Data-driven aero-elastic modeling of flexible blended-wing-body aircraft
Date
2025
Authors
Zeinalzadeh, Aghil
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
As modern aircraft configurations like the BWB feature high-aspect-ratio wings prone to aeroelastic instabilities, there is an increasing need for real-time, constraint-aware flutter suppression methods. Traditional empirical and analytical techniques often struggle to capture time-varying dynamics near flutter speed. This thesis presents a data-driven framework for predicting the nonlinear aeroelastic response of flexible Blended Wing Body (BWB) aircraft near flutter condition. This technique is positioned within the context of active control frameworks such as Model Predictive Control (MPC), which require accurate, low-dimensional models for real-time application.
The study begins with a comprehensive review of aeroelasticity, flutter suppression strategies, and emerging data-driven methods, including Dynamic Mode Decomposition with control (DMDc) and Long Short-Term Memory (LSTM) networks. To support model development, low-fidelity aeroelastic simulations are performed using SHARPy, a nonlinear simulation platform for data generation and flutter analysis. A cantilevered BWB wing is excited with gust profiles near flutter, and transient responses are recorded. This simulation data is then used to build both DMDc and LSTM models. The LSTM model is designed for sequence forecasting, enabling it to capture long-term dependencies in the time series, while DMDc provides a linear approximation of system dynamics influenced by control inputs.
The SHARPy solver setup is validated using a Pazy wing benchmark, confirming its ability to reproduce unsteady behaviors such as Limit Cycle Oscillations (LCO) and transient gust responses. These results establish SHARPy as a reliable platform for data generation and flutter analysis. Building on this, the final chapter develops and evaluates the predictive models of the flexible wing based on the ORCA configuration. Flutter speed is estimated through modal analysis, followed by dynamic simulations using PRBS and sine-sweep velocity inputs.
Results show that LSTM significantly outperforms DMDc near flutter, achieving less than 2% prediction error within the training velocity range. Although accuracy decreases during extrapolation, LSTM maintains a computational advantage making it highly suitable for real-time use. The chapter concludes by conceptually integrating LSTM into an MPC framework, highlighting its potential for flutter suppression and flight envelope enhancement. Overall, this thesis establishes a robust framework for real-time aeroelastic control using machine learning–based models, enabling safer
and more efficient operation of flexible aircraft structures.
Description
Keywords
Aeroelstic, Blended Wing Body Aircraft, LSTM, DMDc