Keith Obenschain

Using CT-Analyst to optimize sensor placement

Networked groups of sensors that detect Chemical, Biological, and Radiological (CBR) threats are being developed to defend cities and military bases. Due to the high cost and maintenance of these sensors, the number of sensors deployed is limited. It is vital for the sensors to be deployed in optimal locations for these sensors to be effectively used to analyze the scope of the threat. A genetic algorithm, along with the instantaneous plume prediction capabilities of CT-Analyst has been developed to meet these goals. CT-Analyst’s time dependant plumes, upwind danger zone, and sensor capabilities are used to determine the fitness of sensor networks generated by the genetic algorithm. The optimization and the requirements for the evaluation of sensor networks in an urban region are examined along with the number of sensors required to detect these plumes.

Designing Microfluidic Components for Enhanced Surface Delivery Using a Genetic Algorithm Search

*† ‡ § The Toolbox approach to the automated design of microfluidic components is extended to include a genetic algorithm search of candidate designs. Performance metrics for characterizing surface delivery are described, and the software is applied to choose sequences of grooves to add to a rectangular microchannel in order to optimize surface delivery for pressure-driven flow. Initial searches using five groove shapes identify designs that perform much better than standard mixers found in the literature. These initial searches produced two sets of competing designs, and each set was dominated by a different subset of the allowable groove shapes. Additional targeted searches that limited the groove choices to each of these two subsets produced additional significant improvements in the designs.

Using CT-Analyst as an integrated tool for CBR analysis

Modern information systems that are designed to plan for and respond to Chemical, Biological, and Radiological (CBR) attacks are now attempting to include airborne contaminant plume models in their application package. These plume models are a necessary component for the emergency personnel responding to an actual event and to those charged with developing an effective response plan in advance. The capabilities to create a variety of CBR-event scenarios quickly, to determine possible contaminant agent release locations from reports and sensor data, and to predict the path of a plume before it gets there, are functions that many involved in the military and homeland security will find beneficial. For this reason CT-Analysts ability to generate accurate, time-dependant plumes that can be rendered much faster than real-time and can be adjusted and modified on the screen, is an incredible asset to developers of these civil defense systems. The value of the fast, accurate CT-Analyst computer models for complex urban terrain is greatly increased when the capabilities can be imported into platforms users and developers are already taking advantage of. CT-Analysts strengths can now be accessed through other applications, GIS tools, and development environments. This process will be described to show how this is possible, to which systems it can be applied, and what benefits can result.

Using GPUs on HPC Applications to Satisfy Low-Power Computational Requirements

*† The High-performance, low-power computing is required to reduce the computer infrastructure needed for large multi-physics calculations for reactive flow, high-resolution urban aerodynamics, deforming geometry fluid dynamics, etc. If computer infrastructure and costs can be reduced sufficiently, highly accurate calculations currently being performed only in large computer centers can be moved to operational centers and even into the field. The resulting ability will shorten reaction time and add new capabilities in the field. The approach taken will be to investigate the use of Graphical Processing Units that are making their way into PC based systems. These systems provide much more computational capability per watt than current CPUs and thus are suited to forward deployed military applications. These systems, however, are also low cost and have not been designed for scientific computing. We will examine this novel architecture by investigating new implementations of a key CFD computational kernel.

HPC Enhancement of Plume Modeling for Use by Military Simulators

An application programming interface (API) was developed to make CT-Analysts high fidelity physics based plume predictions available to modeling and simulation (M&S) tools. This plume API was used with the M&S application one semi-automated forces (OneSAF). OneSAF is a next generation entity level simulation. For a 10 by 5 km region of Baghdad, OneSAF made use of CT-Analyst through the application programming interface (API) to determine the plume locations, the plume concentration at points within the simulation, and to determine the attenuation of visibility along a line. Before the simulation is run, FAST3D-CT generates a pre-computed Nomograf database. During the simulation, OneSAF calls a CT-Analyst process that in turn interprets the Nomograf database to generate plumes.

Large Eddy Simulation of Accidental Releases

First responders need a more or less instant estimate of danger zones resulting from accidentally released hazardous materials in order to take immediate action, to coordinate rescue teams and to protect human population and critical infrastructure. To fulfill the need for a sufficient dispersion modeling accuracy while maintaining efficient access to reliable results in a first responders environment, systematic high resolution pre-accidental LES modeling can be combined with ’physical data reduction’ in an emergency assessment tool. A typical example of such an approach adjusted to the geometry of the Hamburg inner city area will be presented. It gives a glimpse into the application of LES-modeling for real-world problems.

Analysis and Design of Novel Column Geometries for Micro- Gas Chromatography

Novel column geometries for micro-gas chromatography applications are studied using traditional computational fluid dynamics and the numerical Toolbox, a suite of codes designed to model the flow of low-velocity liquids and gases through microfluidic components. Transport is modeled in columns with circular cross-section undergoing a 90ϒ bend to determine what effect sequences of such bends have on analyte transport. We also explore flow in columns with a semi-circular cross section that include grooves cut into the flat column surface to promote transport across the channel. The modeling demonstrates that, even at low Reynolds number, sequences of left and right turns redistribute the fluid within the cross-section, and grooves added to a flat channel wall can generate a secondary flow that is strong enough to redistribute the fluid within the cross section in a relatively short length of channel.

Performance of Unstructured Finite Volume Code on a Cluster with Multiple GPUs per Node

This paper will investigate the performance of an unstructured finite volume code on a multi-CPU, multi-GPU cluster. This cluster attempts to balance IO, GPU, and CPU performance to accommodate a wide variety of codes. A new, unstructured finite volume code running in parallel using MPI/OpenMP and MPI/CUDA is presented. The performance of this code on a purpose-built GPU cluster is examined under a number of operating conditions. The GPU cluster is a collection of 24 compute nodes, each consisting of two, 6-core Intel Core i7 Processors and two NVIDIA GPUs with one to two QDR Infiniband ports connected to a switch. Eight of the compute nodes have two NVIDIA Fermi Tesla GPUs well connected with two QDR Infiniband cards. The remaining 18 have two NVIDIA Fermi Video GPUs. The use of multiple chipsets creates non-uniform access to both the GPUs and Infiniband, potentially creating bottlenecks when transferring data between the CPU and the GPU and between nodes. This paper will also explore these issues as well as potential solutions.

Use of the NRL DHPI System to Transfer Dispersion Nomograf Capabilities to the Field

As the ability to generate Dispersion Nomografs™ on the Dedicated High Performance Computing Project Investment (DHPI)-provided machine matures, the focus of this project has moved to the transition of CT-Analyst. CT-Analyst provides near-instantaneous urban plume prediction with unprecedented accuracy and ease. This paper will focus on the transition efforts to deploy CT-Analysts unique Dispersion Nomograf to military and civilian training and field applications. Overall, it has been easier to transition CT-Analysts capabilities to training tools versus field deployable applications. We will describe the users and types of applications CT-Analyst has been transitioned to, as well as the difficulties encountered during the transition process.

Using real-time chemical plume models in virtual training systems

This paper describes the design, implementation, and use of integrated chemical plume-models in virtual training systems. The US Naval Research Laboratory has linked its CT-Analyst® software, a high-fidelity real-time plume modeling tool, with VBS2, a widely-used virtual gaming and training program, to produce new training capabilities that were previously unavailable. This work benefits two different but overlapping training scenarios: 1) tactical training for large-scale chemical gas attacks with a specific focus on crowd management, and 2) handler-focused training for users interested in working with IED-detecting dogs. The use of accurate, faster-than-real-time plume modeling enhances the virtual training systems to provide broader realistic support to the simulation and training communities.