Reliability of Active Parallel Systems

Two-state random processes of Borges-Castanheta and filtered Poisson types are used to model the performance of units of an active parallel system. The states of the probabilistic model of a unit correspond to situations when the unit is operable or is undergoing repairs. The system is disabled when all l or k out of l units are simultaneously inoperable. It is shown that estimates of the probability of successful performance of the system during a reference period, or the reliability, based on the filtered Poisson model are attractively simple and accurate. Results obtained are used to find the reliability of very safe two-unit active parallel systems such as the computer system controlling the reactor power in nuclear power stations.