In human reliability analysis (HRA) there is seldom a sufficient amount of representative data available to base the human error probability (HEP) estimates on. Instead, most HRA methods apply combinations of theoretical considerations and common sense judgements (International Atomic Energy Agency (2020)). Recent efforts in the nuclear field to collect performance data from control room simulators for the use of HRA, e.g. SACADA (Chang et al. (2014)) and HuREX (Jung et al. (2014)), have focused on detailed task decompositions and observations on the number of successes, failures, and near failures of the tasks. The main drawbacks of this approach are 1) that it requires a large amount of data in order to produce statistically justifiable results, and 2) that it is not straightforward to re-aggregate the low-level tasks into human failure events (HFEs) for use in a probabilistic safety assessment (PSA) due to e.g. various feedback cues and possibilities for recovery of failed actions.
This paper explores the suitability of performance measures used in main control room simulators to inform the estimation of HEPs. To overcome the drawbacks of the aforementioned databases, the outlined approach focuses on HFEs of a level of abstraction appropriate for PSA. Instead of counting failures, more detailed evidence of the performance level is used. The suitability of potential performance measures, e.g. the Process Awareness and Situation Understanding (PASU) measure (Braarud (2017)), developed in the OECD Halden Reactor Project, are explored and evaluated. The methodology to estimate the HFE probabilities based on the performance data is also outlined.