Numerical optimization methods can support the aircraft engine blade design process. Expensive aerodynamic and structural design evaluations and high-dimensional variables make it a challenging task. This thesis focuses on Gaussian process surrogate models used in Bayesian optimization, which can efficiently handle the first point. The three proposed method enhancements include dimensionality reduction, trust region, and problem decomposition strategies. They enable an efficient solution of high-dimensional optimization tasks for blade design and beyond.
«
Numerical optimization methods can support the aircraft engine blade design process. Expensive aerodynamic and structural design evaluations and high-dimensional variables make it a challenging task. This thesis focuses on Gaussian process surrogate models used in Bayesian optimization, which can efficiently handle the first point. The three proposed method enhancements include dimensionality reduction, trust region, and problem decomposition strategies. They enable an efficient solution of high...
»