2024-08-012024-08-012024-05-24SILVA, Jonatas Cruz da. Rede neural convolucional para o diagnóstico de rolamentos em baixa rotação. Orientador: Walter dos Santos Sousa; Coorientador: Thiago Barroso Costa. 2024. 58 f. Trabalho de Curso (Bacharelado em Engenharia Mecânica) – Faculdade de Engenharia Mecânica, Campus Universitário de Tucuruí, Universidade Federal do Pará, Tucuruí, 2024. Disponível em: https://bdm.ufpa.br/jspui/handle/prefix/7093. Acesso em:.https://bdm.ufpa.br/jspui/handle/prefix/7093This work focuses on the application of Convolutional Neural Networks (CNNs) in diagnosing faults in low-speed bearings, a crucial area for ensuring safety and efficiency, particularly in the mining industry. The use of vibration signals presents challenges due to low-speed bearings operating under considerable loads and occasionally under non-stationary conditions. Furthermore, because the energy associated with the bearing fault signal is weak, its monitoring is hindered by noise from other sources that can mask the fault signal. Therefore, signal processing tools and machine learning algorithms have been proposed to address these issues. Among machine learning techniques, deep learning-based diagnostic models have become popular in recent years, especially the Convolutional Neural Network model due to its performance. Thus, this work proposes the development of a diagnostic model in low-speed bearings, based on a convolutional neural network, where the input data is vibration signals that are subsequently transformed into a grayscale image, called a vibration image, and finally used for model training. For model training and validation, vibration signals from a bulk mixer operating at 50 rpm were used, with signals collected from healthy bearings and subsequently from defective bearings with varying degrees of severity and different machinery loading levels. After model training, accuracy is analyzed to evaluate the performance and precision of the proposed model. Therefore, this work aims to create an automated diagnostic tool for low-speed bearing failure with a high level of accuracy, thus facilitating the monitoring of machinery using these types of bearings, ensuring efficiency and safety in industrial sectors. As a result, two automated fault diagnosis models were obtained, one multiclass and the other binary, both achieving excellent results.Acesso AbertoRedes neurais convolucionaisDiagnóstico de falhaRolamento de baixa rotaçãoImagem de vibraçãoConvolutional neural networksFault diagnosisLow-speed bearingVibration imageCNPQ::ENGENHARIAS::ENGENHARIA MECANICA::PROJETOS DE MAQUINAS::ESTATICA E DINAMICA APLICADACNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAOTrabalho de Curso - Graduação - Monografia