Título: Identificação De Sistemas Dinâmicos Não Lineares Usando Back-Propagation Com Teacher Forcing
Autores: Paucar, V. Leonardo; Rider, Marcos J.; Morelato, André L.; Vuono, Evandro B.
Resumo: In this paper it is presented the algorithm and the numerical results for identification of non-linear dynamic systems using multilayer perceptrons artificial neural networks (ANN) trained with back-propagation with teacher forcing (BPTF). There have been analyzed several ANN configurations containing two neurons layers, one hidden and the other one in the output. The proposed artificial neural networks have been applied to double pendulum system and to the identification problem of the third-order model of the induction motor. Results obtained from applications suggest that artificial neural networks using BPTF for non-linear dynamic systems simulation and identification are very useful at least for the first steps after the training time period of the proposed neural networks.
Código DOI: 10.21528/CBRN2001-083
Artigo em pdf: 5cbrn_083.pdf
Arquivo BibTex: 5cbrn_083.bib