II. PROBLEM DESCRIPTION

The problem investigated is the rendering of curvilinear data sets. Visualizing irregularly structured data is more difficult than the rectilinear case. In this case, the grid lacks a well-defined (known) structure, so the relation between pixels on the screen and points in the grid is not immediately obvious. As such, this research focuses on a new method to resample curvilinear data sets to a rectilinear grid. A renderer then generates the 3D image based on this rectilinear grid. The particular application of interest is the visualization of computation fluid dynamics (CFD) simulation data.

The visualization of CFD data is an active research area in computer graphics. As a tool, visualization of CFD data has great potential to aid scientists and researchers in both understanding and interpreting large 3D data sets. Useful visualization of CFD data should be interactive, to allow the researcher to vary parameters and immediately see the effects. An interactive visualization tool also benefits the researcher during the modeling stage. Further, the visualization system should incorporate the vast and multivaried data a CFD application generates.

Before this is possible though, considerable improvement must be made in the current state of renderers for CFD data. Toward that end, much effort has concentrated on improving the abilities and performance of (curvilinear data) volume renderers.

However, several problems remain [Eb]. Current systems lack the ability to display accurately the entire scope of the 3D CFD data generated. Similarly, many systems cannot display multiple parameters simultaneously. As a result, the researcher must view separately the density, pressure, velocity, or temperature aspects of the CFD data. Displaying all variables together aids in understanding the relationships between them.

The particular CFD application studied here is that of visualizing air flow around the blades of a turbo jet compressor. The simulated flow produces a large data set, including all the variables mentioned above. The purpose of visualizing this data is to analyze critical regions of flow under varied circumstances, and to check for shock or expansion waves. Other papers [Eb] [Bu] [Yag] describe the problem more fully.

Considering the size of the data, the curvilinear grid itself is not terribly large: its dimensions are 18 X 46 X 60. In this research, the algorithm resamples the grid to a regular rectilinear grid. Having resampled the curvilinear data to voxels, the algorithm processes the regular voxels in a straightforward manner to produce a vector field, which is then rendered (using a realistic volume-tracing algorithm).

---