Jacqueline H. Chen is a Distinguished Member of Technical Staff at the Sandia National Laboratories, and a director on the Board of Directors of the Combustion Institute. She has contributed to research in terascale simulations of turbulent combustion focusing on turbulence-chemistry interactions in combustion based on the S3D code. She has worked closely with computer scientists on timevarying visualization of terascale simulated data, topology of combustion features, and parallel feature detection and tracking algorithms for combustion. She received the DOE INCITE Award in 2005 and 2007 and the DOE Office of Science Leadership Computing Facility Award in 2006.

S3D is the Sandia 3D Direct Numerical Solver [1], which presently is capable of generating tens of terabytes of raw data per run in a parametric study of several runs, presenting a significant challenge for subsequent analysis and interpretation. The current approach is to store the raw data at prescribed time intervals, and then either analyze it on a supercomputer or access it via a local analysis and visualization cluster at Sandia. The raw data is archived so that various modeling groups in the combustion community can interrogate it to test model assumptions. To understand the correlation of scalar fields such as temperature, mixing rates, and species concentrations in turbulent flames, scientists must be able to visualize two or more scalars simultaneously. Conventional visualization tools do not directly support such a capability. Scientists often must make side-by-side comparisons of images of different variables by hand, which is tedious and time-consuming. Furthermore, the information that scientists can derive by looking at separate images is quite limited. Thus, they need effective methods for simultaneously visualizing multiple time-varying variables from large data sets in an interactive fashion [2][3].

The data rates of S3D are expected to increase by tenfold in the next two years, demanding new analysis and visualization capabilities to extract physical insights from such large time-varying, multi-scale complex data. Due to the enormous data volume, scientists need parallel feature detection, extraction, and tracking tools to automate reduction of the data for analysis of intermittent combustion phenomena. An extensible set of feature identification and tracking algorithms needs to be developed into a parallel feature analysis pipeline that is efficient and scalable to petascale computers and beyond so that the automated data analysis can be performed online or offline. The algorithms need to be efficient so as not to impose a significant runtime penalty. Furthermore, a library of parallel turbulent combustion analysis tools needs to be developed for structured grids to understand turbulence interaction with flames and ignition kernels. The tools need to be accessible as part of the feature borne analysis and also in a postprocessing mode. The library of tools must support flame surface analysis, chemistry analysis, conditional statistics, turbulence and scalar spectra, multiscale representation of combustion data, Lagrangian particle tracking, and scalar topology.

[Download]

If you publish your work using this data set, please make acknowledgment of VisFiles: The data set is made available by Dr. Jackqueline Chen at Sandia Laboratories through US Department of Energy's SciDAC Institute for Ultrascale Visuaization.

[Combustion Data]

 
Return to Meet the Scientists