Aerodynamic Shape Optimization of Forebodies Based on the GEK and Discrete Adjoint Methods

The aerodynamic shape optimization of air-breathing hypersonic vehicles often involves a high computational load due to the large number of design variables. Additionally, the complex shockwave structures and shockwave-geometry interactions can lead to a multimodal design space, which may result in local optima or unreasonable optimization directions. To address these issues, this paper introduces a hybrid optimization approach based on the discrete adjoint and gradient-enhanced kriging (GEK) methods for hypersonic vehicle forebody optimization. A global search based on the GEK model is first applied for initial designs, followed by gradient-based optimization to efficiently refine the aerodynamic characteristics. Compared to traditional global optimization methods, this approach enables one to avoid the curse of dimensionality and shows a high optimization efficiency; meanwhile, compared to a gradient-based optimization method, one can avoid issues related to local optima or other unfeasible optimization directions stemming from inappropriate initial individuals. Additionally, the method utilizes the gradient information for design variables to construct the surrogate model, which significantly reduces the sampling points, thereby enhancing the optimization efficiency. This method achieves favorable optimization results when applied to a forebody optimization problem.