Modelling the Release of Volatile Fission Product Cesium from CANDU Fuel Under Severe Accident Conditions Using Artificial Neural Networks

Abstract

An artificial neural network (ANN) model has been developed to predict the release of volatile fission products from CANDU he1 under severe accident conditions. The model was based on data for the release of 134~s measured during three annealing experiments (Hot Cell Experiments 1 and 2, or HCE- 1, HCE-2 and Metallurgical Cell Experiment 1, or MCE- 1) at Chalk River Laboratories. These experiments were comprised of a total of 30 separate tests. The ANN established a correlation among 14 separate input variables and predicted the cumulative fractional release for a set of 386 data points drawn from 29 tests to a normalized error, %, of 0.104 and an average absolute error, Eab, of 0.064. Predictions for a blind validation set (test HCE2-CM6) had an E, of 0.064 and an Eabs of 0.054. A methodology is presented for deploying the ANN model by providing the connection weights. Finally, the performance of an ANN model was compared to a he1 oxidation model developed by Lewis et al. and to the U.S. Nuclear Regulatory Commission's CORSOR-M.