Dependability assessment of network-based safety-related system

Advances in electronic integration and radio communication have led to the emergence of a new kind of safety systems, i.e. Wireless Sensor Network (WSN). This network-based safety-related system is becoming more and more present in the domain of safety due to its easy deployment. It does not need a wire infrastructure and its range of applications is wide. Usually, such a system is composed of various nodes (sensors) collaborating to monitor a targeted phenomenon. In most cases, nodes are battery powered and this is the weakness of the system makes it necessary to design an energy saving policy. The present paper gives a dependability viewpoint of such a system. A modeling framework is suggested integrating the interdependency of the components. Stochastic Petri nets are used to implement this model and two heuristics to schedule component activity. The first is based on the proposed importance measures, the second on a multi-objective genetic algorithm. The aim is to conserve energy and so to extend WSN dependability.

► The dependability of new safety instrumented systems based upon sensor network is studied.
► An energy consumption modelling of these battery powered sensors is presented along with the variable neighborhood. An energy saving policy is necessary in order to extend protective system lifetime.
► Safety function availability, Quality of Service and Coverage level are suggested criteria to assess the dependability.
► Stochastic Petri nets implement this modelling framework. Two heuristics are employed to schedule sensor activity built upon a sleep/wakeup protocol. The first uses defined importance measures and the second multi-objective algorithm with the aim to preserve energy and so to extend network lifetime.