An efficient method for evaluating the effect of input parameters on the integrity of safety systems

• We define a sensitivity index to measure effect of a parameter for safety system.
• We analyze the physical meaning of the sensitivity index.
• We propose an efficient simulation method to assess the sensitivity index.
• We derive the formulations of this index for lognormal and beta distributions.
• Results identify important parameters on exceedance probability of safety system.

Safety systems are significant to reduce or prevent risk from potentially dangerous activities in industry. Probability of failure to perform its functions on demand (PFD) for safety system usually exhibits variation due to the epistemic uncertainty associated with various input parameters. This paper uses the complementary cumulative distribution function of the PFD to define the exceedance probability (EP) that the PFD of the system is larger than the designed value. Sensitivity analysis of safety system is further investigated, which focuses on the effect of the variance of an individual input parameter on the EP resulting from epistemic uncertainty associated with the input parameters. An available numerical technique called finite difference method is first employed to evaluate the effect, which requires extensive computational cost and needs to select a step size. To address these difficulties, this paper proposes an efficient simulation method to estimate the effect. The proposed method needs only an evaluation to estimate the effects corresponding to all input parameters. Two examples are used to demonstrate that the proposed method can obtain more accurate results with less computation time compared to reported methods.