Supersonic Business Jet Design Using Knowledge-Based Genetic Algoithm with Adaptive, Unstructured Grid Methdology

Abstract

In the design of supersonic low-boom aircraft, it is important to balance the aerodynamic performance and sonic boom requirements in a way that represents the best compromise for the overall design. Since the ground sonic boom is typically not a smooth function of the design variables and may actually contain multiple local minima, it is important to select an optimization algorithm that is able to cope with this kind of design space. In this work, we study the use of Kriging approximation models for both boom and performance and use them in conjunction with genetic algorithm techniques to investigate the computational cost and characteristics of such an approach for the design optimization of a low-boom supersonic business jet (SBJ). Direct use of genetic algorithms with high-fidelity CFD analysis tools has been limited by the inherently large computational cost of genetic algorithms (GAs). The use of computationally inexpensive approximation models in lieu of high-fidelity CFD greatly improves the robustness and efficiency of the design process for searches in relatively large design spaces. In order to improve the performance of this method, a new hybridization strategy that combines a GA with gradient information is proposed and its improved convergence rate is demonstrated. Regardless, the proposed procedures still require a large number of evaluations of the flow and boom patterns for different points in the design space. For this purpose we have built two automated Euler analysis tools that use a CAD-based geometry engine, and both multiblock-structured and unstructured, adaptive meshing techniques (they are named QSP107 and QSP-UA respectively). QSP-UA, has been developed to handle the geometric detail of the complete configuration including wing, fuselage, nacelles, diverters, empennage, etc., and to provide accurate performance and boom data. Results of sample test problems, and a 15-dimensional design case are presented and discussed.