Intelligent In-Situ Feature Detection, Tracking and Visualization For Turbulent Flow Simulations

JMSI Inc. and the University of California, Davis, shall develop a methodology they call - Intelligent In-Situ Feature Detection, Tracking and Visualization For Turbulent Flow Simulations. This method utilizes a Python-based framework that enables any Python wrapped flow solver to share, without redundant memory penalties, pertinent data structures with intelligent feature detection-tracking tools and visualization software. The intelligent feature detection and tracking algorithm requires knowledgeable domain experts to use an interactive graphical front end tool to identify features rendered in a few initial and intermediary time steps of the solution. Then, as the computation progresses in time, the system self trains and adapts the transfer functions that allow for the feature to be tracked. The second method is to have the domain expert specify quantitative parameters that identify features within the flow. The transfer function adapts the feature detection to the user's parameters and trains the system. A system that has been trained with inputs from a domain expert utilizing these two methods could then be utilized by non-expert users to detect and track features in another similar flow field and geometry. BENEFIT: This innovation will allow non-expert users to detect and track flow features contained within large unsteady datasets using knowledge of domain experts that has been captured within the trained intelligent system.