A Model-Based Human Reliability Analysis Framework Ali Mosleh a , John A. Forester b , Ronald L. Boring c , Stacey M. L. Hendrickson c , April M. Whaley c , Song-Hua Shen d , , Dana L. Kelly c , James Y.H. Chang d , Vinh N. Dang e Johanna H. Oxstrand f , Erasmia L. Lois d a University of Maryland, College Park, MD, USA b Sandia National Laboratories, Albuquerque, NM, USA c Idaho National Laboratory, Idaho Falls, ID, USA d US Nuclear Regulatory Commission, Washington, DC, USA e Vinh N. Dang, Paul Scherrer Institute, Villigen PSI, Switzerland f Vattenfall Ringhals AB, Väröbacka, Sweden __________________________________________________________________________________ Abstract: In response to a Staff Requirements Memorandum (SRM) to the Advisory Committee on Reactor Safeguards (ACRS), the US Nuclear Regulatory Commission (NRC) has undertaken a research effort to create a consensus approach to human reliability analysis (HRA). This paper provides an overview of the approach being developed. The approach introduces the “crew response tree” (CRT) concept, which depicts the human failure events in a manner parallel to the PRA event tree process, provides a structure for capturing the “context” associated with the human failure events under analysis, and uses the Information Processing Model as a platform to identify potential failures. It incorporates behavioral science knowledge by providing the decompositions of human failures/failure mechanisms/failure factors built from a top-down and bottom-up approach, the latter reflecting those findings from scientific papers that document theories and data of interest. The structure provides a roadmap for incorporating the phenomena with which crews would be dealing, the plant characteristics (e.g., design, indications, procedures, training), and human performance capabilities (awareness, decision, action). In terms of quantification, the approach uses the typical PRA conditional probability expression, which is delineated to a level adequate for associating the probability of a human failure event with conditional probabilities of the associated contexts, failure mechanisms, and the underlying factors (e.g., performance shaping factors). Such mathematical formulation can be used to directly estimate HEPs using various data sources (e.g., expert estimations, anchor values, simulator or historical data), or can be modified to interface with existing quantification approaches. 1 Keywords: HRA, Failure Mechanism, Cognitive Model, IDA. __________________________________________________________________________________ 1. INTRODUCTION In a Staff Requirements Memorandum (SRM) (SRM-M061020) to the Advisory Committee on Reactor Safeguards (ACRS), the Commission directed the ACRS to “work with the staff and external stakeholders to evaluate the different human reliability models in an effort to propose a single model for the agency to use or guidance on which model(s) should be used in specific circumstances.”[1] As a first step toward meeting this directive, a survey of various methods and user needs was performed, followed by a workshop of experts in human reliability analysis (HRA) and related domains to discuss the outline of an approach to address the SRM concerns. The workshop in particular identified a set of desirable attributes of a robust HRA method, which could also be used as a set of criteria in evaluating various methods. The attributes are: 1 The information presented in this paper does not currently represent an agreed-upon NRC staff position. The NRC has neither approved nor disapproved its technical content.