Optimising 3D Printed Concrete Structures Using Topology Optimisation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/322168562

Optimising 3D Printed Concrete Structures Using Topology Optimisation

Chapter · January 2017


DOI: 10.1007/978-3-319-59471-2_37

CITATIONS READS

7 865

4 authors:

Pascal Martens Maarten Mathot


Delft University of Technology 4 PUBLICATIONS   8 CITATIONS   
1 PUBLICATION   7 CITATIONS   
SEE PROFILE
SEE PROFILE

Freek Bos Jeroen Coenders


Eindhoven University of Technology White Lioness technologies
85 PUBLICATIONS   1,503 CITATIONS    38 PUBLICATIONS   99 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

3TU.Bouw Lighthouse 2016 project on Optimising 3D printed concrete View project

3D Concrete Printing View project

All content following this page was uploaded by Jeroen Coenders on 17 April 2018.

The user has requested enhancement of the downloaded file.


Proceedings of the IASS Annual Symposium 2017
“Interfaces: architecture.engineering.science”
25 - 28th September, 2017, Hamburg, Germany
Annette Bögle, Manfred Grohmann (eds.)

Optimising 3D printed concrete structures using topology


optimisation
Pascal A. MARTENS*, Maarten MATHOTa, Jeroen L. COENDERSb, Freek P. BOSc, Jan G. ROTSd,

*White Lioness technologies & Delft University of Technology


Linge 63, 2911EJ Nieuwerkerk aan den IJssel, The Netherlands
pascalmartens1@gmail.com
a
White Lioness technologies & Delft University of Technology
b
White Lioness technologies
c
Eindhoven University of Technology
d
Delft University of Technology

Abstract
Additive manufacturing and 3D printing are rapidly developing digital fabrication techniques (Lu et al.
[10]). After the first developments in small scale printing of metals (Frazier [6]) and plastics (Gibson et
al. [7]) have been made, research from various groups around the world is now also focusing on large
scale printing in concrete (Lim et al. [9]) and making this technology more suitable for the construction
scale. The use of this technology in shell and spatial structures will enable the realisation of complex
and/or customised concrete designs, with an expected reduction in cost and construction time in
comparison with current common practice. Additionally, this new technology will provide opportunities
to create more efficient structures. Structures can already be optimised in the early stages of the design
for weight and structural performance, however the resulting optimised structures are often difficult to
manufacture due to the complexity of the resulting geometry. Additive manufacturing can be the key to
enable the realisation of complex designs without high costs for moulds and labour. Optimising designs
for additive manufacturing is a promising design step on the interface of architecture and engineering.
This paper will present a novel methodology to include material performance and manufacturing
constraints of 3D printed concrete in design optimisation processes. The study examines the possibility
to optimise concrete structures in the design phase. In order to save material and thus create more
sustainable and more cost efficient structures, a topology optimisation tool has been created specifically
for 3D printed concrete. Traditional topology optimisation methods consider isotropic material and will
not necessarily produce realisable and reliable optimised structures. In the algorithm presented
constraints of the printing process and material properties from physical testing of this layered material
have been considered in the optimisation. By adopting this methodology more realistic and feasible
optimal concrete structures can be designed.
Keywords: Conceptual design, structural optimisation, topology optimisation, concrete shell structures, additive
manufacturing, 3D concrete printing

1. Introduction
Shell and spatial structures in challenging shapes have been designed for many years. Architects and
engineers initially started the design of these kind of structures by creating physical models to simulate
the behaviour of the final structure or they designed merely based on their experience and knowledge
from previous projects. In the late 20th century the development of computers and engineering software
started. In addition to physical modelling it now became possible to easily design complex geometrical
structures on the computer: computational modelling. The translation from these conceptual digital

Copyright © 2017 by P. Martens, M. Mathot, J. Coenders, F. Bos and J. Rots


Published by the International Association for Shell and Spatial Structures (IASS) with permission.
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

designs towards actual structures is still a challenging step. Available architectural software now provide
for the possibility to design in shapes only limited by the imagination of the designers. Structural
software is developed to calculate and check structures by means of finite element analysis. However,
the next phase in the design process, the manufacturing of these structures (or parts of the structure) still
needs attention. Manufacturing of challenging structures is still labour and material-intensive.
Automation in construction can improve this.
In addition to the design freedom that computational modelling offers designers and engineers,
computational techniques can also be used to optimise structures. Optimisation in the design phase can
help design efficient structures and therefore save material. Structures can be optimised in many
different ways. One method is the “topology optimisation” method (Bendsøe and Sigmund [1]). In this
method a computational technique is used to distribute a certain limited amount of material in a
predetermined design space. The material is placed in such a way that the resulting structure complies
with the boundary conditions, such as the acting loads and the supports. Using this technique it is
possible to improve the topology of connections, parts of structures or structures as a whole. By using
topology optimisation in the design savings in material can thus be realised, which can lead to more
sustainable and cost effective structures than structures resulting from conventional design methods.
In topology optimisation material can be removed from every position in the design space. The resulting
designs from the technique are therefore characterised by a complex geometry. To enable manufacturing
of these special and in some cases spatial structures additive manufacturing or 3D printing can be used
to realise the final design. Using traditional manufacturing techniques to realise the optimised designs
would lead to waste material and additional labour. Additive manufacturing only places material on the
desired positions and is therefore more material efficient and because of the use of a printing robot less
labour intensive.
This paper will shortly present some promising and ground-breaking concrete additive manufacturing
projects around the globe as well as some useable optimisation techniques. The possibilities and
challenges of the combination of concrete additive manufacturing and topology optimisation in the
building industry are considered and subsequently an algorithm is presented that takes into account some
material specific and manufacturing constraints.

2. Concrete additive manufacturing


The name “additive manufacturing” describes the technologies that build 3D objects by adding material
layer by layer. Research in the field of this manufacturing technique started with the invention of stereo-
lithography by Chuck Hall in the late 1980s (Melchels et al. [11]). In this first 3D printing method an
ultraviolet curable resin was used. This resin was hardened during the manufacturing process by a light
source. After this first attempt many different materials and processes were investigated. In the late
1990s research started focussing on concrete and concrete like materials as the printed material (Pegna
[12]).
3D printing of metals and plastics can be considered small scale additive manufacturing. To make this
technology applicable to structures or buildings a shift to a larger scale is necessary. Various groups
around the globe are therefore now focusing on printing in concrete and moving to larger scales.
Contour Crafting is the first important concrete additive manufacturing project. Behrokh Khoshnevis of
the University of Southern California wanted to automate the construction process, in order to improve
the speed, labour efficiency and durability, whilst lowering the accident rate and the cost of construction
(Khoshnevis [8]). A movable, computer controlled nozzle, or extruder, was used to place layers of
concrete on top of each other. In the footsteps of Khoshnevis his ex-co-worker Ma YiHe started the
company WinSun and used a technique similar to Contour Crafting to actually build 3D printed houses.
A different approach, more comparable to stereo-lithography, was developed by Enrico Dini. In his D-

2
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

Shape project he also manufactured a structure layer wise, however an extruder was not used in this
process in comparison to the earlier mentioned projects. Dini placed a layer of sand-like material on a
baseplate. Using a fluid mixture or hardener he solidifies the material at the required positions.
Subsequently the next layers of sand are placed and hardened at the determined positions. When the top
of the design space is reached, the loose material is removed which leaves the actual structure.
After these ground-breaking projects many research projects started as well. 3DCP (3D Concrete
Printing) at the Loughborough University (Buswell et all. [3]) and at the University of Technology in
Eindhoven are examples of projects focussing on understanding and improving the printed material, the
used 3D printer and other aspects of the printing process.

3. Topology Optimisation
In order to design more sustainable structures material needs to be saved. To determine the minimum
required amount of material in a structure to still fulfil its functions many different techniques are
available. Common important variables in optimising structures are the used material and the section
sizes of structural elements. For example in composite floors steel is used at the bottom, the area under
tension, and concrete is used at the top, the part of the section that is in compression. The dimensions of
the section can be varied as well to obtain a structure that is safe using the least material possible.
To reduce the used amount of material in a structure also different approaches are followed. It is also
possible to consider the shape of a structure and make adaptations to that shape to obtain a uniform
stress distribution in the entire structure. This approach is called shape optimisation. The topology of a
structure remains the same, however the geometry and subsequently the load carrying behaviour can be
improved.
A more technically challenging and economically rewarding optimisation technique is topology
optimisation. Rather than limiting the changes to the sizes of structural components, topology
optimisation provides much more freedom and allows the designer to create totally novel and highly
efficient conceptual designs (Bendsøe and Sigmund [1]). In this method the shape and the sizes are
completely variable. The predefined parameters in topology optimisation are the available design
space, the acting forces and the boundary conditions. An algorithm distributes material over the design
space in the most efficient way.

Figure 1: Initial design (left) and topology optimised design (right)


Different topology optimisation methods can be used to distribute a limited amount of material over the
available space. The optimisation method used in this paper is the modified SIMP or Solid Isotropic
Material with Penalisation method (Bendsøe and Sigmund [1]). In this method of optimisation the
prescribed design space is divided into small parts, the elements. Every element is given a variable
density between 0 and 1. The algorithm steers every single element toward a void (0) or a solid element
(1). The objective of the method is to find a distribution of the mass over the design space that minimises
the work done by the external forces that were prescribed. In other words, the algorithm minimises the
compliance and thus maximises the stiffness of the structure whilst only making use of a chosen fraction
of the initial volume.

3
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

4. Combination of the techniques


Topology optimisation can help design more efficient special and spatial structures. Concrete additive
manufacturing can be a solution to the manufacturing difficulties of the complex outcomes of these
optimisations. By combining the two mentioned techniques material can be saved and manual labour
can be minimised, which can lead to cheaper and more sustainable structures. Important aspects that
need to be considered in the combination of the two methods are the material properties of printed
concrete and the manufacturing process.
Topology optimisation methods that are currently available assume linear elastic and isotropic material
behaviour. When printing in layers, instead of casting, the material behaves in a different way. The
tested specimens seem to be orthotropic (Doomen [5]) due to the layer wise built-up. The material
properties also differ from regular concrete because of the printing process. A different concrete mixture
is needed to assure workability, bonding between the layers and to make sure the mixture hardens fast
enough in order to support the next printed layer. These properties should be incorporated into the
optimisation.
The manufacturing of concrete using 3D printing also comes with some constraints. The angle of
overhanging layers is very small when printing in concrete. This printing constraint should be considered
in the algorithm in order to be able to actually print the final outcome.

5. Algorithm
The algorithm written for this paper is using a density-based approach: the modified SIMP method. In
this approach the design space in which the final design should fit is divided into small elements and
they are all given a virtual density (𝑥𝑥𝑖𝑖 ) between 0 and 1, for example 0.5. This normalised value relates
to the mechanical properties of the material element. The relation between this density and the Young’s
modulus 𝐸𝐸𝑖𝑖 in an element is given by:
𝑝𝑝
𝐸𝐸𝑖𝑖 = 𝐸𝐸𝑖𝑖 (𝑥𝑥𝑖𝑖 ) = 𝐸𝐸𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑥𝑥𝑖𝑖 (𝐸𝐸0 − 𝐸𝐸𝑚𝑚𝑚𝑚𝑚𝑚 ), 𝑥𝑥𝑖𝑖 ∈ [0,1], (1)
in which 𝐸𝐸0 is the Young’s modulus of a solid element, 𝐸𝐸𝑚𝑚𝑚𝑚𝑚𝑚 is the Young’s modulus of a void element,
which is not equal to zero to avoid singularity of the stiffness matrix used for the finite element analysis.
The penalisation power 𝑝𝑝 is larger than 1 and makes intermediate density values steer towards 0 or 1,
so that in the final configuration each element is either filled with solid material or left void.
When running this algorithm it possible to encounter some numerical difficulties. It is possible to reach
unsolvable or partially optimised configurations. Examples of these problems are the existence of
checkerboard patterns, mesh-dependency problems and local minima (Bendsøe and Sigmund [1]).
To avoid the mentioned problems a density filter is used that is proposed by Bruns and Tortorelli [4]:
∑𝑗𝑗∈𝑁𝑁 𝐻𝐻𝑖𝑖𝑖𝑖 𝑣𝑣𝑗𝑗 𝑥𝑥𝑗𝑗
𝑥𝑥�𝑖𝑖 = 𝑖𝑖
∑𝑗𝑗∈𝑁𝑁 𝐻𝐻𝑖𝑖𝑖𝑖 𝑣𝑣𝑗𝑗
, (2)
𝑖𝑖

where 𝑣𝑣𝑗𝑗 is the volume of element 𝑗𝑗, 𝑁𝑁𝑖𝑖 is the neighbourhood of an element i defined by:
𝑁𝑁𝑖𝑖 = {𝑗𝑗: 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑(𝑖𝑖, 𝑗𝑗) ≤ 𝑅𝑅}, (3)
where 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑(𝑖𝑖, 𝑗𝑗) is the distance between the centre of an element 𝑖𝑖 and the centre of a different element
𝑗𝑗 and 𝑅𝑅 is the filter size. The larger the filter size, the larger the neighbourhood. The element 𝐻𝐻𝑖𝑖𝑖𝑖 in the
filter density function is defined as:
𝐻𝐻𝑖𝑖𝑖𝑖 = 𝑅𝑅 − 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑(𝑖𝑖, 𝑗𝑗), (4)

4
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

Figure 2: Optimised result without (left) and with (right) density filter
The filtered density 𝑥𝑥�𝑖𝑖 is now used in the topology optimisation instead of 𝑥𝑥𝑖𝑖 .

The objective of the optimisation is to minimise the compliance. In order to calculate that value first the
displacements of the eight nodes that bound the elements should be determined using finite element
analysis.

Figure 3: The used element: eight-node hexahedron with its natural coordinates

For this calculation the elemental stiffness matrix per element needs to be determined. This stiffness
matrix 𝑘𝑘𝑖𝑖 (𝑥𝑥�𝑖𝑖 ) is the volume integral of the constitutive or stiffness matrix 𝐶𝐶𝑖𝑖0 and the strain-displacement
matrix B following:
𝑘𝑘𝑖𝑖 (𝑥𝑥�𝑖𝑖 ) = ∫𝑉𝑉 𝐵𝐵𝑇𝑇 𝐶𝐶𝑖𝑖0 𝐵𝐵 𝑑𝑑𝑑𝑑 (5)
𝑒𝑒

where for an eight-node hexahedron with natural coordinates 𝜉𝜉1 , 𝜉𝜉2 𝑎𝑎𝑎𝑎𝑎𝑎 𝜉𝜉3 the matrix 𝐵𝐵 is given by:

(6)

The constitutive matrix 𝐶𝐶 is the inverse matrix of the compliance matrix 𝑆𝑆. For elastic materials the
compliance matrix 𝑆𝑆 is described in the following equation:
𝜀𝜀 = 𝑆𝑆𝑆𝑆 (7)
We assume the printed concrete used in manufacturing of the final result is orthotropic. To determine
the matrix 𝑆𝑆 nine engineering constants have to be determined in tests: the three Young’s moduli, three
shear moduli and the Poisson’s ratios in the x-, y- and z-direction.

5
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

(8)

𝜈𝜈𝑦𝑦𝑦𝑦 𝜈𝜈𝑧𝑧𝑧𝑧 𝜈𝜈𝑧𝑧𝑧𝑧 𝜈𝜈𝑥𝑥𝑥𝑥 𝜈𝜈𝑥𝑥𝑥𝑥 𝜈𝜈𝑦𝑦𝑦𝑦


where 𝐸𝐸 𝑦𝑦
= ,
𝐸𝐸 𝑧𝑧 𝐸𝐸 𝑧𝑧
= ,
𝐸𝐸 𝑥𝑥 𝐸𝐸 𝑥𝑥
= 𝐸𝐸 𝑦𝑦
.

Now 𝐶𝐶𝑖𝑖0 = 𝑆𝑆 −1 and the elemental stiffness matrix 𝑘𝑘𝑖𝑖 (𝑥𝑥�𝑖𝑖 ) (or global version 𝐾𝐾 0𝑖𝑖 ) can be used to obtain
the total global stiffness matrix 𝐾𝐾(𝑥𝑥�) by assembling the elemental parts:
𝑝𝑝
𝐾𝐾(𝑥𝑥�) = ∑𝑛𝑛𝑖𝑖=1�𝐸𝐸𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑥𝑥�𝑖𝑖 (𝐸𝐸0 − 𝐸𝐸𝑚𝑚𝑚𝑚𝑚𝑚 )�𝐾𝐾0𝑖𝑖 . (9)
Using the equilibrium equation the global displacements 𝑈𝑈(𝑥𝑥�) of the nodes are calculated:
𝐾𝐾(𝑥𝑥�)𝑈𝑈(𝑥𝑥�) = 𝐹𝐹, (10)
where 𝐹𝐹 is the force vector, which is independent of the filtered virtual densities of the elements.
Using the displacements, the compliance can be calculated and the optimisation problem can be solved.
The objective is to minimise the compliance under prescribed loading and support conditions for a
certain design space and a chosen fraction of the total volume. The total compliance is defined as:
𝑐𝑐(𝑥𝑥�) = 𝐹𝐹 𝑇𝑇 𝑈𝑈(𝑥𝑥�) (11)
The chosen fraction of the total volume is the volume constraint. So when we take into account that
volume constraint the optimisation problem can be stated as:
find 𝑥𝑥 = [𝑥𝑥1 , 𝑥𝑥2 , … , 𝑥𝑥𝑒𝑒 , … , 𝑥𝑥𝑛𝑛 ]𝑇𝑇
minimize 𝑐𝑐(𝑥𝑥�) = 𝐹𝐹 𝑇𝑇 𝑈𝑈(𝑥𝑥�) = 𝑈𝑈(𝑥𝑥�)𝑇𝑇 𝐾𝐾(𝑥𝑥�) 𝑈𝑈(𝑥𝑥�) = ∑ 𝑢𝑢𝑖𝑖𝑇𝑇 𝑘𝑘𝑖𝑖 𝑢𝑢𝑖𝑖
subject to 𝑣𝑣(𝑥𝑥�) = 𝑥𝑥� 𝑇𝑇 𝑣𝑣 − 𝑣𝑣� ≤ 0
𝑥𝑥 ∈ 𝜒𝜒, 𝜒𝜒 = {𝑥𝑥 ∈ ℝ𝑛𝑛 : 0 ≤ 𝑥𝑥 ≤ 1}
𝜕𝜕𝜕𝜕(𝑥𝑥�)
The compliance 𝑐𝑐(𝑥𝑥�), the derivative of the compliance and the derivative of the volume constraint
𝜕𝜕𝑥𝑥𝑒𝑒
𝜕𝜕𝜕𝜕(𝑥𝑥�)
with respect to the design variable 𝑥𝑥𝑒𝑒 can be calculated:
𝜕𝜕𝑥𝑥𝑒𝑒
𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝜕𝜕𝑥𝑥�𝑖𝑖 𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝜕𝜕𝑥𝑥�𝑖𝑖 𝐻𝐻𝑖𝑖𝑖𝑖 𝑣𝑣𝑒𝑒
= ∑𝑖𝑖∈𝑁𝑁𝑒𝑒 where = 𝑣𝑣𝑖𝑖 and =∑ (12)
𝜕𝜕𝑥𝑥𝑒𝑒 𝜕𝜕𝑥𝑥𝑖𝑖 𝜕𝜕𝑥𝑥𝑒𝑒 𝜕𝜕𝑥𝑥𝑖𝑖 𝜕𝜕𝑥𝑥𝑒𝑒 𝑗𝑗∈𝑁𝑁𝑖𝑖 𝐻𝐻𝑖𝑖𝑖𝑖 𝑣𝑣𝑗𝑗

in which 𝑣𝑣𝑖𝑖 = 𝑣𝑣𝑗𝑗 = 𝑣𝑣𝑒𝑒 = 1 because the used elements have equal sides with a length of 1.
𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝜕𝜕𝑥𝑥�𝑖𝑖 𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝑝𝑝−1
= ∑𝑖𝑖∈𝑁𝑁𝑒𝑒 where = −𝑢𝑢𝑖𝑖 (𝑥𝑥�)𝑇𝑇 �𝑝𝑝𝑥𝑥�𝑖𝑖 (𝐸𝐸0 − 𝐸𝐸𝑚𝑚𝑚𝑚𝑚𝑚 )𝑘𝑘𝑖𝑖0 �𝑢𝑢𝑖𝑖 (𝑥𝑥�) (13)
𝜕𝜕𝑥𝑥𝑒𝑒 𝜕𝜕𝑥𝑥𝑖𝑖 𝜕𝜕𝑥𝑥𝑒𝑒 𝜕𝜕𝑥𝑥𝑖𝑖

To obtain an improved structure the design variables (𝑥𝑥 𝑘𝑘 ) are updated during multiple iteration steps.
This optimisation problem is a non-linear programming problem, which can be solved using different
methods such as the MMA (Method of Moving Asymptotes) and the Optimality Criteria (OC) method.
The OC method is a relatively simple method which is able to find a convex approximation of the
original problem from which an improved set of variables (𝑥𝑥 𝑘𝑘+1 ) can be derived.

6
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

According to the OC method convergence is achieved when the following condition holds:
𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝜕𝜕𝜕𝜕(𝑥𝑥�)
+ 𝜆𝜆 = 0, (14)
𝜕𝜕𝑥𝑥𝑒𝑒 𝜕𝜕𝑥𝑥𝑒𝑒

where 𝜆𝜆 is the Lagrange multiplier and the condition can also be denoted as:
𝜕𝜕𝜕𝜕(𝑥𝑥�) 𝜕𝜕𝜕𝜕(𝑥𝑥�) −1
𝐵𝐵𝑒𝑒 = − 𝜕𝜕𝑥𝑥𝑒𝑒
�𝜆𝜆 𝜕𝜕𝑥𝑥𝑒𝑒
� (15)

𝜆𝜆 is found using the bisection method (Burden and Faires [2]), and the condition above is solved using
a scheme proposed by Bendsøe and Sigmund [1] :
𝜂𝜂
max(0, 𝑥𝑥𝑒𝑒 − 𝑚𝑚) , 𝑖𝑖𝑖𝑖 𝑥𝑥𝑒𝑒 𝐵𝐵𝑒𝑒 ≤ max(0, 𝑥𝑥𝑒𝑒 − 𝑚𝑚) ,
𝑥𝑥𝑒𝑒𝑛𝑛𝑛𝑛𝑛𝑛 = �min(1, 𝑥𝑥𝑒𝑒 + 𝑚𝑚) , 𝑖𝑖𝑖𝑖 𝑥𝑥𝑒𝑒 𝐵𝐵𝑒𝑒𝜂𝜂 ≥ min(1, 𝑥𝑥𝑒𝑒 − 𝑚𝑚) , (16)
𝜂𝜂
𝑥𝑥𝑒𝑒 𝐵𝐵𝑒𝑒 , 𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒
where 𝑚𝑚 is the positive move-limit and 𝜂𝜂 is the numerical damping coefficient. For minimum
compliance problems the choice for 𝑚𝑚 = 0.2 and 𝜂𝜂 = 0.5 is recommended by Bendsøe and Sigmund.
In the current printing process it is not possible yet to create overhanging layers. This printing constraint
is added to the optimisation within the OC process. In the algorithm it is possible to define a baseplate
and thus a printing direction. Whenever there is an element in the design space with a virtual filtered
density higher than the supporting element, the density of the supporting element becomes equal to the
density of the considered element. Because this adaption is done inside the iteration loop, a real but
constrained optimised design is found.

Figure 4: Optimised results without (left) and with (right) printing constraint
The final step of the optimisation is a comparison of all the virtual densities in the design space. The
differences in density between the last two iterations is determined. If the largest difference is relatively
small enough or the prescribed maximum amount of iterations is exceeded, the iterations stop and the
final and optimal design has been found.

6. Conclusions and recommended further research


In this paper some important concrete additive manufacturing projects are mentioned, as well as some
structural optimisation methods. The main part presents a topology optimisation algorithm that takes
into account the orthotropic properties of printed concrete and a manufacturing constraint of the current
3D concrete printer. The algorithms uses the modified SIMP approach to distribute the material in an
efficient way and makes use of eight-node cubic elements. Further research should focus on a finer
distribution of the design space in order to obtain more precise and smooth results.
The elementals stiffness matrix is derived assuming orthotropic material instead of isotropic material.
Further tests should determine if this assumption is correct and if the optimised results will actually
behave similar to the non-optimised test specimens.
The current optimisation is optimising the structure for stiffness only for a determined fraction of the
initial volume. Further research needs to focus on minimising the stress in the structure as well. A
structure does not only need to be stiff, it needs to perform for strength as well.

7
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

The printing constraints are still limited and should have more attention in future research. Now only
the baseplate is a parameter in this constraint. In future results it should be possible to assign a certain
overhang angle in different directions, because future printers will be able to print cantilevering layers
as well.

The developed topology optimisation algorithm is the first in which both the orthotropic nature of
printed concrete as well as certain practical limitations of the concrete additive manufacturing process
can be taken into account, providing a first step towards structures which are not only optimised, but
structurally sound and printable as well.

Figure 5: Optimised results in 3D for a floor with distrubuted load and corner supports without (left) and with (right) printing
constraint (baseplate is the top surface, so the floor needs to be rotated after printing)

Acknowledgements
The authors would like to acknowledge the 4TU.Bouw Lighthouse project organisation, Eindhoven
University of Technologies, White Lioness technologies and Delft University of Technology for their
generous support of this research project.

References
[1] Bendsøe MP, Sigmund O. Topology optimization: theory, methods, and applications. Springer
Science & Business Media; 2013 Apr 17.
[2] Burden RL, Faires JD, Reynolds AC. Numerical analysis (1985). Prindle, Weber & Schmidt. 1985.
[3] Buswell RA, Soar RC, Gibb AG and Thorpe A. Freeform construction: mega-scale rapid
manufacturing for construction. Automation in construction, 2007; 16(2), 224-231.
[4] Bruns TE, Tortorelli DA. Topology optimization of non-linear elastic structures and compliant
mechanisms. Computer Methods in Applied Mechanics and Engineering. 2001 Mar
16;190(26):3443-59.
[5] Doomen C,. The effect of layered manufacturing on the strength properties of printable concrete.
Master’s thesis, Eindhoven University of Technology, 2016
[6] Frazier WE. Metal additive manufacturing: a review. Journal of Materials Engineering and
Performance. 2014 Jun 1;23(6):1917-28.
[7] Gibson I, Rosen D, Stucker B. Additive manufacturing technologies: 3D printing, rapid
prototyping, and direct digital manufacturing. Springer; 2014 Nov 26.
[8] Khoshnevis B, Automated construction by contour crafting—related robotics and information
technologies. Automation in construction, 2004;13(1), 5-19.
[9] Lim S, Buswell RA, Le TT, Austin SA, Gibb AG, Thorpe T. Developments in construction-scale
additive manufacturing processes. Automation in construction. 2012 Jan 31;21:262-8.
[10] Lu B, Li D, Tian X. Development Trends in Additive Manufacturing and 3D Printing. Engineering.
2015 Mar 31;1(1):085-9.

8
Proceedings of the IASS Annual Symposium 2017
Interfaces: architecture.engineering.science

[11] Melchels FP, Domingos MA, Klein TJ, Malda J, Bartolo PJ and Hutmacher DW. Additive
manufacturing of tissues and organs. Progress in Polymer Science. 2012;37(8):1079-1104.
[12] Pegna J. Exploratory investigation of solid freeform construction. Automation in construction.
1997 Feb 1;5(5):427-37.

View publication stats

You might also like