Neural-Network-based Viscosity Closure for Non-Newtonian Multiphase Flows
Abstract
Materials used in polymer-based additive manufacturing processes, such as Digital Light Processing (DLP) and direct ink writing (DIW), typically exhibit non-Newtonian rheology. Carreau-Yasuda and power-law models describe basic shear-thinning and shear-thickening behavior well, but applying them to a new material requires choosing a functional form, deriving it, and re-implementing it inside the flow solver. We present a deployment workflow in which a neural network trained on experimental rheometry data serves as the viscosity closure inside a Cahn-Hilliard-Navier-Stokes (CHNS) finite element solver. The learned closure is generalized Newtonian: viscosity depends only on the local shear rate at the current instant. Lipschitz regularization during training produces smooth viscosity predictions, and the trained network is exported in the Open Neural Network Exchange (ONNX) format and queried by the solver at runtime via the ONNX runtime, without solver modification or network reimplementation. The framework is built on a parallel octree-based adaptive mesh refinement infrastructure that concentrates resolution at the fluid interface. We validate the CHNS solver against benchmark shear-thinning bubble-rise cases from the literature, reproducing reported bubble shapes across varying power-law indices and Weber numbers. We characterized two silicone ink formulations, recorded their rise dynamics in perfluorodecalin on high-speed video, and used the resulting data to test the full workflow. Simulated rise velocities fall within the experimentally measured spread, and the simulated steady-state droplet shape agrees with the observed one. This work contributes to a growing body of literature on integrating neural constitutive closures into multiphysics simulations, and demonstrates a practical path for deploying experimentally trained rheological surrogates inside finite element solvers.
Related
- Load-Balanced Neural-Network-Aided Adaptive Tabulation of Phase Equilibria for Transcritical Multiphase Flows
- Sharp-interface Simulations of Energetic Multiphase Flows with Large Density and Viscosity Ratios
- A hydraulic constitutive closure for reduced haemodynamic models with generalised-Newtonian rheology
- Online Learning of a Spectral Eddy Viscosity Closure for Large Eddy Simulation Using Ensemble Kalman Inversion
- Gravity-Informed Neural Networks for Post-Newtonian Binary Dynamics
BibTeX
@misc{murugaiyan2026neuralnetworkbased,
title = {{Neural-Network-based Viscosity Closure for Non-Newtonian Multiphase Flows}},
author = {Suresh Murugaiyan and Claire L. Nelson and Dhruv Gamdha and Austin Cunniff and Cheng-Hau Yang and Abraham Wiletsky and Kaitlyn W. Dilley and Patrick Babb and Andrew Rhode and Christopher M. Bates and Angela A. Pitenis and Michael L. Chabinyc and Adarsh Krishnamurthy and Baskar Ganapathysubramanian},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2605.30659},
eprint = {2605.30659},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.30659},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Neural-Network-based Viscosity Closure for Non-Newtonian Multiphase Flows
AU - Suresh Murugaiyan
AU - Claire L. Nelson
AU - Dhruv Gamdha
AU - Austin Cunniff
AU - Cheng-Hau Yang
AU - Abraham Wiletsky
AU - Kaitlyn W. Dilley
AU - Patrick Babb
AU - Andrew Rhode
AU - Christopher M. Bates
AU - Angela A. Pitenis
AU - Michael L. Chabinyc
AU - Adarsh Krishnamurthy
AU - Baskar Ganapathysubramanian
PY - 2026
DA - 2026/09/08
PB - arXiv
DO - 10.48550/arxiv.2605.30659
UR - https://arxiv.org/abs/2605.30659
AB - Materials used in polymer-based additive manufacturing processes, such as Digital Light Processing (DLP) and direct ink writing (DIW), typically exhibit non-Newtonian rheology. Carreau-Yasuda and power-law models describe basic shear-thinning and shear-thickening behavior well, but applying them to a new material requires choosing a functional form, deriving it, and re-implementing it inside the flow solver. We present a deployment workflow in which a neural network trained on experimental rheometry data serves as the viscosity closure inside a Cahn-Hilliard-Navier-Stokes (CHNS) finite element solver. The learned closure is generalized Newtonian: viscosity depends only on the local shear rate at the current instant. Lipschitz regularization during training produces smooth viscosity predictions, and the trained network is exported in the Open Neural Network Exchange (ONNX) format and queried by the solver at runtime via the ONNX runtime, without solver modification or network reimplementation. The framework is built on a parallel octree-based adaptive mesh refinement infrastructure that concentrates resolution at the fluid interface. We validate the CHNS solver against benchmark shear-thinning bubble-rise cases from the literature, reproducing reported bubble shapes across varying power-law indices and Weber numbers. We characterized two silicone ink formulations, recorded their rise dynamics in perfluorodecalin on high-speed video, and used the resulting data to test the full workflow. Simulated rise velocities fall within the experimentally measured spread, and the simulated steady-state droplet shape agrees with the observed one. This work contributes to a growing body of literature on integrating neural constitutive closures into multiphysics simulations, and demonstrates a practical path for deploying experimentally trained rheological surrogates inside finite element solvers.
ER -