Uma abordagem neuro- nebulosa para otimização de sistema e indentificação robusta
AUTOR(ES)
Ivan Nunes da Silva
DATA DE PUBLICAÇÃO
1997
RESUMO
The ability of artificial neural networks to solve complex and diversified problems make them attractive for application in many áreas of engineering and science. A neural network is basically composed of many simple processing elements with a high degree of connectivity among them. This thesis presents an architecture of artificial neural network to apply in optimization problems with constraints. More specifically, a topology based in the Hopfield networks is proposed to solve some optimization problems, including dynamic programming and combinatorial optimization, and also robust parametric identification problems with unknown-but-bounded disturbance. In this case, the network is used to calculate the parametric uncertainty intervals and the feasible membership set for model parameters. A rule-based fuzzy system has been developed in order to improve the efficiency and the network convergence to the equilibrium points. The internal parameters of the network, associated with the constraints imposed by a specific problem, are explicitally computed using a valid-subspace technique. Simulation results are provided to illustrate the performance of the proposed approach
ASSUNTO(S)
redes neurais (computação) inteligencia artificial identificação de sistemas otimização matematica
ACESSO AO ARTIGO
http://libdigi.unicamp.br/document/?code=vtls000122872Documentos Relacionados
- Uma abordagem de otimização robusta no planejamento agregado de produção na indústria cítrica
- Uma abordagem nebulosa para solução de problemas de diagnóstico, investigação e tratamento de desordens.
- Abordagem robusta para controladores preditivos
- Uma arquitetura neuro-genética para otimização não-linear restrita
- Algoritmos determinístico e evolucionário intervalares para otimização robusta multi-objetivo