A Knowledge Acquisition Approach for Off-Nominal Behaviors

Natural language requirements often ignore unexpected or off-nominal behaviors (ONBs), which can result in catastrophic accidents in safety-critical systems. While some existing techniques can help identify ONBs, most of them are not systematic and algorithmic, and also they require a lot of human effort. In this paper, we propose an algorithmic and systematic approach for knowledge acquisition of ONBs in component based systems using a modified Apriori algorithm. Our approach analyzes component state transition rules to identify dependencies among components, which are used to group components that are dependent on each other into component sets. These sets are used for analysis of possible ONBs. We conducted an empirical study to evaluate our approach. Our results indicate that the component sets generated using our approach are able to expose missing dependencies and ONBs with much less human effort when compared to CCM.