PREDICTING ROLLING-ELEMENT BEARING FAILURES USING DIGITAL TWIN AND MACHINE LEARNING
Digital Twin (DT). Machine Learning. K-means. ANN.
The present work aims to solve the challenge of premature bearing failure in a reducer of a bar rolling mill at Gerdau steelworks – Usina Barão de Cocais. The proposed approach is based on machine learning techniques, particularly clustering using the K-means algorithm, and the application of artificial neural networks (ANN) for proactive prediction and monitoring. The study highlights the importance of predicting failures in critical industrial equipment to ensure operational efficiency and prevent unforeseen shutdowns. In addition, the concept of Digital Twin (DT) is introduced as a fundamental tool in the process, enabling real-time simulation and monitoring of the physical system, which contributes to anticipating potential issues. The applied methodology included the execution of various ANN tests, exploring different activation functions, the number of neurons in the hidden layer, loss metrics, and optimizers. The analysis of the confusion matrices will play a fundamental role in this process, providing valuable insights into the model's performance and indicating possible improvements or necessary adjustments.