This paper describes an extension to the manifold mapping (MM) optimization algorithm for aerodynamic shape design to include gradients through the calculation of adjoint sensitivities. The gradient information is used to enhance the MM surrogate, and to speed up the optimization process. Moreover, the addition of gradients based on adjoint sensitivity allows for handling of large-dimensional design spaces. The gradient-enhanced manifold mapping (GMM) algorithm replaces the direct optimization of a computationally expensive model by an iterative updating and re-optimization of a fast physics-based replacement surrogate model. During the optimization process the GMM surrogate is constructed using a low-fidelity model which is corrected using high-fidelity data (evaluated once per design iteration). The performance of the method is demonstrated on a benchmark case involving drag minimization in two-dimensional transonic inviscid flow. For the demonstration case, the results show that GMM obtains comparable designs as MM with 70% less computing time.
展开▼
