On steady state probabilities of renewable system with Marshal-Olkin failure model (Q1757273)

The author deals with the problem of stability (insensitivity or low sensitivity) of steady-state characteristics of heterogeneous double redundant hot-standby renewable systems with Marshal-Olkin failure models. The paper by \textit{B. A. Sevast'yanov} [Teor. Veroyatn. Primen. 2, 106--116 (1957; Zbl 0082.13101)] should be mentioned among the earliest ones where insensitivity of Erlang formulas to the service time distribution was investigated. Later on, \textit{I. N. Kovalenko} [Investigations on analysis of complex systems reliability. Naukova Dumka, Kyiv (1976)] (see also [\textit{V. Rykov}, Commun. Comput. Inf. Sci. 356, 147--156 (2013; Zbl 1274.60278)] for more details and references) obtained the necessary and sufficient conditions of insensitivity of stationary reliability characteristics of a redundant renewable system with exponential life time and general repair time distributions of its components to the service time distribution. The obtained conditions depend on the sufficient amount of repairing facilities, i.e. in possibility to immediately start the repair of any of the failed components. The asymptotic insensitivity of the reliability characteristics of cold-standby double redundant heterogeneous systems to the life and repair time distributions under ``quick'' restoration follows from results published in the papers by by \textit{B. V. Gnedenko} [Izv. Akad. Nauk SSSR, Tekh. Kibern. 1964, No. 4, 3--12 (1964; Zbl 0203.21603); Izv. Akad. Nauk SSSR, Tekh. Kibern. 1964, No. 5, 111--118 (1964; Zbl 0142.16502)] and \textit{A. D. Solov'ev} [Izv. Akad. Nauk SSSR, Tekh. Kibern. 1970, No. 1, 56--71 (1970; Zbl 0195.49701)]. In the book by \textit{V. Kalashnikov} [Geometric sums: bounds for rare events with applications: risk analysis, reliability, queueing. Dordrecht: Kluwer Academic Publishers (1997; Zbl 0881.60043)] estimates of the convergence rate were investigated in terms of moments of corresponding distributions. Most of these investigations deal with systems, whose components fail independently. In their paper \textit{A. W. Marshall} and \textit{I. Olkin} [J. Am. Stat. Assoc. 62, 30--44 (1967; Zbl 0147.38106)] (see also [\textit{A. W. Marshall} and \textit{I. Olkin}, J. Multivariate Anal. 139, 219--226 (2015; Zbl 1320.62027)]) proposed an approach based on bivariate distribution with dependent components. The model can be used as a failure model for a two-component reliability system. For references on papers dealing with this distribution and its generalizations see, for example, the paper by \textit{D. Kozyrev} et al. [``Reliability function of renewable system under Marshall-Olkin failure model'', Reliability: Theory Appl. 13, No. 1, 39--46 (2018; \url{doi:10.24411/1932-2321-2018-00004})]. However most of the papers consider only the bivariate distributions and their properties and use the Marshal-Olkin model only for the first failure. In the present paper the Marshal-Olkin failure model is used to investigate properties of a heterogeneous double redundant standby renewable system. The steady state characteristics of the system are calculated. The obtained characteristics demonstrate an evident sensitivity to repair time distributions. But under rare failures these characteristics become asymptotically insensitive to distributions of the system components.