PARALLEL ADAPTIVE SEARCH TECHNIQUES FOR STRUCTURAL OPTIMIZATION

AUTOR(ES)
FONTE

IBICT - Instituto Brasileiro de Informação em Ciência e Tecnologia

DATA DE PUBLICAÇÃO

199611

RESUMO

This thesis investigates the capabilities of two alternative search techniques from the field of Artificial Intelligence (AI), the Simulated Annealing (SA) and the Genetic Algorithm (GA), as general purpose procedures for optimum structural design. A number of studies from different researchers have proved that both SA and GA are capable of solving a variety of real-world structural optimization problems. Examples also are presented in this thesis that demonstrate the robustness and reliability of these techniques. One drawback of these techniques, the high computational time required to obtain refined solutions, is addressed by careful modifications of the standard algorithms in order to accelerate the convergence process while preserving the good features of such techniques. Moreover, the rapid evolution of hardware capabilities is substantially reducing the computational cost of these optimization processes and consequently these techniques should no longer be excluded due to their \"burdensome\" computational cost. Following the hardware advances, this thesis presents a range of possibilities for parallel implementations of SAs and GAs. A number of alternative parallel schemes are proposed to match different parallel hardware architectures. In addition, this thesis shows that when properly implemented in suitable parallel hardware, these optimization techniques are capable of rapid solution rates, thereby lending themselves to real-time applications.

ASSUNTO(S)

genetic algorithm ciencia da computacao artificial intelligence simulated annealing search techniques

ACESSO AO ARTIGO

Documentos Relacionados