Content uploaded by Norman Hack
Author content
All content in this area was uploaded by Norman Hack on Sep 21, 2015
Content may be subject to copyright.
N. Gu, S. Watanabe, H. Erhan, H. Haeusler, W. Huang (eds.), Rethinking Comprehensive Design: Specu-
lative Counterculture, Proceedings of the 19th International Conference of the Association of Computer-
Aided Architectural Design Research in Asia CAADRIA 2014, 000–000. © 2014, The Association for
Computer-Aided Architectural Design Research in Asia (CAADRIA), Kyoto, JAPAN
MESH MOULD: DIFFERENTIATION FOR ENHANCED
PERFORMANCE
NORMAN HACK¹, WILLI VIKTOR LAUER², FABIO GRAMAZI-
O¹, MATTHIAS KOHLER¹
¹ETH Zurich, Zurich, Switzerland, ²SEC, Future Cities Laboratory
hack@arch.ethz.ch, lauer@arch.ethz.ch, gramazio@arch.ethz.ch,
kohler@arch.ethz.ch
Abstract. Mesh-Mould is a novel robotic fabrication system for com-
plex, non-standard concrete structures. The system folds together
formwork and reinforcement, the two most labour intensive aspects of
concrete constructions and offers an alternative approach to the cur-
rent modes of prefabrication by suggesting an in-situ fabrication pro-
cess (Figure 1). The paper outlines the development of the Form-
work/Reinforcing Meshes through several iterations of physical and
digital tests. Initially starting from simple triangulated 3D lattices, the
structures evolved to become more complex and differentiated. The
incorporating of flow enhancing ducts and surface perimeters with di-
verse surface aperture densities facilitates an optimal concrete flow
and material distribution within the mesh.
Keywords. Robotic fabrication, concrete formwork, differentiation,
spatial extrusion.
1. Introduction
Recent advances in automation technology and computational power offer
the opportunity to take a new perspective towards digital fabrication on the
scale of construction. In terms of hardware, the industrial Robot, a generic
and customizable machine, has become a cheap and reliable off-the shelve
product, applicable for a wide range of processes. The ability to be equipped
with almost any tool makes it particularly interesting for the application in
the domain of architecture. With the rapid development of computational
power, it is now possible to integrate multiple sensing devices, and enable
real-time on-site coordination of mobile robotic units, even in the complex
and ever-changing environment of the construction site (Helm, 2012). An
2 N. HACK, W. LAUER, F. GRAMAZIO AND M. KOHLER
increasing number of research institutions have recognized the robot’s poten-
tial for the realization of a new, Digital Materiality (Gramazio and Kohler,
2008) and innovative Material Ecologies (Oxman, 2013). Recently this led
to a global collective endeavour in an emerging field of fabrication-
informed, non-standard architecture. On the downside however, the use of
small machines imposes constraints that are related to the limited payload
capacity of standard industrial robots. At first glance the requirements for a
robotically fabricated material system for loadbearing constructions are
somehow conflicting. The material handled by the robot needs to be small
and light, which is usually not the case for loadbearing structures. Conse-
quently construction processes have to be redeveloped in order to match the
abilities of the machine and to fully capitalize on the robot’s potentials.
Figure 1. Conventional Formwork and Mesh Mould
2. Mesh-Mould, Concept and State of the Research
Based on the finding that a construction process for robots needs to be light-
weight, the decision was taken to focus on the robotic fabrication of concrete
formwork. Instead of having the robot transfer the entire mass of the build-
ing it will be used to define the shape of the structure, and take over those
tasks that are coordinatively complex, highly labour and cost intensive. One
specific formwork system called leaking formwork was found to be particu-
larly interesting and offered great potential to be adapted for a robotic pro-
cess. Its basic principle works as follows: Concrete is poured into a perforat-
ed formwork, which is built up from flat plastic panels. The concrete
protrudes through the perforations and covers up the panels. In a final step
the surface gets manually troweled leaving behind a smooth concrete surface
(Figure 2)
MESH MOULD, DIFFERENTIATION FOR ENHANCED PERFORMANCE 3
Figure 2. Forma–tech formwork system
This simple and efficient material system holds a great potential when cross-
bred and augmented with the logics of robotic fabrication (Hack and Lauer,
2013). A custom designed extruder head was developed and attached to a
mobile robot in order to build up porous 3Dimensional mesh structures by
spatially extruding the polymer directly on the building site (Figure 3).
Figure 3. Mobile Robot extruding mesh
3. Two Research Strands: Strength and Morphology
At this point of time the project pursuits two parallel strands of research. The
first is an investigation into improving the tensile capacity of the meshes, in
order to activate them as reinforcement. Here, the development of a suitable
polymer, as well as the change to a metal based process is considered. The
second strand focuses on the morphological differentiation of the meshes
with the goal to improve aspects such as structural integrity, concrete flow,
fabrication time and material use.
4 N. HACK, W. LAUER, F. GRAMAZIO AND M. KOHLER
4. Morphological Optimization of the Meshes
The following section retraces the development from simple triangulated 3D
lattices towards more differentiated spatial structures in accordance to the
diverse requirements and pressures acting upon these meshes. The require-
ment to differentiate the meshes has been an assumption that was confirmed
by the first concrete experiments. Due to the high complexity and the mani-
fold of interacting parameters the differentiation of the mesh structures hap-
pened through several iterations of physical experiments, rather than through
digital simulation. More generally the diverse, partly conflicting partly over-
lapping requirements can be classified into the following domains.
1. Optimized flow of concrete: On one hand the structure and aperture
size of the mesh need to be sufficiently dense in order to keep the
concrete from bleeding out, on the other hand it needs to be wide
enough so that the concrete can evenly flow around the structure
without causing congestions or voids.
2. Structural considerations: The four load cases that influence the
meshes are firstly the dynamic loads acting on the mesh during the
process of pouring, secondly the hydrostatic pressure building up in-
side the formwork as long as the concrete is in its liquid state, thirdly
the wet load of the concrete before curing and lastly the tensional
loads acting on the element after the concrete has cured and the ele-
ment is structurally loaded.
3. Material use and fabrication time: The mesh topology, density and
the extrusion path have a significant impact on the amount of mate-
rial that is used and the time that is needed to fabricate the meshes.
The following experiments started with simple, undifferentiated meshes de-
scribing complex surface geometries. The following concrete tests gave an
impression about the behaviour of the mesh under wet load, while succes-
sively more complexity was discovered when examining the rheological be-
haviour of the concrete within the mesh.
MESH MOULD, DIFFERENTIATION FOR ENHANCED PERFORMANCE 5
5. Experiments
5.1 Mesh and concrete calibration
5.1.1 Doubly curved surface with generic 3D triangulation
The main objective of the first prototypes was to proof that a spatial extru-
sion even for complex double curved surfaces is achievable. The concrete
for these experiments was not applied by pouring it inside the mesh and let-
ting it protrude to the outside but rather, as in classical ferrocement applica-
tions, by pressing a fairly viscous mortar from the outside through the mesh,
successively densifying the interior. Even though this strategy proofed suc-
cessful it was considered too labour intensive for larger building elements.
After applying 35 kg of wet concrete no mesh deformations were recogniza-
ble, even though the mesh itself accounts for only 1/100th of the wet load of
the concrete (Figure 4)
Figure 4. First mesh prototype filled with concrete
5.1.2 Pouring Test cubes
The objective of the following experiments was to test the same mesh typol-
ogy, but with filling the concrete in from atop. The tests were conducted in
order to find an appropriate relationship of concrete viscosity and mesh aper-
ture size as well as to examine the rheological behaviour of concrete within
the mesh. Several test cubes with an edge length of 15 cm and various aper-
ture sizes have been tested with a specific concrete mix, once with the addi-
tion of fibres, once without. The admixture of fibres resulted in clogging
within the dense interior structure. A mix without fibres worked best at a
slump flow of 19 cm measured with the Haegermann mini cone. Generally
the Interior structure was regarded as too dense in relation to exterior struc-
ture, not allowing the concrete to flow evenly around the structure (Figure 5)
6 N. HACK, W. LAUER, F. GRAMAZIO AND M. KOHLER
Figure 5. Test cubes after curing
5.1.3 Pressure columns
Directly following, and building up on the previous experiments, a third se-
ries aimed to test the structural integrity of the meshes during the process of
pouring. It is examined if the structures can withstand the high pressure of
the wet concrete, and if the meshes are actually capable of holding back the
wet concrete with increasing pressure from above. A series of test cubes,
again with an edge length of 15 cm and varying number and sizes of aper-
tures was placed at the bottom of one 15 x 15 x 150 cm pressure column. A
specified concrete is then filled into the column from the top. The maximum
possible infill height is measured before the mesh cannot hold back the con-
crete, or breaks due to over loading.
The experiments showed that none of the meshes broke due to the weight of
the concrete. However, despite the high pressure, the interior structure of the
meshes was too dense. Concrete without fibres and a slump flow of approx-
imately 19 cm worked well up to the maximum infill height. A combination
of the 3Dimensionality of the mesh and the thixotropic effect of the concrete
sealed the mesh surface immediately after initial troweling, allowing to fill
in more concrete from the top without any concrete leaking out.
5.1.4 Findings of calibration experiments
The experiments gave an insight about the required concrete and mesh pa-
rameters. With regards to the concrete, a combination of an aperture size of
approximately 30 x 17 mm and a slump flow of 19 cm was found to work
best. The concrete should stay at a specific slump flow for the duration of the
pouring process. The thixotropic effect of the concrete is highly beneficial to
the leaking formwork system, and could further be stimulated by adding
thixotropic agents. However the general finding was that the inner structure
was too dense compared with the outside structure. The assumption is made
that through reducing the inner flow resistance of the meshes a better distri-
bution of the concrete and less honeycombing can be achieved.
MESH MOULD, DIFFERENTIATION FOR ENHANCED PERFORMANCE 7
5.2 Mesh refinement and differentiation
Based on the findings of the previous experiments a new series of differenti-
ated meshes was developed, aiming to improve the structural integrity, en-
hance concrete flow and material distribution within the meshes and to re-
duce fabrication time and material consumption. These goals can be
achieved by a variety of means and strategies. The reduction of flow re-
sistance in the interior, for instance might enhance the concrete flow, reduce
fabrication time and material use but at the same time weaken the structural
capacity of the mesh. Like in many multi-parameter optimization tasks, here
the goal is to find a well-balanced solution for the various, competing pa-
rameters. The samples measure 80 x 30 x 15 cm and represent a section of a
single curved wall (Figure 6). The various measures taken to achieve the
aforementioned goals are listed below:
SOS1: This mesh resembles the original, undifferentiated pattern and acts as
a sample to compare the differentiated meshes with.
DOD: The pattern works with a reduced interior structure. Instead of cross
connecting each point, the interior pattern follows a trapezoidal connection,
leaving more space for the concrete to flow through.
2D2D: The interior pattern cross-connects only every 4th point, leaving the
interior sparse. The outer surfaces are not flat anymore, but are folded to
form beams with increased structural depth.
SOS_Ducts: The interior structure follows a rectangular pattern with straight
instead of triangular connections. Ducts for filling in the concrete are intro-
duced. The missing cross connections are expected to cause higher hydrostat-
ic pressures when the concrete is filled in; therefore the pattern on the outside
of these ducts is densified.
DD_Ducts: The density of the interior structure is reduced by cross connect-
ing only every other point. At the location of the ducts the outer surfaces are
folded and the pattern is densified.
DD_Ducts_Hex: The cross connections are kept dense, the pattern on the
outside surfaces is sparse and follows a hexagonal pattern, where the ducts
appear, the outer surfaces are folded for more structural integrity.
8 N. HACK, W. LAUER, F. GRAMAZIO AND M. KOHLER
Figure 6. From left to right: SD, DOD, 2D2D, SOS_Ducts, DD_Ducts, DD_Ducts_Hex
The fabrication time for each sample is taken by stopwatch and the amount
of used material is determined by the mesh’s weight. The ratio of material
used for the perimeter defining surfaces and inner structure is determined by
the quotient of the path-lengths and describes where the flow resistance is
higher. Structural tests are undertaken for two different load scenarios: a
combination of compression and shear force as well as a 3-point bending test
(Figure 7). Since the samples are not representing the actual material (see
chapter 3) these results solely compare the geometry related structural per-
formance. A beam with the dimensions of 50 x 5 x 10 cm is fabricated for
each of the six patterns and deflections are measured under the influence of a
constant force. The performance of concrete flow inside the mesh is evaluat-
ed by sectioning the cured sample and measuring area proportion of voids.
The leakage behaviour, defined as the amount of concrete protruding from
the inside to the outside is measured according to a rating system, whereas 0
means no protrusion and 10 complete leakage. Consequently a 5 marks the
best performance.
Figure 7. SOS_Ducts 3point bending and shear-pressure deflection
MESH MOULD, DIFFERENTIATION FOR ENHANCED PERFORMANCE 9
Table 1. Performance evaluation of the different mesh typologies
At this point of time the samples are not yet evaluated upon their rheological
behaviour (concrete flow and leakage). Due to the limited availability of a
concrete laboratory these tests had to be shifted back towards the beginning
of January 2014.
6. Conclusion
In biological systems differentiation, adaptation and specialization are
known to effectively contribute to the performance and the survival of an or-
ganism. However the current modes of industrial production, do not allow
realizing these concepts economically in the context of construction. The
costs to fabricate a material efficient geometry that is well adapted to a spe-
cific load case, usually exceeds the savings in material. The presented re-
search demonstrated that with the proposed robotic fabrication system, case
specific differentiation becomes a feasible strategy also in the domain of ar-
chitecture. Whereas the paper highlighted the benefits of differentiation on
the scale of the meshes, it has a much larger influence, ranging from the ma-
terial level to the urban scale. The liberation from a construction paradigm
based on repetition and simplification implies that it will be possible to de-
sign and construct a more sustainable, material efficient and diverse urban
environment. However at its recent stage the research relies on the results of
the simultaneously ongoing research on the material-scale. Only a material
Weight
per sam-
ple
Fabrica-
tion time
per sam-
ple
Ratio In-
ner to
outer
structure
3 point
bending
cm; 1 kg
load
Deflection
shear and
pressure
cm, 2kg
load
SD
649 gr
9h 3min
1.91
0.8 cm
X =0.2 cm
Y = 0 cm
DOD
352 gr
5h 36min
0.53
7.5 cm
X= 3 cm
Y=10cm
2D2D
446 gr
7h 18min
0.21
1.5 cm
X=2 cm
Y=11cm
DD_Du
cts
418 gr
6h 33min
0.63
2 cm
X= 0.2 cm
Y=0.6 cm
SOS_Du
cts
464 gr
6h 51min
0.81
8 cm
X= 1.6 cm
Y=0.6 cm
DD_Du
cts_Hex
538 gr
7h 42min
1.23
9.5 cm
X= 0.5 cm
Y=0.1 cm
10 N. HACK, W. LAUER, F. GRAMAZIO AND M. KOHLER
with high tensional capacity allows to fully capitalizing on the freedom that
is gained through robotic control.
6. Next steps
The evaluation of the samples will be completed shortly by conducting the
missing concrete pouring experiments. In a following step new prototypes,
with the dimensions of approximately 2 x 1 x 0. 2 meters will be fabricated
on the basis of the most successful spatial patterns. These prototypes are ex-
pected to deliver more insights about the behaviour on the real scale of con-
struction. The differentiation studies and the parallel research about improv-
ing the tensile capacity will converge in the first half of 2014. With the
integration of the actual, tensile active material a new series of prototypes
can be evaluated upon their real structural performance. A higher capacity to
compensate for tensile forces moreover allows testing gradual transitions
from vertical surfaces to horizontal planes, as tensional forces particularly
occur at these transitions. Finally the development of a holistic architectural
demonstrator will take place in 2015 and a 1:1 prototype in form of a small
architectural intervention will be fabricated.
Endnotes
1. The abbreviations describe the pattern of the interior structure. ”S” stands for a
“straight” connection e.g. from the first point of the right perimeter to the first point of
the left perimeter. “O” describes an “On beam” connection, e.g. for two consecutive
points on the same perimeter, while “D” represents a diagonal connection of two
points, e.g. the first point on the left perimeter and the second on the right perimeter.
Acknowledgements
We would like to thank Ena Lloret Christensen for her valuable help with the concrete exper-
iments, Heinz Richner for his support with the concrete Lab in Zurich, Prof. Dr. Robert Flatt
for sharing his knowledge and infrastructure and Dr. Norman Blank from Sika for the intense
discussions. Furthermore we would like to thank Dr. Silke Langenberg, Volker Helm, Dr. Jan
Willmann and Michal Budig for their steady support from Zurich and Singapore.
References
Gramazio, F. and Kohler, M., 2008. Digital Materiality in Architecture.: Lars Müller Publish-
ers.
Hack, N. and Lauer, W., 2013. Overcoming Repetition: Robotic Fabrication Processes at a
large Scale. International Journal of Architectural Computing , 11(3), pp. 286-299.
Hill, D., 2003. Building Component. USA, Patent No. WO 03/040487 A1.
Oxman, N., 2013. Towards a Material Ecology. San Francisco, Association for Computer
Aided Design in Architecture .
Helm, V., 2012. Mobile Robotic Fabrication on Construction Sites: dimRob. Vilamoura,
IEEE, pp. 4335 - 4341.
