Review of Two-Scale Particle Simulation Paper by: Barbara Solenthaler ETH Zurich Markus Gross ETH Zurich.

Review of Two-Scale Particle Simulation

Paper by:

Barbara SolenthalerETH Zurich

Markus GrossETH Zurich

Overview

The authors explain that they have a dual-scale method for particle-based fluidic simulations that

assigns CPU resources to simulation regions where complex flow behavior occurs. Their

method uses low and high resolution simulations that run at the same time. The coarser fluid

simulation is represented by large particles, while the fine level simulates only a subset of the total

fluidic region with smaller particles.

Overview

Fluid simulations demand a high degree of discrete resolution (number of particles) to obtain realistic simulations. Often, small details such as surface ripples, thin sheets and small droplets are not reproduced if the scale is too high. Alternately, if the scale is too low, the computational cost can be too large.

Two-Scale Model

In this model, you can see the low resolution (blue) and high resolution (yellow) particles so you can get more detail around the collisions with the cylinders.

Two-Scale Model

This shows some of the models’ splash simulations around the cylinders in the water.

Flood Corridor Example

In the next four pictures, you can see the high resolution (yellow) defined according to the camera view as the picture is rotated around.

Flood Corridor Example

In this shot, you can see how the higher resolution allows for more realistic flows where the water hits the edge.

Flood Corridor Example

As the water flattens out, the higher resolution allows you to see ripples and splashes in the water that you wouldn’t get with the lower resolution.

Flood Corridor Example

And again, you can see ripples and splashes in the view closest to the camera, while the detail in the view away from the camera is at a lower resolution.

Two-Scale Model Overlay

In this diagram, you can see how they overlay a high resolution simulation over top of a low resolution simulation. The particles of both simulation are merged for the final rendering.

Smoothing Equation

This equation describes the smoothing over a neighborhood with radius h by using a kernel W(xij, h) to weight the contributions according to the distance xij between two particles i and j.

Complex Regions

This diagram shows how their model can deal with dynamically changing, complex regions such as the boundary between two different fluids with different densities.

Region Transitions

Abrupt density force changes are avoided during a region transition by the way the particle forces are modeled.

Particle Distribution

Diagram A shows various sampling schemes (low and high resolutions), while diagram B shows the particle creation at boundary walls with low and high resolutions.

Feedback

The diagram, from left to right, shows L without feedback, L with feedback from H and H. Feedback provides a type of dampening effect that makes the fluid simulation appear more realistic.

Conclusion

Their two-scale model seems to reduce the overall computational cost of fluid simulation while still maintaining the small scale surface details that make the simulation more believable. Their model also allows more computational resources to be allocated for the higher resolution areas, where more detail is desired. A very effective model, overall.