Lessons learned from dependency usage in HERA: Implications for THERP-related HRA methods

Abstract

Abstract-Dependency occurs when the probability of success or failure on one action changes the probability of success or failure on a subsequent action. Dependency may serve as a modifier on the human error probabilities (HEPs) for successive actions in human reliability analysis (HRA) models. Discretion should be employed when determining whether or not a dependency calculation is warranted: dependency should not be assigned without strongly grounded reasons. Human reliability analysts may sometimes assign dependency in cases where it is unwarranted. This inappropriate assignment is attributed to a lack of clear guidance to encompass the range of scenarios human reliability analysts are addressing. Inappropriate assignment of dependency produces inappropriately elevated HEP values. Lessons learned about dependency usage in the Human Event Repository and Analysis (HERA) system may provide clarification and guidance for analysts using THERP-based dependency models. This paper presents the HERA approach to dependency assessment and discusses considerations for dependency usage in HRA, including the cognitive basis for dependency, direction for determining when dependency should be assessed, considerations for determining the dependency level, temporal issues to consider when assessing dependency, (e.g., considering task sequence versus overall event sequence, and dependency over long periods of time), and diagnosis and action influences on dependency.

Using Expert Models in Human Reliability Analysis-A Dependence Assessment Method Based on Fuzzy Logic

“…One of these is an attempt to analyze operating experience and to obtain in this way empirical relationships between the factors and the observed human failure events: the human error repository and analysis project and database sponsored by the U.S. Nuclear Regulatory Commission. ( 39,40 ) A second effort is the International HRA Empirical Study being performed by an international group of organizations jointly with the Organisation for Economic Co‐operation and Development (OECD) Halden Reactor Project, in which the predictions of HRA methods are being compared with simulator data. ( 41,42 ) While both quantitative predictions (the HEPs) and qualitative predictions (the “driving” or most important input factors identified in the HRA) will be addressed in this work, the number of data points will not be sufficient to validate comprehensively the relationship between the input variables and the predicted failure probabilities, a relationship represented by each HRA method.…”