Automotive Soiling Simulation Based On Massive Particle Tracing

1
Automotive Soiling SimulationBased On Massive
Particle Tracing

• Stefan Röttger
• Martin Schulz
• Wolf Bartelheimer
• Thomas Ertl
• Visualization and Interactive Systems Group
  University of Stuttgart

2
Introduction

• Where did we start?
• Lattice-Boltzmann CFD solver (PowerFlow) used at
  BMW
• Hierarchical cartesian grids
• Fast tri-linear interpolation

3
Standard Flow Visualization Tools

• Interactive and immersive navigation in virtual
  windtunnel
• Stream lines
• Stream ribbons
• Glyphs
• Cutting planes

4
Goals

• What did we want?
• Massive particles for simulation of dust
  particles
• Interactively animated particles for more
  intuitive flow visualization
• Automotive soiling simulation

5
Massive Particles

• Each particle is treated as ideal sphere with
  specific mass, diameter and initial position and
  velocity
• Particle drag, gravity and electrostatic forces
  affect acceleration of particles
• Stokes approximation of the drag is not good
  enough
• Particle drag in the flow is approximated by the
  formula of OSeen for low Reynolds numbers

6
Massive Particle Tracing

• Second order differential equations
  a(t)gtv(t)gtx(t)
• Adaptive, embedded Runge-Kutta tracer of order
  4(3)
• Below 3 µm the differential equations are
  becoming stiff, but then massive and massless
  particle tracing is almost equivalent gt no
  soiling
• Average dust particle diameter in real world
  evalutions is 5 to 500 µm

7
Animated Massive Particles

• Define emitters that generate particles at a
  certain frequency
• Emitters can be sized and positioned
  interactively
• Initial parameters of particles are assigned
  stochastically
• Particle stream is traced and displayed step by
  step
• The stream is displayed in slow motion at a given
  target frame rate
• Camera exposure model gt particles leave short
  traces

8
Animated Massive Particles

• Intuitive visualization analogue to smoke probes
• Simultaneous display of multiple particles
• Particle velocity is visible implicitly
• Better three-dimensional impression due to
  animation
• Life time is color coded

9
Automotive Soiling Simulation

• Stochastically generate and trace massive
  particles
• Check for collision with the 70,000 surface
  triangles by utilizing an octree
• Color the hit points on the car body (nearest
  mesh vertex)
• Color code the number of hits on the surface
  (blue no hits)

10
Simulation Performance

• Hit probability is low
• Large number of particles
• Simulation can be parallelized efficiently (e.g.
  64 CPU SGI Onyx)
• On an SGI Octane with 2x250MHz MIPS R10K
  approximately 1000 particles can be traced
  simultaneously at 7 Hz
• Scales well with CPUs

1 hour
3 hours
11
Simulation Quality

• Now using stationary flow fields (120M per time
  step)
• Turbulences are smoothed away in time averaged
  flow fields and hit probability is reduced even
  further
• gt Instationary flow fields
• Electrostatic forces influence dust aggregation
  as well

Stokes
OSeen
12
Conclusion

• Animated massive particle streams for intuitive
  data set exploration
• Massive particle tracing used to compute
  automotive soiling simulation by employing
  collision detection
• Good coincidence with real world soiling
  situation
• More accurate simulations require instationary
  flow fields and research about near surface
  effects
• Massive particle tracing can be applied to other
  regions of interest like smog or droplet
  distribution simulations

13
Discussion
Questions?