Multi-Fidelity Machine Learning for Predicting Startup Stress Growth and Strain Hardening in Wormlike Micellar Solutions
Abstract
Wormlike micellar solutions exhibit formulation-dependent nonlinear startup responses, including stress overshoot, undershoot, and strain hardening, making prediction across compositions and shear rates challenging. We developed a multi-fidelity machine-learning framework to predict stress-growth curves in cetylpyridinium chloride (CPyCl)-based wormlike micelles from sparse experimental measurements. The framework combines a low-fidelity analytical model describing the overall startup response with high-fidelity experimental data capturing formulation-specific nonlinear features. This approach addresses the limited and uneven coverage of salt chemistry, concentration, and shear rate without relying solely on data-intensive neural-network training. The trained model predicts complete stress-growth responses at untested shear rates and enables extraction of strain-hardening characteristics from interpolated curves. By augmenting sparse rheological measurements, the framework provides a basis for exploring relationships among formulation variables, linear viscoelastic properties, and nonlinear startup behavior, while supporting more efficient characterization of wormlike micellar solutions.
Related
- Passage of particles through matter and the effective straggling-function: High-fidelity accelerated simulation via Physics-Informed Machine Learning
- A unified description of flow-induced scission of wormlike micelles under shear and extensional flows
- Probing Flow-Induced Chain Scission in Shear-Banding Wormlike Micelles
- An exact density-functional closure for the two-dimensional discrete wormlike chain
- Quantum Representation Learning Beyond Pairwise Fidelity
BibTeX
@misc{zolfaghari2026multifidelity,
title = {{Multi-Fidelity Machine Learning for Predicting Startup Stress Growth and Strain Hardening in Wormlike Micellar Solutions}},
author = {Saghar Zolfaghari and Nunzia Lauriello and Ilaria Cusano and Marcello Ferranti and Rossana Pasquino and Safa Jamali},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.04115},
eprint = {2611.04115},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.04115},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Multi-Fidelity Machine Learning for Predicting Startup Stress Growth and Strain Hardening in Wormlike Micellar Solutions
AU - Saghar Zolfaghari
AU - Nunzia Lauriello
AU - Ilaria Cusano
AU - Marcello Ferranti
AU - Rossana Pasquino
AU - Safa Jamali
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.04115
UR - https://arxiv.org/abs/2611.04115
AB - Wormlike micellar solutions exhibit formulation-dependent nonlinear startup responses, including stress overshoot, undershoot, and strain hardening, making prediction across compositions and shear rates challenging. We developed a multi-fidelity machine-learning framework to predict stress-growth curves in cetylpyridinium chloride (CPyCl)-based wormlike micelles from sparse experimental measurements. The framework combines a low-fidelity analytical model describing the overall startup response with high-fidelity experimental data capturing formulation-specific nonlinear features. This approach addresses the limited and uneven coverage of salt chemistry, concentration, and shear rate without relying solely on data-intensive neural-network training. The trained model predicts complete stress-growth responses at untested shear rates and enables extraction of strain-hardening characteristics from interpolated curves. By augmenting sparse rheological measurements, the framework provides a basis for exploring relationships among formulation variables, linear viscoelastic properties, and nonlinear startup behavior, while supporting more efficient characterization of wormlike micellar solutions.
ER -