New AI Method Discovers Material Models from Full-Field Data

Researchers have developed CANN-EUCLID, a novel unsupervised method for discovering constitutive artificial neural network models directly from full-field data. This approach combines Constitutive Artificial Neural Networks (CANNs) with the EUCLID framework, enabling the identification of sparse hyperelastic laws without requiring local stress measurements or a predefined model structure. The method was evaluated on isotropic and anisotropic benchmarks, demonstrating its ability to accurately recover ground-truth laws when representable by the chosen CANN basis, and to approximate missing contributions otherwise. This framework shows promise for interpretable full-field constitutive model identification, particularly for complex materials like soft biological tissues where traditional methods face limitations. AI

IMPACT This research introduces a novel AI-driven approach for discovering material constitutive models, potentially accelerating research in material science and engineering by enabling more accurate and interpretable model identification from experimental data.

RANK_REASON The cluster contains a research paper detailing a new AI methodology for material model discovery.

Read on arXiv cs.LG →

• arXiv
• CANN-EUCLID
• Constitutive artificial neural networks
• EUCLID
• Mathias Peirlinck

• paper
• other

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →