Language selection

Search


Statistical Analysis of Common-Cause Failure Events Using ICDE Data

An abstract of the technical paper presented at:
Probabilistic Safety Assessment & Management (PSAM) Conference
Honolulu, Hawaii, U.S.
June 22–27, 2014

Prepared by:
Mahesh Pandey and Shuo Yu
University of Waterloo

Smain Yalaoui and Yolande Akl
Canadian Nuclear Safety Commission

Abstract

Common-cause failure (CCF) events are a subset of dependent events in which two or more components fail within a short interval of time, as a result of a shared (or common) cause. Common-cause events are highly relevant to probabilistic safety assessments (PSAs) because of their potential adverse impact on the safety and availability of critical safety systems at a nuclear plant. An accurate estimation of CCF rates is therefore important for a realistic PSA of a nuclear power plant.

Because of lack of data, CCF rates/probabilities were estimated by expert judgment in early years of PSA. Over the years as more data was collected by utilities and regulators worldwide, more formal statistical analysis methods for data analysis and estimation of CCF rates emerged. Therefore, the estimates of CCF rates consistent with operating experience should be used in PSAs, in place of generic or expert judgment estimates. The International Common-Cause Failure Data Exchange (ICDE) is a concerted effort undertaken by many countries to compile CCF event data in a consistent manner. CCF event data are compiled by a third party (OECD/NEA) in a database, with a copy maintained by the Canadian Nuclear Safety Commission.

This paper presents a review the current status of CCF modeling techniques and international best practices, and explains the basic principles of underlying probability theory.

The Empirical Bayes method and data mapping are used in the statistical analysis. A detailed case study based on CCF data for motor operated valves is presented.

To obtain a copy of the abstract’s document, please contact us at cnsc.info.ccsn@cnsc-ccsn.gc.ca or call 613-995-5894 or 1-800-668-5284 (in Canada).  When contacting us, please provide the title and date of the abstract.

Page details

Date modified: