Deep Learning Based Transonic Flutter Prediction Model for Multidisciplinary Design Optimization of Medium Range Truss Braced Wing Aircraft

View Video Presentation: https://doi.org/10.2514/6.2023-3944.vid This research focuses on the development and implementation of a deep-learning-based Transonic Flutter Constraint Model for the Multidisciplinary Design Optimization (MDO) of Truss-Braced Wing (TBW) aircraft. TBW configurations, as next-generation aircraft, demand accurate flutter analysis in the transonic flight regime. High-fidelity Computational Fluid Dynamics (CFD) methods, although accurate, are very computationally expensive and unsuitable for MDO. The study investigates the use of extended indicial function-based flutter analysis via aerodynamic strip theory for 3D wings as a potential replacement for CFD. However, this analytical method still entails significant computation time. As a solution, a deep-learning-based flutter model is developed as a surrogate for reducing computational costs. The research highlights the challenges of generating datasets for deep-learning-based flutter models and the implementation of the surrogate model within the MDO framework. Emphasis is placed on the importance of the deep-learning-based flutter model due to the significant time consumed by flutter analysis in structural modules during the design process. Results obtained from the MDO framework with the surrogate model reveal promising reduction of computational time in the prediction of flutter constraint for TBW design, showcasing the effectiveness of the proposed method. The the total wall-clock time of the new DNN method was reduced by 1500 times compared to the previous method. Overall, this study contributes to the development of more efficient and accurate MDO processes for TBW aircraft, paving the way for the future of sustainable aviation.