High-fidelity Aerodynamic and Aeroacoustic Multi-Objective Bayesian Optimization

Abstract

Employing high-fidelity numerical simulations in engineering design problems, particularly in aerodynamic and aeroacoustic optimization problems, introduces a significant challenge in terms of time constraints, which cannot be resolved by conventional methods such as evolutionary algorithm-based methods that require a considerably large number of objective function evaluations. This work substantiates the idea that this limitation can be overcome by using surrogate based optimization. More precisely, multi-objective Bayesian optimization that takes Kriging as a surrogate model and Expected Hypervolume Improvement as an infill criterion. With this approach, it is possible to obtain a Pareto front at a relatively small computational budget as demonstrated by solving well-known analytical optimization problems. The objective of this work is to assess the applicability of a multi-objective Bayesian optimization framework to an aerodynamic-aeroacoustic shape optimization problem where the objective functions are evaluated by means of high-fidelity computational fluid dynamics and acoustic simulations. Here, the proposed optimization method shows its capability to return Pareto fronts that contain various design trade-offs that result in reduced drag, reduced pitching moment, and reduced aerodynamic noise with a reasonable number of function evaluations.