Global Sensitivity Analysis for Event-driven Indicators Used in Probabilistic RAMI Analysis

In this paper a novel conceptual design of a method for performing Global Sensitivity Analysis (GSA) for discrete event-driven indicators used in standard probabilistic Reliability, Availability, Maintainability and Inspectability (RAMI) analysis for a generic industrial plant is presented. The main objective of the GSA is to assess how sensitive the operability of the facility (output) is to various failure events (inputs) in the system. This is done in four steps. Firstly, suitable output indicators of plant's operability are defined given the predefined collection of basic events, which serve as inputs to the system. Secondly, the sources of uncertainty associated with basic events are detected. Thirdly, the events that are the most detrimental to the functionality of the facility are identified. Finally, the sensitivity of output indicators to input events selected in previous step is evaluated so that the most sensitive events are identified. The global sensitivity is determined with Sobol indices, which accounts for higherorder interactions between inputs. Information derived from the proposed GSA procedure, which complements standard RAMI analysis, is meant to support design and decisionmaking processes by identifying components and failure events with the highest impact on the system's availability (also through nonlinear interactions). This knowledge is further used to improve reliability and robustness of the analyzed system with optimal allocation of resources. The functionality of the proposed GSA method is visualized on a subsystem of the International Fusion Materials Irradiation Facility (IFMIF).