Equipment failure prediction

CONTEXT
In Small Hydroelectric Power Plants (SHPs), the mechanical equipment called speed multiplier was the component with the highest rate of failures that, in some cases, generated operational unavailability of the generation units.

 

SOLUTION
Data were collected from different sources, such as operational, field sensors, shutdown and maintenance events. Feature engineering techniques were applied to these data, which, combined with unsupervised learning techniques, helped to identify anomalous patterns and failure patterns, allowing for early prediction of equipment failure events.


RESULTS
Predictability of up to 60 days before failures occur, allowing the maintenance area to take preventive and mitigating actions, contributing to increased availability.