Título: Identificação de uma Unidade de Destilação Atmosférica e a Vácuo da Refinaria Gabriel Passos/Petrobrás Utilizando Redes Neurais e Lógica Nebulosa
Autores: Bomfim, Carlos Henrique Morais; Caminhas, Walmir Matos
Resumo: In an integrated petroleum company we have the need of taking important decisions concerning the production planning and program in order to maximize the profits. These tasks, production planning and program, demand mathematics models of the process plants. Using these models and with the support of mathematics programs is possible to choose the optimum production strategy or that one which fits better. In this work computer intelligence techniques (neural networks and fuzzy systems) were used to model one of the REGAP’s process unit. To achieve this purpose sample datas from the unit were used to get process models. These models were addressed to provide us with the information concerning the operation quality and efficiency. Models were got and these models estimate products yields with enough accuracy when validated against actual data. The approach to model this process combines neural network with fuzzy system (used to define process variable not easy to measure directly such us quality of feedstocks and operation targets). Some important intermediate products had these qualities modeled too. The results show that the proposed methodology is a suitable alternative to be considered in distillation process modeling. In this article we will show the methodology to identify inference models of product qualities.
Código DOI: 10.21528/CBRN2001-054
Artigo em pdf: 5cbrn_054.pdf
Arquivo BibTex: 5cbrn_054.bib