Desenvolvimento de um sistema inteligente de monitoramento prescritivo para severidade das condições de funcionamento de um redutor do tipo coroa sem-fim

• Ano de defesa: 2023
• Autor(a) principal: Oliveira Neto, João Manoel de
• 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: Engenharia Mecânica; Severidade de vibrações; Transformada de wavelet; Estatística; Rede neural artificial; Análise sonora; Redutor coroa sem-fim; Mechanical Engineering; Vibration severity; Wavelet transform; Statistic; Artificial neural network; Sound analysis; Worm gear reducer; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/30202

Industrial machines, in general, are characterized by the operation coming from an electric motor associated with a speed reduction system or mechanical power transmission, such as through gears. Several other elements make up the operating system of a machine, and due to the imposed cyclic load and the poor conditions of use to which this set is imposed, preventive monitoring is carried out or predictive maintenance techniques are adopted in order to predict the appearance of failures. This research aims to develop an intelligent system, through the collection of data via analysis of sound signals, to carry out the prescriptive diagnosis on the severity related to bad operating conditions in a rotating system, whose transmission system is given by gears of the worm type crown, where the operating severity was classified as “light”, “medium” and “severe”. The sound signals were collected with a microphone and at the same time the vibration analysis was carried out in order to validate the obtained results. The extraction of the characteristics of the signals was carried out by multi-resolution wavelet analysis, using the information contained in the coefficient of detail 4, as well as statistical tools, these being standard deviation, variance and kurtosis coefficient. Once the operating patterns were identified, the architecture of a perceptron-type multilayer artificial neural network was elaborated, with a backpropagation algorithm for classifying these signals. As a result, an ANN with a general efficiency of 99.7% was obtained. It was concluded that the development of the prescriptive intelligent system was able to detect the severity resulting from poor operating conditions inserted in the prototype in the laboratory and in industrial equipment, and can serve as an auxiliary tool in maintenance routines.