Multi-objective aerodynamic design exploration using multi-fidelity methods and pareto set identification techniques

The research objective of this work is to accelerate the process of multi-objective aerodynamic design exploration when using computationally intensive simulation models. The design of aerodynamic surfaces is important for modern engineered systems such as unmanned aerial systems and turbomachinery. Physics-based simulations are needed for capturing the nonlinear system behavior and nonlinear interactions between disciplines. The key challenges with using high-fidelity physics-based simulations as part of aerodynamic design include (1) the high computational cost of the simulations (ranging from few hours to days or weeks on high performance computing clusters), (2) large numbers of conflicting objectives and constraints, and design variables, and (3) the repetitive evaluations during the design exploration phase. The main contributions of this thesis are the adaptation and integration of multi-fidelity methods and Pareto set identification techniques to rapidly determine the best possible trade-offs of the aerodynamic characteristics. The proposed multi-fidelity aerodynamic Pareto set identification techniques use sequential domain patching and point-by-point exploration. The algorithms are validated using analytical problems and demonstrated on aerodynamic design problems involving transonic airfoils and subsonic wings. The proposed algorithms are benchmarked against a surrogate-assisted multi-objective evolutionary algorithm. It is found that approaches produce comparable Pareto fronts. Furthermore, the proposed multi-fidelity point-by-point aerodynamic MOO algorithm is over 50% more efficient than the benchmark method. The value of the proposed algorithms is more visible in cases where designers have a limited computational budget and only a few Pareto optimal points are required in the vicinity of a target design