Extended pseudo-spectral physics-informed neural networks for phase-field models

2026-06-23 · Source: Takara TLDR - Daily AI Papers · Field: Science & Research — Physical Sciences & Chemistry, Mathematics & Computational Sciences · Depth: Expert, medium

Summary

An extended pseudo-spectral physics-informed neural network (ESPINN) framework has been developed for the inverse identification of phase-field models. This framework enables the simultaneous recovery of both the bulk chemical potential and unknown gradient coefficients from transient snapshot data. Numerical experiments, specifically on the one-dimensional Cahn-Hilliard equation, demonstrated accurate and statistically stable reconstruction in noiseless conditions, even when using only a single snapshot pair. When noise was present, the reconstruction accuracy degraded gracefully, but increasing the number of snapshots significantly improved robustness by reducing variance across multiple runs. These findings establish ESPINN as a data-efficient and physically consistent approach for learning the free-energy structure within continuum models of phase separation.

Key takeaway

For research scientists developing or applying continuum models of phase separation, ESPINN offers a data-efficient method for inferring unknown constitutive quantities. If you need to identify bulk chemical potential or gradient coefficients from limited transient observations, consider integrating ESPINN. This approach maintains physical consistency and performs robustly with noisy data. Increasing available snapshots further reduces variance.

Key insights

ESPINN accurately infers phase-field model parameters and free-energy structure from limited transient data, even with noise.

Principles

• Inverse identification is feasible with limited data.
• Noise robustness improves with more snapshots.
• Physical consistency aids model learning.

Method

ESPINN uses an extended pseudo-spectral physics-informed neural network to simultaneously recover bulk chemical potential and gradient coefficients from transient snapshot data for phase-field models.

In practice

• Apply ESPINN for unknown constitutive quantity inference.
• Increase snapshot count to mitigate noise effects.

Topics

• Phase-field Models
• Physics-informed Neural Networks
• Inverse Problems
• Cahn-Hilliard Equation
• Continuum Mechanics
• Spectral Methods

Code references

• PredictiveIntelligenceLab/GradientPathologiesPINNs
• arthur-bizzi/neusa
• deeplearningmethods/PISD

Best for: AI Scientist, Research Scientist

Related on AIssential

• Differential Equations V — Physics Informed Neural Networks (PINNs) — Physics Informed Neural Networks embed physical laws into neural networks for fast, accurate, and physically consistent simulations.
• WaveGraphNet: Physics-Consistent Guided-Wave Damage Localization through Coupled Inverse-Forward Graph Learning — Coupling inverse graph localization with a physics-consistent forward model improves damage localization extrapolation in unseen regions.
• Sparse probes and murky physics: a case study of interpretability challenges in a foundation model for continuum dynamics — Interpreting scientific foundation models reveals intermittent internal structures misaligned with known physics, despite accurate outputs.
• EPINN: Enhanced Physics-Informed Neural Network for Solving Continuous Integral Equations — EPINN offers a robust, scalable solution for high-dimensional integral equations by integrating physical principles into neural networks.
• Functional-prior-based Bayesian PDE-constrained inversion using PINNs — Functional priors in physics-informed neural networks improve Bayesian PDE-constrained inversion and uncertainty quantification.
• Physics-Informed Neural Network Modeling of Biodegradable Contaminant Transport through GCL/SL Composite Liners — Hard-constrained PINNs significantly improve accuracy and stability for modeling contaminant transport in composite liners.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.