Reevaluating conceptual design fidelity: An efficient supersonic air vehicle design case

Abstract

Multifidelity and multidisciplinary design optimization (MDO) results for an efficient supersonic air vehicle (ESAV) are presented. This work builds upon previous published work that created a multifidelity and multidisciplinary analysis and design optimization framework. Based on the usual aircraft conceptual design assumption that low-fidelity analyses are capable of converging upon a design through MDO that would at least be representative of that which would be predicted by higher fidelity methods, a process is created to allow low-fidelity multidisciplinary analysis runs to guide the higher fidelity optimization in an effort to reduce the total required computational expense. Due to disparate physical simulation results between the different fidelity levels, the expected reduction in computational effort was not realized. A separate low-fidelity optimization and a higher fidelity optimization were then subsequently performed. A MDO method that required an order of magnitude fewer objective function evaluations than the low-fidelity MDO method was developed to perform the more computationally expensive higher fidelity MDO in the full ESAV design space. Although optimal low-fidelity and higher fidelity ESAV configurations are identified, the low-fidelity optimal solution is infeasible when analyzed with the higher fidelity framework, and vice versa. This was a surprising and unexpected result as these two fidelity level model suites were developed by the same team for the same aircraft. The common assumption that low-fidelity MDO design methods will find a feasible design close to the eventual higher fidelity design did not hold for the ESAV case. Sensitivity studies were performed around the optimal design to gain insight into this important region of the design space with respect to the particular fidelity level models used.