Abstract: Uncertainty always exists in any design problems, including aircraft conceptual design. The effect of such uncertainty on aircraft performances, therefore, must be accounted for during the optimization of aircraft conceptual design. In this article, a conceptual design of a passenger jet is used to demonstrate optimization under uncertainty. Different from deterministic optimization, optimization under uncertainty is much more computationally expensive. Hence sequential global surrogate models are developed to alleviate the computational demand. By sequentially adding sampling points both globally and locally in the design space, the accuracy of the surrogate models is gradually improved until reaching the expected level of accuracy. The cheaper surrogate models are then used in the optimization. In the optimization of the passenger jet conceptual design, there are five design variables and the take-off weight is minimized under four performance constraints. With the consideration of uncertainty, the objective (take-off weight) is much less sensitive to uncertainties, and the constraints (performance requirements) are also satisfied with much higher probabilities.

Key words: aircraft design, conceptual design, uncertainty, optimization, passenger jet, Monte Carlo methods