Aeroelastic reliability based optimisation of composite plates via surrogate modeling

ifasd2024 Tracking Number 17

In aerospace engineering, optimizing composite laminates in aeronautical structures can be a complex problem which often results in intricate models that are computationally demanding. Addressing this, our work introduces a reduced model approach, employing a composite plate as a representative element of an aircraft wing [1]. The primary objective is to enhance aeroelastic behaviour in response to wind gusts, focusing on minimizing the bending moment generated by such gusts. Concurrently, flutter analysis is conducted, serving as the primary constraint in our optimization problem. The optimization of composite materials presents unique challenges, particularly due to the non-convex nature of the stacking sequence design space and the presence of numerous local minima. Furthermore, the variable number of design parameters (angles of plies) adds to the complexity given that the possible stacking error must be considered within the optimisation process [2]. To navigate these challenges, our approach utilizes material homogenization, transforming the problem into one within a convex design space with a fixed number of variables. Building upon this foundation, we implement a surrogate model-based optimization strategy. The surrogate models, are developed within the homogenized design space using constrained Efficient Global Optimization (C-EGO) [3]. Following the establishment of these models, a Genetic Algorithm (GA) is employed to determine the optimal stacking sequence [4]. The genetic algorithm attempts to find a stacking sequence that optimises the aeroelastic problem with a reliability-based approach. Considering the possible uncertainties in the ply orientation of the composite laminate the GA manages to tailor the composite to the aeroelastic constraints. The results of this two-step optimization process, combined with the proposed dual design space approach, demonstrate the efficacy of surrogate models in exploring the full spectrum of optimal composite configurations. Moreover, the coupling of this methodology with the aeroelastic problem at hand permits for a cost-efficient optimisation reducing notably the number of calls to the aeroelastic solver compared to gradient based optimisation methods. This methodology not only streamlines the design process for composite aeroelastic structures but also showcases its possible applicability in more complex aeroelastic scenarios.