Optimum Design of Steel Frames Using Different Variants of Differential Evolution Algorithm

D. Safari, Mahmoud R. Maheri* (Corresponding Author), A. Maheri

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)


A differential evolution (DE)-based algorithm is used for discrete optimization of steel frames. DE is a simple yet efficient population-based search algorithm, originally proposed for continuous optimization problems. It is based on the same search philosophy as most evolutionary algorithms (EA), utilizing mutation, crossover and selection operators. However, unlike traditional EAs, the DE creates new candidate solutions by perturbing the parent individual with the weighted difference of several other randomly chosen individuals of the same population. In this study, performance of DE in optimal design of steel frames is investigated. Eleven different variants of DE are tested through three benchmark problems. The comparison results between the DE and other metaheuristic algorithms, such as genetic algorithm (GA), ant colony optimizer (ACO) and particle swarm optimizer (PSO) methods, taken from the literature, show that in most cases the DE can perform as well as other techniques. It is found that two particular variants of DE, namely ‘DE/best/1’ and ‘DE/best/1 with jitter’ provide better results compared to the other variants. In both variants, the best in the iteration is selected as the base vector of DE algorithm for perturbation in mutation stage. Also, results show that low values for control parameter CR, can yield better designs.
Original languageEnglish
Pages (from-to)2091-2105
Number of pages15
JournalIranian Journal of Science and Technology, Transactions of Civil Engineering
Early online date26 Jul 2021
Publication statusPublished - 1 Dec 2021


  • Diferential evolution
  • Discrete optimization
  • Variants of differential evolution
  • Steel frames optimization


Dive into the research topics of 'Optimum Design of Steel Frames Using Different Variants of Differential Evolution Algorithm'. Together they form a unique fingerprint.

Cite this