Additive Manufacturing 37 (2021) 101656

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Experimental prediction of material deformation in large-scale additive

manufacturing of concrete
Negar Ashrafi a, Shadi Nazarian a, Nicholas A. Meisel b, José Pinto Duarte a, *
a
Stuckeman School of Architecture and Landscape Architecture, The Pennsylvania State University, University Park, PA 16802, USA
b
School of Engineering Design, Technology and Professional Programs, The Pennsylvania State University, University Park, PA 16802, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Additive manufacturing (AM) of cementitious material has become a popular subject over the last decade. The
Cementitious material multidisciplinary nature of this topic has led researchers from multiple areas of expertise such as architecture,
Digital fabrication engineering, and materials science to collaborate to improve the technology, which does not permit yet to print
Additive manufacturing of concrete
mixtures with coarse aggregates, but is often referred to as AM of “concrete” or “concrete” printing. An important
3d printing
aspect of research in the area is finding a Portland cement-based mortar with adequate rheological, hardening
and strength properties for printing architectural structures. In addition, the properties of fresh and hardened
mortar and its deformation behavior affect the shape accuracy of the printed geometries and require designers to
adjust the toolpaths and technology to account for issues in the printing. This paper is aimed at studying the
deformation of a printed concrete mix, which previous studies have shown to be printable. It is focused on the
effect of the number of layers, the number of beads and time on layer height and width. It proceeds through a
series of experimental tests and it uses regression analysis to model material behavior. The resulting equations
can be used in toolpath design to compensate for such deformation and have more accurate printed geometries
subsequently. Future studies will be concerned with linking material properties with material deformation and
use results to develop a more generic toolpath generator.

1. Introduction and motivation built, as well as the printing time, it is one of the crucial tasks in process
planning. Designing a toolpath is comprised of two main steps, including
Recently, using AM techniques in the building industry became more “interior filling” and “linking sequence” [4]. Interior filling concerns the
popular, as architects and engineers attempt to use their potential to method according to which layers are filled up with filaments, while
build free-form, unsupported structures automatically. Even though AM linking sequence refers to the order in which each layers are printed
technology in the building industry is in its early stage of adoption, it can without interrupting the deposition process [5]. The main objectives of
be cost effective and time efficient while improving accuracy in con­ toolpath design for material extrusion-based AM include: (1) obtain
struction [1]. These potential advantages have motivated many recent satisfactory extrusion quality; (2) increase printing efficiency; and (3)
studies aimed at improving the technology, particularly, applied to decrease the number of sub-paths to reduce travel time used in linking
cementitious materials. Although, it does not permit to print ready- the paths [6].
mixed concrete with coarse aggregates, the printed material is often Unlike plastic 3D-printing, one of the main challenges in AM of
referred to as “concrete” and the technology as “concrete” AM or concrete is the deformable characteristic of extruded fresh concrete.
printing in the literature on the topic, and so, this terminology is also Fresh concrete used in AM should be “flowable” and “buildable” at the
used in this study, where the term concrete refers to Portland cement- same time, which means it should be deformable enough to be extruded
based mortar. Most of concrete AM processes rely on material extru­ through the nozzle but retain its shape during the building process,
sion techniques [2] which deposit linear filaments or beads of concrete resisting deformation as much as possible [7]. Generally, there are three
through an extrusion print head along a programmed path. main causes of deformation that affect extruded concrete layers
The AM toolpath determines the trajectory for the extruder when including, self-weight, the weight of subsequent layer(s) printed on the
depositing the material layer by layer [3]. Since the design of toolpaths top of it, and extrusion pressure (Fig. 1). No major self-weight defor­
affects the geometric quality, strength and stiffness of the model to be mation can be detected in a layer when the quality of the printing

* Corresponding author.

https://doi.org/10.1016/j.addma.2020.101656
Received 10 June 2020; Received in revised form 14 September 2020; Accepted 6 October 2020
Available online 13 October 2020
2214-8604/© 2020 Elsevier B.V. All rights reserved.
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Fig. 1. Diagram illustrating material deformation of layers in concrete printing under the load of subsequent layer(s).

Fig. 2. Effect of the number of beads and layers on shape deformation: a) designed shape, and b) deformed shape.

Fig. 3. Compensating for shape deformation through toolpath design: a) deformed shape, and b) compensated shape.

material is adequate. However, the weight of subsequent layers and deformation in toolpath design can lead to better shape quality and
extrusion pressure may cause unwanted deformations once the next increased strength.
layer(s) are deposited [7]. Without considering the effect of time on material deformation, the
The printing time gap between consecutive layers is an important number of beads and the number of layers will have a cumulative effect
factor in predicting deformation. The longer the time gap between two on shape deformation due to the increasing loads and extrusion pres­
layers the harder the bottom layer can become thus maintaining its sure, as illustrated on Fig. 2. In addition, studying the effect of the
shape with less deformation [7]. On the other hand, some studies [8,9] number of beads and the number of layers on shape deformation and
have shown that longer interlayer time gaps lead to lower bond strength, compensating for such an adverse effect in the design of the toolpath can
which negatively impacts the structural properties of printed elements. result in more accurate printed geometries, as shown in Fig. 3. The work
In this regard, finding the interlayer time gap that can results in mini­ presented in this paper is aimed at studying the combined effect of time,
mum deformation without compromising structural properties is the number of beads and the number of layers on material deformation
essential in toolpath design. As such, considering the amount of material and use this information to improve shape quality through toolpath

2
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Table 1
List of case studies using cement-based AM, adapted from Kontovourkis et al., [15.].
Project Location Architects On Site / Prefab Material Printer Method

Una Casa Tutta Di Un Pezzo Milan Marco Ferreri On Site Inorganic binder + sand Crane 3D Printer D-shape
Radiolaria Pavilion Pontedera Andrea Morgante On Site Inorganic binder + sand Crane 3D Printer D-shape
3DP Office Dubai Killa Design Prefab Cementitious N/A CC
10×3DP House Shanghai Yingchuang Prefab Cementitious Crane 3D Printer CC
Stupino House Moscow Apis Cor & PIK On Site Cementitious Crane 3D Printer CC
3DP House Milan CLS Architect & Arup On Site Cementitious Robotic 3D Printer CC
Vulcan House Austin ICON On Site Cementitious Crane 3D Printer CC
3DP House Valencia Be-More-3D On Site Cementitious Crane 3D Printer CC
Penn state Den at Mars[16] Peoria PSU Team On Site Cementitious Robotic 3D Printer RAM
3D Printed Tiny Houses[17] Mexico ICON On Site Cementitious Crane 3D Printer CC

design. The term “beads” refers to the linear filaments extruded by the • open time, which is the time span after mixing during which the
nozzle, labelled B1, B2, and so on, in Fig. 2a, and “layers” refers to material can be pumped and extruded, and it depends on the main­
collection of adjacent filaments that are deposited on same level, tenance of the viscosity and yields stress of the mixture; [17]
labelled as L1, L2, and so on, in the same figure. • setting time, which is the time required for the material to keep its
Section 2 summarizes the state of art in large scale additive shape before adding successive layers;
manufacturing of concrete. Section 3 briefly describes the printing sys­ • layer cycle-time, which is the time needed to build one layer and it
tem and the properties of the concrete mix used in the study. Section 4 will affect the building rate; [16]
describes three tests designed to determine: (1) the effect of the number • rheological properties and how they affect print qualities; [18–20]
of beads and layers on layer width deformation, (2) the time interval • structural build-up, which refers to the property of fresh concrete
until the printed concrete mix no longer deforms, and (3) the effect of getting stronger with time because of chemical and/or physical re­
different interlayer time intervals on layer height deformation. Section 5 actions; [21,22]
summarizes the current study, presents the conclusions and identifies • interlayer adhesion, which refers to the bonding between extruded
future work. layers; [18,20,23,24] and,
• layer deformation, as consecutive layers are being added on top
2. State of the art in large-scale additive manufacturing [16,25].

In the architectural field of interest, AM has been used for concept In 2012, Le et al. [8] studied the mix design and fresh properties of
modeling [10]. However, scaling up AM techniques for full-scale auto­ high-performance, fiber-reinforced, fine-grained aggregate concrete for
mated building construction can have a strong impact on the construc­ 3D printing applications. The cementitious mixture was comprised of
tion industry with increasing customization and design flexibility, sand, cement, fly ash, silica fume, and polypropylene fiber. The most
reducing construction time, and reducing manpower and construction critical fresh properties of printing concrete were defined to be extrud­
cost [11]. Although the use of AM for this purpose include experimen­ ability and buildability, which are linked to workability and time. In
tation with various materials such as plastic, metal, and clay [12], the their study, the final mixture was tested by constructing a 2 m long
focus of this paper is on the use of AM with concrete mixtures as the most concrete bench.
common building material, particularly Portland cement-based ones. Extrudability refers to the ability of the concrete mixture to be
Lim et al. [13] claimed that construction-scale 3D-printing is extruded through a printing head containing a nozzle to form a
currently performed by the use of only three main processes in the public continuous and intact concrete bead [8,26]. Although the essential
domain: D-shape, which has been invented by Enrico Dini in Italy, is a flowability is a requirement for proper extrudability, it cannot guarantee
large binder-jet 3D-printer that utilizes materials such as sand, inorganic the desired extrusion through nozzles. If the geometry size of the
seawater, and magnesium-based binder to form materials that are printing hose is not consistent with the nozzle head or if they have
similar to stone in terms of properties and appearance; Contour Crafting different cross-sections, such as the cases of conical or rectangular
(CC), which was initiated by Behrokh Khoshnevis in the US, is a layered nozzles, it may lead to the blocking of the nozzle head.
fabrication technology uses automation to mimic the conventional The material also needs to have sufficient buildability features to be
construction process to build both whole structures and subcomponents able to be deposited correctly, remain in position, and be stiff enough to
for assembly [14]; and Concrete Printing, developed at Loughborough support upper layers without collapsing [8,27,28]. In addition, the
University in the UK is a technique that uses a 3D-printer comprised of a material should bond layers together well enough [18,20,23,24].
frame and a nozzle mounted on a moving horizontal beam for Buildability also relies on the workability of the material mixture, which
manufacturing full-scale construction and architectural components refers to the capacity of concrete to be transported, placed, compacted,
[13]. Recently, another printing technique called Robotic Additive and finished without any segregation. Insufficient buildability may
Manufacturing (RAM) has been introduced to the field, gaining result in deforming the extruded layer in horizontal and vertical di­
considerable attention. The term (RAM) refers to the process of layer by rections and, as a result, changing the geometry of the printed specimen
layer material extrusion, leading to the creation of a solid product using or collapsing the structure [28]. Buildability failure can be caused by
industrial robots [15]. Table 1 presents a list of current case studies plastic collapse, when during the printing process the stress that the
using Portland cement-based AM. bottom layer bears reaches the yield stress of the material. It can also
Since the properties of the material significantly affect printability cause by elastic buckling which may occur due to progressive lateral
and print quality, research on the fresh and hardened properties of deformations or unusual layer placement. To guarantee a stable struc­
Portland cement-based mortar for additive manufacturing is increas­ ture, the modulus of elasticity of the printed material needs to exceed the
ingly gaining more attention. critical value. Buildability failures were studied and modelled by several
According to Buswell et al. [16], the most important rheological and researchers [18,29–32].
physical parameters concerning the fresh material) can be categorized In 2017, Kazemian et al. [7] presented a framework to test the per­
into the following groups: formance of cement-based mixtures in large-scale AM. Their research
evaluated the workability of a fresh ‘‘printing mixture” regarding quality

3
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

of print, stability of shape, and printability. Print quality was explained Table 2
considering both surface and layer quality, and methods were provided GCT material composition [40.].
to determine them. Also, detailed information about the two testing Components Chemical Name Content
procedures developed to assess stability, i.e., “layer settlement” and
Pulverized Limestone <2%–6%
“cylinder stability”, were presented. The proposed framework was used Lime 30%
for characterizing material deformation in their study. Crystalline Silica <50%–70%
Roussel [18] discussed the rheological requirements for printable Portland Cement <50%
concrete such as yield stress, viscosity, elastic modulus, critical strain Calcium Sulfoaluminate Cement 5%–12%
Cellulose 0.2%–2%
and building rate to prevent the critical deformation of layers and Starch 0.2%–2%
guarantee it could carry its own weight. He also studied the elastic
behavior of the extruding material to control the final geometrical di­
mensions of one layer. Some other studies attempted to manipulate the material mixture
In 2015, Malaeb et al. [33] studied the effect of superplasticizer on design to improve bonding. Panda et al. [37] developed a 3D printable
both flowability and buildability of a cementitious mortar. It was geopolymer mortar as a reference material with high yield stress and low
claimed that in a mixture with a relatively lower water to cement (W/C) viscosity properties by adding nano-clay. They then investigated the
ratio, the compressive strength and flowability of the mixture increased effect of adding slag to the reference material to improve interlayer
with the amount of superplasticizer. This would, however, negatively bonding. The results of this study indicated that, using slag reduced the
affect the buildability of the material. The best W/C ratio was reported workable time period and also adversely affected the bonding strength
to be about 0.39 and the best amount of superplasticizer was reported to as it caused the printed layer to lose moisture from its surface before
be about 1.9%, respectively. A 10-cm height wall was constructed using deposition of the next layer up.
with a 2-cm diameter nozzle. During the deposition of each layer, a slight deformation of each
Throughout the layer-by-layer construction procedure, it is very layer helps the bonding to the previous layer [2]. On the other hand, the
important that previously- deposited layers can withstand the load from overall deformation increases as the height of the structure increases.
the subsequent layers. In this regard, Perrot et al. [34] attempted to Since it is common to have equal layer height during the printing pro­
optimize the building rate by proposing a theoretical framework based cess, the designed distance between the nozzle and the previous layer
on the rheological properties of Portland cement-based mixtures such increases and it causes deformation of the printed filament, which can
that each layer could sustain the weight of subsequent layers during the result in weakening the bonding between layers and, in some cases, the
deposition processes. Reiter et al. [35] studied the role of chemically collapse of the structure [38]. To address this issue two approaches have
controlling cement hydration by adding different type of admixtures to been proposed: Wolfs et al. [39] attempted to adjust the nozzle height
control the structural build-up. dynamically during the printing process. While Wangler et al. [17] and
Interlayer bonding is also another critical factor in AM of concrete. Perrot et al. [19] tried to control the buildup rate by injecting additives
To have a homogenous structure, all the successive layers must be to the material right before extrusion, to accelerate the hardening pro­
bonded together perfectly and, since there is no vibration or external cess so that each layer could maintain its shape under the load of upper
force during the AM process, bonding must occur during fabrication. In layers.
this regard, Zareiyan and Khoshnevis [23] studied the effect of extrusion Most studies on material aspects of Portland cement-based mixtures
rate, layer thickness, and layer width on the bonding between layers. for AM are focused on eliminating the deformability of the fresh con­
They concluded that decreasing the maximum aggregate size and crete as much as possible, while keeping flowability and printability.
increasing the cement to aggregate ratio lead to a better interlayer However, despite this body of work, current mixes still deform during
adhesion and as a result higher structural strength. They also realized printing. As such, instead of manipulating material properties to avoid
that although the increasing layer thickness resulted in better interlayer deformation, this study is aimed at studying concrete deformation
bonding, it reduced the overall compressive strength of the printed during and after the printing process and using it as a decisive factor in
specimen. Likewise, the shorter setting time increases the buildability of toolpath design. The deformation of each printed layer results from layer
the material but it may affect interlayer bonding negatively, as it may height deformation and layer width deformation, which is also depen­
produce cold joints at layer interface. Roussel [18] also discussed the dent on layer height deformation. Ashrafi et al. [25] studied layer height
necessity of the interlayer bonding and concluded that a long time- deformation under the load of the subsequent layers. By measuring layer
interval between layers can cause drying of the upper surface of the height deformation of one printed bead in each layer, they designed a
layer and, as the result, weaken the interlayer bonding. Keita et al. [24] new toolpath that compensated for the deformation to print layers of
showed that the tensile splitting strength in uncovered specimens equal height. The current study aims at modeling layer width defor­
reduced up to 50% in comparison to samples that were protected from mation, not only for a different number of layers but also for a different
drying. number of beads, under the load of upper layers. Since concrete defor­
Tay et al. [20] also investigated the effect of time interval on the mation depends on time, our study also attempted to determine the time
bond strength at microscopic scale. They concluded that the rheological interval after which deposited concrete no longer deformed, model
properties of the initial layer increased the time interval between layers deformation during such interval, and then design toolpaths to
and weakened interlayer bonding. Consequently, it negatively affected compensate for such deformation and print specimens with the desired
the tensile strength of the printed specimen. Also, as the time interval shape.
increased, the voids between filaments increased, which weakened the
bond strength exponentially. Wolfs et al. [36] studied the impacts of 3. Printing material and printing system
different process parameters on interlayer bonding strength and how it
affected the mechanical behavior of 3D printed concrete. It was reported The material used in all the experiments described in this study was
that layer orientation did not affect bonding strength except in cases in developed by Gulf Concrete Technologies in cooperation with our
which the time-interval between layers was very short. They also research team. The mixture was a blend of Portland cement, lime, pul­
confirmed the fact that, an increase in time interval affected layer verized limestone, specially graded masonry sand, fibers and admixtures
bonding negatively. Also, unprotected (uncovered) surfaces during the (Table 2). The maximum particle size in GCT concrete was 1 mm.
time interval had weaker bonding strength. They also did not find any Mastersizer 3000 system, which applies laser diffraction technique
significant relation between the height of the nozzle and bonding was used to measure the particle size and particle size distribution of the
strength. GCT material. The particle size distribution of GCT material is illustrated

4
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Fig. 4. Particle Size Distribution of GCT Material.

26
24 24.55

22
Strength (MPa)

20
18 17.95
16
14 15.12

12
10
3 7 28
Age (days)
Fig. 5. Compressive strength test of GCT material.

Table 3 Table 4
GCT Material setting time. GCT Material Flowability.
Setting Time Initial Set (min) Final Set (min) Flowability Flow (cm)
Values Ave Values Ave Ave values for each test/ (SD) Average of two test
75.3 80.7 143.0 143.0 22.97/ (0.39) 23.25
86.0 143.0 23.5/( 0.24)

in Fig. 4. The material shows a range of particle size from 0.460 µm to A flow table test (ASTM C-1437) was conducted to evaluate the
27.4 µm. D10: 0.930 µm (i.e., 10% of particles are smaller than flowability of the mixture. The value provided in Table 4 corresponds to
0.930 µm); D50: 3.07 µm (50% of particles are smaller than 3.07 µm); the average obtained from four readings for each test and their standard
D90: 8.35 µm (90% of particles are smaller than 8.35 µm). The specific deviation. Two tests were conducted to guarantee the repeatability of
surface area of GCT was determined to be 2805 m2/kg. the test results, since according to the standard, the results of two
The compressive strength of the material was tested in accordance properly conducted tests by the same operator on the same batch of
with ASTM C-109. Fig. 5 presents the average compressive strengths material, should not vary by more than 11%.
measured at 3, 7, and 28 days with standard deviation reported. The AM system used in this study, diagrammed in Fig. 6, consists of a
The Vicat Needle test (ASTM C-191) was performed to measure the mixer-pump (m-tec Duomix 2000) for mixing and extruding the mate­
initial and final setting times presented in Table 3. Two tests were rial, a silo that contains the dry mix and feeds the pump, and an in­
conducted to evaluate the repeatability of the test results. According to dustrial 6-axis robotic arm (ABB IRB 6640). With this system, it was
the ASTM C-191 test, the results of two properly conducted tests on a possible to control variables related to the pump and robotic arm, which
similar batch of material by the same operator for initial setting time depend on the properties of the concrete mixture and can directly affect
should not differ by more than 34 min and for the final setting time print quality. These variables included the dry mix feed and water flow
should not vary by more than 56 min, and so the results reported in rates, which determine the proportion of water to dry mix, and together
Table are compliant. with the pump paddle rotation speed and the nozzle section determine

5
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Fig. 6. 1. Computer for generating the toolpath and the RAPID code; 2. Robot Controller for reading the RAPID code and controlling the robot; 3. Robot for moving
the nozzle based along the generated toolpath; 4. Silo for storing the dry mix, 5. Duomix 2000 pump for mixing the dry mix with water and pumping the fresh
material; 6. Nozzle for depositing the material; and 7. Printed specimen.

Table 5
results of compression Test for 3D Printed Cylinder with GCT material.
Compression Test

Compressive Strength f c [psi (MPa)]

′
Maximum Load [lb (kN)]
2150 (14.82) 60,830 (270.58)

accordance with ASTM C39 (Fig. 7). The specimen was printed using the
GCT material and a nozzle to printed layer distance of 15 mm, as in all
the tests included in this study. The results of the test (Table 5)
demonstrated that the printing material and the printing system could
be used to print structures with adequate strength.

4. Experimental setup and methods

In this study three different set of experiment were conducted. Test 1

targeted the effect of the number of adjacent beads and layers on layer
width deformation, Test 2 was aimed at finding the time interval after
Fig. 7. Compression test for 3D printed cylinder. which the printed material stops deforming, and Test 3 studied the effect
of different interlayer time intervals on layer height deformation.
the robotic arm speed. Gulf Concrete Technology (GCT), the company
that supplied the Duo-mix 2000, manufactured in Germany by m-tec, 4.1. Test 1- Effect of the number of beads and layers on layer width
disclosed basic information regarding the relationships between these deformation
variables, which served to determine adequate printing settings,
including the pump flow rate and the robot speed for a given nozzle size. 4.1.1. Experimental setup
The toolpaths were computed and generated using Rhino and Grass­ The main goal of the undertaken series of tests was to study layer
hopper. The Grasshopper plugin HAL was used to convert the toolpaths width deformation under the load of upper layers for different numbers
to RAPID, which is the high-level programming language for ABB in­ of layers and beads. In a previous study, Ashrafi et al. [20] determined
dustrial robots. ABB robots have two operating modes: manual mode, in that the printing orientation along X and Y axes affected material
which mode the manipulator movement is under manual control and the deformation. In the current study, it was realized that in manual mode,
speed is reduced to a maximum 250 mm/s; and automatic mode, in built-in limitations in the robot’s movement along each axis caused it to
which the safety function of the three-position enabling switch (one of move at slightly different speeds, depending on the printing direction
the two safety functions of ABB robots) is bypassed so that the manip­ and orientation, whereas in automatic mode it always moved at the
ulator is allowed to move without human intervention and the robot same speed. Considering that the robotic arm speed affected material
moves at full speed completely “autonomous”. In all the experiments extrusion and, therefore, its shape, it was decided to run the current tests
described in this study, the robot was used in automatic mode. in automatic mode. In addition, Test 1 was designed in a way to verify
A 2-day compressive strength test was conducted on a 3D printed whether robot speed remained the same in automatic mode and so
cylinder with dimensions of 150 mm diameter × 300 mm height in printing orientation did not affect material deposition and shape.
For this purpose, one to five beads of material were printed in five

6
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Fig. 8. Diagram of the toolpath and printing setup in Test 1.

Fig. 9. Preparation of specimen to determine layer width deformation: a) the printing toolpath and cutting points, b) the printed specimen, c) the cured specimen
after cutting.

layers along X and Y orientations to determine the effect of orientation

on layer width (Fig. 8). The length of the printed beads was 100 cm and
they were connected to enable continuous printing (Fig. 9a and b). A
15 mm layer height was used in the toolpath design in all the experi­
ments. Each specimen was printed thrice to determine the mean (μ) and
standard deviation (σ ). The total layer width of each layer for a different
number of beads was measured after curing. For measuring the total
layer width, the hardened 100 cm beads were cut in ten equal 10 cm

Table 6
Values of printing variables in Test 1.
Pump Flow Nozzle Robot Nozzle Lab Relative
Speed Rate size Speed distance Temperature Humidity
Fig. 10. Measuring the layer width of each segment. (kg/ (mm) (mm/ to the (C◦ ) %
h) sec) bed
(mm)

7 750 24.5 169 15 18.3 56

7
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Table 7 represents the statistical significance level (i.e., p-value) of the inde­
ANOVA analysis results, Tests of Between-Subjects Effects. pendent variables and their three-way interaction. As can be seen, the p-
Source Type lll Sum df Mean F Sig. value for the number of beads and the number of layers is less than 0.05,
of Squares Square which means that statistically they can significantly affect the layer
Orientation 2.884 1 2.884 0.319 0.574 width. The p-value for printing orientation is larger than 0.05, which
Number of Beads 324,662.564 4 81,165.641 8964.749 0.000 indicates that it does not significantly affect layer width. Therefore, for
Number of Layers 3010.788 4 752.697 83.135 0.000 further analysis, all values from x and y orientations were combined to
Orientation*Number 117.527 16 7.34 0.811 0.670 provide more statistically reliable results. The statistical significance
of Beads*Number of
the Layers
level of the three-way interaction term is 0.670, meaning that there is
not a statistically significant three-way interaction effect of triad ori­
R square: 0.997, Dependent variable: Layer width entation*number of beads*number of the layers.
Table 8 presents the average percentage error in each test. Results
show that layer 5 with five beads has the least percent error in all the
Table 8
variations. This can be explained by the fact that layer 5 has the least
Average percentage error in Test 1 - Effect of the number of beads and layers on
layer width deformation. amount of load on top, and the five number of beads will provide a
longer interlayer time interval for printing the following layer, meaning
Number of layers
that the printed concrete has more time to harden, and therefore de­
1 2 3 4 5 forms less.
Number of beads 1 35.78 29.78 25.00 21.00 17.89 Fig. 11 shows the percentage error in width for each set of beads. The
2 30.83 26.00 23.11 19.94 16.56 trend lines show a linear relationship between the data points. As it can
3 27.85 22.78 21.56 17.70 13.48 be observed, trend lines have a negative slope, meaning that as the layer
4 21.17 19.08 13.64 11.81 8.36
5 18.33 16.98 12.89 9.98 6.16
number increases, the percentage error in width decreases because the
layer bears a smaller load on top compared to the previous layer. Also,
the slope of trend lines decreases with the number of beads meaning that
long segments (Fig. 9a and c) and the total layer width was measured for as the number of beads increase, the deformation of the layers decreases.
each segment (Fig. 10). The average layer width for each specimen was A linear regression analysis was conducted on the results to develop
considered as the total layer width for that specimen. Four variables an equation that could predict layer width based on number of layers
(dependent and independent) were under consideration in this experi­ and number of beads. A regression analysis is a statistical model for
ment: (1) layer width (dependent), in which the deformation was estimating the relationships between a dependent variable (layer width
measured based on the percent error of the printed and the designed in this case) and one or more independent variables (number of layers
layer; (2) orientation (independent), which had two values, "X" and "Y"; and number of beads). Based on the result of the regression analysis (R2
(3) number of beads (independent), which had five values, from 1 to 5; = 0.99), Eq. 1 was developed:
and (4) , number of layers (independent), which had five values, from 1
zLW = axNL + byNB + c (1)
to 5.
Table 6 presents the values of the relevant variables of the printing
WL = ( − 3.0213 × NL ) + (32.7086 × NB ) + 16.46
system. Results show that the average percentage error of layer width
decreases with increasing the number of layers and beads in both X and where, WL= Layer width, NL = number of layers; and NB = Number of
Y orientations. beads.
The experimental results and the values estimated by Eq. 1 are
compared in Table 9. This shows that the obtained equation from
4.2. Data analysis and discussion of Test 1 - Effect of the number of beads
regression analysis can predict the layer width for different number of
and layers on layer width deformation
beads and number of layers with the average percentage error of 2.93.
Since the deformation of concrete is time-dependent so that after a
To verify whether the three independent variables affected layer
certain time concrete gets hard enough that no deformation occurs, the
height, a three-way ANOVA analysis was used on the collected data.
next test was conducted to determine the time interval within which Eq.
Results of the analysis are shown on Table 7. The "Sig." column

Fig. 11. The % error in width for each set of beads.

8
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Table 9 2 is valid.
Comparison of the average layer width estimated by Eq. 2 and in Test 1 - Effect of
the number of beads and layers on layer width deformation.
Number of Number of Observed LW in Estimated LW by % 4.3. Test 2 - Time interval for zero deformation
layers beads Test 1 Eq. 2 error
4.3.1. Experimental setup
1 1 40.73 46.15 13.31
2 1 38.93 43.13 10.78 The objective of Test 2 was to establish the time interval needed to
3 1 37.50 40.11 6.96 print the next layer with no deformation of the layer underneath. For
4 1 36.30 37.09 2.17 that, a 100 cm by 100 cm square was printed with 50 layers. This square
5 1 35.37 34.07 3.67
geometry enabled a continuous toolpath while preventing printing on
1 2 78.50 78.86 0.46
2 2 75.60 75.84 0.32 the top of fresh material, right after extrusion. The intention was to
3 2 73.87 72.82 1.42 measure layer height deformation after printing each subsequent layer.
4 2 71.97 69.80 3.01 For this purpose, one of the layers (layer 3) was selected. The problem
5 2 69.93 66.78 4.51 with measuring layer height from the elevation viewpoint is that the
1 3 115.07 111.57 3.04
fresh concrete may dip on the edges, making it impossible to accurately
2 3 110.50 108.55 1.77
3 3 109.40 105.53 3.54 measure layer height (Fig. 12a). To overcome this difficulty, two thin
4 3 105.93 102.51 3.23 plastic cards were placed during printing under and on the top of layer 3
5 3 102.13 99.49 2.59 (Fig. 12b). These cards were placed at three different points of the third
1 4 145.40 144.28 0.77
layer to determine μ and σ . Fig. 13 shows the placement of the thin cards
2 4 142.90 141.26 1.15
3 4 137.47 138.24 0.56 on layer 3. A video camera was placed in front of the main front edge,
4 4 134.17 135.22 0.78 and the video was captured while the layers were deposited. The video
5 4 130.03 132.19 1.66
1 5 177.50 176.99 0.29
Table 10
2 5 175.47 173.97 0.85
3 5 169.33 170.95 0.95
Values of printing variables in Test 2 - Time interval for zero deformation.
4 5 164.97 167.92 1.79 Pump Flow Nozzle Robot Nozzle Lab Relative
5 5 159.23 164.90 3.56 Dial Rate size Speed distance Temperature Humidity
Ave % error of Eq. 1 in comparison to value obtained from Test1 2.93 (kg/ (mm) (mm/ to the (C◦ ) %
h) sec) bed
(mm)

7 850 24.5 118 15 19.4 28

Fig. 12. Diagram of Test 2 - Time interval for zero deformation, layer height
measuring strategy: a) deformation of layer edges by dipping, and b) mea­
surement of layer height by placing thin rigid cards in between layers dur­
ing printing.

Fig. 14. The printed specimen after printing all the 50 layers.

Fig. 13. The placement of thin rigid cards under and on the third layer at three different points.

9
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Fig. 15. Adjusting the printed bead properties for the new batch of material.

measured to establish the time interval needed to print the next layer to
Table 11
avoid deformation of the current layer. Table 11 presents the average
The average layer height of layer 3 after printing subsequent layers.
layer height of layer 3 after printing the subsequent layers on top.
Number of layers on Time Ave layer height Deformation The results show that, after printing the second layer on top of layer
top (sec) (mm) (mm)
3, the deformation remained the same (0.1 mm) thereby indicating that
0 0.00 15.0 0.0 after 33.9 ≈ 34 s the deformation of layer 3 stopped. This test was the
1 33.90 14.9 0.1 preliminary test for conducting test 3.
2 67.80 14.9 0.1
3 101.69 14.9 0.1
4 135.60 14.9 0.1
4.4. Test 3; Effect of different interlayer time intervals on layer height
deformation
was then used to extract a picture from each of the layers after it was
printed and before the next layer was deposited. These pictures were 4.4.1. Experimental setup
used for measuring the layer height at the three selected points. Table 10 After establishing a time interval after which deformation becomes
presents the values of relevant variables of the printing system. Fig. 14 zero, the next series of tests were undertaken to study the effect of
shows the printed specimen after printing all the 50 layers. different interlayer time intervals on layer height deformation under the
Although the material used in Test 2 had the same mix design as the weight of subsequent layer(s). For this purpose, 6 squares of different
material used in Test 1, trial tests showed a slight difference in behavior sizes with 5 layers were printed (Table 12). The 6 squares represented 6
between the materials batches used in each test. The two batches were different interlayer time intervals and the goal was to measure the
prepared at a different times of the year with different relative humidity nozzle distance to the previous printed layer (in short, nozzle to previous
and temperature, with the batch in Test 1 produced in winter and the layer distance) on one side of the square at multiple selected points
one in Test 2 in summer. In addition, the environmental conditions on (every 10 cm). Each of the 6 squares was printed at least 3 times to
the days of the two tests were also different. To compensate for such determine μ and σ . Table 13 shows the values of relevant variables of the
different behavior, the values of variables of the printing system, namely printing system in Test 3, which were adjusted following the same
water flow rate and robot speed, were adjusted so that the printed beads procedure described for Test 1.
had similar properties (layer width and layer height) in comparison to
the printed beads in Test 1. For this purpose, the water flow rate was Table 13
adjusted so the material was extrudable. Then a single bead was printed Values of printing variables in Test 3 - Effect of interlayer time interval on layer
in one layer with same nozzle distance to bed (15 mm). The layer width height deformation.
of the single bead was measured and with changing the robot speed Pump Flow Nozzle Robot Nozzle Lab Relative
adjusted to have the same width as a single bead in Test 1 (Fig. 15). Dial Rate size Speed distance Temperature Humidity
(kg/ (mm) (mm/ to the (C◦ ) %
h) sec) bed
4.3.2. Data Analysis and Discussion of Test 2 - Time interval for zero
(mm)
deformation
In Test 2, the material deformation in one layer (layer 3) was 7 850 24.5 157 15 26.6 38

Table 12
Printed samples specifications.

10
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Fig. 16. The average layer height for different interlayer time intervals.

4.5. Data Analysis and Discussion of Test 3 - Effect of interlayer time DN = (0.02823 × NL ) + ( − 0.004 × t) + 2.75
interval on layer height deformation
where, DN = nozzle distance to previous layer; NL = number of layers;
In Test 3, the nozzle distance to the previous printed layer was and t = time.
measured for each of the printed specimens, which represent 6 different The experimental results and the values estimated by Eq. 2 are
interlayer time intervals, to study its effect on layer height deformation. compared in Table 14. The results indicated that Eq. 2 could predict the
As shown in Fig. 16, maximum deformation occurred in sample 1 (S1) nozzle distance to previous layer for different number of layers and time
which is the smallest sample and as a result has the minimum time in­ with the average percentage error of 1.23.
terval between two consecutive layers. Moreover, deformation de­ In the next step, the printing time needed to print each of the samples
creases as the time interval increases, meaning that longer interlayer was divided by 34 s, which was shown to be a time interval after which
time intervals result in smaller deformations. deformation becomes zero. The outcome indicated the number of layers
A linear regression analysis was conducted on the results to develop needed for deformation to become zero in each sample. The values were
an equation that could predict the nozzle distance to previous layer (and round up in order to have a full layer. As can be seen in Table 15, in all
layer height deformation) based on number of layers and printing time. samples after S2, deformation becomes zero after printing 2 layers. This
Using regression analysis, it was attempted to estimate the relationships means that for any given layer, deformation is caused only by the next
between a dependent variable (nozzle distance to previous layer in this following layer, after which it stops.
case) and independent variables (number of layers and printing time). For instance, in sample 3, based on the results presented in Table 16,
Based on the result of the regression analysis (R2 = 0.87), Eq. 2 was the average nozzle distance to the previous layer is 15.1 mm for the
developed: second layer. This means there is a 0.1 mm deformation in the first layer
zND = axNL + byt + c (2) (Fig. 17a). Since there is no more deformation after printing two layers,
after printing layer 3, the first layer will have no more deformation and

11
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Table 14
Comparison between the average nozzle distances to previous layer estimated by
Eq. 2 and in Test 3 - Effect of interlayer time interval on layer height
deformation.
NL Time Observed ND in Test 3 Estimated ND byEq. 2 % error

1 12.74 15.00 15.30 1.97

2 12.74 15.47 15.73 1.72
3 12.74 16.17 16.18 0.11
4 12.74 17.03 16.65 2.27
5 12.74 17.93 17.12 4.51
1 15.29 15.00 15.12 0.83
2 15.29 15.30 15.56 1.68
3 15.29 15.78 16.00 1.39
4 15.29 16.47 16.46 0.03
5 15.29 17.37 16.93 2.50
1 17.83 15.00 14.96 0.30
2 17.83 15.17 15.38 1.43
3 17.83 15.68 15.82 0.90
4 17.83 16.23 16.28 0.27
5 17.83 16.65 16.74 0.56
1 20.38 15.00 14.79 1.41
2 20.38 15.08 15.21 0.89
3 20.38 15.52 15.65 0.80
4 20.38 16.00 16.09 0.59
5 20.38 16.42 16.56 0.81
1 22.93 15.00 14.62 2.52
2 22.93 15.06 15.04 0.09
3 22.93 15.42 15.47 0.32
Fig. 17. Diagram depicting layer by layer deformation.
4 22.93 15.88 15.91 0.22
5 22.93 16.25 16.37 0.74
1 25.48 15.00 14.46 3.61 second layer on top remained the same (0.2 mm) which confirms the
2 25.48 15.02 14.87 0.99 hypothesis that 34 s after printing a given layer the deformation of that
3 25.48 15.30 15.30 0.01
layer is zero.
4 25.48 15.56 15.74 1.17
5 25.48 15.83 16.19 2.24 The time interval of 33.9 s ≈ 34 s is, thus, the time after which no
Ave % error of Eq. 2 in comparison to value obtained from Test3 1.23 deformation occurs. Knowing this is quite helpful, as in the application
of additive manufacturing of concrete at construction scale large scale,
most layers would take more than 34 s to print. However, the exact time
at which the deformation becomes zero might be less than 33.9 s. To
Table 15
obtain a more precise value, Test 2 was repeated a specimen 5% smaller
The number of layers needed for deformation to become zero in each sample.
than S6. A 95 cm by 95 cm square was thus printed with multiple layers
S1 S2 S3 S4 S5 S6 (Fig. 19). The average layer height of layer 3 was measured after
Size 50*50 6*600 70*70 80*80 90*90 100*100 printing subsequent layers. As shown in Table 18, layer 3 still experi­
Printing time for one 12.74 15.29 17.83 20.38 22.93 25.48 enced deformation after printing the subsequent layer, which took
layer (Sec)
24.2 s. Together the result of this test and that of the previous test show
34/Printing time 2.66 2.22 1.90 1.66 1.48 1.33
Number of layers 3 3 2 2 2 2 that deformation must stop between 24.2 s and 33.9 s. It would be
needed for possible to carry our further tests with specimens with sides between
0 deformation 95 cm and 100 cm to zero in on the time interval at which deformation
stops, but the uncertainty acceptable as in most construction scale
printing layers likely that take more than 34 s to print.
Table 16
Average nozzle distance (ND) to the previous layer in sample 3. 5. Conclusions and future work
S3 Layer Number Average ND to the previous layer (mm)
1 15 This study is part of a larger study whose goal is to study and model
2 15.1 material deformation in construction scale additive manufacturing of
concrete to compensate for such a deformation in toolpath design. For
this purpose, a printing system consisting of a robotic arm and an in­
will keep the height of 14.9 mm, but layer 2, which had 15.1 mm ND to
dustrial scale mixer and pump were used, together with a Portland
previous layer after printing, will deform under the load of layer 3 and
cement-based concrete mix developed on purpose for 3D printing. Pre­
become 15.0 mm with 0.1 mm deformation (Fig. 17b). While printing
vious work addressed the effect of printing orientation and direction and
layer 4, the deformation of layer 2 will become zero and layer 3 which
the number of layers on layer height [20]. The current study was focused
was 15.1 mm in the beginning (because the ND was 15.1) will deform
on the effect of the number of layers and beads on layer width and the
0.1 mm, and this cycle will continue until all the layers are printed
effect of time on layer quality. Three tests were undertaken. The goal of
(Fig. 17c). It can be concluded that the deformation in the specimens
Test 1 was to model the effect of the number of beads and layers on layer
larger than 70 × 70 cm is negligible. This also applies to any specimen
width deformation. Results showed they have a cumulative impact that
in which each layers takes more than ≈18 s to print.
can be modeled by a linear equation. As concrete takes time to harden,
In order to validate this assumption, Test 2 was repeated for sample 3
the goal of Test 2 was to determine the time interval after which
(S3). For that, a 70 × 70 cm square was printed with multiple layers
deposited concrete no longer deforms. Results show this time interval to
(Fig. 18). After placing thin rigid cards under and on top of the third
be 33.9 s. The goal of Test 3 was then to determine the effect of different
layer, layer height was measured after printing each of the following
interlayer time intervals on layer height deformation. Regression anal­
layers. As can be observed in Table 17, the deformation after printing the
ysis was applied to model the deformation and determine the adequate

12
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

Fig. 18. Still images from video of the layer by layer printing process of the specimen printed to validate experimental results.

then compensation is necessary. This is the case of small building parts,

Table 17 such as wall sections between openings or construction details. This
The average layer height of layer 3 for the S3 sample. aspect is important when designing the toolpath and it needs to be
Number of layers on Time Ave layer height Deformation of layer 3 considered in the context of other toolpath design features, such as
top (sec) (mm) (mm) travel moves, which may require the printing machine to be equipped
0 0.00 15.0 0.0 with a stop and go feature. Analyzing the tradeoffs between alternative
1 17.61 14.9 0.1 toolpath designs and its impact on printing accuracy and speed is of the
2 35.22 14.8 0.2
utmost importance in designing for the AM of concrete, as it interferes
3 52.83 14.8 0.2
4 70.44 14.8 0.2 with the design of the structure and the strategy for decomposing it into
5 88.05 14.8 0.2 parts, layers, and beads for 3D printing,
6 105.66 14.8 0.2 The current work presents two main limitations. The first concerns
7 123.27 14.8 0.2 the time for zero deformation. Our work showed that before 24.2 s
deformation still occurs but after 33.9 s it stops. It would be possible to
distance of the printing nozzle to previous layers as a function of the carry our further tests to obtain a more precise value for the time interval
number of layers and the layer printing time. A specimen was then at which deformation stops, but the uncertainty is acceptable as most of
produced with several layers, with each two layers taking about 35 s to construction scale printing will likely require layers that take longer
print. Results show that each layer only deformed under the weight of
the layer immediately on top and in the predicted way, thereby vali­ Table 18
dating the study. These results can be used as a decisive factor in tool­ The average layer height of layer 3 for 95 cm by 95 cm specimen.
path design and in the design of whole parts or structures. It can dictate Number of layers on Time Ave layer height Deformation of layer 3
the scale, wall thickness, infill pattern and inevitable travel moves. top (sec) (mm) (mm)
In cases in which the path length is such that the time travelled for 0 0.00 15.00 0.00
the nozzle to reach a certain point on the above layer is bigger than the 1 24.20 14.92 0.08
time interval for zero deformation, there is no need for deformation 2 48.40 14.91 0.09
compensation. This is the case of toolpaths for printing large construc­ 3 145.20 14.91 0.09
4 96.80 14.91 0.09
tion elements, such as walls and pavements. If path length is shorter,

Fig. 19. Still images from video of the layer by layer printing process of a 95 cm by 95 cm specimen.

13
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

than 34 s to print. The second limitation is that the obtained time in­ [7] A. Kazemian, X. Yuan, E. Cochran, B. Khoshnevis, Cementitious materials for
construction-scale 3D printing: laboratory testing of fresh printing mixture, Constr.
terval for zero deformation is valid for the material used, with its specific
Build. Mater. 145 (2017) 639–647, https://doi.org/10.1016/j.
mix design, or for materials with similar rheological and hardening conbuildmat.2017.04.015.
properties. By studying the relationship between the rheological and [8] T.T. Le, S.A. Austin, S. Lim, R.A. Buswell, A.G.F. Gibb, T. Thorpe, Mix design and
strength properties of different printable concrete mixtures and the fresh properties for high-performance printing concrete, Mater. Struct. Constr. 45
(2012) 1221–1232, https://doi.org/10.1617/s11527-012-9828-z.
corresponding material deformation over time, it will be possible to [9] T.T. Le, S.A. Austin, S. Lim, R.A. Buswell, R. Law, A.G.F. Gibb, T. Thorpe, Hardened
develop a generic model to predict material deformation based on the properties of high-performance printing concrete, Cem. Concr. Res. 42 (2012)
rheological and strength properties of fresh concrete and take it into 558–566, https://doi.org/10.1016/j.cemconres.2011.12.003.
[10] I. Gibson, T. Kvan, L.W. Ming, Rapid prototyping for architectural models, Rapid
account in toolpath design. Additionally, it will be necessary to conduct Prototyp. J. 8 (2002) 91–99, https://doi.org/10.1108/13552540210420961.
the material rheological tests in a controlled lab environment to find the [11] P. Wu, J. Wang, X. Wang, A critical review of the use of 3-D printing in the
effect of temperature and humidity on material behavior. construction industry, Autom. Constr. 68 (2016) 21–31, https://doi.org/10.1016/j.
autcon.2016.04.005.
In summary, the work presented in this paper represents one step [12] R. Rael, V. San Fratello, Printing Architecture: Innovative Recipes for 3D Printing,
towards modeling the interplay between different variables connected Chronicle Books, 2018.
to materials, printing system and design in the AM of concrete. It focused [13] S. Lim, T. Le, J. Webster, R. Buswell, S. Austin, A. Gibb, T. Thorpe, Fabricating
construction components, 2009.
on the experimental prediction of material deformation for a given [14] B. Khoshnevis, Automated construction by contour crafting - related robotics and
mixture to guarantee shape accuracy (print quality) of the printed information technologies, Autom. Constr. 13 (2004) 5–19, https://doi.org/
structures. Future research will be twofold. First, it will target the 10.1016/j.autcon.2003.08.012.
[15] O. Kontovourkis, G. Tryfonos, C. Georgiou, Robotic additive manufacturing (RAM)
enlargement of this predictive model to encompass mixtures with other
with clay using topology optimization principles for toolpath planning: the
rheological properties and the effect of environmental variables. Second, example of a building element, Archit. Sci. Rev. 63 (2020) 105–118, https://doi.
it will address the study of the relationship between printing features org/10.1080/00038628.2019.1620170.
and the strength of the printed structure. The ultimate goal will be a [16] R.A. Buswell, W.R. Leal de Silva, S.Z. Jones, J. Dirrenberger, 3D printing using
concrete extrusion: a roadmap for research, Cem. Concr. Res. 112 (2018) 37–49,
model relating printing features to shape accuracy and structural https://doi.org/10.1016/j.cemconres.2018.05.006.
strength, so that printing variables can be manipulated to achieve [17] T. Wangler, E. Lloret, L. Reiter, N. Hack, F. Gramazio, M. Kohler, M. Bernhard,
desired levels of quality and strength. B. Dillenburger, J. Buchli, N. Roussel, R. Flatt, Digital concrete: opportunities and
challenges, RILEM Tech. Lett. 1 (2016) 67, https://doi.org/10.21809/
rilemtechlett.2016.16.
CRediT authorship contribution statement [18] N. Roussel, Rheological requirements for printable concretes, Cem. Concr. Res. 112
(2018) 76–85, https://doi.org/10.1016/j.cemconres.2018.04.005.
[19] D. Rangeard Perrot, A. Pierre, Structural built-up of cement-based materials used
The paper study the effect of time, the number of beads and the for 3D-printing extrusion techniques, Mater. Struct. Constr. 49 (2016) 1213–1220,
number of layers on deformation of a printed concrete mix. Negar is the https://doi.org/10.1617/s11527-015-0571-0.
[20] Y.W.D. Tay, G.H.A. Ting, Y. Qian, B. Panda, L. He, M.J. Tan, Time gap effect on
PhD student who carried out the research, advised by J. Duarte in design
bond strength of 3D-printed concrete, Virtual Phys. Prototyp. 14 (2019) 104–113,
computing, S. Nazarian in material matters, and N. Meisel in engineer­ https://doi.org/10.1080/17452759.2018.1500420.
ing design. [21] Q. Yuan, D. Zhou, B. Li, H. Huang, C. Shi, Effect of mineral admixtures on the
structural build-up of cement paste, Constr. Build. Mater. 160 (2018) 117–126,
https://doi.org/10.1016/j.conbuildmat.2017.11.050.
[22] V.N. Nerella, M.A.B. Beigh, S. Fataei, V. Mechtcherine, Strain-based approach for
Declaration of Competing Interest measuring structural build-up of cement pastes in the context of digital
construction, Cem. Concr. Res. 115 (2019) 530–544, https://doi.org/10.1016/j.
The authors declare that they have no known competing financial cemconres.2018.08.003.
[23] B. Zareiyan, B. Khoshnevis, Interlayer adhesion and strength of structures in
interests or personal relationships that could have appeared to influence contour crafting - effects of aggregate size, extrusion rate, and layer thickness,
the work reported in this paper. Autom. Constr. 81 (2017) 112–121, https://doi.org/10.1016/j.
autcon.2017.06.013.
[24] E. Keita, H. Bessaies-Bey, W. Zuo, P. Belin, N. Roussel, Weak bond strength
Acknowledgements between successive layers in extrusion-based additive manufacturing:
measurement and physical origin, Cem. Concr. Res. 123 (2019), 105787, https://
doi.org/10.1016/j.cemconres.2019.105787.
This research was financially sponsored by The Raymond A. Bowers
[25] N. Ashrafi, J.P. Duarte, S. Nazarian, N.A. Meisel, Evaluating the relationship
Program for Excellence in Design and Construction of the Built Envi­ between deposition and layer quality in large-scale additive manufacturing of
ronment, The Pennsylvania State University, Autodesk, Inc. ®, and Golf concrete, Virtual Phys. Prototyp. 14 (2019) 135–140, https://doi.org/10.1080/
17452759.2018.1532800.
Concrete Technology (GCT). The authors express their gratitude to Dr.
[26] Y.W. Tay, B. Panda, S.C. Paul, M.J. Tan, S.Z. Qian, K.F. Leong, C.K. Chua,
Sven Bilén, Dr. Ali Memari, Dr. Aleksandra Radlińska, Mr. Jamie Heil­ Processing and properties of construction materials for 3D printing, Mater. Sci.
man, Mr Zhanzhao Li, and Mr. Nathan Watson, for their valuable in­ Forum 861 (2016) 177–181, https://doi.org/10.4028/www.scientific.net/
sights and contributions to this research. MSF.861.177.
[27] S. Lim, R.A. Buswell, T.T. Le, S.A. Austin, A.G.F. Gibb, T. Thorpe, Developments in
construction-scale additive manufacturing processes, Autom. Constr. 21 (2012)
References 262–268, https://doi.org/10.1016/j.autcon.2011.06.010.
[28] Z. Li, M. Hojati, Z. Wu, J. Piasente, N. Ashrafi, J.P. Duarte, S. Nazarian, S.G. Bilén,
A.M. Memari, A. Radlińska, Fresh and hardened properties of extrusion-based 3D-
[1] P. Carneau, R. Mesnil, N. Roussel, O. Baverel, An exploration of 3D printing design
printed cementitious materials: a review, Sustain 12 (2020) 1–33, https://doi.org/
space inspired by masonry, Proc. IASS Annu. Symp. (2019) 1–9, https://doi.org/
10.3390/su12145628.
10.5281/zenodo.3563672.
[29] J. Kruger, S. Zeranka, G. van Zijl, 3D concrete printing: a lower bound analytical
[2] R.A. Buswell, W.R. Leal de Silva, S.Z. Jones, J. Dirrenberger, 3D printing using
model for buildability performance quantification, Autom. Constr. 106 (2019),
concrete extrusion: a roadmap for research, Cem. Concr. Res. 112 (2018) 37–49,
102904, https://doi.org/10.1016/j.autcon.2019.102904.
https://doi.org/10.1016/j.cemconres.2018.05.006.
[30] A.S.J. Suiker, Mechanical performance of wall structures in 3D printing processes:
[3] D. Ding, Z. Pan, D. Cuiuri, H. Li, A practical path planning methodology for wire
theory, design tools and experiments, Int. J. Mech. Sci. 137 (2018) 145–170,
and arc additive manufacturing of thin-walled structures, Robot. Comput. Integr.
https://doi.org/10.1016/j.ijmecsci.2018.01.010.
Manuf. 34 (2015) 8–19, https://doi.org/10.1016/j.rcim.2015.01.003.
[31] R.J.M. Wolfs, F.P. Bos, T.A.M. Salet, Early age mechanical behaviour of 3D printed
[4] S.H. Choi, W.K. Zhu, A dynamic priority-based approach to concurrent toolpath
concrete: numerical modelling and experimental testing, Cem. Concr. Res. 106
planning for multi-material layered manufacturing, CAD Comput. Aided Des. 42
(2018) 103–116, https://doi.org/10.1016/j.cemconres.2018.02.001.
(2010) 1095–1107, https://doi.org/10.1016/j.cad.2010.07.004.
[32] R.J.M. Wolfs, A.S.J. Suiker, Structural failure during extrusion-based 3D printing
[5] Y. Jin, J. Du, Z. Ma, A. Liu, Y. He, An optimization approach for path planning of
processes, Int. J. Adv. Manuf. Technol. 104 (2019) 565–584, https://doi.org/
high-quality and uniform additive manufacturing, Int. J. Adv. Manuf. Technol. 92
10.1007/s00170-019-03844-6.
(2017) 651–662, https://doi.org/10.1007/s00170-017-0207-3.
[33] Z. Malaeb, H. Hachem, A. Tourbah, T. Maalouf, N. El Zarwi, F. Hamzeh, 3D
[6] Y. an Jin, Y. He, J. zhong Fu, W. feng Gan, Z. wei Lin, Optimization of tool-path
concrete printing: machine and mix design, Int. J. Civ. Eng. Technol. 6 (2015)
generation for material extrusion-based additive manufacturing technology, Addit.
14–22, http://www.researchgate.net/profile/Farook_Hamzeh/publication/
Manuf. 1 (2014) 32–47, https://doi.org/10.1016/j.addma.2014.08.004.

14
N. Ashrafi et al. Additive Manufacturing 37 (2021) 101656

280488795_3D_Concrete_Printing_Machine_and_Mix_Design/links/ Proc. Int. Conf. Prog. Addit. Manuf. 2018-May 2018 62–67. 〈https://doi.org/1
55b608c308aec0e5f436d4a1.pdf. 0.25341/D4QG6D〉.
[34] A. Perrot, D. Rangeard, A. Pierre, Structural built-up of cement-based materials [38] B. Panda, S.C. Paul, N.A.N. Mohamed, Y.W.D. Tay, M.J. Tan, Measurement of
used for 3D-printing extrusion techniques, Mater. Struct. Constr. 49 (2016) tensile bond strength of 3D printed geopolymer mortar, Measurement 113 (2018)
1213–1220, https://doi.org/10.1617/s11527-015-0571-0. 108–116.
[35] L. Reiter, T. Wangler, N. Roussel, R.J. Flatt, The role of early age structural build- [39] R.J.M. Wolfs, F.P. Bos, E.C.F. van Strien, T.A.M. Salet, A Real-Time Height
up in digital fabrication with concrete, Cem. Concr. Res. 112 (2018) 86–95, Measurement and Feedback System for 3D Concrete Printing, Springer
https://doi.org/10.1016/j.cemconres.2018.05.011. International Publishing, Cham, 2018, pp. 2474–2483. https://doi.org/10.1007/9
[36] R.J.M. Wolfs, F.P. Bos, T.A.M. Salet, Hardened properties of 3D printed concrete: 78-3-319- 59471-2_282.
the influence of process parameters on interlayer adhesion, Cem. Concr. Res. 119 [40] GCT Structural Mortal Mix 4000 PSI FS; Gulf Concrete Technology: Long Beach,
(2019) 132–140, https://doi.org/10.1016/j.cemconres.2019.02.017. MS, November 2018.
[37] B. Panda, N.A. Noor Mohamed, Y.W.D. Tay, L. He, M.J. Tan, Effects of slag addition
on bond strength of 3D printed geopolymer mortar: an experimental investigation,

15