Improving the efficiency of genetic algorithms for frame designs

S. Y. Chen, Subramaniam Rajan

Research output: Contribution to journalArticle

19 Citations (Scopus)

Abstract

The focus of this paper is on the development of a design software system that has enough flexibility and capability to search for the most economical steel roof truss design in a reasonable amount of time. This objective is achieved by improving the efficiency and robustness of the genetic algorithm (GA) methodology developed earlier. The effects of schema representation, schema survival, type of crossover, problem definition, the size of the population, and the number of design iterations on the computational expense and the value of the objective function are studied. The research results show that while the final AISI (American Iron and Steel Institute) code-conforming 'best' designs are very close to each other when different starting designs (ground structures) are used, the use of some GA strategies can lead to either highly non-optimal designs or design processes that are computationally expensive. The results also show some other interesting conclusions. The size of the population and the maximum number of design iterations (or generations) need to be at least the size of the chromosome. The schema representation is perhaps one of the most important factors. Depending on the complexity of the initial design (density of the ground structure) and the size of the chromosome, a newly developed Association String strategy has led to a computationally effective GA process when combined with the elitist, one-point and uniform crossover strategies.

Original languageEnglish (US)
Pages (from-to)281-307
Number of pages27
JournalEngineering Optimization
Volume30
Issue number3-4
StatePublished - 1998

Fingerprint

Genetic algorithms
Genetic Algorithm
Chromosomes
Schema
Steel
Roofs
Chromosome
Crossover
Iron
Iteration
Design
Genetic algorithm
Software System
Objective function
Strings
Flexibility
Robustness
Methodology
Strategy

Keywords

  • AISI
  • Frame design
  • Genetic algorithm
  • Optimal design
  • Roof truss
  • Structural optimization

ASJC Scopus subject areas

  • Management Science and Operations Research
  • Engineering (miscellaneous)

Cite this

Improving the efficiency of genetic algorithms for frame designs. / Chen, S. Y.; Rajan, Subramaniam.

In: Engineering Optimization, Vol. 30, No. 3-4, 1998, p. 281-307.

Research output: Contribution to journalArticle

@article{9c50898e5bd047e4b23b09c7b68ebf31,
title = "Improving the efficiency of genetic algorithms for frame designs",
abstract = "The focus of this paper is on the development of a design software system that has enough flexibility and capability to search for the most economical steel roof truss design in a reasonable amount of time. This objective is achieved by improving the efficiency and robustness of the genetic algorithm (GA) methodology developed earlier. The effects of schema representation, schema survival, type of crossover, problem definition, the size of the population, and the number of design iterations on the computational expense and the value of the objective function are studied. The research results show that while the final AISI (American Iron and Steel Institute) code-conforming 'best' designs are very close to each other when different starting designs (ground structures) are used, the use of some GA strategies can lead to either highly non-optimal designs or design processes that are computationally expensive. The results also show some other interesting conclusions. The size of the population and the maximum number of design iterations (or generations) need to be at least the size of the chromosome. The schema representation is perhaps one of the most important factors. Depending on the complexity of the initial design (density of the ground structure) and the size of the chromosome, a newly developed Association String strategy has led to a computationally effective GA process when combined with the elitist, one-point and uniform crossover strategies.",
keywords = "AISI, Frame design, Genetic algorithm, Optimal design, Roof truss, Structural optimization",
author = "Chen, {S. Y.} and Subramaniam Rajan",
year = "1998",
language = "English (US)",
volume = "30",
pages = "281--307",
journal = "Engineering Optimization",
issn = "0305-215X",
publisher = "Taylor and Francis Ltd.",
number = "3-4",

}

TY - JOUR

T1 - Improving the efficiency of genetic algorithms for frame designs

AU - Chen, S. Y.

AU - Rajan, Subramaniam

PY - 1998

Y1 - 1998

N2 - The focus of this paper is on the development of a design software system that has enough flexibility and capability to search for the most economical steel roof truss design in a reasonable amount of time. This objective is achieved by improving the efficiency and robustness of the genetic algorithm (GA) methodology developed earlier. The effects of schema representation, schema survival, type of crossover, problem definition, the size of the population, and the number of design iterations on the computational expense and the value of the objective function are studied. The research results show that while the final AISI (American Iron and Steel Institute) code-conforming 'best' designs are very close to each other when different starting designs (ground structures) are used, the use of some GA strategies can lead to either highly non-optimal designs or design processes that are computationally expensive. The results also show some other interesting conclusions. The size of the population and the maximum number of design iterations (or generations) need to be at least the size of the chromosome. The schema representation is perhaps one of the most important factors. Depending on the complexity of the initial design (density of the ground structure) and the size of the chromosome, a newly developed Association String strategy has led to a computationally effective GA process when combined with the elitist, one-point and uniform crossover strategies.

AB - The focus of this paper is on the development of a design software system that has enough flexibility and capability to search for the most economical steel roof truss design in a reasonable amount of time. This objective is achieved by improving the efficiency and robustness of the genetic algorithm (GA) methodology developed earlier. The effects of schema representation, schema survival, type of crossover, problem definition, the size of the population, and the number of design iterations on the computational expense and the value of the objective function are studied. The research results show that while the final AISI (American Iron and Steel Institute) code-conforming 'best' designs are very close to each other when different starting designs (ground structures) are used, the use of some GA strategies can lead to either highly non-optimal designs or design processes that are computationally expensive. The results also show some other interesting conclusions. The size of the population and the maximum number of design iterations (or generations) need to be at least the size of the chromosome. The schema representation is perhaps one of the most important factors. Depending on the complexity of the initial design (density of the ground structure) and the size of the chromosome, a newly developed Association String strategy has led to a computationally effective GA process when combined with the elitist, one-point and uniform crossover strategies.

KW - AISI

KW - Frame design

KW - Genetic algorithm

KW - Optimal design

KW - Roof truss

KW - Structural optimization

UR - http://www.scopus.com/inward/record.url?scp=0031618448&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0031618448&partnerID=8YFLogxK

M3 - Article

VL - 30

SP - 281

EP - 307

JO - Engineering Optimization

JF - Engineering Optimization

SN - 0305-215X

IS - 3-4

ER -