This research presents a predictive maintenance framework for assessing the health state of centrifugal pumps, utilizing machine learning algorithms to predict failures using collected data. A dataset of 5,118 records was analyzed to enhance safety and productivity in industrial applications, with findings indicating the effectiveness of the XGBoost algorithm for health state prediction. The study demonstrates the importance of feature selection and the use of statistical analysis to optimize predictive maintenance models, aiming to reduce operational costs and downtime.