Network architecture follows coupling in multiphysics systems: single vs. multiple branches in DeepONet and S-DeepONet
5 affiliations
- 1National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign, Urbana, IL USA
- 2The Grainger College of Engineering, Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, IL USA
- 3The Grainger College of Engineering, Nuclear, Plasma & Radiological Engineering, University of Illinois at Urbana-Champaign, Urbana, IL USA
- 4Department of Industrial and Manufacturing Systems Engineering, Kansas State University, Manhattan, KS USA
- 5Civil and Urban Engineering Department, New York University Abu Dhabi, Abu Dhabi, UAE
Abstract
Real-time prediction of complex physical systems requires surrogate models that learn from data while representing strong multiphysics coupling. Deep Operator Networks have shown success in single-physics problems, yet their effectiveness in capturing nonlinear interactions in coupled systems (such as thermo-mechanical or electro-thermal coupling) remains underexplored. Here we pose a practical question: should the architecture of a neural operator reflect the strength of physical coupling it aims to model? We compare single-branch and multi-branch designs, in both feedforward and sequential recurrent forms, across three representative systems: a reaction-diffusion problem with heterogeneous sources, a nonlinear thermo-electrical problem with temperature-dependent conductivity and Joule heating, and a viscoplastic thermo-mechanical model of steel solidification. Single-branch networks consistently outperform multi-branch variants in tightly coupled regimes by encouraging shared latent representations, whereas multi-branch designs remain favorable for decoupled or single-physics tasks. Once trained, these surrogates deliver full-field predictions up to 1.8 × 104 times faster than physics-based solvers.
Related
- A multiphysics-coupled DeepONet framework for predicting sodium dendrite growth in wide-temperature sodium-ion capacitors
- QPI-DeepONet-MAC: A Scalable and Stable Hybrid Classical-Quantum Architecture for Physics-Informed Deep Operator Networks
- Ferroelectric Switching in ZnO/Zn1-xMgxO Heterostructures: Atomistic Insights into Interfacial Coupling and Layer Architecture
- Coherent and dissipative coupling of multiple photon-magnon hybrids in a chiral microwave resonator
- Interest of Multiphysics and Multilevel Simulation Approaches to Enhance the Machining Process
BibTeX
@article{park2026network,
title = {{Network architecture follows coupling in multiphysics systems: single vs. multiple branches in DeepONet and S-DeepONet}},
author = {Jaewan Park and Kazuma Kobayashi and Qibang Liu and Diab Abueidda and Seid Koric and Syed Bahauddin Alam},
journal = {Communications Engineering},
year = {2026},
month = jul,
doi = {10.1038/s44172-026-00714-4},
url = {https://doi.org/10.1038/s44172-026-00714-4},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Network architecture follows coupling in multiphysics systems: single vs. multiple branches in DeepONet and S-DeepONet
AU - Jaewan Park
AU - Kazuma Kobayashi
AU - Qibang Liu
AU - Diab Abueidda
AU - Seid Koric
AU - Syed Bahauddin Alam
PY - 2026
DA - 2026/07/03
JO - Communications Engineering
DO - 10.1038/s44172-026-00714-4
UR - https://doi.org/10.1038/s44172-026-00714-4
AB - Real-time prediction of complex physical systems requires surrogate models that learn from data while representing strong multiphysics coupling. Deep Operator Networks have shown success in single-physics problems, yet their effectiveness in capturing nonlinear interactions in coupled systems (such as thermo-mechanical or electro-thermal coupling) remains underexplored. Here we pose a practical question: should the architecture of a neural operator reflect the strength of physical coupling it aims to model? We compare single-branch and multi-branch designs, in both feedforward and sequential recurrent forms, across three representative systems: a reaction-diffusion problem with heterogeneous sources, a nonlinear thermo-electrical problem with temperature-dependent conductivity and Joule heating, and a viscoplastic thermo-mechanical model of steel solidification. Single-branch networks consistently outperform multi-branch variants in tightly coupled regimes by encouraging shared latent representations, whereas multi-branch designs remain favorable for decoupled or single-physics tasks. Once trained, these surrogates deliver full-field predictions up to 1.8 × 104 times faster than physics-based solvers.
SN - 2731-3395
ER -