The Highly Active Liquid Effluent and Storage plant at Sellafield, UK, currently uses three evaporators to reduce the volume of active liquor stored within the facility before being vitrified for long term storage. This liquor is highly corrosive and the lifetime of the evaporators is potentially limited by the corrosion loss from the heating elements, comprising an external jacket and a number of internal coils, all heated by low pressure steam. Inspection of the heating coils inside the evaporators is possible and measurement data is available of their thicknesses by depth at various inspection intervals. This inspection data has been combined with operational data and thermal models for the heating elements. Our theoretical understanding from laboratory measurements suggests that corrosion is related to temperature through an Arrhenius relationship. As such we have been able to develop a predictive model for the thickness profiles and remaining useful life of the uninspected components. This model is a non-linear mixed effects (multilevel) model and has undergone significant developmental work to account for a number of practical data issues. This paper will briefly outline the various components of the model, whilst discussing issues relevant to any statistical model such as complexities of data collection, approaches to handling correlated data, selecting appropriate model formulations and data transformations. The inclusion of uncertainties in prediction via Monte-Carlo simulation will also be discussed.

This content is only available via PDF.
You do not currently have access to this content.