Speaker
Description
Parametric reduced-order models (ROMs) for linear structural dynamics commonly focus on time-invariant parametric settings. However, when parameters vary in time, the governing operators become time-dependent and ROM accuracy must be ensured over a family of admissible parameter trajectories rather than a bounded parameter domain. Direct sampling of this trajectory space is computationally prohibitive, and approximating the solution manifold using a single global basis often leads to an impractically high-dimensional reduced subspace. We therefore adopt a regime-based ROM library approach, where each local ROM is constructed by computing a proper orthogonal decomposition (POD) basis from full-order snapshots at selected constant parameter values and applying a Galerkin projection onto the resulting reduced subspace. The parameter values are chosen using a low-cost library refinement indicator based on discrepancies between adjacent local ROMs. In the online phase, the model selects the appropriate local ROM according to the active parameter regime and a mass-orthogonal projection transfers the reduced states across subspaces at regime transitions. The method is demonstrated on a Kirchhoff-Love plate with a time-varying boundary stiffness parameter, which induces a time-dependent stiffness matrix affinely. Results show that this method can achieve sufficient accuracy with substantially smaller reduced dimensions than a single global ROM.