Abstract Chen

Visualization and Analysis of Petascale Combustion Simulation Data

Direct numerical simulations of turbulent combustion have generated enormous volumes of data (over 50TB last year) that require new analysis and visualization capabilities. We need to deploy effective strategies to extract physical insights from such large data. Our effort in this area will focus on three distinct areas that will be coupled together.

First, parallel feature detection, extraction, and tracking tools will be developed to automate reduction of the data for analysis of intermittent combustion phenomena. We propose to extend the Feature Detection & Tracking Library, FDTools, a serial framework to support the easy assembly of an extensible set of feature identification and tracking algorithms, into a parallel feature analysis pipeline collaborating with Professor Kwan-Liu Ma, PI of the SciDAC Ultrascale Visualization Institute.

Second, a library of parallel turbulent combustion analysis tools will be developed for structured grids to understand turbulence interaction with flames and ignition kernels. The library of tools will focus on flame surface analysis, chemistry analysis, conditional statistics, turbulence and scalar spectra, multi-scale representation of combustion data, Lagrangian particle tracking, and scalar topology. Portions of this effort require close collaboration with Valerio Pascucci of the SciDAC VACET Center and with Ramanan Sankaran of NCCS/ORNL.

Third, we will implement a novel approach to simultaneous visualization of multiple scalars in large time-dependent structured data using volume rendering. Professor Kwan-Liu Ma of the SciDAC Ultrascale Visualization Institute will collaborate with us to develop an intuitive user interface for his volume rendering software, and implement his interactive volume visualization system on a new Sandia BES visualization cluster that supports graphics accelerated rendering to facilitate interactive temporal browsing of large data.