A selective genetic algorithm for multiobjective optimization of cross sections in 3D trussed structures based on a spatial sensitivity analysis

• Autores: Mroginski, Javier Luis; Beneyto, Pablo Alejandro; Gutiérrez, Guillermo José; Di Rado, Héctor Ariel
• Año de publicación: 2016
• Idioma: inglés
• Tipo de recurso: artículo
• Fuente: Multidiscipline Modeling in Materials and Structures, 2016, vol. 12, no. 2, p. 423-435.
• Materia: Sensitivity analysis; Genetic algorithm; Finite element method; 3D bars structure; Multiobjective optimization
• Institución: Universidad Nacional del Nordeste

Fil: Mroginski, Javier Luis. Universidad Nacional del Nordeste. Facultad de Ingeniería; Argentina.
Fil: Mroginski, Javier Luis. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina.
Fil: Beneyto, Pablo Alejandro. Universidad Nacional del Nordeste. Facultad de Ingeniería; Argentina.
Fil: Gutiérrez, Guillermo José. Universidad Nacional del Nordeste. Facultad de Ingeniería; Argentina.
Fil: Di Rado, Héctor Ariel. Universidad Nacional del Nordeste. Facultad de Ingeniería; Argentina.
Purpose – There are many problems in civil or mechanical engineering related to structural design. In such a case, the solution techniques which lead to deterministic results are no longer valid due to the heuristic nature of design problems. The purpose of this paper is to propose a computational tool based on genetic algorithms, applied to the optimal design of cross-sections (solid tubes) of 3D truss structures. Design/methodology/approach – The main feature of this genetic algorithm approach is the introduction of a selective-smart method developed in order to improve the convergence rate of large optimization problems. This selective genetic algorithm is based on a preliminary sensitivity analysis performed over each variable, in order to reduce the search space of the evolutionary process. In order to account for the optimization of the total weight, the displacement (of a specific section) and the internal stresses distribution of the structure a multiobjective optimization function was proposed. Findings – The numerical results presented in this paper show a significant improvement in the convergence rate as well as an important reduction in the relative error, compared to the exact solution. Originality/value – The variables sensitivity analysis put forward in this approach introduces a significant improvement in the convergence rate of the genetic algorithm proposed in this paper.