An advanced method for the sensitivity analysis of safety system

Safety system always involves variation on account of the epistemic uncertainty of the inputs. Therefore, it is significant to identify the uncertainty source of the output for the safety system. Sensitivity analysis (SA) which measures the effect of variance variation of inputs on the absolute change in the variance of system unsafety is a useful tool for identifying the importance of inputs. A finite difference method has been proposed to estimate the effect in the existing work. This method may be numerical unstable or inaccuracy and is computational heavy. In order to overcome these issues, an advanced method which combing the simulation method and the analytical deduction of the partial derivative is established in this paper to estimate the SA indices. Discussion and several examples are introduced to illustrate the efficiency and accuracy of the proposed method when comparing with the finite difference method.