In this paper, we present a study on the aerodynamic shape optimization of a three-dimensional subsonic engine using computational fluid dynamics simulations. Gaussian process-based surrogate modeling (kriging) and screening techniques are combined to tackle the high cost associated with both computational fluid simulations and the large number of design variables involved, with a multi-objective genetic algorithm used to obtain the Pareto fronts. The primary goal of the study was to identify the tradeoff between performance and noise effects associated with various geometric features within practical costs. The fan face total pressure recovery is used to measure the aerodynamic performance, and the angle is used as an indicator of the noise impact on the ground. The geometry is modeled using a feature-based computer-aided design package. An unstructured tetrahedral mesh is generated for the subsequent using the Reynolds averaged Navier–Stokes flow equations. Analyses of variance techniques are used to the dominant geometry parameters, thereby reducing the number of design variables and computational in the trade study. Multiple Pareto fronts are constructed using progressively built kriging models based on data with the reduced parameter set. A full-scale search was also carried out for comparison with the produced using the reduced parameter set. The procedures outlined can be further applied to other problems with significant numbers of parameters and high-fidelity analysis codes.
No takes yet. Share an insight, caveat, or question.
Song et al. (2007) studied this question.