MACHINE LEARNING APLICADO À GESTÃO DE EQUIPAMENTOS EM UMA INDÚSTRIA CIMENTEIRA: ANÁLISE DE PARÂMETROS TRIBOLÓGICOS UTILIZANDO LINGUAGEM DE PROGRAMAÇÃO PYTHON

Maintenance and asset management in the cement industry present complex challenges due to the intensive wear on equipment such as the vertical roller mill, essential in cement production. This work addresses the application of Machine Learning techniques to predict wear and monitor faults in the mill to increase asset availability and efficiency. Two programs were developed: one for compliance and fault analysis and another for wear control, utilizing Artificial Neural Networks (ANN) and Long Short-Term Memory (LSTM) networks. The input variables for the models were obtained from equipment manuals and maintenance plans, and the data were processed and modeled in Python using specific libraries. The results showed that the LSTM model demonstrated superior predictive performance at various roller positions, with higher coefficients of determination (𝑅2) and lower mean squared error (MSE) compared to the ANN model. This system enables predictive monitoring, aligning with Predictive Maintenance concepts and contributing to the reduction of unscheduled downtime. It was possible to conclude that Machine Learning techniques hold significant potential for optimizing maintenance in the cement industry, particularly in monitoring of wear and faults, as indicated by the results obtained in the simulations.