Coupled Aerostructural Optimization using High-Fidelity Aerodynamics and Surrogate Modeling for the Structure

Abstract

View Video Presentation: https://doi.org/10.2514/6.2022-3359.vid A new multidisciplinary design optimization (MDO) strategy is presented for the coupled aerostructural optimization of an aircraft wing at the preliminary design stage. The new strategy features high-fidelity aerodynamic analysis and a surrogate model of the structure, integrated with the coupled-adjoint method to compute gradients. In this method, the surrogate is trained only once, prior to the optimization, without being updated during the optimization. This gives the new strategy a monolithic architecture, and compartmentalizes the structural discipline, for ease of implementation in an industrial environment with distinct teams of disciplinary experts. The surrogate model approximates a full structural sizing process, returning the structural weight and equivalent stiffness of an optimized wing structure, given inputs of global geometry parameters and sizing loads. Utilizing this type of surrogate eliminates the need for numerous structural constraints in the aerostructural optimization problem, thereby removing the need for the aggregation of constraints in the coupled-adjoint formulation. Furthermore, by outputting the wing stiffness, the surrogate enables the multipoint design optimization of flexible wings, as the stiffness may be used to determine the elastic response for multiple design load cases. The sizing loads are passed to the surrogate as a parameterized loading envelope to keep the number of surrogate inputs small, and allow the surrogate to be trained using a wide variety of representative load envelopes to cover the range of expected designs. The new method is being implemented in the Isight process integration framework, using the NSU3D Reynolds-Averaged Navier-Stokes code for the aerodynamic analysis and adjoint implementation, and a proprietary wing structure sizing code from Bombardier Aviation. The new method is applied to the aerostructural optimization of the CRM configuration.