123 Methodology for Improving HRA by Simulator Studies (PSAM-0391)

Data on human reliability will considerably improve the basis for and the use of applied Human Reliability Analysis (HRA) methods within Probabilistic Risk/Safety Assessment (PRA/PSA). Simulator experiments can produce important basic information for HRA method development and data for informing use of existing HRA methods [1]. In simulators, one can study variations in context and how this impacts human performance. The frequencies of task failures as defined in PRA are generally low, failures might not be that easily definable or observable and the number of simulator runs are limited in terms of practically available resources. One way to obtain useful results on human reliability is to study human performance under adverse context in difficult scenarios, where a relatively high frequency of failures or degraded performance is expected, and compare this to advantageous context in nominal conditions where good performance is expected. This information can be used to improve HRA methods, both in structure and use.

Simulator studies can inform HRA in various ways. They can inform HRA practitioners in the use of HRA methods, both relating to the occurrence of context and to the influence of context on human performance. They can inform HRA method development, and they can provide input to generic data bases or repositories.

In this paper, a methodology for HRA studies is presented, proposing to do controlled studies and manipulate certain variables that can be related directly to Performance Shaping Factors (PSFs) in several HRA methods. When patterns of good or bad performance are found, e.g., manifested in response times, detailed qualitative analyses are done in order to gain insights in the causes for the observed performance. As important as the results directly related to PSFs, are the insights and results related to crew characteristics that are gained by this method.

The OECD Halden Reactor Project is currently undertaking a range of studies to support HRA, utilizing a simulation facility called HAMMLAB (HAlden huMan-Machine LABoratory) and this paper outlines the needs, relevance, use, and the indicated main methodology used in and under development for HAMMLAB.