Design space exploration using uncertainty-based bounding methods in computational fluid dynamics

A method is proposed to use computational fluid dynamics (CFD) as the analysis tool in a sequential decision process (SDP). The SDP efficiently explores and reduces the trade space by repeatedly bounding the prediction of the design parameters through multiple analysis iterations of increasing fidelity. In the context of fluid-dynamic shape design with CFD, the trade space containing the optimal design is reduced using a sequence of computational meshes, each having reduced error bounds compared to those prior. Earlier iterations, with higher numerical uncertainty but lower computational time, are used to eliminate regions not of interest within the trade space. The reduced subset is then further evaluated using CFD with tighter bounds, achieved through a more costly, refined computational mesh. This process is demonstrated on an aerodynamic shape design in a two-parameter, drag minimization study of a generic fuselage pod.