Thermodynamically Consistent Machine Learning for Fracture Prediction in Heterogeneous Materials

Abstract

Understanding fracture behavior is critical for assessing the structural integrity and longevity. Highly time-consuming physics-based fracture simulations motivate the development of machine-learning-based fracture evolution prediction with surrogate models. However, most existing machine-learning models for fracture prediction operate as black boxes and require large training datasets to generalize reliably. In this work, Thermodynamically Consistent Latent Dynamics Identification (tLaSDI) for parametric systems is introduced to predict fracture patterns and evolution in heterogeneous materials. The model learns the thermodynamic state variable in a lower-dimensional latent space using parametric GENERIC Formalism-Informed Neural Networks (pGFINNs) that enforce key thermodynamic principles, such as total energy conservation and entropy generation. This model embeds physics into the neural network architecture using a hard-constraint method that inherently complies with the underlying thermodynamic principles. The proposed tLaSDI-pGFINNs is further enhanced by a Geometry-Conditioned Autoencoder in learning the geometry configuration for effective geometry-guided learning. The energy and entropy evolution histories demonstrate that the model preserves the thermodynamic constraints on unseen testing samples. The results further show that tLaSDI-pGFINNs can more accurately capture both fracture patterns and the subsequent fracture evolution than the purely data-driven Autoencoder Long Short-Term Memory (AE-LSTM) model.

A thermodynamically consistent and conservative phase-field model for surfactant contact line dynamics with solid adsorption
Jiangxu Huang, Zhenhua Chai, Xi Liu +1
arXiv22 Sep 2026PreprintarXiv licenceFull textFluid dynamicsarXiv
A thermodynamically consistent framework for finite growth of multi-constituent mixtures with application to tumor growth
Jonathan Stollberg, Marco F. P. ten Eikelder, Dominik Schillinger
arXiv8 Sep 2026PreprintCC BY-NC-SA 4.0Full textPhysiology Biological & medical physicsarXiv
Thermodynamically consistent thixotropic viscoelastic modeling of human blood from a conformational free energy
Thomas Martinod, Andrés Córdoba, Jay D. Schieber
arXiv2 Nov 2026PreprintCC BY-NC-SA 4.0Full textCondensed matterarXiv
Learning shared hyperelastic constitutive representation from heterogeneous full-field experiments across multiple materials
Lucas Mabileau, Maxence Barberet-Pinto, Thomas Ferchal +9
Computer Methods in Applied Mechanics and Engineering30 Oct 2026Peer-reviewedCC BY 4.0Full textGeneral physicsHAL
A thermodynamically consistent theory of multicomponent porous media: finite-deformation mechanics, compositional flow, electrostatics, reactions, and phase transformations
John T. Foster
EarthArXiv10 Oct 2026PreprintCC BY 4.0Full textGeophysics & seismology Hydrology & soilsEarthArXiv

Cite

BibTeX RIS

BibTeX

@misc{alim2026thermodynamically,
  title = {{Thermodynamically Consistent Machine Learning for Fracture Prediction in Heterogeneous Materials}},
  author = {Muhamad Alim and Xiaolong He and Mei-Na Tsao and Chi-Hua Yu and J. S. Chen},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.20614},
  eprint = {2610.20614},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.20614},
}

RIS

TY  - UNPB
TI  - Thermodynamically Consistent Machine Learning for Fracture Prediction in Heterogeneous Materials
AU  - Muhamad Alim
AU  - Xiaolong He
AU  - Mei-Na Tsao
AU  - Chi-Hua Yu
AU  - J. S. Chen
PY  - 2026
DA  - 2026/10/01
PB  - arXiv
DO  - 10.48550/arxiv.2610.20614
UR  - https://arxiv.org/abs/2610.20614
AB  - Understanding fracture behavior is critical for assessing the structural integrity and longevity. Highly time-consuming physics-based fracture simulations motivate the development of machine-learning-based fracture evolution prediction with surrogate models. However, most existing machine-learning models for fracture prediction operate as black boxes and require large training datasets to generalize reliably. In this work, Thermodynamically Consistent Latent Dynamics Identification (tLaSDI) for parametric systems is introduced to predict fracture patterns and evolution in heterogeneous materials. The model learns the thermodynamic state variable in a lower-dimensional latent space using parametric GENERIC Formalism-Informed Neural Networks (pGFINNs) that enforce key thermodynamic principles, such as total energy conservation and entropy generation. This model embeds physics into the neural network architecture using a hard-constraint method that inherently complies with the underlying thermodynamic principles. The proposed tLaSDI-pGFINNs is further enhanced by a Geometry-Conditioned Autoencoder in learning the geometry configuration for effective geometry-guided learning. The energy and entropy evolution histories demonstrate that the model preserves the thermodynamic constraints on unseen testing samples. The results further show that tLaSDI-pGFINNs can more accurately capture both fracture patterns and the subsequent fracture evolution than the purely data-driven Autoencoder Long Short-Term Memory (AE-LSTM) model.
ER  -

Abstract

Related

Cite