Numerical dispersed flow simulation of fire-flake particle dynamics and its learning representation.

Area of Science:

• Computational fluid dynamics
• Artificial intelligence
• Particle simulation

Background:

• Simulating fire-flake particle dynamics is complex due to their chaotic nature and sensitivity to flame movement.
• Existing methods struggle to accurately capture the intricate interactions between particles and flame.

Purpose of the Study:

• To develop efficient methods for simulating detailed fire-flake particle flow.
• To create a synthetic dataset of fire-flake particle movement.
• To extend simulation capabilities using AI-driven approaches.

Main Methods:

• Utilized chaotic advection and controlled buoyant flow techniques for particle simulation.
• Developed a stochastic solver for subgrid interactions during particle advection.
• Employed neural networks to learn and represent fire-flake particle motion.

Main Results:

• Successfully generated a synthetic dataset of detailed fire-flake particle movement.
• Achieved efficient simulation of particles in chaotic regions with enhanced detail.
• Demonstrated stable particle representation without extensive parameter tuning.

Conclusions:

• The proposed methods offer a significant improvement over traditional random walk approaches.
• The integration of AI enables more sophisticated and adaptable fire-flake particle simulations.
• This work provides a robust framework for realistic simulation of complex particle dynamics.