Learning Subsystem Dynamics in Nonlinear Systems via Port-Hamiltonian Neural Networks

Authors

G.J.E. Van Otterdijk, S. Moradi, S. Weiland, R. Tóth, N.O. Jaensson, M. Schoukens

Abstract

Port-Hamiltonian neural networks (pHNNs) are emerging as a powerful modeling tool that integrates physical laws with deep learning techniques. While most research has focused on modeling the entire dynamics of interconnected systems, the potential for identifying and modeling individual subsystems while operating as part of a larger system has been overlooked. This study addresses this gap by introducing a novel method for using pHNNs to identify such subsystems based solely on input-output measurements. By utilizing the inherent compositional property of the port-Hamiltonian systems, we developed an algorithm that learns the dynamics of individual subsystems, without requiring direct access to their internal states. On top of that, by choosing an output error (OE) model structure, we have been able to handle measurement noise effectively. The efficiency of the proposed approach is demonstrated through tests on interconnected systems, including multi-physics scenarios, highlighting its potential for identifying subsystem dynamics and facilitating their integration into new interconnected models.

Citation

• Journal: 2025 IEEE 64th Conference on Decision and Control (CDC)
• Year: 2025
• Volume:
• Issue:
• Pages: 2071–2076
• Publisher: IEEE
• DOI: 10.1109/cdc57313.2025.11312257

BibTeX

@inproceedings{Van_Otterdijk_2025,
  title={{Learning Subsystem Dynamics in Nonlinear Systems via Port-Hamiltonian Neural Networks}},
  DOI={10.1109/cdc57313.2025.11312257},
  booktitle={{2025 IEEE 64th Conference on Decision and Control (CDC)}},
  publisher={IEEE},
  author={Van Otterdijk, G.J.E. and Moradi, S. and Weiland, S. and Tóth, R. and Jaensson, N.O. and Schoukens, M.},
  year={2025},
  pages={2071--2076}
}

Download the bib file

References

• Nelles O (2020) Nonlinear Dynamic System Identification. Nonlinear System Identification 831–89 – 10.1007/978-3-030-47439-3_19
• Greydanus, Hamiltonian neural networks. Advances in neural information processing systems (2019)
• Bertalan T, Dietrich F, Mezić I, Kevrekidis IG (2019) On learning Hamiltonian systems from data. Chaos: An Interdisciplinary Journal of Nonlinear Science 29(12). https://doi.org/10.1063/1.512823 – 10.1063/1.5128231
• Mattheakis M, Sondak D, Dogra AS, Protopapas P (2022) Hamiltonian neural networks for solving equations of motion. Phys Rev E 105(6). https://doi.org/10.1103/physreve.105.06530 – 10.1103/physreve.105.065305
• Moradi S, Jaensson N, Tóth R, Schoukens M (2023) Physics-Informed Learning Using Hamiltonian Neural Networks with Output Error Noise Models. IFAC-PapersOnLine 56(2):5152–5157. https://doi.org/10.1016/j.ifacol.2023.10.10 – 10.1016/j.ifacol.2023.10.108
• Cherifi, Numerical methods to compute a minimal realization of a port-hamiltonian system. (2019)
• Benner P, Goyal P, Van Dooren P (2020) Identification of port-Hamiltonian systems from frequency response data. Systems & Control Letters 143:104741. https://doi.org/10.1016/j.sysconle.2020.10474 – 10.1016/j.sysconle.2020.104741
• Control 1(2–3):173–378. https://doi.org/10.1561/260000000 – 10.1561/2600000002
• Duindam V, Macchelli A, Stramigioli S, Bruyninckx H (2009) Modeling and Control of Complex Physical Systems. Springer Berlin Heidelber – 10.1007/978-3-642-03196-0
• van der Schaft A, Jeltsema D (2021) On Energy Conversion in Port-Hamiltonian Systems. 2021 60th IEEE Conference on Decision and Control (CDC) 2421–242 – 10.1109/cdc45484.2021.9683292
• Desai SA, Mattheakis M, Sondak D, Protopapas P, Roberts SJ (2021) Port-Hamiltonian neural networks for learning explicit time-dependent dynamical systems. Phys Rev E 104(3). https://doi.org/10.1103/physreve.104.03431 – 10.1103/physreve.104.034312
• Cheng X, Shi S, Lestas I, Hof PMJV den (2023) A Necessary Condition for Network Identifiability With Partial Excitation and Measurement. IEEE Trans Automat Contr 68(11):6820–6827. https://doi.org/10.1109/tac.2023.323982 – 10.1109/tac.2023.3239829
• Van den Hof PMJ, Ramaswamy KR, Fonken SJM (2023) Integrating data-informativity conditions in predictor models for single module identification in dynamic networks. IFAC-PapersOnLine 56(2):2377–2382. https://doi.org/10.1016/j.ifacol.2023.10.121 – 10.1016/j.ifacol.2023.10.1210
• Kivits EMM, Van den Hof PMJ (2023) Identification of Diffusively Coupled Linear Networks Through Structured Polynomial Models. IEEE Trans Automat Contr 68(6):3513–3528. https://doi.org/10.1109/tac.2022.319140 – 10.1109/tac.2022.3191406
• Classens K, Schoukens M, Oomen T, Noël J-P (2025) Locating nonlinearities in mechanical systems: A frequency-domain dynamic network perspective. Mechanical Systems and Signal Processing 224:112124. https://doi.org/10.1016/j.ymssp.2024.11212 – 10.1016/j.ymssp.2024.112124
• Fonken SJM, Ramaswamy KR, Van Den Hof PMJ (2023) Local Identification in Dynamic Networks using a Multi-Step Least Squares Method. 2023 62nd IEEE Conference on Decision and Control (CDC) 431–43 – 10.1109/cdc49753.2023.10384234
• Lizan Kivits EMM, Van den Hof PMJ (2022) Local identification in diffusively coupled linear networks. 2022 IEEE 61st Conference on Decision and Control (CDC) 874–87 – 10.1109/cdc51059.2022.9993111
• Moradi, Port-hamiltonian neural networks with output error noise models. (2025)
• Khosrovian, Port-hamiltonian neural networks for learning coupled systems and their interactions.
• Neary, Compositional learning of dynamical system models using port-hamiltonian neural networks. Proc. of the Learning for Dynamics and Control Conference
• Schaft A van der (2021) Port-Hamiltonian Systems: From Modeling to Control. Encyclopedia of Systems and Control 1753–175 – 10.1007/978-3-030-44184-5_100073
• Beintema, Continuous-time identification of dynamic state-space models by deep subspace encoding. Proc. of the International Conference on Learning Representations
• Cockburn B, Shu C-W (2001) Runge–Kutta Discontinuous Galerkin Methods for Convection-Dominated Problems. Journal of Scientific Computing 16(3):173–261. https://doi.org/10.1023/a:101287391088 – 10.1023/a:1012873910884
• Beintema GI, Schoukens M, Tóth R (2023) Deep subspace encoders for nonlinear system identification. Automatica 156:111210. https://doi.org/10.1016/j.automatica.2023.11121 – 10.1016/j.automatica.2023.111210
• Kivits, Modelling and identification of physical linear networks. PhD thesis (2024)
• Kivits EMM (Lizan), Van den Hof PMJ (2023) Identifiability of diffusively coupled linear networks with partial instrumentation*. IFAC-PapersOnLine 56(2):2395–2400. https://doi.org/10.1016/j.ifacol.2023.10.121 – 10.1016/j.ifacol.2023.10.1213
• Kingma, Adam: A method for stochastic optimization. Proc. of the International Conference on Learning Representations