Quantifying the impacts of uncertainties on design performance,i.e.,uncertainty quantification (UQ),is the most important stage in uncertainty-based design optimization (UBDO) for air vehicles.Crude Monte Carlo (MC) simulation with easy operation and strong generality has been widely used for UQ: however,the convergence rate for this approach is slow and requires an enormous number of simulations,which becomes infeasible for the expensive function evaluations,e.g.CFD simulation.