Authors
Paul Rosenthal, Tran Van Long, Lars Linsen
Introduction
Data sets resulting from physical simulations typically contain a multitude of physical variables. In most cases they are highly dependent on each other and most phenomena can only be explained with a view in most of the attributes. Nevertheless we have only at most four dimensions as domain for visualizations. Hence scientific approaches are necessary to convert the high-dimensional data to human understandable pictures and animations.
The provided time-varying data set shows the simulation of the propagation of an ionization front instability. It includes a variety of different attributes, including density, temperature, mass abundances of eight chemical species and velocity. The size of each data set of the 200 time steps is 37 · 106 points resulting in 1.7 GB of data per time step.
Our goal was to present a visualization method that takes into account the entire multi-field volume data rather than concentrating on one variable. We propose a combined approach based on nonequidistant resampling, multi-dimensional clustering and surface extraction.