Thesis
A robust data-driven Bayesian approach for complex nonlinear aeroelastic system identification
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2026
- Thesis identifier
- T17595
- Person Identifier (Local)
- 202183247
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- This thesis presents a computational framework for the analysis of nonlinear aeroelastic phenomena, with particular emphasis on sustained oscillatory behaviour that poses significant challenges in aerospace design. Traditional time-domain methods often prove computationally intensive and limited in scope. To address these challenges, this research introduces innovative frequency-domain and data-driven techniques that substantially reduce computational demands while ensuring robust treatment of uncertainty. The proposed methodology integrates three core developments: a harmonic balance approach enhanced by operator-based stability analysis, a surrogate modelling strategy informed by probabilistic inference, and a hierarchical learning architecture that leverages models of varying accuracy. These innovations enable the efficient exploration of complex system dynamics without reliance on extensive time-domain simulations, while accurately capturing critical features such as limit cycle oscillations and stability transitions. A probabilistic framework is employed for the estimation of system parameters and model structures, grounded in statistical evidence. This allows for the construction of bifurcation diagrams augmented with an interval of probability, offering new capabilities for visualising uncertainty in nonlinear dynamic systems. To overcome the limitations imposed by sparse experimental data, a novel learning architecture is introduced that efficiently synthesises information across models of differing resolution. This approach effectively captures both inherent variability and knowledge-based uncertainty, enabling accurate predictions with significantly reduced data requirements. The framework is demonstrated on a representative aeroelastic configuration, where it performs robustly under both idealised and noisy conditions. It delivers high predictive fidelity in identifying critical dynamic transitions, while alternative strategies within the framework offer enhanced computational efficiency with controlled trade-offs in accuracy. A notable innovation includes the use of normalised continuation parameters, which facilitate the tracking of intricate nonlinear behaviours across multiple solution branches. Overall, this work achieves substantial gains in computational efficiency while preserving, and in some cases improving, predictive performance. It supports a wide range of practical applications, from early-stage aeroelastic design to real-time system monitoring, and contributes novel theoretical insights through the integration of advanced continuation methods, hierarchical uncertainty quantification, and Bayesian system identification. This thesis thus establishes a versatile foundation for the analysis of nonlinear aeroelastic systems, bridging the gap between computational tractability and modelling accuracy. Its modular and extensible architecture positions it for broader application to related phenomena such as flutter and gust response, and provides a pathway towards future innovations in digital twin technologies and certification strategies for next-generation aerospace systems.
- Advisor / supervisor
- Feng, Jinglang
- Yuan, Jie
- Resource Type
- DOI
Relations
Items
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
PDF of thesis T17595 | 2026-02-09 | Public | Download |