Estudo da viabilidade de uso de redes neurais artificiais para a predição do rendimento da fermentação alcoólica

• Ano de defesa: 2012
• Autor(a) principal: Amgarten, Ederson Luis
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Positivo; Brasil; Pós-Graduação; Programa de Pós-Graduação em Biotecnologia Industrial; UP
• Palavras-chave em Português: Fermentação alcoólica; Rendimento de fermentação; Modelo neuronal perceptron de multicamadas; CNPQ::CIENCIAS BIOLOGICAS::BIOLOGIA GERAL
• Link de acesso: https://repositorio.cruzeirodosul.edu.br/handle/123456789/2560

The use of ethanol as fuel is explored in Brazil as an additive to gasoline since 1931 when its was officially regulated, and since 1975, with the program PROÁLCOOL. Currently the importance of ethanol production is justified by its renewable character, for its environmental and economic advantages as a substitute for fossil fuels. Only in the last 30 years, the use of ethanol as fuel, the country saved more than one billion barrels of oil. The growing worldwide interest in producing renewable energy drives the development of new technologies to improve production processes reducing losses inherent in the process which will produce more without necessarily raising the cane growing areas, previously reserved for food production . Have been in modern computer systems available today a strong ally in the pursuit of perfection. The work presented here shows the use of artificial neural networks to construct a mathematical model that predicts the fermentation efficiency. The entries used the model were selected by the statistical method of linear correlation from a series of parameters monitored in plants of ethanol and sugar production. A theoretical analysis based on information provided by anothers authors was necessary to confirm the statistical results. Finally the chosen parameters were tested on three different combinations within a neural network multilayer perceptron (MLP Mult-Layer Perceptron) and the results presented and discussed in this paper.