System safety aspects

The following definition is presented: a safe system is n-resilient at m% to faults if: there exist at least n+1 distinct configurations of its subsystems; any n faults in n of the n+1 configuration maintains a safe system; and the highest probability of any set of n faults of the n+1 configurations does not exceed m%. It is shown how this type of resilient configuration can be achieved, either in time (repetitions) or physically (alternatives). A fault-model based on a graph-based model of the system has been built. Both the components and their interrelation characteristics which constitute safe behavior were defined. Monitoring relations between isolated components for runtime safety verification as well as for resiliency compensation are introduced. A detailed design example demonstrates how a safe system is constructed and how common-mode faults are dealt with.<
>