VROC's power generation client wished to investigate the ability for AI predictive analytics to assist with their asset health monitoring and predictive maintenance on critical assets, including their boiler pumps.
VROC set up AI models to monitor the pump health by ingesting all the data for the critical sensor tags corresponding to pump and motor bearing temperatures, as well as pump flowrate. The AI models would show any deviation from predicted values.
VROC's models detected a trend of deviation of actual temperatures increasing against predicted AI temperatures on both DE and NDE bearing temperatures. This trend existed for almost 2 months. In addition VROC's platform detected a deviation of the pump suction flowrate, which was higher than predicted, and meant that pump flowrate was reducing. A sudden DE temperature spike occurred which caused a trip.
Due to the insight provided by the model, the client was able to ensure the availability of a back-up pump to mitigate down time.
The maintenance team revealed that wear signs on a balance disk and seat caused a change of pump shaft axial displacement, causing a decrease in the flowrate with increased temperatures and subsequent failure
The estimated savings due to the prediction of the failure and the deployment of a backup
pump has been estimated at 150,000 USD.
VROC'S LEAD DATA SCIENTIST EXPLAINS THIS CASE STUDY IN FURTHER DETAIL - WATCH BELOW