Avaliação da Predição de Algoritmos de Treinamento Supervisionado de Redes Neurais Artificiais Aplicado a Qualidade de Biodiesel

• Ano de defesa: 2015
• Autor(a) principal: Sousa, Raquel Machado de
• Orientador(a): LABIDI, Sofiane
• Banca de defesa: Abdelouahab, Zair
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA DE ELETRICIDADE/CCET
• Departamento: Engenharia
• País: BR
• Palavras-chave em Português: biodiesel; índice de iodo; predição; redes neurais artificiais; viscosidade
• Palavras-chave em Inglês: artificial neural networks; biodiesel; iodine value; prediction; viscosity
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS::ENGENHARIA BIOMEDICA::ENGENHARIA MEDICA::BIOMATERIAIS E MATERIAIS BIOCOMPATIVEIS
• Link de acesso: http://tedebc.ufma.br:8080/jspui/handle/tede/292

To ensure efficient combustion and emissions quality, as well as safety in the transport and handling of biodiesel, the National Agency of Petroleum, Natural Gas and Biofuels (ANP) establishing, through Resolution No. 14 of 2012, quality standards and specifications for this biofuel, and for that many official or alternative methods may be used. In literature, it is possible to identify an increasing use of linear methods and non - linear in the recognition and classification standards applied to the monitoring of biodiesel quality. In this context, the Artificial Neural Networks (ANN) have shown to be quite viable, as a tool non - linear, in predicting biofuel properties. The present work proposes to assess the prediction of biodiesel quality properties using supervised training algorithms of ANNs. In order to contribute to a study to provide a network structure with a training algorithm that can perform better with good results in the prediction. Through the prediction of the properties of the biodiesel from the composition of the esters of the raw material, it is possible to assess the feasibility of using such raw materials for the synthesis of a quality biodiesel. In this work we obtained a better ANN architecture for iodine value prediction and viscosity. The results of the simulations showed that the ANNs are a technology that can be used to predict these properties, like other related composition of fatty acid esters.