Improvement of the Representativity of the Morris Method for Air-Launch-to-Orbit Separation
详细信息 查看全文
- 作者：Henri Sohier* ; henri.sohier@onera.f" class="auth_mail" title="E-mail the corresponding author ; Jean-Loup Farges** ; Helene Piet-Lahanier***
- 关键词：air-launch-to-orbit ; store separation ; sensitivity analysis ; screening ; elementary effects ; sampling
- 刊名：IFAC-PapersOnLine
- 出版年：2014
- 出版时间：2014
- 年：2014
- 卷：47
- 期：3
- 页码：7954-7959
- 全文大小：496 K
文摘
This work is part of a technical study of the French space agency CNES (Centre National d'Etudes Spatiales) and the French aerospace laboratory Onera for the development of an air-launch-to-orbit system. The separation between the space launcher and the aircraft must be studied with a sensitivity analysis to estimate the role of the different uncertainties in the risk of collision and improve the robustness of the controller. As the number of factors is too large to apply a quantitative method, the qualitative Morris method is first used for factors fixing. The Morris method is generally used with limited calculations to roughly estimate the factors' impact, but the quality of its results is rarely discussed. The Morris method is based on the random sampling of so-called trajectories or, less frequently, radial points. The representativity of the trajectories and radial points is analyzed and improved with a new solution based on a discrete Latin hypercube. Extensive tests are used to compare the quality of the results of various procedures with five common test functions. A new quality indicator, more sensitive, is developed. It shows that the new solution greatly improves, at no additional cost, the estimation of the factors' impact by the Morris method. The classic radial points appear to be more efficient than the commonly used classic trajectories. From a more global perspective, this study raises the question of the convergence in the Morris method. The application of the new solution to the air-launch-to-orbit separation provides results with a much better quality indicator than in the previous campaign of sensitivity analysis.