Development of a simplified generic PRA model for regulatory application

An abstract of a technical presentation presented at:
Workshop on Advanced Code Suites for Design, Safety Analysis and Operation of Heavy Water Reactors
Ottawa, Canada
October 2–5, 2012

Prepared by:
Amin Patel, Raducu Gheorghe, Yolande Akl
Probabilistic Safety Assessment and Reliability Division
Canadian Nuclear Safety Commission (CNSC)
Ottawa, Ontario, Canada

Abstract

The Canadian regulator, namely the Canadian Nuclear Safety Commission (CNSC), has the mandate to manage the risk in the public’s interest as defined in the Nuclear Safety Control Act (NSCA) and regulations. The regulatory fundamentals policy aims at basing the regulatory actions on levels of risk. The CNSC has developed the risk-informed decision making (RIDM) approach to be applied to the nuclear power reactor activities. In this approach, the probabilistic safety assessment (PSA) is recognized as a valuable instrument that bring insights that complement the traditional safety approach. Over the last few years, the two safety assessment methods, probabilistic and deterministic, have been used increasingly, as their strengths, limits, and complementary values are better understood. This article describes how the CNSC has started working on a simplified PSA of a CANDU 6 as a training vehicle for staff and to enable regulatory applications to be carried out.

This effort is the first among several milestones towards a generic regulatory PRA model, which may be used for internal applications, and a tool to help support the regulatory framework. The model is small enough for easy interrogation, the results can be quantified instantly and is able to replicate the key results from a given licensee without the need to resort to a high-performance computer. What is described here is the regulatory challenges of a given PSA review and the effort to achieving PSAs potential within the organization. The experience of developing a semi-independent proof-of-concept simplified risk model is hereby presented and the potential for further work is identified. The simplified model is to serve as an engine for certain Canadian regulatory applications, retaining the vital functionality of existing but excessively large and complicated licensee models, yet small enough to reproduce key results.