Ali Kaveh · Majid Ilchi Ghazaan

Meta-heuristic
Algorithms for
Optimal Design
of Real-Size
Structures
Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures
Ali Kaveh Majid Ilchi Ghazaan
•

Meta-heuristic Algorithms
for Optimal Design
of Real-Size Structures

123
Ali Kaveh Majid Ilchi Ghazaan
Department of Civil Engineering Department of Civil Engineering
Iran University of Science and Technology Iran University of Science and Technology
Tehran Tehran
Iran Iran

ISBN 978-3-319-78779-4 ISBN 978-3-319-78780-0 (eBook)

https://doi.org/10.1007/978-3-319-78780-0
Library of Congress Control Number: 2018937691

© Springer International Publishing AG, part of Springer Nature 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by the registered company Springer International Publishing AG
part of Springer Nature
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

The importance of optimization is permanently increasing in today’s world due to

limitations in available resources and increase in human population. Engineers
always strive to design efficient structural systems which must be as economical as
possible yet strong enough to withstand the most demanding functional require-
ments arising during their service life. The traditional trial-and-error structural
design approach is not sufficient to achieve designs satisfying economical and
safety criteria simultaneously.
There are a large number of developed methods that may be used to determine
the solution of optimum design problems. These can be collected under two broad
categories of analytical methods and numerical optimization techniques. Analytical
methods are usually employed for finding minimum and maximum values of a
function by using classical mathematical tools. These methods find the optimum
solution as the exact solution of the system of equations which expresses the
conditions for optimality. Although analytical methods are suitable tools for fun-
damental studies of single structural components, these are not suitable to determine
the optimum solution of large-scale structural systems. Recently developed
stochastic search algorithms, however, have provided an efficient tool for solving
large-scale problems. These stochastic search algorithms usually make use of the
ideas taken from the nature and do not require gradient computations of the
objective function and constraints as is the case in mathematical programming
based optimum design methods.
Structural optimization has matured from a narrow academic discipline, where
researchers focused on optimum design of small idealized structural components
and systems, to be the basis for modern design of complex structural systems. The
present book contributes to this expanding discipline by focusing on the opti-
mization of large-scale structures with numerous design variables. Four of the most
well-known and efficient optimization algorithms are presented, and their results are
compared to propose a suitable optimization technique for this class of problems.
These algorithms are Colliding Bodies Optimization, Enhanced Colliding Bodies
Optimization, Vibrating Particles System and a hybrid algorithm based on
Vibrating Particles System, Multi Design Variable Configurations Cascade

v
vi Preface

Optimization, and Upper Bound Strategy. From these four algorithms, the last one
is highly suitable for large-scale problems. In the cascade optimization strategy,
several optimizers can be used, one followed by another in a specified sequence, to
solve a large-scale problem. In this procedure, the first optimizer starts from a
user-specified design known as “cold-start”. The optimal solution achieved in the
first cascade stage is called a “hot-start” and is used to initiate the second opti-
mization stage. Accordingly, each optimization stage of the cascade procedure
starts from the optimum design achieved in the previous stage. Therefore, each
cascade stage except the first one initiates from a hot-start and produces a new
hot-start for the next stage. In general, the optimization algorithm implemented at
each stage of a cascade process may or may not be the same.
In this book, optimal design of different space structures is performed with
different types of limitations such as strength, buckling, displacement, and natural
frequencies. The considered structures consist of double-layer grids, barrel vaults,
domes, antennas, and steel frames.
This book can be considered as an application of meta-heuristic algorithms to
optimal design of skeletal structures. The present book is addressed to those sci-
entists, engineers, and students who wish to explore the potentials of newly
developed meta-heuristics. The concepts presented in this book are not only
applicable to skeletal structures and finite element models but can equally be used
for designing other systems such as hydraulic and electrical networks. The author
and his graduate students have been involved in various developments and appli-
cations of different meta-heuristic algorithms to structural optimization in the last
two decades. This book contains part of this research suitable for various aspects of
optimization for skeletal structures. This book is likely to be of interest to civil,
mechanical, and electrical engineers who use optimization methods for design, as
well as to those students and researchers in structural optimization who will find it
to be necessary professional reading.
In Chap. 1, a short introduction is provided for the development of optimization
and different meta-heuristic algorithms. Chapter 2 presents an explanation of the
meta-heuristic algorithms utilized in this book. In Chap. 3, optimal design of
well-known structural optimization benchmark problems is discussed. Chapter 4
considers optimum design of large-scale special truss structures. Chapter 5 deals with
optimal design of large-scale double-layer grids. Chapter 6 provides optimal design
of large-scale barrel vaults. Chap. 7 deals with optimal design of dome structures.
Chapter 8 discusses optimal design of steel lattice transmission line towers. Chapter 9
provides optimum seismic design of 3D steel frames. Appendix A provides the
computer codes developed for configuration processing of the structures.
We would like to take this opportunity to acknowledge a deep sense of gratitude
to a number of colleagues and friends who, in different ways, have helped in the
preparation of this book. Professor Ch. Bucher encouraged and supported the first
author to write this book. Our special thanks are due to Ms. Silvia Schilgerius, the
senior editor of the Applied Sciences of Springer, for her constructive comments,
editing, and unfailing kindness in the course of the preparation of this book.
Preface vii

My sincere appreciation is extended to our Springer colleague Mr. J. Komala for

excellent typesetting of the book.
We would like to thank Dr. A. Zolghadr for proofreading of this book. We
would like to thank the publishers who permitted some of our papers to be utilized
in the preparation of this book, consisting of Springer Verlag, Elsevier, and Wiley.
Our warmest gratitude is due to our family for their continued support in the course
of preparing this book.
Every effort has been made to render the book error free. However, the authors
would appreciate any remaining errors being brought to their attention through the
following email addresses:
[email protected]
[email protected]

Tehran, Iran Ali Kaveh

April 2018 Majid Ilchi Ghazaan
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . ................ 1
1.1 Structural Optimization Using Meta-heuristic Algorithms . . . . . . . 1
1.2 Goals and Organization of the Present Book ................ 3
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................ 4
2 Optimization Algorithms Utilized in This Book . . . . . . . . . . . . . . . . 7
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Colliding Bodies Optimization Algorithm . . . . . . . . . . . . . . . . . . 7
2.2.1 Theory of Collision Between Two Bodies . . . . . . . . . . . . . 8
2.2.2 Presentation of CBO . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 The Enhanced Colliding Bodies Optimization
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Vibrating Particles System Algorithm . . . . . . . . . . . . . . . . . . . . . 13
2.4.1 Damped Free Vibration . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.2 Presentation of VPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 The MDVC-UVPS Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5.1 The Multi-Design Variable Configurations
Cascade Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5.2 The Upper Bound Strategy . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5.3 Presentation of MDVC-UVPS . . . . . . . . . . . . . . . . . . . . . 20
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3 Optimal Design of Usual-Size Skeletal Structures . . . . . . . . . . . . . . . 23
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Numerical Examples with Frequency Constraints . . . . . . . . . . . . . 24
3.2.1 A 72-Bar Space Truss Problem . . . . . . . . . . . . . . . . . . . . 24
3.2.2 A Spatial 120-Bar Dome-Shaped Truss Problem . . . . . . . . 25
3.2.3 A 200-Bar Planar Truss Problem . . . . . . . . . . . . . . . . . . . 28

ix
x Contents

3.3 Numerical Examples with Strength Constraints . . . . . . . . . . . . . . 32

3.3.1 A Spatial 120-Bar Dome-Shaped Truss Problem . . . . . . . . 32
3.3.2 A 3-Bay 15-Story Frame Problem . . . . . . . . . . . . . . . . . . 34
3.3.3 A 3-Bay 24-Story Frame Problem . . . . . . . . . . . . . . . . . . 40
3.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4 Optimal Design of Large-Scale Special Truss Structures . . . . . . . . . 45
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2 Optimum Design Problem of Steel Trusses . . . . . . . . . . . . . . . . . 45
4.3 Design Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3.1 A Spatial 582-Bar Tower Truss . . . . . . . . . . . . . . . . . . . . 47
4.3.2 A Spatial 942-Bar Tower Truss . . . . . . . . . . . . . . . . . . . . 50
4.3.3 A Spatial 2386-Bar Tower Truss . . . . . . . . . . . . . . . . . . . 50
4.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5 Optimal Design of Double-Layer Grids . . . . . . . . . . . . . . . . . . . . . . . 65
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2 Optimal Design of Double-Layer Grids . . . . . . . . . . . . . . . . . . . . 66
5.3 Design Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.3.1 A 520-Bar Double-Layer Grid . . . . . . . . . . . . . . . . . . . . . 69
5.3.2 A 672-Bar Double-Layer Grid . . . . . . . . . . . . . . . . . . . . . 69
5.3.3 A 800-Bar Double-Layer Grid . . . . . . . . . . . . . . . . . . . . . 74
5.3.4 A 1016-Bar Double-Layer Grid . . . . . . . . . . . . . . . . . . . . 75
5.3.5 A 1520-Bar Double-Layer Grid . . . . . . . . . . . . . . . . . . . . 77
5.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6 Optimal Design of Double-Layer Barrel Vault Space Structures . . . 85
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.2 Optimal Design of Double-Layer Barrel Vaults . . . . . . . . . . . . . . 86
6.3 Design Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.3.1 A 384-Bar Double-Layer Barrel Vault . . . . . . . . . . . . . . . 89
6.3.2 A 693-Bar Double-Layer Barrel Vault . . . . . . . . . . . . . . . 89
6.3.3 A 1536-Bar Double-Layer Barrel Vault . . . . . . . . . . . . . . . 92
6.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7 Optimal Design of Dome-Shaped Trusses . . . . . . . . . . . . . . . . . . . . . 101
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.2 Frequency Constraint Optimization Problem . . . . . . . . . . . . . . . . . 102
7.3 Design Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Contents xi

7.3.1 A 600-Bar Dome Truss . . . . . . . . . . . . . . . . . . . . . . . . . . 104

7.3.2 A 1180-Bar Dome Truss . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.3.3 A 1410-Bar Dome Truss . . . . . . . . . . . . . . . . . . . . . . . . . 115
7.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8 Optimal Design of Steel Lattice Transmission Line Towers . . . . . . . 123
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.2 Optimal Design of Transmission Line Towers . . . . . . . . . . . . . . . 124
8.3 Design Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.3.1 A 47-Bar Power Transmission Tower . . . . . . . . . . . . . . . . 125
8.3.2 A 160-Bar Power Transmission Tower . . . . . . . . . . . . . . . 129
8.3.3 A 244-Bar Power Transmission Tower . . . . . . . . . . . . . . . 133
8.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
9 Optimal Seismic Design of 3D Steel Frames . . . . . . . . . . . . . . . . . . . 139
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
9.2 Optimum Design Problem of Steel Space Frame . . . . . . . . . . . . . 140
9.2.1 Design Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.2.2 The Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.3 Design Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
9.3.1 A Four-Story 132-Member Steel Frame . . . . . . . . . . . . . . 145
9.3.2 A Four-Story 428-Member Steel Frame . . . . . . . . . . . . . . 146
9.3.3 A Twelve-Story 276-Member Steel Frame . . . . . . . . . . . . 150
9.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Appendix A: Configuration Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Chapter 1
Introduction

1.1 Structural Optimization Using Meta-heuristic

Algorithms

The action of making the best or most effective use of a situation or resource is
called optimization. Optimization problems are studied in different fields and var-
ious steps need to be taken to achieve an optimal solution for a problem. These
steps are as follows: The parameters of the problem, which can be either continuous
or discrete, should be recognized. The objective function(s) and the constraints of
the problem have to be identified. At the end, a suitable optimizer should be chosen
and employed to solve the problem.
Structural optimization is a critical and challenging field that has received
considerable attention by engineers and researchers. These optimization problems
can be classified as follows: (1) obtaining optimal size of structural members (sizing
optimization); (2) finding the optimal form for the structure (shape optimization);
and (3) achieving optimal size and connectivity between structural members
(topology optimization). The main concept of this book is to propose a suitable
optimization technique for size optimization of real-size structures where the search
space has a large size and a great number of design constraints must be controlled.
In the past, the most commonly used optimization techniques were gradient-
based algorithms which utilized gradient information to search the solution space
near an initial starting point [1, 2]. In general, gradient-based methods converge
faster and can obtain solutions with higher accuracy compared to stochastic
approaches. However, the acquisition of gradient information can be either costly or
even impossible to obtain the minima. Moreover, these kinds of algorithms only
guarantee convergence to local minima. Furthermore, a good starting point is quite
vital for a successful execution of these methods. In many optimization problems,
prohibited zones, side limits, and non-smooth or non-convex functions should be
taken into consideration. As a result, these non-convex optimization problems
cannot be solved easily by these methods.

© Springer International Publishing AG, part of Springer Nature 2018 1

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_1
2 1 Introduction

On the other hand, other types of optimization methods, known as meta-heuristic

algorithms, do not suffer the abovementioned restrictions. These methods are
suitable for global search due to their capability of exploring and finding promising
regions in the search space in an affordable computational time. Meta-heuristic
algorithms tend to perform well for most of the optimization problems [3, 4]. This is
because these methods refrain from simplifying or making assumptions about the
original problem. Evidence of this is provided by their successful applications to a
vast variety of fields, such as engineering, physics, chemistry, art, economics,
marketing, genetics, operations research, robotics, social sciences, and politics.
A heuristic method can be considered as a procedure that is likely to discover a
very good feasible solution, but not necessarily an optimal solution, for a consid-
ered specific problem. No guarantee can be provided about the quality of the
solution obtained, but a well-designed heuristic method usually can provide a
solution that is at least nearly optimal. The procedure also should be sufficiently
efficient to deal with very large problems. Heuristic methods are often considered as
iterative algorithms, where each iteration involves conducting a search for a new
solution that might be better than the best solution found previously. After a rea-
sonable time when the algorithm is terminated, the solution it provides is the best
one that was found during all iterations. A meta-heuristic is formally defined as an
iterative generation process which guides a subordinate heuristic by combining
intelligently different concepts for exploring (global search) and exploiting (local
search) the search space; learning strategies are used to structure information in
order to find efficiently near-optimal solutions [5–7].
Since the 1970s, many meta-heuristic algorithms have been developed and
applied to different optimization problems. In terms of the source of inspiration,
meta-heuristics can be generally classified into three categories:
1. Evolutionary algorithms: They mimic the concepts of evolution in nature like
Genetic Algorithm (GA) [8], Differential Evolution (DE) [9], Evolutionary
Strategy (ES) [10], Evolutionary Programming (EP) [11, 12], and
Biogeography-Based Optimization (BBO) algorithm [13].
2. Swarm algorithms: these techniques mimic the processes of decentralized,
self-organized systems, which can be either natural or artificial in nature like Ant
Colony Optimization (ACO) [14], Particle Swarm Optimization (PSO) [15],
Artificial Bee Colony (ABC) algorithm [16], Cuckoo Search (CS) algorithm
[17], Firefly Algorithm (FA) [18], Bat Algorithm (BA) [19], Dolphin
Echolocation (DE) [20], and Cyclical Parthenogenesis Algorithm (CPA) [21].
3. Physical algorithms: These methods are inspired by the physical laws like
Simulated Annealing (SA) [22], Charged System Search (CSS) [23], Ray
Optimization (RO) [24], Water Evaporation Optimization (WEO) [25], and
Thermal Exchange Optimization (TEO) [26].
A complete explanation of various recently developed meta-heuristic algorithms
and their applications can be found in Kaveh [27, 28].
1.2 Goals and Organization of the Present Book 3

1.2 Goals and Organization of the Present Book

Sizing optimization problems are very popular design problems and can be found
frequently in literature. The contribution of this book is concerned with size opti-
mization of real-size structures with numerous design variables. Various types of
structures with different design constraints are studied. Four efficient optimization
algorithms will be presented and their results are compared to propose a suitable
optimization technique for this class of problems.
The remaining chapters of this book are organized in the following manner:
Chapter 2 presents the rules of the optimization algorithms employed in this
book. These algorithms consist of Colliding Bodies Optimization (CBO) [29],
Enhanced Colliding Bodies Optimization (ECBO) [30], Vibrating Particles System
(VPS) [31], and a hybrid algorithm called MDVC-UVPS [32].
Chapter 3 presents a comparison between the results obtained by proposed
algorithms and some well-known state-of-the-art meta-heuristics for usual-size
skeletal structures. Four truss design examples and two frame design examples with
continuous and discrete sizing variables are studied in this chapter. Different types
of constraints (i.e., natural frequency constraints and strength and displacement
constraints) are considered for benchmarks.
Chapter 4 presents optimal design of three spatial tower trusses. The examples
have 582, 942, and 2386 elements and contain 32, 76, and 220 variables, respec-
tively. All of the structures are designed according to AISC-ASD provisions [33]
and the cross-sectional areas of the bar elements are selected from W-shape profile
list.
Chapter 5 deals with the design optimization of five double-layer grids with
different configurations. The examples contain larger square on square, square on
larger square, square on square, square on diagonal, and diagonal on diagonal
double-layer grids. Strength and slenderness limitations are imposed according to
AISC-LRFD provisions [34].
Chapter 6 deals with size optimization of three double-layer barrel roof struc-
tures. These examples consist of a 384-bar double-layer barrel vault, a 693-bar
double-layer barrel vault, and a 1536-bar double-layer barrel vault. The structures
are subjected to stress, stability, and displacement limitations according to the
provisions of AISC-ASD [31]. The design variables are the cross-sectional areas of
the bar elements which are selected from steel pipe sections.
Chapter 7 examines the abilities of the proposed methods for size optimization
of dome-shaped trusses. Three dome truss design examples with 600, 1180, and
1410 elements are studied in this chapter. Two different constraint cases are con-
sidered for each example (i.e., natural frequency constraints and strength and dis-
placement constraints).
Chapter 8 presents the application of the proposed algorithms for optimization of
steel lattice transmission line towers. Three benchmark structural examples
including 47-bar, 160-bar, and 244-bar power transmission towers are studied in
this chapter. The design variables are the cross-sectional areas of the bar elements
4 1 Introduction

and in all problems, solution candidates are allowed to select discrete values from a
permissible list of cross sections.
Chapter 9 deals with the optimization of 3D steel frames under seismic loads
based on response spectra. Three irregular steel frame problems (i.e., four-story
132-member, four-story 428-member, and twelve-story 276-member steel frames)
are considered to evaluate the performance of the proposed algorithms. The frames
are designed according to the LRFD-AISC design criteria [35]. Load combinations
recommended by ASCE 7-10 [36] are considered and the frames are Intermediate
Moment Frames (IMF).

References

1. Majid KI (1974) Optimum design of structures. Newness-Butterworth, UK

2. Kirsch U (1993) Structural optimization: fundamentals and applications. Springer, Berlin
3. Gonzalez TF (ed) (2007) Handbook of approximation algorithms and metaheuristics.
Computer and information science series. Chapman & Hall/CRC, Boca Raton, FL
4. Yang X-S (2010) Nature-inspired metaheuristic algorithms, 2nd edn. Luniver Press, UK
5. Glover F, Kochenberger GA (eds) (2003) Handbook of metaheuristics. Kluwer Academic
Publishers, Dordrecht
6. Voß S, Martello S, Osman IH, Roucairol C (eds) (1999) Metaheuristics: advances and trends
in local search paradigms for optimization. Kluwer Academic Publishers, Norwell, MA
7. Osman IH, Laporte G (1996) Metaheuristics: a bibliography. Ann Oper Res 63:513–623
8. Goldberg DE, Holland JH (1988) Genetic algorithms and machine learning. Mach Learn
3:95–99
9. Storn R, Price K (1997) Differential evolution—a simple and efficient heuristic for global
optimization over continuous spaces. J Global Optim 11:341–359
10. Rechenberg I (1973) Evolution strategy: optimization of technical systems by means of
biological evolution, vol 104. Fromman-Holzboog, Stuttgart, pp 15–16
11. Fogel LJ, Owens AJ, Walsh MJ (1966) Artificial intelligence through simulated evolution.
Wiley, Hoboken
12. Yao X, Liu Y, Lin G (1999) Evolutionary programming made faster. Evol Comput IEEE
Trans 3:82–102
13. Simon D (2008) Biogeography-based optimization. Evol Comput IEEE Trans 12:702–713
14. Colorni A, Dorigo M, Maniezzo V (1991) Distributed optimization by ant colonies. In:
Proceedings of the first European conference on artificial life, pp 134–142
15. Eberhart RC, Kennedy J (1995) A new optimizer using particle swarm theory. In: Proceedings
of the sixth international symposium on micro machine and human science, pp 39–43
16. Karaboga D, Basturk B (2007) A powerful and efficient algorithm for numerical function
optimization: artificial bee colony (ABC) algorithm. J Global Optim 39:459–471
17. Yang XS, Deb S (2009) Cuckoo search via Lévy flights. In: Nature & biologically inspired
computing, pp 210–214
18. Yang XS (2010) Engineering optimization: an introduction with metaheuristic applications,
Wiley, PP. 221–230
19. Yang XS (2010) A new metaheuristic bat-inspired algorithm. In: Nature inspired cooperative
strategies for optimization (NICSO 2010). Springer, Berlin, 65–74
20. Kaveh A, Farhoudi N (2013) A new optimization method: dolphin echolocation. Adv Eng
Software 59:53–70
21. Kaveh A, Zolghadr A (2017) Cyclical parthenogenesis algorithm for guided modal strain
energy based structural damage detection. Appl Soft Comput 57:250–264
References 5

22. Kirkpatrick S, Gelatt CD, Vecchi MP (1983) Optimization by simulated annealing. Science
220(4598):671–680
23. Kaveh A, Talatahari S (2010) A novel heuristic optimization method: charged system search.
Acta Mech 213(3):267–289
24. Kaveh A, Khayatazad M (2012) A new meta-heuristic method: ray optimization. Comput
Struct 112:283–294
25. Kaveh A, Bakhshpoori T (2016) Water evaporation optimization: a novel physically inspired
optimization algorithm. Comput Struct 167:69–85
26. Kaveh A, Dadras A (2017) A novel meta-heuristic optimization algorithm: thermal exchange
optimization. Adv Eng Softw 110:69–84
27. Kaveh A (2017) Advances in metaheuristic algorithms for optimal design of structures, 2nd
edn. Springer International Publishing, Switzerland
28. Kaveh A (2017) Applications of metaheuristic optimization algorithms in civil engineering.
Springer, Switzerland
29. Kaveh A, Mahdavi VR (2014) Colliding bodies optimization: a novel meta-heuristic method.
Comput Struct 139:18–27
30. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies optimization for design
problems with continuous and discrete variables. Adv Eng Softw 77:66–75
31. Kaveh A, Ilchi Ghazaan M (2017) Vibrating particles system algorithm for truss optimization
with multiple natural frequency constraints. Acta Mech 228(1):307–322
32. Kaveh A, Ilchi Ghazaan M (2018) A new hybrid meta-heuristic algorithm for optimal design
of large-scale dome structures. Eng Optimiz 50(2):235–252
33. American Institute of Steel Construction (AISC) (1989) Manual of steel construction:
allowable stress design, Chicago, USA
34. American Institute of Steel Construction (AISC) (1994) Manual of steel construction load
resistance factor design, USA
35. AISC 360-10 (2010). Specification for structural steel buildings. American Institute of Steel
Construction, Chicago, Illinois 60601-1802, USA
36. ASCE 7-10 (2010) Minimum design loads for building and other structures
Chapter 2
Optimization Algorithms Utilized
in This Book

2.1 Introduction

The main features and rules of the optimization algorithms utilized in this book are
explained in this chapter. These algorithms consist of Colliding Bodies
Optimization (CBO) [1], Enhanced Colliding Bodies Optimization (ECBO) [2],
Vibrating Particles System (VPS) [3], and a hybrid algorithm called MDVC-UVPS
[4]. All of the algorithms considered here are recently developed and are
multi-agent meta-heuristic methods. These algorithms start with a set of randomly
selected candidate solutions of the optimization problem and according to a series
of simple rules, mainly inspired by the nature, the existing solutions are perturbed
iteratively in order to improve their cost function values. Many other recently
developed meta-heuristic algorithms and their applications can be found in the
recently published books of Kaveh [5, 6].

2.2 Colliding Bodies Optimization Algorithm

Colliding Bodies Optimization (CBO) is an efficient meta-heuristic algorithm which

is recently introduced by Kaveh and Mahdavi [1]. CBO is based on the governing
physical laws of one-dimensional collision between two bodies which states that
when an object collides with another one, they move toward minimum energy
levels. CBO algorithm is simple in concept, depends on no internal parameters, and
does not require memory for saving the best-so-far solutions.

© Springer International Publishing AG, part of Springer Nature 2018 7

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_2
8 2 Optimization Algorithms Utilized in This Book

2.2.1 Theory of Collision Between Two Bodies

Collisions between bodies are governed by the laws of momentum and energy.
When a collision occurs in an isolated system between two objects (Fig. 2.1), the
total momentum of the system of objects is conserved. Provided that there are no
net external forces acting upon the objects, the total momentum of the objects
before the collision equals the total momentum of the objects after the collision.
Conservation of total momentum can be expressed by the following equation:

m1 v1 þ m2 v2 ¼ m1 v01 þ m2 v02 ð2:1Þ

Likewise, the conservation of the total kinetic energy is expressed as

1 1 1 1
m1 v21 þ m2 v22 ¼ m1 v02 02
1 þ m 2 v2 þ Q ð2:2Þ
2 2 2 2

where v1 is the initial velocity of the first object before impact, v2 is the initial
velocity of the second object before impact, v01 is the final velocity of the first object
after impact, v02 is the final velocity of the second object after impact, m1 is the mass
of the first object, m2 is the mass of the second object, and Q is the loss of kinetic
energy due to the impact [7].

Fig. 2.1 The collision

between two bodies, a before
collision, b at the time of
collision, and c after collision
2.2 Colliding Bodies Optimization Algorithm 9

The formulas for the velocities after a one-dimensional collision are

ðm1  e m2 Þv1 þ ðm2 þ e m2 Þv2

v01 ¼ ð2:3Þ
m1 þ m2

ðm2  e m1 Þv2 þ ðm1 þ e m1 Þv1

v02 ¼ ð2:4Þ
m1 þ m2

where e is the Coefficient Of Restitution (COR) of the two colliding bodies, defined
as the ratio of the relative velocity of separation to the relative velocity of approach:

0


v  v0
v0
e¼ 2 1
¼ ð2:5Þ
j v2  v1 j v

According to the coefficient of restitution, there are two special cases of any
collision as follows:
1. A perfectly elastic collision is defined as the one in which there is no loss of kinetic
energy in the collision ðQ ¼ 0 and e ¼ 1Þ. In reality, any macroscopic collision
between objects will convert some kinetic energy to internal energy and other
forms of energy. In this case, after collision, the velocity of separation is high.
2. An inelastic collision is the one in which part of the kinetic energy is changed to
some other forms of energy in the collision. Momentum is conserved in inelastic
collisions (as it is for elastic collisions), but one cannot track the kinetic energy
through the collision since some of it will be converted to other forms of energy.
In this case, the coefficient of restitution is not equal to one ðQ 6¼ 0 and e\1Þ,
and after collision the velocity of separation is low.
For most of the real objects, the value of e is between 0 and 1.

2.2.2 Presentation of CBO

In CBO, each solution candidate is considered as a Colliding Body (CB) and these
massed objects are composed of two main equal groups, i.e., stationary and moving
objects, where the moving objects move to follow stationary objects and a collision
occurs between pairs of objects. This is done for two purposes: (i) to improve the
positions of moving objects; and (ii) to push stationary objects toward better
positions. After the collision, the new positions of colliding bodies are updated
based on their new velocities using the collision laws as discussed in the previous
subsection. The procedure of CBO can be outlined as follows and its pseudocode is
provided in Fig. 2.2.
Step 1 The initial positions of all colliding bodies are determined randomly in
the search space.
10 2 Optimization Algorithms Utilized in This Book

procedure CBO
Initialize algorithm parameters
Initial positions are created randomly
The values of objective function and masses are evaluated
while maximum iterations is not fulfilled
Stationary and moving groups are created
for each CB
The velocity before the collision is evaluated by Eq. (2.7) or Eq. (2.8)
The velocity after the collision is evaluated by Eq. (2.9) or Eq. (2.10)
New location is updated by Eq. (2.12) or Eq. (2.13)
end for
The values of objective function and masses are evaluated
end while
end procedure

Fig. 2.2 Pseudocode of the CBO algorithm

Step 2 Each CB has a specified mass defined as

1
fitðkÞ
m k ¼ Pn 1
; k ¼ 1; 2; . . .; n ð2:6Þ
i¼1 fitðiÞ

where fit(i) represents the objective function value of the ith CB and n is
the number of colliding bodies. In order to select pairs of objects for
collision, CBs are sorted according to their mass in an decreasing order
and they are divided into two equal groups: (i) stationary group and
(ii) moving group (see Fig. 2.3). Moving objects collide with stationary
objects to improve their positions and push stationary objects toward
better positions.

Fig. 2.3 The pairs of objects for collision

2.2 Colliding Bodies Optimization Algorithm 11

Step 3 The velocity of stationary bodies before collision is zero and moving
objects move toward stationary objects:
vi ¼ 0; i ¼ 1; 2; . . .; n2 ð2:7Þ

vi ¼ xin2  xi ; i ¼ n2 þ 1; n2 þ 2; . . .; n ð2:8Þ

Step 4 The velocities of stationary and moving bodies after the collision (v′i) are
evaluated by
0 ðmi þ n2 þ emi þ n2 Þvi þ n2 n
vi ¼ i ¼ 1; 2; :::; ð2:9Þ
mi þ mi þ n2 2
 
mi e min vi
0
vi ¼
2
i ¼ n2 þ 1; n2 þ 2; . . .; n ð2:10Þ
mi þ min
2

e is the coefficient of restitution that decreases linearly from unit to zero.

Thus, it is stated as
iter
e¼1 ð2:11Þ
itermax

where iter is the current iteration number and itermax is the total number
of iterations for optimization process.
Step 5 The new position of each stationary CB is
n n
xnew
i ¼ xin2 þ rand  v0i ; i ¼ þ 1; þ 2; . . .n ð2:12Þ
2 2

where xnew
i , xi, and v′i are the new position, previous position, and the
velocity after the collision of the ith CB, respectively. rand is a random
vector uniformly distributed in the range of [−1,1] and the sign ‘‘°’’
denotes an element-by-element multiplication. The new position of each
moving CB is calculated by
0
xnew
i ¼ xin2 þ rand  vi ; i ¼ n2 þ 1; n2 þ 2; . . .; n ð2:13Þ

Step 6 The optimization process is terminated after a fixed number of iterations.

If this criterion is not satisfied, go to Step 2 for a new round of iteration.

2.3 The Enhanced Colliding Bodies Optimization

Algorithm

In order to get faster and more reliable solutions, Enhanced Colliding Bodies
Optimization (ECBO) is developed [2]. Memory which saves a number of the
best-so-far solutions is used in ECBO to improve the algorithm’s performance.
12 2 Optimization Algorithms Utilized in This Book

A mechanism is also defined to change some components of CBs randomly to

afford a chance for the CBs to escape from local minima and prevent probable
premature convergence. The steps of this technique are outlined in the following
and its pseudocode is provided in Fig. 2.4.
Step 1 The initial position vectors of all CBs are created randomly in the search
space.
Step 2 The value of mass for each CB is evaluated according to Eq. (2.6).
Step 3 Colliding Memory (CM) is utilized to save a number of historically best
CB vectors and their related mass and objective function values.
Solution vectors which are saved in CM are added to the population and
the same number of current worst CBs are deleted. Finally, CBs are
sorted according to their objective function values in an increasing order.
Using this mechanism can improve the algorithm’s performance without
increasing the computational cost.
Step 4 CBs are composed of two main equal groups, i.e., stationary and moving
groups. The pairs of CBs for collision are selected according to Fig. 2.3.
Step 5 The velocity of stationary bodies before the collision is zero (Eq. (2.7)).
Besides, moving objects move toward stationary objects and their
velocities before the collision are calculated by Eq. (2.8).

procedure ECBO
Initialize algorithm parameters
Initial positions are created randomly
The values of objective function and masses are evaluated
while maximum iterations is not fulfilled
Colliding memory is updated
The population is updated
Stationary and moving groups are created
for each CB
The velocity before the collision is evaluated by Eq. (2.7) or Eq. (2.8)
The velocity after the collision is evaluated by Eq. (2.9) or Eq. (2.10)
New location is updated by Eq. (2.12) or Eq. (2.13)
If rnj < pro
k ← random_int (1,m) /* m is the number of variables */
kth dimension is regenerated randomly in its allowable range
end if
end for
The values of objective function and masses are evaluated
end while
end procedure

Fig. 2.4 Pseudocode of the ECBO algorithm

2.2 Colliding Bodies Optimization Algorithm 13

Step 6 The velocities of stationary and moving bodies after collision are eval-
uated by Eqs. (2.9) and (2.10), respectively.
Step 7 The new position of each CB is calculated by Eqs. (2.12) and (2.13).
Step 8 To improve the exploration capabilities of the standard CBO and to
prevent premature convergence, a stochastic approach is employed in
ECBO. A parameter like Pro within (0, 1) is introduced to specify
whether a component of each CB must be changed or not. For each
colliding body, Pro is compared with rni (i = 1,2,…,n) which is a ran-
dom number uniformly distributed within (0, 1). If rni < pro, one
dimension of the ith CB is selected randomly and its value is regenerated
by
 j j 
xij ¼ xmin
j
þ random xmax  xmin ð2:14Þ

where xij is the jth variable of the ith CB. xmin

j j
and xmax are the minimum
and maximum limits of the jth variable. In order to protect the structures
of CBs, only one dimension is changed.
Step 9 The process of optimization is terminated after a predefined maximum
number of objective function evaluations. If it is not fulfilled, go to Step
2 for a new round of iteration.

2.4 Vibrating Particles System Algorithm

Vibrating Particles System (VPS) algorithm is a population-based algorithm which

simulates free vibration of single degree of freedom systems with viscous damping
[3]. Similar to other multi-agent methods, VPS has a number of individuals (or
particles) consisting of the variables of the problem. The solution candidates
gradually approach to their equilibrium positions which are determined using a
combination of the current population and the historically best position in order to
maintain a proper balance between diversification and intensification.

2.4.1 Damped Free Vibration

A vibration is the oscillating motion of a particle or a body about a position of

equilibrium. In general, there are two types of vibrations: (1) free vibration and
(2) forced vibration. When the motion is maintained by the restoring forces only,
the vibration is said to be a free vibration and when a force is applied to the system,
the resulting motion is described as a forced vibration. In the study of a vibrating
system, the effects of friction can be neglected resulting in an undamped vibration.
However, all vibrations are actually damped to some degree by friction forces.
14 2 Optimization Algorithms Utilized in This Book

These forces can be caused by dry friction, or Coulomb friction, between rigid
bodies, by fluid friction when a rigid body moves in a fluid, or by internal friction
between the molecules of a seemingly elastic body. In this section, the free
vibration of single degree of freedom systems with viscous damping is studied. The
viscous damping is caused by fluid friction at low and moderate speeds. Viscous
damping is characterized by the fact that the friction force is directly proportional
and opposite to the velocity of the moving body [8].
The vibrating motion of a body or system of mass m having viscous damping
can be characterized by a block and a spring of constant k as it is shown in Fig. 2.5.
The effect of damping is provided by the dashpot connected to the block, and the
magnitude of the friction force exerted on the plunger by the surrounding fluid is
equal to c_x (c is the coefficient of viscous damping, and its value depends on the
physical properties of the fluid and the construction of the dashpot). If the block is
displaced a distance x from its equilibrium position, the equation of motion can be
expressed as

m€x þ c_x þ kx ¼ 0 ð2:15Þ

Before presenting the solutions for this differential equation, the critical damping
coefficient cc is defined as

cc ¼ 2m xn ð2:16Þ
rffiffiffiffi
k
xn ¼ ð2:17Þ
m

where xn is the natural circular frequency of the vibration.

Depending on the value of the coefficient of viscous damping, three different
cases of damping can be distinguished: (1) over-damped system (c > cc), (2) criti-
cally damped system (c = cc), and (3) under-damped system (c < cc). The solutions
of over-damped and critically damped systems correspond to a non-vibratory
motion. Therefore, the system only oscillates about its equilibrium position when
c < cc .

Fig. 2.5 Free vibration of a system with damping

2.2 Colliding Bodies Optimization Algorithm 15

Fig. 2.6 Vibrating motion of under-damped system

The solution of Eq. (2.15) for under-damped system is as follows:

xðtÞ ¼ q enxn t sinðxD t þ /Þ ð2:18Þ

qffiffiffiffiffiffiffiffiffiffiffiffiffi
xD ¼ xn 1  n2 ð2:19Þ

c
n¼ ð2:20Þ
2m xn

where q and / are constants generally determined from the initial conditions of the
problem. xD and n are damped natural frequency and damping ratio, respectively.
Equation (2.18) is shown in Fig. 2.6, and the effect of damping ratio on vibratory
motion is illustrated in Fig. 2.7.

2.4.2 Presentation of VPS

In the VPS, the position of each particle is updated by learning from the historically
best position of the entire population, a good particle, and a bad particle. By
controlling the weights of these terms, a proper balance between the diversification
and the intensification inclinations can be achieved. The VPS procedure can be
outlined as follows and its pseudocode is provided in Fig. 2.8.
Step 1 The initial positions of all particles are determined randomly in search
space.
Step 2 The objective function value is calculated for each particle.
Step 3 For each particle, three equilibrium positions with different weights are
defined that the particle tends to approach: (1) the best position achieved
so far across the entire population (HB), (2) a good particle (GP), and
16 2 Optimization Algorithms Utilized in This Book

Fig. 2.7 Free vibration of systems with four levels of damping: a n ¼ 5%, b n ¼ 10%,
c n ¼ 15%, and d n ¼ 20%

procedure VPS
Initialize algorithm parameters
Initial positions are created randomly
The values of objective function are evaluated and HB is stored
while maximum iterations is not fulfilled
for each particle
The GP and BP are chosen
if P<rand
w3=0 and w2=1-w1
end if
for each component
New location is obtained by Eq. (2.22)
end for
Violated components are regenerated by harmony search-based handling
approach
end for
The values of objective function are evaluated and HB is updated
end while
end procedure

Fig. 2.8 Pseudocode of the VPS algorithm

(3) a bad particle (BP). In order to select the GP and BP for each
candidate solution, the current population is sorted according to their
objective function values in an increasing order, and then GP and BP are
chosen randomly from the first and second half, respectively.
2.2 Colliding Bodies Optimization Algorithm 17

Figure 2.7 shows the important effect of damping level in the vibration. In order
to model this phenomenon in the optimization algorithm, a descending function that
is proportional to the number of iterations is proposed as follows:
 a
iter
D¼ ð2:21Þ
itermax

where iter is the current iteration number and itermax is the total number of iterations
for optimization process. a is a constant and Fig. 2.9 shows the effect of this
parameter on D.
According to the mentioned concepts, the positions are updated by

xij ¼ w1 :½D:A:rand 1 þ HB j  þ w2 :½D:A:rand 2 þ GP j  þ w3 :½D:A:rand 3 þ BP j 

ð2:22Þ

A ¼ ½w1 :ðHB j  xij Þ þ ½w2 :ðGP j  xij Þ þ ½w3 :ðBP j  xij Þ ð2:23Þ

w1 þ w2 þ w3 ¼ 1 ð2:24Þ

where xji is the jth variable of particle i. w1, w2, and w3 are three parameters to
measure the relative importance of HB, GP, and BP, respectively. Here, rand1,
rand2, and rand3 are random numbers uniformly distributed in the range of [0,1].
The effects of A and D parameters in Eq. (2.22) are similar to that of q and enxn t in
Eq. (2.18), respectively. Also, the value of sinðxD t þ /Þ is considered as unity
(xðtÞ ¼ q enxn t is shown in Fig. 2.6 by dotted line).
A parameter like p within (0, 1) is defined to specify whether the effect of BP
must be considered in updating position or not. For each particle, p is compared

Fig. 2.9 The influence of a on function D

18 2 Optimization Algorithms Utilized in This Book

with rand (a random numbers uniformly distributed in the range of [0,1]) and if p <
rand, then w3 = 0 and w2 = 1−w1.
Three essential concepts consisting of self-adaptation, cooperation, and com-
petition are considered in this algorithm. Particles move toward HB so the
self-adaptation is provided. Any particle has the chance to have influence on the
new position of the other one, so the cooperation between the particles is supplied.
Because of the p parameter, the influence of GP (good particle) is more than that of
BP (bad particle), and therefore the competition is provided.
Step 4 The particle moves in the search space to find a better result and may
violate the side constraints. If any component of the system violates a
boundary, it must be regenerated by harmony search-based side con-
straint handling approach [9]. In this technique, there is a possibility like
HMCR (harmony memory considering rate) that specifies whether the
violating component must be changed with the corresponding compo-
nent of the historically best position of a random particle or it should be
determined randomly in the search space. Moreover, if the component of
a historically best position is selected, there is a possibility like PAR
(pitch adjusting rate) that specifies whether this value should be changed
with the neighboring value or not.
Step 5 Steps 2 through 4 are repeated until a termination criterion is fulfilled.
Any terminating condition can be incorporated; however, in this book,
the optimization process is terminated after a fixed number of iterations.

2.5 The MDVC-UVPS Algorithm

MDVC-UVPS is a hybrid algorithm based on VPS, Multi-Design Variable

Configurations cascade optimization (Multi-DVC) and Upper Bound Strategy
(UBS) [4]. The VPS is used as the main engine of the hybrid algorithm; Multi-DVC
cascade optimization is utilized to handle the large number of variables and UBS is
employed to reduce the computational time. The reason for selecting VPS for
hybridization is that CBO, ECBO, and VPS algorithms have been used previously
for optimal design of many structures, and the VPS has shown its superiority
compared to CBO and ECBO (Kaveh [5, 6]). This will be further clarified by the
examples of the subsequent chapter.

2.5.1 The Multi-Design Variable Configurations Cascade

Optimization

The cascade optimization strategy is proposed to use several optimizers, one fol-
lowed by another in a specified sequence, to solve a large-scale problem [10]. In
this procedure, the first optimizer starts from a user-specified design known as
2.2 Colliding Bodies Optimization Algorithm 19

“cold-start”. The optimal solution achieved in the first cascade stage is called a
“hot-start” and is used to initiate the second optimization stage. Accordingly, each
optimization stage of the cascade procedure starts from the optimum design
achieved in the previous stage. Therefore, each cascade stage except the first one
initiates from a hot-start and produces a new hot-start for the next stage. In general,
the optimization algorithm implemented at each stage of a cascade process may or
may not be the same [11, 12].
In the multi-DVC cascade optimization, a series of appropriate Design Variable
Configurations (DVCs) for the optimization problem under consideration is con-
structed to use a different configuration at each cascade optimization stage. The
coarsest DVC which avoids confusing the employed optimizer with huge design
spaces is utilized in the first stage of the cascade procedure. Therefore, the areas of
appropriate design variable values are identified by detecting near-optimum solu-
tions among the relatively limited design options provided. As the number of design
variables processed in the cascade stages becomes larger, more detailed represen-
tation of the full design space is offered and the optimizer is given the opportunity
to improve the quality of the optimal solution reached. In the final cascade stages
utilizing the finest DVC, relatively small adjustments to an already good-quality
design occur in an effort to identify (or at least approach) the globally optimum
design. In summary, the first optimization stage of the cascade procedure serves the
purpose of basic design space exploration, while the last stages aim at fine-tuning
the achieved optimal solution.

2.5.2 The Upper Bound Strategy

An upper bound strategy is proposed as a simple, yet efficient strategy, to reduce the
total number of structural analyses through avoiding unnecessary analyses during
the course of optimization. The key issue in the UBS is to detect those candidate
designs which have no chance to surpass the best design found so far during the
iterations of the optimum design process. After identifying those non-improving
designs, they are directly excluded from the structural analysis stage, resulting in a
significant saving in the computational effort. The current best design can usually be
considered as the upper bound for the forthcoming candidates to eliminate
unnecessary structural analysis and associated fitness computations for those can-
didates that have no chance of surpassing the best solution. Basically, the key
feature in this approach is to define the penalized weight of the current best solution
found during the previous iterations as an upper bound for the net weight of the
newly generated candidate solutions. Thus, any new candidate solution with a net
weight greater than this upper bound will not be analyzed and this will reduce the
computational burden of the optimization algorithm [13].
20 2 Optimization Algorithms Utilized in This Book

2.5.3 Presentation of MDVC-UVPS

According to the above concepts, the following steps are provided to introduce the
MDVC-UVPS algorithm. Pseudocode of this algorithm is provided in Fig. 2.10.
Step 1 A series of design variable configurations are constructed and sorted
according to the size in an increasing order. The coarsest DVC is utilized
in the first stage of the cascade procedure and the initial positions of all

procedure MDVC-UVPS
Initialize algorithm parameters
A series of design variable configurations are defined
The coarsest DVC is selected and initial positions are created randomly based on it
The values of objective function are evaluated and HB is stored
while maximum iterations is not fulfilled
The upper bound is selected
for each particle
The GP and BP are chosen
if P<rand
w3=0 and w2=1-w1
end if
for each component
New location is obtained by Eq. (2.22)
end for
Violated components are regenerated by harmony search-based handling
approach
Net weight is computed
if net weight > upper bound
The position of particle is replaced to its historically best solution
else
The objective function value is calculated
end if
end for
if the termination criterion of current DVC is fulfilled
The next DVC is used
The population is updated
end for
HB is updated
end while
end procedure

Fig. 2.10 Pseudocode of the MDVC-UVPS algorithm

2.2 Colliding Bodies Optimization Algorithm 21

particles are randomly set based on this configuration. The objective

function values are calculated for the initial population.
Step 2 Since the benchmark examples utilized in this book are large size
problems, the value of the so far best design is not selected as the upper
bound. In order to select it, the median of the historically best objective
function values is chosen.
Step 3 New positions are determined by Eq. (2.22) and violated components are
regenerated by the harmony search-based handling approach.
Step 4 Net weight of each solution candidate is computed (note that it is not
necessary to consider design constraints). Any particle with a net weight
greater than upper bound will not be analyzed and its position is replaced
to its historically best design. Other particles should be analyzed and
their objective function values should be calculated (design constraints
are considered for these particles).
Step 5 The optimization process in each stage except the last one is terminated
after a fixed number of iterations with no improvement. If this criterion is
satisfied, the next DVC is used in the new stage and particles are ini-
tialized based on the finally attained optimum design of the previous
stage. In the new population, the HB is one of the particles and the other
ones are randomly selected from the neighboring region of HB by the
normal distribution with the mean HB j and the standard deviation
j j
xmax xminj j
C (xmin and xmax are the minimum and maximum limits of the jth
variable and C is a constant). After a predefined number of iterations, the
cascade process is terminated.

References

1. Kaveh A, Mahdavi VR (2014) Colliding bodies optimization: a novel meta-heuristic method.

Comput Struct 139:18–27
2. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies optimization for design
problems with continuous and discrete variables. Adv Eng Softw 77:66–75
3. Kaveh A, Ilchi Ghazaan M (2017) Vibrating particles system algorithm for truss optimization
with multiple natural frequency constraints. Acta Mech 228(1):307–322
4. Kaveh A, Ilchi Ghazaan M (2018) A new hybrid meta-heuristic algorithm for optimal design
of large-scale dome structures. Eng Optimiz 50(2):235–252
5. Kaveh A (2017) Advances in metaheuristic algorithms for optimal design of structures, 2nd
edn. Springer International Publishing, Switzerland
6. Kaveh A (2017) Applications of metaheuristic optimization algorithms in civil engineering.
Springer, Switzerland
7. Tolman RC (1979) The principles of statistical mechanics. Clarendon Press, Oxford
(Reissued)
8. Beer FP, Johnston ER Jr, Mazurek DF, Cornwell P, Self BP (2013) Vector mechanics for
engineers. McGraw-Hill, New York, USA
9. Kaveh A, Talatahari S (2010) A novel heuristic optimization method: charged system search.
Acta Mech 213(3):267–289
22 2 Optimization Algorithms Utilized in This Book

10. Patnaik SN, Coroneos RM, Hopkins DA (1997) A cascade optimization strategy for solution
of difficult design problems. Int J Numer Methods Eng 40:2257–2266
11. Charmpis DC, Lagaros ND, Papadrakakis M (2005) Multi-database exploration of large
design spaces in the framework of cascade evolutionary structural sizing optimization.
Comput Methods Appl Mech Eng 194:3315–3330
12. Lagaros ND (2014) A general purpose real-world structural design optimization computing
platform. Struct Multidiscip Optim 49(6):1047–1066
13. Kazemzadeh Azad S, Hasançebi O, Kazemzadeh Azad S (2013) Upper bound strategy for
metaheuristic based design optimization of steel frames. Adv Eng Softw 57:19–32
Chapter 3
Optimal Design of Usual-Size Skeletal
Structures

3.1 Introduction

Sizing optimization of truss and frame structures are frequent structural design
problems that are subjected to various constraints such as displacements, stress,
buckling, and natural frequencies. A great number of papers has been published in
literature, where different meta-heuristic search algorithms have been applied to this
class of problems [1, 2]. The aim of this chapter is to examine the ability of the CBO,
ECBO and VPS which have been utilized in the next chapters for comparison with
MDVC-UVPS. The results of well-known state-of-the-art meta-heuristics are also
provided and compared here. The reason for selecting VPS for hybridization is that
the VPS has shown its superiority compared to CBO and ECBO for optimal design of
many structures, Kaveh [1, 2].
There are many design constraints that should be fulfilled in structural opti-
mization. In the literature, the benchmark design examples usually studied with
displacements, stress, and buckling constraints or displacements and frequencies
constraints. Therefore, we decided to follow the same path in this book. Structural
design examples with frequency constraints are studied in the next section and after
that, benchmark examples with strength constraints are optimized.
The optimization problem can formally be stated as

Find fXg ¼ ½x1 ; x2 ; . . .; xng 

Png P
nmðiÞ
to minimize WðfXgÞ ¼ xi qj Lj
( i¼1 j¼1 ð3:1Þ
gj ðfXgÞ  0; j ¼ 1; 2; . . .; nc
subjected to:
ximin  xi  ximax

where {X} is the vector containing the design variables; W({X}) presents the weight of
the structure; ng is the number of design groups; nm(i) is the number of members for the

© Springer International Publishing AG, part of Springer Nature 2018 23

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_3
24 3 Optimum Design of Usual-Size Skeletal Structures

ith group; and qj and Lj denote the material density and the length of the jth member,
respectively. ximin and ximax are the lower and upper bounds of the design variable xi,
respectively. gj({X}) denotes design constraints, and nc is the number of the constraints.
For constraints handling, a penalty approach is utilized. For this purpose, the
objective function (Eq. (3.1)) is redefined as follows:

PðfXgÞ ¼ ð1 þ e1 : tÞe2  WðfXgÞ ð3:2Þ

where P({X}) is the penalized cost function or the objective function to be mini-
mized and t denotes the sum of the violations of the design constraints. Here, e1 is
set to unity, and e2 is calculated by

iter
e2 ¼ 1:5 þ 1:5  ð3:3Þ
itermax

where iter is the current iteration number, and itermax is the total number of itera-
tions for optimization process [3].

3.2 Numerical Examples with Frequency Constraints

3.2.1 A 72-Bar Space Truss Problem

The schematic of a 72-bar space truss is shown in Fig. 3.1 as the first design
example. The elements are divided into 16 groups, considering the symmetry. The
material density is 2767.99 kg/m3 and the elastic modulus is 68.95 GPa for all the
members. Four nonstructural masses of 2268 kg are attached to the nodes 1 through
4. The allowable minimum cross-sectional area of all elements is set to 0.645 cm2.
This example has two frequency constraints. The first frequency is required to be
f1 = 4 Hz and the third frequency is required to satisfy f3  6 Hz.
The optimized designs found by CSS-BBBC (hybridization of Charged System
Search and Big Bang-Big Crunch with trap recognition capability) [4], HALC-PSO
(transplants a Harmony search-based mechanism to Particle Swarm Optimization
with an Aging Leader and Challengers) [5], TWO (Tug of War Optimization) [6],
CBO [7], EBO [7], and VPS [8] are compared in Table 3.1. The CSS-BBBC
obtained the lightest design; however, the best designs of all methods are
approximately identical. Besides, the elastic modulus of 69.8 GPa was used in [4]
that generally results in lighter structures. The average optimized weight and the
standard deviation on average weight of the VPS are less than those of all other
methods. Frequency constraints are satisfied by all methods (see Table 3.2).
The VPS requires 4720 structural analyses to find the optimum solution while
HALC-PSO, CBO, and ECBO require 8000, 4000, and 14,800 structural analyses,
respectively.
3.2 Numerical Examples with Frequency Constraints 25

Fig. 3.1 Schematic of a spatial 72-bar truss

3.2.2 A Spatial 120-Bar Dome-Shaped Truss Problem

Figure 3.2 shows the schematic of a 120-bar dome truss. The members are cate-
gorized into seven groups considering the symmetry. The material density is
7971.810 kg/m3 and the modulus of elasticity is 210 GPa for all elements.
Nonstructural masses are attached to all free nodes as follows: 3000 kg at node 1,
500 kg at nodes 2–13, and 100 kg at the remaining nodes. Element cross-sectional
areas can vary between 1 and 129.3 cm2. The frequency constraints are as

f1  9 Hz and f2  11 Hz

The comparison of the outcomes of different algorithms can be seen in

Table 3.3. The VPS [8] yields the least weight. The best weight of this method is
8888.74 kg while it is 9046.34 kg for CSS-BBBC [4], 8889.96 kg for the
HALC-PSO [5], 8890.69 kg for the CBO [7] and 8896.50 kg for the ECBO [7].
Moreover, it can be seen that the lightest average optimized weight is found by the
VPS technique. Table 3.4 reports the natural frequencies of the optimized
26 3 Optimum Design of Usual-Size Skeletal Structures

Table 3.1 Performance comparison of the 72-bar space truss problem

Design Members Areas (cm2)
variable in the CSS-BBBC HALC-PSO TWO CBO [7] ECBO VPS [8]
group [4] [5] [6] [7]
1 1–4 2.854 3.3437 3.380 3.7336 3.5498 3.5017
2 5–12 8.301 7.8688 8.086 7.9355 7.8356 7.9340
3 13–16 0.645 0.6450 0.647 0.6450 0.645 0.6450
4 17–18 0.645 0.6450 0.646 0.6450 0.645 0.6450
5 19–22 8.202 8.1626 8.890 8.3765 8.1183 8.0215
6 23–30 7.043 7.9502 8.136 8.0889 8.1338 7.9826
7 31–34 0.645 0.6452 0.654 0.6450 0.645 0.6450
8 35–36 0.645 0.6450 0.647 0.6450 0.6450 0.6450
9 37–40 16.328 12.2668 13.097 12.9491 12.6231 12.8175
10 41–48 8.299 8.1845 8.101 8.0524 8.0971 8.1129
11 49–52 0.645 0.6451 0.663 0.6450 0.6450 0.6450
12 53–54 0.645 0.6451 0.646 0.6450 0.645 0.6450
13 55–58 15.048 17.9632 16.483 16.6629 17.3908 17.3362
14 59–66 8.268 8.1292 7.873 8.0557 8.0634 8.1010
15 67–70 0.645 0.6450 0.651 0.645 0.645 0.6450
16 71–72 0.645 0.6450 0.657 0.645 0.645 0.6450
Weight (kg) 327.507 327.77 328.83 327.740 327.653 327.649
Average N/A 327.99 336.1 328.20 327.76 327.670
optimized
weight (kg)
Standard N/A 0.19 5.8 0.54 0.06 0.018
deviation on
average
weight (kg)

Table 3.2 Natural frequencies (Hz) evaluated at the optimum designs of the 72-bar space truss
problem
Frequency Natural frequencies (Hz)
number CSS-BBBC HALC-PSO TWO CBO ECBO VPS
[4] [5] [6] [7] [7] [8]
1 4.000 4.000 4.000 4.000 4.000 4.0000
2 4.000 4.000 4.000 4.000 4.000 4.0002
3 6.004 6.000 6.000 6.000 6.000 6.0000
4 6.2491 6.230 6.259 6.267 6.246 6.2428
5 8.9726 9.041 9.082 9.101 9.071 9.0698

structures, and it is clear that none of the frequency constraints are violated. The
HALC-PSO, CBO, ECBO, and VPS algorithms obtain the optimal solution after
17,000, 3,700, 7700, and 6860 analyses, respectively.
3.2 Numerical Examples with Frequency Constraints 27

Fig. 3.2 Schematic of the spatial 120-bar dome-shaped truss

28 3 Optimum Design of Usual-Size Skeletal Structures

Table 3.3 Performance comparison of the spatial 120-bar dome-shaped truss problem
Design variable Areas (cm2)
CSS-BBBC HALC-PSO CBO [7] ECBO [7] VPS [8]
[4] [5]
1 17.478 19.8905 19.7738 19.8290 19.6836
2 49.076 40.4045 40.6757 41.4037 40.9581
3 12.365 11.2057 11.6056 11.0055 11.3325
4 21.979 21.3768 21.4601 21.2971 21.5387
5 11.190 9.8669 9.8104 9.4718 9.8867
6 12.590 12.7200 12.2866 13.0176 12.7116
7 13.585 15.2236 15.1417 15.2840 14.9330
Weight (kg) 9046.34 8889.96 8890.69 8896.50 8888.74
Average N/A 8900.39 8945.64 8920.16 8896.04
optimized
weight (kg)
Standard N/A 6.38 38.33 20.12 6.65
deviation on
average weight
(kg)

Table 3.4 Natural frequencies (Hz) evaluated at the optimum designs of the spatial 120-bar
dome-shaped truss problem
Frequency number Natural frequencies (Hz)
CSS-BBBC [4] HALC-PSO [5] CBO [7] ECBO [7] VPS [8]
1 9.000 9.000 9.000 9.001 9.0000
2 11.007 11.000 11.000 11.001 11.0000
3 11.018 11.000 11.000 11.003 11.0000
4 11.026 11.010 11.010 11.010 11.0096
5 11.048 11.050 11.049 11.052 11.0491

3.2.3 A 200-Bar Planar Truss Problem

The last structural optimization problem solved in this class is the optimal design of
a 200-bar planar truss schematized in Fig. 3.3. Due to the symmetry, the elements
are divided into 29 groups. The modulus of elasticity and the material density of
members are 210 GPa and 7860 kg/m3, respectively. Nonstructural masses of
100 kg are attached to the upper nodes. A lower bound of 0.1 cm2 is assumed for
the cross-sectional areas. The first three natural frequencies of the structure must
satisfy the following limitations (f1  5 Hz, f2  10 Hz, f3  15 Hz).
Table 3.5 presents the results of the optimal designs utilizing CSS-BBBC [4],
HALC-PSO [5], CBO-PSO (a hybrid of CBO and PSO algorithms) [9], CBO [7],
ECBO [7], and VPS [3]. The weight of the best result obtained by VPS is
3.2 Numerical Examples with Frequency Constraints 29

Fig. 3.3 Schematic of the 200-bar planar truss

2156.62 kg that is the best among the compared methods. The average optimized
weight of the HALC-PSO is 2157.14 kg, which is less than those of all other
methods. Table 3.6 reports the natural frequencies of the optimized structures, and
Table 3.5 Performance comparison of the 200-bar planar truss problem
30

Element group Members in the group Areas (cm2)

CSS-BBBC HALC-PSO CBO-PSO CBO [7] ECBO [7] VPS [3]
[4] [5] [9]
1 1,2,3,4 0.2934 0.3072 0.2797 0.3059 0.2993 0.3031
2 5,8,11,14,17 0.5561 0.4545 0.6968 0.4476 0.4497 0.4496
3 19,20,21,22,23,24 0.2952 0.1000 0.1000 0.1000 0.1000 0.1002
4 18,25,56,63,94,101,132,139,170,177 0.1970 0.1000 0.1000 0.1001 0.1000 0.1000
5 26,29,32,35,38 0.8340 0.5080 0.5796 0.4944 0.5137 0.5086
6 6,7,9,10,12,13,15,16,27,28,30,31,33, 0.6455 0.8276 0.8213 0.8369 0.7914 0.8204
34,36,37
7 39,40,41,42 0.1770 0.1023 0.1279 0.1001 0.1013 0.1000
8 43,46,49,52,55 1.4796 1.4357 1.0152 1.5514 1.4129 1.4210
9 57,58,59,60,61,62 0.4497 0.1007 0.1000 0.1000 0.1019 0.1002
10 64,67,70,73,76 1.4556 1.5528 1.5647 1.5286 1.6460 1.5900
11 44,45,47,48,50,51,53,54,65,66,68,69, 1.2238 1.1529 1.6465 1.1547 1.1532 1.1530
71,72,74,75
12 77,78,79,80 0.2739 0.1522 0.2296 0.1000 0.1000 0.1277
13 81,84,87,90,93 1.9174 2.9564 2.9007 2.9980 3.1850 2.9160
14 95,96,97,98,99,100 0.1170 0.1003 0.1000 0.1017 0.1034 0.1009
15 102,105,108,111,114 3.5535 3.2242 3.0133 3.2475 3.3126 3.2826
16 82,83,85,86,88,89,91,92,103,104,106, 1.3360 1.5839 1.6142 1.5213 1.5920 1.5856
107,109,110,112,113
17 115,116,117,118 0.6289 0.2818 0.2755 0.3996 0.2238 0.2794
18 119,122,125,128,131 4.8335 5.0696 5.0951 4.7557 5.1227 5.0680
19 133,134,135,136,137,138 0.6062 0.1033 0.1000 0.1002 0.1050 0.1004
(continued)
3 Optimum Design of Usual-Size Skeletal Structures
Table 3.5 (continued)
Element group Members in the group Areas (cm2)
CSS-BBBC HALC-PSO CBO-PSO CBO [7] ECBO [7] VPS [3]
[4] [5] [9]
20 140,143,146,149,152 5.4393 5.4657 5.5172 5.1359 5.3707 5.4760
21 120,121,123,124,126,127,129,130,141, 1.8435 2.0975 2.2032 2.1181 2.0645 2.1169
142,144,145,147,148,150,151
22 153,154,155,156 0.8955 0.6598 0.8659 0.9200 0.5443 0.6939
23 157,160,163,166,169 8.1759 7.6585 7.6477 7.3084 7.6497 7.6912
24 171,172,173,174,175,176 0.3209 0.1444 0.1000 0.1185 0.1000 0.1332
25 178,181,184,187,190 10.98 8.0520 8.1273 7.6901 7.6754 7.9972
26 158,159,161,162,164,165,167,168,179, 2.9489 2.7889 2.9665 3.0895 2.7178 2.7859
180,182,183,185,186,188,189
27 191,192,193,194 10.5243 10.4770 10.2386 10.6462 10.8141 10.4331
28 195,197,198,200 20.4271 21.3257 20.6364 20.7190 21.6349 21.2289
3.2 Numerical Examples with Frequency Constraints

29 196,199 19.0983 10.5111 11.6468 11.7463 10.3520 10.7392

Weight (kg) 2298.61 2156.73 2195.469 2161.15 2158.08 2156.62
Average optimized N/A 2157.14 N/A 2447.52 2159.93 2159.46
weight (kg)
Standard deviation on N/A 0.2413 N/A 301.29 1.57 2.79
average weight (kg)
31
32 3 Optimum Design of Usual-Size Skeletal Structures

Table 3.6 Natural frequencies (Hz) evaluated at the optimum designs of the 200-bar planar truss
problem
Frequency number Natural frequencies (Hz)
CSS-BBBC [4] HALC-PSO [5] CBO-PSO CBO [7] ECBO [7] VPS [3]
1 5.010 5.000 5.003 5.000 5.000 5.0000
2 12.911 12.254 12.281 12.221 12.189 12.2086
3 15.416 15.044 15.125 15.088 15.048 15.0153
4 17.033 16.718 16.613 16.759 16.643 16.6946
5 21.426 21.461 21.331 21.419 21.342 21.4046

Fig. 3.4 Convergence curves for the 200-bar planar truss problem

it is clear that none of the frequency constraints are violated. Comparison of the
convergence rates between CBO, ECBO, and VPS is illustrated in Fig. 3.4.
The VPS requires 16,420 structural analyses to find the optimum solution while
HALC-PSO, CBO-PSO, CBO, and ECBO require 13,000, 9000, 10,500, and
14,700 structural analyses, respectively. It should be noted that the designs found
by VPS at 9,000th, 10,500th, 13,000th, and 14,700th analyses are 2158.35,
2158.06, 2157.74, and 2157.72 kg, respectively.

3.3 Numerical Examples with Strength Constraints

3.3.1 A Spatial 120-Bar Dome-Shaped Truss Problem

The schematic and element grouping of a spatial 120-bar dome truss are shown in
Fig. 3.5. This structure is divided into seven groups of elements because of sym-
metry (for the sake of clarity, not all the element groups are numbered in Fig. 3.5).
3.3 Numerical Examples with Strength Constraints 33

Fig. 3.5 Schematic of the spatial 120-bar dome-shaped truss

The modulus of elasticity is 30,450 ksi (210 GPa) and the material density is
0.288 lb/in.3 (7971.810 kg/m3). The yield stress of steel is taken as 58.0 ksi
(400 MPa). The dome is considered to be subjected to vertical loading at all the
unsupported joints. These loads are taken as −13.49 kips (−60 kN) at node 1,
−6.744 kips (−30 kN) at nodes 2 through 14, and −2.248 kips (−10 kN) at the
remaining nodes. Element cross-sectional areas can vary between 0.775 in.2
34 3 Optimum Design of Usual-Size Skeletal Structures

(5 cm2) and 20.0 in.2 (129.032 cm2). Displacement limitations of ±0.1969 in.
(± 5 mm) are imposed on all nodes in x, y, and z coordinate directions. Constraints
on member stresses are imposed according to the provisions of the AISC [10] as
follows:
The allowable tensile stresses for tension members are calculated as

riþ ¼ 0:6Fy ð3:4Þ

where Fy is the yield strength.

The allowable stress limits for compression members are calculated depending
on two possible failure modes of the members known as elastic and inelastic
buckling. Therefore
8 h  i h i
< 1 k2i
Fy = 53 þ 3ki k3
 8Ci 3 for ki \Cc
r
2Cc2 8Cc
¼ c
ð3:5Þ
i
: 12p 2E
2
for ki  Cc
23ki

where E is the modulus of elasticity; ki is the slenderness ratio ðki ¼ kli =ri Þ; Cc
denotes the slenderness ratio dividing the elastic and inelastic buckling regions
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Cc ¼ 2p2 E=Fy ; k is the effective length factor (k is set equal to 1 for all truss
members); Li is the member length; and ri is the minimum radius of gyration.
This truss is optimized by MSPSO (Multistage Particle Swarm Optimization)
[11], TWO (Tug of War Optimization) [12], WEO (Water Evaporation
Optimization) [13], CBO, ECBO, and VPS [3]. Comparison of the optimal designs
is given in Table 3.7. It can be seen that the lightest design (i.e., 33,249.98 lb) and
the best average optimized weight (i.e., 33,253.56 lb) are found by the VPS method.
Figure 3.6 compares the convergence curves of the best results obtained by the
CBO, ECBO, and VPS. The VPS converges to the optimum solution after 8280
analyses. The MSPSO, TWO, WEO, CBO, and ECBO obtain the optimal solution
after 15,000, 16,000, 19,510, 12,080, and 19,800 analyses, respectively.

3.3.2 A 3-Bay 15-Story Frame Problem

The schematic of a 3-bay 15-story frame is represented in Fig. 3.7. The applied loads
and the numbering of the member groups are also shown in this figure. The modulus
of elasticity is 29 Msi (200 GPa) and the yield stress is 36 ksi (248.2 MPa).
The effective length factors of the members are calculated as kx  0 for a
sway-permitted frame and the out-of-plane effective length factor is specified as
ky= 1.0. Each column is considered as non-braced along its length, and the
non-braced length for each beam member is specified as one-fifth of the span length.
Limitation on displacement and strength are imposed according to the provisions of
the AISC [14] as follows:
Table 3.7 Performance comparison of the spatial 120-bar dome-shaped truss problem
Element group Optimal cross-sectional areas (in.2)
MSPSO [11] TWO [12] WEO [13] CBO ECBO VPS [3]
1 3.0244 3.0247 3.0243 3.0260 3.0234 3.0244
2 14.7804 14.7261 14.7943 14.8237 14.8569 14.7536
3 5.0567 5.1338 5.0618 5.1576 4.8649 5.0789
4 3.1359 3.1369 3.1358 3.1310 3.1319 3.1371
5 8.4830 8.4545 8.4870 8.3350 8.5716 8.4829
3.3 Numerical Examples with Strength Constraints

6 3.3104 3.2946 3.2886 3.3872 3.3967 3.3012

7 2.4977 2.4956 2.4967 2.4938 2.5042 2.4963
Weight (lb) 33,251.22 33,250.31 33,250.24 33,256.15 33,268.56 33,249.98
Average optimized weight (lb) 33,257.29 33,282.64 33,255.55 33,284.19 33,303.53 33,253.56
Standard deviation on average weight (lb) 4.29 25.38 N/A 31.40 26.958 4.36
35
36 3 Optimum Design of Usual-Size Skeletal Structures

Fig. 3.6 Convergence curves for the spatial 120-bar dome-shaped truss problem

(a) Maximum lateral displacement

DT
 R0 ð3:6Þ
H

where DT is the maximum lateral displacement, H is the height of the frame

structure, and R is the maximum drift index which is equal to 1/300.
(b) The inter-story displacements

di
 RI  0; i ¼ 1; 2; . . .; ns ð3:7Þ
hi

where di is the inter-story drift, hi is the story height of the ith floor, ns is the total
number of stories, and RI is the inter-story drift index (1/300).
(c) Strength constraints

(
2/c Pn þ /b Mn  1  0; for /PPu n \0:2
Pu Mu
c
ð3:8Þ
/c Pn þ 9/b Mn  1  0; for /PPu n  0:2
Pu 8Mu
c

where Pu is the required strength (tension or compression); Pn is the nominal axial

strength (tension or compression); /c is the resistance factor (/c = 0.9 for tension,
3.3 Numerical Examples with Strength Constraints 37

Fig. 3.7 Schematic of the 3-bay 15-story frame

38 3 Optimum Design of Usual-Size Skeletal Structures

/c = 0.85 for compression); Mu is the required flexural strengths; Mn is the nominal

flexural strengths; and /b denotes the flexural resistance reduction factor
(/b = 0.90).
The nominal tensile strength for yielding in the gross section is calculated by

Pn ¼ Ag : Fy ð3:9Þ

The nominal compressive strength of a member is computed as

Pn ¼ Ag : Fcr ð3:10Þ

where
8  
< Fcr ¼ 0:658k2c Fy ; for kc  1:5
  ð3:11Þ
: Fcr ¼ 0:877 Fy ; for kc [ 1:5
k2
c

rffiffiffiffiffi
kl Fy
kc ¼ ð3:12Þ
rp E

where Ag is the cross-sectional area of a member, and k is the effective length factor
that is calculated by Dumonteil [15]:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1:6GA GB þ 4:0ðGA þ GB Þ þ 7:5
k¼ ð3:13Þ
GA þ GB þ 7:5

where GA and GB are stiffness ratios of columns and girders at the two end joints A
and B of the column section, respectively.
Additionally, in this example, the sway of the top story is limited to 9.25 in.
(23.5 cm).
Table 3.8 presents the comparison of the results of different algorithms.
The VPS algorithm yields the least weight for this example, which is 86,985 lb. The
other design weights are 92,723 lb by CSS [16], 93,315 lb by ES-DE (Eagle
Strategy with Differential Evolution) [17], 91,248 lb by DSOS (Discrete Symbiotic
Organisms Search) [18], 93,795 lb by CBO [19] and 86,986 lb by ECBO [19]. The
best design of VPS has been achieved in 19,600 analyses. The CSS, ES-DE, CBO,
and ECBO require 5,000, 10,000, 9520, and 9000 structural analyses to find the
optimum solutions, respectively. It should be noted that the proposed method
achieved about 92,000 lb (the best weight among the other methods except ECBO)
after 10,800 structural analyses. Element stress ratio and inter-story drift evaluated
at the best design optimized by VPS are shown in Fig. 3.8. The maximum stress
ratio is 99.88% and the maximum inter-story drift is 45.41.
3.3 Numerical Examples with Strength Constraints 39

Table 3.8 Performance comparison of the 3-bay 15-story frame problem

Element Optimal W-shaped sections
group CSS [16] ES-DE [17] DSOS [18] CBO [19] ECBO [19] VPS [3]
1 W21  147 W18  106 W16  100 W24  104 W14  99 W14  90
2 W18  143 W36  150 W32  152 W40  167 W27  161 W36  170
3 W12  87 W12  79 W12  79 W27  84 W27  84 W14  82
4 W30  108 W27  114 W27  114 W27  114 W24  104 W24  104
5 W18  76 W30  90 W21  93 W21  68 W14  61 W21  68
6 W24  103 W10  88 W12  79 W30  90 W30  90 W18  86
7 W21  68 W18  71 W21  55 W8  48 W14  48 W21  48
8 W14  61 W18  65 W14  61 W21  68 W14  61 W14  61
9 W18  35 W8  28 W14  22 W14  34 W14  30 W12  30
10 W10  33 W12  40 W14  43 W8  35 W12  40 W10  39
11 W21  44 W21  48 W21  48 W21  50 W21  44 W21  44
Weight (lb) 92,723 93,315 91,248 93,795 86,986 86,985
Average N/A 98,531 N/A 98,738 88,410 90,066
optimized
weight (lb)
Standard N/A 3294 N/A N/A N/A 2533
deviation on
average
weight (lb)

Fig. 3.8 Constraint margins for the best design obtained by VPS for the 3-bay 15-story frame
problem: a element stress ratio and b inter-story drift
40 3 Optimum Design of Usual-Size Skeletal Structures

3.3.3 A 3-Bay 24-Story Frame Problem

The last structural optimization problem solved in this chapter is the weight min-
imization of a 3-bay 24-story frame shown in Fig. 3.9. Frame members are col-
lected in 20 groups (16 column groups and 4 beam groups). Each of the four beam
element groups is chosen from all 267 W shapes, while the 16 column element
groups are limited to W14 sections. The material has a modulus of elasticity equal
to E = 29.732 Msi (205 GPa) and a yield stress of fy = 33.4 ksi (230.3 MPa). The
effective length factors of the members are calculated as kx  0 for a
sway-permitted frame and the out-of-plane effective length factor is specified as
ky = 1.0. All columns and beams are considered as non-braced along their lengths.
Similar to the previous example, the frame is designed following the LRFD-AISC
specification and uses an inter-story drift displacement constraint (AISC [14]).
This steel frame structure was optimized by CSS [16], ES-DE [17] and DSOS
[18], CBO [19], ECBO [19], and VPS [3]. Table 3.9 presents a comparison between
these results. The lightest design (i.e., 201,618 lb) is found by ECBO algorithm and
after that, the best design belongs to VPS (i.e., 202,998 lb). Figure 3.10 shows the
convergence curves of the best results found by CBO, ECBO, and VPS. The best
design has been achieved at 16,220 analyses by VPS and it obtained a weight of
209,532 lb after 8800 analyses, which is the best result compared to the
weight achieved by the other method. The CSS, ES-DE, DSOS, CBO, and ECBO
get the optimal solution after 5500, 12,500, 7500, 8280, and 15,360 analyses,
respectively.

3.4 Concluding Remarks

In this chapter, the CBO, ECBO, and VPS algorithms are examined in the context
of size optimization of skeletal structure designed for minimum weight. Four
trusses and two frames subjected to different constraints are employed, and their
final results are compared with results of the state-of-the-art algorithms from lit-
erature. The VPS achieved the lightest designs for most of the benchmarks and after
that, ECBO performed better among the compared methods. The small values of the
standard deviation on average weights prove the robustness of these techniques.
Moreover, they show appropriate convergence rates. Although the results found by
CBO are not better than ECBO and VPS but it has no internal parameter to adjust.
Therefore, it can be performed more easily on various problems. Generally, com-
parison of the results proves the efficiency of the proposed algorithms and shows
that they are suitable methods for comparison with the MDVC-UVPS.
3.4 Concluding Remarks 41

Fig. 3.9 Schematic of the 3-bay 24-story frame

42 3 Optimum Design of Usual-Size Skeletal Structures

Table 3.9 Performance comparison of the 3-bay 24-story frame problem

Element Optimal W-shaped sections
group CSS [16] ES-DE [17] DSOS [18] CBO [19] ECBO [19] VPS [3]
1 W30  90 W30  90 W30  90 W27  102 W30  90 W30  90
2 W21  50 W21  55 W21  62 W8  18 W6  15 W8  18
3 W21  48 W21  48 W21  48 W24  55 W24  55 W21  48
4 W12  19 W10  45 W21  55 W6  8.5 W6  8.5 W6  8.5
5 W14  176 W14  145 W14  176 W14  132 W14  145 W14  176
6 W14  145 W14  109 W14  109 W14  120 W14  132 W14  145
7 W14  109 W14  99 W14  120 W14  145 W14  99 W14  99
8 W14  90 W14  145 W14  82 W14  82 W14  90 W14  82
9 W14  74 W14  109 W14  61 W14  61 W14  74 W14  82
10 W14  61 W14  48 W14  99 W14  43 W14  38 W14  38
11 W14  34 W14  38 W14  34 W14  38 W14  38 W14  30
12 W14  34 W14  30 W14  38 W14  22 W14  22 W14  30
13 W14  145 W14  99 W14  120 W14  99 W14  99 W14  90
14 W14  132 W14  132 W14  109 W14  109 W14  99 W14  99
15 W14  109 W14  109 W14  90 W14  82 W14  99 W14  99
16 W14  82 W14  68 W14  90 W14  90 W14  82 W14  90
17 W14  68 W14  68 W14  82 W14  74 W14  68 W14  61
18 W14  43 W14  68 W14  38 W14  61 W14  61 W14  61
19 W14  34 W14  61 W14  38 W14  30 W14  30 W14  34
20 W14  22 W14  22 W14  22 W14  22 W14  22 W14  26
Weight 212,364 212,492 209,795 215,874 201,618 202,998
(lb)
Average 215,226 N/A N/A 225,071 209,644 212,289
optimized
weight (lb)
Standard 2448 N/A N/A N/A N/A 8292
deviation
on average
weight (lb)
References 43

Fig. 3.10 Convergence curves for the 3-bay 24-story frame problem

References

1. Kaveh A (2017) Advances in metaheuristic algorithms for optimal design of structures, 2nd
edn. Springer, Switzerland
2. Kaveh A (2017) Applications of metaheuristic optimization algorithms in civil engineering.
Springer, Switzerland
3. Kaveh A, Ilchi Ghazaan M (2017) A new meta-heuristic algorithm: vibrating particles system.
Sci Iranica, Trans A, Civil Eng 24(2):551–566
4. Kaveh A, Zolghadr A (2012) Truss optimization with natural frequency constraints using a
hybridized CSS–BBBC algorithm with trap recognition capability. Comput Struct
102–103:14–27
5. Kaveh A, Ilchi Ghazaan M (2015) Hybridized optimization algorithms for design of trusses
with multiple natural frequency constraints. Adv Eng Softw 79:137–147
6. Kaveh A, Zolghadr A (2017) Truss shape and size optimization with frequency constraints
using Tug of war optimization. Asian J Civil Eng 7(2):311–333
7. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies algorithm for truss optimization
with frequency constraints. J Comput Civil Eng 29(6)
8. Kaveh A, Ilchi Ghazaan M (2017) Vibrating particles system algorithm for truss optimization
with multiple natural frequency constraints. Acta Mech 228(1):307–322
9. Kaveh A, Mahdavi VR (2015) A hybrid CBO–PSO algorithm for optimal design of truss
structures with dynamic constraints. Appl Soft Comput 34:260–273
10. American Institute of Steel Construction (AISC) (1989) Manual of steel construction:
allowable stress design
11. Talatahari S, Kheirollahi M, Farahmandpour C, Gandomi AH (2013) A multi-stage particle
swarm for optimum design of truss structures. Neural Comput Appl 23:1297–1309
12. Kaveh A, Zolghadr A (2016) A novel metaheuristic algorithm: tug of war optimization. Int J
Optim Civil Eng 6(4):469–492
13. Kaveh A, Bakhshpoori T (2016) Water evaporation optimization: a novel physically inspired
optimization algorithm. Comput Struct 167:69–85
14. American Institute of Steel Construction (AISC) (2001) Manual of steel construction: load
and resistance factor design
15. Dumonteil P (1992) Simple equations for effective length factors. Eng J AISC 29(3):111–115
44 3 Optimum Design of Usual-Size Skeletal Structures

16. Kaveh A, Talatahari S (2012) Charged system search for optimal design of frame structures.
Appl Soft Comput 12:382–393
17. Talatahari S, Gandomi AH, Yang XS, Deb S (2015) Optimum design of frame structures
using the eagle strategy with differential evolution. Eng Struct 91:16–25
18. Talatahari S (2016) Symbiotic organisms search for optimum design of and grillage system.
Asian J Civil Eng 17(3):299–313
19. Kaveh A, Ilchi Ghazaan M (2015) A comparative study of CBO and ECBO for optimal
design of skeletal structures. Comput Struct 153:137–147
Chapter 4
Optimal Design of Large-Scale Special
Truss Structures

4.1 Introduction

A truss is a two- or three-dimensional structure composed of linear members

connected at nodes to sustain concentrated loads with the members being subjected
to tension or compression. Optimum design problems of steel trusses are known as
benchmarks in the field of structural optimization due to the presence of many
design variables, large search spaces, and multiple constraints. Truss optimization
problems can be classified in three categories: (1) size optimization (obtaining the
optimal cross sections of the structural members); (2) shape optimization (finding
the optimal form of the structure); and (3) topology optimization (achieving the
optimal size and connectivity of the structural members). Truss structures can be
considered as a suitable means to investigate the efficiency of optimization algo-
rithms. This branch of structural optimization has been extensively developed in the
last three decades [1–16].
In this chapter, sizing optimization of large-scale tower trusses is studied. Steel
truss members are adopted from a predetermined list of available sections; there-
fore, a discrete optimization is performed in order to obtain the optimum or a near
optimum solution. These types of structures are typically considered as high-rise
and large-scale structures composed of several hundred elements. These towers
have important applications in telecommunication and broadcasting industries.

4.2 Optimum Design Problem of Steel Trusses

The aim of optimizing a structure is generally to find a set of design variables

corresponding to the minimum weight structure while satisfying certain constraints.
Truss structures are often designed to carry multiple loading conditions under static
constraints on the nodal displacements, stresses in the members, and critical

© Springer International Publishing AG, part of Springer Nature 2018 45

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_4
46 4 Optimum Design of Large-Scale Special Truss Structures

buckling loads. The mathematical formulation of a structural optimization problem

can be expressed as follows:


Find f X g ¼ x1 ; x2 ; ::; xng
X
ng nmðiÞ
X
to minimize W ð f X gÞ ¼ xi qj Lj
i¼1 j¼1 ð4:1Þ
(
gj ðfX gÞ  0; j ¼ 1; 2; . . .; nc
subjected to :
ximin  xi  ximax

where {X} is a vector containing the design variables; W({X}) presents the weight
of the structure; ng is the number of design groups; nm(i) is the number of members
for the ith group; and qj and Lj denote the material density and the length of the jth
member, respectively. ximin and ximax are the lower and upper bounds of the design
variable xi, respectively. gj({X}) denotes the set of design constraints, and nc is the
number of constraints.
For constraint handling, a penalty approach is utilized. For this purpose, the
objective function (Eq. 4.1) is redefined as follows:

Pðf X gÞ ¼ ð1 þ e1 :tÞe2  W ðf X gÞ ð4:2Þ

where P({X}) is the penalized cost function or the objective function to be mini-
mized and t denotes the sum of the violations of the design constraints. Here, e1 is
set to unity, and e2 is calculated by
iter
e2 ¼ 1:5 þ 1:5  ð4:3Þ
itermax

where iter is the current iteration number, and itermax is the total number of iter-
ations for the optimization process.
The constraint conditions for truss structures studied here are briefly explained in
the following. Limitations on stress and stability of truss elements are imposed
according to the provisions of the ASD-AISC [17] as follows.
The allowable tensile stress for tension members are calculated by

riþ ¼ 0:6Fy ð4:4Þ

where Fy stands for the yield strength.

The allowable stress limits for compression members are calculated depending
on two possible failure modes of the members known as elastic and inelastic
buckling. Thus
8 h  i  
< 1  ki 2 Fy = 5 þ 3ki  ki 3
2 3

 2Cc 3 8Cc 8Cc for ki \Cc

ri ¼ 12p2 E ð4:5Þ
: 2 for ki  Cc
23ki
4.2 Optimum Design Problem of Steel Trusses 47

where E is the modulus of elasticity; ki is the slenderness ratio ðki ¼ kli =ri Þ; Cc
denotes the slenderness ratio dividing the elastic and inelastic buckling regions
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Cc ¼ 2p2 E=Fy ; k is the effective length factor (k is set to 1 for all truss
members); li is the member length; and ri is the minimum radius of gyration.
According to the provisions of ASD-AISC, the maximum slenderness ratios are
limited to 300 and 200 for tension and compression members, respectively. Nodal
displacements in all coordinate directions must be less than ±3.15 in.
(i.e., ±8 cm).

4.3 Design Examples

In this section, three tower truss optimization benchmark problems are optimized
using CBO, ECBO, VPS, and MDVC-UVPS algorithms. These optimization
examples are as follows:
• A spatial 582-bar tower truss
• A spatial 942-bar tower truss
• A spatial 2386-bar tower truss.

4.3.1 A Spatial 582-Bar Tower Truss

The schematic of the 582-bar tower truss is shown in Fig. 4.1. The members of the
truss are grouped into 32 independent sizing variables considering its symmetry
about x- and y-axes. A single load case is considered consisting of lateral loads of
1.12 kips (5.0 kN) applied in both x- and y-directions and vertical loads of
−6.74 kips (−30 kN) applied in z-direction to all free nodes of the tower. A discrete
set of standard steel sections selected from the list of W-shape profiles is considered
for the sizing variables based on the cross-sectional areas and radii of gyration.
Cross-sectional areas of the elements can vary between 6.16 and 215 in.2 (i.e.,
between 39.74 and 1387.09 cm2). Limitations on stress and stability of truss ele-
ments and nodal displacements are defined in Sect. 4.2.
This problem is optimized in three stages by MDVC-UVPS. The number of
design variables in stages 1, 2, and 3 is 8, 15, and 32, respectively. Table 4.1
presents the design variables configurations. Table 4.2 presents the results obtained
by CBO [18], ECBO [18], VPS, and MDVC-UVPS. The best design obtained by
MDVC-UVPS is better than those of the other methods (1,295,038 in.3). The best
volumes found by CBO, ECBO, and VPS are 1,334,994, 1,296,776, and
1,304,569 in.3, respectively. MDVC-UVPS is the most robust optimizer, achieving
the lowest average volume over the independent optimization runs. The CBO,
48 4 Optimum Design of Large-Scale Special Truss Structures

Fig. 4.1 Schematic of the 582-bar tower truss

Table 4.1 Design variable configurations utilized for the 582-bar tower problem
Number of design Design variables in the group (design variable
variables in stages configurations)
Stage 1 8 [1 6 9]; [2 4 7 10]; [3 5 8 11]; [12 13 14]; [19 22 25 28
31]; [32]; [15 17 20 23 26 29]; [16 18 21 24 27 30]
Stage 2 15 [1 6 9]; [2 4]; [7 10]; [3 5]; [8 11]; [12]; [13]; [14]; [19 22
25]; [28 31]; [32]; [15 17 20]; [23 26 29]; [16 18 21]; [24
27 30]

ECBO, VPS, and MDVC-UVPS algorithms obtained the optimal solutions after
17,700, 19,700, 17,540, and 15,480 analyses, respectively. Stress ratios and nodal
displacements in all directions evaluated for the best design optimized by
MDVC-UVPS are shown in Fig. 4.2. The maximum stress ratio and the maximum
nodal displacement obtained by MDVC-UVPS are 99.88% and 3.1493 in.
respectively.
4.3 Design Examples 49

Table 4.2 Performance comparison for the spatial 582-bar tower truss problem
Element group Optimal W-shaped sections
CBO [18] ECBO [18] VPS MDVC-UVPS
1 W8  21 W8  21 W8  21 W8  21
2 W14  82 W14  90 W14  90 W14  90
3 W8  28 W8  24 W8  24 W8  24
4 W12  50 W14  61 W21  62 W14  61
5 W8  24 W8  24 W8  24 W8  24
6 W8  21 W8  21 W8  21 W8  21
7 W12  53 W10  49 W10  39 W10  45
8 W12  26 W8  24 W8  24 W8  24
9 W8  21 W8  21 W8  24 W8  21
10 W14  48 W14  43 W10  33 W14  43
11 W8  24 W8  24 W8  24 W8  24
12 W14  61 W12  72 W14  74 W10  68
13 W14  82 W12  72 W10  77 W12  72
14 W12  50 W10  54 W10  49 W10  49
15 W14  74 W12  65 W10  77 W14  82
16 W8  40 W8  31 W8  31 W8  31
17 W12  53 W10  60 W21  62 W21  62
18 W6  25 W8  24 W8  24 W8  24
19 W8  21 W8  21 W10  22 W8  21
20 W8  40 W14  43 W10  49 W14  38
21 W8  24 W8  24 W8  24 W8  24
22 W8  21 W8  21 W8  21 W8  21
23 W12  26 W8  21 W10  22 W6  25
24 W12  26 W8  24 W8  24 W8  24
25 W10  22 W8  21 W8  21 W8  21
26 W10  22 W8  21 W8  21 W8  21
27 W6  25 W8  24 W8  24 W8  24
28 W8  21 W8  21 W8  21 W8  21
29 W8  21 W8  21 W10  22 W8  21
30 W8  24 W8  24 W8  24 W8  24
31 W8  21 W8  21 W10  22 W8  21
32 W6  25 W8  24 W8  24 W8  24
Volume (in.3) 1,334,994 1,296,776 1,304,569 1,295,038
Average optimized volume 1,345,429 1,306,728 1,324,086 1,302,422
(in.3)
Standard deviation on 9116 7536 12,218 4347
average volume (in.3)
50 4 Optimum Design of Large-Scale Special Truss Structures

Fig. 4.2 Constraint margins for the best design obtained by MDVC-UVPS algorithm for the
582-bar tower truss problem: a element stress ratio and b nodal displacements

4.3.2 A Spatial 942-Bar Tower Truss

Figure 4.3 shows the schematic of a 942-bar tower truss. This example has been
investigated by many researchers considering 59 design variables [19]. In this
study, the design variables are increased to 76, and the performance constraints,
material properties, and other conditions are the same as those of the first example.
Figure 4.4 shows the member groups. Three stages with 16, 28, and 76 design
variables are considered to solve this problem using MDVC-UVPS. The design
variable configurations are shown in Table 4.3.
The optimized designs found by different algorithms are compared in Table 4.4.
The volume of the best result obtained by the hybrid algorithm is 3,263,387 in.3
that is the best among the compared methods. The average optimized volume and
the standard deviation on average volume obtained by MDVC-UVPS are less than
those of all other considered methods. This algorithm requires 14,587 structural
analyses to find the optimum solution, while CBO [20], ECBO [20], and VPS [21]
require 29,600, 19,960, and 26,180 structural analyses, respectively. The amount of
saving in structural analyses in each iteration of the MDVC-UVPS is shown in
Fig. 4.5.

4.3.3 A Spatial 2386-Bar Tower Truss

The schematic of a 2386-bar tower truss is shown in Fig. 4.6. This example is
studied here for the first time. The performance constraints, material properties, and
other conditions are the same as those of the first example. The elements are divided
4.3 Design Examples 51

Fig. 4.3 Schematic of the 942-bar tower truss

Fig. 4.4 Member groups of 942-bar tower truss

into 220 groups and member groups are presented in Fig. 4.7. Four stages are
considered to optimize this example using MDVC-UVPS. The number of design
variable in stages 1, 2, 3, and 4 is 21, 42, 84, and 220, respectively. Table 4.5 lists
the design variable configurations.
Table 4.6 presents the optimum designs obtained by the proposed algorithms.
The lightest design (i.e., 12,165,572 in.3) is achieved by MDVC-UVPS algorithm
52 4 Optimum Design of Large-Scale Special Truss Structures

Table 4.3 Design variable configurations utilized for the 942-bar tower problem
Number of design Design variables in the group (design variable
variables in stages configurations)
Stage 16 [1]; [2–6]; [7–12]; [13–18]; [19–24]; [25–29]; [30–35];
1 [36]; [37–43]; [44–51]; [52–58]; [59]; [60–64]; [65–70];
[71–75]; [76]
Stage 28 [1]; [2 3]; [4–6]; [7–9]; [10–12]; [13–15]; [16–18]; [19–
2 21]; [22–24]; [25 26]; [27–29]; [30–32]; [33–35]; [36];
[37–39]; [40–43]; [44–47]; [48–51]; [52–54]; [55–58];
[59]; [60 61]; [62–64]; [65–67]; [68–70]; [71 72]; [73–
75]; [76]

Table 4.4 Performance comparison for the spatial 942-bar tower truss problem
Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
1 W14  145 W12  190 W12  170 W12  170
2 W36  280 W36  230 W36  260 W14  257
3 W24  250 W40  199 W44  262 W24  279
4 W14  257 W24  229 W30  235 W33  241
5 W33  241 W36  150 W36  245 W40  249
6 W44  262 W30  173 W24  229 W14  211
7 W30  211 W24  250 W40  199 W24  192
8 W33  201 W27  258 W14  193 W24  192
9 W24  176 W14  159 W40  174 W14  176
10 W24  162 W30  191 W24  162 W40  174
11 W21  147 W18  158 W14  145 W12  136
12 W12  136 W18  119 W18  119 W24  131
13 W33  221 W24  250 W12  279 W36  230
14 W6  25 W14  30 W8  21 W8  24
15 W10  54 W8  21 W10  22 W8  21
16 W8  21 W8  21 W12  26 W10  22
17 W10  22 W8  21 W10  22 W8  21
18 W8  21 W8  21 W10  22 W10  22
19 W10  22 W8  21 W10  22 W8  21
20 W8  21 W8  21 W10  22 W10  22
21 W8  21 W8  21 W6  25 W8  21
22 W10  22 W8  21 W8  24 W8  21
23 W10  22 W8  21 W10  22 W8  21
24 W14  145 W24  117 W14  145 W21  147
25 W14  34 W12  50 W8  31 W12  30
26 W8  24 W14  30 W8  24 W6  25
27 W8  24 W10  33 W8  24 W6  25
(continued)
4.3 Design Examples 53

Table 4.4 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
28 W6  25 W6  25 W8  24 W8  24
29 W12  79 W8  31 W12  26 W8  24
30 W10  22 W8  31 W10  22 W10  22
31 W6  25 W8  21 W8  21 W10  22
32 W10  22 W12  26 W10  22 W10  22
33 W8  21 W8  21 W8  21 W10  22
34 W6  25 W8  21 W10  22 W10  22
35 W10  22 W8  21 W8  21 W10  22
36 W14  90 W18  86 W16  89 W12  87
37 W40  174 W30  191 W30  211 W12  170
38 W21  147 W30  116 W14  109 W36  150
39 W12  136 W27  178 W24  131 W21  147
40 W10  100 W24  131 W21  101 W24  104
41 W14  74 W18  86 W10  88 W16  100
42 W14  82 W10  88 W10  77 W21  93
43 W14  61 W21  62 W12  50 W21  62
44 W24  117 W12  136 W27  114 W18  119
45 W8  31 W8  21 W10  22 W8  21
46 W6  25 W8  21 W10  22 W10  22
47 W12  30 W8  21 W10  22 W8  21
48 W12  26 W8  21 W6  25 W6  25
49 W8  21 W8  21 W10  22 W10  22
50 W10  22 W8  21 W8  40 W8  21
51 W18  76 W27  94 W12  58 W21  62
52 W8  24 W10  22 W6  25 W8  24
53 W8  24 W6  25 W10  22 W8  24
54 W8  24 W8  21 W10  22 W10  22
55 W10  22 W8  21 W10  22 W8  21
56 W10  22 W8  21 W10  22 W8  21
57 W8  24 W8  21 W8  21 W8  21
58 W10  22 W8  21 W10  22 W8  21
59 W14  48 W21  62 W14  43 W12  50
60 W21  111 W12  152 W24  117 W12  96
61 W12  87 W14  120 W18  119 W21  93
62 W10  60 W12  65 W14  38 W14  48
63 W6  25 W14  30 W10  77 W14  38
64 W8  24 W8  21 W14  61 W8  24
65 W10  22 W8  21 W10  22 W8  21
66 W14  34 W8  21 W10  22 W8  21
(continued)
54 4 Optimum Design of Large-Scale Special Truss Structures

Table 4.4 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
67 W24  146 W8  21 W8  21 W8  21
68 W12  58 W8  21 W10  22 W10  22
69 W8  21 W8  21 W10  22 W12  26
70 W8  21 W8  21 W10  22 W10  22
71 W8  24 W8  24 W8  31 W8  21
72 W8  24 W8  24 W10  22 W8  21
73 W8  24 W8  21 W12  26 W10  22
74 W16  36 W8  21 W10  22 W10  22
75 W10  22 W8  21 W8  21 W8  21
76 W6  25 W8  21 W8  28 W14  30
Volume (in.3) 3,414,157 3,376,968 3,296,202 3,263,387
Average optimized 3,601,976 3,429,516 3,346,822 3,276,876
volume (in.3)
Standard deviation on 47,864 45,121 41,617 13,823
average volume (in.3)

Fig. 4.5 Saving in structural analyses using the MDVC-UVPS algorithm for the 942-bar tower
truss problem

after 13,385 analyses. The best design obtained by the CBO [20], ECBO [20], and
VPS [21] is 15,587,709 in.3, 14,086,857 in.3, and 12,989,713 in.3, respectively.
These values are found after 29,970, 29,670, and 29,980 analyses. MDVC-UVPS is
also the most robust optimizer, achieving the lowest average design over the
independent optimization runs. Convergence history diagrams are depicted in
4.3 Design Examples 55

Fig. 4.6 Schematic of the 2386-bar tower truss

Fig. 4.7 Member groups of the 2386-bar tower truss

Fig. 4.8 and demonstrate that the intermediate designs found by MDVC-UVPS are
always better than those found by the other considered algorithms.

4.4 Concluding Remarks

Truss structures are often designed to carry multiple loading conditions under static
constraints on the nodal displacements, stresses in the members, and critical
buckling loads. This class of problems is highly nonlinear and also widely studied
in the field of practical structural engineering. In this chapter, three tower trusses
with a large number of design variables are studied to test and verify the efficiency
56 4 Optimum Design of Large-Scale Special Truss Structures

Table 4.5 Design variable configurations utilized for the 2386-bar tower problem
Number of design Design variables in the group (design variable
variables in stages configurations)
Stage 21 [1–10]; [11–20]; [21–31]; [32–41]; [42–51]; [52–62];
1 [63–72]; [73–82]; [83–92] [93–103]; [104–113]; [114–
124]; [125–135]; [136–146]; [147–156]; [157–167] [168–
178]; [179–188]; [189–199]; [200–210]; [211–220]
Stage 42 [1:4]; [5:10]; [11:15]; [16:20]; [21:25]; [26:31]; [32:36];
2 [37:41]; [42:46]; [47:51]; [52:56]; [57:62]; [63:67];
[68:72]; [73:77]; [78:82]; [83:87]; [88:92]; [93:97];
[98:103]; [104:108]; [109:113]; [114:118]; [119:124];
[125:129]; [130:135]; [136:140]; [141:146]; [147:151];
[152:156]; [157:161]; [162:167]; [168:172]; [173:178];
[179:183]; [184:188]; [189:193]; [194:199]; [200:204];
[205:210]; [211:215]; [216:220]
Stage 84 [1]; [2–4]; [5–7]; [8–10]; [11 12]; [13–15]; [16 17]; [18–
3 20]; [21 22]; [23–25]; [26–28]; [29–31]; [32 33]; [34–36];
[37 38]; [39–41]; [42 43]; [44–46]; [47 48]; [49–51]; [52
53]; [54–56]; [57–59]; [60–62]; [63 64]; [65–67]; [68 69];
[70–72]; [73 74]; [75–77]; [78 79]; [80–82]; [83 84]; [85–
87]; [88 89]; [90–92]; [93 94]; [95–97]; [98–100]; [101–
103]; [104 105]; [106–108]; [109 110]; [111–113]; [114
115]; [116–118]; [119–121]; [122–124]; [125 126]; [127–
129]; [130–132]; [133–135]; [136 137]; [138–140]; [141–
143]; [144–146]; [147 148]; [149–151]; [152 153]; [154–
156]; [157 158]; [159–161]; [162–164]; [165–167]; [168
169]; [170–172]; [173–175]; [176–178]; [179 180]; [181–
183]; [184 185]; [186–188]; [189 190]; [191–193]; [194–
196]; [197–199]; [200 201]; [202–204]; [205–207]; [208–
210]; [211 212]; [213–215]; [216 217]; [218–220]

Table 4.6 Performance comparison for the spatial 23862-bar tower truss problem
Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
1 W14  605 W14  730 W14  665 W14  730
2 W14  730 W14  730 W14  605 W14  730
3 W14  730 W14  730 W14  665 W14  730
4 W14  605 W14  665 W14  665 W14  730
5 W14  730 W14  730 W14  605 W14  730
6 W14  605 W14  730 W14  665 W14  730
7 W40  249 W14  730 W14  665 W14  730
8 W14  730 W40  215 W14  665 W14  730
9 W14  665 W14  665 W14  605 W14  730
10 W14  665 W14  500 W14  665 W14  730
11 W14  665 W12  279 W14  665 W14  455
12 W27  194 W33  318 W14  426 W14  455
13 W27  194 W14  605 W14  665 W14  455
(continued)
4.4 Concluding Remarks 57

Table 4.6 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
14 W14  730 W14  730 W14  426 W14  455
15 W30  235 W14  455 W14  605 W14  455
16 W33  241 W33  221 W14  550 W14  370
17 W12  136 W44  335 W36  245 W36  328
18 W12  279 W14  426 W33  291 W12  279
19 W27  258 W33  221 W33  263 W33  291
20 W12  230 W24  229 W30  292 W12  279
21 W30  235 W14  145 W33  221 W30  191
22 W36  150 W12  252 W18  158 W24  192
23 W36  328 W27  194 W18  158 W12  136
24 W14  211 W36  245 W12  136 W27  94
25 W27  258 W27  161 W14  109 W24  131
26 W10  54 W33  118 W44  335 W14  61
27 W24  146 W33  201 W18  86 W27  94
28 W27  161 W8  21 W14  30 W12  65
29 W14  34 W14  90 W10  33 W12  26
30 W10  39 W8  21 W8  21 W10  45
31 W8  28 W8  35 W14  61 W16  36
32 W14  730 W30  211 W14  605 W14  500
33 W12  79 W14  120 W14  120 W10  112
34 W8  21 W16  67 W10  22 W8  24
35 W18  97 W10  100 W14  30 W8  24
36 W12  79 W12  26 W10  22 W6  25
37 W14  176 W8  31 W6  25 W10  22
38 W10  112 W33  118 W10  22 W10  22
39 W12  45 W10  68 W10  22 W10  22
40 W10  60 W8  21 W12  26 W10  22
41 W10  22 W8  35 W10  22 W10  22
42 W14  82 W14  74 W10  22 W10  22
43 W14  159 W8  24 W12  26 W8  21
44 W14  159 W14  120 W12  26 W8  21
45 W21  62 W8  24 W8  24 W8  21
46 W12  53 W10  39 W8  21 W8  21
47 W36  359 W16  36 W6  25 W8  21
48 W14  132 W8  21 W10  49 W8  21
49 W10  45 W8  21 W8  21 W8  21
50 W8  21 W12  40 W10  22 W8  21
51 W14  74 W14  34 W12  26 W8  21
52 W14  43 W12  26 W10  22 W8  21
53 W12  26 W8  21 W10  22 W8  21
(continued)
58 4 Optimum Design of Large-Scale Special Truss Structures

Table 4.6 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
54 W8  21 W8  21 W10  22 W8  21
55 W12  106 W8  21 W12  26 W8  21
56 W14  34 W8  21 W10  22 W8  21
57 W8  40 W8  21 W8  31 W8  21
58 W10  22 W8  24 W10  22 W8  21
59 W8  21 W14  34 W6  25 W8  21
60 W21  62 W10  22 W10  22 W8  21
61 W8  21 W16  36 W10  22 W8  21
62 W12  45 W8  35 W8  21 W8  21
63 W14  176 W33  318 W10  112 W18  97
64 W27  114 W12  136 W16  89 W16  100
65 W18  158 W21  147 W10  68 W10  54
66 W21  101 W18  86 W12  79 W12  65
67 W12  50 W10  88 W10  60 W16  67
68 W36  300 W14  82 W14  61 W16  67
69 W21  111 W12  152 W14  43 W14  61
70 W14  159 W10  49 W16  67 W16  67
71 W10  54 W10  60 W21  62 W10  54
72 W12  87 W12  136 W12  58 W12  50
73 W33  241 W16  89 W14  61 W12  50
74 W33  263 W14  90 W21  62 W12  53
75 W30  116 W14  38 W16  36 W14  48
76 W12  106 W12  65 W10  68 W10  33
77 W24  146 W14  90 W10  60 W10  49
78 W24  192 W12  65 W14  34 W8  40
79 W16  67 W30  116 W14  43 W10  45
80 W12  50 W14  90 W8  35 W14  43
81 W27  94 W18  76 W21  62 W16  36
82 W10  68 W14  48 W12  45 W8  40
83 W14  61 W10  68 W12  26 W8  31
84 W16  36 W8  28 W12  50 W8  24
85 W12  65 W10  60 W6  25 W14  34
86 W8  24 W14  38 W10  22 W12  26
87 W12  72 W10  45 W10  22 W8  31
88 W12  40 W12  50 W10  22 W16  36
89 W8  21 W14  82 W16  36 W8  28
90 W12  65 W8  40 W6  25 W10  22
91 W12  26 W10  22 W12  26 W10  22
92 W8  21 W8  21 W10  22 W10  22
93 W8  21 W8  21 W10  22 W8  21
(continued)
4.4 Concluding Remarks 59

Table 4.6 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
94 W10  22 W12  40 W8  21 W8  21
95 W8  21 W12  40 W14  82 W8  21
96 W14  61 W8  21 W10  22 W8  21
97 W12  58 W10  39 W12  79 W8  21
98 W12  30 W14  30 W21  93 W8  24
99 W14  43 W14  48 W14  30 W14  30
100 W10  39 W10  88 W10  22 W10  22
101 W18  97 W12  50 W14  48 W8  24
102 W18  76 W14  34 W8  24 W10  22
103 W10  39 W14  43 W12  26 W10  22
104 W27  84 W12  65 W8  31 W10  22
105 W12  87 W12  53 W12  26 W10  22
106 W30  116 W12  26 W12  45 W21  62
107 W12  87 W8  21 W14  38 W8  24
108 W8  31 W6  25 W14  38 W6  25
109 W14  82 W10  39 W10  22 W10  45
110 W14  30 W8  28 W10  39 W14  48
111 W12  53 W10  39 W16  89 W8  28
112 W14  34 W8  21 W14  34 W14  74
113 W8  21 W10  22 W8  21 W12  26
114 W8  40 W10  49 W14  38 W8  21
115 W12  30 W10  33 W12  30 W8  21
116 W12  40 W8  31 W10  22 W8  21
117 W8  24 W10  22 W8  28 W8  21
118 W8  24 W8  21 W6  25 W8  21
119 W10  22 W8  28 W12  58 W10  22
120 W10  22 W14  30 W24  279 W10  22
121 W8  28 W12  26 W14  38 W14  43
122 W14  34 W10  49 W8  31 W6  25
123 W12  40 W8  21 W10  22 W8  31
124 W10  49 W18  86 W10  22 W10  77
125 W36  260 W33  118 W18  158 W10  22
126 W10  22 W8  21 W8  21 W10  22
127 W6  25 W10  22 W21  182 W10  22
128 W8  21 W12  26 W8  31 W12  40
129 W16  36 W10  22 W10  22 W10  22
130 W8  28 W8  24 W14  48 W10  22
131 W14  74 W8  21 W16  36 W10  22
132 W10  45 W8  21 W12  30 W10  22
133 W10  33 W8  21 W8  21 W10  22
(continued)
60 4 Optimum Design of Large-Scale Special Truss Structures

Table 4.6 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
134 W10  22 W8  21 W8  21 W10  22
135 W10  22 W8  21 W10  22 W10  22
136 W14  38 W12  26 W10  22 W8  21
137 W12  26 W10  22 W10  49 W12  26
138 W12  50 W10  22 W12  106 W8  21
139 W10  45 W8  21 W10  22 W12  26
140 W8  21 W10  22 W6  25 W12  26
141 W8  21 W8  21 W10  22 W8  21
142 W6  25 W8  21 W10  22 W8  21
143 W12  50 W8  21 W10  22 W8  21
144 W8  31 W8  21 W10  22 W8  21
145 W8  21 W14  30 W12  30 W8  21
146 W8  21 W10  22 W12  40 W8  21
147 W14  30 W8  21 W14  550 W10  22
148 W8  21 W8  21 W10  22 W10  22
149 W8  24 W8  21 W6  25 W10  22
150 W8  21 W14  34 W10  22 W10  22
151 W10  22 W10  22 W10  22 W10  22
152 W14  30 W12  30 W8  24 W6  25
153 W8  21 W8  21 W12  26 W14  30
154 W21  101 W10  22 W8  28 W14  38
155 W14  43 W8  24 W8  31 W12  26
156 W36  230 W27  146 W12  79 W27  94
157 W12  53 W14  48 W10  22 W8  21
158 W10  22 W8  21 W10  22 W8  21
159 W14  38 W14  34 W8  24 W8  21
160 W10  54 W8  21 W10  45 W8  21
161 W14  30 W10  22 W33  201 W8  21
162 W8  21 W6  25 W14  34 W8  21
163 W10  22 W8  21 W12  65 W8  21
164 W12  30 W8  24 W12  30 W8  21
165 W8  28 W10  22 W10  22 W8  21
166 W8  21 W8  24 W10  22 W8  21
167 W8  21 W8  21 W6  25 W8  21
168 W10  22 W8  21 W8  28 W10  33
169 W14  34 W14  34 W14  30 W14  30
170 W10  22 W10  22 W8  24 W10  22
171 W8  21 W8  31 W10  22 W10  22
172 W8  21 W6  25 W8  21 W10  22
173 W8  21 W10  22 W12  26 W8  21
(continued)
4.4 Concluding Remarks 61

Table 4.6 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
174 W8  28 W8  21 W8  21 W8  21
175 W8  21 W6  25 W10  22 W8  21
176 W8  21 W8  24 W10  22 W8  21
177 W8  21 W8  21 W8  24 W8  21
178 W10  22 W8  21 W12  45 W8  21
179 W8  21 W8  21 W8  24 W8  24
180 W8  21 W8  21 W14  38 W8  24
181 W8  21 W8  21 W10  22 W8  21
182 W8  24 W10  22 W12  26 W8  21
183 W8  21 W6  25 W10  22 W8  21
184 W8  21 W8  21 W12  58 W8  21
185 W6  25 W8  21 W14  34 W8  21
186 W8  21 W14  30 W10  22 W10  77
187 W12  26 W10  22 W6  25 W16  36
188 W14  605 W14  605 W14  605 W30  326
189 W10  22 W16  36 W8  21 W10  22
190 W8  28 W8  24 W8  24 W10  22
191 W8  21 W14  38 W10  22 W10  22
192 W12  26 W8  21 W10  22 W10  22
193 W8  21 W10  22 W8  24 W10  22
194 W8  21 W8  21 W12  26 W8  21
195 W10  22 W8  21 W10  22 W8  21
196 W10  22 W8  21 W10  22 W8  21
197 W10  22 W8  28 W6  25 W8  21
198 W8  21 W8  21 W12  30 W8  21
199 W8  21 W8  21 W8  24 W8  21
200 W8  21 W10  22 W10  22 W8  21
201 W8  21 W12  26 W10  22 W8  21
202 W8  21 W8  21 W12  26 W8  21
203 W10  22 W8  21 W10  22 W8  21
204 W8  21 W6  25 W10  22 W8  21
205 W14  48 W10  22 W8  24 W8  21
206 W8  28 W8  21 W6  25 W8  21
207 W12  26 W10  22 W8  21 W8  21
208 W8  21 W8  21 W8  24 W8  21
209 W8  21 W8  21 W12  26 W8  21
210 W6  25 W8  21 W10  22 W8  21
211 W8  21 W8  21 W12  26 W8  21
212 W8  21 W6  25 W10  22 W8  21
213 W8  21 W8  21 W10  22 W8  21
(continued)
62 4 Optimum Design of Large-Scale Special Truss Structures

Table 4.6 (continued)

Element group Optimal W-shaped sections
CBO [20] ECBO [20] VPS [21] MDVC-UVPS
214 W8  28 W8  21 W10  22 W8  21
215 W10  22 W8  21 W8  24 W8  21
216 W8  21 W8  21 W10  22 W10  49
217 W12  26 W12  26 W8  21 W10  49
218 W10  22 W8  31 W10  22 W10  49
219 W12  53 W12  58 W12  53 W10  49
220 W14  211 W14  99 W27  178 W10  49
Volume (in.3) 15,587,709 14,086,857 12,989,713 12,165,572
Average optimized 16,314,541 15,556,672 13,371,681 12,435,451
volume (in.3)
Standard deviation on 390,632 343,118 267,601 145,629
average volume (in.3)

Fig. 4.8 Convergence curves for the 2386-bar tower truss problem

of MDVC-UVPS. In the first example, the design found by ECBO, VPS, and
MDVC-UVPS is approximately identical and is about 3% lighter than the result
obtained by the CBO. In the second one, the design obtained by MDVC-UVPS is
4.5, 3.5, and 1% lighter than the best designs achieved by CBO, ECBO, and VPS,
respectively. These values are 28, 15, and 6.5% for the last example. The average
optimized design and the standard deviation on average design of the hybrid
algorithm are less than those of all other compared methods for all of the examples.
Moreover, this algorithm comes close to the optimum design rapidly.
References 63

References

1. Lee KS, Geem ZW (2004) A new structural optimization method based on the harmony
search algorithm. Comput Struct 82:781–798
2. Camp CV (2007) Design of space trusses using Big Bang-Big Crunch optimization. J Struct
Eng 133:999–1008
3. Rahami H, Kaveh A, Gholipour Y (2008) Sizing, geometry and topology optimization of
trusses via force method and genetic algorithm. Eng Struct 30:2360–2369
4. Lamberti L (2008) An efficient simulated annealing algorithm for design optimization of truss
structures. Comput Struct 86:1936–1953
5. Kaveh A, Talatahari S (2009) A particle swarm ant colony optimization for truss structures
with discrete variables. J Constr Steel Res 65:1558–1568
6. Li LJ, Huang ZB, Liu F (2009) A heuristic particle swarm optimization method for truss
structures with discrete variables. Comput Struct 87:435–443
7. Wu CY, Tseng KY (2010) Truss structure optimization using adaptive multi-population
differential evolution. Struct Multidiscip Optim 42(4):575–590
8. Dede T, Bekiroglu S, Ayvaz Y (2011) Weight minimization of trusses with genetic algorithm.
Appl Soft Comput 11:2565–2575
9. Gandomi AH, Talatahari S, Yang XS, Deb S (2013) Design optimization of truss structures
using cuckoo search algorithm. Struct Des Tall Spec 22(17):1330–1349
10. Kaveh A, Mahdavi VR (2014) Colliding bodies optimization method for optimum discrete
design of truss structures. Comput Struct 139:43–53
11. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies optimization for design
problems with continuous and discrete variables. Adv Eng Softw 77:66–75
12. Ahrari A, Atai AA, Deb K (2014) Simultaneous topology, shape and size optimization of
truss structures by fully stressed design based on evolution strategy. Eng Optimiz 47(8):37–41
13. Hasançebi O, Kazemzadeh Azad S (2015) Adaptive dimensional search: a new metaheuristic
algorithm for discrete truss sizing optimization. Comput Struct 154:1–16
14. Bekdaş G, Nigdeli SM, Yang XS (2015) Sizing optimization of truss structures using flower
pollination algorithm. Appl Soft Comput 37:322–331
15. Ho-Huu V, Nguyen-Thoi T, Vo-Duy T, Nguyen-Trang T (2016) An adaptive elitist
differential evolution for optimization of truss structures with discrete design variables.
Comput Struct 165:59–75
16. Kaveh A, Bakhshpoori T (2016) A new metaheuristic for continuous structural optimization:
water evaporation optimization. Struct Multidiscip Optim 54(1):23–43
17. American Institute of Steel Construction (AISC) (1989) Manual of steel construction:
allowable stress design, Chigago, USA
18. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies optimization for design
problems with continuous and discrete variables. Adv Eng Softw 77:66–75
19. Hasançebi O (2008) Adaptive evolution strategies in structural optimization: enhancing their
computational performance with applications to large-scale structures. Comput Struct
86:119–132
20. Kaveh A, Ilchi Ghazaan M (2016) Optimum design of large-scale truss towers using cascade
optimization. Acta Mech 227(9):2645–2656
21. Kaveh A, Ilchi Ghazaan M (2017) MATLAB codes for vibrating particles system algorithm.
Int J Optim Civil Eng 7(3):355–366
Chapter 5
Optimal Design of Double-Layer Grids

5.1 Introduction

Double-layer grids belong to the category of space structures and consist of two
planar networks of members forming the top and bottom layers parallel to each
other and interconnected by vertical and inclined web members. Double-layer grids
are characterized by ball joints with no moment or torsional resistance; therefore, all
members can only resist tension or compression. Even in the case of connection by
comparatively rigid joints, the influence of bending or torsional moment is
insignificant. A wide variety of double-layer grids can be formed either by altering
the direction of the top and bottom layers with respect to each other or by changing
the relative positions of the nodal points of the top and bottom layers. Additional
variations can also be introduced by utilizing layers of different sizes [1].
Double-layer grids are ideally suited for covering exhibition pavilions, assembly
halls, swimming pools, hangars, churches, bridge decks, and many types of
industrial buildings in which large unobstructed areas are required.
In the past decades, a number of meta-heuristic algorithms have been developed
and used for structural optimization problems, e.g., see Kaveh [2]. Double-layer
grids have a great number of structural elements, and therefore optimization tech-
niques can be rewardingly employed to achieve economic and efficient designs of
them. Here, five different types of double-layer grids are studied and optimized
utilizing the colliding bodies optimization (CBO) [3], enhanced colliding bodies
optimization (ECBO) [4], vibrating particles system (VPS) [5], and a hybrid
algorithm called MDVC-UVPS [6]. The cross-section areas of the grid elements are
considered as discrete design variables and all of them are selected from a list of
tube sections available in AISC-LRFD [7]. Strength, stability, and displacement
constraints are considered for each example.

© Springer International Publishing AG, part of Springer Nature 2018 65

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_5
66 5 Optimal Design of Double-Layer Grids

5.2 Optimal Design of Double-Layer Grids

Here, the aim of the optimization problem is to find a set of design variables that
result in a double-layer grid with the minimum weight while satisfying certain
constraints. This can be expressed as


Find fX g ¼ x1 ; x2 ; ::; xng
X
ng X
nmðiÞ
to minimize W ð f X gÞ ¼ xi qj Lj
i¼1 j¼1 ð5:1Þ
(
gj ðfX gÞ  0; j ¼ 1; 2; . . .; nc
subjected to :
ximin  xi  ximax

where {X} is the vector containing the design variables; ng is the number of design
groups; W({X}) presents the weight of the structure; nm(i) is the number of
members for the ith group; qj and Lj denote the material density and the length of
the jth member, respectively. ximin and ximax are the lower and upper bounds of the
design variable xi, respectively. gj({X}) denotes design constraints; and nc is the
number of constraints.
For handling the constraints, a penalty approach is utilized. For this purpose, the
objective function (Eq. 5.1) is redefined as follows:

Pðf X gÞ ¼ ð1 þ e1 :tÞe2  W ðf X gÞ ð5:2Þ

where P({X}) is the penalized cost function or the objective function to be mini-
mized and t denotes the sum of the violations of the design constraints. Here, e1 is
set to unity and e2 is calculated by

iter
e2 ¼ 1:5 þ 1:5  ð5:3Þ
itermax

where iter is the current iteration number and itermax is the total number of iterations
for the optimization process.
The constraint conditions for grid structures are briefly explained in the
following.
Displacement constraint:

di  dmax
i ; i ¼ 1; 2; . . .; nn ð5:4Þ

Tension member constraint [7]:


/t Fy Ag ; /t ¼ 0:9
pu  p r ; pr ¼ min ð5:5Þ
/t Fu Ae ; /t ¼ 0:75
5.2 Optimal Design of Double-Layer Grids 67

Compression member constraint [7]:

pu  pr ; pr ¼ /c Fcr Ag ; /c ¼ 0:85
8 Fy qffiffiffiffi
>
< ð0:658Fe ÞFy ; KL  4:71 E
r Fy p2 E ð5:6Þ
Fcr ¼ qffiffiffiffi ; Fe ¼
> ðKL 2
: 0:877Fe ; r Þ
KL E
r [ 4:71 Fy

Slenderness ratio constraints [7]:

KL
 200; for compression members
r ð5:7Þ
KL
 300; for tension members
r

where di and dmax

i are the displacement and allowable displacement for the ith node;
nn is the number of nodes; Pu is the required strength (tension or compression); Pr
is the nominal axial strength (tension or compression); Ag and Ae are the gross
cross-sectional area and the effective net cross-sectional area of a member; k is the
effective length factor taken equal to 1; L is the length of member; and r is the
radius of gyration.

5.3 Design Problems

Design optimization of five double-layer grids with different configurations are

investigated in this section. These examples include the following:
• A 520-bar double-layer grid (larger square on square)
• A 672-bar double-layer grid (square on larger square)
• A 800-bar double-layer grid (square on square)
• A 1016-bar double-layer grid (square on diagonal)
• A 1520-bar double-layer grid (diagonal on diagonal).
A span of 40  40 m is considered for all examples and the height is equal to
3 m. All connections are assumed to be ball jointed. The design variables are the
cross-sectional areas of the bar elements which are selected from the list of steel
pipe sections from AISC-LRFD [7]. These pipe sections are shown in Table 5.1.
ST, EST, and DEST abbreviations stand for standard weight, extra strong, and
double-extra strong, respectively. The modulus of elasticity, the yield stress, and the
density of steel are taken as 205 GPa, 248.2 MPa, and 7833.413 kg/m3, respec-
tively. Strength and slenderness limitations are according to AISC-LRFD provi-
sions [7] as discussed earlier and displacement limitations of span/600 were
imposed on all nodes in the vertical direction.
68 5 Optimal Design of Double-Layer Grids

Table 5.1 The steel pipe sections

No. Type Nominal diameter (in.) Area (cm2) Gyration radius (cm)
1 a
ST ½ 1.6129 0.662432
2 b
EST ½ 2.064512 0.635
3 ST ¾ 2.129028 0.846582
4 EST ¾ 2.774188 0.818896
5 ST 1 3.161284 1.066038
6 EST 1 4.129024 1.034542
7 ST 1¼ 4.322572 1.371346
8 ST 1½ 5.16128 1.582166
9 EST 1¼ 5.677408 1.331214
10 EST 1½ 6.903212 2.003806
11 ST 2 6.903212 1.53543
12 EST 2 9.548368 1.945132
13 ST 2½ 10.96772 2.41681
14 ST 3 14.387068 2.955798
15 EST 2½ 14.5161 2.346452
c
16 DEST 2 17.161256 1.782572
17 ST 3½ 17.290288 3.395726
18 EST 3 19.483832 2.882646
19 ST 4 20.451572 3.835908
20 EST 3½ 23.741888 3.318002
21 DEST 2½ 25.999948 2.143506
22 ST 5 27.74188 4.775454
23 EST 4 28.451556 3.749548
24 DEST 3 35.290252 2.65811
25 ST 6 35.999928 5.700014
26 EST 5 39.419276 4.675124
27 DEST 4 52.25796 3.490976
28 ST 8 54.19344 7.462012
29 EST 6 54.19344 5.577332
30 DEST 5 72.90308 4.379976
31 ST 10 76.77404 9.342628
32 EST 8 82.58048 7.309358
33 ST 12 94.19336 11.10361
34 DEST 6 100.64496 5.236464
35 EST 10 103.87076 9.216898
36 EST 12 123.87072 11.028934
37 DEST 8 137.41908 7.004812
a
ST standard weight
b
EST extra strong
c
DEST double-extra strong
5.3 Design Problems 69

Each example has been solved 20 times independently, and 1000 iterations are
considered as the terminal condition. A population of 20 particles is considered for
each algorithm and the other algorithm parameters are set the same as the values
proposed in [4–6]. The optimization algorithms are coded in MATLAB and the
structures are analyzed using the direct stiffness method by our own codes.

5.3.1 A 520-Bar Double-Layer Grid

The larger square on square double-layer grid contains 520 members and
165 nodes shown in Fig. 5.1. The bottom layer is simply supported at the nodes
illustrated in Fig. 5.2a. Each top layer joint is subjected to a concentrated vertical
load of 46 kN. Cross-sectional areas of the members are categorized into 20 groups
as shown in Fig. 5.2. In order to optimize this structure by MDVC-UVPS, two
stages are considered. The design variable configuration utilized for the first stage is
listed as follows: [1], [2 4], [3 5], [6 8], [7 9], [10 11], [12 13], [14 15], [16], [17
18], and [19 20].
The results found by CBO, ECBO, VPS, and MDVC-UVPS algorithms are
summarized in Table 5.2. VPS achieves the lightest design (i.e., 60,018 kg).
MDVC-UVPS has better performance in terms of the average optimized weight and
standard deviation on average weights which are 63,130, and 1932 kg, respectively.
The best designs obtained by CBO, ECBO, and MDVC-UVPS are 64,513, 61,119,
and 61,456 kg, respectively. The maximum stress ratios for the best designs of the
CBO, ECBO, VPS, and MDVC-UVPS are 94.39, 96.76, 96.90, and 98.33%,
respectively. Convergence histories are depicted in Fig. 5.3. The required number
of structural analyses to achieve the best design by CBO, ECBO, VPS, and
MDVC-UVPS are 3400, 19,820, 19,560, and 2804 analyses, respectively.

5.3.2 A 672-Bar Double-Layer Grid

Figure 5.4 shows the 3D view of a square on larger square grid. This structure has
672 members and 205 nodes and the bottom layer is simply supported at the nodes
shown in Fig. 5.5a. Each top layer joint is subjected to a concentrated vertical load

Fig. 5.1 3D view of the 520-bar double-layer grid problem

70 5 Optimal Design of Double-Layer Grids

Fig. 5.2 Top view of the 520-bar double-layer grid problem and member groups: a all members
with simple supports, b bottom layer members, c top layer members, and d web members

of 30 kN. The cross-sectional areas of the members are categorized into 22 groups
as depicted in Fig. 5.5. In order to optimize this structure by MDVC-UVPS, two
stages are considered. The design variable configuration utilized for the first stage is
listed as follows: [1 2], [3 4], [5 6], [7 8], [9], [10 12], [11 13], [14 16], [15 17], [18
19], [20 21], and [22].
Table 5.3 lists the optimal designs found by different methods. MDVC-UVPS
obtained the lightest design compared to other methods that is 53,552 kg.
Moreover, the average optimized weight and the standard deviation on average
weight of MDVC-UVPS (58,589 and 3626 kg) are less than those of all other
methods. The best designs found by the CBO, ECBO, and VPS are 55,621, 54,569,
and 53,704 kg, respectively. The maximum stress ratios for the best designs of the
5.3 Design Problems 71

Table 5.2 Performance comparison for the 520-bar double-layer grid problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 EST 4 ST 4 ST 5 ST 6
2 EST 4 ST 5 ST 5 EST 4
3 ST 4 DEST 3 ST 3½ EST 4
4 ST 4 ST 3½ ST 3½ ST 3
5 DEST 5 EST 5 EST 4 ST 6
6 ST 3½ ST 3½ EST 3 EST 4
7 EST 3½ EST 5 DEST 2½ EST 4
8 EST 4 DEST 4 EST 5 DEST 4
9 EST 6 ST 8 EST 6 EST 6
10 EST 6 EST 6 ST 8 EST 8
11 EST 10 EST 5 EST 5 ST 5
12 EST 8 DEST 5 DEST 5 EST 6
13 EST 8 ST 12 EST 12 ST 10
14 ST 10 DEST 6 DEST 6 EST 12
15 DEST 8 DEST 8 DEST 6 DEST 8
16 ST 6 ST 6 ST 6 ST 5
17 ST 4 ST 4 ST 4 ST 4
18 ST 3½ ST 3½ ST 3½ ST 4
19 EST 3 ST 3 ST 3½ ST 5
20 ST 3½ ST 2½ ST 2½ ST 5
Weight (kg) 64,513 61,119 60,018 61,456
Average optimized weight (kg) 66,906 63,463 63,360 63,130
Standard deviation on average weight (kg) 7167 4052 2446 1932

Fig. 5.3 Convergence curves for the 520-bar double-layer grid problem
72 5 Optimal Design of Double-Layer Grids

Fig. 5.4 3D view of the 672-bar double-layer grid problem

Fig. 5.5 Top view of the 672-bar double-layer grid problem and member groups: a all members
with simple supports, b bottom layer members, c top layer members, and d web members
5.3 Design Problems 73

Table 5.3 Performance comparison for the 672-bar double-layer grid problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 ST 4 ST 4 ST 5 ST 4
2 ST 6 ST 5 ST 4 ST 5
3 ST 2½ ST 3½ ST 4 ST 3½
4 ST 2½ EST 1½ EST 1½ EST 1½
5 ST 2½ ST 4 ST 6 ST 3
6 EST 6 EST 4 EST 6 DEST 4
7 EST 4 EST 6 EST 6 DEST 4
8 EST 6 DEST 4 DEST 5 EST 6
9 EST 5 ST 6 EST 5 ST 6
10 ST 4 ST 3½ ST 3½ ST 3½
11 ST 6 EST 6 ST 5 DEST 4
12 EST 5 EST 4 EST 4 ST 5
13 EST 5 DEST 4 EST 6 EST 6
14 DEST 4 DEST 4 ST 6 EST 5
15 EST 8 EST 6 DEST 4 DEST 4
16 DEST 4 DEST 4 EST 6 EST 5
17 ST 5 DEST 5 ST 8 DEST 5
18 ST 4 ST 4 ST 4 ST 4
19 ST 4 ST 4 ST 3½ ST 3½
20 ST 3½ ST 3½ ST 3½ ST 3½
21 ST 3½ ST 3 ST 3½ EST 2½
22 ST 2½ ST 2½ ST 2½ ST 2½
Weight (kg) 55,621 54,569 53,704 53,552
Average optimized weight (kg) 62,287 59,164 60,850 58,589
Standard deviation on average weight (kg) 9853 5597 5985 3626

CBO, ECBO, VPS, and MDVC-UVPS are 96.59, 95.73, 99.96, and 98.91%,
respectively. Convergence histories are demonstrated in Fig. 5.6. It should be noted
that MDVC-UVPS requires 3772 structural analyses to find the optimum solution
while CBO, ECBO, and VPS require 4640, 14,480, and 19,640 structural analyses,
respectively.
74 5 Optimal Design of Double-Layer Grids

Fig. 5.6 Convergence curves for the 672-bar double-layer grid problem

5.3.3 A 800-Bar Double-Layer Grid

The design of square on square double-layer grid shown in Fig. 5.7 is considered as
the third example. This structure has 800 members and 221 nodes and the bottom
layer is simply supported at the nodes illustrated in Fig. 5.8a. Each top layer joint is
subjected to a concentrated vertical load of 30 kN. Cross-sectional areas of the
members are categorized into 24 groups as shown in Fig. 5.8. In order to optimize
this structure by MDVC-UVPS, two stages are considered. The design variable
configuration utilized for the first stage is listed as follows: [1], [2 4], [3 5], [6 8], [7
9], [10], [11], [12 14], [13 15], [16 18], [17 19], [20 21 22], and [23 24].
Table 5.4 summarizes the results obtained by CBO, ECBO, VPS, and
MDVC-UVPS methods. MDVC-UVPS has better performance in terms of the best
weight, average optimized weight, and standard deviation on average weight which
are 53,590, 57,679, and 3524 kg, respectively. The best designs obtained by CBO,
ECBO, and VPS are 55,714, 53,673, and 53.714 kg, respectively. The maximum
stress ratios for the best designs of the CBO, ECBO, VPS, and MDVC-UVPS are

Fig. 5.7 3D view of the 800-bar double-layer grid problem

5.3 Design Problems 75

Fig. 5.8 Top view of the 800-bar double-layer grid problem and member groups: a all members
with simple supports, b bottom layer members, c top layer members, and d web members

99.94, 93.86, 91.87, and 94.71%, respectively. Convergence histories are depicted
in Fig. 5.9. The required number of structural analyses to achieve the best design by
CBO, ECBO, VPS, and MDVC-UVPS are 11,520, 16,860, 15,600, and 5122
analyses, respectively.

5.3.4 A 1016-Bar Double-Layer Grid

Figure 5.10 shows the 3D view of a square on the diagonal grid. This grid has
1016 members and 320 nodes and simple support conditions are employed for the
bottom layer at the nodes demonstrated in Fig. 5.11a. Each top layer joint is sub-
jected to a concentrated vertical load of 30 kN. The elements are divided into
76 5 Optimal Design of Double-Layer Grids

Table 5.4 Performance comparison for the 800-bar double-layer grid problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 EST 3½ ST 4 ST 4 ST 4
2 ST 6 ST 5 ST 5 ST 5
3 ST 2 EST 2 EST 1½ ST 1½
4 ST 3½ ST 3 EST 3 ST 3
5 ST 2½ EST 2 ST 3½ ST 2½
6 ST 3 ST 2 EST 1½ ST 2
7 EST 3 EST 3½ ST 5 ST 3
8 ST 2½ ST 3 ST 4 DEST 2
9 EST 3 EST 3½ EST 3 ST 5
10 ST 5 ST 3 ST 2 DEST 3
11 ST 8 EST 5 ST 6 DEST 4
12 ST 3½ ST 3½ ST 3½ ST 3½
13 ST 4 ST 6 ST 6 ST 6
14 ST 5 ST 6 ST 6 ST 5
15 ST 6 ST 6 ST 6 ST 5
16 ST 6 ST 6 ST 6 ST 6
17 DEST 4 EST 5 EST 6 DEST 4
18 EST 5 EST 6 EST 5 DEST 4
19 EST 5 DEST 4 DEST 4 DEST 5
20 EST 3½ ST 4 ST 4 ST 4
21 ST 3½ ST 3½ ST 3½ ST 3½
22 ST 3 ST 3½ ST 3 ST 3
23 ST 2½ ST 2½ ST 2½ ST 2½
24 ST 2½ ST 2½ ST 2½ ST 2½
Weight (kg) 55,714 53,673 53,714 53,590
Average optimized weight (kg) 61,464 58,953 57,912 57,679
Standard deviation on average weight (kg) 10,127 4643 4102 3524

25 groups and the member groups are presented in Fig. 5.11. Two stages with 11
and 25 variables are considered by MDVC-UVPS algorithm. The first DVC is as
follows: [1 2], [3 4 5], [6 7 8], [9 10 11], [12], [13 15], [14 16], [17 19], [18 20], [21
22 23], and [24 25].
Table 5.5 presents the optimum designs obtained by proposed algorithms. The
lightest design (i.e., 65,826 kg) is achieved by MDVC-UVPS algorithm after
4234 analyses. The best designs obtained by CBO, ECBO, and VPS are 74,849,
67,839, and 67,229 kg, respectively. These values are found after 9760, 15,760,
and 15,220 analyses. MDVC-UVPS obtains the lost values of average optimized
weight and standard deviation on average weight which are equal to 70,488, and
5018 kg, respectively. The maximum values of the stress ratio for CBO, ECBO,
5.3 Design Problems 77

Fig. 5.9 Convergence curves for the 800-bar double-layer grid problem

Fig. 5.10 3D view of the 1016-bar double-layer grid problem

VPS, and MDVC-UVPS are 93.01, 93.99, 96.07, and 97.10%, respectively.
Convergence history diagrams are depicted in Fig. 5.12.

5.3.5 A 1520-Bar Double-Layer Grid

Figure 5.13 shows the 3D view of a diagonal on diagonal grid. This structure has
1520 members and 401 nodes and the bottom layer is simply supported at the
nodes shown in Fig. 5.14a. Each top layer joint is subjected to a concentrated
vertical load of 16 kN. The cross-sectional areas of the members are categorized
into 31 groups as depicted in Fig. 5.14. In order to optimize this structure by
MDVC-UVPS, two stages are considered. The design variable configuration
78 5 Optimal Design of Double-Layer Grids

Fig. 5.11 Top view of the 1016-bar double-layer grid problem and member groups: a all
members with simple supports, b bottom layer members, c top layer members, and d web
members

utilized for the first stage is listed as follows: [1 2], [3 4 5], [6 7 8], [9 10 11], [12],
[13 14], [15 16], [17 18], [19 20], [21], [22], [23 24 25], [26 27 28], and [29 30 31].
Table 5.6 presents a comparison between the results of the optimal designs
found by different methods. MDVC-UVPS obtained the lightest design compared
to other methods that is 79,571 kg. Moreover, the average optimized weight and the
standard deviation on the average weight of MDVC-UVPS (85,398 and 4407 kg)
are less than those of all other methods. The best designs found by the CBO,
ECBO, and VPS are 93,174, 82,254, and 82,357 kg, respectively. The maximum
5.3 Design Problems 79

Table 5.5 Performance comparison for the 1016-bar double-layer grid problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 EST 5 EST 5 ST 6 DEST 4
2 DEST 3 EST 5 ST 5 DEST 3
3 ST 3½ ST 3 ST 3½ ST 3½
4 ST 2½ ST 3½ ST 2½ ST 2½
5 ST 2½ ST 2½ ST 4 ST 3
6 ST 2 ST 2 EST 1 EST 1½
7 ST 2 DEST 2 EST 2 EST 1½
8 ST 2½ DEST 2 DEST 2 EST 2½
9 DEST EST 2 EST 3 ST 3½
2½
10 DEST ST 6 DEST DEST 2
2½ 2½
11 ST 1½ ST 2 EST 12 DEST 2½
12 DEST 5 EST 8 DEST 5 EST 8
13 EST 3½ EST 3½ ST 4 EST 4
14 EST 3½ ST 5 ST 5 ST 4
15 EST 4 ST 4 ST 5 ST 5
16 ST 6 EST 5 DEST 4 ST 4
17 ST 5 ST 5 EST 4 ST 6
18 EST 4 EST 5 EST 4 ST 6
19 EST 5 EST 5 EST 4 EST 6
20 ST 8 ST 8 DEST 4 EST 6
21 ST 6 ST 5 ST 6 ST 5
22 ST 3 ST 3 ST 3½ ST 3½
23 EST 6 EST 2½ EST 2½ EST 2½
24 ST 3½ ST 5 ST 2½ ST 2½
25 EST 1½ ST 4 EST 1½ ST 2½
Weight (kg) 74,849 67,839 67,229 65,826
Average optimized weight (kg) 79,422 73,042 72,366 70,488
Standard deviation on average weight (kg) 8154 9158 5545 5018

stress ratio for the best designs of the CBO, ECBO, VPS, and MDVC-UVPS are
99.59, 99.40, 99.94, and 99.92%, respectively. Convergence histories are demon-
strated in Fig. 5.15. It should be noted that MDVC-UVPS requires 3142 structural
analyses to find the optimum solution while CBO, ECBO, and VPS require 4360,
18,000, and 12,120 structural analyses, respectively.
80 5 Optimal Design of Double-Layer Grids

Fig. 5.12 Convergence curves for the 1016-bar double-layer grid problem

Fig. 5.13 3D view of the 1520-bar double-layer grid problem

5.4 Concluding Remarks

Optimal design of larger square on square, square on larger square, square on

square, square on diagonal, and diagonal on diagonal double-layer grids using four
meta-heuristic algorithms are studied. Strength and stability constraints of
AISC-LRFD specifications and some displacement limitations are imposed on these
grids. The diagonal on diagonal type has more connections and members and the
optimum weight achieved for this configuration is heavier than those of other grids.
The lightest designs are found for the square on larger square and square on square
grids. For the utilized algorithms, MDVC-UVPS achieved the best designs for four
of the five cases. This algorithm had also better performance in terms of the average
optimized weight and standard deviation on average weight. Convergence history
diagrams are depicted for all design problems confirming the superiority of the
MDVC-UVPS algorithm.
5.4 Concluding Remarks 81

Fig. 5.14 Top view of the 1520-bar double-layer grid problem and member groups: a all members
with simple supports, b bottom layer members, c top layer members, and d web members

Table 5.6 Performance comparison for the 1520-bar double-layer grid problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 EST 5 DEST 5 EST 5 ST 6
2 ST 5 EST 5 ST 5 DEST 3
3 ST 2½ ST 2½ EST 2½ ST 2½
4 EST 1½ EST 1½ EST 1½ ST 2½
5 ST 1½ ST 2 ST 1½ ST 2½
6 EST 1¼ ST 2 ST 1¼ EST 2½
7 ST 3 EST 2½ ST 8 EST 2½
8 ST 4 EST 3 EST 5 EST 2½
(continued)
82 5 Optimal Design of Double-Layer Grids

Table 5.6 (continued)

Element group Sections
CBO ECBO VPS MDVC-UVPS
9 ST 5 DEST EST 3½ EST 4
2½
10 EST 10 EST 4 DEST 3 DEST 3
11 ST 6 ST 6 EST 5 DEST 3
12 DEST 6 DEST 5 EST 10 DEST 6
13 EST 4 ST 2½ ST 2½ ST 3½
14 ST 3½ ST 3½ ST 2½ ST 3½
15 ST 3 ST 4 EST 2½ ST 3½
16 ST 4 ST 5 ST 3 EST 3
17 EST 10 DEST ST 4 DEST 2½
2½
18 EST 3 EST 3½ DEST 3 DEST 3
19 DEST ST 6 DEST 3 EST 5
2½
20 EST 4 DEST 5 DEST 4 ST 6
21 DEST 6 EST 6 EST 10 ST 10
22 EST 5 EST 5 ST 6 ST 6
23 EST 3½ ST 3 ST 2½ ST 2½
24 ST 3 ST 2½ EST 2½ ST 2½
25 ST 2½ ST 2½ ST 3 ST 2½
26 EST 1½ EST 1½ EST 1½ EST 2
27 ST 3 EST 2 ST 2½ EST 2
28 ST 2½ EST 2 ST 3 EST 2
29 ST 2½ ST 2½ ST 2½ EST 1½
30 ST 3 EST 1½ ST 1½ EST 2
31 ST 3½ EST 2 ST 4 ST 2½
Weight (kg) 93,174 82,254 82,357 79,571
Average optimized weight (kg) 97,823 90,752 89,607 85,398
Standard deviation on average weight (kg) 9226 5995 5188 4407
References 83

Fig. 5.15 Convergence curves for the 1520-bar double-layer grid problem

References

1. Lan TT (2005) Space frame structures. Handbook of structural engineering. CRC Press, Boca
Raton, FL
2. Kaveh A (2017) Advances in metaheuristic algorithms for optimal design of structures, 2nd
edn. Springer, Switzerland
3. Kaveh A, Mahdavi VR (2014) Colliding bodies optimization: a novel meta-heuristic method.
Comput Struct 139:18–27
4. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies optimization for design problems
with continuous and discrete variables. Adv Eng Softw 77:66–75
5. Kaveh A, Ilchi Ghazaan M (2016) Vibrating particles system algorithm for truss optimization
with multiple natural frequency constraints. Acta Mech 228(1):307–332
6. Kaveh A, Ilchi Ghazaan M (2018) A new hybrid meta-heuristic algorithm for optimal design of
large-scale dome structures. Eng Optimiz 50(2):235–252
7. American Institute of Steel Construction (AISC) (1994) Manual of steel construction load
resistance factor design. USA
Chapter 6
Optimal Design of Double-Layer Barrel
Vault Space Structures

6.1 Introduction

The brick architecture of the Orient and the masonry construction of the Romans
provide numerous examples of the structural use of barrel vaults. During recent
years, architects and engineers have rediscovered the advantages of barrel vaults as
viable and often highly suitable forms for covering not only low-cost industrial
buildings, warehouses, large-span hangars, and indoor sports stadiums, but also
large cultural and leisure centers. The impact of industrialization on production of
prefabricated barrel vaults has proved to be the most significant factor leading to
lower cost for these structures [1].
Barrel vaults are given different names depending on the way their surface is
formed. The earlier types of barrel vaults were constructed as single-layer struc-
tures. Nowadays, with the increase of the spans, double-layer systems are often
preferred. While the members of single-layer barrel vaults are mainly under the
action of flexural moments, those of double-layer barrel vaults are almost exclu-
sively under the action of axial forces and the elimination of bending moments
leads to a full utilization of strength of all the elements. Double-layer barrel vaults
are generally statically indeterminate. In such systems, due to the rigidity, the risk
of instability can almost be eliminated. The use of this type of barrel vaults
enhances the stiffness of the vault structure and provides structural systems of great
potential, capable of having spans in excess of 100 m [2, 3].
In the recent decades, different meta-heuristic algorithms have been developed
and applied to structural optimization problems. The growing popularity of these
techniques arises from (i) the lack of dependency on gradient information; (ii) in-
herent capability to deal with both discrete and continuous design variables; and
(iii) incorporating global search features to produce reasonable solutions for com-
plicated problems [4]. Double-layer barrel vaults have a great number of structural
elements, and utilizing optimization techniques has a considerable influence on
their economy and efficient structural configuration. For optimal design of

© Springer International Publishing AG, part of Springer Nature 2018 85

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_6
86 6 Optimal Design of Double-Layer Barrel Vault Space Structures

double-layer barrel vaults, Kaveh and Eftekhar presented an improved hybrid Big
Bang-Big Crunch (IBBBC) algorithm [5]. In another study optimal design of some
single-layer barrel vaults and a double arch barrel vault were investigated by Kaveh
et al. [6]. In several studies, the optimal design of a practical model of a braced
barrel vault has been studied by various researchers. Hasançebi and Çarbaş used ant
colony search method [7], Hasançebi et al. employed a reformulation of the ant
colony optimization [8] and Hasançebi and Kazemzadeh Azad utilized a refor-
mulation of Big Bang-Big Crunch algorithm and adaptive dimensional search
method [4, 9, 10].
In this chapter, three double-layer barrel roof structures are optimized to
investigate the performance of the CBO [11], ECBO [12], VPS [13] and
MDVC-UVPS [14] meta-heuristic algorithms. The structures are subjected to
stress, stability, and displacement limitations according to the provisions of
AISC-ASD [15]. The design variables are the cross-sectional areas of the bar
elements which are selected from a list of steel pipe sections.

6.2 Optimal Design of Double-Layer Barrel Vaults

Typically in practical design optimization of skeletal structures, the goal is to find a

minimum cost or weight design by selecting the cross-sectional areas of structural
elements from a list of available sections such that the final design satisfies strength
and serviceability requirements determined by standard design codes. Size opti-
mization of a skeletal structure with its members being collected in ng design
groups can be formulated as follows:

Find fXg ¼ ½x1 ; x2 ; . . .; xng 

Png P
nmðiÞ
to minimize WðfXgÞ ¼ xi qj Lj
i¼1 j¼1 ð6:1Þ

gj ðfXgÞ  0; j ¼ 1; 2; . . .; nc
subjected to:
xi min  xi  xi max

where {X} is the vector containing the design variables; W({X}) presents the weight
of the structure; nm(i) is the number of members for the ith group; qj and Lj denote
the material density and the length of the jth member, respectively. ximin and ximax
are the lower and upper bounds of the design variable xi, respectively. gj({X})
denotes design constraints; and nc is the number of constraints.
For constraint handling, a penalty approach is utilized. For this purpose, the
objective function (Eq. 6.1) is redefined as follows:

PðfXgÞ ¼ ð1 þ e1 :tÞe2  WðfXgÞ ð6:2Þ

6.2 Optimal Design of Double-Layer Barrel Vaults 87

where P({X}) is the penalized cost function or the objective function to be mini-
mized and t denotes the sum of the violations of the design constraints. Here, e1 is
set to unity and e2 is calculated by:

iter
e2 ¼ 1:5 þ 1:5  ð6:3Þ
itermax

where iter is the current iteration number and itermax is the total number of iterations
for optimization process.
The constraint conditions for barrel vaults studied here are briefly explained in
the following. Limitation on stress and stability of barrel vault elements are
imposed according to the provisions of the AISC-ASD [15] as follows:
The allowable tensile stresses for tension members are calculated by

riþ ¼ 0:6Fy ð6:4Þ

where Fy stands for the yield strength.

The allowable stress limits for compression members are calculated depending
on two possible failure modes of the members known as elastic and inelastic
buckling. Thus
8 h  i h i
< 1  ki 2 Fy = 5 þ 3ki  ki 3
2 3
for ki \Cc
r
2C 3 8C 8C
i ¼ ð6:5Þ
c c c
: 12p22E for ki  Cc
23ki

where E is the modulus of elasticity; ki is the slenderness ratio ðki ¼ kli =ri Þ; Cc
denotes the slenderness ratio dividing the elastic and inelastic buckling regions
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðCc ¼ 2p2 E=Fy Þ; k is the effective length factor (k is set 1 for all truss members);
Li is the member length; and ri is the minimum radius of gyration.
AISC-ASD recommends the maximum slenderness ratio of the elements to be
restricted to 300 and 200 for tension and compression members, respectively.

6.3 Design Examples

Three double-layer barrel vault problems with 384, 693, and 1536 bar elements are
considered to evaluate the performance of CBO, ECBO, VPS, and MDVC-UVPS
algorithms. The design variables are the cross-sectional areas of the bar elements
which are selected from the list of steel pipe sections of AISC-LRFD [16]. These
pipe sections are shown in Table 6.1. ST, EST, and DEST abbreviations stand for
standard weight, extra strong, and double-extra strong, respectively. Each example
has been solved 30 times independently and a maximum of 1000 iterations is
considered as the termination condition. A population of 20 particles is considered
88 6 Optimal Design of Double-Layer Barrel Vault Space Structures

Table 6.1 The steel pipe sections

No. Type Nominal diameter (in) Area (in2) Moment of Gyration radius (in)
inertia (in4)
1 ST 1/2 0.25 0.017 0.2608
2 EST 1/2 0.32 0.02 0.2500
3 ST 3/4 0.33 0.037 0.3333
4 EST 3/4 0.43 0.045 0.3224
5 ST 1 0.49 0.087 0.4197
6 EST 1 0.64 0.11 0.4073
7 ST 1 1/4 0.67 0.19 0.5399
8 ST 1 1/2 0.8 0.31 0.6229
9 EST 1 1/4 0.88 0.24 0.5241
10 EST 1 1/2 1.07 0.67 0.7889
11 ST 2 1.07 0.39 0.6045
12 EST 2 1.48 0.87 0.7658
13 ST 2 1/2 1.7 1.54 0.9515
14 ST 3 2.23 3.02 1.1637
15 EST 2 1/2 2.25 1.92 0.9238
16 DEST 2 2.66 1.31 0.7018
17 ST 3 1/2 2.68 4.79 1.3369
18 EST 3 3.02 3.89 1.1349
19 ST 4 3.17 7.23 1.5102
20 EST 3 1/2 3.68 6.28 1.3063
21 DEST 2 1/2 4.03 2.87 0.8439
22 ST 5 4.3 15.2 1.8801
23 EST 4 4.41 9.61 1.4762
24 DEST 3 5.47 5.99 1.0465
25 ST 6 5.58 28.1 2.2441
26 EST 5 6.11 20.7 1.8406
27 DEST 4 8.1 15.3 1.3744
28 ST 8 8.4 72.5 2.9378
29 EST 6 8.4 40.5 2.1958
30 DEST 5 11.3 33.6 1.7244
31 ST 10 11.9 161 3.6782
32 EST 8 12.8 106 2.8777
33 ST 12 14.6 279 4.3715
34 DEST 6 15.6 66.3 2.0616
35 EST 10 16.1 212 3.6287
36 EST 12 19.2 362 4.3421
37 DEST 8 21.3 162 2.7578
ST standard weight, EST extra strong, DEST double-extra strong
6.3 Design Examples 89

for each algorithm and the other algorithm parameters are set similar to the values
proposed in [12–14]. The optimization algorithms are coded in MATLAB and the
structures are analyzed using the direct stiffness method by our own codes.

6.3.1 A 384-Bar Double-Layer Barrel Vault

The first design problem deals with the size optimization of a spatial 384-bar barrel
vault consisting of two rectangular nets as shown in Fig. 6.1. In order to make the
structure stable, the angles of the bottom nets are placed at the center of one of the
above nets. The two nets are connected using diagonal elements [3]. The span of
the barrel vault is 24.82 m, its rise is 5.12 m and its length is 26.67 m. The depth of
the structure, i.e., the distance between the top and bottom layers, is equal to
1.35 m. This structure consists of 111 pinned joints and 384 bar elements, which
are grouped into 31 independent sizing variables as identified in Fig. 6.1. The
structural material properties are assumed as follows: The modulus of elasticity is
considered to be 30,450 ksi (210,000 MPa), the yield stress of steel is taken as 58
ksi (400 MPa), and the density of steel is equal to 0.288 lb per cubic inch
(7833.413 kg/m3). All connections are assumed as ball jointed and the supports are
considered at the two external edges of the top layer of the barrel vault. Vertical
concentrated loads of −20 kips (−88.964 kN) are applied to all free joints (non-
support joints) of the top layer. Strength and slenderness limitations are according
to AISC-ASD provision [15], which are discussed earlier. Displacement constraints
of ±0.1969 in (5 mm) are imposed on all nodes in x, y and z directions.
Table 6.2 presents the comparison of the results of different algorithms.
The VPS algorithm yields the least weight for this example, which is 62,455.30 lb
(28,329.24 kg). The other design weights are 69,448.52 lb (31,501.32 kg) by CBO,
62,486.02 lb (28,343.18 kg) by ECBO, and 62,735.42 lb (28,456.31 kg) by
MDVC-UVPS. The best design of the VPS has been achieved in 12,780 analyses.
CBO, ECBO, and MDVC-UVPS require 4320, 15,980, and 3460 structural anal-
yses to find their optimum solutions, respectively. Figure 6.2 shows the conver-
gence curves of the best results found by CBO, ECBO, VPS, and MDVC-UVPS.

6.3.2 A 693-Bar Double-Layer Barrel Vault

The considered 693-bar braced barrel vault consists of 259 joints and 693 members
with 23 independent design variables. The free span of the barrel vault is 19.03 m,
its rise is 5.75 m and its length is 22.9 m. The geometry and the member grouping
scheme of the structure is shown in Fig. 6.3. The structural material properties are
assumed as follows: The modulus of elasticity is taken as 29,000 ksi
(203,893.6 MPa), the yield stress of steel is assumed to be 36 ksi (253.1 MPa), and
the density of steel is 0.288 lb/in3 (7833.413 kg/m3). It is assumed that the barrel
90 6 Optimal Design of Double-Layer Barrel Vault Space Structures

Fig. 6.1 a 3D view, b plan view with group numbers c flattened cross-sectional view of the
384-bar double-layer barrel vault
6.3 Design Examples 91

Table 6.2 Performance comparison for the 384-bar braced barrel vault problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 ST 1/2 ST 1/2 ST 3/4 ST 1/2
2 EST 2 ST 2 1/2 EST 2 1/2 EST 2
3 EST 2 EST 2 EST 2 1/2 EST 2
4 ST 3 ST 1 1/2 EST 1 1/2 ST 1 1/2
5 DEST 2 1/2 EST 4 DEST 3 DEST 3
6 ST 2 1/2 ST 1 1/2 ST 1 1/2 ST 1 1/2
7 ST 12 ST 12 ST 12 ST 12
8 DEST 4 ST 10 EST 8 DEST 5
9 DEST 5 ST 12 EST 10 EST 10
10 ST 12 DEST 8 EST 10 EST 10
11 DEST 5 DEST 5 DEST 5 DEST 5
12 DEST 6 EST 8 DEST 5 ST 12
13 DEST 3 ST 6 ST 6 ST 6
14 EST 3 1/2 EST 3 1/2 DEST 3 ST 4
15 ST 2 1/2 ST 2 1/2 ST 2 1/2 EST 2 1/2
16 EST 6 ST 5 ST 5 ST 4
17 EST 6 EST 4 DEST 3 ST 6
18 EST 2 EST 1 1/2 EST 1 1/2 EST 1 1/2
19 EST 2 ST 1 1/4 ST 1 1/4 ST 1 1/4
20 EST 2 1/2 EST 1 1/2 EST 1 1/2 EST 1 1/2
21 EST 4 EST 1 1/2 EST 1 1/2 EST 1 1/2
22 ST 3 1/2 ST 1 1/4 EST 1 1/2 ST 1 1/4
23 EST 1 1/2 EST 1 1/2 EST 1 1/2 EST 1 1/2
24 ST 3 1/2 EST 2 1/2 EST 2 1/2 ST 3 1/2
25 ST 2 1/2 ST 2 1/2 EST 2 1/2 EST 2
26 DEST 4 ST 2 1/2 EST 1 1/2 EST 2
27 EST 3 DEST 2 ST 3 ST 3 1/2
28 EST 2 EST 1 1/2 EST 1 1/2 EST 2
29 ST 2 1/2 ST 2 1/2 EST 2 EST 2
30 ST 3 EST 1 1/2 EST 2 EST 2
31 ST 2 1/2 EST 1 1/2 EST 1 1/2 EST 2
Weight (lb) 69,448.52 62,486.02 62,455.30 62,735.42
(31,501.32 kg) (28,343.18 kg) (28,329.24 kg) (28,456.31 kg)
Average optimized 123,397 65,785 67,900 65,738
weight (lb) (55,971.93 kg) (29,839.57 kg) (30,798.92 kg) (29,818.26 kg)
Standard deviation on 103,837 3386 2913 2882
average weight (lb) (47,099.67 kg) (1535.86 kg) (1321.31 kg) (1307.25 kg)
92 6 Optimal Design of Double-Layer Barrel Vault Space Structures

Fig. 6.2 Convergence curves for the 384-bar barrel vault problem

vault is loaded by uniformly distributed vertical loads applied to the top of the roof,
and supports are considered at the two external edges of the top and bottom layers.
The applied loads are considered as follows: a uniform dead load (DL) pressure of
35 kg/m2; a positive wind load (WL) pressure of 160 kg/m2; and a negative wind
load (WL) pressure of 240 kg/m2. For design purposes, these loads are combined
follows:
• Load case 1: 1.5(DL + WL) = 1.5(35 + 160) = 292.5 kg/m2 (2.87 kN/m2)
• Load case 2: 1.5(DL – WL) = 1.5(35 – 240) = –307.5 kg/m2 (3.00 kN/m2)
Stress and slenderness constraints are considered according to AISC-ASD [15]
which are discussed earlier. The nodes are subjected to the displacement limits
of ± 0.1 in (0.254 cm) in every direction.
Table 6.3 presents a comparison between the obtained results. The lightest
design (i.e., 9091.1 lb (4123.65 kg)) is found by MDVC-UVPS algorithm and after
that the best design belongs to VPS (i.e., 9201.4 lb (4173.68 kg)). Figure 6.4 shows
the convergence curves of the best results found by CBO, ECBO, VPS, and
MDVC-UVPS. MDVC-UVPS converges to the optimum solution after 4120
analyses. CBO, ECBO, and VPS obtain their optimal solutions after 4400, 16,720,
and 9800 analyses, respectively. Element stress ratios at the best design optimized
by MDVC-UVPS are shown in Fig. 6.5. The maximum stress ratio is 81.12%.

6.3.3 A 1536-Bar Double-Layer Barrel Vault

The last design example is the size optimization of a 1536-bar double-layer barrel
vault with 413 joints, a span of 40 m, a length of 50 m, and 35 independent variable
6.3 Design Examples 93

Fig. 6.3 a 3D view, b plan view with group numbers of the top layer and c flattened
cross-sectional view with the group number of bracing and the bottom layer elements of the
693-bar double-layer barrel vault
94 6 Optimal Design of Double-Layer Barrel Vault Space Structures

Table 6.3 Performance comparison for the 693-bar barrel vault problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 ST 4 ST 4 EST 3 ST 4
2 ST 1 ST 1 ST 1 ST 1
3 ST 1 1/4 ST 3/4 ST 3/4 ST 3/4
4 ST 1 1/4 ST 1 ST 1 ST 1
5 ST 3/4 ST 3/4 ST 3/4 ST 3/4
6 EST 3 ST 3 ST 3 1/2 ST 3 1/2
7 ST 1 ST 1 ST 1 ST 1
8 ST 3/4 ST 1 ST 3/4 ST 1
9 ST 1 1/2 ST 1 ST 1 ST 1
10 ST 3/4 ST 3/4 ST 3/4 ST 3/4
11 ST 3 EST 2 ST 3 EST 2 1/2
12 ST 1 ST 1 1/4 EST 1 1/4 ST 1
13 ST 1 1/4 EST 2 EST 1 ST 1 1/2
14 ST 1 1/4 ST 1 ST 1 ST 1
15 ST 3/4 ST 3/4 ST 3/4 ST 3/4
16 ST 2 ST 1 EST 1 1/2 EST 1 1/4
17 ST 1 1/2 ST 1 ST 1 ST 1
18 EST 1 1/2 ST 3 EST 1 1/2 EST 2
19 ST 1 1/2 ST 1 ST 1 ST 1
20 ST 3/4 ST 3/4 EST 3/4 ST 3/4
21 ST 2 1/2 ST 3/4 ST 1 ST 1
22 ST 1 ST 3/4 ST 1 ST 1
23 ST 3/4 ST 3/4 ST 3/4 ST 3/4
Weight (lb) 10,221.8 9240.5 9201.4 9091.1
(4,636.53 kg) (4191.42 kg) (4173.68 kg) (4123.65 kg)
Average optimized 15,563 9577 9823 9475
weight (lb) (7,059.26 kg) (4344.05 kg) (4455.64 kg) (4297.79 kg)
Standard deviation on 3976 505 598 765
average weight (lb) (1803.48 kg) (229.06 kg) (271.25 kg) (347.00 kg)

groups [17]. The geometric details and member groups are presented in Fig. 6.6.
The modulus of elasticity is taken as 30,450 ksi (210,000 MPa), the yield stress of
steel is equal to 58 ksi (400 MPa), and the density of materials is considered to be
0.288 lb/in3 (7833.413 kg/m3). The supports are fixed at the two external edges of
the top layer of the structure and all joints of the top layer are subjected to con-
centrated vertical loads of 5 kips. The design constraints (including stress and
slenderness limitation) are considered according to AISC-ASD [15] which are
discussed earlier. The nodal displacements are limited to ±0.1969 in (5 mm) in
every direction.
6.3 Design Examples 95

Fig. 6.4 Convergence curves for the 693-bar barrel vault problem

Fig. 6.5 Stress ratios for the best design obtained by MDVC-UVPS for the 693-bar barrel vault
problem

Table 6.4 presents the results of the optimal designs of different optimization
algorithms. The weight of the best result obtained by MDVC-UVPS is 122,852 lb
(55,724.73 kg) that is the best among the compared methods. The average opti-
mized weight of this method is 146,229 lb (66,328.36 kg) which is less than those
of all other methods. Comparison of the convergence curves of CBO, ECBO, VPS,
and MDVC-UVPS is illustrated in Fig. 6.7. MDVC-UVPS requires 4762 structural
analyses to find the optimum solution while CBO, ECBO, and VPS require 4540,
15,060, and 18,080 structural analyses, respectively.
96 6 Optimal Design of Double-Layer Barrel Vault Space Structures

Fig. 6.6 a 3D view, b top plan view of quarter of the barrel vault with the related group numbers
and c flattened cross-sectional view of the 1536-bar double-layer barrel vault
6.3 Design Examples 97

Table 6.4 Performance comparison for the 1536-bar barrel vault problem
Element group Sections
CBO ECBO VPS MDVC-UVPS
1 DEST 3 EST 6 DEST 4 DEST 4
2 ST 4 DEST 3 ST 3 1/2 ST 4
3 ST 4 EST 3 1/2 ST 4 EST 3 1/2
4 DEST 2 1/2 EST 4 DEST 3 EST 5
5 EST 6 DEST 4 EST 6 EST 5
6 DEST 3 DEST 5 EST 5 EST 5
7 EST 1 1/4 ST 1/2 EST 1/2 ST 1/2
8 EST 2 ST 1 1/2 EST 2 ST 1 1/2
9 EST 2 1/2 ST 1 1/2 ST 3 1/2 ST 1 1/2
10 ST 2 ST 1 ST 1 1/2 ST 1 1/4
11 EST 2 EST 1 1/2 ST 1 1/4 ST 2 1/2
12 EST 4 ST 2 1/2 ST 3 EST 2 1/2
13 ST 1 1/4 EST 3 1/2 ST 3 1/2 EST 3
14 ST 5 EST 4 ST 5 DEST 3
15 ST 8 EST 2 1/2 DEST 3 ST 6
16 ST 8 DEST 2 1/2 DEST 3 ST 5
17 EST 4 ST 3 1/2 ST 3 DEST 2 1/2
18 DEST 2 1/2 EST 1 1/2 EST 2 1/2 EST 2
19 ST 3 1/2 EST 1 1/2 EST 1 1/2 EST 2
20 EST 2 ST 1 1/2 ST 2 1/2 ST 1 1/2
21 DEST 2 1/2 ST 1 1/2 EST 1 1/2 ST 1 1/2
22 EST 2 ST 1 1/2 EST 2 ST 1 1/2
23 ST 2 1/2 ST 1 1/2 ST 1 1/2 EST 2
24 ST 1 1/2 ST 1 1/2 DEST 2 1/2 ST 1 1/2
25 EST 2 1/2 ST 1 1/2 ST 1 1/2 EST 2
26 ST 3 1/2 DEST 2 ST 3 ST 2 1/2
27 EST 2 EST 1 1/2 ST 1 1/2 EST 1 1/2
28 EST 2 ST 1 1/2 ST 1 1/2 ST 1 1/2
29 EST 1 1/2 EST 1 1/2 EST 2 ST 1 1/2
30 ST 2 ST 1 1/4 ST 2 EST 1 1/2
31 EST 1 1/2 ST 2 1/2 EST 1 1/2 EST 1 1/2
32 ST 3 1/2 ST 1 1/4 EST 1 1/2 ST 2
33 EST 2 EST 2 ST 1 1/2 EST 1 1/2
34 ST 2 ST 1 1/2 EST 1 1/2 ST 2
35 EST 1 1/2 EST 1 1/2 ST 1 1/2 ST 2
Weight (lb) 152,229 128,111 129,117 122,852
(69,049.91 kg) (58,110.17 kg) (58,566.49 kg) (55,724.73 kg)
Average optimized 215,621 149,002 147,089 146,229
weight (lb) (97,804.04 kg) (67,586.17 kg) (66,718.45 kg) (66,328.36 kg)
Standard deviation on 36,322 16,775 14,644 14,552
average weight (lb) (16,475.38 kg) (7,609.01 kg) (6,642.41 kg) (6600.68 kg)
98 6 Optimal Design of Double-Layer Barrel Vault Space Structures

Fig. 6.7 Convergence curves for the 1536-bar barrel vault problem

6.4 Concluding Remarks

Barrel vaults are effective semicylindrical structural systems that are commonly
used to provide long-span and economical roofs for multipurpose facilities
including warehouses, rail stations, pools, sport centers, airplane hungers, and
community centers.
In this chapter, three barrel vault optimization problems with discrete variables
are considered under stress, stability, and displacement limitations. In the 384-bar
barrel vault problem, the design found by the ECBO, VPS, and MDVC-UVPS are
approximately identical and are about 10% lighter than the result obtained by CBO.
In the 693-bar barrel vault problem, the design obtained by MDVC-UVPS is 12.4,
1.6, and 1.2% lighter than the best design achieved by the CBO, ECBO, and VPS,
respectively. These values are 24, 4.3, and 5 for the 1536-bar barrel vault problem.
A suitable average optimized weight and also an acceptable standard deviation from
the mean value of the independent runs show the robustness of the proposed
algorithms. Also, comparison of the convergence curves indicates that the
MDVC-UVPS comes close to the optimum design rapidly.

References

1. Makowski ZS (2006) Analysis, design and construction of braced barrel vaults. Elsevier
Applied Science Publishers, London, New York
2. Kaveh A, Mirzaei B, Jafarvand A (2014) Optimal design of double layer barrel vaults using
improved magnetic charged system search. Asian J Civil Eng 15(1):135–154
3. Kaveh A, Moradveisi M (2016) Optimal design of double-layer barrel vaults using CBO and
ECBO algorithms. Iran J Sci Technol, Trans Civil Eng 40(3):167–178
References 99

4. Hasançebi O, Kazemzadeh Azad S (2015) Adaptive dimensional search: a new metaheuristic

algorithm for discrete truss sizing optimization. Comput Struct 154:1–16
5. Kaveh A, Eftekhar B (2012) Optimal design of double layer barrel vaults using an improved
hybrid big bang-big crunch method. Asian J Civil Eng 13:465–487
6. Kaveh A, Farahani M, Shojaei N (2012) Optimal design of barrel vaults using charged search
system. Int J Civil Eng 10:301–308
7. Hasançebi O, Çarbaş S (2011) Ant colony search method in practical structural optimization.
Int J Optim Civil Eng 1:91–105
8. Hasançebi O, Çarbas S, Saka MP (2011) A reformulation of the ant colony optimization
algorithm for large scale structural optimization. In: Proceedings of the second international
conference on soft computing technology in civil, structural and environmental engineering,
Civil-Comp Press, Stirlingshire
9. Hasançebi O, Kazemzadeh Azad S (2013) Reformulations of big bang-big crunch algorithm
for discrete structural design optimization. World Acad Sci Eng Technol 74
10. Hasançebi O, Kazemzadeh Azad S (2013) Discrete size optimization of steel trusses using a
refined big bang–big crunch algorithm. Eng Optimiz 46(1):61–83
11. Kaveh A, Mahdavi VR (2014) Colliding bodies optimization: a novel meta-heuristic method.
Comput Struct 139:18–27
12. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies optimization for design
problems with continuous and discrete variables. Adv Eng Softw 77:66–75
13. Kaveh A, Ilchi Ghazaan M (2016) Vibrating particles system algorithm for truss optimization
with multiple natural frequency constraints. Acta Mech 228(1):307–332
14. Kaveh A, Ilchi Ghazaan M (2018) A new hybrid meta-heuristic algorithm for optimal design
of large-scale dome structures. Eng Optimiz 50(2):235–252
15. American Institute of Steel Construction (AISC) (1989) Manual of steel construction:
allowable stress design, Chicago, USA
16. American Institute of Steel Construction (AISC) (1994) Manual of steel construction load
resistance factor design, Chicago, USA
17. Kaveh A, Mahjoubi S (2018) Optimum design of double-layer barrel vaults by lion pride
optimization algorithm. Structures 13:213–229
Chapter 7
Optimal Design of Dome-Shaped
Trusses

7.1 Introduction

Domes are one of the oldest and well-established structural forms and have been
used in architecture since the earliest times. These structures are of special interest
to engineers as they enclose large spaces with small surfaces and have proven to be
very economical in terms of consumption of constructional materials [1]. The main
aim of this chapter is frequency constraint optimization of dome truss structures;
however, all the domes are also optimized considering strength, stability, and
displacement constraints. Structural optimization considering natural frequency
constraints is believed to represent nonlinear and non-convex search spaces with
several local optima [2]. In this class of problems, large generalized eigenproblems
should be solved in order to find the natural frequencies of the structure. The size of
the structure affects the dimensions of the matrices involved and thus the required
computational time and effort. On the other hand, as the number of optimization
variables increases, more and more structural analyses are needed to be performed
in order to reach a near-optimal solution [3].
Structural optimization considering natural frequency constraints has been
studied since the 1980s [2] and was approached using mathematical programming
and meta-heuristic algorithms. Lin et al. [4] studied the minimum weight design of
structures under simultaneous static and dynamic constraints proposing a bi-factor
algorithm based on the Kuhn–Tucker criteria. Konzelman [5] considered the
problem using some dual methods and approximation concepts for structural
optimization. Grandhi and Venkayya [6] utilized an optimality criterion based on
uniform Lagrangian density for resizing and scaling procedure to locate the con-
straint boundary. Wang et al. [7] proposed an optimality criteria algorithm for
combined sizing–layout optimization of three-dimensional truss structures. In this
method, the sensitivity analysis helps to determine the search direction and the
optimal solution is achieved gradually from an infeasible starting point with a
minimum weight increment, and the structural weight is indirectly minimized.

© Springer International Publishing AG, part of Springer Nature 2018 101

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_7
102 7 Optimum Design of Dome-Shaped Trusses

Sedaghati [8] utilized a new approach using combined mathematical programming

based on the Sequential Quadratic Programming (SQP) technique and a finite
element solver based on the integrated force method. Lingyun et al. [9] combined
the simplex search method and the Niche Genetic Hybrid Algorithm (NGHA) for
mass minimization of structures with frequency constraints. Gomes [10] used the
Particle Swarm Optimization (PSO) algorithm to study simultaneous layout and
sizing optimization of truss structures with multiple frequency constraints. Kaveh
and Zolghadr [11] hybridized Charged System Search and Big Bang-Big Crunch
with trap recognition capability (CSS-BBBC) to solve layout and sizing opti-
mization problems of truss structures with natural frequency constraints. Miguel
and Fadel Miguel [12] employed Harmony Search (HS) and Firefly Algorithm
(FA) to study simultaneous layout and sizing optimization of truss structures with
multiple frequency constraints. A hybrid Optimality Criterion (OC) and Genetic
Algorithm (GA) method was used by Zuo et al. [13] for truss optimization with
frequency constraints. Kaveh and Javadi [14] utilized a hybridization of Harmony
search, Ray optimizer, and Particle Swarm Optimization algorithm (HRPSO) for
weight minimization of trusses under multiple natural frequency constraints. Kaveh
and Ilchi Ghazaan [15] employed Particle Swarm Optimization with an Aging
Leader and Challengers (ALC-PSO) and HALC-PSO that transplants a harmony
search-based mechanism to ALC-PSO as a variable constraint handling approach to
optimize truss structures with frequency constraints. Hosseinzadeh et al. [16] used
hybrid Electromagnetism-like mechanism algorithm and Migration Strategy (EM–
MS) for layout and size optimization of truss structures with multiple frequency
constraints.

7.2 Frequency Constraint Optimization Problem

In this chapter, the main aim of the optimization problem is to minimize the weight
of the structure under multiple frequency constraints while the design variables are
only the cross-sectional areas of bars (sizing optimization). Each variable should be
chosen within a permissible range.
The mathematical formulation can be expressed as follows:

Find fXg ¼ ½x1 ; x2 ; . . .; xng 

Png P
nmðiÞ
to minimize WðfXgÞ ¼ xi qj Lj
8 i¼1 j¼1
ð7:1Þ
< xj  xj
subjected to: x  xk
: k
xi min  xi  xi max

where {X} is a vector containing the design variables; W({X}) presents the weight of
the structure; ng is the number of design groups; nm(i) is the number of members for
7.2 Frequency Constraint Optimization Problem 103

the ith group; qj and Lj denote the material density and the length of the jth member,
respectively. xj is the jth natural frequency of the structure and x*j is its upper bound;
xk is the kth natural frequency of the structure and x*k is its lower bound; ximin and
ximax are the lower and upper bounds of the design variable xi, respectively.
For constraint handling, a penalty approach is utilized. For this purpose, the
objective function (Eq. 7.1) is redefined as follows:

PðfXgÞ ¼ ð1 þ e1  tÞe2  WðfXgÞ ð7:2Þ

where P({X}) is the penalized cost function or the objective function to be mini-
mized and t denotes the sum of the violations of the design constraints. In this
chapter, e1 is set to unity and e2 is calculated by
iter
e2 ¼ 1:5 þ 1:5  ð7:3Þ
itermax

where iter is the current iteration number and itermax is the total number of iterations
for optimization process.

7.3 Design Examples

The efficiencies of CBO, ECBO, VPS, and MDVC-UVPS are studied through three
dome truss structures. These examples contain
• A 600-bar dome truss
• A 1180-bar dome truss
• A 1410-bar dome truss
Two constraint cases are considered for each example. In Case 1, natural fre-
quency constraints are incorporated. In Case 2, limitation on stresses and stability of
truss elements are considered according to the provisions of the ASD-AISC [17] as
follows.
The allowable tensile stresses for tension members are calculated by

riþ ¼ 0:6Fy ð7:4Þ

where Fy stands for the yield strength.

The allowable stress limits for compression members are calculated depending
on two possible failure modes of the members known as elastic and inelastic
buckling. Thus
8 h  i  
>
< 1  2C k2i k3i
2 Fy = 3 þ 8C  8C 3
5 3ki
for ki \Cc
c
r
c c
i ¼ ð7:5Þ
>
: 12p22E for k  C
23k i c
i
104 7 Optimum Design of Dome-Shaped Trusses

where E is the modulus of elasticity; ki is the slenderness ratio ðki ¼ kli =ri Þ; Cc
denotes the slenderness ratio dividing the elastic and inelastic buckling regions
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(Cc ¼ 2p2 E=Fy ); k is the effective length factor (k is set to 1 for all truss
members); li is the member length; and ri is the minimum radius of gyration.
In this design code provisions, the maximum slenderness ratio is limited to 300
for tension members, and this limit is recommended to be 200 for compression
members. Nodal displacements in all the coordinate directions must be less
than ±8 cm.

7.3.1 A 600-Bar Dome Truss

The 600-bar single-layer dome structure is schematized in Fig. 7.1. The entire
structure is composed of 216 nodes and 600 elements. Figure 7.2 shows a sub-
structure in more detail for nodal numbering and coordinates. Each of the elements
of this substructure is considered as a design variable. Thus, this is a size opti-
mization problem with 25 variables. The elastic modulus is 200 GPa and the
material density is 7850 kg/m3 for all elements. The yield stress of steel is taken as
400 MPa. A nonstructural mass of 100 kg is attached to all free nodes. The dome is
considered to be subjected to vertical loading at all the unsupported joints for the
second constraint case. According to symmetry, one only needs to define loading
conditions of the labeled nodes shown in Fig. 7.2. These loads are −100, −500,
−1500, −2500, −3500, −4500, −5500, −5000, and 0 kN, respectively. The
allowable minimum and maximum cross-sectional areas of all elements are set to
1  10−4 and 100  10−4 m2, respectively. There are two constraints on the

Fig. 7.1 Schematic of the 600-bar dome truss

7.3 Design Examples 105

Fig. 7.2 Details of a substructure of the 600-bar dome truss

natural frequencies, which are x1  5 and x3  7 Hz. In order to optimize this

structure by MDVC-UVPS, two stages are considered. The design variable con-
figuration utilized for the first stage is listed as follows: [7 10 13], [16 19 22], [9 12
15], [18 21 24], [11 14 17], [20 23 25], [1 3 4 6], and [2 5 8]. Thus eight variables
are used in the first stage.

7.3.1.1 Constraint Case 1

Table 7.1 presents the optimum designs obtained by CBO, ECBO [18], VPS [19],
and MDVC-UVPS [20]. It can be seen that the best design was found by
MDVC-UVPS. Also, the average optimized weight and standard deviation on
average weight of this method are the lowest which are 6119.95 and 16.23 kg.
Table 7.2 reports the natural frequencies of the optimized structures and it is clear
that none of the frequency constraints are violated. Convergence histories for these
techniques are depicted in Fig. 7.3. The required structural analyses to achieve the
best designs by the CBO, ECBO, VPS, and MDVC-UVPS, respectively, are
17,940, 19,020, 25,040, and 17,513 analyses, respectively. It should be noted that
MDVC-UVPS obtained the best design of VPS after 10,295 analyses.
106 7 Optimum Design of Dome-Shaped Trusses

Table 7.1 Performance comparison for the 600-bar dome truss problem—constraint case 1
Element number (nodes) Areas (cm2)
CBO ECBO VPS [19] MDVC-UVPS
[18] [20]
1 (1–2) 1.2404 1.4305 1.3155 1.2575
2 (1–3) 1.3797 1.3941 1.2299 1.3466
3 (1–10) 5.2597 5.5293 5.5506 4.9738
4 (1–11) 1.2658 1.0469 1.3867 1.4025
5(2–3) 17.2255 16.9642 17.4275 17.3802
6 (2–11) 38.2991 35.1892 40.1430 37.9742
7(3–4) 12.2234 12.2171 12.8848 13.0306
8 (3–11) 15.4712 16.7152 15.5413 15.9209
9 (3–12) 11.1577 12.5999 12.2428 11.9419
10 (4–5) 9.4636 9.5118 9.3776 9.1643
11 (4–12) 8.8250 8.9977 8.6684 8.4332
12 (4–13) 9.1021 9.4397 9.1659 9.2375
13 (5–6) 6.8417 6.8864 7.1664 7.2213
14 (5–13) 5.2882 4.2057 5.2170 5.2142
15 (5–14) 6.7702 7.2651 6.5346 6.7961
16 (6–7) 5.1402 6.1693 5.4741 5.2078
17 (6–14) 5.1827 3.9768 3.6545 3.4586
18 (6–15) 7.4781 8.3127 7.6034 7.6407
19 (7–8) 4.5646 4.1451 4.2251 4.3690
20 (7–15) 1.8617 2.4042 1.9717 2.1237
21 (7–16) 4.8797 4.3038 4.5107 4.5774
22 (8–9) 3.5065 3.2539 3.5251 3.4564
23 (8–16) 2.4546 1.8273 1.9255 1.7920
24 (8–17) 4.9128 4.8805 4.7628 4.8264
25 (9–17) 1.2324 1.5276 1.6854 1.7601
Weight (kg) 6182.01 6171.51 6120.01 6115.10
Average optimized weight (kg) 6226.37 6191.50 6158.11 6119.95
Standard deviation on average 60.12 39.08 28.49 16.23
weight (kg)

Table 7.2 Natural frequencies (Hz) evaluated at the optimum designs of the 600-bar dome truss
problem
Frequency number Natural frequencies (Hz)
CBO ECBO [18] VPS [19] MDVC-UVPS [20]
1 5.000 5.002 5.000 5.000
2 5.000 5.003 5.000 5.000
3 7.000 7.001 7.000 7.000
4 7.000 7.001 7.000 7.000
5 7.001 7.002 7.000 7.000
7.3 Design Examples 107

Fig. 7.3 Convergence curves for the 600-bar dome truss problem—constraint case 1

7.3.1.2 Constraint Case 2

The results found by the proposed algorithms are summarized in Table 7.3.
MDVC-UVPS achieves the lightest design (i.e., 7338.37 kg). It also has better
performance in terms of the average optimized weight and the standard deviation on
average weight, which are 7694.13 and 122.67 kg, respectively. The best designs
obtained by CBO, ECBO, and VPS are 7464.76, 7463.78, and 7379.63 kg,
respectively. The maximum stress ratios for the best designs of CBO, ECBO, VPS,
and MDVC-UVPS are 99.99, 98.89, 99.99, and 99.99%, respectively. The required
numbers of structural analyses to achieve the best design by CBO, ECBO, VPS,
and MDVC-UVPS are 18,520, 16,400, 18,720, and 15,406 analyses, respectively.

7.3.2 A 1180-Bar Dome Truss

For the second example, size optimization of a 1180-bar dome truss structure is
considered. The configuration of the structure is depicted in Fig. 7.4. The structure
consists of 400 nodes and 1180 elements. A substructure is illustrated in Fig. 7.5 in
more detail for nodal numbering. Each of the elements of this substructure is
considered as a design variable. Thus, this is a size optimization problem with 59
variables. Table 7.4 summarizes the coordinates of the nodes in Cartesian coordi-
nate system. The elastic modulus is 200 GPa and the material density is 7850 kg/
m3 for all elements. The yield stress of steel is taken as 400 MPa. A nonstructural
mass of 100 kg is attached to all free nodes. The dome is considered to be subjected
to vertical loading at all the unsupported joints for the second constraint case.
According to symmetry, only we need to define loading conditions of labeled nodes
108 7 Optimum Design of Dome-Shaped Trusses

Table 7.3 Performance comparison of the 600-bar dome truss problem—constraint case 2
Element number (nodes) Areas (cm2)
CBO ECBO VPS MDVC-UVPS
1 (1–2) 1.0007 1.1545 1.1552 1.0023
2 (1–3) 2.6144 4.6875 4.5998 4.5959
3 (1–10) 1.0267 1.4021 1.1522 5.368
4 (1–11) 1.0036 1.0025 1.0004 1.0208
5(2–3) 4.5976 4.8782 4.611 4.5967
6 (2–11) 1.1161 2.7684 1.3882 1.3896
7(3–4) 4.752 5.1739 4.752 4.8037
8 (3–11) 4.7657 4.7672 5.0116 4.7688
9 (3–12) 4.6437 1.4847 1.5148 1.4038
10 (4–5) 5.0566 5.7881 5.4851 5.0869
11 (4–12) 5.8028 5.7286 5.5706 5.576
12 (4–13) 8.041 3.8644 3.4961 3.4982
13 (5–6) 6.5194 6.2631 6.0269 6.0143
14 (5–13) 7.588 6.8744 6.8286 7.0211
15 (5–14) 4.688 4.7951 4.686 4.6903
16 (6–7) 7.0441 7.0731 7.0437 7.0399
17 (6–14) 8.463 8.5365 9.257 8.5434
18 (6–15) 5.7627 5.7799 5.8364 5.7635
19 (7–8) 8.5035 8.814 8.5035 8.5032
20 (7–15) 10.9017 10.9549 10.839 10.8387
21 (7–16) 7.7492 7.8738 7.7491 7.7481
22 (8–9) 6.1231 6.2894 6.0047 6.0053
23 (8–16) 13.7034 13.6983 13.6386 13.6388
24 (8–17) 5.4526 5.4853 5.4506 5.4509
25 (9–17) 12.4331 12.5248 12.4329 12.4327
Weight (kg) 7464.76 7463.78 7379.63 7338.37
Average optimized weight (kg) 8093.16 7739.41 7701.80 7694.13
Standard deviation on average weight 934.45 154.61 487.09 122.67
(kg)

shown in Fig. 7.5. These loads are −1000, −2500, −3500, −3500, −3500, −2000,
−1000, −100, −500, 0, −2000, −3000, −3500, −3500, −2500, −1500, −100, −200,
−500, and −200 kN, respectively. The allowable minimum and maximum
cross-sectional areas of all elements are set to 1  10−4 and 100  10−4 m2,
respectively. There are two constraints on the natural frequencies, which are
x1  7 and x3  9 Hz. This problem is optimized in three stages by
MDVC-UVPS. The numbers of design variables considered in the first and second
stages are 8 and 18, respectively. These DVCs contain
7.3 Design Examples 109

Fig. 7.4 Schematic of the 1180-bar dome truss

Fig. 7.5 Details of a substructure of the 1180-bar dome truss

110 7 Optimum Design of Dome-Shaped Trusses

Table 7.4 Coordinates of the nodes of the 1180-bar dome truss problem
Node number Coordinates (x, y, z) Node number Coordinates (x, y, z)
1 (3.1181, 0.0, 14.6723) 11 (4.5788, 0.7252, 14.2657)
2 (6.1013, 0.0, 13.7031) 12 (7.4077, 1.1733, 12.9904)
3 (8.8166, 0.0, 12.1354) 13 (9.9130, 1.5701, 11.1476)
4 (11.1476, 0.0, 10.0365) 14 (11.9860, 1.8984, 8.8165)
5 (12.9904, 0.0, 7.5000) 15 (13.5344, 2.1436, 6.1013)
6 (14.2657, 0.0, 4.6358) 16 (14.4917, 2.2953, 3.1180)
7 (14.9179, 0.0, 1.5676) 17 (14.8153, 2.3465, 0.0)
8 (14.9179, 0.0, −1.5677) 18 (14.4917, 2.2953, −3.1181)
9 (14.2656, 0.0, −4.6359) 19 (13.5343, 2.1436, −6.1014)
10 (12.9903, 0.0, −7.5001) 20 (3.1181, 0.0, 13.7031)

Stage 1: [1 6 11 15 19]; [23 27 31 35]; [4 10 14 18 22]; [26 30 34 38 40]; [2 7 41

42 8 12 43 44 13 16 45 46]; [17 20 47 48 21 24 49 50 25 28 51 52]; [29 32 53 54 33
36 55 56 37 39 57 58]; [3 5 9 59].
Stage 2: [1 6 11]; [15 19]; [23 27]; [31 35]; [4 10]; [14 18 22]; [26 30]; [34 38
40]; [2 7 41 42]; [8 12 43 44]; [13 16 45 46]; [17 20 47 48]; [21 24 49 50]; [25 28
51 52]; [29 32 53 54]; [33 36 55 56]; [37 39 57 58]; [3 5 9 59].

7.3.2.1 Constraint Case 1

Table 7.5 presents a comparison between the results of the optimal designs.
MDVC-UVPS obtained the lightest design compared to other methods that is
37,451.77 kg. Moreover, the average optimized weight and the standard deviation
on average weight of MDVC-UVPS are less than those of all other methods which
are 37,545.53 and 64.85 kg. Table 7.6 shows the optimized structural frequencies
(Hz) for various methods. None of the frequency constraints were violated. The
MDVC-UVPS algorithm requires 19,391 structural analyses to find the optimum
solution while CBO, ECBO, and VPS require 19,960, 19,860, and 24,780 structural
analyses, respectively. MDVC-UVPS obtained the best design of ECBO after 5873
analyses. The amount of saving in structural analyses at each iteration of the
MDVC-UVPS is shown in Fig. 7.6.

7.3.2.2 Constraint Case 2

Table 7.7 lists the optimal designs found by different methods. MDVC-UVPS
obtained the lightest design compared to other methods that is 17,909.10 kg.
Moreover, the average optimized weight and the standard deviation on average
weight of MDVC-UVPS (18,417.05 and 427.44 kg) are less than those of the other
methods. The best designs found by the CBO, ECBO, and VPS are 19,869.30,
7.3 Design Examples 111

Table 7.5 Performance comparison for the 1180-bar dome truss problem—constraint case 1
Element number (nodes) Areas (cm2)
CBO ECBO [18] VPS [20] MDVC-UVPS
[20]
1 (1–2) 13.0171 7.6678 6.8743 7.3691
2 (1–11) 10.4346 11.1437 10.0230 9.3399
3 (1–20) 3.0726 1.8520 4.4140 2.7203
4 (1–21) 12.6969 14.5563 13.5515 13.2822
5 (1–40) 3.5654 4.9499 1.8303 3.6758
6 (2–3) 6.5190 6.8095 7.0824 6.1391
7 (2–11) 7.4233 6.6803 6.3960 7.0964
8 (2–12) 6.3471 6.7889 6.5646 6.0208
9 (2–20) 2.3013 1.0630 2.3705 2.1225
10 (2–22) 12.1936 9.1602 13.2621 12.3488
11 (3–4) 7.2877 6.9891 7.0922 6.8578
12 (3–12) 7.0961 6.9881 6.8079 5.7773
13 (3–13) 6.5669 6.9555 6.3815 6.9931
14 (3–23) 7.8257 7.5443 7.3122 7.3355
15 (4–5) 8.6812 9.5431 8.7221 10.5464
16 (4–13) 5.7888 6.9123 6.3680 6.9589
17 (4–14) 21.1342 8.9891 7.3159 8.0977
18 (4–24) 10.0502 6.8926 11.5749 7.7738
19 (5–6) 12.9279 12.6128 14.7985 12.4614
20 (5–14) 9.3212 8.1983 5.5174 7.8154
21 (5–15) 10.1260 11.8358 15.7381 10.2039
22 (5–25) 10.1358 9.7321 8.3419 8.9262
23 (6–7) 15.8585 19.1650 17.5000 16.5275
24 (6–15) 9.9672 10.4682 10.3084 9.0166
25 (6–16) 14.8493 14.1178 15.1958 13.8204
26 (6–26) 11.4909 11.14567 10.9395 11.4021
27 (7–8) 26.2359 23.4125 24.9421 24.2631
28 (7–16) 13.8812 15.5167 13.9614 14.5494
29 (7–17) 18.8857 16.6613 18.4153 17.7753
30 (7–27) 14.0257 15.9631 14.4945 15.4594
31 (8–9) 33.8826 37.0532 36.3529 34.1372
32 (8–17) 25.7142 22.2937 19.6608 19.1254
33 (8–18) 24.8644 22.7409 23.7259 24.1954
34 (8–28) 19.8498 23.5624 22.0297 21.5899
35 (9–10) 53.2630 47.7652 47.3286 49.4717
36 (9–18) 22.7771 22.5066 22.9442 26.2915
37 (9–19) 35.4230 34.6418 30.8229 33.7558
38 (9–29) 57.5480 31.6492 33.1098 29.7608
(continued)
112 7 Optimum Design of Dome-Shaped Trusses

Table 7.5 (continued)

Element number (nodes) Areas (cm2)
CBO ECBO [18] VPS [20] MDVC-UVPS
[20]
39 (10–19) 35.1385 32.7268 32.5526 34.0489
40 (10–30) 10.7300 1.05206 1.7363 1.0024
41 (11–21) 9.2401 11.3681 11.5271 9.0344
42 (11–22) 5.2661 6.5512 8.4571 7.5316
43 (12–22) 6.2415 6.3619 5.4136 6.3726
44 (12–23) 4.4768 5.9296 7.1832 5.7643
45 (13–23) 8.8846 7.8739 5.4066 6.7270
46 (13–24) 7.3710 6.2794 6.2534 6.7021
47 (14–24) 8.2595 7.6206 6.9383 7.8082
48 (14–25) 7.6091 7.2937 10.6872 8.1225
49 (15–25) 11.3030 10.5783 12.8005 10.1777
50 (15–26) 13.8381 10.1173 10.2216 10.1825
51 (16–26) 13.3654 15.1088 11.5330 13.4590
52 (16–27) 13.1836 12.8251 11.6918 13.9788
53 (17–27) 13.5793 17.4375 20.7566 18.1070
54 (17–28) 10.0628 20.1153 18.1341 19.2212
55 (18–28) 24.1197 24.2121 28.2882 23.4359
56 (18–29) 24.2604 23.3175 24.2023 27.6479
57 (19–29) 34.1389 34.6196 48.0180 33.6805
58 (19–30) 38.0340 35.2970 35.6517 35.7035
59 (20–40) 2.6689 8.8569 5.5956 4.7617
Weight (kg) 40,985 37,984.39 38,699.14 37,451.77
Average optimized weight 42,019.10 38,042.15 38,861.82 37,545.53
(kg)
Standard deviation on average 655.72 101.43 385.41 64.85
weight (kg)

Table 7.6 Natural frequencies (Hz) evaluated at the optimum designs of the 1180-bar dome truss
problem
Frequency number Natural frequencies (Hz)
CBO ECBO [18] VPS [20] MDVC-UVPS [20]
1 7.000 7.000 7.000 7.000
2 7.001 7.001 7.001 7.001
3 9.000 9.000 9.000 9.000
4 9.000 9.000 9.000 9.000
5 9.005 9.064 9.177 9.005
7.3 Design Examples 113

Fig. 7.6 Saving in structural analyses using the MDVC-UVPS algorithm in the 1180-bar dome
truss problem—constraint case 1

18,860.43, and 18,903.65 kg, respectively. The maximum stress ratios for the best
designs of CBO, ECBO, VPS, and MDVC-UVPS are 81.11, 86.52, 93.50, and
99.99%, respectively. Convergence history diagrams are depicted in Fig. 7.7.
MDVC-UVPS requires 14,456 structural analyses to find the optimum solution
while CBO, ECBO, and VPS require 18,220, 19,740, and 19,880 structural anal-
yses, respectively.

Table 7.7 Performance comparison of the 1180-bar dome truss problem—constraint case 2
Element number (nodes) Areas (cm2)
CBO ECBO VPS MDVC-UVPS
1 (1–2) 4.8705 5.9493 5.7015 9.4521
2 (1–11) 7.3783 5.1128 7.2102 3.8492
3 (1–20) 1.2699 1.7992 1.0055 8.1978
4 (1–21) 6.1376 5.4755 3.4166 4.2423
5 (1–40) 3.542 4.6291 3.8978 8.1978
6 (2–3) 6.1148 7.084 5.1114 9.4521
7 (2–11) 5.4713 6.473 5.064 3.8492
8 (2–12) 9.4118 7.1685 5.1467 4.8473
9 (2–20) 9.6458 11.6397 9.7255 8.1978
10 (2–22) 4.7464 6.5289 4.6638 4.2423
11 (3–4) 9.2156 9.4237 9.7013 9.4521
12 (3–12) 6.9436 5.8957 10.1884 4.8473
13 (3–13) 5.1492 5.4008 8.9843 5.5341
14 (3–23) 4.0869 5.4558 4.5727 7.112
(continued)
114 7 Optimum Design of Dome-Shaped Trusses

Table 7.7 (continued)

Element number (nodes) Areas (cm2)
CBO ECBO VPS MDVC-UVPS
15 (4–5) 11.5998 8.9965 11.7971 9.4449
16 (4–13) 8.1323 6.7849 6.6183 5.5341
17 (4–14) 6.194 8.0975 7.0562 7.4661
18 (4–24) 5.8227 6.7986 6.3479 7.112
19 (5–6) 11.8302 10.7667 14.1573 9.4449
20 (5–14) 7.7969 7.5666 8.1737 7.4661
21 (5–15) 8.2957 7.9201 7.4761 7.4775
22 (5–25) 7.5279 7.3118 7.3921 7.112
23 (6–7) 12.7894 11.8195 10.3913 10.9878
24 (6–15) 7.3903 7.4026 7.0689 7.4775
25 (6–16) 8.0166 7.7645 8.2977 7.4749
26 (6–26) 8.5456 8.5533 8.2033 8.7224
27 (7–8) 12.2363 12.0751 11.6556 10.9878
28 (7–16) 8.1607 8.2215 9.6521 7.4749
29 (7–17) 8.6099 4.9116 4.4503 7.5663
30 (7–27) 9.9546 10.417 10.9239 8.7224
31 (8–9) 17.7962 17.8845 11.1662 11.0094
32 (8–17) 9.4484 7.9859 7.7067 7.5663
33 (8–18) 6.1548 5.2099 4.5083 4.1094
34 (8–28) 10.1519 8.8044 8.8879 8.7309
35 (9–10) 12.0321 11.7072 11.3688 11.0094
36 (9–18) 4.0536 5.4377 6.8448 4.1094
37 (9–19) 4.9601 8.8425 4.6168 3.8898
38 (9–29) 8.42 8.5341 8.3597 8.7309
39 (10–19) 9.6607 3.7971 3.7932 3.8898
40 (10–30) 1.0388 1.0996 1.0000 8.7309
41 (11–21) 5.25 5.9873 5.7214 3.8492
42 (11–22) 10.7763 4.4572 4.1908 3.8492
43 (12–22) 13.2206 5.2111 6.1135 4.8473
44 (12–23) 10.7522 10.7124 7.8898 4.8473
45 (13–23) 5.5857 8.5213 5.6613 5.5341
46 (13–24) 8.4136 5.7257 5.6122 5.5341
47 (14–24) 6.1712 6.9778 6.5029 7.4661
48 (14–25) 8.5468 7.2017 10.5263 7.4661
49 (15–25) 9.1846 7.4758 9.4067 7.4775
50 (15–26) 8.6247 8.028 7.2511 7.4775
51 (16–26) 8.0938 8.5583 7.949 7.4749
52 (16–27) 8.7015 8.3535 9.3917 7.4749
53 (17–27) 4.5706 4.6425 7.7928 7.5663
(continued)
7.3 Design Examples 115

Table 7.7 (continued)

Element number (nodes) Areas (cm2)
CBO ECBO VPS MDVC-UVPS
54 (17–28) 7.623 7.729 7.6062 7.5663
55 (18–28) 4.2279 5.3966 7.1012 4.1094
56 (18–29) 5.8307 5.1474 8.1847 4.1094
57 (19–29) 6.6752 7.3322 7.8282 3.8898
58 (19–30) 5.1712 5.5059 6.0886 3.8898
59 (20–40) 17.8857 2.7226 8.2244 8.1978
Weight (kg) 19,869.30 18,860.43 18,903.65 17,909.10
Average optimized 22,973.72 19,768.86 20,483.73 18,417.05
weight (kg)
Standard deviation on 3793.88 1195.92 1033.38 427.44
average weight (kg)

Fig. 7.7 Convergence curves for the 1180-bar dome truss problem—constraint case 2

7.3.3 A 1410-Bar Dome Truss

Figure 7.8 shows the 1410-bar double-layer dome truss structure. The entire
structure is composed of 390 nodes and 1410 elements. Figure 7.9 shows a sub-
structure in more detail for nodal numbering. Each of the elements of this sub-
structure is considered as a design variable. Thus, this is a size optimization
problem with 47 variables. Table 7.8 presents the coordinates of the nodes in
Cartesian coordinate system. The elastic modulus is 200 GPa and the material
density is 7850 kg/m3 for all elements. The yield stress of steel is taken as
400 MPa. A nonstructural mass of 100 kg is attached to all free nodes. The dome is
116 7 Optimum Design of Dome-Shaped Trusses

Fig. 7.8 Schematic of the 1410-bar dome truss

Fig. 7.9 Details of a substructure of the 1410-bar dome truss

considered to be subjected to vertical loading at all the unsupported joints for the
second constraint case. According to symmetry, only we need to define loading
conditions of labeled nodes shown in Fig. 7.9. These loads are −200, −600, −1000,
−1500, −2000, −2500, 0, −400, −1000, −1200, −1500, −2000, and −1000 kN,
respectively. The allowable minimum and maximum cross-sectional areas of all
elements is set to 1  10−4 and 100  10−4 m2, respectively. There are two
7.3 Design Examples 117

Table 7.8 Coordinates of the nodes of the 1410-bar double-layer dome truss problem
Node number Coordinates (x, y, z) Node number Coordinates (x, y, z)
1 (1.0, 0.0, 4.0) 8 (1.989, 0.209, 3.0)
2 (3.0, 0.0, 3.75) 9 (3.978, 0.418, 2.75)
3 (5.0, 0.0, 3.25) 10 (5.967, 0.627, 2.25)
4 (7.0, 0.0, 2.75) 11 (7.956, 0.836, 1.75)
5 (9.0, 0.0, 2.0) 12 (9.945, 1.0453, 1.0)
6 (11.0, 0.0, 1.25) 13 (11.934, 1.2543, −0.5)
7 (13.0, 0.0, 0.0)

constraints on the natural frequencies, which are x1  7 and x3  9 Hz. Three

stages with 10, 20, and 47 variables are considered to optimize this example by
MDVC-UVPS. These DVCs contain
Stage 1: [3 7 11 15 19 23]; [1 4 8 12 16 20]; [28 32 36 40 44 47]; [25 29 33 37 41];
[2 26 5 27]; [6 30 9 31]; [10 34 13 35]; [14 38 17 39]; [18 42 21 43]; [22 45 24 46].
Stage 2: [3 7 11]; [15 19 23]; [1 4 8]; [12 16 20]; [28 32]; [36 40 44 47]; [25 29];
[33 37 41]; [2 26]; [5 27]; [6 30]; [9 31]; [10 34]; [13 35]; [14 38]; [17 39]; [18 42];
[21 43]; [22 45]; [24 46].

7.3.3.1 Constraint Case 1

Table 7.9 summarizes the results obtained by CBO, ECBO [18], VPS [20], and
MDVC-UVPS [20] methods. MDVC-UVPS has a better performance in terms of
the best weight, average optimized weight, and standard deviation on average

Table 7.9 Performance comparison for the 1410-bar dome truss problem—constraint case 1
Element number (nodes) Areas (cm2)
CBO ECBO [18] VPS [20] MDVC-UVPS
[20]
1 (1–2) 1.0073 7.7765 5.6333 5.8499
2 (1–8) 2.5808 6.2173 4.7628 4.5115
3 (1–14) 24.3407 23.9162 37.7385 19.4823
4 (2–3) 6.6750 11.2399 7.4927 8.8480
5 (2–8) 3.8881 2.5775 3.1824 5.0084
6 (2–9) 5.0607 1.8559 1.0193 1.3568
7 (2–15) 78.9781 16.9202 8.9475 17.4331
8 (3–4) 9.2944 13.7947 10.4272 9.1098
9 (3–9) 2.6585 5.4502 4.1398 2.8712
10 (3–10) 3.5399 2.9751 3.1408 3.5473
11 (3–16) 10.2473 13.7811 15.4194 12.3768
12 (4–5) 9.6820 9.3870 8.9931 10.1099
(continued)
118 7 Optimum Design of Dome-Shaped Trusses

Table 7.9 (continued)

Element number (nodes) Areas (cm2)
CBO ECBO [18] VPS [20] MDVC-UVPS
[20]
13 (4–10) 2.4435 2.3499 3.1988 2.5797
14 (4–11) 5.0637 4.9125 7.1565 5.8381
15 (4–17) 12.9434 11.8755 17.8564 13.6402
16 (5–6) 6.9073 8.8668 9.2685 9.9096
17 (5–11) 3.1808 3.6304 3.3221 3.6543
18 (5–12) 5.9622 6.2651 6.1486 6.1529
19 (5–18) 13.3195 15.1030 8.4422 11.2448
20 (6–7) 13.2136 13.1091 12.8578 13.1071
21 (6–12) 5.4405 5.2940 5.8031 5.2361
22 (6–13) 8.4703 5.9929 7.5484 7.0691
23 (6–19) 1.8700 1.0000 1.4805 2.0015
24 (7–13) 5.5203 4.9879 4.5332 4.7178
25 (8–9) 2.4492 3.1780 2.0347 2.6101
26 (8–14) 2.2150 5.9226 5.8589 4.5434
27 (8–15) 3.1193 2.4607 2.4401 4.6174
28 (8–21) 8.7508 7.5710 6.9250 9.6758
29 (9–10) 5.1195 4.8616 3.3875 3.6296
30 (9–15) 3.8508 1.5956 1.5024 1.4891
31 (9–16) 4.4435 4.9084 4.0498 3.4020
32 (9–22) 9.1339 11.6118 11.0886 6.2153
33 (10–11) 5.7811 5.2554 5.4639 5.9308
34 (10–16) 3.4510 2.8687 2.8459 3.2334
35 (10–17) 1.8344 2.3286 2.3136 2.7173
36 (10–23) 2.7952 1.6159 3.4370 1.3932
37 (11–12) 7.2668 6.9795 8.0225 6.5660
38 (11–17) 4.7761 5.3159 5.8009 4.8170
39 (11–18) 3.3394 2.9915 4.4004 3.2626
40 (11–24) 1.0001 1.0018 1.0005 1.0165
41 (12–13) 7.3874 4.1091 7.7222 7.2529
42 (12–18) 7.3114 6.0130 5.2574 5.9226
43 (12–19) 4.8773 5.8695 4.5055 5.3115
44 (12–25) 1.0000 1.0000 1.0005 1.0010
45 (13–19) 7.9928 7.7041 7.9383 7.7499
46 (13–20) 3.4989 3.7600 4.7805 4.7836
47 (13–26) 2.0951 1.0006 1.0054 1.0035
Weight (kg) 11,102.84 10,739.19 10,491.83 10,345.12
Average optimized weight (kg) 12,359.41 10,812.20 10,936.34 10,393.83
Standard deviation on average 251.88 64.91 158.39 39.15
weight (kg)
7.3 Design Examples 119

Table 7.10 Natural frequencies (Hz) evaluated at the optimum designs of the 1410-bar dome
truss problem
Frequency number Natural frequencies (Hz)
CBO ECBO [18] VPS [20] MDVC-UVPS [20]
1 7.000 7.008 7.001 7.000
2 7.001 7.008 7.002 7.001
3 9.000 9.001 9.000 9.000
4 9.000 9.012 9.000 9.000
5 9.000 9.012 9.000 9.000

Fig. 7.10 Convergence curves for the 1410-bar dome truss problem—constraint case 1

weight that, respectively, are 10,345.12, 10,393.83, and 39.15 kg, respectively.
Table 7.10 reports the natural frequencies of the optimized structures and it is clear
that none of the frequency constraints are violated. Convergence histories for VPS
and MDVC-UVPS are depicted in Fig. 7.10. The required numbers of structural
analyses to achieve the best designs by the CBO, ECBO, VPS, and MDVC-UVPS,
respectively, are 19,000, 19,420, 25,700, and 17,750 analyses, respectively.
MDVC-UVPS obtained the best design of VPS after 10,697 analyses.

7.3.3.2 Constraint Case 2

Table 7.11 presents the optimum designs obtained by the proposed algorithms. The
lightest design (i.e., 7661.64 kg) is achieved by MDVC-UVPS algorithm after
16,308 analyses. The best designs obtained by CBO, ECBO, and VPS are 8413.46,
7860.01, and 7848.68 kg, respectively. These values are found after 18,940,
120 7 Optimum Design of Dome-Shaped Trusses

Table 7.11 Performance comparison of the 1410-bar dome truss problem—constraint Case 2
Element number (nodes) Areas (cm2)
CBO ECBO VPS MDVC-UVPS
1 (1–2) 5.1214 5.217 4.6048 4.8489
2 (1–8) 2.2479 2.213 1.5208 1.5104
3 (1–14) 1 4.0413 1.4229 4.3939
4 (2–3) 5.6721 5.3523 4.785 4.8489
5 (2–8) 2.5777 2.8635 2.3714 2.3413
6 (2–9) 1.6817 1.8832 2.2803 1.6246
7 (2–15) 1.4126 1.0007 6.0836 4.3939
8 (3–4) 6.8558 6.4681 5.037 4.8489
9 (3–9) 2.1922 1.2068 2.1952 2.1707
10 (3–10) 2.0673 1.738 1.6864 1.6765
11 (3–16) 8.9218 12.5144 2.9786 4.3939
12 (4–5) 6.4513 6.3101 5.8296 7.6688
13 (4–10) 2.5147 1.7218 2.4275 2.4287
14 (4–11) 2.3745 2.4362 4.4668 1.8282
15 (4–17) 4.273 3.5615 3.0016 5.5832
16 (5–6) 6.5994 6.1832 6.1684 7.6688
17 (5–11) 3.3831 2.7977 2.5737 2.5749
18 (5–12) 2.7308 4.1412 4.5709 3.6629
19 (5–18) 8.5163 4.1542 4.2362 5.5832
20 (6–7) 7.834 7.9148 8.7333 7.6688
21 (6–12) 3.6101 5.894 3.3266 3.7234
22 (6–13) 5.0307 3.3083 5.439 3.1638
23 (6–19) 6.127 6.6223 5.8551 5.5832
24 (7–13) 3.8352 3.6804 3.7713 3.64
25 (8–9) 5.3726 4.8207 4.6028 6.1741
26 (8–14) 2.0258 1.5864 1.5129 1.5104
27 (8–15) 5.5215 2.5913 2.3505 2.3413
28 (8–21) 3.6576 1.0843 4.334 4.0242
29 (9–10) 5.638 5.9325 8.0424 6.1741
30 (9–15) 1.7705 3.0351 1.5699 1.6246
31 (9–16) 2.3381 1.2356 2.5573 2.1707
32 (9–22) 3.316 1.708 7.4354 4.0242
33 (10–11) 6.4184 4.8743 4.8246 6.3156
34 (10–16) 5.0152 3.429 1.6796 1.6765
35 (10–17) 2.9268 1.9623 3.3532 2.4287
36 (10–23) 5.7701 2.7079 2.4308 4.8511
37 (11–12) 8.4621 5.0557 5.1426 6.3156
38 (11–17) 1.925 4.1289 1.9981 1.8282
39 (11–18) 3.0442 3.4292 2.5741 2.5749
(continued)
7.3 Design Examples 121

Table 7.11 (continued)

Element number (nodes) Areas (cm2)
CBO ECBO VPS MDVC-UVPS
40 (11–24) 4.4108 4.9348 3.491 4.8511
41 (12–13) 8.4293 7.3564 6.3216 6.3156
42 (12–18) 2.295 4.4329 3.7521 3.6629
43 (12–19) 4.1246 3.3212 7.627 3.7234
44 (12–25) 5.3458 4.9391 4.8609 4.8511
45 (13–19) 3.199 3.7342 7.1805 3.1638
46 (13–20) 4.0629 4.1154 3.7848 3.64
47 (13–26) 3.6865 5.0799 3.7592 4.8511
Weight (kg) 8413.46 7860.01 7848.68 7661.64
Average optimized weight (kg) 9932.11 8250.20 8959.27 8106.52
Standard deviation on average 1726.69 409.09 1277.34 244.08
weight (kg)

19,840, and 19,860 analyses. MDVC-UVPS obtained values for average optimized
weight and standard deviation on average weight which are equal to 8106.52 and
244.08 kg, respectively. The maximum values of the stress ratio for CBO, ECBO,
VPS, and MDVC-UVPS are 96.96, 99.41, 99.53, and 99.99%, respectively.

7.4 Concluding Remarks

Structural optimization with multiple natural frequency constraints is a highly

nonlinear and non-convex dynamic optimization problem since weight reduction
conflicts with the frequency constraints. Moreover, determining the optimum
design of large-scale structures is known as one of the difficult and complex
optimization problems. In this chapter, the performances of CBO, ECBO, VPS, and
MDVC-UVPS algorithms were compared through three dome truss structures with
dynamic frequency constraints. Moreover, all the examples are designed based on
strength, stability, and displacement constraints to evaluate the performance of the
algorithms in another type of optimization problem. MDVC-UVPS converged to
better designs in all of the test problems. Also, the average weights and standard
deviations found by this method in the independent optimization runs were lower in
all of the benchmark examples indicating that the search reliability of the proposed
method is superior. Apart from these characteristics, MDVC-UVPS needs fewer
structural analyses to obtain the best design of other algorithms. Comparison of the
convergence curves showed that the intermediate designs found by this algorithm
were usually better than those found by CBO, ECBO, and VPS.
122 7 Optimum Design of Dome-Shaped Trusses

References

1. Lan TT (2005) Space frame structures. Handbook of structural Engineering. CRC Press,
Boca Raton, FL, pp 24–31
2. Bellagamba L, Yang TY (1981) Minimum mass truss structures with constraints on
fundamental natural frequency. AIAA J 19:1452–1458
3. Kaveh A, Zolghadr A (2016) Optimal design and analysis of large-scale domes with
frequency constraints. Smart Struct Syst 18(4):733–754
4. Lin JH, Chen WY, Yu YS (1982) Structural optimization on geometrical configuration and
element sizing with static and dynamic constraints. Comput Struct 15:507–515
5. Konzelman CJ (1986) Dual methods and approximation concepts for structural optimization.
M.Sc. Thesis, Department of Mechanical Engineering, University of Toronto, Canada
6. Grandhi RV, Venkayya VB (1988) Structural optimization with frequency constraints.
AIAA J 26:858–866
7. Wang D, Zha W, Jiang J (2004) Truss optimization on shape and sizing with frequency
constraints. AIAA J 42:622–630
8. Sedaghati R (2005) Benchmark case studies in structural design optimization using the force
method. Int J Solids Struct 42:5848–5871
9. Lingyun W, Mei Z, Guangming W, Guang M (2005) Truss optimization on shape and sizing
with frequency constraints based on genetic algorithm. Comput Mech 35:361–368
10. Gomes HM (2011) Truss optimization with dynamic constraints using a particle swarm
algorithm. Expert Syst Appl 38:957–968
11. Kaveh A, Zolghadr A (2012) Truss optimization with natural frequency constraints using a
hybridized CSS–BBBC algorithm with trap recognition capability. Comput Struct
102–103:14–27
12. Miguel LFF, Fadel Miguel LF (2012) Shape and size optimization of truss structures
considering dynamic constraints through modern meta-heuristic algorithms. Expert Syst Appl
39:9458–9467
13. Zuo W, Bai J, Li B (2014) A hybrid OC–GA approach for fast and global truss optimization
with frequency constraints. Appl Soft Comput 14:528–535
14. Kaveh A, Javadi SM (2014) Shape and size optimization of trusses with multiple frequency
constraints using harmony search and ray optimizer for enhancing the particle swarm
optimization algorithm. Acta Mech 225:1595–1605
15. Kaveh A, Ilchi Ghazaan M (2015) Hybridized optimization algorithms for design of trusses
with multiple natural frequency constraints. Adv Eng Softw 79:137–147
16. Hosseinzadeh Y, Taghizadieh N, Jalili S (2016) Hybridizing electromagnetism-like mech-
anism algorithm with migration strategy for layout and size optimization of truss structures
with frequency constraints. Neural Comput Appl 27(4):953–971
17. American Institute of Steel Construction (AISC) (1989) Manual of steel construction:
allowable stress design, Chicago, USA
18. Kaveh A, Ilchi Ghazaan M (2017) Optimal design of dome truss structures with dynamic
frequency constraints. Struct Multidiscip Optim 53(3):605–621
19. Kaveh A, Ilchi Ghazaan M (2016) Vibrating particles system algorithm for truss optimization with
multiple natural frequency constraints. Acta Mech 228(1):307–322. (Published online, 1–16)
20. Kaveh A, Ilchi Ghazaan M (2018) A new hybrid meta-heuristic algorithm for optimal design
of large-scale dome structures. Eng Optimiz 50(2):235–252
Chapter 8
Optimal Design of Steel Lattice
Transmission Line Towers

8.1 Introduction

Lattice towers are used for power lines of all voltages and are the most common
type for high-voltage transmission lines. The design optimization of these structures
has always been a difficult task due to a large number of design variables. Some
studies have already been performed in the context of optimization of transmission
line tower structures. For instance, Rao [1] utilized a derivative-free nonlinear
optimization technique for minimum weight design of high-voltage transmission
line towers under a set of control parameters, including geometrical parameters as
well as tensions in conductors and ground-wires. Kaveh et al. [2] employed a
migration genetic algorithm for optimization of transmission towers and they
trained neural networks as analyzers to take part of the computational load. París
et al. [3] studied the shape optimization of a transmission line tower, subjected to
multiple load cases and code constraints. Guo and Li [4] presented an adaptive
genetic algorithm for different optimization models of steel transmission towers
considering size, layout, and topology design variables. Chunming et al. [5] utilized
a genetic algorithm for optimization of a transmission tower, where cross-sectional
areas and material types of the members were selected as design variables. Tort
et al. [6] integrated the simulated annealing optimization algorithm into the com-
mercial PLS-TOWER software to optimize steel lattice towers for minimum weight
using both size and layout design variables.
In this chapter, the efficiency of Colliding Bodies Optimization (CBO) [7],
Enhanced Colliding Bodies Optimization (ECBO) [8], Vibrating Particles System
(VPS) [9], and a hybrid algorithm called MDVC-UVPS [10] are investigated in
optimum design of three latticed steel towers. The procedure considers discrete
values of cross-sectional areas.

© Springer International Publishing AG, part of Springer Nature 2018 123

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_8
124 8 Optimal Design of Steel Lattice Transmission Line Towers

8.2 Optimal Design of Transmission Line Towers

For the sizing optimization of transmission line towers, the cross-section areas of
truss bars are often considered as discrete design variables; therefore, all of them are
selected from a list of discrete cross sections based on production standards. The
optimization problem aims to minimize the weight of the structure while satisfying
strength and serviceability requirements.
Size optimization of a transmission line tower with its members being collected
in ng design groups can be formulated as follows:
Find fXg ¼ ½x1 ; x2 ; . . .; xng 
Png P
nmðiÞ
to minimize WðfXgÞ ¼ xi qj Lj
i¼1 j¼1 ð8:1Þ

gj ðfXgÞ  0; j ¼ 1; 2; . . .; nc
subjected to:
ximin  xi  ximax

where {X} is the vector containing the design variables; W({X}) presents the weight
of the structure; nm(i) is the number of members for the ith group; qj and Lj denote
the material density and the length of the jth member, respectively. ximin and ximax
are the lower and upper bounds of the design variable xi, respectively. gj({X})
denotes the jth design constraint; and nc is the number of constraints.
For constraint handling, a penalty approach is utilized. For this purpose, the
objective function (Eq. 8.1) is redefined as follows:

PðfXgÞ ¼ ð1 þ e1 :tÞe2  WðfXgÞ ð8:2Þ

where P({X}) is the penalized cost function or the objective function to be mini-
mized and t denotes the sum of the violations of the design constraints. Here, e1 is
set to unity and e2 is calculated by
iter
e2 ¼ 1:5 þ 1:5  ð8:3Þ
itermax

where iter is the current iteration number and itermax is the total number of iterations
of the optimization process.

8.3 Design Problems

The following three benchmark structural examples are considered in this section:
• A 47-bar power transmission tower
• A 160-bar power transmission tower
• A 244-bar power transmission tower
8.3 Design Problems 125

The design variables are the cross-sectional areas of the bar element and in all
problems, solution candidates are allowed to select discrete values from a per-
missible list of cross sections (real numbers are rounded to the nearest integer in
each iteration). Each example has been solved thirty times independently and a
maximum of 1000 iterations is considered as the termination condition.
A population of 20 particles is considered for each algorithm and the other algo-
rithm parameters are set similar to the values proposed in [8–10]. The optimization
algorithms are coded in MATLAB and the structures are analyzed using the direct
stiffness method by our own codes.

8.3.1 A 47-Bar Power Transmission Tower

The first design example demonstrated in Fig. 8.1 has 47 members and 22 nodes.
The cross-sectional areas of the members were categorized into 27 groups, as
follows: (1) A1 = A3, (2) A2 = A4, (3) A5 = A6, (4) A7, (5) A8 = A9, (6) A10,
(7) A11 = A12, (8) A13 = A14, (9) A15 = A16, (10) A17 = A18, (11) A19 = A20,
(12) A21 = A22, (13) A23 = A24, (14) A25 = A26, (15) A27, (16) A28,
(17) A29 = A30, (18) A31 = A32, (19) A33, (20) A34 = A35, (21) A36 = A37,
(22) A38, (23) A39 = A40, (24) A41 = A42, (25) A43, (26) A44 = A45, and
(27) A46 = A47. The cross-sectional areas were chosen from the 64 discrete values
listed in Table 8.1. The material of the members has a Young’s modulus of
206.842 kN/mm2 (30,000 ksi) and a density of 8303.97 kg/m3 (0.3 lb/in3). The
nodes of structure are subjected to a combination of three loading cases: (1) 6.0 kips
acting in the positive x-direction and 14.0 kips acting in the negative y-direction at
nodes 17 and 22, (2) 6.0 kips acting in the positive x-direction and 14.0 kips acting
in the negative y-direction at node 17, and (3) 6.0 kips acting in the positive x-
direction and 14.0 kips acting in the negative y-direction at node 22. The first
loading case corresponds to the load applied by the two power lines to the tower at
an angle, while the second and third loading cases occur when one of the two lines
snaps.
Both stress and buckling constraints should be satisfied for all of the members of
the tower. Allowable tensile and compressive stresses are taken as 137.895 MPa
(20 ksi) and 103.421 MPa (15.0 ksi), respectively. Moreover, the Euler buckling
compressive stress for a member with a cross-sectional area of Ai is calculated as
follows:
KEAi
rcr
i ¼ L2 i ¼ 1; 2; 3; . . .; 47 ð8:4Þ
i

where K is a constant parameter that should be selected based on the type of the
cross-sectional geometry; E is the Young’s modulus of the material; and Li is the
length of member i. The buckling constant K is assumed as 3.96 [11].
126 8 Optimal Design of Steel Lattice Transmission Line Towers

Fig. 8.1 Schematic of the 47-bar power transmission tower

Table 8.2 shows the optimum designs found by different algorithms. The lightest
design (i.e., 2374.09 lb) is achieved by MDVC-UVPS algorithm after 4867 anal-
yses. The weights of the best designs obtained by CBO, ECBO, and VPS are
2386.44, 2375.35, and 2374.81 lb, respectively. These values are found after 9760,
Table 8.1 List of available cross-sectional areas for the 47-bar power transmission tower problem
No. Area (mm2) Area (in.2) No. Area (mm2) Area (in.2) No. Area (mm2) Area (in.2) No. Area (mm2) Area (in.2)
8.3 Design Problems

1 71.6130 0.1110 17 1008.3850 1.5630 33 2477.4140 3.8400 49 7419.43 11.5001

2 90.9680 0.1410 18 1045.1590 1.6200 34 2496.7690 3.8700 50 8709.66 13.5000
3 126.4510 0.1960 19 1161.2880 1.8000 35 2503.2210 3.8800 51 8967.724 13.9000
4 161.2900 0.2500 20 1283.8680 1.9900 36 2696.7690 4.1800 52 9161.272 14.2000
5 198.0640 0.3070 21 1374.1910 2.1300 37 2722.5750 4.2200 53 9999.98 15.5000
6 252.2580 0.3910 22 1535.4810 2.3800 38 2896.7680 4.4900 54 10,322.56 16.0000
7 285.1610 0.4420 23 1690.3190 2.6200 39 2961.2480 4.5899 55 10,903.2 16.9000
8 363.2250 0.5630 24 1696.7710 2.6300 40 3096.7680 4.8000 56 12,129.01 18.8000
9 388.3860 0.6020 25 1858.0610 2.8800 41 3206.4450 4.9700 57 12,838.68 19.9000
10 494.1930 0.7660 26 1890.3190 2.9300 42 3303.2190 5.1200 58 14,193.52 22.0000
11 506.4510 0.7850 27 1993.5440 3.0900 43 3703.2180 5.7400 59 14,774.16 22.9000
12 641.2890 0.9940 28 729.0310 1.1300 44 4658.0550 7.2200 60 15,806.42 24.5000
13 645.1600 1.0000 29 2180.6410 3.3800 45 5141.9250 7.9700 61 17,096.74 26.5000
14 792.2560 1.2280 30 2238.7050 3.4700 46 5503.2150 8.5300 62 18,064.48 28.0000
15 816.7730 1.2660 31 2290.3180 3.5500 47 5999.9880 9.3000 63 19,354.8 30.0000
16 939.9980 1.4570 32 2341.9310 3.6300 48 6999.9860 10.8500 64 21,612.86 33.5000
127
128 8 Optimal Design of Steel Lattice Transmission Line Towers

Table 8.2 Performance comparison for the 47-bar power transmission tower problem
Design variable Areas (in.2)
CBO ECBO VPS MDVC-UVPS
1 3.84 3.84 3.84 3.84
2 3.38 3.38 3.38 3.38
3 0.785 0.766 0.766 0.766
4 0.196 0.111 0.111 0.111
5 0.994 0.785 0.785 0.785
6 1.8 1.99 1.99 1.99
7 2.13 2.13 2.13 2.13
8 1.228 1.228 1.228 1.228
9 1.563 1.563 1.563 1.563
10 2.13 2.13 2.13 2.13
11 0.111 0.111 0.111 0.111
12 0.111 0.141 0.111 0.111
13 1.8 1.8 1.8 1.8
14 1.8 1.8 1.8 1.8
15 1.563 1.457 1.457 1.457
16 0.442 0.442 0.442 0.563
17 3.63 3.63 3.63 3.63
18 1.457 1.457 1.457 1.457
19 0.307 0.307 0.307 0.25
20 3.09 3.09 3.09 3.09
21 1.266 1.266 1.266 1.228
22 0.307 0.307 0.307 0.391
23 3.84 3.84 3.84 3.84
24 1.563 1.563 1.563 1.563
25 0.111 0.111 0.111 0.111
26 4.59 4.59 4.59 4.59
27 1.457 1.457 1.457 1.457
Weight (lb) 2386.44 2375.35 2374.81 2374.09
Average optimized weight (lb) 2462.76 2415.51 2415.07 2413.46
Standard deviation on average 44.79 41.01 35.65 38.21
weight (lb)
1 in.2 = 6.4516 cm2

16,240, and 15,540 analyses. The average optimized weight and standard deviation
on average weight achieved by MDVC-UVPS are 2413.46 and 38.21 lb, respec-
tively. Figure 8.2 shows the convergence curves of the best results found by CBO,
ECBO, VPS, and MDVC-UVPS.
8.3 Design Problems 129

Fig. 8.2 Convergence curves for the 47-bar power transmission tower problem

8.3.2 A 160-Bar Power Transmission Tower

The second design example is the 160-bar tower structure shown in Fig. 8.3. The
members of the structure are categorized into 38 design groups. The geometrical
information including member’s connectivity, design groups, and coordinates of
nodes can be found in [12]. The material of the members has a Young’s modulus of
2.047  106 kgf/cm2 and a density of 7850 kg/m3. This tower is subjected to the
combination of eight loading conditions as listed in Table 8.3. Moreover, the design
constraints consist of buckling stresses of the structural members. For a member
under compressive force, the buckling stress can be calculated as follows:
For kl/r  120
ðkl=rÞ2
rall ¼ 1300  ð8:5Þ
24

and if kl/r > 120, then

ðkl=rÞ2
rall ¼ 1300  ð8:6Þ
24

where k is effective length factor, r is radius of gyration, and l is length of member.

For all truss members, k is equal to unity.
For this design example, the available 42 sections and their radius of gyrations
are as follows:
D = {1.84, 2.26, 2.66, 3.07, 3.47, 3.88, 4.79, 5.27, 5.75, 6.25, 6.84, 7.44, 8.06,
8.66, 9.40, 10.47, 11.38, 12.21, 13.79, 15.39, 17.03, 19.03, 21.12, 23.20, 25.12,
27.50, 29.88, 32.76, 33.90, 34.77, 39.16, 43.00, 45.65, 46.94, 51.00, 52.10, 61.82,
61.90, 68.30, 76.38, 90.60, 94.13} (cm2).
130 8 Optimal Design of Steel Lattice Transmission Line Towers

Fig. 8.3 Schematic of the

160-bar bar power
transmission tower

r = {0.47, 0.57, 0.67, 0.77, 0.87, 0.97, 0.97, 1.06, 1.16, 1.26, 1.15, 1.26, 1.36, 1.46,
1.35, 1.36, 1.45, 1.55, 1.75, 1.95, 1.74, 1.94, 2.16, 2.36, 2.57, 2.35, 2.56, 2.14, 2.33,
2.97, 2.54, 2.93, 2.94, 2.94, 2.92, 3.54, 3.96, 3.52, 3.51, 3.93, 3.92, 3.92} (cm).
The optimal designs found by different methods are presented in Table 8.4. The
weight of the best result obtained by MDVC-UVPS is 1336.71 kg that is the best
among the compared methods. The average optimized weight of this method is
1364.56 kg, which is less than those of all other methods. Comparison of the
convergence rates of CBO, ECBO, VPS, and MDVC-UVPS is illustrated in
Fig. 8.4. MDVC-UVPS requires 4518 structural analyses to find its optimum
8.3 Design Problems 131

Table 8.3 Loading conditions for the 160-bar power transmission tower problem
Condition Node Fx Fy Fz Condition Node Fx Fy Fz
1 52 −868 0 −491 5 52 −917 0 −491
37 −996 0 −546 37 −951 0 −546
25 −1091 0 −546 25 −1015 0 −546
28 −1091 0 −546 28 −636 0 −428
2 52 −493 1245 −363 6 52 −917 1259 −491
37 −996 0 −546 37 −572 0 −428
25 −1091 0 −546 25 −1015 1303 −546
28 −1091 0 −546 28 −1015 0 −546
3 52 −917 0 −491 7 52 −917 0 −491
37 −951 0 −546 37 −951 0 −546
25 −1015 0 −546 25 −1015 0 −546
28 −1015 0 −546 28 −636 1303 −428
4 52 −917 0 −546 8 52 −498 1460 −363
37 −572 1259 −428 37 −951 0 −546
25 −1015 0 −546 25 −1015 0 −546
28 −1015 0 −546 28 −1015 0 −546

Table 8.4 Performance comparison for the 160-bar power transmission tower problem
Design variable Areas (cm2)
CBO ECBO VPS MDVC-UVPS
1 19.03 19.03 19.03 19.03
2 5.27 5.27 5.27 5.27
3 19.03 19.03 19.03 19.03
4 5.27 5.27 5.27 5.27
5 19.03 19.03 19.03 19.03
6 5.75 5.75 5.75 5.75
7 15.39 15.39 15.39 15.39
8 5.75 5.75 5.75 5.75
9 13.79 13.79 13.79 13.79
10 5.75 5.75 5.75 5.75
11 5.75 5.75 5.75 5.75
12 12.21 12.21 13.79 12.21
13 6.25 6.25 6.25 6.25
14 5.75 5.75 5.75 5.75
15 4.79 3.88 2.66 3.88
16 6.25 7.44 7.44 7.44
17 4.79 1.84 1.84 1.84
18 8.66 8.66 8.66 8.66
19 2.66 2.66 2.66 2.66
20 3.07 3.07 3.07 3.07
(continued)
132 8 Optimal Design of Steel Lattice Transmission Line Towers

Table 8.4 (continued)

Design variable Areas (cm2)
CBO ECBO VPS MDVC-UVPS
21 2.66 2.66 2.66 2.66
22 8.06 8.06 8.06 8.06
23 5.75 5.27 5.27 5.75
24 6.25 7.44 6.25 6.25
25 6.25 5.75 5.75 5.75
26 1.84 1.84 1.84 1.84
27 4.79 4.79 4.79 5.27
28 2.66 2.66 2.66 2.66
29 3.47 3.47 3.47 3.47
30 1.84 1.84 1.84 1.84
31 2.26 2.26 2.26 2.26
32 3.88 3.88 3.88 3.88
33 1.84 1.84 1.84 1.84
34 1.84 1.84 1.84 1.84
35 3.88 3.88 3.88 3.88
36 1.84 1.84 1.84 1.84
37 1.84 1.84 1.84 1.84
38 3.88 3.88 3.88 3.88
Weight (kg) 1350.37 1339.17 1337.44 1336.71
Average optimized weight (kg) 1403.97 13,776.26 1375.51 1364.56
Standard deviation on average 215.11 109.83 125.43 73.46
weight (kg)

Fig. 8.4 Convergence curves for the 160-bar bar power transmission tower problem
8.3 Design Problems 133

solution while CBO, ECBO, and VPS require 5820, 11,420, and 11,080 structural
analyses, respectively.

8.3.3 A 244-Bar Power Transmission Tower

The 244-bar transmission tower shown in Fig. 8.5 is studied as the final design
problem to demonstrate the efficiency of the compared algorithms. Members of the
structure are categorized into 26 groups as given by Saka [13]. The cross-sectional
areas were chosen from the 45 discrete values listed in Table 8.5. The modulus of
elasticity and yield stress of steel are assumed to be 210 kN/mm2 and 233.3 N/mm2,

Fig. 8.5 Schematic of the

244-bar bar power
transmission tower
 134

Table 8.5 List of available cross sections for the 244-bar power transmission tower problem
No. Section A in.2 (mm2) r in. (mm) Num. Section A in.2 (mm2) r in. (mm)
1 L661 11.0 (7096.76) 1.17 (29.72) 24 L 3 1/2  3 1/2  1/2 3.25 (2096.77) 0.683 (17.35)
2 L 6  6  7/8 9.73 (6277.41) 1.17 (29.72) 25 L 3 1/2  3 1/2  7/16 2.87 (1851.61) 0.684 (17.37)
3 L 6  6  3/4 8.44 (5445.15) 1.17 (29.72) 26 L 3 1/2  3 1/2  3/8 2.48 (1600.00) 0.687 (17.45)
4 L 6  6  5/8 7.11 (4587.09) 1.18 (29.97) 27 L 3 1/2  3 1/2  5/16 2.09 (1348.38) 0.690 (17.53)
5 L 6  6  9/16 6.43 (4148.38) 1.18 (29.97) 28 L 3 1/2  3 1/2  1/4 1.69 (1090.32) 0.694 (17.63)
6 L 6  6  1/2 5.75 (3709.67) 1.18 (29.97) 29 L 3  3  1/2 2.75 (1774.19) 0.584 (14.83)
7 L 6  6  7/16 5.06 (3264.51) 1.19 (30.23) 30 L 3  3  7/16 2.43 (1567.74) 0.585 (14.86)
8 L 6  6  3/8 4.36 (2812.90) 1.19 (30.23) 31 L 3  3  3/8 2.11 (1361.29) 0.587 (14.91)
9 L 6  6  5/16 3.65 (2354.83) 1.20 (30.48) 32 L 3  3  5/16 1.78 (1148.38) 0.589 (14.96)
10 L 5  5  7/8 7.98 (5148.38) 0.973 (24.71) 33 L 3  3  1/4 1.44 (929.03) 0.592 (15.04)
11 L 5  5  3/4 6.94 (4477.41) 0.975 (24.77) 34 L 3  3  3/16 1.09 (703.22) 0.596 (15.14)
12 L 5  5  5/8 5.86 (3780.64) 0.978 (24.84) 35 L 2 1/2  2 1/2  1/2 2.25 (1451.61) 0.487 (12.37)
13 L 5  5  1/2 4.75 (3064.51) 0.983 (24.97) 36 L 2 1/2  2 1/2  3/8 1.73 (1116.13) 0.487 (12.37)
14 L 5  5  7/16 4.18 (2696.77) 0.986 (25.04) 37 L 2 1/2  2 1/2  5/16 1.46 (941.93) 0.489 (12.42)
15 L 5  5  3/8 3.61 (2329.03) 0.990 (25.15) 38 L 2 1/2  2 1/2  1/4 1.19 (767.74) 0.491 (12.47)
16 L 5  5  5/16 3.03 (1954.83) 0.944 (25.25) 39 L 2 1/2  2 1/2  3/16 0.902 (581.93) 0.495 (12.57)
17 L 4  4  3/4 5.44 (3509.67) 0.778 (19.76) 40 L 2  2  3/8 1.36 (877.42) 0.389 (9.88)
18 L 4  4  5/8 4.61 (2974.19) 0.779 (19.79) 41 L 2  2  5/16 1.15 (741.93) 0.390 (9.91)
19 L 4  4  1/2 3.75 (2419.35) 0.782 (19.86) 42 L 2  2  1/4 0.938 (605.16) 0.391 (9.93)
20 L 4  4  7/16 3.31 (2135.48) 0.785 (19.94) 43 L 2  2  3/16 0.715 (461.29) 0.394 (10.00)
21 L 4  4  3/8 2.86 (1845.16) 0.788 (20.02) 44 L 2  2  1/8 0.484 (312.26) 0.398 (10.11)
22 L 4  4  5/16 2.40 (1548.38) 0.791 (20.09) 45 L 1 1/4  1 1/4  3/16 0.434 (280.00) 0.244 (6.198)
23 L 4  4  1/4 1.94 (1251.61) 0.795 (20.19)
8 Optimal Design of Steel Lattice Transmission Line Towers
8.3 Design Problems 135

Table 8.6 The load cases and displacement bounds for the 244-bar bar power transmission tower
problem
Loading conditions Joint number Loads (kN) Displacement
limitations
(mm)
X Z X Z
1 1 10 −30 45 15
2 10 −30 45 15
17 35 −90 30 15
24 175 −45 30 15
25 175 −45 30 15
2 1 – −360 45 15
2 – −360 45 15
17 – −180 30 15
24 – −90 30 15
25 – −90 30 15

respectively [14]. The load cases considered and the bounds imposed on the dis-
placements are shown in Table 8.6. Values of the allowable tensile and compres-
sive stresses are calculated based on ASD-AISC code [15]. Moreover, as per the
recommendation of ASD-AISC, the maximum slenderness ratios are set to 200 and
300 for compression and tension members, respectively.
Table 8.7 shows the comparison of the results of different algorithms.
MDVC-UVPS algorithm yields the least volume design for this example, which is
0.755011 m3. The other designs are 0.813555 m3 by CBO, 0.768159 m3 by ECBO,
and 0.764982 m3 by VPS. The best design of the hybrid method has been achieved
after 6665 analyses. CBO, ECBO, and VPS require 30,000, 4020, 9260, and 19,480
structural analyses to find their optimum solutions, respectively. Figure 8.6 shows
the convergence curves of the best results found by CBO, ECBO, VPS, and
MDVC-UVPS.

8.4 Concluding Remarks

In this chapter, size optimization of transmission line towers under multiple loading
cases is studied. Three steel lattice towers with 26, 27, and 38 discrete variables are
considered. Numerical results demonstrate that the MDVC-UVPS achieved the
lightest designs in all examples. The average and standard deviation values
achieved from independent runs also confirm the efficiency of the hybrid method in
optimal design of real-size structures. Moreover, comparison of the convergence
curves and the required structural analyses indicate that MDVC-UVPS comes close
to the optimum design rapidly.
Table 8.7 Performance comparison for the 244-bar bar power transmission tower problem
136

Design Sections
variable CBO ECBO VPS MDVC-UVPS
1 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16
2 L 4  4  1/4 L 4  4  1/4 L 4  4  1/4 L 4  4  1/4
3 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4
4 L 4  4  5/16 L 3  3  7/16 L 3  3  3/8 L 5  5  7/16
5 L 3  3  7/16 L 2  2  3/8 L 2 1/2  2 1/2  3/8 L 2  2  3/8
6 L 5  5  7/8 L 5  5  7/8 L 5  5  5/16 L 4  4  3/4
7 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16
8 L 4  4  3/4 L 5  5  3/8 L 5  5  7/16 L 4  4  3/4
9 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4
10 L 2  2  3/8 L 2  2  3/8 L 2  2  3/8 L 2  2  3/8
11 L 3 1/2  3 1/2  5/16 L 5  5  3/4 L 5  5  3/8 L 5  5  1/2
12 L 5  5  3/4 L 4  4  5/8 L 4  4  5/8 L 4  4  3/8
13 L 2 1/2  2 1/2  3/8 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4
14 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8
15 L 6  6  3/4 L 6  6  5/8 L 6  6  7/8 L 6  6  3/4
16 L 3 1/2  3 1/2  5/16 L 3 1/2  3 1/2  5/16 L 3 1/2  3 1/2  5/16 L 3 1/2  3 1/2  5/16
17 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8
18 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8
19 L 3  3  3/8 L 3  3  3/8 L 2 1/2  2 1/2  1/2 L 3  3  3/8
20 L 6  6  5/8 L 6  6  9/16 L 6  6  1/2 L 6  6  9/16
21 L 3  3  3/16 L 3  3  3/16 L 3  3  3/16 L 3  3  3/16
22 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4
23 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4 L 2  2  1/4
24 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8 L 2  2  1/8
25 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16
26 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16 L 1 1/4  1 1/4  3/16
Volume (m3) 0.813555 0.768159 0.764982 0.755011
Average optimized volume (m3) 0.899143 0.830075 0.833456 0.818619
Standard deviation on average volume (m3) 0.1462 0.1184 0.1207 0.1053
8 Optimal Design of Steel Lattice Transmission Line Towers
References 137

Fig. 8.6 Convergence curves for the 244-bar bar power transmission tower problem

References

1. Rao GV (1995) Optimum designs for transmission line towers. Comput Struct 57(1):81–92
2. Kaveh A, Gholipour Y, Rahami H (2008) Optimal design of transmission towers using
genetic algorithm and neural networks. Int J Space Struct 23(1):1–9
3. París J, Martínez S, Navarrina F, Colominas I, Casteleiro M (2010) Structural optimization of
high tension towers. In: 2nd International conference on engineering optimization, Portugal
4. Guo HY, Li ZL (2011) Structural topology optimization of high-voltage transmission tower
with discrete variables. Struct Multidiscip Optim 43(6):851–861
5. Chunming W, Tingting S, Bin M, Jing G (2012) Research on the optimal layout of high
strength steel in the transmission tower. Phys Procedia 33:619–625
6. Tort C, Şahin S, Hasançebi O (2017) Optimum design of steel lattice transmission line towers
using simulated annealing and PLS-TOWER. Comput Struct 15(179):75–94
7. Kaveh A, Mahdavi VR (2014) Colliding bodies optimization: a novel meta-heuristic method.
Comput Struct 139:18–27
8. Kaveh A, Ilchi Ghazaan M (2014) Enhanced colliding bodies optimization for design
problems with continuous and discrete variables. Adv Eng Softw 77:66–75
9. Kaveh A, Ilchi Ghazaan M (2016) Vibrating particles system algorithm for truss optimization
with multiple natural frequency constraints. Acta Mech 228(1):307–332
10. Kaveh A, Ilchi Ghazaan M (2018) A new hybrid meta-heuristic algorithm for optimal design
of large-scale dome structures. Eng Optimiz 50(2):235–252
11. Lee KS, Geem ZW, Lee SH, Bae KW (2005) The harmony search heuristic algorithm for
discrete structural optimization. Eng Optimiz 37(7):663–684
12. Groenwold AA, Stander N, Snyman JA (1999) A regional genetic algorithm for the discrete
optimal design of truss structures. Int J Numer Methods Eng 44:749–766
13. Saka MP (1990) Optimum design of pin-jointed steel structures with practical applications.
J Struct Eng 116(10):2599–2620
14. Toğan V, Daloğlu AT (2006) Optimization of 3D trusses with adaptive approach in genetic
algorithms. Eng Struct 28(7):1019–1027
15. American Institute of Steel Construction (AISC) (1998) Manual of steel construction:
allowable stress design, Chicago, USA
Chapter 9
Optimal Seismic Design of 3D Steel
Frames

9.1 Introduction

Steel buildings are preferred in residential as well as commercial buildings because

of being super-quick to build on-site, as a great deal of work can be prefabricated at
the factory. Moreover, these structures are flexible, which makes them quite suit-
able structures for resisting dynamic forces such as earthquake loads. Design of
frame structures necessitates the selection of steel sections for its columns and
beams from a standard steel section tables such that the frame satisfies the ser-
viceability and strength requirements specified by the code of practice while the
economy is taken into account in the overall or material cost of the frame.
Sizing optimization of frame structures which are complicated optimization
problems was studied by many researchers. Hasançebi et al. [1] utilized seven
meta-heuristic search algorithms (genetic algorithms, simulated annealing, evolu-
tion strategies, particle swarm optimizer, tabu search, ant colony optimization, and
harmony search) to optimize real-size rigidly connected steel frames. Evolution
strategies approach has been the most effective algorithm amongst the seven and
simulated annealing could be considered as the second winner of this performance
competition. Aydogdu and Saka [2] used ant colony optimization for optimum
design of irregular steel space frames including element warping effect. Kaveh and
Zakian [3] employed Charged System Search (CSS) and Improved Harmony
Search (IHS) algorithms for optimal design of steel frames under seismic loading.
The results show that the performance of the CSS is better than that of the IHS
considering the same number of structural analyses. Hasançebi and Carbas [4]
examined Bat Inspired (BI) algorithm in the context of discrete size optimization of
real-size steel frames designed for minimum weight. The numerical results of the BI
algorithm are compared to the previously reported solutions using different
meta-heuristics in the literature and indicate robustness and efficiency of the BI
algorithm for this class of problems. Talatahari et al. [5] combined the Eagle
Strategy algorithm with Differential Evolution (ES-DE) to minimize the weight of

© Springer International Publishing AG, part of Springer Nature 2018 139

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0_9
140 9 Optimum Seismic Design of 3D Steel Frames

steel frames. The optimized designs obtained by the proposed algorithm are better
than those found by the standard differential evolution algorithm and also very
competitive with literature. The overall convergence behavior is significantly
enhanced by the hybrid optimization strategy. Artar and Daloglu [6] used a method
based on genetic algorithm for minimum weight design of composite steel frames
with semirigid connections and column bases. Their results show that consideration
of the contribution of concrete on the behavior of the floor beams enables a lighter
and more economical design for steel frames with semirigid connections and col-
umn bases. Kaveh et al. [7] employed cuckoo search algorithm to optimize steel
frame structures under seismic loading based on response spectral and equivalent
static analyses. The effect of lateral seismic loading distribution on the achieved
optimum designs was also investigated. Results show similar weights for optimum
designs using spectral and equivalent static analyses; however, different material
distribution and seismic behaviors are observed. Carbas [8] enhanced the perfor-
mance of the firefly algorithm by suggesting two new expressions for the attrac-
tiveness and randomness parameters of the algorithm and optimized two steel space
frame design examples by this method. Kaveh and Bolandgerami [9] indicated the
efficiency of the cascade enhanced colliding body optimization for optimum design
of large-scale space steel frames. Kaveh et al. [10] employed nine meta-heuristic
algorithms to study the effect of the change in the ductility type on the structural
weight. Results show that the Ordinary Moment Frame (OMF) can produce lighter
designs in most cases, in spite of larger base shear. Kaveh et al. [11] studied three
different types of lateral resisting steel moment frames consisting of ordinary
moment frame, intermediate moment frame, and special moment frame. Optimum
seismic design was performed for 3D steel moment frames with different types of
lateral resisting systems.
The contribution of this chapter is concerned with optimization of steel frames
under seismic loads based on response spectral. Frame members are selected from
available set of steel sections for producing practically acceptable designs according
to Load and Resistance Factor Design-American Institute of Steel Construction
(LRFD-AISC) specification. Three irregular steel frame problems are considered
to evaluate the performance of the CBO, ECBO, VPS and MDVC-UVPS algo-
rithms [12].

9.2 Optimum Design Problem of Steel Space Frame

Structural optimization aims to design structure with minimum weight or minimize

an objective function corresponding to minimal cost of the structure, while the
corresponding design constraints are satisfied. The design constraints and objective
function for investigated problems are explained in the following subsections.
9.2 Optimum Design Problem of Steel Space Frame 141

9.2.1 Design Constraints

The design should be carried out in such a way that the frame satisfies the following
constraint:
1. Strength constraints: Each frame member should have sufficient strength to
resist the internal forces developed due to factored external loading.
2. Serviceability constraints: Beam deflections and lateral displacement of the
frame should be less than the limits specified in the code.
3. Geometric constraints: Three types of geometric constraints are considered to
satisfy practical requirements. The first type ensures that the depths of the
W-shaped sections selected for the columns of two consecutive stories should be
either equal to each other or the one in the above story should be smaller than
the one in the below story. The second and third types of constraint make sure
that if a beam is connected to flange of a column, the flange width of the beam
should be less than or equal to the flange width of the column in the connection
and if a beam is connected to the web of a column, the flange width of the beam
should be less than or equal to (D-2tb), where D and tb are the depth and the
flange thickness of the W-shaped section selected for column, respectively.
Figure 9.1 and the following formula clarify the geometric constraints.

Fig. 9.1 Beam–column geometric constraints

142 9 Optimum Seismic Design of 3D Steel Frames

ðDÞi
1 i ¼ 2; . . .; ns ð9:1Þ
ðDÞi1

ðBf Þbi
 1 i ¼ 1; . . .; nf ð9:2Þ
ðBf Þci

ðBf Þbi
1 i ¼ 1; . . .; nw ð9:3Þ
ðDÞi  2ðtb Þi

where ns, nf, and nw are the number of stories, the total number of joints where
beams are connected to the flange of a column, and the total number of joints where
beams are connected to the web of a column, respectively. (Bf)bi and (Bf)ci are the
flange width of W-section selected for the beam and column at joint i, respectively.

9.2.2 The Objective Function

Sizing optimization of a steel frame with its members being collected in ng design
groups can be formulated as follows:
Find fXg ¼ ½x1 ; x2 ; ::; xng 
Png P
nmðiÞ
to minimize WðfXgÞ ¼ xi qj Lj
ð9:4Þ
 i¼1 j¼1
gj ðfXgÞ  0; j ¼ 1; 2; . . .; nc
subjected to:
ximin  xi  ximax

where {X} is the vector containing the design variables; W({X}) presents the weight
of the structure; nm(i) is the number of members for the ith group; qj and Lj denote
the material density and the length of the jth member, respectively. ximin and ximax
are the lower and upper bounds of the design variable xi, respectively. gj({X})
denotes design constraints; and nc is the number of the constraints.
For constraints handling, a penalty approach is utilized. For this purpose, the
objective function (Eq. 9.4) is redefined as follows:

PðfXgÞ ¼ ð1 þ e1 :tÞe2  WðfXgÞ ð9:5Þ

where P({X}) is the penalized cost function or the objective function to be mini-
mized and t denotes the sum of the violations of the design constraints. In this
chapter, e1 is set to unity and e2 is calculated by
iter
e2 ¼ 1:5 þ 1:5  ð9:6Þ
itermax

where iter is the current iteration number and itermax is the total number of iterations
for optimization process.
9.3 Design Examples 143

9.3 Design Examples

Optimum seismic designs of three steel space frames are investigated in this sec-
tion. These design examples contain:
• The four-story 132-member steel frame
• The four-story 428-member steel frame
• The twelve-story 276-member steel frame
The algorithms select suitable sections from the complete W-section list given in
AISC which consists of 273 sections starting from W6  8.5 to W14  730. These
sections with their properties are used to prepare a design pool. The sequence
numbers assigned to this pool that are sorted with respect to the area of sections are
considered as design variables. In other words, the design variables represent a
selection from a set of integer numbers between 1 and 273. A population of 20
agents is used for all the algorithms. In the first and the last two examples, the
optimization process is terminated after 150 and 200 iterations, respectively.
The frames are subjected to gravity and earthquake loads and are designed
according to the LRFD-AISC design criteria [13]. Load combinations recom-
mended by ASCE 7–10 are considered in Ref. [14] and the frames are Intermediate
Moment Frames (IMF). The gravity loads are as follows: the design dead load is
2.88 kN/m2, the design live load is 2.39 kN/m2, and the ground snow load is
0.755 kN/m2 [15]. Iran has a long history of seismicity and has experienced
destructive earthquakes since ancient times. The history and the existence of faults
indicate the seismicity of metropolitan Tehran and this city are considered as high
seismicity area. Therefore, the response spectrum function of Tehran is utilized in
this study [16]. Figure 9.2 shows the design response spectrum as acceleration
versus period. The details of this figure are reported in Table 9.1. Occupancy
importance and redundancy factor are considered as 1. Modal and directional
combinations of response spectrum load cases are performed by CQC and SRSS
methods, respectively. In all the seismic load combinations, 100% of the design
seismic force for one direction is added to the 30% of the design seismic force for
the perpendicular direction. The vertical seismic load effect is also considered (SDS
is set to 1.05). For the first two examples, story drift is limited to 8.75 cm and for
the last problem, it is limited to 7 cm.
In the following examples, the height of each story is taken as 3.5 m and each
frame member is modeled as a line element. A992 steel is used with modulus of
elasticity equal to E = 200 GPa, yield stress of Fy = 250 MPa, and weight per unit
volume of q = 7850 kgf/m3. Direct analysis and general second-order methods are
used for the models. Tau-b fixed is employed for stiffness reduction method.
144 9 Optimum Seismic Design of 3D Steel Frames

Fig. 9.2 Design response spectrum

Table 9.1 Design response spectrum details

T (S) Sa (g) T (S) Sa (g)
0.000 3.432 2.000 2.789
0.100 8.580 2.100 2.697
0.150 8.580 2.200 2.613
0.200 8.580 2.300 2.537
0.300 8.580 2.400 2.467
0.400 8.580 2.500 2.402
0.500 8.580 2.600 2.343
0.600 7.293 2.700 2.288
0.700 6.374 2.800 2.237
0.800 5.684 2.900 2.189
0.900 5.148 3.000 2.145
1.000 4.719 3.100 2.104
1.100 4.368 3.200 2.065
1.200 4.076 3.300 2.028
1.300 3.828 3.400 1.994
1.400 3.616 3.500 1.961
1.500 3.432 3.600 1.931
1.600 3.271 3.700 1.902
1.700 3.129 3.800 1.874
1.800 3.003 3.900 1.848
1.900 2.890 4.000 1.823
9.3 Design Examples 145

9.3.1 A Four-Story 132-Member Steel Frame

The plan and 3D views of the first example are shown in the Figs. 9.3 and 9.4,
respectively. This is an irregular steel frame with 70 joints and 132 members which
are collected in 16 independent design variables. The member grouping is given in
Table 9.2. In order to optimize this structure by MDVC-UVPS, three stages are
considered. The design variable configuration utilized for the first and second stages
are listed as follows: first stage: [1 2 3 4], [5 6 7 8], [9 10 11 12], and [13 14 15 16];
second stage: [1 2], [3 4], [5 6], [7 8], [9 10], [11 12], [13 14], and [15 16].
The optimal designs found by the different algorithms are compared in Table 9.3
that shows also the corresponding structural weights and required number of
structural analyses. The MDVC-UVPS obtained 304.89 kN after 1775 analyses
which is better than 362.83 kN found by CBO after 2900 analyses, 325.98 kN
achieved by ECBO after 2740 analyses and 349.83 kN obtained by VPS after 2720
analyses. The maximum values of the stress ratio for CBO, ECBO, VPS, and
MDVC-UVPS are 85, 95, 95, and 98%, respectively. Convergence history dia-
grams are depicted in Fig. 9.5. It can be seen that the intermediate designs found by
MDVC-UVPS are always better than those found by other algorithms.

Fig. 9.3 Plan view of the four-story 132-member steel frame problem
146 9 Optimum Seismic Design of 3D Steel Frames

Fig. 9.4 3D view of the four-story 132-member steel frame problem with a U-shape form

Table 9.2 Member grouping Story Corner Side Side Inner

of the four-story 132-member column column beam beam
steel frame problem
1 1 5 9 13
2 2 6 10 14
3 3 7 11 15
4 4 8 12 16

9.3.2 A Four-Story 428-Member Steel Frame

The second design example is selected as a space four-story steel frame with 215
joints and 428 members. Figures 9.6 and 9.7 show the plan and 3D views of the
structure. This is a size optimization problem with 20 variables and the member
grouping of the frame members is listed in Table 9.4. Three stages with 5, 10, and
20 variables are considered by MDVC-UVPS algorithm. These DVCs are as
9.3 Design Examples 147

Table 9.3 Performance comparison of the four-story 132-member steel frame problem
Element group Optimal W-shaped sections
CBO ECBO VPS MDVC-UVPS
1 W21  68 W18  60 W18  97 W18  55
2 W14  82 W16  45 W18  86 W18  46
3 W14  74 W12  50 W18  76 W16  67
4 W14  99 W10  88 W16  77 W14  48
5 W16  50 W21  55 W18  50 W24  55
6 W14  68 W18  65 W16  57 W16  57
7 W12  58 W12  45 W12  50 W16  50
8 W10  88 W12  53 W12  58 W14  48
9 W14  22 W16  26 W14  22 W16  26
10 W12  26 W16  26 W12  26 W16  26
11 W14  22 W12  26 W14  22 W8  21
12 W10  22 W8  24 W10  22 W10  22
13 W12  30 W14  26 W14  26 W16  26
14 W14  26 W8  28 W12  30 W16  31
15 W12  22 W10  22 W12  19 W10  22
16 W6  25 W12  22 W12  22 W14  22
Weight (kN) 362.83 325.98 349.83 304.89
Number of structural analyses 2900 2740 2720 1775

Fig. 9.5 Convergence curves for the four-story 132-member steel frame problem
148 9 Optimum Seismic Design of 3D Steel Frames

Fig. 9.6 Plan view of the four-story 428-member steel frame problem

Fig. 9.7 3D view of the four-story 428-member steel frame problem with one plane of symmetry
9.3 Design Examples 149

Table 9.4 Member grouping of the four-story 428-member steel frame problem
Story Corner column Side column Inner column Side beam Inner beam
1 1 5 9 13 17
2 2 6 10 14 18
3 3 7 11 15 19
4 4 8 12 16 20

Table 9.5 Performance comparison of the four-story 428-member steel frame problem
Element group Optimal W-shaped sections
CBO ECBO VPS MDVC-UVPS
1 W30  90 W14  61 W27  84 W16  50
2 W21  83 W12  58 W14  68 W14  68
3 W14  68 W12  53 W14  68 W14  53
4 W14  90 W10  54 W12  72 W14  48
5 W14  38 W24  62 W18  65 W21  68
6 W14  61 W21  68 W18  65 W21  68
7 W12  87 W18  60 W18  60 W21  50
8 W12  53 W16  67 W16  36 W16  45
9 W18  106 W40  167 W24  62 W27  84
10 W16  100 W40  149 W24  62 W18  65
11 W16  89 W36  160 W14  74 W18  86
12 W14  82 W33  169 W14  61 W14  211
13 W14  26 W16  26 W16  26 W14  26
14 W10  19 W12  30 W12  26 W14  26
15 W24  55 W16  26 W14  30 W16  26
16 W5  16 W12  30 W12  30 W14  26
17 W16  31 W12  26 W16  26 W12  26
18 W12  35 W12  30 W10  26 W12  26
19 W8  24 W10  26 W12  22 W12  22
20 W16  31 W10  22 W10  22 W8  21
Weight (kN) 1247.72 1170.9 1115.68 1064.17
Number of structural 3700 3660 3520 1931
analyses

follows: first stage: [1 2 3 4], [5 6 7 8], [9 10 11 12], [13 14 15 16], and [17 18
19 20]; second stage: [1 2], [3 4], [5 6], [7 8], [9 10], [11 12], [13 14], [15 16], [17
18], and [19 20].
Table 9.5 presents the optimum designs obtained by proposed algorithms. The
lightest design (i.e., 1064.17 kg) is achieved by MDVC-UVPS algorithm after 1931
analyses. The best designs obtained by the CBO, ECBO, and VPS are 1247.72,
150 9 Optimum Seismic Design of 3D Steel Frames

Fig. 9.8 Constraint margins for the best design obtained by MDVC-UVPS algorithm in the
four-story 428-member steel frame problem: a element stress ratio and b story drift

1170.9 and 1115.68 kN, respectively. These values are found after 3700, 3660 and
3520 analyses. Figure 9.8 demonstrates the existing stress ratios and inter-story
drifts for the best designs of MDVC-UVPS. The maximum stress ratio is 93% while
maximum drift is 8.71 cm. The maximum stress ratios for the best designs of the
CBO, ECBO, and VPS are 87, 80, and 90%, respectively.

9.3.3 A Twelve-Story 276-Member Steel Frame

The plan and 3D views of the twelve-story frame in Figs. 9.9 and 9.10 show an
irregular steel space frame composed of 130 joints and 276 members. All members
are categorized into 24 design groups. The member grouping is given in Table 9.6.
Three stages with 8, 16, and 24 variables are considered by MDVC-UVPS algo-
rithm. These DVCs are as follows: first stage [1 2 3 4], [5 6], [7 8 9 10], [11 12], [13
14 15 16], [17 18], [19 20 21 22], and [23 24]; second stage: [1 2], [3 4], [5], [6], [7
8], [9 10], [11], [12], [13 14], [15 16], [17], [18], [19 20], [21 22], [23], and [24].
In Table 9.7, the minimum weight designs of the twelve-story 276-member steel
frame problem obtained by the proposed algorithms are compared. The
MDVC-UVPS obtained the lightest design compared to other methods that is
1644.31 kN. This algorithm requires 1576 structural analyses to find the optimum
solution while CBO, ECBO, and VPS require 3580, 3140, and 3440 structural
analyses, respectively. The maximum values of the stress ratio for the CBO, ECBO,
VPS, and MDVC-UVPS are 84, 97, 79, and 94%, respectively. Convergence his-
tory diagrams are depicted in Fig. 9.11 and demonstrated that the intermediate
designs found by MDVC-UVPS are always better than those found by other
9.3 Design Examples 151

Fig. 9.9 Plan view of the twelve-story 276-member steel frame problem

algorithms. The amount of saving in structural analyses in each iteration of the

MDVC-UVPS is shown in Fig. 9.12.

9.4 Concluding Remarks

Optimum design of real-size steel space frames under design code provisions is a
complicated optimization problem due to the presence of large numbers of highly
nonlinear constraints and discrete design variables. Main purpose of this chapter is
to propose a suitable meta-heuristic optimization algorithm for optimum design of
this class of problems. Three irregular steel frame optimization problems are con-
sidered and the stress constraints of LRFD-AISC, maximum lateral displacement
constraints, and geometric constraints are imposed on all frames. Numerical results
demonstrate that the MDVC-UVPS achieves the lightest designs in all the con-
sidered examples. This algorithm also shows a good convergence rate and comes
close to the optimum design rapidly. In the first example, the best weight found by
this method is 19, 7, and 14% lighter than the weight of the best design obtained
using the CBO, ECBO, and VPS, respectively. In the second example, the optimum
152 9 Optimum Seismic Design of 3D Steel Frames

Fig. 9.10 3D view of the

twelve-story 276-member
steel frame problem with
corner shaped form

design produced by the hybrid method is 17, 10, and 5% lighter than the ones
attained by other techniques, respectively. These values are, respectively, 15, 11,
and 4% for the last example. Besides, average computational time required for
MDVC-UVPS is about 35% less than the other algorithms.
9.4 Concluding Remarks 153

Table 9.6 Member grouping of the twelve-story 276-member steel frame problem
Story Corner column Side column Side beam Inner beam
1–2 1 7 13 19
3–4 2 8 14 20
5–6 3 9 15 21
7–8 4 10 16 22
9–10 5 11 17 23
11–12 6 12 18 24

Table 9.7 Performance comparison of the twelve-story 276-member steel frame problem
Element group Optimal W-shaped sections
CBO ECBO VPS MDVC-UVPS
1 W30  173 W33  130 W33  141 W33  118
2 W27  194 W24  162 W33  130 W21  101
3 W24  146 W24  146 W33  118 W14  132
4 W24  131 W18  192 W30  132 W14  99
5 W16  67 W14  120 W14  68 W14  90
6 W14  145 W10  100 W12  79 W12  170
7 W36  231 W33  152 W33  201 W40  235
8 W33  241 W33  152 W30  211 W40  199
9 W27  258 W27  194 W30  191 W36  232
10 W24  306 W27  194 W27  194 W33  221
11 W14  176 W24  335 W24  335 W33  169
12 W12  230 W24  250 W24  250 W27  217
13 W16  40 W21  57 W21  57 W21  62
14 W16  36 W18  65 W16  67 W24  62
15 W16  45 W21  55 W21  55 W21  55
16 W18  40 W21  55 W18  65 W18  60
17 W16  45 W14  48 W12  50 W16  31
18 W8  21 W12  50 W12  53 W14  48
19 W27  217 W16  40 W16  40 W18  40
20 W24  250 W16  40 W16  45 W16  45
21 W36  182 W12  53 W12  53 W12  53
22 W33  169 W12  50 W12  50 W12  50
23 W8  31 W40  264 W27  129 W33  141
24 W12  26 W8  67 W8  67 W12  35
Weight (kN) 1896.06 1824.64 1716.41 1644.31
Number of structural 3580 3140 3440 1576
analyses
154 9 Optimum Seismic Design of 3D Steel Frames

Fig. 9.11 Convergence curves for the twelve-story 276-member steel frame problem

Fig. 9.12 Saving in structural analyses using the MDVC-UVPS algorithm in the twelve-story
276-member steel frame problem

References

1. Hasançebi O, Çarbaş S, Doğan E, Erdal F, Saka MP (2010) Comparison of non-deterministic

search techniques in the optimum design of real size steel frames. Comput Struct 88
(17):1033–1048
2. Aydogdu I, Saka MP (2012) Ant colony optimization of irregular steel frames including
elemental warping effect. Adv Eng Softw 44(1):150–169
3. Kaveh A, Zakian P (2013) Optimal design of steel frames under seismic loading using two
meta-heuristic algorithms. J Construct Steel Res 82:111–130
References 155

4. Hasançebi O, Carbas S (2014) Bat inspired algorithm for discrete size optimization of steel
frames. Adv Eng Softw 67:173–185
5. Talatahari S, Gandomi AH, Yang XS, Deb S (2015) Optimum design of frame structures
using the eagle strategy with differential evolution. Eng Struct 9:16–25
6. Artar M, Daloglu AT (2015) Optimum design of composite steel frames with semi-rigid
connections and column bases via genetic algorithm. Steel Compos Struct 19(4):1035–1053
7. Kaveh A, Bakhshpoori T, Azimi M (2015) Seismic optimal design of 3D steel frames using
cuckoo search algorithm. Struct Des Tall Special Build 24(3):210–227
8. Carbas S (2016) Design optimization of steel frames using an enhanced firefly algorithm. Eng
Optim 48(12):2007–2025
9. Kaveh A, Bolandgerami A (2017) Optimal design of large-scale space steel frames using
cascade enhanced colliding body optimization. Struct Multidiscip Optim 55(1):237–256
10. Kaveh A, Ghafari MH, Gholipour Y (2017) Optimum seismic design of steel frames
considering the connection types. J Construct Steel Res 130:79–87
11. Kaveh A, Ghafari MH, Gholipour Y (2017) Optimal seismic design of 3D steel moment
frames: different ductility types. Struct Multidiscip Optim 56(6):1353–1367
12. Kaveh A, Ilchi Ghazaan M (2018) Optimum seismic design of 3D irregular steel frames using
modern meta-heuristic algorithms. J Comput Civil Eng, 32(3), 04018015
13. AISC 360-10 (2010) Specification for structural steel buildings. American Institute of Steel
Construction, Chicago, Illinois 60601-1802, USA
14. ASCE 7-10 (2010) Minimum design loads for building and other structures, Chicago, USA
15. Aydoğdu İ, Akın A, Saka MP (2016) Design optimization of real world steel space frames
using artificial bee colony algorithm with levy flight distribution. Adv Eng Softw 92:1–14
16. ICSRDB (2015) Iranian code of practice for seismic resistant design of buildings (Standard
2800), 4rd edn, PN S 253, Building and Housing Research Center of Iran
Appendix A
Configuration Processing

A.1 Introduction
For large systems, configuration processing is one of the most tedious and
time-consuming parts of the analysis. Different methods have been proposed for
configuration processing and data generation, among which the formex algebra of
Nooshin [1, 2] is perhaps the most general tool for this purpose. Behravesh et al. [3]
employed set theory and showed that some concepts of set algebra can be used to
build up a general method for describing the interconnection patterns of structural
systems. There are many other references on the field of data generation; however,
most of them are prepared for specific classes of problem. For example, many
algorithms have been developed and successfully implemented on mesh or grid
generation, a complete review of which may be found in a paper by Thacker [4] and
a book by Thomson et al. [5].
Graph theoretical methods for the formation of structural and finite element
models are developed by Kaveh [6, 7]. In all these methods, a submodel is
expressed in algebraic form and then the functions are used for the formation of the
entire model. The main functions employed consist of translation, rotation,
reflection, and projection, or combination of these functions. On the other hand,
many structural models can be viewed as the graph products of two or three
subgraphs, known as their generators. Many properties of structural models can be
obtained by considering the properties of their generators. This simplifies many
complicated calculations, particularly in relation with eigen solution of regular
structures, as shown by Kaveh and Rahami [8, 9] and Kaveh and Mirzaie [10]. Four
undirected and directed graph products are presented in [11] for the formation of
structural models. The undirected products are extensively used in graph theory and
combinatorial optimization, however, the directed products are more suitable for the
formation of practical structural models.
Different structures are studied in this book and all of them, except the 3D
frames, are analyzed using the direct stiffness method by our own codes. In this
appendix, MATLAB codes for generation of the coordinates and connectivities of
three different types of structures utilized in this book are developed. These

© Springer International Publishing AG, part of Springer Nature 2018 157

A. Kaveh and M. Ilchi Ghazaan, Meta-heuristic Algorithms for Optimal Design
of Real-Size Structures, https://doi.org/10.1007/978-3-319-78780-0
158 Appendix A: Configuration Processing

structures consist of a 1016-bar double-layer grid, a 693-bar double-layer barrel

vault, and a 600-bar dome truss. Using the concepts employed in these codes, all
the structures studied in the book can be simply generated.

A.2 Examples

In this section, configuration processing of three types of space structures are

discussed in detail. These structures are as follows:
• A double-layer grid
• A double-layer barrel vault
• A double-layer dome.

A.2.1 Example 1

Consider a double-layer grid as shown in Fig. A.1. It is a 40
40 m grid with

certain bays of equal length in two directions and the height is equal to 3 m. The
structure consists of 1016 members and 320 joints. The geometry and the mem-
ber’s labels of the structure are depicted in Fig. A.2. The MATLAB code for
generation of coordinates and connectivities of this grid is listed below.

Fig. A.1 3D view of the 1016-bar double-layer grid

Appendix A: Configuration Processing 159

Fig. A.2 a bottom layer, b top layer, and c web members of the 1016-bar double-layer grid
160 Appendix A: Configuration Processing

% MATLAB codes for generation of coordinates and connectivities

% example1.m

% NUM_X: number of parts in x direction bottom top layer

% NUM_Y: number of parts in y direction bottom top layer
% LEN_X: length of bays in x direction
% LEN_Y: length of bays in y direction
NUM_X = 10;
NUM_Y = 10;
LEN_X = 4;
LEN_Y = 4;

% Node coordinates of bottom-layer

% nodes: node coordinates
nodes = [];
for ny=1:NUM_Y
x = [LEN_X / 2 + LEN_X * (0:NUM_X-1), LEN_X * (0:NUM_X)];
y = [repmat(LEN_Y * (ny-1), 1, NUM_X), repmat(LEN_Y / 2 + LEN_Y *
(ny - 1), 1, NUM_X + 1)];
z = zeros(1, 2 * NUM_X + 1);
nodes = [nodes; [x; y; z]'];
end
nodes = [nodes; [LEN_X / 2 + LEN_X * (0:NUM_X-1); repmat(LEN_Y * NUM_Y, 1,
NUM_X); zeros(1,NUM_X)]'];

% Node coordinates of top-layer

xTop = LEN_X/2 + repmat(LEN_X * (0:NUM_X-1), 1, NUM_Y);
yTop = LEN_Y/2 + reshape(repmat(LEN_Y * (0:NUM_Y-1), NUM_X, 1), [], 1)';
zTop = repmat(3, 1, NUM_X * NUM_Y);
nodes = [nodes; [xTop; yTop; zTop]'];

% Bottom-layer elements
% elements: connections at elements
elements = [];
elements = [elements; [1:NUM_X-1; 2:NUM_X]'];
for ny=1:NUM_Y
firstNodeBottomRow = (2 * NUM_X + 1) * (ny - 1) + 1;
firstNodeTopRow = (2 * NUM_X + 1) * (ny - 1) + NUM_X + 1;
for nx=1:NUM_X
elements = [elements; [firstNodeBottomRow + nx - 1, firstNodeTopRow +
nx - 1]];
Appendix A: Configuration Processing 161

elements = [elements; [firstNodeBottomRow + nx - 1, firstNodeTopRow +

nx]];
end
for nx=1:NUM_X+1
if nx ~= 1
elements = [elements; [firstNodeTopRow + nx - 1,
firstNodeTopRow + nx - 1 + NUM_X]];
elseif ny ~= NUM_Y
elements = [elements; [firstNodeTopRow, firstNodeTopRow +
(2 * NUM_X + 1)]];
end
if nx ~= NUM_X+1
elements = [elements; [firstNodeTopRow + nx - 1,
firstNodeTopRow + nx - 1 + NUM_X + 1]];
elseif ny ~= NUM_Y
elements = [elements; [firstNodeTopRow + NUM_X, firstNodeTopRow +
NUM_X + (2 * NUM_X + 1)]];
end
end
end
elements = [elements; (2 * NUM_X + 1) * NUM_Y + [1:NUM_X-1; 2:NUM_X]'];

% Top-layer elements
firstNodeTopLayer = (2 * NUM_X + 1) * NUM_Y + NUM_X + 1;
for ny=1:NUM_Y
firstNode = firstNodeTopLayer + (ny - 1) * NUM_X;
elements = [elements; [firstNode:firstNode+NUM_X-2;
firstNode+1:firstNode+NUM_X-1]'];
if ny ~= NUM_Y
elements = [elements; [firstNode:firstNode+NUM_X-1;
firstNode+NUM_X:firstNode+2*NUM_X-1]'];
end
end

% Web elements
for ny=1:NUM_Y
firstNodeBot = (2 * NUM_X + 1) * (ny - 1) + 1;
firstNodeTop = firstNodeTopLayer + (ny - 1) * NUM_X;
elements = [elements; [firstNodeBot:firstNodeBot+NUM_X-1;
firstNodeTop:firstNodeTop+NUM_X-1]'];
for nx=1:NUM_X
elements = [elements; [firstNodeBot + NUM_X + nx - 1, firstNodeTop +
162 Appendix A: Configuration Processing

nx - 1]];
elements = [elements; [firstNodeBot + NUM_X + nx, firstNodeTop + nx –
1]];
end
elements = [elements; [firstNodeBot+2*NUM_X+1:firstNodeBot+3*NUM_X;
firstNodeTop:firstNodeTop+NUM_X-1]'];
end

A.2.2 Example 2

The double-layer barrel vault presented in Fig. A.3 consists of 693 members and
259 joints. The geometry and the member’s labels of the structure are shown in
Fig. A.4.
In geometry, a circular segment is a region of a circle which is “cut off” from the
rest of the circle by a chord. Let R be the radius of the circle, h the central angle,
c the chord length, s the arc length, h the height of the segment, and d the height of
the triangular portion (Fig. A.5); therefore, we have

h c2
R ¼ hþd ¼ þ ðA:1Þ
2 8h
c
h ¼ 2 arctan ðA:2Þ
2d

For the top layer of the barrel vault, hT = 5.75 m and cT = 22.9 m. Therefore, RT
and hT are 14.275 m and 1.862 rad, respectively. For the bottom layer, hB = 4.25 m
and cB = 19.03 m. Thus, RB and hB are 12.776 m and 1.680 rad, respectively.
The connections at elements and node coordinates of the double-layer barrel
vault can be generated by the following MATLAB code.

Fig. A.3 3D view of the 693-bar double-layer barrel vault

Appendix A: Configuration Processing 163

Fig. A.4 a flatten cross-sectional view and b plan view of the 693-bar double-layer barrel vault

Fig. A.5 A circular segment is enclosed between a chord (the dashed line) and the arc whose
endpoints equal the chord
164 Appendix A: Configuration Processing

% MATLAB codes for generation of coordinates and connectivities

% example2.m

%NUM_X: number of bays in x direction for top layer

%NUM_Z: number of bays in z direction
NUM_X = 18;
NUM_Z = 6;

% Node coordiantes of top-layer

% H_TOP: height of segment for top layer
% C_TOP: chord length for top layer
% R_TOP: radius of circle for top layer
% THETA_TOP: central angle for top layer
% nodes: node coordinates
H_TOP = 5.75;
C_TOP = 22.9;
R_TOP = (H_TOP/2) + (C_TOP^2)/(8*H_TOP);
THETA_TOP = 2 * atan(C_TOP/(2*(R_TOP-H_TOP)));
nodes = [];
for nz=1:NUM_Z+1
x = sin(THETA_TOP/2) - cos((pi - THETA_TOP)/2 + (0:NUM_X) *
(THETA_TOP / NUM_X));
y = -cos(THETA_TOP/2) + sin((pi - THETA_TOP)/2 + (0:NUM_X) *
(THETA_TOP / NUM_X));
nodes = [nodes; [R_TOP * x; R_TOP * y; repmat(9 / NUM_Z * (nz - 1), 1,
NUM_X+1)]'];
end

% Node coordinates of bottom-layer

% H_BOT: height of segment for bottom layer
% C_BOT: chord length for bottom layer
% R_BOT: radius of circle for bottom layer
% THETA_BOT: central angle for bottom layer
H_BOT = 4.25;
C_BOT = 19.03;
R_BOT = (H_BOT/2) + (C_BOT^2)/(8*H_BOT);
THETA_BOT = 2 * atan(C_BOT/(2*(R_BOT-H_BOT)));
for nz=1:NUM_Z+1
x = sin(THETA_BOT/2) - cos((pi - THETA_BOT)/2 + (0:NUM_X-1) *
(THETA_BOT / (NUM_X - 1)));
y = -cos(THETA_BOT/2) + sin((pi - THETA_BOT)/2 + (0:NUM_X-1) *
(THETA_BOT / (NUM_X - 1)));
nodes = [nodes; [1.935 + R_BOT * x; R_BOT * y; repmat(9 / NUM_Z *
Appendix A: Configuration Processing 165

(nz - 1), 1, NUM_X)]'];

end

% Top-layer elements
% elements: connections at elements
elements = [];
for nz=1:NUM_Z+1
firstNode = (nz - 1) * (NUM_X + 1) + 1;
elements = [elements; [firstNode:firstNode+NUM_X-1;
firstNode+1:firstNode+NUM_X]'];
if nz < NUM_Z+1
for nx=1:NUM_X
if (nx <= NUM_X/2 && nz <= NUM_Z/2) ||
(nx > NUM_X/2 && nz > NUM_Z/2)
elements = [elements; [firstNode + nx, firstNode + nx +
NUM_X]];
else
elements = [elements; [firstNode + nx - 1, firstNode + nx +
NUM_X + 1]];
end
if nx ~= NUM_X
elements = [elements; [firstNode + nx, firstNode + nx +
NUM_X + 1]];
end
end
end
end

% Bottom-layer elements
for nz=1:NUM_Z+1
firstNode = (NUM_X + 1) * (NUM_Z + 1) + (nz - 1) * NUM_X + 1;
elements = [elements; [firstNode:firstNode+NUM_X-2;
firstNode+1:firstNode+NUM_X-1]'];
end

% Web elements
for nz=1:NUM_Z+1
firstNodeBot = (NUM_X + 1) * (NUM_Z + 1) + (nz - 1) * NUM_X + 1;
firstNodeTop = (nz - 1) * (NUM_X + 1) + 1;
for nx=1:NUM_X-1
elements = [elements; [firstNodeBot + nx - 1, firstNodeTop + nx]];
elements = [elements; [firstNodeBot + nx, firstNodeTop + nx]];
end
end
166 Appendix A: Configuration Processing

A.2.3 Example 3

The dome truss shown in Fig. A.6 consists of 600 elements and 216 joints. Fig. A.7
presents a substructure in more detail for nodal numbering and coordinates. The
following MATLAB code is developed to generate the dome from its substructure.

Fig. A.6 3D view of the 600-bar dome truss

Fig. A.7 A substructure of

the 600-bar dome truss
Appendix A: Configuration Processing 167

% MATLAB codes for generation of coordinates and connectivities

% example3.m

% NUM_S: number of substructures

% NUM_NODE: number of substructure nodes
% THETA: angle of revolution
% RADIUS: distance of the the substructure nodes from z axis
% % Z_COORDS: z coordinates of substructure nodes
NUM_S = 24;
NUM_NODE = 9;
THETA = 2 * pi / NUM_S;
RADIUS = [1 1 3 5 7 9 11 13 14];
Z_COORDS = [7 7.5 7.25 6.75 6 5 3.5 1.5 0];

% Node coordinates
% nodes: node coordinates
nodes = [];
for ns=1:NUM_S
nodes = [nodes; [RADIUS * cos((ns-1) * THETA); RADIUS * sin((ns-1) *
THETA); Z_COORDS]'];
end

% Elements
% elements: connections at elements
elements = [];
for ns=1:NUM_S
firstNodeCur = (ns - 1) * NUM_NODE + 1;
if ns ~= NUM_S
firstNodeNext = firstNodeCur + NUM_NODE;
else
firstNodeNext = 1;
end
elements = [elements; [firstNodeCur:firstNodeCur+NUM_NODE-2;
firstNodeCur+1:firstNodeCur+NUM_NODE-1]'];
elements = [elements; [firstNodeCur, firstNodeCur + 2]];
elements = [elements; [firstNodeCur:firstNodeCur+NUM_NODE-2;
firstNodeNext:firstNodeNext+NUM_NODE-2]'];
elements = [elements; [firstNodeCur+2:firstNodeCur+NUM_NODE-1;
firstNodeNext+1:firstNodeNext+NUM_NODE-2]'];
elements = [elements; [firstNodeCur, firstNodeNext + 1]];
end
168 Appendix A: Configuration Processing

References
1. Nooshin H (1975) Algebraic representation and processing of structural configurations.
Comput Struct 5(2):119–130
2. Nooshin H (1984) Formex configuration processing in structural engineering. Elsevier
Applied Science Publishers Ltd, Essex, U.K.
3. Behravesh A, Kaveh A, Nani M, Sabet S (1988) A set theoretical approach for configuration
processing. Comput Struct 30(6):1293–1302
4. Thacker WC (1980) A brief review of techniques for generating irregular computational grids.
Int J Numer Methods Eng 15(9): 1335–1341
5. Thompson JF, Warsi ZU, Mastin CW (1985) Numerical grid generation: foundations and
applications. Elsevier, Amsterdam
6. Kaveh A (1993) A graph-theoretical approach to configuration processing. Comput Struct 48
(2):357–363
7. Kaveh A (2004) Structural mechanics: graph and matrix methods. 3rd edn. Research Studies
Press, Somerset, UK
8. Kaveh A, Rahami H (2006) Block diagonalization of adjacency and Laplacian matrices for
graph product; applications in structural mechanics. Int J Numer Methods Eng 68(1):33–63
9. Kaveh A, Rahami H (2005) A unified method for eigendecomposition of graph products. Int J
Numer Methods Biomed Eng 21(7):377–388
10. Kaveh A, Mirzaie R (2008) Minimal cycle basis of graph products for the force method of
frame analysis. Int J Numer Methods Biomed Eng 24(8):653–669
11. Kaveh A, Koohestani K (2008) Graph products for configuration processing of space
structures. Comput Struct 86(11–12):1219–1231