摘要
The launch process of a multi-stage launch vehicle is significantly influenced by uncertain parameters,including air density,aerodynamic parameters,and engine thrust,which often exhibit deviation.Predicting the trajectory range of the launch vehicle under the influence of uncertainty is essential before launch,and uncertainty quantification serves as a crucial method to address this challenge.In traditional uncertainty quantification for launch vehicles,unknown parameters are often assigned specific distributions based on prior knowledge.However,prior knowledge is sometimes subjective,and unknown parameters are often assigned conservative ranges to meet safety margins.In addition,the flight data of the past launch is precious,especially in quantifying the uncertainty of reusable or same-type launch vehicles.This paper utilizes flight data to estimate parameters base on Bayesian methods and integrates the estimation results with prior knowledge,which can more objectively set the distribution of uncertain parameters.Reasonable distribution has a positive impact on uncertainty quantification,which can avoid control strategies that are not robust enough or overly redundant.Therefore,the uncertainty quantification for launch vehicles is discussed under different information sources.In addition,the algorithm is accelerated based on Gaussian process regression and polynomial chaos expansions.
基金
supported by the National Natural Science Foundation of China(62273119,62173103).
