Lifetime modeling of equipment at nuclear power plants

One of the on-going projects at the OECD Halden Reactor Project (HRP) focuses on condition based maintenance at nuclear power plants and development of lifetime models for plant components and equipment. A methodology is being developed for systematic and cost-effective development of lifetime models. A key factor in this modelling is to find a suitable combination of data-driven and first principles based elements, so that the relative advantages of the approaches are balanced with cost to serve the goals of each application. This hybrid approach is especially beneficial when the phenomena to be studied involve multivariate non-linear effects, e.g., multiple simultaneously acting degrading factors amplifying each other's impacts. Hybrid modelling also alleviates sensitivity to the validity of assumptions made in modelling work, especially the assumption of the available data set being representative - an assumption that is difficult to avoid in purely data- driven modelling. This approach has been applied to several real world case studies. In these case studies the remaining lifetime was estimated for clogged water filters for heat exchangers, valve erosion and air filters. The methodology has been refined and further developed based on the lessons learnt from each case study. For example, in the air filter case, estimation accuracy was improved by modelling phenomena with different time scales (gradual accumulation of particles, sporadic aerosol emissions, and seasonal variability) separately and taking them into account in the remaining useful life (RUL) estimation.