AI for Computational Mechanics


Data-driven and physics-informed modeling for accelerated simulation

This research area integrates machine learning with computational mechanics to accelerate prediction, optimization, and design exploration. We develop hybrid pipelines that combine physics-based simulation data with surrogate models, graph-based learning, and physics-informed neural networks for high-dimensional engineering systems.

Research Topics

Physics-Informed & Surrogate Modeling

  • Physics-Informed Neural Networks (PINNs)
  • Gaussian Process and response surface models
  • Reduced-order and surrogate-assisted simulation

Graph Neural Networks for Physical Systems

  • Pore-network and microstructure graph representation
  • GNN-based effective property prediction
  • Contact-network learning in particle systems
  • Message-passing models for transport and percolation

Hybrid Simulation + AI Pipelines

  • AI-assisted microstructure optimization
  • Multi-objective design under physical constraints
  • Data-driven parameter inference and calibration

Digital Twin & Intelligent Engineering

  • Simulation-driven digital twin frameworks
  • Real-time prediction and control integration
  • AI-enhanced engineering decision systems

AI and FVM GNN