Affiliation:
1. Institute for Numerical Analysis TU Braunschweig Braunschweig Germany
Abstract
AbstractPort‐Hamiltonian neural networks can be used to capture the interactions between small interconnected port‐Hamiltonian systems from data. We now examine different composed models and check how accurate we can train the modified composed systems. The learning of the dynamics of the composed systems is investigated for cases in which the interconnection is known beforehand and also for cases without knowledge about the interconnection. In the latter cases, only small additional data is required.
Reference6 articles.
1. Neary C. &Topcu U.(2023).Compositional learning of dynamical system models using port‐Hamiltonian neural networks.Proceedings of The 5th Annual Learning for Dynamics and Control Conference.PMLR (211) 679–691.
2. Modeling and Control of Complex Physical Systems
3. MORpH: Model reduction of linear port-Hamiltonian systems in MATLAB
4. Control of port-Hamiltonian differential-algebraic systems and applications
5. Port‐controlled Hamiltonian systems: Towards a theory for control and design of nonlinear physical systems;Van der Schaft A.;Journal of The Society of Instrument and Control Engineers,2000