Graph Neural Networks accurately model thermoplastic composite mechanics

Researchers have developed a novel data-driven surrogate framework to predict the mechanical behavior of additively manufactured short-fiber thermoplastic composites. This framework utilizes a hybrid Graph Neural Network-Long Short-Term Memory (GNN-LSTM) architecture, trained on data from microstructures reconstructed from micro-computed tomography. The model accurately predicts stiffness and stress-strain behavior, achieving a high R-squared value of approximately 0.98 compared to high-fidelity simulations, while reducing computational cost by over two orders of magnitude. This approach offers a physics-informed and data-efficient method for identifying mechanically weak areas in components and accelerating digital-twin development. AI

IMPACT This research demonstrates a novel application of GNN-LSTM for simulating complex material behaviors, potentially accelerating design and analysis in fields like aerospace and automotive engineering.

RANK_REASON Academic paper detailing a new methodology for modeling material properties. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• Graph Neural Networks
• Long Short-Term Memory
• Pharindra Pathak

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →