Spatio-temporal Dynamical System Analysis Tools for Very Large Data Sets
Small Business Information
5100 Springfield Street, Suite 301, Dayton, OH, -
AbstractABSTRACT: This Phase-I research effort will enable propulsion engineers to design and control high-pressure combustors/augmentors operating on various aerospace fuels and emerging alternatives by developing a set of nonlinear analysis tools that can achieve two overarching goals: efficiently analyze large numerical or experimental data sets and provide advanced data mining techniques to more quickly identify relevant dynamical system behavior. This will be accomplished through four key technical objectives: (a) to remove noise from large data sets thereby reducing the quantity of data needs to be further analyzed by separating the noise from nonlinear dynamics, (b) to remove redundant information from large sets of data to reduce the equivalent of 1 TB of raw data to ~200 GB depending on the reacting flow conditions, (c) to identify dynamically significant information within a very short amount of time to understand the underlying physics of the spatio-temporal dynamical systems, and (d) to develop a set of dynamical invariant tools to compare experimental and numerical data in order to influence the development of next generation war fighters as well as extending the life of existing war fighters. The development of the data analysis tools are required to exploit the current state-of-the-art laser-based measurement systems, which routinely provide 2D temperature and species concentration measurements at a rate of 10 to 50 kHz, as well as to analyze reacting flow data generated by state-of-the-art supercomputers. The proposed research effort will focus on analyzing 1 TB of imaging data within a day using standard laptop computers, or even faster using high-bandwidth GPU processors, thereby enabling real-time engineering analysis, design, optimization, and control. BENEFIT: This research effort on the development nonlinear spatio-temporal data analysis tools would allow investigation of turbulent reacting flow phenomena in a real-time basis. This tool set should also allow devising intelligent control strategies for the development of intelligent engines and will have a major impact on understanding of high-speed time-evolving phenomena related to ignition, flame growth, and stability in turbine engine augmentors/afterburners. Hence, this research effort along with methods to analyze high-speed planar and tomographic images will make the concept of real time sensing and control of combustion phenomena a reality. The proposed software system could be easily integrated with any on-board sensing and control system, which would have a significant impact on engine health and stability for the war fighter. Furthermore, in the modeling of reacting flows related to combustors and augmentors, it is of great importance to study the interactions among various temporal and spatial scales to understand and predict the complex behavior under turbulent conditions. This understanding requires information on local flame-front phenomena, global heat release, and acoustic feedback at frequencies of 5 kHz or greater. The objective of this effort is to develop various nonlinear analysis tools for the interpretation of high-bandwidth (generally 5 kHz or greater) numerical data for comparison and validation with laser diagnostics signals acquired with fs-CARS, fs-LIF, PLIF, PIV, and time-division-multiplexed (TDM) laser systems. The advent of new laser technologies is enabling high-bandwidth temperature and species-concentration measurements in chemically reacting flows that were unimaginable just a few years ago. For example, particle image velocimetry and planar laser-induced fluorescence are now performed at a rates of 10 kHz or greater; diode laser-based thermometry and species-concentration measurements are now routinely performed at a rate of 50 kHz; and femtosecond laser-based point or line CARS spectroscopy is performed at rates of 10 kHz or greater. The tremendous bandwidths afforded by these new technologies allow, for the first time, the identification and thorough characterization of spatio-temporal instability modes in chemically reacting flows. The software system should be applicable for any high-bandwidth experimental or numerical data, and as such will have a very broad commercial application covering most of the Universities, Government laboratories, engine companies, etc. Renewed interest in diverse and alternative fuels makes the proposed data analysis tools particularly timely for addressing a number of important research and development needs in high-performance, clean power production from sustainable sources.
* information listed above is at the time of submission.