Thèse Nonlinear Reduced-Order Structural Models For High-Fidelity Partitioned Fluid-Structure Interaction Solvers H/F

Doctorat.Gouv.Fr

Offshore wind turbines are subjected to complex unsteady aerodynamic loads that induce large-amplitude vibrations of their flexible blades. Accurately predicting the vibratory response of these structures is of central importance for optimising their design and assessing their fatigue life. High-fidelity partitioned Fluid-Structure Interaction (FSI) solvers, in which a Computational Fluid Dynamics (CFD) solver is coupled to a structural solver, are capable of capturing these complex aeroelastic phenomena. However, the complexity of the implementation and production deployment of such a co-simulation remain a barrier for routine use in design loops. One way to mitigate this issue is to used a reduced order model (ROM) for the structure, therefore drastically speeding up the computations due to the possibility to use specific methodology for the data transfer and the coupling as it is done with the modal approach. Beyond computational efficiency, non-intrusive ROMs offer a further advantage: since they are independent of the structural solver at runtime, they can be directly coupled to any CFD code without requiring the integration of two complex solvers, which significantly simplifies the simulation process. However, accurate simulations require accounting for geometric nonlinearities when dealing with the slender, highly flexible blades of next-generation offshore wind turbines. This complex FSI problem therefore demands the use of a nonlinear ROM.

This project aims to develop and validate such a nonlinear ROM, to be coupled into the existing partitioned FSI solver of the LHEEA Lab. (ISISCFD), currently using a linear ROM. The scientific approach builds on recent advances in invariant manifold theory to construct nonlinear reduced bases of higher quality and larger validity domain than current state-of-the-art methods based on enriched modal bases. A key challenge is the adaptation of these nonlinear concepts, usually used in the frequency domain, to the time-domain context required by the CFD solver. The methodology will be developed first on a 2D flexible beam, then extended to a 3D model representative of an offshore wind turbine blade.

• co-supervision shared between LHEEA and GeM laboratories
• access to local and national computer resources
• experimental data for validation purposes

1. Develop a ROM for geometrically nonlinear structures based on invariant manifold theory, operating in the time domain and under arbitrary time-varying aerodynamic forcing.
2. Coupling this original ROM to the existing partitioned CFD solver of LHEEA, replacing the current linear ROM, while preserving the FSI numerical techniques and coupling algorithm.
3. Validate the resulting FSI solver against reference nonlinear FE simulations and experimental data on blade sections and lab-scale wind turbine acquired at LHEEA, demonstrating both accuracy and computational speed-up.