The adaptive cycle engine can vary the cycle in wider range and meet the requirement of mixed missions. It is expected to be the propulsion system for the next generation of aircraft and has therefore attracted a lot of attention. The uncertainty in component performance can offset the performance benefits of adaptive cycle engine. Therefore, the robust design should be applied at the engine performance design stage to reduce the effect of component performance uncertainty. Compared with conventional engines, the adaptive cycle engine has more variable geometry components. It causes two major difficulties for the robust design: one is the high computation cost and the other is the complex control law design. In this paper, a robust design method based on a two-level surrogate model and a two-stage design process is presented to overcome these difficulties. In this method, the probabilistic method is applied to quantify uncertainty and the selection principle of optimization target is studied. The cycle parameters and control law are optimized through a two-stage design process. The optimization is developed based on a two-level surrogate model to reduce the computational cost. This robust design method is validated by the Monte Carlo method. According to this study, performance under confidence probability is a better optimization target than the signal-to-noise ratio of performance. Using this method, the probability of engine performance meeting demand can be improved from about 50% to over 96%. The computational cost of uncertainty quantification is only 0.5% of the Monte Carlo method and the computational cost of whole optimization design process is less than the cost of once uncertainty quantification through Monte Carlo method. This method can improve the robustness of adaptive cycle engine performance and reduce the computational cost significantly. It can also be applied to the robust design of other complex gas turbine engines performance.

自适应循环引擎可以在更大范围内改变循环并满足混合任务的要求。预计它将成为下一代飞机的推进系统，因此引起了很多关注。组件性能的不确定性可能会抵消自适应循环引擎的性能优势。因此，应在发动机性能设计阶段应用稳健的设计，以减少部件性能不确定性的影响。与常规发动机相比，自适应循环发动机具有更多的可变几何部件。对于鲁棒性设计造成两个主要困难：一个是高计算成本，另一个是复杂的控制律设计。在本文中，为了克服这些困难，提出了一种基于两级代理模型和两阶段设计过程的鲁棒设计方法。该方法采用概率方法对不确定性进行量化，研究了优化目标的选择原则。通过两阶段的设计过程优化了循环参数和控制规律。该优化是基于两级代理模型开发的，以降低计算成本。这种稳健的设计方法已通过蒙特卡洛方法进行了验证。根据这项研究，置信概率下的性能比性能的信噪比是更好的优化目标。使用这种方法，发动机性能满足要求的可能性可以从大约50％提高到超过96％。不确定性量化的计算成本仅为0。5％的蒙特卡洛方法和整个优化设计过程的计算成本低于一次通过蒙特卡洛方法进行不确定性量化的成本。该方法可以提高自适应循环发动机性能的鲁棒性，并显着降低计算成本。它也可以应用于其他复杂燃气涡轮发动机性能的稳健设计。