摘要：AbstractThe industrial maintenance activities represent an increase in production costs, mainly caused by unnecessary production stops. Recent technologies approaches are handling the continuous monitoring of industrial machines, storing sensors data, and also maintenance history. More data analysis is necessary specifically for rotating machines presenting methodologies to reduce the maintenance. In order to handle this problem, a comparative analysis of machine learning methods is presented. The strategy aims to predict failures and then indicates the maintenance necessity before a break occurs. Thus, it is applied and analyzed the specific machine learning algorithms, Gradient Boosting and Random Forest, using a dataset of rotation machines. The results show that both methods have an excellent performance (metrics accuracy, precision, and recall), with slightly better results in Gradient Boosting (hit rate of 99.93%) indicating the prominent application of these algorithms in this industrial scenario.
