Otimização robusta multiobjetivo para o projeto de sistemas em Engenharia

Abstract

The aim of this work was the development of a robust multiobjective optimization algorithm by using as a reference the Fireflies Colony Algorithm associated with the concept of Effective Mean. The main operators for the extension of the algorithm for multiobjective case were the ordering of Pareto curves by means of ranking procedures and the truncation of solutions through the crowding distance operator. To insert robustness to the optimization process, the mean effective definition was used instead the commonly used expectancy measures, as suggested by the literature. The proposed methodology was tested in mathematical problems whose nominal and robust Pareto curves were known. In addition, so as to evaluate the quality of the algorithm proposed metrics of convergence and diversity were taken into account. The proposed algorithm proved to be very efficient with respect to convergence and diversity of solutions. The methodology was also applied to design classical mechanical systems, including the design a flexible rotor with respect to the position of the critical speeds. The main contributions of this thesis was the development of a computational tool for the treatment of robust multi-objective optimization problems, the analysis and interpretation of the influence of robustness parameter on robust Pareto curves compared with the nominal curves and the formation of a bank data for future comparisons with other robust multi-objective optimization strategies. The results indicate that the approach proposed arises as an interesting strategy for the robust design of engineering systems.