Conference PaperPDF Available

Abstract and Figures

Recent advances in computation allow for the integration of design and simulation of highly interrelated systems, such as hybrids of structural membranes and bending active elements. The engaged complexities of forces and logistics can be mediated through the development of materials with project specific properties and detailing. CNC knitting with high tenacity yarn enables this practice and offers an alternative to current woven membranes. The design and fabrication of an 8m high fabric tower through an interdisciplinary team of architects, structural and textile engineers, allowed to investigate means to design, specify, make and test CNC knit as material for hybrid structures in architectural scale. This paper shares the developed process, identifies challenges, potentials and future work.
Content may be subject to copyright.
VI International Conference on Textile Composites and Inflatable Structures
E. Oñate, K.-U.Bletzinger, and B. Kröplin (Eds)
, TH.
Centre for Information Technology and Architecture (CITA)
The Royal Danish Academy of Fine Arts, Schools of Architecture, Design and Conservation
Copenhagen, Denmark,
Department for Structural Design and Technology (KET),
University of Arts Berlin,
Fibrenamics, Universidade do Minho
Guimarães, Portugal,
AFF a. ferreira & filhos, sa
Caldas de Vizela, Portugal,
Essener Labor für Leichte Flächentragwerke, Universität Duisburg-Essen
Essen, Germany,
Key words: Bespoke Textiles, CNC Knit, Form Finding, Material Design, Testing
Summary. Recent advances in computation allow for the integration of design and simulation
of highly interrelated systems, such as hybrids of structural membranes and bending active
elements. The engaged complexities of forces and logistics can be mediated through the
development of materials with project specific properties and detailing. CNC knitting with
high tenacity yarn enables this practice and offers an alternative to current woven membranes.
The design and fabrication of an 8m high fabric tower through an interdisciplinary team of
architects, structural and textile engineers, allowed to investigate means to design, specify,
make and test CNC knit as material for hybrid structures in architectural scale. This paper
shares the developed process, identifies challenges, potentials and future work.
Current structural membranes are based on coated fabrics or films, produced in
homogenous lanes, which are cut in pattern and seemed, before details for the interfacing with
other parts are applied. The current approach towards the manufacturing of a highly bespoke
membrane is rooted in the limitation of the underlying process of weaving. Knitting on the
other hand is characterised by the ability to create highly heterogeneous materials and
bespoke figures. With industrial CNC knitting machines, it is for instance possible to knit to
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
shape, introduce details directly into the material, combine and change fibres of different
strength or change the structure of the material at any place. The material allows hence for
highly bespoke textile surfaces [1]. Within knitted surfaces, the complexity of detailing,
material properties and production can be collapsed into the material itself.
The design and production of a 8m high textile tower (Fig.1) made from bending active
load carrying GFRP rods and restraining membrane surfaces explored the potential of CNC
knit surfaces in architectural application, especially for temporary installations. The Tower
was developed and produced in the period autumn 2014 to spring 2015 and assembled and
exhibited in the courtyard of the Danish Design Museum in April to May 2015. The
development of the tower is based on previous research and expertise of the partners in their
respective fields. The actual production of a physical demonstrator required however to
synthesise the respective fields and created insights in the systems, tools and processes, which
need to come into place to realise a novel set of hybrid structures of bespoke CNC knit
membranes and active bending elements.
Figure 1: The Tower in the Courtyard of the Danish Design Museum.
Figure 2: The interior of the Tower is characterised by the tensioning system and the resulting cone-like
membranes. Photo: Anders Ingvartsen.
Active bending is the intentional use of elastic deformation as a shaping process for
linear or planar structural elements. The main motivation for its use lies in the simplicity of
creating curved elements for exciting and versatile architectural forms but with a low demand
on production energy and material consumption [2, 3]. The Tower extends previous research
into bending active structures with membrane surfaces on the level of scale and through the
integration of details into the membrane. The most prominent examples of bending active
structures, the Multihalle Mannheim (1979) by Mutschler, Langer and Otto [4] and recent
continuation of this work by Bavarell [5], used braced grids made from bend element. The
membrane provides enclosure, but does not contribute on the structural level. Ahlquist and
Lienhard have introduced structures on installation scale, as in the Tour de l’Architecte (2012)
in Monthoiron/France [6], where a limited number of surfaces stabilise bending members.
The structural pattern of the tower under discussion in this paper consist however of 64
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
individual membrane surfaces and even more bending active rods. Our research comes hence
closer to the amount of interacting discreet elements found on building scale.
As elastic bending generates residual stresses, the choice of material, cross-section height and
curvature is limited. However, in hybrid systems the selection of the profiles’ properties can
be optimized for the bending process, since under external loads the restraining elements
drastically improve the global stiffness of the hybrid structure. Previous research has shown
that the application of membranes as restraining elements has several advantages in
comparison to typical cable based restrained systems [7,8]. Physical and numerical test have
emphasized the importance of the construction details- especially the joining of membrane
and actively bent rods- of such hybrid systems for their global bearing behavior. The stiffness
of membrane restrained systems is i.e. significantly reduced, if the connections to the bent
element and the edge-cable allows sliding a strong connection and channels for the bending
active element are recommended.
The tower explores furthermore textile joints made from high tenacity knit as softer
alternative to common steel joints in active bending structure. These create obstructive force
maxima within coupled bending active elements [9].
The basic concept of the hybrid structure is to create a membrane reinforced by a dense grid
of overlapping bending and compression stiff GFRP rods, continuously integrated in the
membranes pockets and channels. This necessitates a fine balance between softness and
rigidity in the material, in order to create a structure flexible and bendable - capable of
responding to changing conditions of the environment.
The structural system of the tower consist of a double curved grid- shell like structure
made of slender overlapping rods, embedded in a membrane made from high tenacity yarn in
the surface of the shell in combination with a cable based restraining system perpendicular to
the surface of the shell (Fig. 2). This second restraining system pulls the centre of each
membrane patch radially to the tower central axis. This results in a spoke wheel effect which
provides horizontal stiffness, braces the cylindrical gridshell and increase the load capacity of
the overall system significant. The dominant load cases for such a vertical oriented
lightweight structure are the agency of wind with an exponential increase of loading in height
and a risk for dynamic complications. In the horizontal section the cylindrical shape of the
tower leads to typical wind load distribution with less pressure windward, high suction
sidewise and lower suction leewards (Fig 3).
Figure 3: FE simulation of the tower with wind load distribution and bracing system. Deflected state under
windload from front in the middle and relaxed state on the right.
Under such loading the membrane is supposed to act in two ways: as a tensile element to
transfer tensile forces on the windward side and as restraining systems to prevent all curved
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
compression elements for further bending and buckling in plane. The radial restraining cable
system has also multifunctional purposes: as a form-stabilisation system the spoke wheels are
preventing any further bending out of plain of the rods while they give the membrane patches
a strong local double curvature, increasing the pre- stress of membrane.
The tower project assigns the textile material key roles:
Reduction of fabrication effort through: Knitting of patches with bespoke shapes
instead of cutting and patching
Reduced assembly time through minimisation of amounts of seams and integration
of detail
Bracing of structure under load: use of high tenacity yarn
Constraining the movement of bending active rods: Integration of detailing into the
membrane (tunnels and pockets)
Minimisation of pattern cut: Use of material stretch to achieve double curved
Designed material behaviour under forces: development of bespoke knitting
Commonly used knit is however characterised by high stretch, low strength, antistrophe
behaviour and non-linear behaviour. To develop sufficient behaviour and detail through the
local variation of knit structure and fibre, motivated the research on the material level.
Weft knitting technology was used to produce the fabric for the tower. As the tower is
exposed to various weather conditions and submitted to various stresses and strains, high
performance yarns based on polyamide (PA), polyester (PES) and polypropylene (PP) are
good candidates The lower moisture absorption, average elasticity and very high resistance
makes high tenacity polyester here an especially well suited.
In terms of knit structure single cardigan, half cardigan, rib, interlock, jersey and piquet
Lacoste were investigated in tests, where only the latter two showed the required transparency
and low level of elasticity. During the knitting of samples on a on an Stoll CMS 320 TC
electronic flat knitting machine it became obvious that the loop length and the PES yarn linear
density (55 and 110 tex) (Table 1), were the crucial parameters to control the fabrics elasticity
and porosity. The tensile behavior of the samples was studied by Grab method, according to
NP EN ISO 13934-2 standard, on a Hounsfield universal testing machine. Five specimens
(100x200 mm) of each sample were tested in coursewise and walewise directions, with a
crosshead speed of 100 mm/min.
The obtained results show, that the structures produced with higher linear density yarns
present higher tensile strength. When comparing Jersey and piquet Lacoste structures, for the
same linear density, it appears that, in the coursewise direction, piquet Lacoste structure has
higher tensile strength and lower elongation – making it the structure of choice for mechanical
behavior, but also for visual aspects.
The pockets, channels and reinforced areas in the membrane were developed on the base of
a ground structure (piquet Lacoste) using the machines abilities for individual needle
selection, take-down adjustment during knitting, loop transfer and racking.
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
NP: 10,5 NP: 11 NP: 12 NP:11 NP: 10,5 NP: 11 NP: 12 NP:11
PES HT 100 PES HT 100 PES HT 100 PES HT 100 PES HT 100 PES HT 100 PES HT 100 PES HT 100
multifilament multifilament multifilament multifilament multifilament multifilament multifilament multifilament
55 55 55 110 55 55 55 110
0,48 0,53 0,63 0,6 3,77 4,16 5,08 4,59
( 5%)
Maximum strength
at break [N] -
walewise direction
Extension at break
[%] - walewise
Maximum strength
at break [N] -
Extension at break
[%] - coursewise
Piquet Lacoste
Tensile Strength
Loop length (lu)/100 wales [cm]
Composition [%]
Yarn Type
Linear density [tex]
Table 1: Mechanical and Structural Characterization of the produced samples
It was crucial for the design and simulation of the tower to know the behaviour of the knit
under biaxial load. Currently no testing procedure exists for the biaxial testing of knitted
fabrics, which was hence performed according to the testing procedure of the Japanese
Standard MSAJ/M-02-1995 [10] for biaxial testing of woven fabrics. Herewith, the
“fictitious” elastic constants of the material could be determined.
The main characteristic of MSAJ/M-02-1995 is that five different predefined stress ratios
warp:fill 1:1, 2:1 1:2, 1:0 and 0:1 are consecutively applied on a cross shaped test
specimen with the yarns parallel to the arms of the cross. During the loading and unloading
procedure, the load ratio warp:fill is held constant. The maximum tensile test load is fixed to
¼ of the maximum strip tensile strength of the fabric direction with the lower strength. The
result of this test procedure is a stress-strain-diagram. From this complete set of test data ten
stress-strain-paths can be extracted one for each yarn direction for the five load ratios.
MSAJ/M-02-1995 recommends to determine on the base of the five ratios one single design
set of elastic constants from the extracted stress-strain-paths stepwise in a double step
correlation analysis.
In the frame of this project two different types of Piquet Lacoste Structure CNC knitted
fabrics were tested in the Essen Laboratory for Lightweight Structures (ELLF), University of
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
Duisburg-Essen: NP11 with linear densities of 110 tex and 55 tex. For two reasons it was not
possible to perform “classical” biaxial tests according to MSAJ/M-02-1995 with this material:
The geometrical dimensions of the available material were limited to a length of 1600 mm
(wales, machine direction) and a width of 700 mm (course, transverse direction). Herewith,
the ELLF-standardized cross shaped test specimens (1500 mm x 1500 mm) to be used in the
ELLF biaxial testing machines could not be cut.
As the material is knitted, cutting it in a cruciform shape was not possible without any
additional auxiliary seams in order to avoid unravelling of the material.
In total, four biaxial tests were carried out, three for the 110 tex and one for the 55 tex
material. Due to the aforementioned reasons, the “classical” cruciform geometry of the biaxial
test specimen was modified to the test specimen geometry presented in Figure 4. As the
material behaviour was unknown, a stepwise procedure was chosen to identify the biaxial
material characteristic of the CNC knitted fabric. In a first step the 110 tex material was
investigated by performing two MSAJ-tests with an upper maximum tensile test load of ¼ of
the maximum strip tensile strength of the material and finally one with a reduced maximum
tensile test load of ca. 17.5 % of the maximum strip tensile strength. The reduction became
necessary due to the fact that the material exhibited very high strains and transverse
contractions in both directions in the load ratios 2:1, 1:2, 1:0 and 0:1. As the uniaxial load
ratios still showed very high strains and transverse contractions, it was finally decided to
perform additional uniaxial tests in the biaxial testing machine whereby the zero stress
direction was unclamped. For the 55 tex material the last test procedure was applied with one
modification: the 1:0 and 0:1 load ratios were only applied in the pure uniaxial testing with
unclamped zero stress direction.
Figure 4: Biaxial testing of the CNC 55 tex knitted fabric with modified test specimen after applying the
prestress [© ELLF]
Exemplary for the 55 tex material, Figure 5 presents the loading sequence according to
MSAJ/M-02-1995 without the 1:0 and 0:1 load ratios and the stress-strain-diagram as a result
of the biaxial test.
From the complete set of test data six stress-strain-paths were extracted – one for each yarn
direction for the three load ratios. Together with the uniaxial tests, one single design set of
elastic constants from the extracted stress-strain-paths were determined stepwise in a double
step correlation analysis. In the first step each curved loading path was substituted by a
straight line. In the second step the slopes of the straight lines obtained in the first step were
modified in such a way that they satisfy the equations of the assumed linear-elastic
constitutive law using a correlation analysis routine programmed at the ELLF, University of
Duisburg-Essen [11, 12, 13].
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
Fig 5.1 & 5.2: Exemplary loading sequence according to MSAJ/M-02-1995 without the 1:0 and 0:1 load ratios (left)
and the stress-strain-diagram as a result of the biaxial test (right) [© ELLF]
The Piquet Lacoste, NP 11, 55 tex material shows an approximately bilinear behaviour
with a significant change of stiffness in the stress range of 0.5 - 2 kN/m, so that two
evaluations for different stress intervals were conducted:
(1) stress interval 0.1 - 3 kN/m and
(2) stress interval 1.0 - 3 kN/m.
As a result, fictitious elastic constants were determined from the two sets of diagrams as
given in Figure 6. Overall, the tensile stiffness is extremely low with Young’s moduli of
approximately E = 5 kN/m for wales and course for the full stress interval. With E
10 kN/m and E
= 26 kN/m it is only slightly higher for the stress interval 1 - 3 kN/m. In
contrast, the Poisson’s ratios are very high with minor ν = 0.83 and 0.66, respectively. Strains
are large with up to 70 % under uniaxial loading. It can be stated that some of the calculated
linear graphs do not correlate well with “their” measured stress-strain-paths.
Stress interval 0.1 3 kN/m
Stress interval 1 3 kN/m
Figure 6: Correlation between measured and calculated stressstrain paths and resulting stiffness parameters for
Piquet Lacoste, NP 11, 55 tex for two stress intervals: 0.1-3 kN/m (left) and 1-3 kN/m (right) [© ELLF]
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
The interaction between the towers two form active systems membrane and GFRP
determines the towers design and behaviour. A design process is required, which can predict
this and feedback to the designer the resulting shape, but especially whether the material and
elements specifications are sufficient. We use a two stage approach, combining a new
developed particle spring simulation for an almost realtime form finding with hundred and
more interacting bending members and a FE simulation for detailed analysis.
5.1 Formfinding
Established methods for the form finding of membranes, such as particle spring or dynamic
relaxation, do not consider material properties or requirements from fabrication, such as
patching [14]. During form finding the membrane is set to a fraction of its real stiffness in
order to facilitate large deformations. The resultant shape represents the pure flow of forces.
Our hybrid structural system consist however of interacting members, where, unlike the
membrane, the rods maintain their bending stiffness during form finding and influence the
final shape and stresses. The higher the curvature and the diameter of the rods, the higher the
stresses in the material. The material of the rods need to have a high strength and a low
Young’s modulus to be able to perform accordingly. It is important to create a curvature high
enough to tension the membrane without risking breaking the rod under excessive bending.
External loading, such as wind, causes quite big deflections of the structure, which exceeds
easily the bearing capacity of the structure’s elements. Structural analysis is hence essential to
design the tower within its bearing capacity.
Figure 7: Formfinding tool in Rhino/Grasshopper with inbuild comparative bending radii analysis of differently
dimensioned towers. Note the relationship between macro shape and bending radii.
The developed fomfinding tool is based on the Kangaroo 2 particle spring solver in
Grasshopper/Rhino, operating on discrete piecewise linear geometries for modelling the
behaviour of bending members and coupled discreet meshed for the tensile membranes in one
unified and interactive system. The form finding and dimensioning process has three stages:
generating, exercising, and refining constraints. The designer is guided during the process by
inbuilt lightweight analysis features, providing feedback on the bending stress, utilisation and
reserve of the member in isolation (Fig. 7), as the size and shape of the resulting membrane
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
patches in the plane (Fig 8). Optimisation algorithms help to determine the best fit of the
unrolled flat knitting patch in the max. width of the CNC knitting machine and constraints
coded into the tool, prevent the design to exceed i.e. the max. length of bending elements. The
process outputs the form found geometries and solver statistics.
Fig. 8 Steps of refinement and deviations during development of strips of the doubly curved membrane into flat
knit patterns.
5.2 FE Analysis
As the formfinding tool does not provide the resulting stresses, caused through the
construction and external forces an FE environment (Sofistik) is used for subsequent analysis.
It is well suited to the real-time form finding and provides a precise mathematical
definition of the global stiffness matrix. Large deformations must be simulated using an
incremental process. Current approaches for the simulation of complex structural systems
with large deformations, such as the elastic cable approach, developed by Julian Lienhard
[15], reach their limit, in a system like the tower, where many elements and membranes are
interacting. In our approach the form found geometry is imported as lines and surfaces into
FE and converted into structural beam and membrane elements and materials, cross-sections
and support conditions are defined. The results from the bi-axial tests defined the
characteristics of the knitted membrane. The material behavior in terms of elasticity where
nonlinear and linearised for two stress intervals to be used as elastic constants in the FE-
The expected levels of prestress in the hand assembled tower did only allow for the lower
range of 0.1-3kN/m, so an isotropic Young’s modulus of 10 Mpa was ascertained for the FE-
simulation. The analysis confirmed the viability of the structural system, indicated however,
that the membrane would have to be multiple times stiffer in order to balance the levels of
applicable prestress, remaining structural capacity in the rods and external loading.
5.2 Preparation of Fabrication data
The specification for the different Knit patches is directly derived from the form finding
tool. The form found patches are however curved and represent the knit in a pretensioned
state. As this prohibits a direct unroll we develop the knit patterns in several steps:
1) The topology of the doubly curved patch is projected into the XY plane
2) A relaxation algorithm creates a best fit of the original length of the edges within the
subdivided polygon to the flattened version. The numeric results showed overall
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
deviations in the range of millimeters. Extensive testing of the assembly of the derived
cutting patterns onto physical models proofed the validity of the approach (Fig. 8). The
pretension of the knit and the anisotrophy of the material is compensated through non-
uniform scaling of the pattern. The necessary ratio is based on the bi-axial tests and
measurements on a limited set of test patches and prototypes, which allowed
determining the behaviour of the assembly rather than the constituting elements.
3) The compensated polygon representation of the unrolled stripes, are converted to
vector representations and automatically refined with details, such as areas with
reinforced textile (110tex), pockets, channels, holes and elements for the tensioning.
The refined vector pattern is non-uniform scaled to compensate the difference between the
computational representation of a knitted loop for CNC knitting (square pixel) and the
physical reality (rectangular).
The form finding tools provides finally the production information for the GFRP rods and
for the CNC knit (fig 9). The information for the patches are bespoke to the level of local knit
structure and overall shape.
Fig. 9: Production information with all details included is generated in Pixel format. White= Piquet, Grey=
interlock, Pink= Tubular Jersey. Red pixels = holes. The figure shows the state before compensation for knit and
machine parameters.
Each patch has three distinct knitting structures. The central part, using Piquet, is more
elastic, and transparent with greater aptitude towards stretching. The peripheral and other
reinforced areas use Interlock structures and Tubular Jersey (Fig. 10).
Fig. 10: Structural details on the textile. Piquet / Interlock / Tubular Jersey / Rod placement & sewing holes
Fig. 11: Each patch is reinforced in the centre and holds a detail for the tensioning to the towers centre axis.
Three pairs of holes are allowing cables to pass from the outside to the inside without interfering with the
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
Both structures use front and rear needle beds simultaneously in order to increase the
stability and functionality. Interlock produces perfectly intertwined double knit, while the
shift to tubular jersey produces two independent faces, creating a pocket or cavity the
channels and pockets for the GFRP rods. The same principle is used for circular channels in
the centre of the membrane patches, where steel rings are inserted, which provide in concert
with CNC milled plastic details the interfaces to the towers inner tensioning system. Within
the circles, three pairs of Ajour knitting patterns create holes, which allow tensioning
cables to pass from the outside pulling plastic ring to the center of the tower (Fig 11).
This knitting technique is as well used in the boundary regions of the up to 8m long knitted
strips and creates here holes for interfacing seems with neighbouring strips. These holes use
only two needles width (line 2 of Figure 12), while a hole with four needles width (line 4)
creates holes wide enough to introduce rods into the channels.
Fig. 12: Ajour Knitting Pattern is used in order to create a larger hole (approximately 4 needle width). Note that
this representation is made on a Jersey basic structure in order to simplify the drawing. Here each line is
represented by a pair of dots, each dot represents a needle and each line of dots represents a needle bed, rear and
front, respectively.
Fg.13: Weft-knitting electronic machine Shima Seiki SSR112
The final production of the membrane took place on a weft-knitting machine Shima Seiki
SSR112 (Figure 6), with 1150 mm of total width, 10 (needles/inch) gauge, 450 available
needles and 55 take down. This machine was selected due to the high quality production, high
memory capacity and a gauge similar to the one used during Fibrenamics´s developments,
which should assure repeatability of the mechanical properties of the knit. Due to the
complexity of the fabrics design and the elastic behaviour of the used yarns, the production
speed was lower than for conventional textiles. Higher speed may lead to faults of the textiles
structure, due to characteristics of flatbed knitting needles and as this would complicate the
internal cutting system for the yarn.
It is challenging to CNC knit as bespoke and with the amount of changing detail, as in the
tower :
The precise behaviour and dimension of knit is hard to foresee before the actual
production despite all measurements. The channels should for instance all have the
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
same width in any angle on the surface. They should be tight enough to fix rods in the
determined position, and wide enough for larger diameters of rods. Extensive tests
were necessary.
Trade-off between the integration of detail and the stability of the structure. This
is especially true for areas around wider holes with Ajour knitting pattern. The sewing
holes are made through transfer of loops between neighbouring needles (see Fig. XX
Line 2, transferring C to B and D to E). However the placement holes couldn’t simply
be made by replication of this transfer, as this would compromise the structures
stability. The increase in size of the hole was hence accomplished through creating a
weakening of the surrounding structure. In this way the rest of the transfer could be
introduced later, obtaining only a slightly larger but more elastic opening of the
channel, so that rods could be forced in. Grinding the edges on the tip of the rods
eased the insertion. Bigger diameter rods damaged however often loops and led to
larger defects in the knit.
The inner logics of knit and fabrication limit the complexity in the structure.
Constraints emerge in the fabrication end, which make it difficult to realize the
theoretically possible overlay of varied width of the knit, a varying amount of details
and a stable structure. The special challenge in the pattern for the tower, was to place
all crucial details into the knitting.
Knowledge, Process and Technological barriers. In the production of knitted fabrics
the synthesis of size, fabric’s stability and experience and knowledge of technicians
are decisive aspects. The needle bed determines i.e. the width of each part, whilst the
achievable complexity, detailing and size is dependent on the storage capacity pattern
without compromising texture and dimensional stability. Methods to untangle these
constraints and create a more systematic and scientific approach are needed.
The validity of the projects decisions were tested throughout the development process by
means of physical models in increasing scale up to 1:1 prototypes. These models verified i.e.
the precision of the formfinding tool through comparative studies of 3d scanned physical
models and their simulation.
For the tower project the interaction of membrane and GFRP rods, the ability to pretension
the knit through the rods manually, as well as the functionality of inserting and fixation of the
rods in the textile through channels and pockets could only be verified in 1:1 scale prototypes.
As CNC knitting on large scale machines means a considerable effort, the project had only
limited possibilities to test and improve the taken decisions iteratively. The test of a single
patch was possible prior to the actual production, verified the approach, but demonstrated as
well, that the structure in total was more elastic and less precise than anticipated. Due to time
constraints, the effective length of the rods was increased through sewing the pockets shorter.
This led to imprecisions on the global scale and deviations from the simulated geometry.
These were increased by further imprecision induced through the fabric connections of the
rods. While the system of pockets and channels was generally strong enough to prevent rods
from poking through the fabric, the design of the detail allowed for a torsion between the
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
rods. A more fixed and precise solution would be necessary.
The final assembly took place on the ground, starting with the top layer, filling in each story
underneath as soon as it was completed - a strategy, which made it possible to abstain from
expensive temporary scaffolding. Simply a centered scissor lift was used to push the build
part a level upwards. It turned however, out that hanging the structure created distortions,
which could not be adjusted in this state. Having smaller more flexible jacks around the
perimeter would be a better strategy, especially as the structure proved stable in intermediate
The monitoring of the tower over a four week period showed, that it was able to withstand
wind forces up to 11 m/s.
The project demonstrates the potentials of material specification in CNC knit for
architectural structures. The project found several practical challenges in the process of
specifying material properties in knit, their implementation into fabrication and the
verification of these. These have been discussed in the chapters above. The project is first of
all a case for a future integrated and interdisciplinary design practice, where a building
structure can only be understood as an interacting multi-scalar system.
Feedback on tool and process level
The design of the towers hybrid structure required the integration of constraints, that
determine the making of a single loop of knit, with those that operate on the largest scale. This
is a call for an exchange and feedback between all partners from design to fabrication.
On the level of design computation, the projects demonstrates successfully, how feedback
from all levels can be implemented in digital design tools and how these are the key
component to design with highly interdependent hybrid systems. Information from the
structural, fabrication and level of assembly was successfully integrated in the form finding
tool and allowed to design the overall structure, as the discretisation of the surfaces, so that
they stayed within the fabrication limits.
However, the tools need to be informed about the relations and properties of the
interplaying components and materials and the processes to specify and fabricate them. The
reciprocal nature of structures and processes was - and could - not be well understood in the
beginning of the project. And while early design processes find usual a resort in “standard”
material specifications, these assumptions are void, when design demands a high level of
material behaviour and specification.
In the case of CNC knit this is obviously highly dependent on the limitations of the actual
CNC production in both size and data capacity, as the inner constraints of the knit
structure. These were not known early on in the tower project, which took hence resort in
small-scale prototypes with placeholder materials. This approach did not scale well, as the
material level is decisive for the hybrid structure. This is methodologically different to the
geometric form finding of pure membrane structures, where small-scale physical models
provide a sufficient point of departure. The developed CNC knit structure sufficed the
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
determined requirements on sample size, with the exception of the material stretch. The shift
to production level machines, the engagement with large knit structures with full details and
the assembly on these within a bending active structure proved to be challenging. Earlier full-
scale physical prototypes with the real materials would have enabled a better understanding of
the physically achievable material and system behaviour.
The necessary approaches to test and verify the bi-axial behaviour of the designed knit
and obtain a numeric base for simulation are currently experimental and time consuming and
it has to be emphasized, that the obtained “fictitious” elastic constants are results of
preliminary, but nevertheless quite practicable biaxial tests, which were modified for the
knitted material in the first time. In future, the test and evaluation procedures have to be
progressed in order to meet the requirements of the special knitted material. Furthermore, the
constitutive law used for the determination of the “fictitious” elastic constants is actually
limited to small strains which were not observed for the investigated knitted material.
The obtained values are sufficient to predict the overall structural capacity through FE
simulation. The simulation is however sensitive and difficult to set-up to give a timely
feedback to the design process. A coupling of the quick and stable form finding processes
with particle based systems and precise FE analysis are hence necessary.
Feedback between disciplines
The project exemplifies the problem, that highly integrated projects with interdepended
systems and scales pose. In the case of the tower design decisions were only possible in an
interdisciplinary dialogue. An exchange of all partners was necessary at a point in time, when
a design consists merely of ambitions and intentions - not requirements. This demands a
designerly conversation between disciplines, which might not be used to the wicked and
open-ended nature of design, the related symptoms of initial insecurity and the many loops
and iterations, which are necessary to from the vagueness into a secure territory. Fast cycles
between making, testing and specifying of material and structure are here an appropriate
means to solidify the design and finally engage the potential, which resides within bespoke
Potentials of CNC knit in hybrid structures
The benefits of CNC knit for The Tower were the integration of details into the knit. This
allows to collapse the complexities from fabrication into the material itself. The activation of
the material stretch allowed the knit furthermore to take on geometries, which would be
difficult to achieve with patches made from weave. The elasticity of the material challenges
however the construction and calculation of its behaviour. The knitted fabric shows in the
conducted test, as in the final tower a non-linear behaviour, were a considerable stretch has to
be introduced, before the knitted structure “locks” and is actually able to convey forces. This
behaviour is different from the usual woven PVC coated membranes. It meant, that the
stabilising pre-stress could only from the “locking” point onward be introduce into the
membrane. As the geometry of the restraining system and the pre-stress in the membrane
influence each other it was challenging to find their appropriate configuration in the pre-
construction phases.
M. Tamke, A. Deleuran, C. Gengangel, M. Schmeck, R. Cavalho, R. Fangueiro, F. Monteiro, N.
Stranghöner, J. Uhlemann, Th. Homm , M. Ramsgaard Thomsen
A better understanding of the actual forces in the tensioned areas would allow to enact on
these with a better fitting choice of the diameter of the bending active element or even with
local change of the knit structure. Such fabrication of membranes with graded material
properties is another promising area of CNC knitting. It has only be touched upon in this
project. We focused on the creation of reinforced zones in the membrane, a variation of knit
structure offers even the potential to influence the membrane shape through locally stiffer
textile material (Henrysson 2012). These processes require a complete understanding, control
and coordination of the different processes in all disciplines.
[1] Ramsgaard Thomsen, M., Karmon, A., Listener: a probe into information based material specification, in
Conference Proceedings Ambience 2011, Borås
[2] Lienhard J., Knipper, J., Considerations on the Scaling of Bending-Active Structures, International Journal
of Space Structures, Volume 28, Issue 3, page. 137-147, Multi Science Publishing December 2013
[3] Gengnagel C., Alpermann H., Lafuente E: Active Bending in Hybrid Structures, FORM RULE | RULE
FORM 2013, Günther H. Filz, Rupert Maleczek, Christian Scheiber (eds.), innsbruck university press,
2014, page 12 27
[4] Happold, E.; Liddell, W.I.: Timber lattice roof for the Mannheim Bundesgartensschau, The Structural
Engineer, No. 3, Vol 53, 1975
[5] Douthe C, Baverel O, Caron J-F, Form-finding of a grid shell in composites materials, Journal of the
International association For Shells and Spatial Structures:IASS,2006. VOL. 47, 2006.
[6] Ahlquist, S., Lienhardt, J.: 2013, Textile Hybrid M1 at La Tour de l´Architecte, in Tensinews, Vrije
Universiteit (VUB), Brussels, Belgium, 2013(1), pp. 6-9. (ISSN 1784-5688)
[7] H. Alpermann, C. Gengnagel: Shaping actively-bent elements by restraining systems, Conference
Proceedings IASS- APCS Symposium : From Spatial Structures to Space Structures, 2012 Seoul
[8] Alpermann H. and Gengnagel C., Restraining actively-bent structures by membranes. International
Conference on Textile Composites and Inflatable Structures, 2013.
[9] Nicholas P., Lafuente Hernandez E. and Gengnagel C., The Faraday Pavilion: activating bending in the
design and analysis of an elastic gridshell. Symposium on Simulation for Architecture and Urban Design,
[10] MSAJ/M-02-1995 Testing Method for Elastic Constants of Membrane Materials, Membrane Structures
Association of Japan, 1995.
[11] Uhlemann, J., Stranghöner, N., Saxe, K., Tensile structures: investigations into elastic constants'
determination of fabrics, Journal of the International Association for Shell and Spatial Structures 56
(2015), issue 1, pp. 25-35.
[12] Uhlemann, J., Stranghöner, N., Saxe, K., Stiffness parameters for architectural fabrics: an analysis of two
determination procedures, Structural Engineering International 25 (2015), issue 1, pp. 9-19.
[13] Uhlemann, J., Stranghöner, N., Schmidt, H., Saxe, K., Effects on Elastic Constants of Technical
Membranes Applying the Evaluation Methods of MSAJ/M-02-1995, Proceedings of the STRUCTURAL
MEMBRANES 2011, Barcelona, 2011.
[14] Henrysson E., Conceptual Design and Analysis of Membrane Structures. Chalmers University of
Technology, 2012.
... The production and erection of the demonstrator revealed a set of major challenges, all of which originate from a lack of knowledge and time in the practice to develop, test and verify the material [9]: ...
... The detailing in the knit of the tower needed as well improvement. The placement of the entrances to for the rods into the channels in the knit (a cavity made by interlock textile structure using front and rear needle beds) needed to be bigger and wider [9]. The developed solution allows not only for openings into the channels of nearly any sizes, but provides as well a different visual appearance. ...
Conference Paper
Full-text available
Membrane architecture uses currently off the shelf materials and produces the shapes and details through cutting and laborsome joining of textile patterns. This paper discusses investigations into an alternative material practice - knit - which engages bespoke membrane materials. A practice which allows for customised and graded material properties, the direct fabrication of shaped patterns and the integration of detailing directly into the membrane material. Based on two demonstrators built as hybrids of bespoke CNC knit and bending active GFRP rods this paper discusses the affordances and procedures, which this new practice of digital fabrication of membrane material requires. The central focus is set on the interaction between the involved disciplines and the emerging iterative process of design, material specification, prototyping, evaluation, (re-)design and (re-)specification. We discuss how design and engineering practices change, when material properties move from given and constant into the area of design and gradient.
... This customizability not only influences the membrane form, but also enables the integration of external structural elements. Such strategies can be found in the domain of CNC-knitted membrane installations, as seen in the Hybrid Tower and Isoropia by the Centre for Information Technology and Architecture (CITA) [9,10]. In these examples, fabric membranes were fully fashioned and integrated with continuous channels. ...
Full-text available
This paper describes a design-to-fabrication workflow for knit membrane tensegrity shells, a novel class of tensegrity structures that substitute discrete cables with a continuous machine-knitted membrane as the main tensile element. The workflow integrates (a) a simplified, simulation-driven design method for membrane tensegrity shells, (b) the conversion of the membrane geometry into digital inputs for Computer Numerical Control (CNC) knitting, and (c) the rationalization of the assembly connection details. This workflow begins by iteratively parameterizing reciprocal strut patterns, connecting the strut endpoints to form a mesh to represent the membrane, and implementing a dynamic relaxation algorithm to form-find the tensegrity shell geometry. The digital model then undergoes optimization procedures to negotiate between structural performance and fabrication constraints. Subsequently, the membrane geometry is extracted and converted into machine instructions to be CNC knitted with the following features: (i) integrated pockets to hold compressive struts at intended locations within the membrane, (ii) different types of yarn to create localized stiffer regions in response to stress concentrations imposed by the struts, and (iii) alterations of the shape of the membrane to adhere to the digital geometry. Several physical prototypes, including a 4-meter-diameter pavilion that was exhibited at the International Association for Shell and Spatial Structure’s Form&Force Expo 2019, serve as case studies for assessing the fidelity, benefits, and limitations of the proposed workflow.
... Recent projects have also introduced knitting on an architectural scale for multi-performative hybrid structures (Ahlquist and Menges 2013;Tamke 2015) or in combination with pneumatic actuators for reinforced and programmed knitted structures (Baranovskaya et al. 2016). Knitectonics (Chaturvedi et al. 2011) also tackles the question of programmed material distribution -with differentiated properties -using domestic craft tools, such as small circular flat weft knitting devices, conveniently hacked to be computer numerically controlled. ...
Full-text available
CITA Complex Modelling investigates the infrastructures of architectural design models. By questioning the tools for integrating information across the expanded digital design chain, the book asks how to support feedback between different scales of design engagement moving from material design, across design, simulation and analysis to specification and fabrication. The book conveys the findings of the Complex Modelling research project a five-year framing project supported by the Independent Research Fund Denmark. Undertaken at CITA, the Centre for Information Technology and Architecture, The Royal Danish Academy of Fine Arts, School of Architecture, Complex Modelling asks how new interdisciplinary methods for adaptive parametrisation, advanced simulation, machine learning and robotic fabrication can be orchestrated within novel workflows that expand the agency of architecture. CITA Complex Modelling privileges the ‘how to’ of research and the maker-based inquisitiveness of design creation and gives insight into the nature and depth of considerations incorporated into the bespoke workflows developed in each project. By foregrounding deeper design-based explanations of the underlying computational and material methods, the book reveals the complex network of processes that merge design agency with analysis, simulation with prediction, parametrisation with open topological modelling and the practices of physical prototyping, fabrication testing and troubleshooting that form the projects.
This paper shows a computational approach for knitting net shape preforms with bespoke 3D shapes and patterns. The approach takes partial knitting as the major shaping technique and as the fabrication constraints to generate multicolored pixel-based knitting maps based on given 3D meshes. The generation process includes five steps: 1 generation of wales, 2 generation of courses, 3 generation of 2D knitting maps, 4 stitch placement optimizations, and 5 pattern variations. At final stage, users can get a knittable 3D mesh with each face representing each stitch, as well as a 2D pixel-based knitting map. The knittable 3D mesh allows designers to further design pattern variations; the 2D knitting map can be directly used for generating knitting information in knitting software or easily followed by users.
Conference Paper
Full-text available
This paper shows a computational approach for knitting net shape preforms with bespoke 3D shapes and patterns. The approach takes partial knitting as the major shaping technique and as the fabrication constraints to generate multi-coloured pixel-based knitting maps based on given 3D meshes. The generation process include 5 steps: 1 generation of wales, 2 generation of courses, 3 generation of 2D knitting maps, 4 stitch placement optimizations, and 5 pattern variations. At final stage, users can get a knittable 3D mesh with each face representing each stitch, as well as a 2D pixel-based knitting map. The knittable 3D mesh allows designers to further design pattern variations, the 2D knitting map can be directly used for generating knitting information in knitting software or easily followed by users.
Full-text available
Knitted fabrics have unique characteristics compared to conventional architectural membranes. Conventional architectural membranes are composite materials consisting of woven or non-woven textiles and layers of coating. Characteristics of membrane behaviour are a high tensile strength, no bending resistance and a certain shear stiffness. Their resistance against environmental impact is derived from a form found shape with double curvature and prestress keeping the system under tension in all conditions. When combined to a hybrid system with an actively bent GFRP rod, the membrane stabilizes the rod, as the rod tensions the membrane. The doubly curved surface is usually subdivided into developable pieces and sewn together. Knitted fabrics are very flexible and have a high elasticity allowing for complex shapes and double curvature without the need of cutting them in special patterns. CNC-driven knitting-processes allow for integration of channels, pockets and structural details, potentially reducing the effort in further processing.
Conference Paper
This paper presents an enquiry into how to inform material systems that allow for a high degree of variation and gradation of their material composition. Presenting knit as a particular system of material fabrication, we discuss how new practices that integrate material design into the architectural design chain presents new opportunities and challenges to how we understand and create cycles of design, analysis, specification and fabrication. By tracing current interdisciplinary efforts in establishing simulation methods for knitted textiles, our aim is to question how these efforts can be understood and extended in the context of knitted architectural textiles. The paper draws on a number of projects that prototype methods for using simulation and sensing as grounds for informing the design of complex, heterogeneous and performative materials. It asks how these methods can allow feedback in the design chain and be interfaced with highly craft based methods of fabrication.
Full-text available
One important and challenging aspect of the design process for tensile membrane structures is the determination of biaxial material stiffness parameters. Coated textiles that are used as architectural fabrics display highly nonlinear and anisotropic stress–strain behaviour under biaxial tensile stresses. Nevertheless, in state-of-the-art structural analyses, the behaviour of these coated textiles is often simplified to a linear-elastic plane stress relationship, where the elastic constants are “tensile stiffness” and “Poisson's ratio.” The elastic constants must be determined for each material using biaxial tensile tests. Several different biaxial test procedures to determine the elastic constants exist worldwide, and these procedures yield various sets of elastic constants with a wide spectrum of stiffness parameters. At the same time, design engineers have no guidelines to assess which set of parameters is appropriate for a specific design situation. This paper compares two different methods of determining stiffness parameters using theoretical and experimental analyses. The variation in structural analysis results due to stiffness parameters that were determined using different techniques is demonstrated using three types of PES/PVC materials from two material producers. Furthermore, this paper provides guidance regarding the manner of evaluation of sets of elastic constants and the modification of the evaluation of experimental biaxial tests, if required. Information is also provided concerning the applicability of the investigated procedures.
Conference Paper
Full-text available
In actively-bent elements the curved end-geometry is achieved by elastic bending. The shape depends of the boundary conditions and its curvature- and stress-distribution is normally not constant. Local stress maxima restrict the design-load of the elastically bent element. Restraining systems can be used to optimise the shape of the elastically-bent elements to get a continuous curvature and therefore a homogeneous stress-distribution. In addition they improve the stiffness of the overall hybrid system significantly. Beside a structural optimization restraining systems can as well be used to shape the bent element for architectural reasons.
Conference Paper
Full-text available
Membranes can be used to restrain actively-bent structures. The paper explains the structural behaviour behind using membranes to couple actively-bent elements and analyses the influence of different parameters.
Conference Paper
Full-text available
Active Bending describes curved beams or surfaces that base their geometry on the elastic deformation of initially straight elements. The bending stress in the bent elements depends on the Young‘s modulus of the material, the height of the cross section and the curvature. To achieve sufficient curvature the height of the cross section and the Young‘s modulus need to be small. Therefore, the stiffness of the lastically-bent elements is small. Additional elements offer the opportunity to create a hybrid structure, whose stiffness is significantly higher than the stiffness of the elastically-bent elements alone. The general influence of Active Bending on stiffness and stress is discussed and numerically analysed. The advantages of and different ways to use elastically bent elements are presented in five case studies. The differences in their structural behaviour compared to curved elements are analysed. It can be shown that, despite the initial stresses and the limitation on the profile’s choice, the structural capacity of a given construction is hardly affected by the elastic bending, which is therefore advantageous over the use of straight elements in terms of fabrication, transport and/or assembling.
Conference Paper
Full-text available
This paper reports the architectural and engineering design, and construction, of The Faraday Pavilion, a GFRP elastic gridshell with an irregular grid topology. Gridshell structures are self-formed through an erection process in which they are elastically deformed, and the prediction and steering of this aspect becomes a central part of both architectural and engineering design processes. While there are existing architectural approaches to determining the geometry of other kinds of form-active structure, as well as new engineering approaches to the simulation of bending active structures, a fast and light-weight design approach to bending active structures is not currently developed. In this paper, we introduce an approach to the architectural design of a bending active structure whereby the shell form and grid topology are determined by simulation. Particular features are that the grid topology is not pre-described, but rather emerges as part of the simulation, and that different calculative models relating to the material, element and structural scales are solved and synthesized by extending the technique of dynamic relaxation. Secondly, the results of this design modelling are provided as the input for a quantitative engineering simulation, where the equilibrium shape and load-bearing capacity of the elastically-bent structure is calculated with a non-linear, three-dimensional finite element model within the FEM-package Sofistik. We compare the advantages and disadvantages of both methods, which while complementary have important differences relating to the interaction with the design of the structure, differences in the definition of supports, connections and elements, the speed of calculation and the magnitude and precision of the results.
Full-text available
Bending-active structures are composed of curved beam or shell elements which base their geometry on the elastic deformation of an initially straight or planar configuration. In bending-active structures the moment of inertia has a direct influence on the residual stress and is therefore limited by a given minimal curvature in the system and the permissible bending stress capacities of the chosen material. These interdependencies may lead to a scaling issue that limit bending-active structures to a certain size range. This range may be widened if the system's reliance on elastic stiffness is compensated by other stiffening factors such as coupling of structural elements and stress stiffening effects. This paper will analyse the scaling effects in bending-active structures. Some basic systems are studied by means of dimensional analysis and FEM parameter studies to clarify at which power each influencing factor effects scaling. Based on these findings some more complex structures are studied for their scalability. This will offer the basis for some more general conclusions.
Full-text available
SUMMARY The advantages of the use of glass fiber composites for grid shell are presented. The shape of grid shells results from a post-buckling state of tubes. To bypass the difficulty to predict the geometry of the final equilibrium state, the large rotations which occur during the erection process are modelled using the dynamic relaxation algorithm. This paper proposes an adaptation of this method for structures prestressed by bending through the development of a computer program. It includes the validation of this numerical tool through comparisons with a finite elements software. Then an application to the form-finding of a grid shell and the study of its stability under standard loading conditions will be presented. Finally the authors conclude on the technical and economic feasibility of this composite grid shell.
The aim of the present paper is to provide a state of the art report of currently published biaxial tensile test procedures together with correlated evaluation procedures. The investigations show a great spectrum of resulting fictitious elastic constants (tensile stiffness and Poisson's ratio) that can be determined for one fabric material on the basis of the different established procedures. In the analysis of tensile membrane structures this range of fictitious elastic constants results in a spectrum of computed stress and deformation values for one and the same structure and load case. The great range of structural analysis results due to that variety of possible stiffness parameters is demonstrated for one exemplary membrane structure with different magnitudes of curvature. Furthermore, this paper provides guidance regarding how to evaluate sets of elastic constants and how to modify the evaluation of experimental biaxial tests in order to minimize the uncertainties left to the design engineer. Information is also provided concerning the applicability of the investigated procedures. Copyright
This paper describes the engineering design of the shells constructed by lifting a flat square lattice into a doubly curved shape. The authors attempt to show how the loads were defined, how the structure was modeled and tested both physically and mathematically, and how these models were used to determine the construction details.