Extending the survival signature paradigm to complex systems with non-repairable dependent failures

George-Williams, Hindolo and Feng, Geng and Coolen, Frank PA and Beer, Michael and Patelli, Edoardo (2019) Extending the survival signature paradigm to complex systems with non-repairable dependent failures. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 233 (4). pp. 505-519. ISSN 1748-006X

[img]
Preview
Text (George-Williams-etal-JRR2019-Extending-survival-signature-paradigm-complex-systems-non-repairable-dependent-failures)
George_Williams_etal_JRR2019_Extending_survival_signature_paradigm_complex_systems_non_repairable_dependent_failures.pdf
Accepted Author Manuscript

Download (528kB)| Preview

    Abstract

    Dependent failures impose severe consequences on a complex system’s reliability and overall performance, and a realistic assessment, therefore, requires an adequate consideration of these failures. System survival signature opens up a new and efficient way to compute a system’s reliability, given its ability to segregate the structural from the probabilistic attributes of the system. Consequently, it outperforms the well-known system reliability evaluation techniques, when solicited for problems like maintenance optimisation, requiring repetitive system evaluations. The survival signature, however, is premised on the statistical independence between component failure times and, more generally, on the theory of weak exchangeability, for dependent component failures. The assumption of independence is flawed for most realistic engineering systems while the latter entails the painstaking and sometimes impossible task of deriving the joint survival function of the system components. This article, therefore, proposes a novel, generally applicable, and efficient Monte Carlo Simulation approach that allows the survival signature to be intuitively used for the reliability evaluation of systems susceptible to induced failures. Multiple component failure modes, as well, are considered, and sensitivities are analysed to identify the most critical common-cause group to the survivability of the system. Examples demonstrate the superiority of the approach.