Abstract
Characteristics of safety assessment for geological disposal of radioactive waste are as follows.
・Considering long-term safety that far exceeds our experience and history
・Treating spatial heterogeneity of natural strata.
International experiences accumulated until recently have significantly improved our knowledge and confidence. However, our knowledge is not necessarily complete in the present, prediction of phenomena in disposal system inheres uncertainties associated with scenarios, models, and parameters. From the view point of safety assessment, it is important that the disposal system has enough safety, even though considering above uncertainties. Therefore it is necessary to identify the realistic scenario and the scenarios including uncertainties that should be considered. On the other hand, for rational design, it is essential to identify the range of parameters that conducts favorable scenario and factors that have influence on bifurcation of scenarios.
This paper presents a hybrid simulation system that consists of cellular automata, large-scale numerical analysis, and neural networks to simulate evolution of the near field after closure of the repository efficiently. This system was applied to a test case. As a result, several qualitatively different scenarios were identified by cellular automata, and for a number of representative cases were solved by large-scale numerical analysis using PC clusters. In addition, a comprehensive parametric study using neural networks, which “learned” mappings obtained by accurate numerical simulations. Therefore, it is possible that near-field performance is confirmed efficiently, and information about bifurcation of scenarios is presented. This hybrid simulation system could be applied to define the specifications for engineered barrier system rationally and robustiously.
