Identification of influential uncertainties in Monte Carlo analysis

Monte Carlo simulation is a powerful and a practical tool for evaluating nonlinear systems. Its advantage is that it allows the effects of combinations of uncertainties to be taken into account. When the result of a Monte Carlo simulation is unsatisfactory, further investigations of both the system model and the control system are necessary, and it is important to identify those uncertain parameters that significantly influence the outcome of the simulation. However, the influential parameters are usually difficult to identify because multiple uncertain parameters are incorporated into a simulation simultaneously. A methodology is presented for identifying influential parameters in Monte Carlo analysis. When a Monte Carlo simulation yields an unsatisfactory result, the influential uncertainties are identified by further Monte Carlo simulations incorporating test vectors derived from the original uncertain parameter vector and by a statistical hypothesis test. The method is applied to the simulation results of an unmanned flight system, demonstrating its effectiveness in a practical application.