ABSTRACT
The purpose of this manuscript is to propose a novel methodology for conducting uncertainty analysis for event-driven indicators of system’s availability. In this approach the system’s availability depends on the frequency of components’ failures and their duration. For all the components the parameters that determine their behavior are obtained from statistical analysis, which means they are given with certain uncertainty, hence the system’s availability is also uncertain. In this paper we present a numerically effective algorithm that assesses uncertainty about system’s availability. A modular approach has been adopted, in which the uncertainty about the complete system’s availability is estimated by combining the information obtained for separate subsystems using the knowledge about the functional relationships between them. Furthermore, to assure numerical efficiency, a novel approach has been adopted, which combines screening approach with quasi-random Sobol sampling approach. The functionality of the proposed method is visualized on an illustrative example.
