This work presents cases of monitored generation and transmission equipment at Itaipu Binacional and a methodology using machine learning to estimate equipment health and support condition-based and predictive maintenance. Data from SCADA/EMS and CMMS are integrated through the AVEVA PI System and fused to calculate real status of equipments. Models such as Recurrent Neural Network, Long Short-Term Memory, and Random Forest are trained through a internal developed analysis tool, and used to classify conditions and predict failures based on a estimated health index. Results are visualized in PI Vision, improving and supporting maintenance decisions.