High-fidelity aerodynamic and aerostructural optimization of UAV propellers using the adjoint method

Unmanned aerial vehicles (UAVs) have seen widespread usage in commercial and military applications because they can significantly reduce the cost and human risk in a mission. The propeller is a crucial component for UAVs and its aerodynamic efficiency impacts the vehicle's overall performance. Aerodynamic optimization is a powerful technique that can use computer simulations to find the best possible design for UAV propellers automatically. However, existing propeller aerodynamic optimization commonly uses low-fidelity models, e.g., blade element momentum, and it is not clear to what extent high-fidelity optimization can benefit the UAV propeller design. In this paper, we develop the capability to conduct high-fidelity aerodynamic and aerostructural optimization for UAV propellers. We use finite-volume computational fluid dynamics and finite-element structural dynamics solvers to simulate the fluid and solid domains, respectively. We then use the discrete adjoint approach to compute the derivatives, which allows us to conduct gradient-based optimization with a large number of design variables. To consider fluid-structure interaction and its derivative computation, we utilize OpenMDAO/MPhys, an open-source framework that facilitates high-fidelity multidisciplinary design optimization. We conduct aerodynamic and aerostructural optimizations with shape and planform variables (e.g., span and chord). The objective function is the propeller shaft power, and the constraints include propeller thrust, mass, von-mises stress, and propeller geometry (e.g., thickness and curvature). Compared with the baseline design, the shape-only aerodynamic, shape+planform aerodynamic, and shape+planform aerostructural optimizations exhibit 7.4%, 12.8%, and 11.8% power reduction, respectively, and all the constraints are satisfied. This study has the potential to significantly reduce the time period for designing high-performance UAV propellers.