Skip to main content Skip to footer site map
Trident Scholar Program
Trident Scholars 2007

Trident Scholar Abstracts 2007


Madeline E. Boe

Midshipman First Class
United States Navy

Modeling Geophysical Fluid Flow Problems Using COMSOL®:
Working Towards a Hydrodynamic Model of the Chesapeake Bay

            The goal of this Trident project is to lay the groundwork for a hydrodynamic model of the Chesapeake Bay. It is a constant challenge to develop increasingly accurate models of a body of water. In the case of the Chesapeake Bay, the goal is not only to make the models more accurate, but to also make them adaptable enough that they may be used for several different applications. A partial differential equation solver like COMSOL® is designed to work with the nonlinear equations associated with fluid flow, particularly the Navier-Stokes equations. A working model of the Chesapeake takes these equations, combines them with the forcing functions for currents in the Bay, like winds and tides, and implements them in the full geometry of the Chesapeake Bay. In order to arrive at the end result of a full model, the project must first be broken down into two separate sets of problems. One side of the project runs small scale models in basic rectangular geometries. Some of these models adapt Stommel’s and Munk’s linear equations for western intensification and allow COMSOL® to solve them using a built-in classical PDE mode. Other models implement the full non-linear Navier-Stokes equations into rectangular or cubic geometries. Each one of the small models deals with some aspect of the Chesapeake Bay. The other half of the project addresses the full complex geometry of the Chesapeake Bay. Instead of immediately using the complete non-linear equations of motion, this part of the project uses COMSOL® to run classical linear partial differential equations, for example Poisson’s equation, within the boundaries of the Bay. When the two halves of the project are completed, they are combined together to create a model using both the complex geometry of the Chesapeake Bay and the full non-linear Navier-Stokes equations.

FACULTY ADVISORS
Professor Reza Malek-Madani
Mathematics Department
Associate Professor David R. Smith
Associate Professor Mario E.C. Vieira
Oceanography Department


 
Mark E. Daniel
Midshipman First Class
United States Navy

Fully Interconnected Optical Network Architecture (FIONA)

        The goal of this Trident Scholar Project is to develop a reconfigurable optical network architecture that connects existing local area networks (LANs) to create a single extended LAN. This includes the design of a network backbone, construction of a testable network and implementation of a control algorithm. The final goal of this project is a proof-of-concept demonstration using a transmitted digital signal over an autonomously controlled network. Recently, network evolution has been driven by a need for increased performance, as characterized by increased data rates and larger bandwidth. The network architecture in this project will allow for both increased data rates and a larger available bandwidth while reducing weight and increasing reconfigurability and reliability.

        This fully interconnected optical network architecture (FIONA) uses dense wavelength division multiplexing (DWDM), wavelength conversion, and out-of-band control to achieve all-optical routing within the network. The architecture uses all-optical wavelength conversion to route data across the backbone or to reconfigure the backbone in the event of link failure. The use of DWDM reduces the number of connections and links while maintaining connectivity between the local area networks. The reconfigurability of the network improves survivability and fault tolerance without requiring redundant systems or links. The use of out-of-band control allows for mixed signal transmission and data transparency across the backbone. Fiber optics allow for higher data rates and larger available bandwidth compared to existing copper networks. FIONA is an improvement over existing backbone network architectures, such as those used in shipboard and avionics applications. It provides greater connectivity between local area networks, resulting in increased network performance.

FACULTY ADVISORS
Captain Robert J. Voigt, USN
Associate Professor R. Brian Jenkins
Electrical Engineering Department


 
Alex G. Dulude
Midshipman First Class
United States Navy

An Examination of Vortex Ring Entrainment at Low Reynolds' Number

        In recent years, the utility and relative ubiquity of vortex rings in diverse fluid flows have become topics of interest to researchers in many fields. The exploitation of vortex rings in industrial mixing devices and large pulsed jet engines is well-established, and their employment in artificial hearts and aircraft boundary layer control devices has recently been shown to be beneficial. Studies have shown that this phenomenon is the driving mechanism in the propulsion of small marine and flying animals (e.g. squid and hummingbirds). The Reynolds’ dynamic scaling similarities between these creatures and the flow regimes of next generation micro unmanned aerial vehicles (MAVs) and unmanned underwater vehicles (UUVs) has suggested that efficient propulsion systems for these new vehicles could be centered on the generation of vortex rings. For this application, the generation of rings of maximum size and momentum would be critical to propulsive efficiency and system viability. This optimization of vortex rings for propulsion has been linked to the delay of vortex pinch-off, wherein the vortex ring separates from its generator pulse prior to entrainment of all fluid ejecta and momentum. This study examined the effects of generator pulse shaping and duration on the onset of pinch-off using dye injection flow visualization and stereoscopic particle image velocimetry (SPIV).

        The design, construction, and implementation of a novel, piston-cylinder-type, vortex ring generator was accomplished. The results of previous studies were validated at a Reynolds’ Number (Re) of 13,000, where the ability delay pinch-off by 20% of pulse length was demonstrated. The examination of novel pulse shapes at this Re indicated that pinch-off may be delayed by 10% more than previously believed, through the use of the new profiles.

        The effects of pulse shaping and duration were further examined at Re as low as 250 through the use of aqueous glycerin solutions with elevated viscosities. Pulse lengths five times larger than those which resulted in pinch off at a Reynolds’ Number of 13,000 were observed to not experience pinch off at a Reynolds’ Number of 250. Analysis of the SPIV data suggests that the structure of the vortex ring, specifically the trailing portion, is fundamentally different from that of rings at higher Reynolds’ Number. These SPIV images of vortex rings produced with varying pulse lengths are believed to be the first of their kind at Reynolds’ numbers this low.

FACULTY ADVISORS
Assistant Professor David S. Miklosovic
Assistant Professor Eric N. Hallberg
Aerospace Engineering Department


 
Sean A. Genis
Midshipman First Class
United States Navy

Nonlinear Acoustic Landmine Detection:
Profiling Soil Surface Vibrations and Modeling Mesoscopic Elastic Behavior

        This Trident Scholar project involves an investigation of nonlinear acoustic landmine detection, which has shown tremendous promise in recent years. Traditional detection mechanisms are often unreliable: metal detection does not work for landmines with nonmetallic cases and linear acoustic detection is poor at discriminating false alarms. When a buried landmine is exposed to sound energy, the flat top plate of the landmine flexes like a drumhead. This interaction at the soil-mine interface produces a significant, unique, and detectable nonlinear response at the soil surface directly over a buried landmine.

        The observed nonlinearity over a buried landmine results from the interaction that takes place at the soil-mine interface. This boundary is modeled using the soil-plate oscillator (SPO). The SPO is an electro-mechanical resonator in which an electromagnetic inductor is used to drive a clamped polycarbonate plate that simulates the top plate of a landmine. Measurements of the plate’s motional impedance without soil loading allow its stiffness and magnetic forcing factor to be determined. From these elastic parameters the effective mass, damping, and stiffness contributions of a combined soil-plate system is ascertained. The effective system measurements enable the nonlinear contributions of the soil and plate to be separated and examined independently.

        Experiments in nonlinear acoustic landmine detection are characterized by measuring the nonlinear tuning curve behavior of soil surface vibration plotted as a function of frequency. One and two-dimensional profiles of a buried VS 1.6 anti-tank mine and an acrylic mine simulant are conducted inside the anechoic chamber for comparison.    A Laser Doppler Vibrometer (LDV) is configured to pause at specific locations as it scans across a buried mine. At each location the sand is subjected to a series of increasing amplitude acoustic sound pressure level sweeps across a bandwidth that includes the resonant frequency of the buried mine. In all cases, observed tuning curves show relatively small nonlinear effects from “off target” locations, while measurements of the nonlinear response at “on target” locations are identifiably strong.

        Backbone curves (plots of peak amplitude versus corresponding resonant frequency) from locations over the buried mine simulant exhibit linear softening. Classically nonlinear systems, i.e., systems with only atomic elasticity, cannot predict linear backbone curves. This linear behavior, however, is observed in mesoscopic elastic geomaterials such as sandstone. In these materials, the nonlinearity is due to a hysteretic slip-stick mechanism. Observed nonlinearity in the landmine detection experiments is compared to a variety of elastoplastic models for hysteresis. One particular model proposed by Scalerandi et. al. [J. Acoust. Soc. Am., 113, 2003] explains the mesoscopic nanoscale “slip-stick” phenomenon between individual grains in the soil lattice structure using a bi-state protocol of hysteretic elastic elements. An attempt to theoretically predict the experimentally observed nonlinearity is being made using the Scalerandi model.

FACULTY ADVISOR
Professor Murray S. Korman
Physics Department


 
Jordan P. Kehrer
Midshipman First Class
United States Navy

Six-Dimensional Anti-de Sitter Space Solutions to Einstein's Field Equation with Scalar Fields

        The purpose of this Trident Scholar project is to study a scalar field in six-dimensional Anti-de Sitter space by extending the Randall-Sundrum model. The model includes a single scalar field and two compactified extra dimensions. One of these extra dimensions is defined by periodic boundary conditions. The other extra dimension is compactified and stabilized by a scalar field in the space. The shape of the six-dimensional space is defined by its metric, a mathematical structure that describes how the length scale changes as a function of position in space and time. The metric is required to satisfy a differential equation known as the Einstein Field equation. In addition to this requirement, the scalar field must satisfy its own differential equation, the Klein-Gordon equation. Perturbation methods were used to simultaneously solve the Einstein Field equation and the Klein-Gordon equation to find the back reaction of the energy due to the scalar field on the Anti-de Sitter space metric. The physical characteristics of the newly calculated space were explored to ensure that it satisfied the hierarchy problem as well as to determine how the laws of physics were affected by the warping of the space.

FACULTY ADVISORS
Assistant Professor Adam J. Lewandowski
Associate Professor Paul T. Mikulski
Physics Department


 
Daniel C. Ku
Midshipman First Class
United States Navy

Multi-phase Converters and Passive Filtering to Improve Power Harmonics in an Integrated Power System

        This Trident Project’s objective was to compare competing strategies that seek to reduce bus voltage harmonics in a naval warship Integrated Power System (IPS). The subsidiary benefit of this task was to improve the efficiency and minimize the derating factor for engines, generators, and transformers.

        Integrated electric ship propulsion can provide the United States Navy with new war-fighting capabilities and ship design opportunities. An IPS is one in which a common electric bus delivers power to both the ship’s propulsion system and ship service electric system.

        One of the main challenges to the IPS/all-electric ship is the introduction of significant harmonic distortion into the main AC distribution bus caused by power electronic equipment. Power electronic equipment is necessary to implement the variable speed motor drive for propulsion and for power conversion associated with distribution. The harmonic distortion leads to de-rating distribution equipment and degrading the performance of various system loads. As a result, every system attached to the main distribution bus must be able to accommodate the harmonics.

        The method of reducing harmonics was a system based on multi-pulse rectifiers and passive filtering. A multi-pulse rectifier is a power electronic device that converts AC power into DC power. Six and twelve-pulse rectifier systems were simulated and evaluated, as well as constructed and tested in the laboratory on a reduced scale. The effect of adding passive filtering was also analyzed.

        Size, weight, and acquisition cost estimates were derived from vendor data and assessed for feasibility of implementation on an actual destroyer-class warship. This project demonstrates the feasibility of improving power harmonics in an IPS using a system of multi-pulse converters and filtering.

FACULTY ADVISORS
Associate Professor John G. Ciezki
Assistant Professor Thomas E. Salem
Electrical Engineering Department


 
Scott F. Lord
Midshipman First Class
United States Navy

Remote Measurement of High Temperatures in the Presence of a Strong Magnetic Field

        Electromagnetic railguns generate a large amount of heat, enough to melt the surface of the rails upon firing. As such, thermal management systems are crucial in the development of railguns as effective sustainable weapons. In creating a thermal management system, it is important to have knowledge of the temperature distribution throughout the rail as it heats; however, the environment inside a railgun makes conventional temperature sensing techniques ineffective. Large time-varying magnetic fields induce noise into sensors with electrical connections. The high rate of change of temperature requires a fast thermal response and a fast sampling rate. Finally, the intense heat generated requires a sensor that is thermally stable over a large range of temperatures. To overcome such environmental challenges this project proposes an interferometric technique where the temperature is measured remotely with a low power laser and a thin sapphire sensor bonded to the rail. In developing such a sensor the following goals were set upon: no physical connections to the sensor, temperature measurement between 27°C and 800°C, a sampling rate of 10 KHz, and an accuracy of 10% of the total temperature range.

        To meet such criteria an interferometer was constructed from a thin sapphire wafer with a nickel oxide coating on the top side and a nickel coating on the bottom side. A 633nm wavelength laser was directed at the wafer at normal incidence, and the reflection from the sensor was collected with a photodiode. As the sapphire sensor changed temperature its reflectance changed due to variations in the optical properties in the sapphire, nickel, and nickel oxide.

        A computer model was developed that predicts the reflectance at various temperatures utilizing a series of matrices. The model was constructed in a general format so that predictions can be made for adding more layers or changing the optical properties of a layer with relative ease. Furthermore, the model provided a means to determine the optimal thickness of the top nickel oxide layer to maximize the sensor’s sensitivity.

        An optical experiment was configured which demonstrated the operation of the sensor between 26°C and 355°C. The results indicated the sensor response was similar to the computer model predictions. Also, the data collected in the experiment provided insight into possible means to improve the sensor.

FACULTY ADVISORS
Associate Professor Samara L. Firebaugh
Electrical Engineering Department
Associate Professor Andrew N. Smith
Mechanical Engineering Department


 
Ashley S. Myers
Midshipman First Class
United States Navy

Variability and Model Adequacy in Simulations of Store-induced Limit Cycle Oscillations

        This Trident Scholar project investigates the available computational aeroelasticity methods to determine which tool is best suited for the study of limit cycle oscillation of wings carrying external stores.

        A prominent phenomenon of the transonic flight regime is the potential for limit cycle oscillation (LCO), which is a stable oscillation produced by aeroelastic interactions within a component of the aircraft. Such oscillation shortens the fatigue life of aircraft and increases the amount of maintenance necessary. These aspects are of concern to the aerospace industry, particularly with high performance military aircraft. The research here will focus specifically on the aircraft wing and the influence of its external stores on its aeroelastic properties. The heavy Goland wing provides the basic model. External stores are modeled using non-structural masses with ignored aerodynamic effects.

        The research began with a linear structural analysis of a standard aeroelastic wing finite element model. The structural model was modified to allow variations in the stiffness properties of each individual structural element as well as the size and location of non-structural masses to better model a variety of different external store configurations. Specifically, the mass and location of the wing-tip external store were varied to investigate the store’s influence on the occurrence of LCO in the wing.

        The flutter and LCO characteristics of the structural models were assessed using successively more complex computational methods, beginning with the linear tools available in NASTRAN® and ZONA6® before advancing into the transonic small disturbance theory analysis possible within ZTRAN® and finally CAP-TSDv®, the most complex method explored in this research.

        By nature, more complex computational models require more time for processing. They do not, however, always provide more accurate or reliable results. By reusing the same structural model with increasingly complex and refined aerodynamic models, this study develops an understanding of the necessary level of complexity in the computational model to produce reliable results with the least amount of processing time. It also provides baseline assessment of the ability of intermediate fidelity analysis tools to predict the likelihood of LCO.

FACULTY ADVISORS
Professor Gabriel N. Karpouzian
Assistant Professor Chris L. Pettit
Aerospace Engineering Department


 
Jeremie A. Papon
Midshipman First Class
United States Navy

Autonomous Detection and Imaging of Abandoned Luggage in Real World Environments

        The principal goal of this Trident Project is to develop a system that will be able to detect when a bag is left unattended in a crowded scene, such as an airport, using neural networks and other video processing techniques. Video surveillance is commonplace in today’s public areas, but as the number of cameras increases, so do the human resources required to monitor them. Additionally, current surveillance networks are restricted by the low resolution of their cameras. For example, while there is an extensive security camera network in the London Underground, its low resolution prevented it from being used to autonomously identify when terrorists entered the train stations in July 2005. With this in mind, this project endeavors to develop a surveillance system that is able to autonomously monitor a scene for suspicious events by combining a low resolution camera for surveillance (a webcam) with a moving high resolution camera (a 6 mega-pixel digital still-frame camera) to provide a greater level of detail. This enhanced capability can be used to determine whether or not the event is a threat. For the purposes of this research, suspicious events are defined as a person leaving a piece of luggage unattended for an extended period of time.

        Initial analysis of the surveillance video involves separating the foreground (such as people carrying luggage) from the background. Objects identified as foreground are then tracked over time. If an object being tracked separates into multiple objects, a neural network is used to determine if luggage has been left unattended. When an individual separates from the luggage they entered the scene with, an event timer begins. If they do not return within a preset amount of time, a high resolution image of the abandoned item is taken. The final system will be able to detect abandonment events, and using the high resolution camera mounted on a motorized mount, be able to take images of the luggage and flag them for a human supervisor to investigate. This security system is adaptable to many other applications, and could be modified to address the problem revealed by the London subway bombings of identifying terrorists before they are able to cause harm.

FACULTY ADVISORS
Associate Professor Robert W. Ives
Electrical Engineering Department
Assistant Professor Randy P. Broussard
Weapons and Systems Engineering Department


 
Matthew P. Shearer
Midshipman First Class
United States Navy

Simulation and Evaluation of Marine Propeller Crashback Through Computational Fluid Dynamics (CFD)

        This Trident Scholar project focuses on the modeling of the flow field during crashback using computational fluid dynamics (CFD). Crashback occurs when a ship or submarine traveling in the forward direction reverses its propeller to aid in slowing or stopping. This maneuver results in unsteady forces and moments that act on the vessel, significantly decreasing maneuverability and controllability. The goals of this project are to utilize the CFD results to visualize and evaluate the flow field during crashback in order to determine the physical source of the unsteady forces and moments.

        Computational methods provide the capability to investigate larger flow field volumes compared to experimental methods such as Laser Doppler Velocimetry (LDV) and Particle Image Velocimetry (PIV), enabling researchers to evaluate results throughout the control volume, rather than only small portions of the flow. In the fall of 2006, crashback results were obtained for an advance ratio of -0.5 using a CFD technique known as Large Eddy Simulation (LES). These results were compared with force data (including thrust, side force and torque) obtained through water tunnel experiments performed in the 36-inch Variable Pressure Water Tunnel located at the Naval Surface Warfare Center in Carderock, MD. The mean, root mean square, and standard deviation values for the forces calculated using LES correlated as expected with the forces measured in the water tunnel, indicating that the CFD model simulated crashback effectively. Spectral analysis was also performed on both the LES results and the water tunnel data, which showed that the frequencies with which the side forces acted on each propeller were very similar as well.

        Once the LES was shown to accurately represent the effects of the fluid flow during crashback, the flow field itself was evaluated to extract the physics of the flow in hopes of determining the source of the unsteady side forces. This stage of the project consists of generating three-dimensional animations of the flow field from the LES results. One series of animations focuses on the flow separation over the propeller blades at several radii to determine whether the separation has any association with the unsteady forces. The second animation shows the formation and dissipation of a pressure isosurface, representing the formation and shedding of a vortex ring near the propeller blade tips, and its correlation to the side force acting on the propeller. The final animation shows an isosurface of circulation, in hopes of determining whether there is a circulation limit at which the vortex ring is shed. The evaluation of these animations in three dimensions proves to be a significant advantage over PIV and LDV data (which can only be obtained for very small portions of the control volume) and provides more insight into the physics of the flow field during crashback. The LES has already been shown to accurately represent the effects of the crashback flow field, and it will hopefully present some insight into the physical cause of the unsteady forces acting on the propeller.

FACULTY ADVISORS
Professor Gabriel N. Karpouzian
Aerospace Engineering Department
Commander Jeffrey W. Stettler, USN
Naval Architecture and Ocean Engineering Department

PROJECT COLLABORATOR
Dr. Peter A. Chang, III
Naval Surface Warfare Center (NSWC) - Carderock Division


 
Erik T. Smith
Midshipman First Class
United States Navy

Swarm Manipulation of Large Surface Vessels

        The goal of this Trident project is to develop an independent control scheme to allow a team of autonomous tugboats to move a large disabled vessel, such as a barge, to a desired position and orientation. Independence refers to the extent to which each tugboat’s actions are free from exact knowledge of the locations and actions of other tugboats. Performance of the team will be quantified by measuring the time required to affect the motion, while respecting maximum power constraints on the thrust. Applications of the project include difficult or dangerous tasks such as moving disabled vessels or vessels “not under command” through hostile or dangerous areas, and transportation of large objects such as marine construction equipment, off-shore bases, drilling platforms, and sonar arrays.

        Although it would be ideal to increase both the independence and performance of the system, it must be realized that by increasing one of these, the other is typically degraded. In order to maximize performance, a centralized decision maker must know the position of every boat, and distribute optimal thrust commands throughout the group. However, this architecture is not desirable, since it is less independent of inaccurate tugboat position information or the failure of a single tugboat. In contrast, true distributed operation implies that each tugboat knows nothing about the position or thrust vector of its teammates. While offering the ultimate with regards to independence; true distributed operation comes at the expense of performance. A series of scenarios ranging from centralized to truly distributed operation are being investigated. In the course of solving them, the tradeoff between performance and independence are being quantified.

        It appears that this is the first study of its kind and complexity. Although previous work has studied some of the scenarios in this project, its extent and focus is not close to this research. Each tugboat uses on-line adaptive control methods to compensate for the unknown actions of other swarm members. The analysis is being verified in simulation. In addition, an experimental proof-of-concept device is being built and in-water experiments are being used to validate the results. An incremental approach to experiment design is being used to mitigate the challenges of in-water experimentation.

FACULTY ADVISORS
Assistant Professor Matthew G. Feemster
Assistant Professor Joel M. Esposito
Weapons and Systems Engineering Department


 
Reid W. Smythe
Midshipman First Class
United States Navy

An Examination of Acceptable Navigation Accuracy for LISA Orbits

        This project assesses the accuracy with which a particular constellation of satellites, the Laser Interferometer Space Antenna (LISA), must be placed into orbit. The LISA formation will consist of three satellites orbiting the Sun at approximately 1 AU, forming an equilateral triangle with leg lengths of 5 million kilometers while maintaining an angular separation of 22.5 degrees from the Earth. Three distinct phases composed this assessment.

        The first phase dealt with formation parameters (leg length, leg length time rate of change, interior leg angle, and formation, sun, earth angle) as a function of time. Duplication of plots contained in other papers on the topic using independently written orbit data analysis scripts in MATLAB® validated the output of those scripts. Preliminary graphical analysis indicated the values of each parameter varied in a sinusoidal manner, and changing the initial conditions tended to reduce the time each parameter fell within the acceptable tolerance values.

        The second phase dealt with analyzing formation parameters as a function of the error size of specific initial conditions. Up to two of the eighteen state variables were varied simultaneously, resulting in a surface indicating how long a particular formation parameter was out of specification over the lifetime of the mission. These surfaces indicated general trends in terms of the sensitivity of the formation to certain specific errors. The formation appeared to be most sensitive to velocity errors in the in-track direction, with position errors in the radial direction having the next largest effect. Furthermore, it appeared that one error can counter-act another error, agreeing with other papers on this topic.

        The third phase analyzed the effect of varying all eighteen variables simultaneously. The normalized time out of specification for leg length time rate of change over the life of the formation was used as the metric. A plot was created to show how the value of this metric changes as the initial state variable tolerance changes. As the maximum variation in position and velocity errors increased, there tended to be a corresponding increase in the amount of time the formation parameter was out of specification. Using the data from this plot, the LISA team can ensure that the propulsion vehicle placing the satellites into orbit will have the accuracy required to meet the requirements for the science mission.

        Further work will involve the development of an analytic solution to complement the numerical analysis, as well as a numerical study on the effect of errors in tracking in terms of the differences between the perceived formation parameters and the actual formation parameters.

FACULTY ADVISORS
Professor Daryl G. Boden
Visiting Professor Richard P. Fahey
Aerospace Engineering Department


 
Andrew J. Turo
Midshipman First Class
United States Navy

Dielectric Studies in Materials for High Energy Density Capacitors

        This Trident Scholar project investigates the electrical properties of dielectric materials considered for use in high energy density capacitors and seeks to improve upon these properties. The energy requirements of technologies today are more demanding than ever. The requirement for large energy storage and a quick release of that energy is becoming a greater and greater problem. Limited by size, cost, and the speed of chemical reactions, batteries need to be replaced as the primary component for energy storage. Capacitors have the ability to perform in all the areas batteries and other energy sources fall short.

        A basic capacitor is two parallel conductive plates in a vacuum on which charge is placed, creating an electric field which stores energy. One way to increase the amount of charge that can be placed on the plates is by inserting a dielectric material. The dielectric materials polypropylene and Ultem® were the focus of this study, polypropylene as a control and Ultem® due to its ability to handle high temperatures, its structural stability, and its good dielectric characteristics. Ultem® was studied in both its pure form and with the addition of a filler particle called mesoporic silica, which theoretically should lead to an increase in dielectric performance while maintaining structural stability. This was done through the casting of Ultem® films which underwent dielectric breakdown testing and broadband dielectric spectroscopy.

        Broadband dielectric spectroscopy was done by attaching electrodes to the material and analyzing the charge and current flow through the material at different frequencies and temperatures. The data produced allowed the calculation of the dielectric constant and loss and gave insight into these values. Dielectric strength measurement is a new experiment at the U.S. Naval Academy, so part of this project was to develop the methods to be used. This involved developing an acceptable sample holder, sample preparation techniques, voltage ramping procedures, dielectric breakdown test procedures, and validations of the test equipment.

        Expected results include a verification of the developed dielectric breakdown technique by the study of polypropylene and commercially produced Ultem® and comparison with published data. Also, comparisons of solution cast Ultem® and commercially produced materials should validate the solution casting process used to create samples. The remainder of the study is an investigation of the effects of mesoporic silica upon the dielectric properties of solution cast Ultem®.

        The overall hope for this project is that the addition of mesoporic silica will produce an improvement in the dielectric properties of Ultem® resulting in it reaching a level considered acceptable for high energy naval applications. However, regardless of the effects of the filler particles, this project will result in an in-depth characterization of the material Ultem®, which serves as a basis for future research and application of the material.

FACULTY ADVISOR
Associate Professor Charles A. Edmondson
Physics Department


 
David G. Underhill
Midshipman First Class
United States Navy

Exploring Dimensionality Reduction for Text Mining

        The goal of this project is to determine how various data dimensionality reduction techniques affect the quality of conclusions which can be drawn from text mining. This is being achieved by comparing the results of a variety of classification and automated discovery experiments. These results will be important because they will help further the understanding of how dimensionality reduction can be used in conjunction with text mining to effectively handle large quantities of data which cannot be analyzed to their full potential due to information overload.

        Information overload is common; millions of potentially relevant database records, web pages, communications, and other documents are in danger of being overlooked. A number of modern Naval missions require the prompt fusion and accurate interpretation of information from disparate sources. These missions are challenging because the information that needs to be fused is not simply mathematical, like target track information, but information about changing groups of loosely organized individuals. In this situation, as with many others, the information that needs to be fused exists primarily as free-form text.

        The free form nature of this information necessitates the use of text mining methodologies for analysis. Text mining is the process of automatically extracting information from unstructured text. One application, Literature-Based Discovery (LBD), is intended to automatically uncover interesting relationships between documents. A challenge with text mining is the complexity of a document and the process by which this complexity is simplified into a more manageable representation. This process, known as dimensionality reduction (DR), has been studied in many contexts. Recently, several promising new techniques for performing this reduction have been proposed. These techniques may help effectively reduce vast amounts of information while maintaining interesting details, but have yet to be adequately evaluated for text mining.

        This investigation evaluated the effectiveness of several methods for DR as they relate to various text mining processes. Experiments were conducted to determine how document conditions affect each of these approaches to DR. Evaluation of these methods was based on the ability of standard algorithms to effectively classify documents among known categories. This study also evaluated these techniques with regards to LBD. In particular, the LBD process was used to identify pairs of documents that appear similar after DR has been performed, yet are known to come from different categories. Such pairs are candidate LBD discoveries. These candidates are being evaluated by automatically extracting keywords and using those keywords to query public web services. The quality and number of related documents return by the web services are being used to evaluate LBD effectiveness. It is expected that some newer DR methods which stress local relationships will perform best. These results obtained in this project will facilitate future evaluation of DR on LBD, and may also provide insights into the best ways to use DR for text mining processes in general.

FACULTY ADVISOR
Assistant Professor Lucas K. McDowell
Computer Science Department
go to Top