Safe Cracking: Monte Carlo Nonlinear Coupled Analysis of Nuclear Reactor Bricks
Advanced Gas Cooled Reactors (AGRs), owned and operated by EDF Energy, generate around 15% of electrical power in the UK. As AGRs approach their end of life, cracking is liable to occur in the graphite bricks which make up a major part of the reactor core. Safety cases to justify continued operation of the reactors generally include certain limits on the number, type and distribution of cracked bricks. It is therefore vital to be able to predict the rate of cracking to support the continued safe operation of AGRs.
Quintessa has developed a COMSOL Multiphysics® model of individual bricks to provide a route to predict cracking rates for EDF Energy. Using the Solid Mechanics interface coupled with ODE interfaces to model highly specialized thermo-mechanical processes present in graphite under an intense radiation field, a model of brick shape and stress is used to simulate bricks throughout the AGR lifespan. Statistical models for properties such as dynamic Young’s modulus and flexural strength are incorporated into the model, based on data from reactors. Monte Carlo simulation is performed to account for variability in the bricks by using a Batch Sweep, and data are exported to facilitate further analysis.
Parameter distributions for brick properties are initially defined to represent realistic ranges based on AGR measurements, without global model calibration. We will present a method which uses a response surface fitted to COMSOL Multiphysics® outputs in order to optimize the fit of brick shapes and cracking rates against observations, and hence calibrate the parameter distributions used in the model without having to run many thousands of model realizations. The method provides a straightforward and efficient means of comparing input data with the required calibration, building transparency and confidence in the modelling process. It also provides a means to estimate confidence bounds for comparing model predictions with measurement data using limited samples, and to estimate the probability of a brick cracking based on its shape and the core burnup.
