Keskeiset käsitteet
The core message of this work is to develop a nonintrusive data-driven method, called Lagrangian operator inference (L-OpInf), that can learn reduced-order models (ROMs) of large-scale Lagrangian dynamical systems while preserving the underlying Lagrangian structure.
Tiivistelmä
The key highlights and insights of this work are:
The authors develop a nonintrusive physics-preserving method called Lagrangian operator inference (L-OpInf) to learn ROMs of large-scale Lagrangian systems, including nonlinear wave equations and mechanical systems with external nonconservative forcing.
The proposed L-OpInf method exploits knowledge about the space-time continuous Lagrangian at the PDE level to define and parametrize a Lagrangian ROM form, which is then learned from trajectory data via a constrained linear least-squares problem.
Numerical results demonstrate that the learned Lagrangian ROMs can accurately predict the physical solutions both far outside the training time interval and for unseen initial conditions, unlike standard operator inference methods that violate the underlying Lagrangian structure.
The authors learn Lagrangian ROMs for a high-dimensional soft-robotic fishtail model with dissipation and time-dependent control input, showcasing the versatility and robustness of the proposed method to unknown control inputs.
Unlike structure-preserving Hamiltonian approaches that require both trajectory and momentum data, the presented L-OpInf method can learn accurate and stable ROMs with bounded energy error from high-dimensional trajectory data alone.
Tilastot
The authors do not provide any specific numerical data or metrics in the content. The focus is on the methodology and numerical demonstrations.
Lainaukset
There are no direct quotes from the content that are relevant to the key logics.