Aplicação da rede neural convolucional para a extração de características e classificação de falhas no motor de indução trifásico

• Ano de defesa: 2020
• Autor(a) principal: Nogueira, Eduardo de Andrade
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Cornelio Procopio; Brasil; Programa de Pós-Graduação em Engenharia Elétrica; UTFPR
• Palavras-chave em Português: Motores elétricos de indução; Convoluções (Matemática); Redes neurais (Computação); Localização de falhas (Engenharia); Electric motors, Induction; Convolutions (Mathematics); Neural networks (Computer science); Fault location (Engineering); CNPQ::CIENCIAS EXATAS E DA TERRA; Engenharia Elétrica
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/5302

The three-phase induction motor is the main element of electromechanical energy conversion in the industry. Thus, due to wear and power quality problems, this engine may fail and thus lead to an unexpected stop in industrial processes. This work presents an alternative methodology for the extraction of fault characteristics from the voltage and current signals of the three phase phases of the three-phase induction motor using the convolutional neural network. The data used in this work come from experimental tests in the laboratory, where the faults were artificially introduced in the motor. These tests were performed with data with sample rates of 15ksps, 7.5ksps, 3ksps and 1.5ksps, using four standard classifiers, namely: Perceptron Multi-Layer, Naive Bayes, Support Vector Machine and K-Nearest Neighbors. The results obtained demonstrate that the strategy proposed proved to be viable for the classification of failures in the three-phase induction motor.