Modelagem de sensor virtual para medição de vazão em uma usina do setor sucroenergetico baseado em redes neurais artificiais

• Ano de defesa: 2022
• Autor(a) principal: Lima, Jayne dos Santos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Engenharia Elétrica; Programa de Pós-Graduação em Engenharia Elétrica; UFPB
• Palavras-chave em Português: Usina 4.0; Setor sucroenergético; Sensores virtuais; Inteligência artificial; Redes neurais artificiais; Dropout; Avaliação de incerteza; Factory 4.0; Sugarcane setor; Soft sensors; Artificial inteligency; Artificial neural networks; Uncertainty assessment; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/25791

In the search for increased productivity, the industry has developed technological devices to achieve this goal; one of these strategies is called Industry 4.0. In the sugar-energy sector, industrial plants are looking for tools capable of improving processes and reducing the time of unscheduled stoppages combined with low maintenance costs. In this work, a virtual sensor (soft sensor) was developed to measure the flow rate of the inlet juice of a decanter (caleado broth) using the technique of artificial neural networks. The flow of lime juice is an essential variable in the process of manufacturing sugar and ethanol, as it directly influences the thermal balance of the plant, in addition to determining the number of inputs needed to guarantee the quality of the sugar. In this approach, data from a sugar-energy plant located in Camutanga, in the interior of Pernambuco, are used to create a knowledge bank for the system through the history of the supervisory system of the juice treatment of the unit. In this way, a virtual sensor model was built capable of measuring the flow of lime juice to be used as a possible redundancy in order to guarantee measurement efficiency in cases of failures and/or non-availability of the physical equipment. The results presented by the model from the tests performed in two different scenarios showed the robustness of the proposed model, and in all scenarios, the standard deviation was below 3%. In addition, after analyzing the uncertainty of the meter, it was found that the proposed model has a measurement error of 20 m³/h, which for the proposed application is a very acceptable value.