Metodologia de aquisição de dados para captura de sinais de vibração mecânica de máquinas

• Ano de defesa: 2025
• Autor(a) principal: Duarte, Ana Karine Gomes
• Tipo de documento: Dissertação
• 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: Vibrações mecânicas; Processamento de sinais; Frequências; Defeitos em máquinas; Manutenção preditiva; Mechanical vibrations; Signal processing; Frequencies; Machine defects; Predictive maintenance; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/36931

The growing interest in the application of predictive maintenance to mechanical machines has encouraged the academic community to explore techniques for acquiring and processing signals from sensors installed in this equipment. The signals captured by the sensors are essential for assessing the operating condition of the machine and determining the ideal time for a planned shutdown, allowing less invasive maintenance interventions. However, for this assessment to be reliable, it is essential that the signal acquisition accurately represents the physical phenomena to be monitored. Among the various indicators of potential failures, mechanical vibration signals stand out, as they carry frequency signatures characteristic of defects that may be developing in the machine components. Given the importance of correctly capturing these signals, this work proposes the development of an interactive data acquisition methodology, capable of dynamically adjusting itself to ensure maximum fidelity in the collection of vibration transducer signals. The methodology is based on the sampling theorem, making automatic adjustments to the sampling rate and/or the number of points acquired, according to the measurement situation. The strategy is to seek the greatest possible range of frequencies associated with failures, even in the early stages of severity, allowing more accurate monitoring of the evolution of the defect over multiple measurements. To validate the proposal, this work presents experimental results obtained by capturing signals from an electrical signal generator, simulating real machine failures.