Artificial neural network models for volatile fission product release during severe accident conditions

Artificial neural network (ANN) models have been developed to predict the r elease of volatile fission products from both Canada deuterium uranium (CAN DU) and light water reactor (LWR) fuel under severe accident conditions. Th e CANDU model was based on data for the release of Cs-134 measured during t hree annealing experiments (Hot Cell Experiments 1 and 2, or HCE-1, HCE-2 a nd metallurgical cell experiment 1, or MCE-1) at Chalk River Laboratories. These experiments were comprised of a total of 30 separate tests. The ANN e stablished a correlation among 14 separate input variables and predicted th e cumulative fractional release for a set of 386 data points drawn from 29 tests to a normalized error, E-n, of 0.104 and an average absolute error, E -abs, Of 0.064. Predictions for a blind validation set (test HCE2-CM6) had an E-n of 0.064 and an E-abs Of 0.054. From this 14 variable ANN model, a p runed version utilizing only the 6 most significant variables was trained t o provide comparable predictions. An ANN model was also developed for LWR f uel, based on data from the vertical induction CVI) series of tests (VI-2 t o VI-5) conducted at Oak Ridge National Laboratory. Predictions for data no t used in ANN training had an E-n of 0.045 and an E-abs of 0.059. A methodo logy is presented for deploying the ANN models by providing the algorithms for trained ANNs and the corresponding connection weights. Finally, the per formance of the full ANN CANDU model was compared to a fuel oxidation model developed by Lewis et al, and to the US Nuclear Regulatory Commission's CO RSOR-M. (C) 1999 Elsevier Science B.V. All rights reserved.