Probabilistic safety assessment of high-level radioactive waste repository with the mechanistically modeled failure rate

Abstract

For the treatment of uncertainties arising in the safety assessment of high-level radioactive waste repository, a new probabilistic safety assessment methodology is developed. The present methodology contains two major parts: one is for the treatment of uncertainties involved in the mechanistic models describing the performance of radioactive waste repository and the other is for the treatment of input parameter uncertainties. The present methodology is based on the system reliability analysis method, and the radioactive waste repository is considered as a stand-by redundant system is continuous operation. The risk of radioactive waste repository is presented as a product of the failure probability density of waste repository and the radionuclide inventory in the repository when the failure of waste repository occurs. The failure probability density of waste repository is calculated by combining those of unit components that are estimated using the mechanistic models describing their performances. This probabilistic methodology can be a valuable alternative to the deterministic safety assessment for the high-level radioactive waste repository and provides the quantitative informations on the contribution of unit components to the overall performance of waste repository. The present methodology is applied to the reference repository for the numerical illustration. The results show that the present methodology yields more realistic results as compared with those using the random failure model, and among the unit components of waste repository, the geosphere greatly contributes to the overall performance of waste repository for sorbing radionuclides, and as the retardation factor of nuclide is decreased, the overall performance is dominated by the waste matrix. To evaluate how the safety assessment results are affected by the uncertainties involved in input parameters, the Monte Carlo simulation with Latin Hypercube Sampling and a statistical estimation are use...