Statistical methods for establishing safety-system margins. Final report

This report discusses the development of a statistical methodology which provides a sound, realistic basis for the evaluation of nuclear reactor safety system margins. Stochastic simulation (Monte Carlo), including advanced stratified sampling techniques, is applied to the calculation of Departure from Nucleate Boiling Ratio (DNBR) thermal margin. A detailed investigation of uncertainty components is presented. These include, algorithm modeling errors resulting from on-line rather than off-line design codes being used for computation, instrumentation uncertainties, and operating condition uncertainties. Measured on-site instrumentation variations are investigated to ascertain possible interrelationships among the process variables and to validate the overall DNBR uncertainty assessments. Advanced statistical techniques are developed to improve the Monte Carlo sampling efficiency and to facilitate the treatment of the multivariate aspects of the problem. A comparison of state-of-the-art sampling techniques is presented and the development of new superior techniques is detailed