Statistical modelling of graphite brick cracking in advanced gas-cooled reactors

Abstract

Graphite cores in advanced gas-cooled nuclear reactors contain bricks which are subject to cracking and periodic inspections provide information on the state of a sample of the bricks. Statistical models are fitted to the historic data and used to predict the numbers of cracked bricks that are expected to be seen at reactor inspections. As more data have become available the complexity of the best performing models has increased. By repeatedly making blind predictions of core behaviour the predictive performance of the models can be quantified and they can then be used to forecast behaviour of the core over longer periods. © 2011 Royal Statistical Society.