Physics-Informed Neural Network Frameworks for the Analysis of Engineering and Biological Dynamical Systems Governed by Ordinary Differential Equations
American Journal of Undergraduate Research · 27.06.2026 · 23 (3) · 41-52
DOI: 10.33697/ajur.2026.175 ↗Kurz gefasst
Fachartikel in der Zeitschrift American Journal of Undergraduate Research, online erschienen am 27.06.2026. Verfasst von Andrew Particka und 5 weiteren Autor:innen.
Abstract (Originalsprache)
This study presents and validates the predictive capability of the Physics-Informed Neural Networks (PINNs) methodology for solving a variety of engineering and biological dynamical systems governed by ordinary differential equations (ODEs). While traditional numerical methods are effective for many ODEs, they often struggle to achieve convergence in problems involving high stiffness, shocks, irregular domains, singular perturbations, high dimensions, or boundary discontinuities. Alternatively, PINNs offer a powerful approach for handling challenging numerical scenarios. In this study, classical ODE problems are employed as controlled testbeds to systematically evaluate the accuracy, training efficiency, and generalization capability of the PINNs framework under controlled conditions. Although not a universal solution, PINNs can achieve superior results by embedding physical laws directly into the learning process. The existence and uniqueness properties of several benchmark problems are first analyzed, and the PINNs methodology is subsequently validated on each model system. The results demonstrate that, for complex problems to converge to correct solutions, the loss function components data loss, initial condition loss, and residual loss must be appropriately balanced through careful weighting. It was further established that systematic tuning of hyperparameters including network depth, layer width, activation functions, learning rate, optimization algorithms, weight initialization schemes, and collocation point sampling plays a crucial role in achieving accurate solutions. Additionally, embedding prior knowledge and imposing hard constraints on the network architecture, without losing the generality of the ODE system, significantly enhances the predictive capability of PINNs. KEYWORDS: Physics-Informed Neural Networks; Ordinary Differential Equations; DeepXDE; Dynamical Systems; Activation functions; Adams Algorithm; Picard–Lindelöf Theorem; Grönwall’s Inequality
Zitieren
APA
Particka, A., Whitman, T., Diers, C., Griffin, I., Wickramasinghe, C., & Ranaweera, P. (2026). Physics-Informed Neural Network Frameworks for the Analysis of Engineering and Biological Dynamical Systems Governed by Ordinary Differential Equations. American Journal of Undergraduate Research, 23(3), 41-52. https://doi.org/10.33697/ajur.2026.175
Vancouver
Particka A, Whitman T, Diers C, Griffin I, Wickramasinghe C, Ranaweera P. Physics-Informed Neural Network Frameworks for the Analysis of Engineering and Biological Dynamical Systems Governed by Ordinary Differential Equations. American Journal of Undergraduate Research. 2026;23(3):41-52. doi:10.33697/ajur.2026.175
BibTeX
@article{Particka2026,
author = {Particka, Andrew and Whitman, Tyrus and Diers, Christopher and Griffin, Ian and Wickramasinghe, Charuka and Ranaweera, Pradeep},
title = {Physics-Informed Neural Network Frameworks for the Analysis of Engineering and Biological Dynamical Systems Governed by Ordinary Differential Equations},
journal = {American Journal of Undergraduate Research},
year = {2026},
volume = {23},
number = {3},
pages = {41--52},
doi = {10.33697/ajur.2026.175}
}
EUVIMED