Interval parameter sensitivity analysis based on interval perturbation propagation and interval similarity operator

An interval parameter sensitivity analysis is developed to quantify the impact of simulation model parameters on the model outputs. This sensitivity analysis contains two main steps: the interval uncertainty propagation and the interval sensitivity index. The interval perturbation method is introduced to estimate the extreme bounds of model outputs according to the interval input parameters, which significantly reduces the computation cost of extensive Monte Carlo simulations. Since the output of the interval model are interval quantities, the traditional probabilistic sensitivity method and its sensitivity index are inappropriate as we only have the bounds of samples without inner data points. Hence, this work proposes an interval similarity operator based on the relative interval position operator, which is applicable to measure the variation of interval outputs. This interval sensitivity operator mainly quantifies the discrepancy between intervals based on six typical cases of the interval relative position. Finally, an academic case and a satellite structure case are analyzed to verify the feasibility and efficiency of the proposed method.