Learning Physics-Informed Surrogate Model of Linear Elastic Displacement Fields from Geometry

2026-07-10 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Engineering & Applied Sciences, Physical Sciences & Chemistry · Depth: Expert, quick

Summary

This work introduces a physics-informed DeepONet framework designed to create a fast and physically consistent surrogate model for real-time structural health monitoring in fractured elastic domains. The model predicts displacement fields by integrating both boundary conditions and fracture geometry, utilizing a specialized encoding strategy for the latter. A key innovation is its ability to operate without relying on finite-element-generated training data. Furthermore, the framework weakly imposes the traction-free condition on fracture boundaries through a localized penalty term. A numerical example, focusing on a single representative fracture geometry, successfully demonstrates the formulation's feasibility, establishing a foundation for future applications involving a wider range of fracture geometries.

Key takeaway

For Machine Learning Engineers developing real-time structural health monitoring, this DeepONet approach offers a path to create fast, physically consistent surrogate models. You can predict displacement fields from geometry and boundary conditions without needing extensive finite-element training data. Consider implementing localized penalty terms for weak boundary condition imposition to enhance physical consistency in your models. This method reduces computational overhead and accelerates model deployment for fractured elastic domains.

Key insights

A physics-informed DeepONet predicts elastic displacement fields from geometry and boundary conditions without FEM data, using a localized penalty for fracture conditions.

Principles

• Physics-informed models can reduce data dependency.
• Weak imposition of boundary conditions is effective.
• Geometry encoding is crucial for complex domains.

Method

The DeepONet framework encodes fracture geometry and boundary conditions to predict displacement fields. It weakly imposes traction-free conditions on fractures using a localized penalty term, avoiding finite-element training data.

In practice

• Monitor structural health in real-time.
• Develop surrogate models without FEM data.
• Extend to diverse fracture geometries.

Topics

• Physics-Informed Neural Networks
• DeepONet
• Structural Health Monitoring
• Fracture Mechanics
• Elasticity
• Surrogate Modeling

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• A Physics-Informed Neural Network Framework for Elastodynamic Wave Propagation in Bimaterial Systems — PINNs combined with explicit FEA accurately model elastodynamic wave propagation in bimaterial systems.
• Uncertainty-aware damage identification in short-span bridges via physics-informed variational autoencoder — A physics-informed Gaussian copula VAE robustly identifies bridge damage by quantifying uncertainty and ensuring physical consistency.
• Adaptive Distance-Aware Trunk Deep Operator Learning for Long-Span Roadway Bridges — AD-DeepONet accurately predicts localized bridge responses 60x faster than FEM by adaptively focusing on influence zones and using physics-informed reconstruction.
• Physics-Guided Spatiotemporal Learning for Coastal Wave Peak Period Estimation from Video — Physics-guided deep spatiotemporal learning from video can accurately and consistently estimate coastal wave peak periods, overcoming traditional monitoring limitations.
• 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.
• Latent Space Reinforcement Learning for Inverse Material Estimation in Food Fracture Simulation — Latent space reinforcement learning effectively estimates complex material parameters from desired fracture behaviors.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.