Computational engineering design under uncertainty: an aircraft conceptual design perspective

Presented in this thesis is a novel methodology for aircraft design optimization in the presence of uncertainty, with emphasis on the conceptual design stage. In the initial part of the thesis, the uncertainty typologies of interest for aircraft design are identied within a broader epistemological framework. The main implications for non-deterministic computational design are also outlined. The focus is then restricted to uncertainties that can be modeled by probability theory. In this context, a methodology is developed to enhance robust design optimization (RDO). Firstly, the problem is formulated in order to relax, when required, the common RDO assumption about the normality of objectives and constraints. Secondly, starting from engineering considerations about the risk related with design unfeasibility, suitable estimates of tail conditional expectation are introduced in the set of robustness metrics. The proposed formulation requires the estimation of mean and variance of objec¬tives and constraints. To calculate such moments, a novel uncertainty propaga¬tion technique is proposed, which achieves a favorable trade-obetween the ac-curacy of the estimates and the required computational cost. Peculiar features of the propagation technique are exploited to couple the propagation and the opti¬mization phases for the classes of gradient-based methods and the derivative-free pattern search methods. Also analyzed are the possible advantages achievable when the two types of algorithms are hybridized. The usefulness of the proposed methodology for conceptual design optimization is demonstrated with the aid of two engineering design problems, concerning the sizing of passenger aircraft and the design of transonic airfoils.