁbre rods of varying thickness
and strength. Detailed questions of digital design and analysis workﬂows in Isoropia have been dis-
cussed in a previous publication (La Magna, Fragkia, Noël, Baranovskaya, & Thomsen, 2018).
The structure creates a spatial and structural continuum through the Danish Pavilion, forming
diﬀerentiated outdoor canopy structures on the two outer sides and a vaulted space in the interior
in a reaction to the speciﬁc programme and sites (Figure 4).
Further forming requirements have been:
.the status of the Danish Pavilion as a quasi-listed building, which prohibits irreversible modiﬁ-
cations to the construction
.the need to connect the structure to ‘Danish ground’only, which prohibited any ties to the directly
surrounding ‘Italian’ground and the large amount of ca. 150,000 visitors, which required a sturdy
Figure 4. The Architectural concept of Isoropia. (1) The southern exterior part (left) welcomes visitors to the Biennale and creates a
canopy-like structure guiding visitors into the Danish Pavilion. (2) The interior passage (middle) is the only entrance to the Danish
Pavilion. The aim was to create a dense textile space, which creates curiosity on the side of the visitors, allows them to slow down,
study movies and text about the installation and ﬁnally redirects them into the further exhibition of the pavilion. (3) The northern
exterior part (right) is directed towards the lively cafe zone of the Biennale and creates a shading entrance canopy, which adapts to
the rhythm of the colonnade of the existing Danish Pavilion.
4M. TAMKE ET AL.
construction and detailing in compliance with a structural and historical conservation code that
had to be documented in a building permission to the authorities
.the pavilion had to work at day and night, which required the integration of artiﬁcial lighting.
Most importantly, the time given from commissioning of the installation to its completion was
only 4.5 months.
The conceptual answer to the design challenge was to create a structure, which can morph and adopt
between the adjacent spaces and as the structure is in arm-length of the visitors to set a focus on a
high level and quality of textile surfaces and detail. Our developed solutions are highly intercon-
nected in terms of the processes in design, analysis and fabrication, and they operate simultaneously
on several scales (Figure 5).
Structural system –bending active textile Hybrid
The structural system employed for the canopy falls into the category of bending-active textile
hybrids which work on the interdependence of form and force of the mechanically prestressed
textile membranes and bending-active ﬁbre-reinforced polymers (Lienhard & Knippers, 2015). Such
systems achieve equilibrium through the combination of bending and tension. Figure 6 shows
how the GFRP masts are elastically bent and locked into their curved shape the position tension
cable and pre-stressed textile. The reciprocal dependency of the bending-active elements pre-stres-
sing the tensile elements, which, in turn, lock the bent rods into position, turns the system into a
highly versatile hybrid where the shape of the canopy is controlled by pre-stress ratios and the
cutting pattern of the knit. Additionally, the system of the Isoropia changes with its boundary con-
ditions: the cable net systems on the exterior work as cantilevering arches, while on the interior
the arches are locked into shape by pressing against the adjacent walls and ceilings where three
dimensionality is given to the textile with a tensegrity-like system of internal compression rods
The outdoor areas derive their shape from the mutual force interaction between the bent rods,
tensile membrane and cables. The dedicated steel ﬁxtures on the buildings’wall are the only
Figure 5. The three interdependent scales of structure, element, material.
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 5
support. The pultruded GFRP rods are changing in diameter and wall thickness (22/17mm, 24.3/
20.3mm and 26/19mm), bespoke to local performance requirements. The tensile elements, both
cables and knitted fabric, are made of Dyneema® because of its relatively high stiﬀness and UV stab-
ility. Since the biaxial stiﬀness behaviour of knitted fabrics is highly dependent on the nature of the
knit and changing with pre-stress levels, the knit stiﬀness-dependent behaviour of the hybrid is very
diﬃcult to include in the computational form-ﬁnding and analysis models. Therefore, a very stiﬀﬁbre
with a knit that is pre-stressed to its locking point was chosen to enable a relatively accurate form-
ﬁnding and reliable structural model for the wind analysis and design criteria according to Eurocode
(EN 1990, 2002; EN 1991-1-1, 2002) in combination with the current work-in-progress Prospect for the
Structural Design of Tensile Membrane Structures (Stranghöner et al., 2016).
Having managed to lessen most of the tension and bending forces within the textile hybrid
system helped to reduce the interventions on the surroundings to the bare minimum. This was essen-
tial due to the listed building status of the Danish pavilion, which pushed for solutions that minimise
the support positions used to secure the structure.
In order to be able to design with the highly interdependent hybrid system, we developed a dedi-
cated design system (Gengnagel, La Magna, Ramsgaard Thomsen, & Tamke, 2018) that seamlessly
connected the digital design pipeline used for geometrical exploration, based on Projection
Dynamics (Bouaziz, Martin, Liu, Kavan, & Pauly, 2014), with an intermediate Isogeometric Analysis
tool (Längst, Bauer, La Magna, & Lienhard, 2018), which provided frequent feedback on the structural
performance of the canopy and ﬁnally a robust and detailed analysis using more established tools for
simulation (Julian Lienhard, Bergmann, La Magna, & Runberger, 2017). In this way, the behaviour of
the structure could be constantly monitored during the development of the project, providing valu-
able information to all the parties involved throughout the conceptual and development phases.
Digital design workﬂow
The developed continuous membrane system is based on a single principle structural unit, consisting
of a pair of bend GFRP rods (Beams), which carry a connecting membrane, stressed by cable nets or
compression sticks. Through variations of parameters, in these basic units, such as the width between
Figure 6. Development of the structural system: (a) mast with a clamped support, (b) reducing the clamping stress through a pair
of support forces, (c) bending the mast, (d) locking the bent shape with a tension chord, (e) tie back the chord to the bent mast, (f)
replacing the 1D tie backs with knitted fabric.
6M. TAMKE ET AL.
the supporting rods, their lengths and the interposition of the fabrics in relation to the rods, a con-
tinuous yet adapting structure is possible.
In order to design with the few, yet interconnected parameters, we set up a generative design pipe-
line, where at ﬁrst, based on the geometrical input, a static geometry is generated using Grasshopper.
Subsequently this providesthe input to the particle position-based dynamic relaxation solver from Kan-
garoo2 (K2) (Quinn et al., 2016)(Figure 7). Based on our more detailed Tensinet Symposium paper
(Thomsen et al., 2019) we outline the pipeline below using the terminology from the software tools.
At ﬁrst, points for beams allocation were manually distributed, becoming an input for the auto-
matic generation of beams in a predeﬁned length at deﬁned z-coordinates, as well as for the auto-
matic generation of mesh quads, representing membrane elements (Figure 7(A –C)). Each textile
segment in Isoropia is deﬁned to be of a particular design, based on its functionality and the intended
geometrical eﬀect –either a single-layered cable net-based design or double-layered tensegrity-like
design. Therefore, interior segments receive an additional layer. All the textile quads get discretised in
custom resolutions and large openings are placed in the interior lower membranes (Figure 7(D)). For
the outdoor cable net-restrained patches, the amount of contact points between the rope and the
fabric is parametrically deﬁned (Figure 7(E)). Compression sticks are introduced and iteratively opti-
mised in order to ﬁnd an equilibrium position in between the layers of the interior textiles (Figure 7
(F)). External constraints for the dynamic relaxation with the K2 Solver are introduced in the model:
anchors, external tension ropes, links to the ceiling in the interior and back tension ropes for the non-
cable net units. Anchors are set up as points in 3d with ﬁxed positions. External ropes, which link the
installation to the anchors, are modelled as springs with a certain stiﬀness value (Figure 7(G)). Once
the entire pre-relaxation model is generated, the aforementioned K2 solver is used to calculate the
resulting abstract force vectors within the K2 model in each node of the system and to move the
nodes iteratively until the abstract force equilibrium is found. This represents the ﬁnal form-found
design (Figure 7(H)).
The digital design tool provided an agile platform for quick design explorations and geometrical vari-
ations of the canopy throughout its conception and development. Due to the high level of documen-
tation requested by the authorities, a detailed analysis of the structure was necessary in order to
assess its performance under high wind loads and its eﬀects on the surroundings, especially the reac-
tion forces exerted by the canopy on the support areas ﬁxed to the building. The speciﬁc wind load
for the location was derived from the Italian NTC2008 code (NormeTecniche per le Costruzioni),
which prescribes a basis wind velocity of 25 m/s. According to the topography and exposition of
the area, this corresponded to a peak velocity pressure of 0.44 kN/m
. The pressure coeﬃcients
were corrected with reference to DIN EN 13782–184.108.40.206 (Design of temporary structures), which
specify a 0.7c
value for wind suction and 0.3 c
value for wind pressure. In turn, this corresponded
to characteristic wind pressures of 0.13 kN/m
for wind suction and 0.31 kN/m
for wind pressure.
Finally, a dedicated workﬂow between design and analysis was set up to quickly provide feedback
to the design process.
Kiwi3d, a new tool for Isogeometric Analysis, was used for intermediate quick analysis. In particu-
lar, Kiwi3d incorporates modules for linear and non-linear analysis, as well as form-ﬁnding based on
the URS (Updated Reference Strategy) method (Philipp, Breitenberger, D’Auria, Wüchner, & Bletzin-
ger, 2016). To speed up the transfer between the two platforms, a geometry processing workﬂow
was set up which took care of the discrepancies between the geometric models. This meant convert-
ing discrete lines and surfaces into continuous spline and NURBS patches through interpolation of
the nodes. This geometry conversion was robust and reliable, and an analysis of the full building
process of the canopy could be simulated in three main steps: 1. Bending of the GFRP rods (geometri-
cal non-linear analysis); 2. Attaching and form-ﬁnding of the membrane pattern (form-ﬁnding);
3. Linking the cables to the membrane and pre-stressing them (form-ﬁnding). Due to their slender-
ness, the GFRP rods were modelled using a Bernoulli beam formulation. As a consequence of the
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 7
temporary nature of the structure, all material non-linearities were disregarded. Besides the complete
form-ﬁnding and simulation of the building process, an initial assessment of the structural behaviour
under wind loading (930 mm vertical deformation for wind suction, 640 mm vertical deformation for
wind pressure) and the corresponding reaction forces took place in Kiwi3D.
Figure 7. Steps of the workﬂow of the Digital Design Tool.
8M. TAMKE ET AL.
The analysis of the canopy was completed by running a Finite Element simulation on the ﬁnal
design (SOFiSTiK coupled with the dedicated Grasshopper plugin STiKbug) (Lienhard, La Magna, &
Knippers, 2014). This step was necessary to validate the intermediate results using well-established
tools, which have been extensively tested both in research and in practice. The Finite Element
tools allow for a very detailed description and simulation of the structural behaviour, giving the
analyst the possibility to incorporate advanced aspects such as long-term behaviour and plasticity.
Speciﬁc to this case, compressive springs were added in the ﬁnal simulation model to take into
account the contact between the bending-active rods and the building’s walls, an aspect that
needed to be veriﬁed due to the scrupulous requirements of the organisation.
In Isoropia, we further develop the inter-scalar approach ﬁrst suggested in the Hybrid Tower projects
(Holden Deleuran et al., 2015; Thomsen et al., 2015). Here, design takes place at multiple scales from
the overall structural system, to the single patch, the implemented knit structure and down to ﬁbre
selection and ﬁbre surface.
Knitted membrane: an inter-scale approach
Contrary to Hybrid Tower, in which a single patch design is repeated to achieve the rotational geo-
metry, Isoropia works with non-standard patches. As the structure morphs from canopy to vault and
back again and as it twists through the Danish Pavilion building, each patch is diﬀerentiated in size,
shape and relative surface detailing. This diﬀerentiation creates variances in the patch design from
the variation of protruding cones and slits (Figure 8).
Isoropia also further develops the membrane design from Hybrid Tower, where we detailed the
membrane through tubular jersey for double surface channels, interlocking for reinforcement
parts and holes for tying and pre-stressing (Tamke et al., 2016). Isoropia further diﬀerentiates
between various knit patterns within the patch surface (Figure 9). Initial tests using only one base
knit pattern (Piquet Lacoste) that revealed to be too tight to achieve the strong three dimensionality
Figure 8. Outline drawings of all patches produced for the three zones of Isoropia. The graph lines surrounding the bounding
boxes indicate the amount of more expandable stitch type in the areas of the patches through combination of diﬀerent knit
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 9
needed for the cones provides structural strength to the beams. To increase cone depth we deﬁned a
fourth stitch pattern, in which the interlocking between needles is less, therefore allowing more ﬂexi-
bility to the yarn and better deformation. This fourth stitch pattern is introduced in star-shaped zones
around the cones grading the material locally for performance.
The double patches are assembled using linking (Figure 10). A process from the garment industry,
where two single layers are linked into a double surface. In Isoropia we link with a knit stitch and use
the same ﬁbres as in the patches and achieve a similar performance in the seam, as in the overall
patch, allowing better pre-stressing of the membrane.
Fibre system –manufacturing bespoke knit
At ﬁbre level the material, machine requirements and ﬁnal textile structure had to be considered in
order to achieve a promising textile system, which could handle the variations in pattern and mech-
anical stress that characterises Isoropia. It was further important to create a welcoming and soft
haptic experience for visitors who touch the fabric, unlike the plastic nature of state-of-the-art build-
ing membranes. From a limited range of available high performance yarns, the ultra high molecular
weight polyethylene yarn Dyneema® SK65 was selected due to its superior mechanical properties
speciﬁcally 3.3–3.9 GPa of Tensile Strength; 109–132 GPa of Tensile Modulus and 3–4% Elongation
at Break (Figure 11).
Figure 9. Knit patterns in Isoropia: tubular, piquet lacoste, interlock, piquet).
Figure 10. Linking inner and outer part of double-layered patches using Dunkermotoren linker (BG83 14GG).
10 M. TAMKE ET AL.
In Isoropia the development of design, knit speciﬁcation and fabrication system ran in parallel, and
each step of this only 8-week long process was evaluated in 1:1 prototypes. In order to iterate quickly
and precisely, an automatic link between the digital design tool at CITA and the CNC knitting
machines at the textile producer AFF had to be established early. The objective extended the tech-
nological aspect, as the development of the interface required a direct exchange between the design
and code development at CITA and the personnel operating and creating the knit code for the
machines on the manufacturing ﬂoor. Only this direct collaboration allowed the project to
succeed, as usual administrative layers were bypassed and the knowledge to overcome the chal-
lenges of complex knit structures could be directly picked up in the digital interface.
Producing CNC knit
When manufacturing complex knit structures, as in Isoropia, special considerations have to be taken
regarding the interdependent factors of knit structure, geometry, surface detailing, tension stability
between diﬀerent knitted structures and the resulting quality of the knitted patterns. The aim is here
always to maximise the mechanical performance of knit as well as the user experience. In Isoropia this
list of requirements was extended, as the knit system had to be able to communicate with an external
system at the designers side and changes had to be made rapidly. Therefore, a Shima Seiki M183514
14GG ﬂat knitting machine and APEX3-SDS ONE software were selected (Barﬁeld, 2015).
APEX3 oﬀers a vast library of pre-programmed conventional knit structures and patterns (Jersey,
Pique, Interlock, etc); this means that the user does not need to programme these from scratch, but
can still manipulate them if necessary (Thomsen et al., 2016). The knit structure can, in this way, be
adapted to fulﬁll non-standard requirements, and customised complex patterns can be produced
with the necessary tension stability and elasticity control between the various structures.
In terms of knit production challenges, especially regarding the yarn ﬂow, quantity of yarn per
needle, structural detail and the non-standard shape were faced.
Yarn ﬂow. Yarn ﬂow is directly related to the nature of Dyneema® SK65, which receives its high
strength, through continuous ﬁlaments with the lowest twisting possible (Rao & Farris, 2000).
When using this type of ﬁlament extra care is needed, mostly when knitting at higher gauges. The
machine used for Isoropia has a gauge of 14 (GG14-the needle bed is composed of 14 needles
per inch) and therefore the needle hooks are small and delicate (Figure 12). A strong ﬁbre introduces
high levels of stress in the needle hook, which can easily destroy it and even spread to neighbouring
needles, causing severe defects on the textile and waste of components. Extreme care and a good
production protocol are necessary.
Figure 11. Bobbin with Dyneema® SK75 220DTex yarn used in Isoropia.
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 11
Quantity of yarn per needle. The knitting process can be easily obstructed and even cause the
failure of internal components of the machine by ﬁlaments, which splice from the yarn and get
trapped along the path from the bobbin to the needle. This is likely to happen with the SK65
used in Isoropia, which is about 0.19 mm and made from 192 ﬁlaments, each with a diameter of
only ca. 0.9 µm according to the datasheet by DSM. (Finally ca. 0.38 mm yarn is in each needle
hook, as Isoropia was knitted with two feeders (2 bobbins feeding each needle)). Tests were made
at extreme low knitting velocity in order to evaluate the behaviour of SK65 in terms of ﬂow. Fortu-
nately we found that Dyneema® has a smooth, low-friction surface, which makes it possible to knit
safely, though always with vigilance.
Structural detail and non-standard shape. The structural detail and non-standard outer shape
proved to be challenging when working with Dyneema® SK65, due to the strength of the ﬁbre
and the absence of elastic properties. There is a need to balance the internal stresses of the yarn,
mostly when passing from high density structures, like Interlook, to lower density structures like
Pique Lacoste and vice versa. Here the typical knots of each structure are associated with diﬀerent
levels of internal stress in the textile, and therefore, mixed structures could induce the ﬁnal patch
to change geometry, shrink or expand in the interface areas. The geometric accuracy was key, as
the patches needed to ﬁt precisely to the form-found geometries and in the case of the linked
double surface patches they even needed to match a pair precisely. Again ﬁne adjustments by the
knit-programmer allowed us to achieve balanced structures and equilibrium between all interfaces
within the same patch and a precise outer shape.
Digital fabrication interface
The interface between design and production takes place through bitmap ﬁles (Thomsen & Tamke,
2016). These ﬁles are automatically generated from planarised meshes of the digital design system
(Figure 13). Planarisation of the form-found meshes takes action, following the general method of
geometrically forcing the mesh to the plane, while keeping the edge length values as in the original
three-dimensional mesh. As knitted textiles are known for their elasticity, this method is appropriate.
At ﬁrst, a 2D mesh with identical topology to the form-found 3D mesh is generated (Figure 13(A)).
This allows to skip a large geometrical movement of projection of a 3D mesh to the 2D plane, which
has a high risk of errors and sometimes could be also impossible. The protrusion of each cone in a 3D
Figure 12. Needle and needle size used in Isoropia.
12 M. TAMKE ET AL.
mesh is ‘relaxed’into a simpliﬁed 3D mesh, as the cones are formed as the result of textile stretch only
(Figure 13(B)). The outer edge lengths are maintained. The internal and external edges of the 3D
mesh are measured and used as constraints for a kangaroo2 relaxation of the 2D mesh with a
0.01m tolerance for the output edge lengths (Figure 13(C)). A best-ﬁtting bounding box is generated
and tested against the max width of the CNC knitting machine (Figure 13(D)). The 2D mesh and the
transposed cone centres serve as a base for an automated speciﬁcation of boundaries of areas with
diﬀerent knit structures (rod channels, details for the lighting details, reinforcement edge and the
expandable cone stars, large slits). Each line ﬁgure is highly dependant on the other ones, hence
the hierarchical relation between lines is established (Figure 13(E)).
As the segments of the structure were divided into 4 surface types, 4 line-drawing algorithms were
developed, allowing to map the surface design to any input mesh (Figure 10). Visual feedback is pro-
vided on the relation between zones of diﬀerent stretch (Piquet and Piquet Lacoste). Colours for the
line boundaries are assigned that are later used to deﬁne the colours of areas. Each colour corre-
sponds to a particular knitting structure to be applied (Figure 13(F)). In order to accommodate the
non-square nature of knit stitches, a non-uniform scaling is performed. The coeﬃcient for scaling
derives from the physical prototyping (Figure 13(H)). Line works are processed with Squid plugin
(developed at CITA by Mateusz Zwierzycki), ﬁlled with predeﬁned colours and exported as BMP
(Figure 13(I)). On the side of the receiving APEX system this colour information informed the auto-
matic conversion into the smaller scale pixel pattern of the NC knitcode, which was parsed to the
machine (Figure 14).
Prototyping as means of system calibration
Projects like Isoropia are characterised by a high interdependence of parameters across scales
(Thomsen & Tamke, 2016). Isoropia is as well a high risk research project, where despite similar pre-
cursor projects, many parameters can be determined during the project. Some can, currently, not be
determined at all as the mechanical properties of knit, especially when composed out of multiple knit
structures within a single patch and the precise simulation of it (Tamke et al., 2016). A further dimen-
sion of complexity is added through the non-standard approach, where no textile element is the
Figure 13. Workﬂow for the Digital Fabrication Tool.
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 13
same, and it is hard to measure and predict with current methods, how the change of size aﬀects the
mechanical behaviour of the knit.
We, therefore, employed strategic physical prototyping and evaluation of the build through
measurements –mainly 3D scanning (Figure 15)–from the very beginning of the project with the
aim to understand the interplay of elements and material across all scales and feedback into the
design and structural systems. We started with isolated material probes and increased size and com-
plexity, so that we could study the behaviour of composite elements and ﬁnally (Figure 15) construct
several 1:1 prototypes of parts of the system, which are the direct precursors of the ﬁnal product
(Kamrani & Nasr, 2009).
In the following sections we describe, with a material focus, the diﬀerent scales and employed
methods of investigation. We outline how each allowed the project to leap to a more reliable level
and gain validated correction values for the parameters in the design and fabrication workﬂow.
Prototyping phase 1: sampling/probes
Initial investigations took place on design level through scale models in Lycra (Figure 16)andon
material level through small swatch sampling following established methods in knitwear fashion
and the ﬁeld of technical textiles (Thomsen & Tamke, 2016).
The aim was to verify at ﬁrst that the Dyneema® yarn is the appropriate choice, as besides clear
beneﬁts of the yarn, many uncertainties regarding the process of knitting existed –in particular
the low friction of the yarn in the pull-down mechanisms and its high tenacity eﬀect onto the mech-
anics of the machine needles. Through the production of small samples, not only the knit integrity
and quality could be tested and reﬁned, but also the knitting machine could be adjusted to
produce as fast as possible, while keeping the machine mechanically sound and avoiding defects
on the textile patch.
Figure 14. Processing of the CNC knitting ﬁles for the membranes at AFF and ﬁnal production resolution.
Figure 15. 1:1 Prototyping of the membrane units. Evaluation through 3D scanning.
14 M. TAMKE ET AL.
In the next sampling round the suitability of Dyneema® for larger scale membranes was veriﬁed
through the knitting of several 1 sqm large samples in Dyneema®, based on the existing ﬁles from
Hybrid Tower 2 (2016). Here we could evaluate, how the knit and the detailing (Channels, joint
details, edges) behave in comparison to the formerly used technical yarn. A clear mechanical
improvement was observed and allowed to experience, for the ﬁrst time, the visual and tactile fea-
tures of the membrane and provided crucial feedback on the machine settings and conﬁguration
needed for knitting with Dyneema®.
Prototyping phase 2: introducing non-standard approaches
The aim of the second phase was to engage full scale and evaluate and reﬁne the assumptions about
the behaviour of the knit in asymmetric patches and the detailing of the membrane and the other
parts of the structure. For this purpose a full element of the front section of Isoropia was produced
and erected (Figure 17).
It demonstrated previously undetected challenges concerning the behaviour of the knit in large
scale and under full tension. Here are two examples:
.It was observed that the channels contracted under tension created high friction between the
knitted sleeve and the GFRP, preventing the textile patch to stretch to the desired length along
the rods. An increase in the width and a decrease in the tension of knit in the channels allowed
the fabric to stretch to its geometrical amplitude without wrinkles in the next sample.
.The initial centre surfaces of the patches consisted of Pique Lacoste structure only and were stiﬀer
than anticipated: the cones would not stretch in all patches to the desired depth under the rods,
resulting in lower structural height and therefore less structural performance. As we could as well
observe a relation between the size and position of the patch on the rod and the stretch of the
knit, we introduced a second knit structure which is more stretchy and developed a way to
grade within a patch between the two knit structures in order to achieve the desired stretch of
Figure 16. Lycra physical models investigating design possibilities within the single unit. Exploration of a single or double mem-
brane proposals with the various cable net layouts and membrane openings.
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 15
Prototyping phase 3a: fully digital workﬂow –assembly of connected elements
The next prototyping phase veriﬁed the developed technologies for non-standard design and fabri-
cation and introduced the full connection between the design model and the CNC knitting machine.
For this, 3 connected elements were produced (Figure 18) and evaluated (Figure 19). Here the dimen-
sions and the shaped geometry of all parts were taken from the calibrated digital model and pro-
cessed into CNC ﬁles for the knitting machines. A special challenge was the introduction of the
new stitch type –piquet –in order to grade the cones. Due to the lack of time, this transition
could only be developed as a direct step. Nevertheless multiple design options were tested and
ﬁnally the 4-end stars were chosen as they have a clear logic in terms of grading, increased the
depth of the cones suﬃciently and highlighted the cones of the elements nicely. The dimension
of the stars were empirically tested in 3 neighbouring prototypes, where various percentages
(from 20 to 38% of width) of the textile area were assigned a new piquet stitch type.
Prototyping phase 3b: digitally generated shaped double-layer interior prototype
The necessity to physically prototype the interior membrane was crucial as these elements had the
biggest concentration of new untested design features: double-layer membranes that were pre-
stressed through compression lighting rods and large openings in the middle of the textile. The accu-
rate dimensions of both the fabric and the openings were essential to a successful assembly as the
fabric would else start to wrinkle and sack near the edges of apertures. Due to time constraints only a
single double layer patch could be produced (Figures 20 and 21). It displayed a mismatch of the
opening dimensions to the digitally predicted ones. As a result, luckily a correct compensation
ratio was determined for the digital design model for the ﬁnal CNC production. The need, but lack
of time, to do a further more complex prototype, was revealed in the ﬁnal pieces, where double
patches were linked wrong, which resulted in non-matching assembly. The lack of markers integrated
Figure 17. Illustrations of the second phase of prototyping towards scale, system performance and textile details and features
16 M. TAMKE ET AL.
into the fabric meant the positions to link the fabric were not always obvious to the workers. Fortu-
nately the issue could be solved by delinking the fabrics and linking them once again in the right
places, which caused a delay to the project, which was, however, not threatening the deadline.
Conclusion and potentials
Our research demonstrates, through Isoropia, the potential of an integrated digital workﬂow with CNC
knit, combining Structural, Design, Material and Fabrication systems for at least the ﬁeld of small-scale
membrane structures. The transdisciplinary and interscalar approach has reached a
maturity, which allowed us to design and build a non-standard structural membrane system in only
Figure 18. Three digitally generated shaped single layer patches are selected from the overall design for 1:1 scale prototyping.
Figure 19. Comparison of the digital beam bending geometry to the scanned physical prototype point cloud.
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 17
Figure 20. Two selected testing patches for prototyping the double layer interior module. The prototype was made out of black
Dyneema® due to an intermediate lack of white yarn and in order to test the visual look of black yarn for future projects.
Figure 22. Interior view. The knitted membrane ceiling with integrated light creates a radically new spatial experience and its
haptic qualities invited guests to interact with the textile.
Figure 21. Digital layover of the digital mesh and the 3D scanned geometry of the interior prototype.
18 M. TAMKE ET AL.
4.5 months, with new spatial experiences and a new level of detailing. To date, Isoropia is the largest
structure made in pre-programmed CNC knit, with every single membrane being unique (Figure 22).
The developed process is prototypical and needs further research, we see, however, that our
approach to computational knit oﬀers potential for new designs of structural membranes, as well
for the work with CNC knit in a more general digital context.
Key is here the computation of knit, meaning a generative approach and integration of design,
analysis and making of knit. We observe that only this link makes textiles truly digital, as despite
all eﬀorts, neither membrane nor technical textiles nor fashion is digitally integrated.
In contrast to weave, we ﬁnd through Isoropia and our research that CNC knit is an ideal base tech-
nology for this digital push. CNC knit is essentially an additive manufacturing process that shares
many challenges and potentials with 3D printing. In Isoropia computational knit allows to produce
textile membranes with custom behaviours and geometries. Despite the existing challenges in
terms of predictability and control of customised knit, Isoropia demonstrates the potential when
the direct programming of material behaviour is used strategically for the grading and design of
fabrics elements and overall structure.
For textile architecture our research in Computational knit opens up opportunities for new design
expressions. The ability to morph provides the structure with modes to adapt and create highly local
interfaces to the existing ones, as needed, when building in existing or historic context.
Through the integration of details into the material, computational knit oﬀers potentials for faster
building processes, as the overall complexity of the structure is minimised, leading to less on-site fab-
rication, assembly and the handling of many parts. Isoropia demonstrates how the large amount of
knit structures to choose from allows us to tune speciﬁc areas of a single surface to fulﬁll highly
specialised and high performance tasks.
The ability to change the structure, yarn type or simply the colour on a loop to loop basis rep-
resents in our understanding the biggest potential and freedom for design. On a mundane level it
allows the integration of markings –which we unfortunately didn’t do for the linkage of double sur-
faces in Isoropia. And on an advanced level it allows for direct knitting of 3D shapes and grading of
material behaviours. Currently a way to predict these behaviours computationally is missing, but as
rapid production is possible with CNC knitting this drawback can be countered and physical proto-
typing allows for quick design progression. In Isoropia 3D scanning technology proved to be a valid
approach for quick veriﬁcation of 1:1 scale physical prototypes and the comparison of point cloud to
the digital mesh allowed for direct feedback in the design environment.
However, the development of methods to computational design and predict the behaviour of knit
materials across scales is still a boundary for computational knit.
The realisation of Isoropia builds on an interdisciplinary collaboration between CITA, Centre for IT and Architecture
(design and computation), str.ucture (engineering), AFF –A. Ferreira & Filhos, SA (knit fabrication), DSM Dyneema B.V
(ﬁbre) and alurays lighting technology GmbH (lighting). The project is kindly sponsored by Topglass Italy (GFRP
tubes), Soﬁstik (FE analysis), WK-Led Netherlands (ﬂexible LEDs) and SIKA (Glue) and ﬁnancially supported by the
Danish Ministry of Higher Education and the Sapere Aude: Advanced Grant for Elite Researchers, Danish Council for Inde-
pendent Research –Complex Modelling, Grant No. 12-125688.
No potential conﬂict of interest was reported by the author(s).
This work was conducted in the frame of the Complex Modelling project funded by the Sapere Aude: Advanced Grant for
Elite Researchers, Danish Council for Independent Research (Grant No. 12-125688).
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 19
Martin Tamke http://orcid.org/0000-0003-1209-4967
Ahlquist, S., & Menges, A. (2013). Frameworks for computational design of textile micro-architectures and material behavior.
33rd Conference of ACADIA (pp. 281–292).
Barﬁeld, W. (2015). Fundamentals of wearable computers and augmented reality. Boca Raton, FL: CRC Press.
Bernert, K. (2018). Facade made of blue and white membrane Panels, Russia Nizhniy Novgorod Stadium. TensiNews,35,
Bouaziz, S., Martin, S., Liu, T., Kavan, L., & Pauly, M. (2014). Projective dynamics: Fusing constraint projections for fast simu-
lation. ACM Transactions on Graphics,33(4), 154:1–154:11.
Dupont, S. (2014). Architectural shrouding: A second skin for façade and ceiling. Riyadh. TensiNews,27, 12c13.
EN 1990. (2002). Eurocode –Basis of structuraldesign, The European Union Per Regulation 305/2011, Directive 98/34/EC,
EN 1991-1-1. (2002). Eurocode 1: Actions onstructures –Part 1–1: General actions –Densities,self-weight, imposed loads
for buildings. Authority: TheEuropean Union Per Regulation 305/2011, Directive 98/34/EC, Directive 2004/18/EC.
Gengnagel, C., La Magna, R., Ramsgaard Thomsen, M., & Tamke, M. (2018). Shaping hybrids-form ﬁnding of new material
systems. IJAC,16(2), 91–103.
Hensel, M. U., & Menges, A. (2008). Membrane spaces. Architectural Design,78(192), 74–79.
Holden Deleuran, A., Schmeck, M., Quinn, G., Gengnagel, C., Tamke, M., & Ramsgaard Thomen, M. (2015). The tower:
Modelling, analysis and construction of bending active tensile membrane hybrid structures. IASS Symposium 2015,
Kamrani, A., & Nasr, E. A. (2009). Engineering design and rapid prototyping. New York, NY: Springer.
Knippers, J., Cremers, J., Gabler, M., & Lienhard, J. (2011). Construction manual for polymers + membranes. Basel: Birkhäuser
La Magna, R., Fragkia, V., Noël, R., Baranovskaya, Y. S., & Thomsen, M. R. (2018). Isoropia: An encompassing approach for
the design, analysis and form-ﬁnding of bending-active textile hybrids. IASS Symposium 2018.
Längst, P., Bauer, A. M., La Magna, R., & Lienhard, J. (2018). Isogeometric methods at the interface of architecture and
engineering. IASS Symposium 2018.
Lienhard, J., Bergmann, C., La Magna, R., & Runberger, J. (2017). A collaborative model for the design and engineering of a
textile hybrid structure. IASS Symposium 2017.
Lienhard, J., & Knippers, J. (2015). Bending-active textile hybrids. Journal of the International Association for Shell and
Spatial Structures,56(1), 37–48.
Lienhard, J., La Magna, R., & Knippers, J. (2014). Form-ﬁnding bending-active structures with temporary ultra-elastic con-
traction elements. In N. De Temmerman & C. A. Brebbia (Eds.), Mobile and rapidly assembled structures IV (Wit trans-
actions on the built environment, pp. 107–115. Southampton: WIT Press.
Lyall, S., & Herron Associates. (1992). Imagination headquarters: Herron associates. Vienna: Phaidon.
Nicholas, P., Zwierzycki, M., Nørgaard Clausen, E., Hutchinson, C. R., & Thomsen, M. (2017). Adaptive robotic fabrication for
conditions of material inconsistency increasing the geometric accuracy of incrementally formed metal panels. In A.
Menges, B. Sheil, R. Glynn, & M. Skavara (Eds.), Fabricate (pp. 114–121). London: UCL Press.
Philipp, B., Breitenberger, M., D’Auria, I., Wüchner, R., & Bletzinger, K.-U. (2016). Integrated design and analysis of
structural membranes using the Isogeometric B-Rep analysis. Computer Methods in Applied Mechanics and
Popescu, M., Reiter, L., Liew, A., Van Mele, T., Flatt, R. J., & Block, P. (2018). Building in concrete with an ultra-lightweight
knitted stay-in-place formwork: Prototype of a concrete shell bridge. Structures,14, 322–332.
Quinn, G., Deleuran, A. H., Piker, D., Tamke, M., Brandt-Olsen, C., Thomsen, M. R., & Gengnagel, C. (2016). Calibrated and
interactive modelling of form-active hybrid structures. IASS Symposium 2016.
Rao, Y., & Farris, R. J. (2000).A modeling and experimental study of the inﬂuence of twist on the mechanical properties of
high-performance ﬁber yarns. Applied Polymer,77(9), 1938–1949.
Sabin, J. E. (2013). My thread pavilion: Generative fabrication in knitting processes. 33rd ACADIA Conference (pp.
Solly, J., Frueh, N., Saﬀarian, S., Prado, M., & Menges, A. (2018). ICD/ITKE research pavilion 2016/2017: Integrative design of a
composite lattice cantilever. IASS Symposium 2018.
Stranghöner, N., Uhlemann, J., et al. (2016). Prospect for European guidance for the structural design of tensile membrane
structures –Support to the implementation, harmonization and further development of the Eurocodes (European
Commission Joint Research Centre Science and policy report). Luxembourg: Publications Oﬃce of the European
20 M. TAMKE ET AL.
Tamke, M., Baranovskaya, Y., Deleuran, A. H., Monteiro, F., Stranghöhner, N., Uhlemann, J., & Thomsen, M. R. (2016).
Bespoke materials for bespoke textile architecture. IASS Symposium 2016, Tokyo, Japan.
Tamke, M., & Thomsen, M. R. (2009). Implementing digital crafting: Developing: It’s a small world. In C. Gengnagel (Ed.),
Design modelling symposium (pp. 321–329). Berlin: University of the Arts.
Tamke, M., Thomsen, M. R., Sinke, Y., Gatz, S., La Magna, R., Lienhard, J., …Wisemann, M. (2019). Longterm behaviour and
geometrical precision of graded CNC knitted Bending Active Textile Hybrids under real world conditions. Proceedings of
IASS Annual Symposia, IASS 2019, Barcelona, Symposium: Elastic Gridshells and Bending Active Systems, 1–8.
Thomsen, M. R., Baranovskaya, Y. S., Monteiro, F., Lienhard, J., La Magna, R., & Tamke, M. (2019). Systems for transformative
textile structures in CNC knitted fabrics –Isoropia. In A. Zanelli, C. Monticelli, M. Mollaert, & B. Stimpﬂe (Eds.), Softening
the Habitats –Proceedings of the TensiNet Symposium 2019 (pp. 95–110). Milano: Politechnico Di Milano.
Thomsen, M. R., & Tamke, M. (2016). Prototyping practice: Merging digital and physical enquiries. In C. Gengnagel, E.
Nagy, & R. Stark (Eds.), Rethink! Prototyping (pp. 49–62). Cham: Springer International Publishing.
Thomsen, M. R., Tamke, M., Deleuran, A. H., Tinning, I. K. F., Evers, H. L., Gengnagel, C., & Schmeck, M. (2015). Hybrid tower,
designing soft structures. In M. R. Thomsen, M. Tamke, C. Gengnagel, B. Faircloth, & F. Scheurer (Eds.), Modelling behav-
iour (pp. 87–99). Cham: Springer.
Thomsen, M. R., Tamke, M., Karmon, A., Underwood, J., Gengnagel, C., Stranghöner, N., & Uhlemann, J. (2016). Knit as
bespoke material practice for architecture. 36th ACADIA conference.
ARCHITECTURAL ENGINEERING AND DESIGN MANAGEMENT 21