Towards Automated Fault Pruning with Petri Nets

Embedded systems design is starting considering dependability issues even for mass-market systems. Soft error consequences must in particular be carefully analyzed. Usually, fault injection campaigns are run to analyze the consequences of transient faults, but the length of a comprehensive evaluation often collides with the severe requirements on design cycle times. We propose a new fault pruning technique to identify harmless components and computation cycles as soon as possible, thus avoiding useless fault injection experiments. The technique is based on a formal model of the system and we show that it can be used for both SEUs and SETs.