Forecasting the remaining useful life of filters in nuclear power plants

Research output: Contribution to journalArticlepeer-review

3 Downloads (Pure)


To function effectively, nuclear power plants rely on the effective filtration of air, water, and process fluids, examples of which include inlet sea water, reactor coolant, plant drinking water, and moderator purification. Filtration assets degrade over time, which impairs their filtering performance and reduces the flow rate. Being able to determine the remaining useful life (RUL) of a filter could result in benefits, particularly when moving from a time-based to a condition-based maintenance strategy that would optimize the filter replacement procedure and reduce early replacement of filters that are still fit for purpose. For many filter applications, a time-based strategy is sufficient. For strategically important assets, such as fueling machines, there are benefits to be gained from the development of predictive maintenance strategies. In this paper, we propose a predictive condition-based strategy using differential pressure data as a proxy for filter health. The key objective in this work was the creation of a model that could predict a filter asset RUL. The differential pressure for 7 to 14 days is predicted by a heuristic-based regression model of the history of each filter. This approach has been demonstrated using a civil nuclear generation application but could be applied to wider applications. While this model is still undergoing on-site evaluation, it has been estimated that there will be an operationally significant lifetime cost reduction.

Original languageEnglish
Number of pages11
JournalNuclear Technology
Early online date10 Jun 2024
Publication statusE-pub ahead of print - 10 Jun 2024


  • condition-based maintenance
  • predictive analytics
  • remaining useful life
  • nuclear power plants
  • filters


Dive into the research topics of 'Forecasting the remaining useful life of filters in nuclear power plants'. Together they form a unique fingerprint.

Cite this