You are on page 1of 16

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

net/publication/226183049

Geometry and topology optimization of geodesic domes using charged system


search

Article  in  Structural and Multidisciplinary Optimization · February 2011


DOI: 10.1007/s00158-010-0566-y

CITATIONS READS

68 3,613

2 authors:

A. Kaveh Siamak Talatahari


Iran University of Science and Technology University of Tabriz
925 PUBLICATIONS   20,958 CITATIONS    151 PUBLICATIONS   8,762 CITATIONS   

SEE PROFILE SEE PROFILE

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

Emergency Management Systems View project

combinatorial force ethod View project

All content following this page was uploaded by A. Kaveh on 28 June 2015.

The user has requested enhancement of the downloaded file.


Struct Multidisc Optim
DOI 10.1007/s00158-010-0566-y

RESEARCH PAPER

Geometry and topology optimization of geodesic domes


using charged system search
Ali Kaveh · Siamak Talatahari

Received: 11 December 2009 / Revised: 12 June 2010 / Accepted: 11 August 2010



c Springer-Verlag 2010

Abstract Dome structures provide cost-effective solutions early times, sports stadiums, assembly halls, exhibition cen-
for covering large areas without intermediate supports. In ters, swimming pools, shopping centers and industrial build-
this article, simple procedures are developed to reach the ings have been typical examples of structures with large
configuration of the geodesic domes. A new definition of unobstructed areas nowadays (Makowski 1993). Dome
dome optimization problems is given which consists of structures are the most preferred type of large spanned
finding optimal sections for elements (size optimization), structures. Domes have been of a special interest in the
optimal height for the crown (geometry optimization) and sense that they enclose large areas without intermediate
the optimum number of elements (topology optimization) supports.
under determined loading conditions. In order to find the Although dome structures are economical in terms of
optimum design, the recently developed meta-heuristic algo- consumption of constructional materials compared to the
rithm, known as the Charged System Search (CSS), is conventional forms of structures (Makowski 1984), a more
applied to the optimum design of geodesic domes. The lightweight design can be obtained using optimization
CSS takes into account the nonlinear response of the domes. methods. Optimization of an engineering design is an
Using CSS, the optimum design of the geodesic domes is improvement of a proposed design that results in the best
efficiently performed. properties for minimum cost. Structural optimization can be
categorized as:
Keywords Geodesic domes · Optimization ·
Nonlinear design · Charged system search • Sizing optimization in which the geometry and
topology of the structure remain unchanged but
cross sectional properties are optimized (e.g. Rajeev
1 Introduction and Krishnamoorthy 1992; Camp et al. 1998; Camp
and Bichon 2004; Serra and Venini 2006; Perez and
Structural systems, which enable the designers to cover Behdinan 2007; Lamberti 2008; Kaveh and Shojaee
large spans, have always been a challenging task for struc- 2007; Kaveh et al. 2008a; Kaveh and Talatahari 2009a,
tural engineers. Beginning with the worship places in the b, c, 2010a);
• Geometry optimization which determines the optimum
location of the joints in the structure in addition to
the size of members (e.g. Prager and Rozvany 1977;
A. Kaveh (B)
Hasançebi and Erbatur 2002; Saka 2007; Kaveh and
Centre of Excellence for Fundamental Studies in Structural
Engineering, Iran University of Science and Technology, Talatahari 2010b); and
Narmak, Tehran 16, Iran • Topology optimization that involves finding the num-
e-mail: alikaveh@iust.ac.ir ber of members of the structure and the way in
S. Talatahari
which these members are connected to each other (e.g.
Department of Civil Engineering, Dhingra and Bennage 1995; Rozvany 1998, 2001,
University of Tabriz, Tabriz, Iran 2009; Fourie and Groenwold 2002; Bendsoe and
A. Kaveh, S. Talatahari

Sigmund 2003; Wang et al. 2004; Svanberg and Werme phenomena include the biological evolutionary process,
2005, 2006; Martínez et al. 2007; Kaveh et al. 2008b; animal behavior, or the physical processes.
Rahami et al. 2008; Huang and Xie 2010). Recently, the authors proposed a new optimization
approach, so-called Charged System Search (Kaveh and
The optimal design of topology and geometry has been Talatahari 2010c, d), which utilizes a number of Charged
investigated far less due to its complexity, despite the fact Particle (CP). These particles affect each other based on
that the optimization of topology and geometry greatly their fitness values and separation distances considering
improves the design (Bendsoe and Mota Soares 1992; the governing laws of Coulomb and Gauss from electri-
Kirsch 1989, 1997; Rozvany et al. 1995; Topping 1983). cal physics and the governing laws of motion from the
In general, optimization techniques used in structural Newtonian mechanics. This paper develops the CSS method
problems can be categorized into classical and heuristic to perform an optimum geometrical and topological design
search methods. The disadvantages of traditional opti- of geodesic domes. A simple procedure is developed to cal-
mization methods such as complex derivatives and the culate the joint coordinates and element constructions in
large amount of required memory have forced researchers order to determine the configuration of the domes. Here,
to employ heuristic approaches for solving optimization the joint coordinate equations are formulated for auto-
problems (Lee and Geem 2004). Many of the heuristic matic formation of the configuration. The nodal coordinates
approaches are inspired by the natural phenomena. These are calculated using a simple relationship. The joints are

Fig. 1 Geodesic domes with


a three rings, b four rings, (a) Nr =3 (b) Nr =4
c five rings

3D view 3D view

Top view Top view

Side view Side view


Geometry and topology optimization of geodesic domes using charged system search

divided into two types considering the symmetry, and then the number of elements automatically during the optimiza-
the elements are easily determined. The serviceability and tion process, this number being considered as a design
strength requirements are considered in the design prob- variable. A discussion over the effect of the number of rings
lems as specified in LRFD-AISC (1991). The algorithm and the number of elements on the results is presented in
takes into account the nonlinear response of the dome Section 5.1.
due to effect of axial forces on the flexural stiffness of The remaining sections of this paper are organized as fol-
members. lows: Section 2 contains the statement of the dome design
In this paper, unlike the previous studies on the dome problem. Section 3 reviews the CSS, briefly. Section 4
optimization (Saka 2007; Kaveh and Talatahari 2010b) describes the implementation of CSS methodology to opti-
which consider only size or geometry optimization of the mize geodesic dome structures. The achieved results are
domes by determining the number of rings and the cross- explained in Section 5. Finally, Section 6 presents the
sectional areas of the elements, a method is developed that conclusion of the paper.
has the ability to determine the element configurations and

2 Geometry and topology optimum design


of dome structures
(c) Nr =5
Geometrical optimal design of dome structures consists of
finding optimal sections for elements, optimal height for
the crown and the optimum number of rings under deter-
mined loading conditions. The allowable cross sections are
considered as 37 steel pipe sections taken from LRFD-
AISC (1991) which is also utilized as the code of practice.
The mathematical formulation for optimum design of dome
structures can be expressed as

Find X = [x1 , x2 , ..., xng ], h, Nr


3D view
xi ∈ {d1 , d2 , ..., d37 }
 
h i ∈ h min , h min + h ∗ , ..., h max

nm
to minimize W (X) = ρ · xi · L i (1)
i=1

where X is a vector containing the design variables of the


elements; h is the variable of the crown height; Nr is the
total number of rings which is taken as 3, 4 or 5 as shown in
Fig. 1; d j is the jth allowable discrete value for the design
variables (37 pipe sections taken from LRFD-AISC); h min ,
h max and h ∗ are the permitted minimum, maximum and
increased amounts of the crown height, which in this paper
are taken as 1.00, 9.00 and 0.25 m, respectively; ng is the
number of design variables or the number of size groups;
W (X) is the weight of the structure; L i is the length of mem-
Top view
ber i; ρ is the is the material mass density and nm is the total
number of elements.
Similarly, topological optimization of geodesic dome
structures can be considered as finding optimal sections for
elements, optimal height for the crown, optimum number of
rings and optimum number of the elements. Since the num-
Side view
ber of elements available in each ring is equal to the number
Fig. 1 (continued) of nodes related to that ring, therefore once the number of
A. Kaveh, S. Talatahari

ring nodes as the design variables is determined, the topo- The design constraints are as follows:
logical optimum design will be achieved. In the other hand,
Displacement constraint:
for the geodesic domes the number of nodes in each ring
is considered as the number of joints in the first ring (Nn)
|δi | ≤ δimax i = 1, 2, ..., nn (3)
multiplied by the number associated with that ring. Figure 2
shows the different forms of geodesic domes when Nn is
altered. As a result, if we consider Nn as the new design
Interaction formula constraints:
variable, the complete dome topology optimization problem  
Pu Mux Muy Pu
will be defined as + + ≤ 1 For < 0.2
2φc Pn φb Mnx φb Mny φc Pn
 
Pu 8 Mux Muy Pu
Find X = [x1 , x2 , ..., xng ], h, Nr, N n + + ≤ 1 For ≥ 0.2
φc Pn 9 φb Mnx φb Mny φc Pn
xi ∈ {d1 , d2 , ..., d37 } (4)
 
h i ∈ h min , h min + h ∗ , ..., h max Shear constraint:

nm
to minimize W (X) = ρ · xi · L i (2) Vu ≤ φv Vn (5)
i=1

Fig. 2 Geodesic domes with


different number of elements
(a) Nn =3 (b) Nn =4

3D view 3D view

Top view Top view

Side view Side view


Geometry and topology optimization of geodesic domes using charged system search

Fig. 2 (continued)
(c) Nn =5 (d) Nn =6

3D view 3D view

Top view Top view

Side view Side view

in which δi is the displacement of node i; δimax is the permit- governing laws of motion from the Newtonian mechanics.
ted displacement for the ith node; nn is the total number of This algorithm can be considered as a multi-agent approach,
nodes; Pu is the required strength; Pn is the nominal axial where each agent is a Charged Particle (CP). Each CP is
strength; φc is the resistance factor (φc = 0.9 for tension, considered as a charged sphere with radius a, having a
φc = 0.85 for compression); Mux and Muy are the required uniform volume charge density and is equal to
flexural strengths in the x and y directions, respectively;
Mnx and Mny are the nominal flexural strengths in the x and
y directions; φb is the flexural resistance reduction factor fit(i) − fitworst
qi = i = 1, 2, ..., N (6)
(φb = 0.90); Vu is the factored service load for shear; fitbest − fitworst
Vn is the nominal strength in shear and φv represents the
resistance factor for shear (φv = 0.90). where fitbest and fitworst are the best and the worst fitness
of all the particles; fit(i) represents the fitness of the agent i,
and N is the total number of CPs.
3 Charged System Search algorithm CPs can impose electric forces on the others. The kind
of the forces is attractive, and its magnitude for the CP
The Charged System Search (CSS) algorithm is based on located in the inside of the sphere is proportional to the
the Coulomb and Gauss laws from electrical physics and the separation distance between the CPs, and for a CP located
A. Kaveh, S. Talatahari

Fig. 2 (continued)
(e) Nn =7 (f) Nn =8

3D view 3D view

Top view Top view

Side view Side view

outside the sphere is inversely proportional to the square of Pi j determines the probability of moving each CP toward
the separation distance between the particles the others as
  ⎧
 qi qi ⎪ fit(i)− f itbest
Fj = qj ri j · i 1 + 2 · i 2 pi j (Xi − X j ), ⎨1 > rand ∨ fit( j) > fit(i)
a3 ri j pi j = fit( j)−fit(i) (9)
i,i= j ⎪

0 otherwise
j = 1, 2, ..., N
i 1 = 1, i 2 = 0 ⇔ ri j < a (7) The resultant forces and the motion laws determine the new
i 1 = 0, i 2 = 1 ⇔ ri j ≥ a location of the CPs. At this stage, each CP moves towards
its new position under the action of the resultant forces and
where F j is the resultant force acting on the jth CP; ri j is
its previous velocity as
the separation distance between two charged particles which
is defined as follows Fj
X j,new = rand j1 · ka · · t 2
||Xi − X j || mj
ri j = (8)
||(Xi + X j )/2 − Xbest || + ε + rand j2 · kv · V j,old · t + X j,old (10)
X j,new − X j,old
where Xi and X j are the positions of the ith and jth CPs, V j,new = (11)
t
respectively; Xbest is the position of the best current CP,
and ε is a small positive number. The initial positions of where ka is the acceleration coefficient; kv is the velocity
CPs are determined randomly in the search space and the coefficient to control the influence of the previous velocity;
initial velocities of charged particles are assumed to be zero. and rand j1 and rand j2 are two random numbers uniformly
Geometry and topology optimization of geodesic domes using charged system search

(g) Nn =9 and 1, sets the rate of choosing a value in the new vector
from the historic values stored in the CM, and (1-CMCR)
sets the rate of randomly choosing one value from the
possible range of values. The pitch adjusting process is per-
formed only after a value is chosen from CM. The value
(1-PAR), sets the rate of doing nothing. Here, ‘‘w.p.’’
means ‘‘with probability’’. For more details, the reader may
refer to Kaveh and Talatahari (2009a, c).
In order to have discrete results, a rounding function
is utilized which changes the magnitude of a result to the
3D view nearest discrete value, as follows


Fj
X j,new = Round rand j1 · ka · · t 2 + rand j2 · kv .
mj

· V j,old · t + X j,old (13)

The constraint handling approach is the penalty approach


and the framework of the CSS algorithm is illustrated in
Fig. 3.

Top view
4 The CSS methodology for deign of optimum
dome structures

The behaviour of the dome structures is nonlinear due to


the change of geometry under external loads. The imperfec-
Side view
tions arising either from the manufacturing process and/or
from the construction of the structure can also be the source
Fig. 2 (continued) of nonlinearity. Inclusion of geometric nonlinearity requires
additional considerations in the analysis. Stability check
distributed in the range of (0,1). To save the best design a is also necessary during the analysis to ensure that the
memory (Charged Memory or CM) is considered. If each structure does not lose its load carrying capacity due to
CP exits from the allowable search space, its position is cor- instability (Saka 2007) and furthermore, considering the
rected using the harmony search-based handling. According nonlinear behaviour in the design of domes results in lighter
to this mechanism, any component of the solution vector structural systems. The elastic instability analysis of domes
violating the variable boundaries can be regenerated from involves repeated analysis of the structure at progressively
the CM as increasing load factor. At each increment of the load factor,
⎧ nonlinear analysis of the structure is carried out. For this,

⎪w.p. CMCR the stiffness matrix for a three-dimensional space member



⎪ ==> select a new value for a variable from CM that includes the effect of flexure on axial stiffness and the

⎨ ==> w.p. (1 − PAR) do nothing stiffness against translation is derived. Details of this deriva-
xi, j = ==> w.p. PAR choose a neighboring

⎪ tion and related terms of the nonlinear stiffness matrix of

⎪ value

⎪ (1 − a space member are given in Majid (1972) and Ekhande

⎩w.p. CMCR)
==> select a new value randomly et al. (1989). The stiffness matrix of a stable structure is
(12) positive-definite. During the nonlinear analysis iteration,
the determinant of the overall stiffness matrix is checked to
where xi, j is the ith component of the CP j. The Charged determine whether at any load increment it becomes neg-
Memory Considering Rate (CMCR) varying between 0 ative. This is an indication of a loss of stability in the
A. Kaveh, S. Talatahari

Fig. 3 The flowchart of the


CSS algorithm

structure and the load factor which causes this case, is iden- the first ring (Nn) can be equal to 3 to 10. The distances
tified as the critical load factor. As a result, the geometric between the rings in the dome on the meridian line are gen-
nonlinearity is also included in this study as described in erally of equal length. The structural data for the topology
Kaveh and Talatahari (2010b). of this form of the geodesic dome is a function of the diam-
Figure 4 shows the nodes of the geodesic dome, when eter of the dome (D), the total number of rings (Nr), the
N n = 10. As it can be seen, the number of nodes in each number of nodes in the first ring (Nn) and the height of the
ring is considered as Nn multiplied by the number associ- crown (h). The top joint at the crown is numbered as the first
ated with that ring. In this paper the total number of rings joint as shown in Fig. 4a (joint number 1) which is located
(Nr) is selected from 3 to 5, and the number of nodes in in the centre of the coordinate system in the x − y plane.
Geometry and topology optimization of geodesic domes using charged system search

(a) where n i is the number of ring corresponding to the node i;


R = (D 2 + 4h 2 )/(8h) is the radius of the hemisphere as
shown in Fig. 4b.
The member grouping is determined in a way that mem-
bers between each consecutive pair of rings belong to one
group, and the members on each ring belong to a different
group. The diagonal members between the crown and the
first ring are group 1, the members between ring 1 and 2 are
group 2 and the group number of members on the ring 2 is
3, and so forth. Also, the members on each ring constitute
one group.
The configuration of elements consists of determining
the start and end nodes of each element. For the first group,
the start node for all elements is the joint number 1 and the
end nodes are those on the first ring. The start and end nodes
of elements on each ring are two consecutive nodes on that
ring. As an example, the element (4,5) belongs to the first
ring’ group and the elements (16,17) and (17,18) belong to
the second ring’ group, as illustrated in Fig. 5. The element
with the lower and upper numbers on that ring also corre-
(b) sponds to that group. For the other diagonal elements, we
have

⎪ n i (n i − 1)

⎪ I = Nn × +2

⎪ 2 

⎪ j −1


⎨ + n i × Fi x
⎪ 
3 j = 2, 3, ..., 3 × N n

⎪ n (n + 1) n i = 1, 2, ..., Nr − 1

⎪ J = Nn ×
i i
+j

⎪ 2  

⎪   j −1


⎩ + n i − 2 × Fi x
3
(15)

Fig. 4 Nodal numbering and the related coordinate system. a Top


view of the geodesic dome. b Section of the geodesic domes

The coordinates of other joints in each ring are obtained as


follows:
⎧   
⎪ D.n i 360 N n × n i (n i − 1)

⎪ x i = cos i − − 2

⎪ 2Nr N n · ni 2

⎪   


⎨ D.n i 360 N n × n i (n i − 1)
yi = sin i− −2
⎪ 2Nr N n · ni 2

⎪ 

⎪  

⎪  n i2 D 2

⎪ 
⎩z i = R −2 − (R − h)
4Nr 2
(14) Fig. 5 The dome nodes and their related elements
A. Kaveh, S. Talatahari

⎧    

⎪ ni ni − 1 j −1

⎪ I = Nn × + 3 + n i × Fi x  

⎪ 2 2 ni − 1



⎪       

⎪ j − 2 n i − 1 × Fi x ( j − 1) 2(n i − 1) − 1

⎪ 
⎨ + Fi x
2 j = 1, 2, ..., 2 × N n × (n i − 1)
  (16)

⎪ n i (n i + 1) j −1 n i = 2, 3, ..., Nr − 1


⎪ J = Nn ×


+ 2 × Fi x
2(n i − 1)

⎪  2

⎪ j

⎪ + Fi x +3

⎩ 2

⎧  
⎪ n n −1 A population of 25 individuals is used for the CPs and the
⎨ I = Nn × i i +2
 2   n i = 1, 2, ..., Nr − 1 value of constant a is set to one. The acceleration coefficient

⎩ J = Nn × in + 1 n i +2 ka and the velocity coefficient kv are taken as 0.5. Here,
+1
2 CMCR = 0.95 and PAR = 0.10 are used as suggested by
(17) Kaveh and Talatahari (2010c).
where I and J are the start and end nodal numbers of
the elements, respectively. We divide the nodes into two 5.1 Effects of Nr and Nn on the optimum designs
types: primary and secondary nodes. The primary nodes
are located on the symmetry lines and the number of these Though the number of rings (Nr) and the number of nodes in
nodes is equal to Nn for each ring. Other nodes are consid- the first ring (Nn) are defined as the design variables in our
ered as the secondary ones. The number of the secondary program, and the optimum values for these variables can
nodes for the ring n i is equal to Nn × (n i − 1). As an exam- directly be obtained. However, in order to investigate the
ple, in the third ring, the nodes 32, 35, 38, ..., 59 are the effect of Nr and Nn on the optimum designs, here we con-
primary and the remaining nodes are the secondary ones, sider all possible conditions for these design variables. The
when N n = 10 (Fig. 4a). Equations (15), (16) and (17) dome is considered to be subjected to equipment loading of
present the total elements of the diagonal groups. Using 1,000 kN at its crown in this state.
(15), the elements with the primary start nodes are deter- Different design variables defined in (2) affect the weight
mined. As an example, for n i = 1 and j = 4, 5, 6, the of structures in different manners. Nr and Nn determine the
second group contains (3,13), (3,14) and (3,15), as shown number of elements and the height of dome alters the length
in Fig. 5. Equation (17) presents only one element which of elements and these in turn can change the sum of the
connects the first primary node on the ring n i to the last sec- element lengths. On the other hand, the design variables
ondary node on the ring n i + 1 and for the second group, xi determines the cross-sectional areas of elements without
it will be (2,31). The elements related to the secondary change in the geometry of the structure. In order to inves-
nodes are given by (16). For example, using n i = 3 and tigate the effect of the variables separately, the relation of
j = 1, 2, 3, 4 in this equation, the elements (33,63), (33,64), weight for the dome structure is modified as
(34,64) and (34,65) are specified. Figure 5 shows the pri-

nm
mary and secondary nodes and their related elements when W (X) = ρx · Li (18)
N n = 10 and Nr = 4. i=1

where x is the average cross sectional area of the elements.


5 Results and discussion The sum of the element lengths is a function of the height of
the dome.
In this section, the dome described in the previous Tables 1, 2 and 3 list the optimal designs for the geodesic
section is optimized using the CSS algorithm. The LRFD dome with different Nn obtained by the CSS algorithm
specification and drift constraints are considered as the con- when Nr is equal to 3, 4 and 5, respectively. From these
straints for this structure. The modulus of elasticity for the tables it can be observed that a dome with small number
steel is taken as 205 kN/mm2 . The diameter of the dome of elements (Nn) tends to select the greater height. When
is selected as 20 m. The limitations imposed on the joint Nn increases, almost in the all the tables the height of the
displacements are 28 mm in the all directions for the nodes. dome decreases. For a dome with small Nn, having a large
Geometry and topology optimization of geodesic domes using charged system search

Table 1 Optimum design of the geodesic dome with three rings

Group number Optimum sections (designations)

Nn = 3 Nn = 4 Nn = 5 Nn = 6 Nn = 7 Nn = 8 Nn = 9 N n = 10

1 PIPESTa (12) PIPEST (12) PIPST (12) PIPST (10) PIPST (12) PIPST (12) PIPST (10) PIPST (10)
2 PIPST (6) PIPST (5) PIPST (4) PIPST (3.5) PIPST (3.5) PIPST (3.5) PIPST (3) PIPST (3)
3 PIPST (5) PIPST (4) PIPST (3.5) PIPST (3) PIPST (3) PIPST (2.5) PIPST (2.5) PIPST (2.5)
4 PIPST (12) PIPST (8) PIPST (8) PIPST (10) PIPST (6) PIPST (6) PIPST (8) PIPEST (5)
5 PIPST (6) PIPST (5) PIPST (4) PIPST (3.5) PIPST (3.5) PIPST (4) PIPST (3) PIPST (3)
Height (m) 8.00 7.00 7.25 6.50 5.75 5.00 5.25 5.00
Max. displacement (cm) 2.80 2.80 2.80 2.80 2.80 2.74 2.80 2.80
Max. strength ratio 99.82 86.38 90.19 97.26 91.86 99.98 99.99 97.35
Weight (kg) 7,213 6,312 5,977 5,970 6,105 6,161 6,236 6,454

L i (cm) 238.43 258.92 303.65 330.34 353.46 374.38 415.88 445.55
x (cm2 ) 38.5 31.1 25.1 23.0 22.0 21.0 19.1 18.5

a PIPST, PIPEST, and PIPDEST stand for standard weight, extra strong, and double-extra strong, respectively

height helps the dome to prevent instability. In addition, the very strong and therefore the average cross sectional area
selected sections for the elements in a dome with a small Nn becomes a higher value and for the big values of Nn, the
are stronger than those of a dome with a large value for Nn. sum of the element lengths increases the weight of the dome.
This means that although a dome with small Nn has a small Similarly, for the domes with 4 rings (Table 2) when Nn
value for the sum of the element lengths, however its aver- is 7, 8 and 9, the economical designs are obtained. Opti-
age cross sectional area is a big value. The lowest weight mum value of Nn is 7, 8, 9, and 10 for the domes with
design is the one which has the smallest values for the aver- 5 rings (Table 3). As it can be seen when the number of
age cross sectional area and the sum of the element lengths, rings increases, for optimum design, a bigger values should
simultaneously. be selected for Nn. For a constant Nn, when Nr increases,
From Table 1 related to the domes with three rings, the the height remains almost constant and the weight reduces.
good designs are obtained when Nn is set to 5, 6 and 7. In other words, for a constant Nn, when Nr increases, the
For the smaller values for Nn, as expected the sections are height remains almost constant. For small values of Nn,

Table 2 Optimum design of the geodesic dome with four rings

Group number Optimum sections (designations)

Nn = 4 Nn = 5 Nn = 6 Nn = 7 Nn = 8 Nn = 9 N n = 10

1 PIPEST (12) PIPEST (12) PIPEST (12) PIPEST (12) PIPST (12) PIPST (12) PIPST (12)
2 PIPST (6) PIPEST (3.5) PIPST (3.5) PIPST (3.5) PIPST (4) PIPST (3) PIPST (3)
3 PIPST (3) PIPST (3) PIPST (2.5) PIPST (2.5) PIPST (2.5) PIPST (2.5) PIPST (2.5)
4 PIPST (3) PIPST (2.5) PIPST (2.5) PIPEST (2) PIPST (2) PIPST (2) PIPST (2)
5 PIPST (8) PIPST (6) PIPST (5) PIPEST (2) PIPST (3) PIPST (2.5) PIPST (1.25)
6 PIPST (6) PIPST (5) PIPST (3.5) PIPST (5) PIPST (4) PIPEST (3.5) PIPEST (3)
7 PIPST (3.5) PIPST (3.5) PIPST (3.5) PIPST (2.5) PIPST (3) PIPST (2.5) PIPST (3)
Height (m) 7.50 6.50 6.50 5.00 5.00 4.75 4.50
Max. displacement (cm) 2.80 2.80 2.80 2.80 2.76 2.78 2.80
Max. strength ratio 99.98 99.99 98 95.96 100 93.8 83.57
Weight (kg) 6,813 6,243 6,112 6,070 5,982 6,085 6,259

L i (cm) 371.96 404.65 456.83 466.91 513.79 553.17 591.48
x (cm2 ) 23.3 19.7 17.0 16.6 14.8 14.0 13.5
A. Kaveh, S. Talatahari

Table 3 Optimum design of the geodesic dome with five rings

Group number Optimum sections (designations)

Nn = 4 Nn = 5 Nn = 6 Nn = 7 Nn = 8 Nn = 9 N n = 10

1 PIPEST (12) PIPEST (12) PIPST (12) PIPST (12) PIPST (12) PIPST (12) PIPST (12)
2 PIPST (8) PIPST (5) PIPST (4) PIPST (4) PIPST (3.5) PIPST (3) PIPST (3)
3 PIPST (3) PIPST (2.5) PIPST (3.5) PIPST (3) PIPEST (2) PIPEST (2) PIPEST (2)
4 PIPST (2.5) PIPEST (2) PIPST (2.5) PIPST (2) PIPST (2) PIPST (2) PIPST (2)
5 PIPST (2.5) PIPEST (2) PIPST (2) PIPST (2) PIPST (2) PIPST (2) PIPST (1.25)
6 PIPST (2.5) PIPEST (1.5) PIPST (1.25) PIPST (1.5) PIPST (1.25) PIPST (0.75) PIPST (0.5)
7 PIPEST (5) PIPEST (5) PIPST (5) PIPST (5) PIPEST (3.5) PIPEST (3) PIPST (4)
8 PIPST (4) PIPST (3) PIPST (3) PIPST (3) PIPST (3) PIPST (2.5) PIPST (2.5)
9 PIPST (3) PIPST (3) PIPST (2.5) PIPST (2) PIPST (2) PIPST (2) PIPST (2)
Height (m) 7.00 6.75 6.50 5.50 5.25 5.25 4.75
Max. displacement (cm) 2.80 2.74 2.79 2.80 2.80 2.79 2.80
Max. strength ratio 100 99.75 99.9 91.13 98.5 88.57 87.1
Weight (kg) 6,927 6,332 6,210 5,896 5,848 6,041 6,091

L i (cm) 465.16 524.63 582.56 610.68 662.02 722.56 760.23
x (cm2 ) 19.0 15.4 13.6 12.3 11.3 10.7 10.2


nm
the weight reduces, while for the large values of Nn the Figure 7a, b show the relation between L i and the
weight increases, as shown in Fig. 6. As an example, for i=1
height of the geodesic dome when Nn and Nr are changed,
N n = 8, the weight is equal to 6,161, 5,982 and 5,848 kg
respectively. As expected, raising the height of the dome
when Nr = 3, 4 and 5, respectively, while these are equal
increases the sum of the element lengths; however the rate
to 5,977, 6,243 and 6,332 kg when N n = 5.
of the increment is higher when Nn or Nr has a big value.
When the number of rings increases, a feasible dome is
As an example, according to Fig. 7a, when the height of the
not found with N n = 3. This means that we must either
dome rises from 1 to 9 m, the sum of the element lengths
change the height limits or alter the utilized sections. The
increases 27%, 44% and 53% for N n = 3, 6 and 10, respec-
dominant constraints of the designs are often the displace-
tively, when Nr is kept constant (Nr = 4). Also according
ment constraint; however changing a section of the reported
to Fig. 7b, the element lengths increment is 49%, 51% and
designs to a lighter one often causes swerving the stress
51.5% for Nr = 3, 4 and 5, respectively, for N n = 8. From
constraint in elements.
this investigation, two results can be derived. Firstly with a
higher probability a small height will be a suitable choice
rather than a large one for domes with big Nn. In other
words, we expect when the value of the Nn increases, the
7000
Nr = 5 height of the dome decreases. Secondly, if a height is suit-
Nr = 4 able for a dome with constant Nn, for another dome with a
Nr = 3
6500 different number of rings, the new height does not differ a
great deal.
6000 In the other hand, the sum of the element lengths for
the geodesic dome with four rings is 1.38 times larger than
Weight

5500 that of the dome with three rings in average. This value
becomes 1.75 times when the domes with five and three
5000 rings are compared (Nn is constant). While these differences
are smaller when Nn is investigated (Nr remains constant).
4500
As an example, the sum of the element lengths for the four,
4000
five and six nodes on the first ring is 1.13, 1.26 and 1.39
4 5 6 7 8 9 10 times larger than that of the dome with three nodes on the
Nn
first ring, respectively. Therefore, it is expected that the
Fig. 6 The weight compression when Nn and Nr are varied alterations of x must be smaller when the number of nodes
Geometry and topology optimization of geodesic domes using charged system search

Fig. 7 The relation between the 4


sum of lengths and the height of 9
x 10 (a)
the domes when a Nn is varied
(Nr = 4); b Nr is varied 8
(N n = 8)

Sum of the member lengths


7 Nn = 3
Nn = 4
Nn = 5
6
Nn = 6
Nn = 7
5
Nn = 8
Nn = 9
4
Nn = 10

2
1 2 3 4 5 6 7 8 9
Height of the domes

x 10
4
(b)
9

8
Sum of the member lengths

7
Nr = 3
6 Nr = 4
Nr = 5
5

3
1 2 3 4 5 6 7 8 9
Height of the domes

on the first ring is changed compared to it when the num- Case 2: A multiple loading condition containing: the load-
ber of rings is altered. In addition, the value of x must be ing of the previous case and a non-symmetric
decreased as much as possible when Nn or Nr increases. loading condition including loads acting on all the
Therefore, for the domes with small Nn and/or Nr, we will nodes with the sum equal to 800 kN, in addition to
have stronger sections. These points are supported by the loads with the sum of 400 kN acting only on half
comparisons of the results made in the three Tables 1, 2 of the structure, as shown in Fig. 8.
and 3.

The best weight found for the first case is 2,728 kg which
5.2 Optimum design of domes under multiple loadings
almost is more than 45% economical design than the best
one obtained when 1,000 kN acting as a single load on
In the process of design, often non-symmetric and/or multi-
the crown. For Case 2, due to the non-symmetric loading
ple loading conditions must be taken into account. In order
condition, the weight increases. It is 3,186 kg where both
to investigate the efficiency of the present methodology,
symmetric and non-symmetric loads are utilized. Table 4
here two different loading conditions are considered as
presents the results of these cases. In all loading cases, the
follows:
stress constraints are dominant and the displacement is not
active. The maximum stress ratio is 97.86% and 99.98% for
Case 1: A uniform loading acting on the structural with the cases 1 and 2, respectively. The maximum displacement
sum equal to 1,000 kN. In this Case the equivalent values are very small than its allowable value. For both
load for each joint is equal to 1,000 kN divided by cases the number of ring is equal four and the value of the
the number of the unsupported joints. elements (Nn) is equal to 6 and 7 respectively which are
A. Kaveh, S. Talatahari

(a) 6 Concluding remarks

This paper utilizes the Charged System Search algorithm for


design of geodesic domes. This algorithm determines the
total number of rings, the number of nodes on the first ring,
the optimum height and the optimum steel section designa-
tions for the members of geodesic domes from the available
steel pipe section table and implements the design con-
straints from LRFD-AISC. The CSS is inspired by the laws
from electrostatics and Newtonian mechanics. CSS contains
a number of charged particles. Each CP is considered a
charged sphere of radius a, which can impose an electric
force on other CPs. This force and the laws for the motion
(b)
determine the new location of the CPs. From optimization
point of view, this process provides a good balance between
the exploration and the exploitation paradigms of the algo-
rithm which can considerably improve the efficiency of the
algorithm.
A simple procedure is presented to calculate the joint
coordinates and specify the elements to determine the
configuration of the geodesic domes. First, the joint coor-
dinates are calculated and divided into the primary and
secondary types considering the symmetry of the dome.
Then using some simple relationships, the elements are con-
structed. A complete investigation on the effect of the num-
ber of rings and the number of nodes of the first ring on the
Fig. 8 The multiple load condition a a uniform loading; b non-
final optimum design is performed. It is shown that using
symmetric loading these parameters as the design variables, it is possible to
perform an optimum topological design of dome structures.
Semi-actual load conditions containing non-symmetric and
found directly by the present methodology. In addition, the
multiple loading conditions are also taken into account.
height of the dome is reduced in these cases compared to the
Although the results of the CSS algorithm as other heuris-
case with a single loading at the crown.
tic methods should be considered as suboptimal, however
the results show that the CSS method is a robust technique
Table 4 Optimum design of the geodesic dome with multiple loadings
that can successfully be utilized in the practical optimum
Group number Optimum sections (designations) topology design of the domes.
Since the height of the dome is considered as a design
Case 1 Case 2
variable and the wind load is directly determined using the
1 PIPST (2) PIPST (2.5) height of the dome, therefore the wind load will change in
2 PIPST (2.5) PIPST (2.5) the optimization process and it will be treated as a func-
3 PIPST (2.5) PIPST (2.5) tion of a design variable. Therefore, the load condition will
4 PIPST (2) PIPST (2) be changed by altering the design variables and this makes
5 PIPST (2) PIPST (2.5) the problem more difficult. However, formulation of such
6 PIPST (2) PIPST (2.5) a function gives a more accurate simulation of the real con-
7 PIPST (2.5) PIPEST (2) dition. The future work must consider the exact values for
Height (m) 3.25 3.5 the wind load conditions. Since our presented procedure is
Nr 4 4 utilized for different loading conditions, it is expected that
Nn 6 7 such a real condition can also be included in the present
Max. displacement (cm) 1.14 1.50
methodology.
Max. strength ratio 98.76 99.98
Weight (kg) 2728 3186 Acknowledgement The first author is grateful to Iran National
Science Foundation for the support.
Geometry and topology optimization of geodesic domes using charged system search

References Kirsch U (1989) Optimal topologies of structures. Appl Mech Rev


42:223−239
Kirsch U (1997) Reduction and expansion processes in topology
American Institute of Steel Construction (AISC) (1991) Manual of
optimization. In: Rozvany GIN (ed) Topology optimization in
steel construction-load resistance factor design, 3rd edn. AISC,
structural mechanics. Springer, Berlin, pp 197−206
Chicago
Lamberti L (2008) An efficient simulated annealing algorithm for
Bendsoe MP, Mota Soares C (eds) (1992) Proceedings of NATO ARW
design optimization of truss structures. Comput Struct 86:1936−
on topology design of structures. Kluwer, Dordrecht
1953
Bendsoe MP, Sigmund O (2003) Topology optimization: theory,
Lee KS, Geem ZW (2004) A new structural optimization method based
methods and applications. Springer, Berlin
on the harmony search algorithm. Comput Struct 82:781−798
Camp C, Bichon J (2004) Design of space trusses using ant colony Makowski ZS (1984) Analysis, design and construction of braced
optimization. J Struct Eng ASCE 130(5):741−751 domes. Granada, London
Camp C, Pezeshk S, Cao G (1998) Optimized design of two dimen- Makowski ZS (1993) Space structures−−a review of the developments
sional structures using a genetic algorithm. J Struct Eng ASCE within the last decades. Space Struct 4, Thomas Telford, London
124(5):551−559 1:283−292
Dhingra AK, Bennage WA (1995) Topological optimization of Majid KI (1972) Nonlinear structures. Butterworth, London
truss structures using simulated annealing. Eng Optim 24:239− Martínez P, Martí P, Querin OM (2007) Growth method for size,
259 topology, and geometry optimization of truss structures. Struct
Ekhande SG, Selvappalam M, Madugula KS (1989) Stability func- Multidisc Optim 33:13−26
tions for three-dimensional beam-columns. J Struct Eng ASCE Perez RE, Behdinan K (2007) Particle swarm approach for structural
115(2):467−479 design optimization. Comput Struct 85:1579−1588
Fourie P, Groenwold A (2002) The particle swarm optimization algo- Prager W, Rozvany GIN (1977) Optimization of the structural geom-
rithm in size and shape optimization. Struct Multidisc Optim etry. In: Bednarek AR, Cesari L (eds) Dynamical systems.
23(4):259−267 Academic, New York
Hasançebi O, Erbatur F (2002) Layout optimization of trusses using Rahami H, Kaveh A, Gholipour Y (2008) Sizing, geometry and
simulated annealing. Adv Eng Softw 33:681−696 topology optimization of trusses via force method and genetic
Huang X, Xie YM (2010) Evolutionary topology optimization of algorithm. Eng Struct 30(9):2360−2369
continuum structures with an additional displacement constraint. Rajeev S, Krishnamoorthy CS (1992) Discrete optimization of struc-
Struct Multidisc Optim 40(1−6):409−416 tures using genetic algorithms. J Struct Eng, ASCE 118(5):
Kaveh A, Shojaee S (2007) Optimal design of skeletal structures 1233−1250
using ant colony optimization. Int J Numer Methods Eng Rozvany GIN (1998) Exact analytical solutions for some popular
70(5):563−581 benchmark problems in topology optimization. Struct Optim
Kaveh A, Talatahari S (2009a) Particle swarm optimizer, ant colony 15:42−48
strategy and harmony search scheme hybridized for optimization Rozvany GIN (2001) Aims, scope, methods, history and unified ter-
of truss structures. Comput Struct 87(5−6):267−283 minology of computer-aided topology optimization in structural
Kaveh A, Talatahari S (2009b) Size optimization of space trusses mechanics. Struct Multidisc Optim 21:90−108
using Big Bang−Big Crunch algorithm. Comput Struct 87(17−18): Rozvany GIN (2009) A critical review of established methods of
1129−1140 structural topology optimization. Struct Multidisc Optim 37:217−
Kaveh A, Talatahari S (2009c) A particle swarm ant colony optimiza- 237
tion algorithm for truss structures with discrete variables. J Constr Rozvany GIN, Bendsoe MP, Kirsch U (1995) Layout optimization of
Steel Res 65(8−9):1558−1568 structures. Appl Mech Rev 48:41−119
Kaveh A, Talatahari S (2010a) An improved ant colony optimization Saka MP (2007) Optimum geometry design of geodesic domes using
for design of planar steel frames. Eng Struct 32(3):864−873 harmony search algorithm. Adv Struct Eng 10:595−606
Kaveh A, Talatahari S (2010b) Optimal design of Schwedler and Serra M, Venini P (2006) On some applications of ant colony optimiza-
ribbed domes via hybrid Big Bang-Big Crunch algorithm. J tion metaheuristic to plane truss optimization. Struct Multidisc
Constr Steel Res 66(3):412−419 Optim 32(6):499−506
Kaveh A, Talatahari S (2010c) A novel heuristic optimization method: Svanberg K, Werme M (2005) Hierarchical neighbourhood search
charged system search. Acta Mech 213(3−4):267−289 method for topology optimization. Struct Multidisc Optim 29:
Kaveh A, Talatahari S (2010d) Optimal design of skeletal structures 325−340
via the charged system search algorithm. Struct Multidisc Optim Svanberg K, Werme M (2006) Topology optimization by neighbour-
41(6):893−911 hood search method based on efficient sensitivity calculations. Int
Kaveh A, Farahmand Azar B, Talatahari S (2008a) Ant colony J Numer Methods Eng 67:1670−1699
optimization for design of space trusses. Int J Space Struct Topping BHV (1983) Shape optimization of skeletal structures: a
23(3):167−181 review. J Struct Eng 109:1933−1951
Kaveh A, Hassani B, Shojaee S, Tavakkoli SM (2008b) Structural Wang X, Wang MY, Guo D (2004) Structural shape and topology opti-
topology optimization using ant colony methodology. Eng Struct mization in level-set-based framework of region representation.
30(9):2559−2565 Struct Multidisc Optim 27:1−19

View publication stats

You might also like