Animating Explosions

Abhinav GolasAkram Khan

Under the guidance of Prof. Kalra & Prof. Kumar

Outline The problem How we’ve split the problem Current progress Some comparison

The problem Animating explosions –

Navier Stokes Modeling the pressure

wave Issues

Large quantities Large spatial & temporal

derivatives Inherent shocks (speed of

sound) Large bulk transfer

How we’ve split the problem Abhinav Navier Stokes

modeling Focus on seeing

impact of viscosity on simulation in different media

Models for different solver components – applying known methods for incompressible flows to problem

Akram Euler equation

modeling – inviscid Navier Stokes

Mostly first derivatives – scope for linear algebra based solvers

Focus will be on modeling as sparse linear systems

Current Progress Our implementation

Ready Some methods for larger timestep simulation

ready Original paper implementation (Animating

Explosions Yngve et al, SIGGRAPH 2000) 2D version ready – some parameter tuning

required Rendering

Renderman based particle systems renderer Uses order of 100000 particles

Some comparison Original implementation

0.1 ms timestep for simple blast case Our implementation

1 ms timestep for simple blast case using multistep advection

5 ms timestep for same case also done.

Modifications made Staggered grid (vs. centralized grid)

Better accuracy of spatial derivatives Semi-lagrangian based multi-step

advecton(vs. donor acceptor method) Much more stable – guarantee of output<=input

(but slight dissipation) Allows for larger timestepsStep 1ms 5ms

Frame at 15ms

Future work Complete comparison with earlier

implementation Improve methods to allow larger timesteps Improve rendering Experiment with ENO methods (CFD) Develop solid interaction and fracturing

models

Thanks