The open archive for STFC research publications

Full Record Details

Persistent URL http://purl.org/net/epubs/work/63704
Record Status Checked
Record Id 63704
Title Genetic Algorithm Driven Decision Support System for Facilitating Real-Time Analysis and
Abstract The Orbiter Genetic Algorithm Driven Decision Support System (GADDSS)1,2 is based on many years of conceptualization and development of infrastructure and services that meet the needs of a user base. The neutron beams at the Spallation Neutron Source (SNS) facility located at the ORNL are in high demand. The efficient use of the available beam time requires informed decisions to be made in order for scientists to treat, analyze and model their experiment data. Neutron scientists who are hungry for science-enabling tools and technology now often demand data access and visualization tools that support interactive and programmatic interfaces. Not only are rapid data analysis and simulation capabilities in high demand, but also real-time analysis and visualization of protein crystal and enzyme structures and functions are desired to enable intelligent instrument control. The current commissioning of the neutron science instruments at SNS and HFIR and the schedule over the next several years has led neutron scientists to approach computer and computational scientists for help with identifying enabling technologies for addressing their needs. Common questions are "If I only had ..." and "If I only could ..." followed by "... I could really breakthrough." or "... it would revolutionize my field." Furthermore, a current bottleneck of efficient data collection is the lack of software allowing for real-time tracking of diffraction pattern as it is being collected in an integrated manner. Computer programs for post-mortem data analysis interpreting Bragg diffraction data and refining ordered structures are readily available for structure analysis. This GADDSS real-time system will enable structure refinement to occur while samples are still in the instrument which provides never before attainable real-time determination of experiment duration based on refinement criteria provided by the scientist and verified by the real-time analysis of live experimental data. The base services and infrastructure needed to integrate these capabilities into the design and control of experiments in real-time is currently being developed by the ORNL facility funded Accelerating Data Acquisition, Reduction, and Analysis (ADARA) project. By integrating with this project and properly applying computational resources GADDSS strives to give the instrument scientist real-time control of the neutron instrument experiments. With the ability to guide the experiment using real-time visualization of ongoing experiments, a significant benefit of this research is the efficient utilization of experimental data yielding high-throughput and reduced cost of operation. Dr. Green has focused on user centric solutions and services for over a decade and has delivered grid, cloud, and distributed system infrastructure to the community on a routine basis. In order for the presented technologies to be of real value they must enable the scientist to do and/or perform complicated real-time experiment analysis within a user-friendly environment. The GADDSS environment presents intuitive interfaces that lower the barrier of entrance for the scientist that does not want to become an expert in cyberinfrastructure in order to do their science. In our presentation we will report the development and integration of the peak integration and profile fitting models for optimizing molecular structure refinement, workflow automation, real-time integration of live data processing, the enhanced decision support system, and the implementation of GPU-enabled infrastructure for real-time analysis. 1.Green, ML, Miller, SD, Vazhkudai, SS, and Trater, JR. Doing Your Science While You're in Orbit. Journal of Physics: Conference Series. 251, (2010) DOI: 10.1088/1742-6596/251/1/012095. 2.Green, ML and Miller, R. Evolutionary Molecular Structure Determination Using Grid-enabled Data Mining. Parallel Computing. 30, 9-10, 1057-1071, (2004) DOI: 10.1016/j.parco.2004.07.011.
Keywords NOBUGS2012
Funding Information
Related Research Object(s):
Licence Information:
Language English (EN)
Type Details URI(s) Local file(s) Year
Presentation Presented at NOBUGS 2012 (NOBUGS 2012), RAL, UK, 2012. Mark-L-Green-Tech-X.pptx 2012