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

net/publication/357865935

Optimizing support locations in structural design

Conference Paper · August 2021

CITATIONS READS
3 282

3 authors:

Ting-Uei Lee Xianchuan Meng

RMIT University Nanjing University
32 PUBLICATIONS 184 CITATIONS 5 PUBLICATIONS 17 CITATIONS

SEE PROFILE SEE PROFILE

Yi Min Xie
RMIT University
565 PUBLICATIONS 23,895 CITATIONS

SEE PROFILE

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

Topology Optimization of Binary Structures Using Integer Linear Programming View project

Topology optimization View project

All content following this page was uploaded by Ting-Uei Lee on 16 January 2022.

The user has requested enhancement of the downloaded file.

 Proceedings of the IASS Annual Symposium 2020/21 and
the 7th International Conference on Spatial Structures
Inspiring the Next Generation
23 – 27 August 2021, Guilford, UK
S.A. Behnejad, G.A.R. Parke and O.A. Samavati (eds.)

Optimizing support locations in structural design

Ting-Uei LEEa, Xianchuan MENGa,b, Yi Min XIEa,*
a,*
Centre for Innovative Structures and Materials, School of Engineering,
RMIT University, Melbourne 3001, Australia
[email protected]
b
School of Architecture and Urban Planning, Nanjing University, Nanjing 210093, China

Abstract
Support locations of a structure in traditional structural design procedures are typically prescribed. A
structure with improved structural performance may be obtained by manually adjusting its support
locations through a trial-and-error process. However, such an approach is tedious and time-consuming,
and the results can be far from the optimal solution. This study presents a new method for automatically
finding optimal support locations of a structure under given loading conditions. Computational
algorithms are developed based on a combination of an Optimality Criteria (OC) method and the Bi-
directional Evolutionary Structural Optimization (BESO) technique. Two classes of problems have been
investigated: (I) optimizing the distribution of a given number of vertical supporting columns under a
roof structure; (II) simultaneously designing the topology of a structure and its support locations. A
series of 2D and 3D examples are presented to demonstrate the effectiveness of the new algorithms. It
is shown that treating support locations as additional design variables provides new opportunities to
significantly improve the structural performance and therefore reduce the environmental impact of
bridges, buildings and other artefacts.
Keywords: support location, structural design, topology optimization, BESO

1. Introduction
Topology optimization is an effective strategy to obtain efficient and innovative structural designs by
determining shapes and locations of cavities in continuous geometries. Popular topology optimization
methods include the homogenization approach [1], the Solid Isotropic Material with Penalization
(SIMP) method [2], the Bi-directional Evolutionary Structural Optimization (BESO) method [3,4], and
the level set method [5]. Among many topology optimization methods, the SIMP method and the BESO
method have been increasingly studied due to their unique key advantages; the SIMP method utilizes
an Optimality Criteria (OC) to vary all candidate elements throughout optimization; the BESO method
generates clear 0/1 structural topologies by simultaneously removing and adding inefficient and efficient
elements, respectively. These topology optimization methods are typically performed on a design
domain with pre-determined supports; optimization of supports may be implemented as a separate task
to obtain improved structural performance.
Designing support locations in a structural design can be a challenging task, as support locations have a
direct influence not only on the architectural appearance bust also the structural performance. Diverse
optimization methods have been developed to capture complex requirements from both engineering and
architectural perspectives, including continuum and discrete methods. Continuum methods refer to
topology optimization methods, where complex support structures can be generated from a user-defined
geometry [6]. Discrete methods obtain optimal designs by determining optimal spatial order and
connectivity of line-based representations, which may include the utilization of the Genetic Algorithm
(GA) [7], the Simulated Annealing (SA) algorithm [8], or the Particle Swarm Optimization (PSO) [9].

Copyright © 2021 by the author(s) as listed above. Published in the Proceedings of the IASS Annual Symposium
2020/21 and the 7th International Conference on Spatial Structures, with permission.
Proceedings of the IASS Annual Symposium 2020/21 and the 7th International Conference on Spatial Structures
Inspiring the Next Generation

However, existing discrete methods are limited, which can be computationally expensive if considering
a large number of candidate elements and the objective function for the optimization problem can be
difficult to formulate.
Simultaneous optimization of supports and topology offers new design possibilities to achieve improved
structural performance [10]. However, existing simultaneous optimization techniques require
cumbersome procedures to allow continuous variations of supports. These include utilizing background
filed methods to activate or deactivate support elements through alternations of stiffness matrices [11],
complex packing systems to relocate pre-determined support structures [12], and level-set methods to
obtain curved-based support boundaries [13].
This paper develops a new support optimization method by introducing support locations as additional
design variables into the structural optimization formulation. Two types of support locations are tested,
including line-based and element-based supports. By extension, the new method is combined with the
existing BESO topology optimization method to enable simulations optimization. Section 2 describes
the optimization algorithms. Section 3 presents the optimization results of line-based supports. Section
4 demonstrates the optimization results of element-based supports without and with topology
optimization.

2. Optimization methodology

2.1 Problem definition

Two classes of optimization problems are investigated in this paper. Class I optimization introduces
line-based candidate supports under a roof structure, as shown in Figure 1 (a), to design the distribution
of a given number of vertical supporting columns. These lines represent a predefined column type and
allowable locations [14]. Class II optimization introduces element-based candidate supports into a
design domain, as shown in Figure 1 (b), to design the topology of a structure and its support locations.
These elements not only represent support locations but also consider the size and stiffness of supports.
Both classes of supports are employed as additional design variables into the structural optimization
formulation with the consideration of design requirements, i.e. type, allowable locations, size, and
stiffness.

Figure 1: Introducing support locations as additional design variables into the structural optimization
formulation (a) Class I – Optimization of line-based supports. (b) Class II – Optimization of element-based
support.

2.2 Problem statement

The optimization problem can be understood as compliance minimization of the structural system with
volume constraints. The mathematical problem statement can be written as follows.
Minimize: 𝐶 = 𝐟 𝐮 (1)

2
Proceedings of the IASS Annual Symposium 2020/21 and the 7th International Conference on Spatial Structures
Inspiring the Next Generation

𝑆𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜: 𝑁 ∗ = 𝑦 (2)

0< 𝑦 ≤ 𝑦 ≤1 (3)

𝑉∗ = 𝑉𝑥 (4)

𝑥 = 𝑥 or 1 (5)
𝐊𝐮 = 𝐟 (6)
Equation (1) is the objective function, where 𝐶 is the compliance (which is the inverse of the overall
stiffness), 𝐟 is the global force vector, and 𝐮 is the global displacement vector. Equations (2)-(3) are
constraints for support optimization, where 𝑁 ∗ is the prescribed number of supports, and 𝑦 is the 𝑖-th
design variable that denotes the relative density of support, declared using 𝑦 and 1 to represent its
absence and presence condition, respectively. Equations (4)-(5) are constraints for topology
optimization, where 𝑉 ∗ is the prescribed structure volume, 𝑉 is the individual volume of 𝑖-th structural
element, and 𝑥 denotes the relative density of structural elements declared using 𝑥 and 1 to
represent its absence and presence condition, respectively. Equation (6) is an additional constraint that
ensures the structural equilibrium, where 𝐊 is the global stiffness matrix. Note that equations (4)-(5) are
not triggered in Class I optimization, as it does not include topology optimization.

2.3 Support optimization

This paper utilizes the OC method to update all 𝑦 throughout optimization. Note that 𝑦 can vary
continuously between its lower and upper bound; optimal results include removed and preserved
supports with 𝑦 to be approximately 0 and 1, respectively. The updating scheme is written as
𝑚𝑎𝑥 𝑦𝑚𝑖𝑛 , 𝑦𝑘𝑖 (1 − 𝑚) 𝜂
if 𝑦𝑘𝑖 𝐵𝑖 ≤ 𝑚𝑎𝑥 𝑦𝑚𝑖𝑛 , 𝑦𝑘𝑖 (1 − 𝑚) , (7)
𝑦 = 𝑚𝑖𝑛 1, 𝑦𝑘𝑖 (1 + 𝑚) 𝜂
if 𝑦𝑘𝑖 𝐵𝑖 ≥ 𝑚𝑖𝑛 1, 𝑦𝑘𝑖 (1 + 𝑚) ,
𝜂
𝑦𝑘𝑖 𝐵𝑖 otherwise.
where 𝑦 denotes the 𝑖-th design variable at the 𝑘-th iteration, 𝑚 = 0.2 is the move limit, 𝜂 = 0.5 is a
numerical damping coefficient, and 𝐵 determines the optimality condition.
The initial design variable of supports, 𝑦 , is defined as the ratio of 𝑁 ∗ to the total number of candidate
supports, 𝑁, where
𝑦 = 𝑁 ∗ /𝑁 (8)
The material model of supports is defined as
𝐸(𝑦 ) = 𝑦 𝐸 (9)
where 𝐸(𝑦 ) is the Young’s modulus of 𝑖-th support at the 𝑘-th iteration, 𝐸 is the design Young’s
modulus of supports, and 𝑝 = 3 is the penalty exponent.

2.3.1 Line-based supports

In Class I optimization, the columns are hinged, so they are only loaded in the axial direction. Design
variables are updated according to axial stress [14]; the entire 𝐵 term is defined as
𝐵 = 𝜆𝜎 (10)

where 𝜎 is the axial stress of the 𝑖-th column at the 𝑘-th iteration and 𝜆 is a Lagrange multiplier, which
can be solved using a bi-section method.

3
Proceedings of the IASS Annual Symposium 2020/21 and the 7th International Conference on Spatial Structures
Inspiring the Next Generation

2.3.2 Element-based supports

Design variables in Class II optimization are updated according to element elastic strain energy density;
the entire 𝐵 term is written as
(11)
𝐵 = 𝜆 𝑈

where 𝑈 is the elastic strain energy density of the 𝑖-th support element at the 𝑘-th iteration.

2.4 Topology optimization

This paper uses the soft-kill BESO method to update design variables of structural elements, 𝑥 , and
gradually reduces the structural volume in each iteration. The soft-kill BESO method was developed
previously and fully described in [3]. Structural elements are added or removed according to the
sensitivity threshold procedure, defined as
𝑥 if 𝛼 ≤ 𝛼 , (12)
𝑥 = 1 if 𝛼 > 𝛼 ,
𝑥 otherwise.
where 𝑥 denotes the 𝑖th design variable at the 𝑘th iteration, 𝛼 is the threshold sensitivity based on
the current structure volume and the relative ranking of sensitivities, and 𝛼 is the sensitivity number
obtained through a sensitivity analysis.
The material model of structural elements is defined as
𝐸(𝑥 ) = 𝑥 𝐸 (13)
where 𝐸(𝑥 ) is the Young’s modulus of 𝑖-th structural element at the 𝑘-th iteration, and 𝐸 is the
design Young’s modulus of structural elements.

3. Class I: Optimization of line-based supports

The new method was first tested using a bean-shaped roof, as shown in Figure 2. The optimizer was
used to find 𝑁 ∗ = 7 columns from 𝑁 = 1159 candidate columns in a 9m × 6.5m × 3m design domain.
Line-based supports were assigned with a square cross-section of 100mm × 100mm. The optimization
result is shown in Figure 2 (a). The final configuration represented a roof-column structural system
capable of providing adequate protection against environmental loads and the sunlight, as shown in
Figure 2 (b). It should be highlighted that column locations are fulfilling both engineering and
architectural requirements, where the structural performance was optimized and analysed using the
finite element method, and the number of columns and their allowable locations were pre-determined
by the designer. A set of manual arrangements of columns were tested, as shown in Figure 2 (c). It can
be seen that the optimization result had the smallest compliance, which confirms that the proposed
method can effectively design column locations and improve structural performance.

4. Class II: Optimization of element-based supports

4.1 Without changing topology

For the complex shell structure example, as shown in Figure 3 (a)-(b), the optimizer was used to find
𝑁 ∗ = 8 pin locations from 𝑁 = 100 candidate elements in a 5.8m × 6.9m × 3.2m design domain; shell
thickness was set as 10mm. In this example, the structure itself was treated as a non-designable domain,
meaning the topology optimization algorithm was not triggered. The support optimization result is
shown in Figure 3 (c), which had the smallest compliance compared with a set of manual results shown
in Figure 3 (d). The optimization result has confirmed the validity of the new method for automatically

4
Proceedings of the IASS Annual Symposium 2020/21 and the 7th International Conference on Spatial Structures
Inspiring the Next Generation

finding efficient support locations from many potential support combinations and avoiding extremely
large compliance. The following sub-section further investigates the support optimization method by
triggering the BESO topology optimization algorithm.

Figure 2: Designing column locations under a bean-shaped roof: (a) Optimization process; (b) Optimization
result rendering; (c) Manual results

Figure 3: Designing support locations of a complex shell structure: (a) Design settings; (b) Initial finite
element analysis result; (c) Optimization result; (d) Manual results

5
Proceedings of the IASS Annual Symposium 2020/21 and the 7th International Conference on Spatial Structures
Inspiring the Next Generation

4.2 With topology optimization

4.2.1 2D bridge
Three 2D bridges were designed using the proposed simultaneous optimization method to obtain its
topology and supports, as shown in Figure 4. Details of the loading and boundary conditions of this
example were fully described in [10]. It is seen that the number of supports substantially affects the
formation of the structural topology, and optimizing the support locations reduces the compliance of the
structure. Overall, this example shows that treating support elements as additional design variables is an
effective strategy to design support locations and topology simultaneously.

Figure 4: Optimizing the structural topology and support locations of a 2D bridge example. The algorithm is
optimizing for (a) 2, (b) 4, and (c) 34 support elements, in addition to fixed supports at two ends of the deck.

4.2.2 3D cantilever
Figure 5 shows the optimization results of a 3D cantilever and its support elements. Details of the
loading and boundary conditions of this example can be found in [3]. It is seen that weak support
elements with 𝑦 = 0.5 were formed using 𝑁 ∗ = 6. The formation of weak elements was mainly
attributed to the cantilever’s symmetric characteristic. The design domain had a width made up of eight
elements, and three solid support elements were needed for each supporting region. As a result, four
support elements were formed on each supporting region, including weak and solid elements; support
locations strongly depends on the mesh layout. Nevertheless, this example still demonstrates that the
proposed simultaneous optimization method can be effectively adopted for 3D applications to obtain
efficient and innovate structural designs.

Figure 5: Optimizing the structural topology and support locations of a 3D cantilever example. The algorithm
is optimizing for (a) 4, (b) 6, (c) 8, and (d) 40 support elements

6
Proceedings of the IASS Annual Symposium 2020/21 and the 7th International Conference on Spatial Structures
Inspiring the Next Generation

5. Conclusions
This paper has demonstrated an effective optimization method to design support locations in a structural
design by considering support locations as additional design variables in the structural optimization
formulation. The new method was developed based on an OC algorithm, which allows continuous
variations of all supports throughout the optimization. This paper has shown that the utilization of line-
based and element-based supports can both produce good results and meet the respective optimization
criteria. By extension, it is shown that combining the support optimization algorithm with the BESO
method can effectively obtain efficient and innovative structural designs. Key findings of this paper
include: the number of supports substantially affects the formation of the structural topology, and
optimizing the support locations significantly increases the overall stiffness of the structure.

Acknowledgements
The authors gratefully acknowledge the financial support provided by the Australian Research Council
(FL190100014) and the China Scholarship Council (201906195003).

References
[1] Bendsøe M.P. and Kikuchi N., Generating optimal topologies in structural design using a
homogenization method. Computer Methods in Applied Mechanics and Engineering, 1988; 71;
197–224.
[2] Bendsøe M.P. and Sigmund O., Material interpolation schemes in topology optimization. Archive
of Applied Mechanics, 1999; 69; 635–654.
[3] Huang X. and Xie Y.M., Evolutionary Topology Optimization of Continuum Structures: Methods
and Applications, John Wiley & Sons, 2010.
[4] Xia L., Xia Q., Huang X. and Xie Y.M., Bi-directional evolutionary structural optimization on
advanced structures and materials: a comprehensive review. Archives of Computational Methods
in Engineering, 2018; 25; 437-478.
[5] Luo Z., Tong L. and Kang Z., A level set method for structural shape and topology optimization
using radial basis functions. Computers and Structures, 2009; 87; 425–434.
[6] Kuo Y.H., Cheng C.C., Lin Y.S. and San C.H., Support structure design in additive manufacturing
based on topology optimization. Structural and Multidisciplinary Optimization, 2018; 57; 183–
195.
[7] Ohsaki M., Genetic algorithm for topology optimization of trusses. Computers and Structures,
1995; 57; 219–225.
[8] Lamberti L., An efficient simulated annealing algorithm for design optimization of truss structures.
Computers and Structures, 2008; 86; 1936–1953.
[9] Ray T. and Liew K.M., A swarm metaphor for multiobjective design optimization. Engineering
Optimization, 2002; 34; 141–153.
[10] Buhl T., Simultaneous topology optimization of structure and supports. Structural and
Multidisciplinary Optimization, 2002; 23; 336–346.
[11] Stanford B., Kapania R. and Beran P., Conceptual design of compliant mechanisms for flapping
wings with topology optimization. AIAA Journal, 2011; 49; 855–867.
[12] Zhu J. H. and Zhang W.H., Integrated layout design of supports and structures. Computer Methods
in Applied Mechanics and Engineering, 2010; 199; 557–569.
[13] Xia Q., Wang M.Y., and Shi T., A level set method for shape and topology optimization of both
structure and support of continuum structures. Computer Methods in Applied Mechanics and
Engineering, 2014; 272; 340–353.

7
View publication stats

Proceedings of the IASS Annual Symposium 2020/21 and the 7th International Conference on Spatial Structures
Inspiring the Next Generation

[14] Meng X., Lee, T.U., Xiong Y., Huang X. and Xie Y.M., Optimizing support locations in the roof–
column structural system. Applied Sciences, 2021; 11; 2775.