Sie sind auf Seite 1von 13

Parametric and Shape Optimization of Spatial Trusses Modeled with

Bézier Curves
Waldy J. Torres Z1, Mario P. Rivero G2 and Anderson Pereira3
1
PUC Rio de Janeiro / Mechanical Engineering Department / wtorres19@aluno.puc-rio.br
2
KORDOUT Rio de Janeiro / Engineering Department / m.rivero@kordout.com
3
PUC Rio de Janeiro / Mechanical Engineering Department / anderson@puc-rio.br

Abstract

The design of a structure does not only imply compliance with the requirements of
safety, strength, functionality, and total cost of construction, but also its ability to
respond to optimizations standards pre-defined. Traditionally, the development of a
structural design is based on scientific knowledge, empirical knowledge, and
engineering team experience. This work aims to present a computational methodology
capable of finding the optimal shape of space trusses, through the numerical
implementation of shape optimization and use of parametric curves. Euler's buckling
formula and yield strength constraints must be satisfied to consider the determined
shape to be optimal. An example of a large cross-linked structure is presented to
validate the performance and the applicability of the methodology implemented.

Keywords

Structural optimization; Space trusses; Shape optimization; Parametric Curves.

Introduction

Space structures have been an active area of research since the early 1960s [1]. The
shape and behavior of space structures make this kind of structure attractive in the art of
building façades, roofs, auditoriums, arenas and airports, allowing the development of
differentiated architectural designs. Examples of such structures are shown in Figure 1.

Figure 1. Examples of Lattice Spatial Structures. (a) Salgado Filho International


Airport - Porto Alegre - RS [2]. (b) Anhembi Convention Center - São Paulo - SP
[3].

The term "optimization" is often used in the design area to consider that a model
meets pre-established requirements. Most often designers do not explore other
configurations or design arrangements other possible structural alternative and efficient
solutions to a particular problem, mainly because of cost and time consuming of
iterations and mobilization of different skill inside a project.
According to Arora [4], "optimization" consists of formulating the engineering problem
as an optimization problem, where a performance measure is maximized or minimized,
satisfying the design requirements. Figure 2 presents a flow diagram comparing the
methodology traditionally implemented in the design area with the methodology
considering the same problem as an optimization problem.

Formulate the problem as 0


an Optimization problem

Collect data to describe


1
the system Collect data to describe the 1
system
2 Initial design
Initial design 2
3 Analyze the System
3
Analyze the System
4 Check Performance
Criteria Check Constraints 4

yes yes
Is the designed system Does the design meet the
5 OK! 5
feasible? convergence criteria

Update design based on Upgrade the design using


6 6
experience Optimization concepts

Figure 2. Comparison of Methodologies in the Structural Project. (a)


Conventional Design Methodology [4]. (b) Design Optimization Methodology [4].

This work proposes a computational methodology that allows the determination of


efficient forms in space trusses based on the concepts of mathematical programming
and parameterization of Bezier curves. Finally, an optimized parametric structure model
is presented.

Definition of Parametric Curve

The use of parametric curves is important in several areas, since they allow
modeling and visualization of physical and scientific phenomena. Its main advantage is
the possibility of building efficient and numerically robust algorithms [5].
𝑄(𝑡) Cubic Bézier curve
𝑃3 𝑃4 Control Polygon
Control Points
4 5
3

Symmetry Axis
𝑃2

2 𝑦
𝑥
𝑃1 1
1/4 1/2 3/4 1 𝑡
Figure 3 - Bezier curve.

Figure 3 shows the parameterization of the Bézier curve that passes through the
extreme control points 𝑃1 and 𝑃4, the shape determination is given by the intermediate
control points 𝑃2 and 𝑃3, can be generated by a m degree polynomial. Commonly,
quadratic or cubic Bézier curves are used, since larger-degree polynomials are
computationally more costly to calculate. A cubic Bézier curve can be defined as a
matrix such as follows:

𝑃1𝑥 𝑃1𝑦
2 −2 1 1
−3 3 −2 −1 𝑃2𝑥 𝑃2𝑦
𝑄(𝑡) = [1 𝑡 𝑡2 𝑡 3] [ 0 ] (1)
0 1 0 𝑃3𝑥 𝑃3𝑦
1 0 0 0 [𝑃4 𝑃4𝑦 ]
𝑥

Where to generate 𝑛 points between the initial control point 𝑃1 and the final control
point 𝑃4, it is necessary to divide the parametric space 𝑡 into 𝑛 − 1 range of points
included 0 ≤ 𝑡 ≤ 1. Equation (1) 𝑄(𝑡) represents the Cartesian coordinates of the
interpolated points in the parametric space 𝑡 [6].

Parametric Structure Modeling

Before starting the optimization process, it is necessary to prepare the geometry to


be analyzed. The coordinates in 𝑥 e 𝑦 of the four control points are defined and, by
using equation (1), the new coordinates 𝑎𝑖 interpolated in parametric space are defined
as shown in Figure 4. The nodal positions of the bottom chord and upper chord are
obtained by the following equations:

𝑎𝑖 + 𝑎𝑖+1
𝑝𝑖 = (2)
2
𝑑𝑎𝑖⁄
𝑝𝑖+1 = 𝑎𝑖 + 𝑑𝑡 ℎ (3)
𝑑𝑎𝑖⁄
‖ 𝑑𝑡‖
𝑑𝑎
Since ‖ 𝑖⁄𝑑𝑡‖ is the vector norm tangent at point 𝑎𝑖 and ℎ is the height between
the lower and upper flange.
The total number of nodes and elements in the structure can be determined as
indicated in equations (4) and (5):

𝑁𝑛𝑜𝑑𝑒 = 2(𝑛 + 1) + 1 (4)

𝑁𝐸𝑙 = 4(𝑛 + 1) − 1 (5)

Wherein 𝑛 is the number of divisions of the parametric space.

In the final step, to obtain the shape of the structure it is necessary to establish the
connectivity between the nodes as shown in Figure 4.

𝑁
𝑁4

𝑁3 𝑁1 𝑎 = 𝑁11
𝑎
𝑝𝑝 Chord
𝑁
𝑎4
𝑁2 𝑁

𝑎3 𝑡𝑡 𝑚 Chord
𝑛=4
𝑁1 𝑁 𝑁𝐸𝑙 = 1
𝑎2
𝑎1 = 𝑁 𝑁 𝑜𝑑𝑒 = 11

Figure 4 - Parametric Structure.

Formulation of the Optimization Problem

The solution of an optimization problem is determined through the fundamentals of


mathematical programming, an area that deals with the problems of minimization or
maximization of functions, occupying the theoretical treatment and development of
algorithms for its solution. A mathematical programming problem can be written as [7]:

min 𝑓(𝒙) 𝒙∈ Rn

ℎ𝑘 (𝒙) = 0 𝑘 = 1…𝑚 (6)


𝑠. 𝑡. 𝑔𝑙 (𝒙) ≤ 0 𝑙 = 1…𝑝
𝑥𝑖𝑙𝑏 ≤ 𝑥𝑖 ≤ 𝑥𝑖𝑢𝑏 𝑖 = 1…𝑛

Wherein 𝑓(𝒙) is the objective function to be minimized, ℎ𝑘 (𝒙) and 𝑔𝑙 (𝒙) are the
equality and inequality constraints, respectively. On the other hand, the lateral
constraints 𝑥𝑖𝑙𝑏 and 𝑥𝑖𝑢𝑏 that define the region of possible solution of the optimization
problem are included in the minimum and maximum interval of the design variables 𝑥𝑖 .
In this study, the objective function is represented as the total mass of the structure:

𝐸𝑙

𝑓(𝒙) = 𝑚 𝑇 = ∑ 𝜌𝑖 𝐴𝑖 𝐿𝑖 (7)
𝑖=1

Wherein 𝑁𝐸𝑙 is the total number of bars, 𝜌𝑖 is the material density, 𝐴𝑖 and 𝐿𝑖 are
areas of the cross sections and lengths of each bar member respectively.

The numerical quantities that make it possible to change parameters in an


optimization problem are called design variables, which can be: Bézier curve control
points; nodal coordinates; modulus of elasticity and cross sections of each member of
the structure.

In this study, 8 design variables were considered, with 7 independent design


variables and 1 dependent design variable. Established as follows:
𝑦 𝑦 𝑦
𝒙 = {𝐴𝑏𝑖𝑛 , 𝐴𝑏𝑠𝑢𝑝 , 𝐴𝑑𝑖𝑎𝑔 , 𝑃2𝑥 , 𝑃2 , 𝑃3𝑥 , 𝑃3 , 𝑃4 } (8)

In which, groups 𝐴𝑏𝑖𝑛𝑓 , 𝐴𝑏𝑠𝑢𝑝 and 𝐴𝑑𝑖𝑎𝑔 are variables which store the cross-
sections of the bottom chord, upper chord and diagonals of the structural members, as
shown in Figure 5.

𝐴𝑏𝑖𝑛𝑓
𝐴𝑏𝑠𝑢𝑝
𝐴𝑑𝑖𝑎𝑔
𝑝3 𝑝4

𝑝2

𝑝1

Figure 5 - Area Groups.

𝑦
And the design variables 𝑃2𝑥 to 𝑃4 of equation (8) are defined as the variation of the
𝑦 𝑦
control points in the directions 𝑥 𝑦. Wherein control points 𝑃3 e 𝑃4 have equal
heights.
𝑦 𝑦 𝑦 𝑦
𝑃3 = 𝑃4 = 𝑃3 − 𝑃4 = 0 (9)
Inequality constraints are defined in a standardized manner as:

|𝜎𝑖 (𝒙)|
𝑔𝑙 (𝒙) = −1 ≤0 (10)
𝜎𝑖𝑎 (𝒙)
Wherein 𝜎𝑖 is the axial tension of element 𝑖 and 𝜎𝑖𝑎 is the permissible stress. For 𝜎𝑖 > 0,
traction elements, 𝜎𝑖𝑎 = 𝜎𝑖𝑇 , wherein 𝜎𝑖𝑇 is the admissible tensile stress. For compressed
elements, 𝜎𝑖 < 0, 𝜎𝑖𝑎 (𝒙) = min[𝜎𝑖𝑐 , 𝜎𝑖𝑐𝑟 (𝒙)], wherein 𝜎𝑖𝑐 is the admissible
compressive stress and 𝜎𝑖𝑐𝑟 is the Euler's buckling formula, given by the following
expression:
𝜋 2 𝐸𝑖 𝐼
𝜎𝑖𝑐𝑟 (𝒙) = (11)
𝐴𝑖 𝑙𝑖2
Wherein 𝐸𝑖 the modulus of elasticity of the material, 𝑙𝑖 the length of each bar.

In this study, circular aluminum tubular bars are used where for a mean radius, for
an average radius 𝑅𝑚 and thickness 𝑡𝑠 area and inertia can be written as follows:
𝐴𝑖 = 2𝜋𝑡𝑠 𝑅𝑚 (12)

𝜋𝑅𝑚 𝑡𝑠 2
𝐼= (4𝑅𝑚 + 𝑡𝑠2 ) (13)
4
Optimization Algorithm

Optimization algorithms for linear programming problems have solution in a finite


number of steps. In this study, we used the fmincon function of Matlab® to solve the
optimization problem. The use of the fmincon function uses, sequentially, the
Internal Points (PI) algorithm, which belongs to the class of gradient methods [8],
where the objective function gradients and constraints were provided through the direct
differentiation method, to determine the search direction of the optimization process.

In general, the derivative of the objective function and the constraints are obtained
from the structure response gradient when it is subjected to static loads.

Sensitivity Analysis

Sensitivity analysis, also called as structure response gradients, aims to verify the
behavior of the objective function and the constraints when
a small perturbation is applied in a given design variable, consequently its influence on
the optimization problem is determined.

The sensitivity of the objective function 𝑓(𝒙) and the constraints 𝑔𝑙 (𝒙), can be
determined by differentiating the equations (7) and (10) with respect to the design
variables 𝒙𝑖 , we obtain:

𝜕𝑓(𝒙) 𝜕𝐴𝑖 𝜕𝐿𝑖


= 𝜌𝑖 ( 𝐿𝑖 + 𝐴) (14)
𝜕𝑥𝑖 𝜕𝑥𝑖 𝜕𝑥𝑖 𝑖

𝜕𝑔𝑙 (𝒙) sign[𝜎𝑖 (𝒙)] 𝜕𝜎𝑖 (𝒙) sign[𝜎𝑖 (𝒙)]𝜎𝑖 (𝒙) 𝜕𝜎𝑖𝑎 (𝒙)
= − (15)
𝜕𝑥𝑖 𝜎𝑖𝑎 (𝒙) 𝜕𝑥𝑖 [𝜎𝑖𝑎 (𝒙)]2 𝜕𝑥𝑖

However, the derivative of the axial stress for each bar element is defined as:

𝜕𝜎𝑖 (𝒙) 𝜕𝜎𝑖 (𝒙) 𝜕 𝑖 𝜕𝜎𝑖 (𝒙) 𝜕𝑢𝑖 𝜕 𝑖 𝜕𝑢𝑖


= + = 𝐸𝑖 𝑢 +𝐸 (16)
𝜕𝑥𝑖 𝜕 𝑖 𝜕𝑥𝑖 𝜕𝑢𝑖 𝜕𝑥𝑖 𝜕𝑥𝑖 𝑖 𝑖 𝑖
𝜕𝑥𝑖
Wherein 𝑢𝑖 is the displacement of the structure determined in the finite element
𝜕𝑢
analysis step and 𝜕𝑥𝑖 represents the sensitivity of the displacements as a function of the
𝑖
design variables 𝒙𝒊.

However, if we derive the equation (11) with respect to the design variables 𝑥𝑖 , the
sensitivity of Euler's Buckling formula can be obtained as:

𝜎𝑖𝑐𝑟 (𝒙) 𝜎𝑖𝑐𝑟 (𝒙) 𝜕𝐼 𝜎𝑖𝑐𝑟 (𝒙) 𝜕𝐴𝑖 𝜎𝑖𝑐𝑟 (𝒙) 𝜕𝐿𝑖
= − − (17)
𝜕𝑥𝑖 𝜕𝐼𝑖 𝜕𝑥𝑖 𝜕𝐴𝑖 𝜕𝑥𝑖 𝜕𝐿𝑖 𝜕𝑥𝑖

Or even

𝜎𝑖𝑐𝑟 (𝒙) 𝜋 2 𝐸𝑖 𝜕𝐼 𝜋 2 𝐸𝑖 𝐼 𝜕𝐴𝑖 2𝜋 2 𝐸𝑖 𝐼 𝜕𝐿


= − − (18)
𝜕𝑥𝑖 𝐴𝑖 𝑙𝑖2 𝜕𝑥𝑖 𝐴2𝑖 𝑙𝑖2 𝜕𝑥𝑖 𝐴𝑖 𝑙𝑖3 𝜕𝑥𝑖
Wherein
2
𝜕𝐼 𝜕𝐼 𝜕𝑅𝑚 12𝑅𝑚 + 𝑡𝑠2
= =
𝜕𝑥𝑖 𝜕𝑅𝑚 𝜕𝐴𝑖 8

Scale Factors

The variation in the magnitude of the parameters present in the optimization


problem can cause significant differences between its magnitudes and cause problems in
numerical stability in the algorithm implemented to solve the optimization problem.
Therefore, design variables 𝑥𝑖 are defined as the ratio between the current value of the
design variable and the corresponding initial value, as shown in equation (19).

𝑥𝑖
𝑦𝑖 = (19)
𝑥𝑖

The objective function 𝑓(𝒙) and the constraints 𝑔𝑙 (𝒙) are defined according to the
dimensionless design variable 𝑦𝑖 , rewritten as:
𝑓
𝑓= (21)
𝑓
𝑔𝑙
𝑔𝑙 = (22)
𝑔𝑙
Where 𝑓, 𝑔𝑙 , 𝑓 e 𝑔𝑙 are the current values and corresponding initial values
evaluated in the objective function and constraints respectively.

and lateral constraints as:

𝑥𝑖𝑙𝑏 ≤ 𝑦𝑖 ≤ 𝑥𝑖𝑢𝑏 (23)

In a similar way, the sensitivity of the objective function and the constraints must
be defined as a function of the dimensionless design variable 𝑦𝑖 .
Numbering Example

In this study, we present an example of optimization of the shape of spatial trusses


modeled with Bézier curves, using the methodology developed in the previous sections.
Arc-shaped lattice structure for sugar storage will be analyzed as shown in Figure 6.
The dimensions of the structure are: Internal height 𝐻 = 18,150 𝑚, internal span
𝑉 = 35,0 𝑚 and truss height ℎ = 1,50 𝑚.

ℎ = 1,5 𝑚

𝐻 = 18,150 𝑚
𝐻 = 12,0 𝑚
Stack of Sugar

𝑉 = 35,0 𝑚

Figure 6 - Front View of Initial Structure Configuration.

The following table shows the type of material selected and its mechanical
properties.

Table 1 - Property of Material.


Material Aluminum 6351 - T6
Mechanical Properties Units
Fcy 255 𝑀𝑃𝑎
E 70 𝐺𝑃𝑎
ρ 2647.7956 kg⁄m3

The structure is restricted in directions 𝑥 e 𝑦, in nodal positions 1 and 13. A


symmetry condition around the 𝑦 axis is imposed, which involves restricting nodes 12
and 25 in the direction 𝑥.

In the presented study, it is analyzed with a single load case described as:

1- Load Case 1:

a- Loading distributed of 3 𝑘𝑁/𝑚 acting in the direction of gravity, applied to


the bars of the upper flange.
b- Loading of 50 𝑁 acting in the direction of gravity, applied on all nodes of
the structure.
c- Loading of own weight.
Table 2, in turn, contains the input data of the initial configuration of the structure:

Table 2 - Design Variables, 𝒙𝒍𝒃, 𝒙𝒖𝒃 and Initial Point 𝒙𝟎 .


Variables 𝒙𝒍𝒃 𝒙𝒖𝒃 𝒙𝟎 Units
𝐴𝑏𝑖𝑛𝑓 2.11E-04 2.20E-03 1.21E-3 𝑚2
𝐴𝑏𝑠𝑢𝑝 2.11E-04 2.20E-03 1.21E-3 𝑚2
𝐴𝑑𝑖𝑎𝑔 2.11E-04 1.80E-03 1.01E-3 𝑚2
𝑃2𝑥 0.00 8.00 4.00 𝑚
𝑦
𝑃2 0.00 12.00 6.00 𝑚
𝑥 𝑚
𝑃3 3.50 16.50 10.00
𝑦
𝑃3 14.00 22.30 18.15 𝑚
𝑦
𝑃4 14.00 22.30 18.15 𝑚

In Figure 7 we can see that it consumed thirty iterations for the optimization
algorithm to converge to the optimal solution.

0.9

0.8

0.7

0.6

0.5

0.4

0.3
0 5 10 15 20 25 30 35 40 45 50
Iterations

Figure 7 - Convergence History of the Objective Function.

After you have solved the optimization problem, the design variables are updated
and a new shape in the structure is designed. Figure 8 compares the shape change
between the initial and final arrangement of the structure in analyzes, with final internal
height of 𝐻 = 14,00 𝑚. As can also be seen the distribution of the normalized mass of
the structure in the initial and final configuration with an initial mass of
𝑚𝑖𝑇 = 608,50 𝑘𝑔 and final mass of 𝑚𝑓𝑇 = 185.68 𝑘𝑔, obtaining 6 .45 % of total
initial mass reduction.
Mass (kg)

8
Initial Structure
Final Structure 7

𝐻 = 14.00 𝑚
4

𝑉 = 35.00 𝑚
Figure 8 - Distribution of the Mass in the Initial and Final Structure.
| 𝑔𝑙 (𝒙) |
-0.95

-1.0

-1.05

-1.1

-1.15

-1.2

-1.25

-1.3

Figure 9 - Yield Strength 𝒈𝟏 Constraint.


| 𝑔𝑙 (𝒙) |
0.0
-0.2
-0.4
-0.6
25
-0.8
23
5 -1.0

38 -1.2
37 -1.4

13 -1.6

-1.8

Figure 10 - Euler´s Buckling Formula 𝒈𝟐 Constraint.

Figures 9 and 10 show the plots of constraints associated with Euler's buckling
formula and yield strength constraints, respectively, where values less than zero indicate
that the constraint is being met and values above zero indicate that the constraints are
being violated. It will be seen in Figure 9 that none of the bars are close to their yield
limit. However, it is observed in Figure 10 that the bars 5, 13, 23, 25, 37, 38 and the
bars of the symmetrical right side have constraints close to zero, indicating that they are
active, that is, that the bar is with its Euler´s Buckling formula in the maximum value
allowed, which we can say that the design is being dominated by Euler´s buckling
formula.

In table 3 the convergence of the numerical solution associated with the change of
initial point 𝑥𝑖 is compared, considering any initial point 𝑥𝑖 . However, the numerical
response of the optimization problem tends to converge within a reliable and secure
region with variations in the geometric forms of the structure.

Table 3 - Initial Point Analyzes 𝒙𝟎𝒊 .


𝒇 𝒇 𝒇
Variables 𝒙𝟎𝟏 𝒙𝟏 𝒙𝟎𝟐 𝒙𝟐 𝒙𝟎𝟑 𝒙𝟑 Units
2
𝐴𝑏𝑖𝑛𝑓 2.11E-4 9.30E-4 1.21E-3 5.30E-4 1.11E-3 5.30E-4 𝑚
2
𝐴𝑏𝑠𝑢𝑝 2.11E-4 9.30E-4 1.21E-3 2.22E-4 1.11E-3 2.22E-4 𝑚
2
𝐴𝑑𝑖𝑎𝑔 2.11E-4 7.76E-4 1.01E-3 5.21E-4 1.11E-3 5.21E-4 𝑚
𝑃2𝑥 4.00 3.09 4.00 3.06 4.00 3.06 𝑚2
𝑦
𝑃2 6.00 4.63 6.00 4.58 7.00 4.58 𝑚
𝑥 𝑚
𝑃3 10.00 7.71 10.00 9.50 10.50 9.50
𝑦
𝑃3 18.15 14.00 19.50 14.00 17.00 14.00 𝑚
𝑦
𝑃4 18.15 14.00 19.50 14.00 17.00 14.00 𝑚
Mass 111.86 185.68 624.32 185.68 572.15 185.69 𝑘𝑔
Time 29.7901 22.23 20.92 𝑠
Iterations 49 47 48 𝑁/𝐴

The evaluation of the results was made by comparing the finite element code
developed in Matlab® with Engineering software SAP2000, considering for both cases,
reticulated structure. Table 4 shows the comparison between the two solutions
addressed for the displacements in the directions 𝑥, 𝑦 e 𝑧 of nodal positions 2 and 10, in
addition to comparing the stresses of bars 7 and 41, respectively.

Table 4 - Validation of Results Matlab® Code with SAP2000.


Node / Element Matlab® SAP2000 Relative Error
𝑥 2.86E-04 2.86E-04 0.0
Displacement (𝑚)
𝑦 0.0 0.0 0.0
Node 2
𝑧 -1.7E-3 -1.7E-3 0.0
𝑥 2.1E-3 2.1E-3 0.0
Displacement (𝑚)
𝑦 0.0 0.0 0.0
Node 10
𝑧 -2.05E-2 -2.05E-2 0.0
Stress (Bar 7) (𝑁⁄𝑚 2 ) -4.4221E + 07 -4.4221E + 07 0.0
Stress (Bar 41) (𝑁⁄𝑚 2 ) -4.03431E +07 -4.03431E +07 0.0

The numerical solution presented in Table 4 was satisfactory, with no differences in


displacements and stresses. Therefore, it was demonstrated that the accuracy of the
results obtained in the code implemented in Matlab® is equivalent to the results
provided by the SAP2000 software used for validating the finite element analyzes.
Conclusions

In this work a computational methodology was developed for the parametric and
shape optimization of space trusses modeled with Bézier curves, in order to determine
the optimal configuration of the structure. For this development, structural analysis and
sensitivity analysis, were necessary to the implementation and integration of
mathematical programming. Obtaining the structural response by solving a linear static
problem through a Matlab® implementation. Integration of the Matlab® program with
the SAP2000 was done to verify the optimization results. The design variables
considered were the cross-sectional areas and the coordinates of the control points of the
parametric geometry. The optimization problem is restricted by the yield strength 𝜎𝑖 (𝒙),
Euler´s buckling formula 𝜎𝑖𝑐𝑟 (𝒙) and the lateral constraints 𝑥𝑖𝑙𝑏 and 𝑥𝑖𝑢𝑏 .

Based on the numerical example analyzed in this work, the feasibility and
applicability of parametric and shape optimization in lattice structures can be perfectly
accepted due to the convergence of the objective function to the minimum point. An
increase in the number of area groups can be considered to induce a monotonic response
to the minimum point. Figure 8 shows an improved design in relation to the initial
design, reducing a 6 .45 % of the initial mass,
References

[1] BEHEAD, S. A.; PARKER, G. A. R. Half a Century with the Space Structures Research
Centre of the University of Surrey. IASS – SLTE, 2014.

[2] Image of Salgado Filho International Airport - Porto Alegre, Available at:
<http://www.copa2014.gov.br/pt-br/dinamic/galeria_imagem/14774>, Access on: March
08, 2018.

[3] Image of the Anhembi Convention Center - São Paulo


<http://www.anhembi.com.br/pavilhao-de-exposicoes-anhembi-recebe-30a-feira-
internacional-de-mecanica/>, Access on: March 08, 2018.

[4] ARORA, J. S. Introduction to Optimum Design. 4 ed. Elsevier, 2016.

[5] PALUSZNY, M.; PRAUTZSCH, H.; BOEHM, W. Bézier and B-splines techniques. 1 ed.
Karlsuhe, 2002.

[6] ZOZ J. Estudo de Métodos e Algoritmos de Splines Bézier, Casteljau e B-Spline. 1999
Blumenau, Universidade Regional de Blumenau.

[7] VAZ, L. E.; Pereira, A.; MENEZES, I. F. M. Programação Matemática: Teoria, Algoritmos
e Aplicações na Engenharia. Rio de Janeiro, PUC-Rio, 2012.

[8] HERSKOVITS, J. Advances in Structural Optimization, Kluver Academic Publishers, 1995.

[9] LEE, J. S. A Form-finding Technique for Three-dimensional Spatial Structures.


Architectural Institute of Korea, 2013.

[10] VANDERPLAATS, G. N. Numerical Optimization Techniques for Engineering Design


with Applications. 1 ed. McGraw-Hill, 1984.

[11] FOX, R. L. Optimization Methods for Engineering Design, 1ed. Addison-Wesley


Publishing Company, Inc., 1971.

[12] HANSEN, S. R.; VANDERPLAATS, G. N. Approximation Method for Configuration


Optimization of Trusses. AIAA SDM, 1988.

Das könnte Ihnen auch gefallen