Quantification of Dynamic Human Error Probabilities in Terms of Sampled Timing Data

Abstract

A quantified correlation process to assess those dynamic human error probabilities for which timing data can be sampled from generic occurrences and/or from plant-specific simulator training records is presented. The method is based on a comparison between two random variables, representing the critical time for system response and the action time for operator achievement capability. The maximum entropy estimator is employed to handle the sampled timing data. The operator action in performing rapid cooldown and depressurization following an initiating event of steam generator tube rupture is used to delineate this method. The resulting datum has been successfully applied in a risk analysis project.