ABSTRACT
The methodological approach and results of two empirical studies are reported, which were directed at the goal of identifying interface attributes which are assumed to be crucial for fault detection and diagnosis during supervisory control. In the first study the basic assumption was that due to attributes of standard P&I-interfaces it is likely that operators may fail when they are confronted with unfamiliar scenarios. The results of 11 single-cases reveal that in most cases the information registered by the operators was sufficient, but that in more difficult scenarios only few operators were able to integrate this information adequately. On the basis of these results and taking into account cognitive psychological aspects, two different interfaces have been developed and modified for the second study. Results of an empirical evaluation of these interfaces show specific improvements of control performance and more integration of information. It is concluded that interface attributes and control demands presumably interact, which requires more flexible support.
