Performance Enhancement Architecture for Radiation Dose Computation in Integrated Decommission Process Assessment System

Abstract

This paper proposes a computing architecture suitable for an integrated decommission process assessment system under development by Korea Atomic Energy Research Institute, which provides innovative simulation tools for the dismantling process of main components of nuclear power plants, including features of planning and simulation of dismantling processes to assess the task feasibility, cost estimation of processes, radiation dose estimation of workers. Among those, it is most safety-critical and time consuming to conduct radiation dose estimation of chemically contaminated and highly radioactive components during nuclear facility decommissioning, for usually it uses Monte Carlo simulation methods in order to get accurate results scarifying computing costs. The estimation time further increases as we need to iterate such time-consuming simulation runs for a lot of dismantling scenarios to get the total exposure of moving workers and even changing radioactive sources. Thus, to cope with this situation, we propose a hybrid cluster computing architecture exploiting commodity hardware for cost efficiency and maintainability. We also propose software architecture to effectively enhance the performance of the assessment in both real-time mode and batch mode with fault-tolerance. We implemented a pilot 8-node cluster as proposed and with experiments for a few cases we illustrated that the proposed architecture has the extensibility for the integrated decommission process assessment system.