Design optimisation of controlled aeroelastic aerofoils and wings

Abstract

To aid in the investigation of new simultaneous optimisation strategies for flexible vehicles and their control systems, a two-dimensional aerofoil optimisation which demands minimal computational effort is studied. Although computationally simple, the design allows for optimisation over multiple disciplines: the structure can be designed by varying the stiffness of supporting springs; the control architecture through weightings in a LQR controller; the observer by means of the placement of pressure sensors; and the aerodynamics via the shaping of the compliant trailing edge. Optimising over all these fields simultaneously is compared to a sequential methodology of optimising the open-loop characteristics first and subsequently adding a closed-loop controller. Parametrisation of the design vector and variable selection often require user input and are fixed during optimisation. Our research aims to automate this process. Furthermore, we investigate whether varying the parametrisation and number of design variables during the optimisation can lead to improvements in the final design. This parametrisation is shown to make the optimisation more robust with respect to the initial design, and facilitate an automated variable selection methodology. This variable selection allows for the dimension of the problem to be reduced temporarily and it is shown that this makes the optimisation more robust. The second half of the work focuses on the derivation of a cantilever model. The model consists of a geometrically-nonlinear, slender-beam described by a one-dimensional reference line that can deform in three-dimensional space; a two-dimensional, potential flow model defined over the span of the beam; and trailing edge flaps that can vary in size and position. The intrinsic beam formulation is chosen as it results in equations of motion with at most quadratic nonlinearities, which is exploited for deriving analytic derivatives. These derivatives are used to demonstrate how adjoint-based methods can accelerate aeroelastic calculations.