Otimização extrema generalizada: um novo algoritmo estocástico para o projeto ótimo / x

AUTOR(ES)
DATA DE PUBLICAÇÃO

2002

RESUMO

In this work a new numerical tool for application on optimal design is presented. Based on the theory of Self-Organized Criticality (SOC), it is intended to be used in problems that present complex characteristics such as a non-convex or even disjoint design space, the presence of multiple sub-optimal solutions on it, severe non-linearities on the objective function or on the constraints and the use of a combination of continuos, discrete and integer variables. Called the Generalized Extremal Optimization algorithm (GEO), it extends the Extremal Optimization (EO)method in a way that it can be readly applied to a broad class of optimal design problems. Altought being a stochastic algorithm, it has only one free parameter to be set, diferently to other popular algorithms, such as the Genetic Algorithm (GA)or the Simulated Annealing (SA), that each have at least three of them. This is an ""a priori"" advantage of the GEO over the GA and the SA, since the setting process of the free parameter would present a lesser computional cost for the GEO than for the others. The generalized extremal optimization algorithm is presented here in its canonical form and in an alternative implementation called GEOvar. The performance of both implementations was assessed for a set of test functions and compared to versions of the GA and SA, showing to be competitive to both algorithms. Used to tackle optimal design problems in aerospace engineering, the generalized extremal optimization algorithm showed to be capable to find high quality solutions (in some cases probably the optimal), even starting from infeasible designs. Moreover, it identified characteristics of these problems that were not intuitively obvious at first sight. From the point of view of the theory that inspired the algorithm, a preliminar analisys of the search dynamics to find the optimal, showed that it can present characteristics of SOC. Finaly, it can be said that the generalized extremal optimization algorithm showed to be a highly potential candidate to be incorporated to the optimization numerical tool box of the engineer and of the scientist.

ASSUNTO(S)

criticalidade auto-organizada projeto programação não linear optimization otimização de projeto multidisciplinar self-organized criticaly otimização design nonlinear programming multidisciplinary design optimization

ACESSO AO ARTIGO

Documentos Relacionados