Component-based timed hazard analysis of self-healing systems

Today, self-healing is increasingly used in embedded real-time systems, that are applied in safety-critical environments, to reduce hazards. These systems implement self-healing by reconfiguration, i.e., the exchange of system components during run-time that aims at stopping or removing failures. This reaction is subject to hard real-time constraints because reacting too late does not yield the intended effects. Consequently, it is necessary to analyze the propagation of failures over time and also take into account how the propagation of failures is changed by the reconfiguration. Current approaches do not analyze the propagation times of failures and the changes of structural reconfiguration on the failure propagation. We enhance our hazard analysis approach by extending our failure propagation models by propagation times and taking the system's real-time reconfiguration behavior into account. This allows to analyze how a reconfiguration with certain duration changes the failure propagation of a real-time system and thus whether it is able to prevent a hazard. We show the feasibility of our approach by an example case study from the RailCab project.