Large Scale Testing of Digitally Fabricated

Concrete (DFC) Elements

Freek Bos1(&), Rob Wolfs1, Zeeshan Ahmed1, and Theo Salet1,2

1
Eindhoven University of Technology, Eindhoven, The Netherlands
[email protected]
2
Witteveen+Bos Consulting Engineers, Deventer, The Netherlands

Abstract. Case study projects based on Digitally Fabricated Concrete

(DFC) are presented in an increasing pace around the globe. Generally, though,
it is not reported what structural requirements (if any) these structures meet and
how compliance to these requirements was established. Published material
research is often not connected to the presented case studies, and even when it
is, it is not necessarily obvious their small scale results can be applied to full
scale structures as some scale effects should be anticipated. Caution is required
as DFC related material tests are still under development and scale effects in
DFC have hardly been studied. Therefore, it is recommendable to perform large
scale testing, in the range of 1:5 to 1:1, if DFC is applied to actual use structures.
This paper presents such testing for two projects, a pavilion in Denmark (not
realized) and a bridge in the Netherlands (realized). In both cases, elements
printed with the 3D Concrete Printing facility of the Eindhoven University of
Technology were intended for actual load bearing performance. The conserva-
tive designs past the test requirements, but nevertheless some important findings
with regard to element manufacturing and structural behaviour were experi-
enced. It is concluded that large scale testing remains advisable for DFC
structures as long as not all relevant aspects of the technology are quantitatively
understood, at least when new concepts are being applied.

Keywords: 3D Concrete Printing (3DCP)
Experimental testing

Scale effect
Additive manufacturing

1 Introduction

Case study projects based on Digitally Fabricated Concrete (DFC) are presented in an
increasing pace around the globe, as shown by regular updates on websites and
databases such as www.3ders.org and www.am4ae.com. Generally, though, it is not
reported what structural requirements (if any) these structures meet and how compli-
ance to these requirements was established.
Meanwhile, material research into structural properties of DFC and its dependen-
cies on process specific characteristics such as anisotropy, is also enjoying rapid growth
as evidenced by recent publications [1–3]. However, published material research is
often not connected to the presented case studies, and even when it is, it is not
necessarily obvious their small scale results can be applied to full scale structures as

© RILEM 2019
T. Wangler and R. J. Flatt (Eds.): DC 2018, RILEM Bookseries 19, pp. 129–147, 2019.
https://doi.org/10.1007/978-3-319-99519-9_12
130 F. Bos et al.

some scale effects should be anticipated. Ahmed et al. [4] have already reported initial
findings of the effect of scaling, but they focused on design and construction aspects,
rather than structural performance.
The development of DFC has entered into a hazardous phase. Without a proper
understanding of structural properties and scale effects, the worldwide enthusiasm for
the technology could result in structures that have not been thoroughly proven. If that
were to result in a collapse, that could endanger the (speed of the) development of the
technology as a whole, in addition to the effects on an individual project. As long as
these uncertainties exist, large scale testing of structural elements is an essential tool
that should be applied to prove their validity and avoid structural failures. In Europe,
the Eurocode 0 (in particular Appendix D) [5] provides the legal framework for this
approach, designated as ‘design by testing’.

2 TU/e 3D Concrete Printing and Scale Effects

For two reasons, caution is required when applying results of small scale material tests
on full scale DFC structures. First, these tests themselves are still under development.
In particular, the range of process settings in which test results are valid, is largely
unknown (e.g. environment temperature, pump pressures, interlayer interval time, etc.).
It is thus also unclear which settings need to be recorded. Current studies focus on the
effects of material composition, directional dependency of structural properties (ani-
sotropy), and interlayer interval time. However, many more parameters and settings
could have an influence on the (consistency of) structural properties, such as temper-
ature, post-mixing and friction in the wet phase transport system, print speed, nozzle
height, and so on.
Second, the effect of scale has hardly been investigated. It is already known that the
strength of concrete is scale dependent due to the strain energy release dependency of
developing micro cracks. Shrinkage and creep are also well known phenomena to occur
in concrete that are in themselves not scale dependent, but rather have a more pro-
nounced effect on larger structures. Furthermore, it should be noted that printable
mortars often contain few large aggregates and a high cement content, which con-
tributes to relatively high shrinkage and creep.
In DFC, however, more effects could influence the (distribution of) local material
properties, and thus also the structural response of elements or structures (although not
all of them are necessarily negative). Some of these include:
– effects of DFC system operation, e.g. increasing system temperatures in use,
(partial) flow obstruction by hardening particles, variance in nozzle-to-object dis-
tance in high object prints,
– effects of object geometry, e.g. pressure caused by self-weight of an object, inter-
layer interval time,
– effects of design, e.g. interactions with other materials and joints,
– statistical effects, e.g. due to a varying quality of print material, or scatter in
interlayer strength.
 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 131

A systematic inquiry into these potential scale effects falls outside the scope of this
study, but two projects are being discussed that have been designed, printed, and tested
around the same time in 2017 (June–September):
– a pavilion for the municipality of Nyborg, Denmark (not realized),
– a bicycle bridge in Gemert, the Netherlands (realized).
Both projects were printed with the 3D Concrete Printing (3DCP) facility of the
Eindhoven University of Technology (TU/e) and a custom designed print mortar
described by Bos et al. [6]. The print mortar has been named Weber 3D 115-1. It was
used for the printing of both projects. Initial structural material properties were
experimentally determined by Doomen [7] and Slager [8], and are listed in Table 1.
Directional dependency and interface interval time were taken as research variables.
Characteristic strengths were calculated by Witteveen+Bos consulting engineers from
the average results and sample size, and are given in Table 2.

Table 1. Experimentally determined structural properties of Weber 3D 115-1 print mortar. The
tensile strength in w-direction is valid for an interlayer interval time of up to 2 h approximately.
For the directional dependency, a relative orientation of axis u, v, w is used [6], indicating the
direction parallel to the print in the horizontal plane, perpendicular to the print direction in the
horizontal plane, and vertically perpendicular to the print direction (or parallel to the robot arm),
respectively.
Property Dir. Age Symbol Value
Density 28 days q 2,000 kg/m3
Modulus of elasticity 28 days E 19,000 MPa
Average compressive strength u 28 days fcm,u 23.2 MPa
v 28 days fcm,v 21.5 MPa
w 28 days fcm,w 21.0 MPa
Average tensile strength u 28 days ftm,u 1.9 MPa
v 28 days ftm,v 1.6 MPa
w 28 days ftm,w 1.3 MPa
Creep factor 56 days u 3.0
Shrinkage 56 days e 1.5

It should be noted that at this point experimental sample sizes are still relatively
small. As a result, the characteristic strengths are quite low compared to the average
experimentally determined values.
Even though some variable dependencies have been established, experiences with
printing sessions have learned that several more exist which have not yet been studied
quantitatively, such as temperature (ambient, system), print speed, and the length of the
hose between the pump and the print nozzle. This is an important reason not to be
satisfied with material tests only, but to use large scale tests to validate assumptions
about structural behaviour.
132 F. Bos et al.

Table 2. Characteristic strength values calculated by Witteveen+Bos consulting engineers,

based on average experimental values and sample size. Separate values were determined only for
the u- and w-directions as they are the most distinctly different and most relevant for the
structural response.
Property Dir. Age Symbol Value
Characteristic compressive strength u 28 days fck,u 13.6 MPa
w 28 days fck,w 10.1 MPa
Characteristic tensile strength u 28 days ftk,u 1.48 MPa
w 28 days ftk,w 0.85 MPa

3 Nyborg Studio
3.1 Project and Design
A studio-type structure was planned in the town of Nyborg, Denmark, as part of a
larger construction project for student housing. The Sketch Design by architect I.
Moltke, shown in Fig. 1, featured 3 closed, curved wall elements that were to be
connected with straight timber frames containing the windows and doors. The curved
walls would be manufactured through off-site 3D Concrete Printing, with the TU/e
facility, as this was expected to allow better process and quality control than on-site
printing.

Fig. 1. Sketch design of Nyborg Pavilion by I. Moltke.

The principle of the wall design was elaborated by the architect, structural engineer
Witteveen+Bos and the TU/e research team. Probably the most determining factor was
the stringent insulation requirements. Danish building codes require an R-value of
R = 10 m2K/W, which ruled out an internal zig-zag concrete pattern or solid cast
concrete, as is seen in similar projects [9–11]. Rather, the wall section design consists
of a printed inner and outer face with a thickness t = 60 mm, corresponding to 1
filament. The faces, following a sine curve pattern for architectural effect and to
 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 133

increase object stability during printing, are separated by a 305 mm wide cavity that is
filled with sprayable PUR insulation foam. The faces are connected by GFRP wall ties
(Refus 1 Wall Tie, 365 mm/14”), spaced 285 mm apart horizontally (1 period of the
sine) and 300 mm vertically (30 print layers; effectively corresponding to 11.7 pcs/m2),
to transmit axial loads from one face to another.

3.2 Test Program en Specimens

Subsequently, a test phase with large scale elements was determined and conducted.
The structural principle of the wall is simple: the inner and outer faces are joined by the
wall ties with hinged connections that distribute the acting horizontal loads (Fig. 2).
Thus, the faces may be loaded in flexural tension under wind load or collision. Since
the faces are not joined with shear connections, the faces are slender when loaded in
compression and may be prone to buckling and subsequent flexural tension failure.
Considering the structural design and expected load cases, three tests were considered
appropriate: a compression/buckling test, a bending test, and an impact test.

Fig. 2. Structural principle of wall design (vertical section).

The final pavilion geometry was yet unknown when the test program started.
Because of the intended due date, the design and test phases partly developed in
parallel. The test specimens, nevertheless, had to be designed so that they would yield
useful data. As it was clear the three wall elements would each be different and testing
complete mock-ups of them would be difficult anyway, it was decided to test
approximately 750 mm wide straight segments of the wall section design. An impor-
tant advantage is that results would be relatively easy to understand, calculate, and
compare. Furthermore, in terms of structural response, testing straight segments was a
conservative approach as the overall wall curvature would have a positive effect on the
structural resistance.
The test specimens were obtained from printed ‘rectangular’ sections of 2.5 m
height (Figs. 3 and 4) (the maximum print height of the TU/e gantry facility is 2.5–
2.8 m, depending on the print table design and the print head equipment that is being
134 F. Bos et al.

used). The short sides of the ‘rectangles’ were to be removed by sawing them of, to
obtain a typical wall segment. The use of automatically embedded cable reinforcement
[12, 13] was considered for the final design of the pavilion. However, it was not
included in the test specimens as the direction of the cables would not influence their
response in the test but would complicate the printing and specimen preparation
process.

Fig. 3. Horizontal section of test element.

3.3 Printing
The specimens were printed on June 6th and 7th, 2017. To achieve the filament section
of 10  60 mm2 (instead of the standard 10  40 mm2 section), a customized nozzle
was made that also enabled the application of cable reinforcement later in the project.
This nozzle features a back-printing opening to allow proper placement of the cable in
the filament, rather than the previously applied down-printing opening at the bottom of
a nozzle (see further discussion in Sect. 4). The print speed was reduced by 20% to
80 mm/s to account for the larger filament section. The print time for one object was
approximately 2 h.
The strength and stiffness development of the applied print mortar Weber 3D 115-1
was insufficient to print to a height of 2.5 m, without collapse of the dormant mortar.
Therefore, an accelerator (weber.ad snel) was dissolved in the water that was used to
mix with the mortar. This reduced the setting time of the concrete from 2–3 h to 15–
30 min, which was expected not to negatively influence the filament interlayer bond,
but enough to allow stable printing.
During and after printing, the object was sprayed with a curing agent (BMP Curing
Compound AC) to avoid dehydration of the object. Due to its size, the usually applied
method of wrapping in foil was unpractical. Two objects were printed to their intended
 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 135

Fig. 4. Test element during printing

height of 2.5 m. The printing of the third specimen had to be cancelled at 1.33 m height
due to unexpected inconsistencies in the filament. After initial curing, the insulation
PUR was sprayed into the cavity in a number of stages over several days. The sides
were removed on June 28th, 2017, with a hand-held circular concrete saw operated by
an external professional.

3.4 Observations Before Testing

In pre-test observations of the specimens, several issues were noted. First of all, the
layer thickness t was 70 mm over almost the entire height (except approximately the
final 10 cm), thus exceeding the designed filament width of 60 mm. This difference is
attributed to the pressure of subsequent concrete layers that are being added while the
concrete is still in dormant state.
Furthermore, one week before testing and before the short sides were removed by
sawing, vertical cracks were observed in the short sides of each of the three specimens.
These had not been there before, and must probably be attributed to shrinkage of the
PUR insulation caused by gas exchange. Test samples from a simultaneously running
MSc project [14] show this effect very clearly (Fig. 5). The 1.3 m element suffered 1
vertical crack on the right-side. The 2.5 m elements both suffered a vertical crack on
each short side. Not all cracks extend over the full height, but over a very significant
part nevertheless. However, since all cracks occurred in areas that were removed before
testing, they did not influence the results.
After the short sides were removed, it was also discovered that the test elements
featured several horizontal cracks through the layer interfaces. It is suspected these are
caused by peak stresses around the wall ties that occur during curing of the concrete, in
combination with a low interface adhesion caused by the backward printing nozzle
(further discussed in Sect. 4).
136 F. Bos et al.

Fig. 5. Shrinkage of PUR foam causing fracture in printed concrete (from [14]). The long edges
have broken and moved inward, which seems to indicate a force pulling them in that direction.
This could be explained by PUR shrinkage due to gas exchange.

3.5 Vertical Flexural Test

The vertical flexural test was performed on July 4th, 2017, as a 3-point bending test on
‘standing’ elements, with the load acting horizontally (Fig. 6a). The support span was
224 cm. The load is applied in the middle, and distributed over 60 cm height and the
full width of the element through steel profiles. Due to the sine-shape of the element,
the load is transferred into the concrete over the tops of the sine curves. The forces are
measured at the loading jack and the 2 support points. By deducting one from the other,
the reaction friction on the ground plate can be calculated.

Fig. 6. a, b Vertical flexural test set-up, and resulting load-displacement graph.

 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 137

Specimen 2.5-1 measured 249.5 cm in height, the width of the outer (loaded) face
was 66.5 cm, while the inner (supported) face was 71.0 cm wide (each face included
two full sine waves with a period of 28.5 cm). It was subsequently subjected to the
bending test.
The design moment corresponded with a load of F = 1.13 kN. The load cell dis-
placement was set at 1 mm/min. To assess the development of damage in the element,
the test was stopped when the following load levels were reached: 0.1 kN, 0.5 kN,
1.0 kN, 1.5 kN, 2.0 kN, 2.5 kN, 3.0 kN, 7.5 kN. Due to the nature of the applied load
jacks, the load is released slightly at those points, which can be recognized in the load-
displacement graph of Fig. 6b. Up to a load of F = 3.8 kN, linear load-displacement
behaviour was observed. Slowly, the existing cracks open up when the load increases.
From 2.0 kN load onwards, this becomes observable with the bare eye. The largest
crack is then 0.4 mm wide. At 3.8 kN load, a new crack occurs in the back side, and
the element stiffness is reduced. A new crack occurs at 4.5 kN and from 5.5 kN, the top
part seems to be starting to shear of the bottom parts. The load-displacement behaviour
remains practically linear though. The test was stopped at 7.5 kN (more than 6 times
the design load), and the load was released from the jacks. The element recuperated
practically all of its deformations.
Specimen 2.5-2 was measured and tested in a similar manner, but stopped at a load
of F = 4 kN, to save it for compression testing. The behaviour was comparable to that
of specimen 2.5-1.
Both elements were well capable of carrying the design load without collapse, in
spite of the pre-existing horizontal cracks. The design, therefore, was considered safe
for these load conditions. It was expected that an improvement in the print nozzle could
help to avoid the horizontal cracks in the final manufacturing (see also Sect. 4).

3.6 Compression Test

The compression test was performed once on each face of test element 2.5-2. In the first
test, the outer face was loaded, the inner face was loaded in the second test. This
element had previously been tested in 3-point bending, with the inner face corre-
sponding with the flexural tension side.
Figures 7a and b show the respective test setups. The load is applied by a hydraulic
jack that pushes on a 150 mm steel profile that is placed over the top of the loaded face.
This way, the other face, with wall ties and PUR, acts as stabilizing horizontal
(buckling) support. The design load could not be established since the design was not
yet finalized yet, but was conservatively estimated at 10 kN. The Euler buckling load
of each sheet individually was well above 200 kN, however, excentricities could sig-
nificantly reduce this figure, and buckling could thus not be ruled out completely on
forehand.
The load cell displacement was set at 2 mm/min at the start. The loading speed was
reduced to 1 mm/min after a load of 2 kN was reached. Note that this displacement is
measured over the whole system, including the board interlayer between steel beam
and test element, etc. The total deformation of the concrete itself is much smaller. The
vertical compressive deformation of the loaded sheets was measured over a 1.0 m
length at the center of the element, on each side (2 measurements). Horizontal
138 F. Bos et al.

Fig. 7. a, b Test set-up with load on the outer and inner face of element 2.5-2, respectively.

deformations were measured at the center, +40 and −40 cm at the loaded sheet, and at
the center of the other sheet (4 horizontal measurements). During testing, the vertical
deformation was linear up to a load of 60 kN (or 6 times the design load), at which
point the test was stopped as it was considered to have shown sufficient structural
capacity. No noteworthy observations, visual or audial, were made during the test. No
additional cracks occurred.
Subsequently, the specimen was moved in the set-up and the outer face was loaded,
also to 60 kN. showing comparable behaviour. However, the vertical stiffness was
significantly lower, which can be explained from the smaller concrete section (due to
the sawing process, this face was almost 15% smaller), and the fact that this face had
more cracks as a result of the previous bending test, which allow for some crushing
resulting in the associated deformations.
It was concluded that compression would unlikely be the governing load case,
given that an already damaged element could easily carry 6 times a conservative
estimate of the design load (Fig. 8).

Fig. 8. Element 2.5-2 in compression test on inner face: average load displacement graph for left
and right LVDT.
 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 139

3.7 Impact Test

The 1.33 m element was subjected to pendulum impact loading in a set-up derived
from EN 12600. Since the element shape, size and support do not correspond with the
codified set-up, some adjustments were made (Fig. 9). The element stands on its base
plate. Horizontally, it is supported back to the steel frame by two horizontal timber
beams at the bottom and the top of the element (centre distance 1.24 m). The element
and impact body were positioned so that the impact body would hit the element at mid
height of the top of the middle wave on the front side. It is noted that a pendulum
impact test is not usually required for one-storey building walls. The test was never-
theless performed to acquire an idea of its resistance to dynamic, localized loads and its
sensitivity to damage.

Fig. 9. a and b Pendulum impact test set-up, and element after testing with visible damage
caused by multiple 1200 mm drop height impacts.

The EN 12600 predefines 3 consecutive drop heights: 190 mm, 450 mm and
1200 mm. The element was impacted from each drop height several times. No
noticeable effect was observed after several impacts from 190 mm and 450 mm height.
When the drop height was increased to 1200 mm, increasing damage occurred upon
each impact, beginning with two horizontal cracks upon the first impact, spalling from
the back side on the second impact and spalling at each side on the third impact. A drop
height of 1200 mm corresponds to strict severe impact regulations for balustrades at
height, for which damage is allowed but fall-through is not. The behaviour of the
printed wall element was therefore considered safe for the intended use.

3.8 Overall Conclusion on Structural Performance

The test specimens met the requirements set for each loading condition with consid-
erable margin, and the design was therefore considered safe and sufficiently strong.
Nevertheless, some defects were encountered that called for an improved manufac-
turing design. However, before such studies could be performed, the project was
discontinued due to a shift of priorities with the client.
140 F. Bos et al.

4 Bicycle Bridge Gemert

Around the same period, another project was being developed at the TU/e: a 3D
concrete printed bicycle bridge. This project was initiated by general contractor BAM
as part of a larger infrastructure project. An extensive description of the project and
related testing is given by Salet et al. [15, 16]. This section is largely taken from that
publication.

4.1 Project and Design

To cross a small waterway in Gemert, the Netherlands, a 6.5 m span bicycle bridge
with a width of 3.5 m was designed. It consists of elements printed vertically, that were
subsequently rotated to form a horizontal element. The elements were prestressed
together so that the section always remains in compression (Fig. 10a and b). This
approach is similar to one adopted by Lim et al. [17]. The main advantage is that the
tensile capacity of the printed concrete with hundreds of interfaces, becomes irrelevant.

Fig. 10. a, b 3DCP bicycle bridge, design impression (by BAM). Reproduced from [16].

The zig-zag cross-section of the bridge elements results in a significant material

reduction while maintaining the required flexural and shear resistance. The prestress is
introduced through two solid, cast RC bulkheads. The bulkheads have been placed on
abutments on a pile foundation.
The final design also included zones with automatically embedded reinforcement
cables [12, 13] to provide additional shear capacity and ductility in the lateral direction,
even though the structural concept and dimensions are designed to work without them.
Figure 11a and b show the reinforcement cables and post-tensioned tendons.
Thus, the structural principle of the bridge is quite different from that of the
pavilion: the application of prestress and consequent compression loading in the entire
section avoids a number of uncertainties associated with the tensile strength of the print
mortar and the filament interfaces. It does, however, make the creep and shrinkage
behaviour of the mortar highly relevant as they (partially) determine the prestress loss
that may occur.
 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 141

Fig. 11. a, b Schematic 3D image showing prestress tendons and reinforcement cable zones
(left) and print element section geometry (right). Reproduced from [16].

4.2 Test Program

To validate the structural safety of the design, it was proposed first to test a 50% scaled
mock-up of the bridge destructively in a 4-point bending test, and additionally to
perform an a non-destructive in-situ test of the completed bridged to the design load.
Furthermore, a long term observation would be started to monitor possible prestress
losses. Considering the structural application in an uncontrolled outside environment,
this regime was considered appropriate.

4.3 Printing
Like the pavilion project, a large mouth opening (10  60 mm2) of the nozzle was
required to obtain the appropriate layer width, and the same print speed of 80 mm/s
was used. At 5:20 per layer, it took 48 h to print all bridge elements (Fig. 12). To allow
continuous printing and avoid dry joints, two full wet systems (mixer-pump, hoses and
print nozzle) were used.

Fig. 12. a, b Printing of a bicycle bridge element.

142 F. Bos et al.

To avoid deterioration of the interlayer adhesion, while maintaining proper intro-

duction of the reinforcement cable, a hybrid down/back flow nozzle was developed,
shown in Fig. 13. The development of the nozzle design is discussed more extensively
in [16].

Fig. 13. Down/back-flow nozzle. Reproduced from [16].

4.4 4-Point Bending Test

A mock-up was printed at 50% of the bridge final size, resulting in global dimensions
of l  b  h = 1720  500  460 mm. The mock-up elements were joined by three
post-tensioning tendons at the top and six at the bottom (Dywidag, Ø15.7 mm, Y1860
steel grade). The prestress forces were measured with load cells on one top and one
bottom tendon. The immediate prestress loss was around 20%, while the additional
prestress loss after two weeks of prestressing was an additional 4%. After assembly, the
mock-up was subjected to a load-controlled 4-point bending test until failure (Fig. 14).
The load rate was 30 kN/min. After having reached a load of 120 kN, an alternating
load-unload sequence was applied, at each increase of 30 kN load, until 300 kN.
Finally, the bridge was loaded to 350 kN and the test was stopped. Figure 15 shows the
load-displacement curve.
The load-unload sequence was applied to identify the point at which structural
damage occurs, which is shown by a loss of stiffness at reloading. The test showed damage
first occurred between 240 and 270 kN load. The first visible crack was noted at 300 kN
load, in the mid-span, indicating a flexural failure. Thus, the mock-up easily passed the
pass/fail criterium, which was set at a load carrying capacity of 176 kN (Fig. 16).

4.5 In-Situ Test

An on-site test was performed on the final bridge to ascertain the scaled-up production
process had not introduced unexpected errors. Figure 17 shows the bridge loaded to the
bending moment of the serviceability limit state. The deformations were minimal and
 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 143

Fig. 14. Scale model in 4-point bending test set-up. Reproduced from [16].

Fig. 15. Load-displacement curve of the 4-point bending test on 1:2 scale model. Reproduced
from [16].
144 F. Bos et al.

could not be measured – actually, no response of the bridge to the loading was
observed at all. Consequently, the bridge was approved under the ‘design by testing’
option in the European and Dutch national regulations. The bridge was taken into use
in October 2017 (Fig. 18).

Fig. 16. a, b Bending crack, becoming visible at 300 kN total load. Reproduced from [16].

Fig. 17. In-situ testing of the bridge with water-filled containers.

Fig. 18. Completed bridge at the opening. (photo: Kuppens fotografie)

 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 145

5 Discussion and Conclusion

The large scale experiments on both the pavilion and the bridge showed they were
designed according to adequately robust structural principles as well as with sufficient
margin. Furthermore, no unexpectedly premature failures occurred. Nonetheless, the
two discussed projects yielded a variety lessons related to object scale and structural
response.
Due to the non-static nature of the concrete, prolonged printing may cause failures
due to hardening particles, as was experienced in one of the printed pavilion elements.
Air entrapments or water flow inconsistencies may cause less catastrophic, but
nonetheless undesirable, areas of reduced quality. Large scale printing should include
quality control based on acceptance criteria to ensure no inferior quality printed ele-
ments are approved.
The projects also showed that even the effects of seemingly small alterations to the
manufacturing process on the structural quality should be thoroughly considered. This
was underlined by the experiences with the nozzle designs.
Even though the bridge section elements took over 5 min per layer to print, all
elements stayed well within the limits with regard to deteriorating interface strength
between layers. This is due to the long dormant state of the applied print mortar.
Another geometrical effect did occur, however. The plastic deformation behaviour
of the material in the dormant stage resulted in dimensional deviations from the design
that are not necessarily negligible. Due to the self-weight of the added layers in the
pavilion elements, the face thickness increased to over 70 mm, while 60 mm was
designed. This is an effect that is really particular to DFC technologies, as the final
geometry is not well defined by the mold as with conventional cast concrete. In this
case, the effect was beneficial in structural terms, but the correspondence of the actual
geometry and the designed geometry properties should be considered as they in cases
may also turn out to be detrimental or lead to inefficient material use.
A thoroughly robust design using prestress introduces relatively small risks to the
structural performance. The structural response is relatively easy to predict as it is
hardly dependent on more sensitive parameters like flexural tensile strength. However,
when a composite element is designed, the compatibility with other materials is
extremely important, as was experienced in the pavilion elements that cracked due to
unexpected stresses and loads caused by the insulation foam and the wall ties (in
combination with also unexpected reduced interface strength).
For DFC processes in general, some of the findings may be relevant and others may
not. But the observations do indicate that large scale mock-ups and testing are highly
beneficial to quality control and should form an integral part of approval processes
when new DFC technologies are used for structural applications, as long as the rela-
tions between design, material, process, and product are not fully understood. These
may serve not only to validate the feasibility of designs but also to avoid overly
conservative approaches.
146 F. Bos et al.

Acknowledgements. The following partners were involved in the Nyborg Pavilion project:
- Client: Nyborg municipality, Denmark,
- Architect: I. Moltke, Denmark,
- Structural engineering: Witteveen+Bos, the Netherlands,
- Research, print design and manufacturing of printed elements: Eindhoven University of
Technology, the Netherlands,
- Material supplier print mortar: Saint-Gobain Weber Beamix, the Netherlands.
- In the 3D concrete printed bicycle bridge project, the partners were:
- Client: Province Noord-Brabant, the Netherlands,
- General contractor, initiator: BAM Infra, the Netherlands,
- Structural design and engineering: Witteveen+Bos consulting engineers, the Netherlands,
- Research, print design and manufacturing of printed elements: Eindhoven University of
Technology, the Netherlands,
- Material supplier print mortar: Saint-Gobain Weber Beamix, the Netherlands,
- Prestress system and application: Dywidag, the Netherlands,
- Reinforcement cable supplier: Bekaert N, Belgium.

References
1. Le, T., et al.: Hardened properties of high performance printing concrete. Cem. Concr. Res.
42, 558–566 (2011)
2. Panda, B., Paul, S.C., Tan, M.J.: Anisotropic mechanical performance of 3D printed fiber
reinforced sustainable construction material. Mater. Lett. 209, 146–149 (2017)
3. Sanjayan, J.G., Nematollahi, B., Xia, M., Marchment, T.: Effect of print-time interval on
inter-layer strength of 3D printed concrete. Constr. Build. Mater. (accepted for publication)
4. Ahmed, Z.Y., Bos, F.P., Wolfs, R.J.M., Salet, T.A.M.: Design considerations due to scale
effects in 3D concrete printing. In: Proceedings of 8th International Conference of the Arab
Society for Computer Aided Architectural Design (ASCAAD 2016), London, UK
5. Nederlands Normalisatie Instituut: NEN-EN 1990:2002 Eurocode - Basis of structural
design. Delft, The Netherlands (2002)
6. Bos, F., Wolfs, R., Ahmed, Z., Salet, T.: Additive manufacturing of concrete in construction:
potentials and challenges of 3D concrete printing. Virtual Phys. Prototyp. 11, 209–225
(2016). https://doi.org/10.1080/17452759.2016.1209867
7. Doomen, C.C.M.: The effect of layered manufacturing on the strength properties of printable
concrete. MSc graduation thesis, TU/e, Eindhoven (2016)
8. Slager, G.J.: Influence of the interface between layers on the tensile properties of 3D-printed
concrete. M.Sc. graduation thesis (A-2017.194). TU/e, Eindhoven (2017)
9. 3ders: http://www.3ders.org/articles/20170602-cybe-construction-completes-3d-printing-of-
168-sq-m-rdrone-laboratory-in-dubai.html. Accessed 25 Mar 2018
10. 3ders: https://www.3ders.org/articles/20180312-a-3d-printed-house-for-4000-silicon-valle
ys-new-story-can-build-it.html. Accessed 25 Mar 2018
11. 3ders: http://www.3ders.org/articles/20150118-winsun-builds-world-first-3d-printed-villa-
and-tallest-3d-printed-building-in-china.html. Accessed 25 Mar 2018
12. Bos, F.P., Ahmed, Z.Y., Wolfs, R.J.M., Salet, T.A.M.: 3D printing concrete with
reinforcement. In: Luković, M., Hordijk, D.A. (eds.) High Tech Concrete: Where Technology
and Engineering Meet, pp. 2484–2493. Springer, Cham (2017). https://doi.org/10.1007/978-
3-319-59471-2_283
 Large Scale Testing of Digitally Fabricated Concrete (DFC) Elements 147

13. Bos, F.P., Ahmed, Z.Y., Jutinov, E.R., Salet, T.A.M.: Experimental exploration of metal
cable as reinforcement in 3D printed concrete. Materials 10, 1314 (2017). https://doi.org/10.
3390/ma10111314
14. Vermue, F.: Numerical and experimental assesment of 3D Concrete Printed Sandwich
Panels with different insulating core materials. MSc graduation thesis, TU/e, Eindhoven
(2017)
15. Salet, T.A.M., Ahmed, Z.Y., Bos, F.P., Laagland, H.L.M.: 3D printed concrete bridge. In:
Chua, C.K., Yeong, W.Y., Tan, M.J., Liu, E.J., Tor, S.B. (eds.) 3rd International Conference
on Progress in Additive Manufacturing (Pro-AM 2018), Singapore, 14–17 May 2018
16. Salet, T.A.M., Ahmed, Z.Y., Bos, F.P., Laagland, H.L.M.: Design of a 3D printed concrete
bridge by testing. Virtual Phys. Prototyp. 13(3), 222–236 (2018)
17. Lim, S., Buswell, R., Le, T., Austin, S., Gibb, A., Thorpe, T.: Developments in construction-
scale additive manufacturing processes. Autom. Constr. 21, 262–268 (2012)