Auto-CREAM: Software application for evaluation of HEP with basic and extended CREAM for PSA studies

摘要

Human Reliability Analysis (HRA) enables the estimation of the contribution of erroneous actions from the control room and field operators towards the overall risk, as Human Error Probability (HEP). A challenge usually encountered in HRA is inaccuracy in manual calculation of HEP and variance in the evaluated HEP results for the same case study by different HRA practitioners; owing to unavailability of professional software tools for HRA, unlike plenty of such tools available for Probabilistic Safety Assessment (PSA). CREAM, a widely used second generation HRA method quantifies the conditional failure probability of operator's cognition with respect to the context. Since it is a detailed method based on the concepts of cognition, manual evaluation of HEP with CREAM still remains a cumbersome and error prone process. The objective of this article is to address these challenges by automating and validating the CREAM predictive analysis as a software application. In developing the application, the research contribution of this paper is implementation of uncertainty and sensitivity analysis of the probability of Human Failure Event (HFE) i.e. P(HFE/Context). Uncertainty analysis in Auto-CREAM is performed using a numerical technique, unlike the expert elicitation or empirical approaches adopted by HRA methods.