Modelos de aprendizado de máquina aplicados à manutenção preditiva de pequenas centrais geradoras hidrelétricas

• Ano de defesa: 2023
• Autor(a) principal: Rodrigo Barbosa de Santis
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ENG - DEPARTAMENTO DE ENGENHARIA PRODUÇÃO; Programa de Pós-Graduação em Engenharia de Produção; UFMG
• Palavras-chave em Português: Manutenção preditiva; Pequenas centrais hidrelétricas; Modelagem estatística e computacional; Prognóstico de falhas; Diagnóstico inteligente de falhas; Engenharia de produção; Usinas hidrelétricas; Modelagem; Falha de sistema (Engenharia)
• Link de acesso: http://hdl.handle.net/1843/55113

Maintenance in small hydroelectric plants is fundamental for guaranteeing the expansion of clean energy sources and supplying the energy estimated to be necessary for the coming decades. In the modern industrial context, predictive maintenance guides interventions and repairs based on the state of health of the machine, calculated from statistical and computational techniques. The current work has as main objective to propose a specific maintenance model for small hydroelectric plants. The thesis proposal is presented in the form of a collection of articles, the first being a systematic review on predictive maintenance in the hydroelectric sector, the second on the maintenance and operation profile of the plants and the formulation of the problem, and the third on the application of the method of extended isolation forest for anomaly detection for intelligent fault diagnosis. As a conclusion, we present two lines of action for work for the final thesis: the first in the area of intelligent diagnosis by type of failures and the second in relation to the prognosis of critical variables for the operation.