RSICC CODE PACKAGE PSR-449

1. NAME AND TITLE

BFR: Code System for Common Cause Failure Data Analysis.

2. CONTRIBUTORS

EG and G Idaho Inc., Idaho Falls, Idaho, through the Energy Science and Technology Software Center, Oak Ridge, Tennessee.

3. CODING LANGUAGE AND COMPUTER

FORTRAN IV (95%) and COMPASS (5%); CDC CYBER176 (P00449C017600).

4. NATURE OF PROBLEM SOLVED

BFR calculates both point and interval estimates of common cause failure rates. The first portion of the data analysis deals with plant-to-plant variation in the rates. The second portion of the data analysis consists of Bayesian estimation of all the quantities of interest and the use of residuals to check the binomial assumption.

5. METHOD OF SOLUTION

BFR uses the binomial failure rate model to estimate the rate of simultaneous failures of similar components. Data consists of recorded single and multiple failures. Either Bayesian or non-Bayesian methods can be employed to find point estimates and interval estimates of the quantities of interest, including the rate at which a specific set of k components fails simultaneously, and the rate at which at least k out of m components fail simultaneously. These rates can be used directly in fault trees, or for calculating probabilities. An auxiliary program, PTEST, used to check the goodness of fit between the model and data following BFR execution, is included.

6. RESTRICTIONS OR LIMITATIONS

None noted.

7. TYPICAL RUNNING TIME

No data is available on running times.

8. COMPUTER HARDWARE REQUIREMENTS

BFR was developed on CDC Cyber 176 computers and was not tested at RSICC.

9. COMPUTER SOFTWARE REQUIREMENTS

BFR was written in Fortran IV and COMPASS and was developed under NOS/BE 1.4. BFR makes use of subroutines from the proprietary IMSL library; plots, used only for Bayesian regions when the causes of single failures are not identifiable, are produced by the proprietary DISSPLA graphics package. These routines are not included. Typical run time is 80 CP seconds; however, the time varies greatly, depending on the complexity of the data, and about twice as much time is required if the compiler OPT=0, rather than OPT=2 option, is used.

10. REFERENCES

a) Included in documentation:

C. L. Atwood and W. J. Suitt, User's Guide to BFR, NUREG/CR-2729, EGG-EA-5502 (February 1983).

C. L. Atwood, "Data Analysis Using the Binomial Failure Rate Common Cause Model," NUREG/CR-3437, EGG-2271 (September 1983).

b) Background information:

C. L. Atwood, "Estimators for the Binomial Failure Rate Common Cause Model," NUREG/CR-1401, EGG-EA-5112 (April 1980).

11. CONTENTS OF CODE PACKAGE

Included are the referenced documents in (10.a) and one 3.5" diskette which contains a self-extracting compressed DOS file with the BFR and PTEST source codes and test case input and output.

12. DATE OF ABSTRACT

December 1999.

KEYWORD: PROBABILITY SAFETY ANALYSIS