Derivative particles for simulating detailed movements of fluids.

Area of Science:

• Computational fluid dynamics
• Numerical simulation
• Physics-based animation

Background:

• Conventional Navier-Stokes solvers suffer from numerical dissipation, particularly during the advection step.
• Simulating high-Reynolds-number fluid flows requires methods that minimize nonphysical velocity dissipation.
• Accurate advection is crucial for realistic fluid behavior in simulations.

Purpose of the Study:

• To develop a fluid simulation technique that significantly reduces nonphysical velocity dissipation.
• To improve the accuracy of simulating detailed fluid movements, such as droplets and whirlpools.
• To enhance the performance of large-scale fluid simulations.

Main Methods:

• Hybrid approach combining grid-based and particle-based fluid simulation.
• Replacing the advection step in grid-based solvers with a particle simulator for detailed fluid regions.
• Developing a novel dissipation-suppressing conversion procedure using derivative information for inter-simulator data transfer.
• Utilizing an octree-based constrained interpolation profile (CIP) solver for regions requiring less detail.

Main Results:

• The proposed technique effectively reduces nonphysical dissipation of velocity.
• Demonstrated ability to reproduce detailed movements of high-Reynolds-number fluids (droplets, bubbles, thin sheets, whirlpools).
• Achieved increased accuracy in the advection step, forming the basis for improved simulations.

Conclusions:

• The hybrid particle-grid method offers a significant improvement over conventional fluid solvers by minimizing dissipation.
• The technique is suitable for simulating complex, high-Reynolds-number fluid phenomena with high fidelity.
• Enhanced advection accuracy benefits both detailed and large-scale fluid simulations.