Multi-Fidelity Gradient-Based Optimization for High-Dimensional Aeroelastic Configurations

The simultaneous optimization of aircraft shape and internal structural size for transonic flight is excessively costly. analysis the governing physics expensive, in particular highly flexible aircraft, search optima using samples can scale poorly with design space size. This paper has a two-fold purpose targeting scalable reduction sampling. First, new algorithm explored computing derivatives by analytically linking objective definition, geometry differentiation, mesh construction, analysis. analytic computation enables accurate use more efficient gradient-based methods. Second, scalability multi-fidelity assessed high dimensions. method leverages model during line further sampling costs. demonstrated cases aerodynamic aeroelastic considering both sizing separately combination spaces ranging from 17 to 321 variables, which would be infeasible typical, surrogate-based consistently led high-fidelity evaluations compared single-fidelity problems, but frequently resulted cost penalty involving sizing. While optimizer was successfully applied problems hundreds results underscore importance accurately gradients motivate extension approach constrained