Reliability assessment for complex systems operating in dynamic environment

The paper aims to study a complex repairable system, operating in a dynamic environment. The reliability of the system is assessed by means of statistical modeling. We consider a system with several redundant components, operating in different modes defined on one hand by discrete states of its components and on the other hand by the operational environment conditions. These conditions are defined by continuous physical phenomena. The system's failure is conditioned by a certain combination of its components failures. The dynamic operational environment, the different discrete modes of system's operation and the deterministic and/or stochastic transitions between these modes are simultaneously accounted for by means of the Stochastic Hybrid Automaton (SHA). Monte Carlo simulations are carried out to generate a long trajectory followed by the system. The system's failure behavior is then modeled using the counting process framework applied to the simulated data. Precisely, failures are assumed to be driven by a Non Homogeneous Poisson Process (NHPP). The NHPP parameter estimates are used to assess the reliability of the complete system. The proposed approach is applied to an electrically powered furnace containing two redundant components: two temperature control loops. The temperature evolution is deterministic. The components' failure rates are allowed to be constant or time-dependent. The failure of the system occurs when both control loops fail or the furnace itself fails. The results show the relevance of statistical methods for global system's failure rate assessment in dynamic environment.