Flow Visualization: Intro / Animated Particles / Massive Particles / Soiling Simulation
Automotive flow simulations that are computed by commercial Lattice-Boltzmann applications such as PowerFlow from Exa Corp. need special treatment for visualization. In cooperation with the BMW AG, Munich a tool was developed that is able to handle the large hierarchical data sets, and to view stream lines, ribbons and glyphs as well as slice probes and animated particles at interactive frame rates. This virtual windtunnel project is financed by FORTWIHR, the Bavarian Consortium for High Performance Scientific Computing.

An actual research topic is particle tracing with non-zero mass. To solve the 2nd order PDEs a Runge-Kutta particle tracer has been developed, which is also used to perform automotive soiling simulations at BMW. Another main topic was to provide an immersive environment for ease of navigation and investigation. This lead to implementations of the visualization tool for stereoscopic enviroments like the PowerWall and the CAVE (see also publications).

Overview (Gallery 1):
Here are four screen shots from the Lattice-Boltzmann flow visualization environment. On the left side you can see two images of a BMW 5 series touring with a variety of particle and slice probes positioned in front and at the side of the car. In the middle the flow pressure is color-coded onto the vehicle surface to illustrate the its flow properties. On the right side the projections of the car onto the 4 sides of a CAVE are shown.
Particle Probes (Gallery 2):
This is a series of pictures that shows the vortex originating at the left side of the front window. From left to right that vortex is visualized by means of stream lines, stream ribbons and glyphs. With the aid of a slice probe, which maps the flow speed from blue (low) to red (high), you can see that the velocity inside a vortex is quite low as shown in the left picture. The cutting plane in the middle picture is used to localize areas of low (blue) and high pressure (red). Unsurprisingly, the pressure is especially high in front of the car and at the front window. By utilizing glyphs two or more scalar values can be encoded by a single three-dimensional arrow. In the right picture, for example, the velocity is represented by the length of the glyph, whereas the glyphs color is representing both pressure and total pressure in the flow (see paper for more information).
Animated Particles (Gallery 3):
In turbulent flows stream lines diverge quickly, so it is often very difficult to achieve a good understanding of the flow properties. Here animated particle probes offer a way to interactively trace a number of evolving particles that are generated by a stochastic process. In the above series, for example, the particle stream that is emitted from a small user-defined region behind the car keeps together for a small period of time but eventually splits into two rapidly diverging flows. One implicit advantage of animated particles is that the velocity can be viewed directly. In the samples above the color of the particle traces indicates their life time. On an SGI Octane MXI (2x250 MHz MIPS R10K processor) approximately 500 particles can be animated at a frame rate of 25 Hz. The particle tracing system scales nicely with the number of processors, thus the performance on an 8 processor SGI Onyx2 is quite impressive.
Massive Particles (Gallery 4):
The physically correct tracing of point masses enables us to perform soiling simulations with the animated particle probe system. The involved second order differential equations (PDEs) are solved by an embedded Runge-Kutta tracer of order 4(3). By applying an octree based collision detection the hit points on the car surface are visualized by a color change with red indicating highest dirtyness. The left picture shows a soiling simulation with dusty air. In the middle picture dirt particles were emitted by the front and back wheels of the car, thus simulating the car's self-soiling. In the right image dust particles were generated in the extent of the wireframe box behind the car leading to a dust distribution on the back of the car.
Soiling Simulation (Gallery 4):
A series of five images illustrating the soiling of a BMW 5 series touring. The dust distribution on the sides and the back of the car coincides well with real-world soiling evaluations conducted by BMW (see also VisSym 2001 paper). As we are using a stationary flow for the soiling simulation we expect the results to become even better when migrating to instationary data sets.