A Distributed UMDO Architecture Based on Surrogate Models for Launch Vehicle Design

During the design, manufacturing, testing and flying of launch vehicles, there are numerous uncertainties of the system and operating environment that affect the design. Uncertain Multidisciplinary Design Optimization (UMDO) is an effective method for evaluating these uncertainties, but it can be challenging to apply it in engineering problems due to high computational costs and slow convergence. To address this issue, we propose a high-fidelity distributed UMDO architecture based on Multiple Discipline Feasibility (MDF). In this architecture, surrogate models are used to quickly evaluate off-line aerodynamic forces and loads instead of conducting aerodynamic discipline analysis. Moreover, we propose to construct uncertainty surrogate models by combining the Maximum Entropy (MaxEnt) for the aerodynamic, engine and structure disciplines to quickly evaluate uncertainties, allowing for efficient solution of the UMDO problems. The proposed UMDO architecture is verified with a liquid launch vehicle optimization problem, and the results demonstrate that the distributed UMDO architecture based on surrogate models can effectively obtain the optimum solution.