Statistical modelling of graphite brick weight loss in Advanced Gas Cooled Reactors

• Published in: Nuclear engineering and design Vol. 323; pp. 156 - 165
• Main Authors: Maul, P.R., Robinson, P.C., Burrow, J.F.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.11.2017; Elsevier BV
• Subjects: Advanced Gas Cooled Reactors; Bricks; Core loss; Cores; Gas cooled reactors; Graphite; Mathematical models; Neutrons; Nuclear energy; Nuclear reactors; Oxidation; Performance prediction; Reactor cores; Reactors; Statistical analysis; Statistical models; System variability; Weight loss; Advanced Gas Cooled Reactors; Oxidation; System variability; Graphite; Weight loss
• ISSN: 0029-5493; 1872-759X
• DOI: 10.1016/j.nucengdes.2017.08.017

•Graphite bricks form the core of UK Advanced Gas Cooled Nuclear Reactors.•Weight loss by oxidation may limit reactor life times.•Variability in weight loss is seen at different scales.•Statistical models are used to predict weight loss that will be seen at reactor inspections.•Good comparisons with observations enable long-term forecasts of system behaviour to be made. Demonstrating the physical integrity of graphite cores in UK Advanced Gas Cooled Nuclear Reactors operated by EDF Energy (formerly British Energy) is essential to the demonstration of their continued safe operation. The cores contain around 3000 graphite bricks which are subject to cracking and weight loss due to oxidation. If too much graphite is lost by oxidation the core will cease to act as an efficient moderator for neutrons. Periodic inspections are made of parts of the core to provide information on the state of a sample of the bricks. Statistical models are routinely used to help understand the evolution of brick cracking in the reactors. In this paper details are given of the use of mixed effects statistical models for weight loss to predict what will be seen at reactor inspections. By making blind predictions of core behaviour and then comparing these with observations from inspections, the predictive performance of the models has been quantified and they are then used to produce long-term forecasts of core behaviour. A key feature of the models is the need to represent system variability at different scales. Being able to forecast core behaviour enables reactor lifetimes to be estimated which has major economic implications.