Quantifying organizational factors in human reliability analysis using the big data-theoretic algorithm

Abstract

Probabilistic Risk Assessment (PRA) is an effective tool for estimating riskfrom interactions of equipment failure and human error. Human Reliability Analysis (HRA) focuses on individual error due to internal (e.g., cognitive mode) and external (e.g., physical environmental factors, organizational factors) Performance Shaping Factors (PSFs). Current HL4 techniques include some of the external PSFs related to organizational factors, such as procedures and training, however, the organizational mechanisms associated with these PSFs are not explicitly modeled in NRA. The incorporation of organizational models in this research is done based on the Socio-Technical Risk Analysis (SoTeRiA) theoretical framework. The Integrated PRA (1-PRA) methodology is introduced to connect the combined effects of human error and organizational factors with classical PRA techniques (i.e., Event Trees and Fault Trees). A novel algorithm, the big-data theoretic, is utilized to address wide-ranging, incomplete, and unstructured data. The new algorithm is applied for quanting the organizational mechanisms associated with the "training quality" (in nuclear power plants) that can be used as a surrogate node for "training" PSF, inside HRA. This research helps develop a more realistic and plant-specific estimation of human error. In addition, it facilitates explicit modeling for the sources of dependencies among PSFs.