Presentation is loading. Please wait.

Presentation is loading. Please wait.

Efficient Simulation of Large Bodies of Water by Coupling Two and Three Dimensional Techniques SIGGRAPH 2006 Geoffrey Irving Eran Guendelman Frank Losasso.

Similar presentations


Presentation on theme: "Efficient Simulation of Large Bodies of Water by Coupling Two and Three Dimensional Techniques SIGGRAPH 2006 Geoffrey Irving Eran Guendelman Frank Losasso."— Presentation transcript:

1 Efficient Simulation of Large Bodies of Water by Coupling Two and Three Dimensional Techniques SIGGRAPH 2006 Geoffrey Irving Eran Guendelman Frank Losasso Ronald Fedkiw

2 Abstract ◇ effective simulation of a boat in a large box ◇ Similar to traditional 2D height field, the water is represented by tall cells ◇ Accurate representation of bottom topography

3 Introduction ◇ Motivation : Stormy seas, Sudden floods, cascading rapids ◇ Very expensive effects ◇ Appearance is governed by a surface layer ◇ Interior flow is not visible

4 Octree and Height field ◇ Octree’s drawback - Too much node, Increased numerical dissipation - Large octree cells can’t represent bottom topography ◇ Height field’s merit - Detail refinement in the horizontal direction - Capturing the effects of complex bottom topography well - Not support overturning or other interesting 3D behavior

5 Our method requirement (1) ◇ Detailed 3D behavior near the interface ◇ “Optical depth” is greater than the visible optical depth  uniform MAC grid Navier-Stokes solver near the interface

6 Our method requirement (2) ◇ Represent without losing plausible bulk motion ◇ Represent important details such as bottom topography  Coarsen in the vertical direction obtaining tall thin cells  reach down to the bottom of the domain

7 Advantage ◇ The Octree method suffers from increased dissipation ◇ RLE (Run-Length-Encoded) [Houston et al. 2004; Breen et al. 2004; Wiebe and Houston 2004; Nielsen and Museth 2005; Houston et al. 2005; Houston et al. 2006]  only in the air ◇ Our method greatly reduces the computational cost

8 Previous Work ◇ 3D Navier-Stokes equations - Foster and Metaxas 1997b - Stam 1999 - Fedkiw et al. 2001 ◇ Method for simulating a water surface - Foster and Metaxas 1996 - Foster and Metaxas 1997a - Foster and Fedkiw 2001 ◇ Particle level set - Enright et al. 2002

9 Previous Work ◇ Conservative height field fluid model - Baxter et al. 2004b ◇ Viscoelastic fluid - Goktekin et al. 2004 ◇ Solid fluid coupling - Calson et al. 2004 - Guendelman et al. 2005 ◇ Control - McNamara et al. 2004 - Shi an Yu 2005 ◇ Two phase flow -Hong and Kim 2005

10 Previous Work ◇ Large and deep water simulation - Fournier and Reeves 1986 - Peachey 1986 - Mastin et al. 1987 - Ts’o and Barsky 1987 - Thone et al. 2000 - Hinsinger et al. 2002 ◇ Navier-Stokes Equation update for large bodies of water - Chen and Lobo 1994 - Thon and Ghazanfarpour 2001 - Neyret and Praizelin 2001

11 Grid Structure ◇ Uniform cells and Tall cells ◇ Scalars are stored at cell centers ◇ Velocity components are stored on their respective faces ◇ Tall cells contain uppermost and bottommost pressure values ◇ Pressure values of tall cells can be interpolated

12 Grid Structure ◇ Horizontal pressure derivative ◇ Co-locate horizontal velocities ◇ Vertical pressure derivative ◇ Co-locate vertical velocities

13 Refinement and Coarsening ◇ When change the structure, Interpolate velocities in new grid ◇ The vertical velocities at new tall cells are set to the average value ◇ Those for uniform cells are defined as their interpolated value ◇ This conserves vertical momentum

14 Refinement and Coarsening ◇ For horizontal velocities ◇ Let u(j a ), u(j a +1), …, u(j b )  We desire bottom and top values u a and u b

15 Refinement Analysis and Comparison ◇ With a sufficient optical depth satisfactory results are obtained (middle : ¼ refined case) ◇ 1/16 refined case obtain nonphysical results (right) ◇ Since the computational cost of a tall cell is independent of its height, the tall cells are equivalent in cost regardless of the depth of the water

16 Refinement Analysis and Comparison ◇ Right shows a octree simulation of the splash ◇ Our method (left) is faster and larger simulation

17 Uniform Three-Dimensional Method ◇ Navier-Stokes Equations ◇ Mass conservation ◇ Linear Momentum

18 Mass Conservation

19 Linear Momentum

20 Update u using N-S Eqns

21 Advection ◇ We chose conservative method ◇ The conservation does not cure large truncation errors ◇ it only makes the results more physically plausible

22 Advection ◇ Ignoring the pressure and force terms, we focus on the u velocity ◇ Velocity is updated based on the fluxes across faces ◇ The flux are computed by averaging the velocities to the faces

23 Advection ◇ CFL condition on the time step for stability ◇ We could refine the tall cells into a uniform grid, and re-coarsen ◇ Instabilities appear sharp changes ◇ compute the total weight

24 Laplace Equation ◇ After advection, we solve for the pressure and make the divergence free ◇ We assume that the density is spatially constant

25 Laplace Equation ◇ Consider two adjacent tall cells extending from j 1 to j 2 and j 3 to j 4 ◇ Intersecting in a minimal face from j a = max(j 1, j 3 ) to j b = min(j 2, j 4 ) ◇ Component of discrete gradient is linear with value (P x ) a and (P x ) b ◇ given by interpolating pressure to j a and j b

26 Laplace Equation ◇ Derivative Py = (p 2 -p 1 )/((j 2 -j 1 )dy) vertical pressure p 1 at j 1 and p 2 at j 2 ◇ The discrete volume weighted divergence of a uniform cell ◇ +- sign is chosen the face, A f is the area of the face

27 Laplace Equation ◇ Total flux contribution ◇ We substitute (P x ) a and (P x ) b for u a to u b

28 Examples

29 Example

30 Conclusion ◇ Novel method for the simulation of large bodies of water ◇ The bulk of the water volume is represented with tall cell ◇ Capturing detailed surface motion ◇ Representing detailed bottom topography


Download ppt "Efficient Simulation of Large Bodies of Water by Coupling Two and Three Dimensional Techniques SIGGRAPH 2006 Geoffrey Irving Eran Guendelman Frank Losasso."

Similar presentations


Ads by Google