Desenvolvimento de um soft sensor para estimação da vazão em sistemas de abastecimento de água utilizando redes neurais artificiais

• Ano de defesa: 2022
• Autor(a) principal: Lima, Robson Pacífico Guimarães
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Engenharia Mecânica; Programa de Pós-Graduação em Engenharia Mecânica; UFPB
• Palavras-chave em Português: Medição indireta; Soft sensor; Sistemas de abastecimento de água; Redes neurais artificiais; Indirect measurement; Water supply systems; Artificial neural networks; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/24485

An efficient water supply system is essential to guarantee access to such an essential resource for life, whether in the most diverse consumer segments: residential, commercial and industrial. A typical water supply system is composed of several physical elements (sensors, transducers, valves, pumps, etc.) and some of these elements that make up this scenario have high costs and/or complex installation. In this way, the technique of indirect measurement of a certain quantity from another already known can be used as an alternative in order to reduce costs with equipment that would promote that desired measurement. This technique would dispense with the use of a specific sensor directly in the plant when this device has a high acquisition value or its installation is complex from an operational point of view. Thus, this work aims to implement a virtual instrument (called a soft sensor) using artificial neural networks capable of estimating the flow in a plant that emulates a water supply system. The plant used for the validation of the soft sensor is installed in the Laboratory of Energy and Hydraulic Efficiency in Sanitation of the Federal University of Paraíba (LENHS/UFPB). Computer simulations and several laboratory tests were performed promoting the comparison between the values measured through the soft sensor versus values measured through a real electromagnetic flow sensor and the highest value referring to the maximum error obtained was 4,7 l/s in one of several tests performed with Mean Percentage Absolute Error of up to 0.0023% in another test. Furthermore, other robust parameters were obtained and the results obtained corroborated the efficiency of the technique presented.