MSc Dissertation: AI Surrogate Modeling for Turbulent Flow Simulations

Discovered a novel Grid-Invariant AI architecture combining convolutional autoencoders and adversarial networks to simulate high-fidelity turbulent flows, achieving unprecedented grid independence and scalability.

Key Achievements

• Discovered a novel Grid-Invariant AI architecture combining convolutional autoencoders and adversarial networks to simulate high-fidelity turbulent flows, achieving unprecedented grid independence and scalability
• Conducted 1000+ GPU hours of High-Performance Computing (HPC) for model optimization
• Enhanced long-term stability by 35% and prediction accuracy by 50%
• Research leading to forthcoming publications

Technical Details

• Techniques: Grid-Invariant AI, Convolutional Autoencoders, Adversarial Networks
• Computational Resources: High-Performance Computing (HPC)
• Performance Metrics:
  • 35% improvement in long-term stability
  • 50% increase in prediction accuracy
  • 1000+ GPU hours of optimization