报告题目 (Title):Reduced-order modeling for uncertainty quantification of parameterized fluid-structure interaction problems
中文题目：参数化流固耦合问题不确定性量化的降阶模型
报告人 (Speaker)：孙祥 副教授，中国海洋大学
报告时间 (Time)：2025年6月13日 (周五) 10:15
报告地点 (Place)：校本部 GJ303
邀请人(Inviter)：纪丽洁
摘要：A non-intrusive reduced-order modeling (ROM) method, based on tensor-train decomposition (TTD) and polynomial chaos expansion (PCE), is proposed for parameterized fluid-structure interaction problems. TTD is used to extract the spatial, temporal, and parametric modes into TT-cores to reduce the degrees of freedom. PCE is used to approximate the parameter-dependent TT-cores by utilizing a finite set of polynomials. To validate the proposed TTD-PCE, we considered 1D Burgers’ and diffusion-reaction equations with random force terms. Compared to POD-PCE, TTD-PCE demonstrated superior performance, with eight times faster construction and two times faster prediction for a single sample. Moreover, a TTD-PCE-based uncertainty quantification (UQ) framework involving uncertainty estimation and sensitivity analysis is constructed. Subsequently, flow over a circular cylinder validated the effectiveness of the proposed method for FSI problems. Finally, a flexible filament with various conditions demonstrated the efficacy of the proposed method for UQ analysis. The results indicated a higher level of uncertainty at the free end of the self-propelled filament. Global sensitivity analysis revealed that the impact factor has different effects depending on the computational configurations. The unknown parameter of the filament was identified using TTD-PCE-based Bayesian inference, demonstrating TTD-PCE as a robust UQ framework for both calibration and parameter identification. The great potential of TTD-PCE in UQ analysis demonstrates it as a reliable tool for managing uncertainty in complex dynamical systems, providing valuable insight for inverse problems related to FSI problems.