Gêmeo digital de um sistema de bombeamento de água

• Ano de defesa: 2022
• Autor(a) principal: Nascimento, Caio Mangueira do
• 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: Indústria 4.0; Gêmeos digitais; Planta hidráulica; Redes neurais artificiais; Modelo digital; Sombra digital; Industry 4.0; Digital twins; Hydraulic plant; Artificial neural networks; Digital model; Digital shadow; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/25126

With the need to modernize water supply companies and the arrival of the Industry 4.0 concept, the search for better optimized solutions in the shortest possible time for internal problems has been constantly sought. One of the alternatives used for these situations is the use of Digital Twins (DT). Digital Twins are real-time virtual representations of objects, processes and systems. While Digital Twins can represent purely digital things, they often serve as a bridge between the physical and digital domains. The study related to Digital Twins is recent, in which suitable architectures are still being sought for their construction and the repertoire of applications related to water distribution systems is limited. The present research focused on the development of a Digital Twin of a water pumping system focused on maintaining the integrity of the hydraulic plant from the pressure data of the pipes. For the development of the Digital Twin, the research was divided into three parts: the construction of the Digital Model, the Digital Shadow, and finally the Digital Twin. The system considered as a scenario for the research consists of an experimental bench that is installed in the Laboratory of Energy and Hydraulic Efficiency in Sanitation (LENHS) of the Federal University of Paraiba, having in its constitution actuators and sensors, such as valves, water pump set and several pressure sensors. The development of digital models from the real system was carried out through the use of Artificial Neural Networks, in which two different activation functions were evaluated, the sigmoid and ReLU, and considering the application of feedback or not with output delays. From the developed digital models, it was observed that the greater the delay, the lower the mean square error (MSE) obtained, and the best digital models obtained were used in parallel with the hydraulic plant, as a Digital Shadow. Of the best models found, three use the ReLU activation function and one uses the sigmoid activation function. Using the digital models in parallel with the physical system, a satisfactory mean absolute percentage error (MAPE) was obtained from the use of the digital model with two delays. As Digital Shadow, experiments were carried out with updating the digital model considering two scenarios, one with new and old data or only with new data, in which an improvement in the results was observed in relation to the digital model without updating and the influence of the MSE obtained during the update on MAPE results during experiments with Digital Shadow. For the Digital Twin, a supervisory was developed in ScadaBR, which is a free software, freeware and open source software for the development of Automation, Data Acquisition and Supervisory Control applications, which allows the management of the DT and the physical plant, not allowing entries outside the permissible range or operations that increase pressure in the pipelines. The obtained results show the use of the Digital Twin as being of great value in relation to the protection of the physical system against harmful operations as well as the possibility of performing simulated tests without communication with the physical plant. Keywords: Industry 4.0; Digital Twins; Hydraulic Plant; Artificial neural networks; Digital Model; Digital Shadow.