Selecting significant contextual factors and estimating their effects on operator reliability in computer-based control rooms

Abstract

To realistically quantify human reliability and generate concrete evidence for improving man–machine interfaces, the collection and analysis of empirical human reliability data is essential. In this study, we statistically modeled the reliability of nuclear power plant operators under various contexts using data from fully computerized control rooms. The data, obtained from a full-scope simulator of the APR1400 (Advanced Power Reactor 1400 MW) via the HuREX (human reliability data extraction) framework, were analyzed with multivariable regression. To select the significant variables and estimate the effects of the selected variables among 55 candidates, 4 strategies were employed. From this analysis, insights regarding the variable selection and impact estimation processes were obtained. In addition, the effects of six contextual variables that were commonly selected in the different approaches are discussed in detail.