2007 Bowman Scholars
Division of Engineering and Weapons
- 1/C Travis M. Albright
- Major: Mechanical Engineering
- Title: Monte Carlo Modeling of DT-702 Thermoluminescent Dosimeter (TLD) Response Characteristics
- Advisor: Professor Martin E. Nelson, Mechanical Engineering Department Abstract
Monte Carlo Modeling of DT-702 Thermoluminescent Dosimeter (TLD) Response Characteristics
The objective of this research project is to model the current Navy standard DT-702 thermoluminescent dosimeter response to gamma photons using Monte Carlo modeling techniques. Computer simulation of the DT-702 response will be accomplished using MCNP5 (Monte Carlo N-Particle Transport Code, version 5) modeling software, and will be validated by comparison with experimental data that has been previously collected at the National Institute of Standards and Technology (NIST). This project will expand on the work of a previous Bowman Scholar researcher by analyzing the gamma response in addition to the beta response previously studied.
The project will involve the development and analysis of a detailed MCNP5 model geometry for the DT-702 NIST traceability irradiation procedure. Visual editing software, to include the Mortiz and Sabrina packages, will be used to create the geometry cards for the MCNP5 input file. The dosimeter card and holder will be modeled on phantom within a geometry model of the NIST irradiation facility and source using a parallel cluster MCNP5 Workstation at the Naval Dosimetry Center. The high number of particle histories made possible by the parallel cluster will ensure the statistical significance of the results. An ionization chamber will be modeled within in the MCNP5 geometry model and analyzed to provide a secondary dose metric for comparison to the imperical NIST data and MCNP5 dose tallies. The overarching objective for the project will be to understand more accurately the relationship between different forms of incident radiation and corresponding thermoluminescence of the doped LiF chips. Ultimately, it is hoped this understanding can be applied to suggest improvements to the dose algorithm for the DT-702 system in complex, mixed dosage situations.
- 1/C Natalie E. Alford
- Major: Mechanical Engineering
- Title: Rapid Metal Prototyping for Naval Applications
- Advisor: Professor Angela L. Moran, Mechanical Engineering Department
- Advisor: Associate Professor Richard E. Link, Mechanical Engineering Department Abstract
Rapid Metal Prototyping for Naval Applications
This research project involves a comparative study between the material properties of a naval machine component produced via rapid prototyping and the properties for the component fabricated using traditional manufacturing methods. Specifically, selective laser sintering (SLS) is a rapid manufacturing method which uses a high powered laser to fuse metal or polymer particles into a three dimensional part directly from virtual form. Initially, mechanical property tests will be performed on standard test specimens to establish expected tensile strength, ductility, and impact toughness for the polymeric and metallic materials used in the SLS process. A part with complex geometry will be selected, modeled, fabricated, and inspected in terms of dimensions, tolerances, and finish quality. Solid models of the part will be created using SolidWorks® 3D mechanical design software, and then several copies of the part will be constructed from high strength polymer and steel. The components will be evaluated through tensile testing, impact toughness testing, and microstructural characterization. Results will be compared with the material properties of the same component produced by traditional fabrication routes. This investigation will contribute to the feasibility study of using rapid metal prototyping for part on call manufacturing and the replacement of metal castings on ship and submarine systems with parts fabricated via selective laser sintering. The effectiveness of the process will be judged on ease of use and the ability of the components to meet Navy standards for new materials.
- 1/C Gregory A. Borek
- Major: Mechanical Engineering
- Title: Agent-based Modeling Approach to Complex Adaptive Systems in Humanitarian Operations
- Advisor: Capt Koichi Takagi, USMC, Mathematics Department Abstract
Agent-based Modeling Approach to Complex Adaptive Systems in Humanitarian Operations
This research project explores the domain of possibilities for a humanitarian aid mission, by modeling the effect of a platoon-sized element of individuals trying to distribute food to a crowd that includes some hostile elements representing terrorists or other dissidents. The model will simulate the struggle for influence over the neutral elements of the crowd by the soldiers and the terrorists in order to fulfill their aims; the soldiers’ goal is to distribute the greatest of food among the crowd as possible despite the presence of hostile elements that may attempt to disrupt the process.
The project employs the Map Aware Non-uniform Automata (MANA) program developed by the New Zealand Defense Technology Agency. MANA was developed as part of the software suite known as Project Albert to test the influence of varying agent parameters such as probabilities of kill, range, and communications capabilities on the outcome of conventional military operations. MANA is used at the Marine Corps Warfighting Laboratory and at the Naval Postgraduate School to simulate a wide range of agent-based operations. The term “agent” was adopted from economics by John Holland to describe the behavior of the simplest repeated entity in a system (Hidden Order, 1995).
What makes MANA unique from other military models such as JANUS is its attempt to model “intangibles” in combat. Since its inception, the main focus of science has been to discover mathematical relationships for simple systems that could then be applied to more complex systems. The “Butterfly Theory” by Edward Lorenz in the early 1960’s, however, proposed that small perturbations in the initial conditions for a system bounded by mathematical constraints often lead to dramatic differences in outcomes. In simple terms, the world is often much more complex than the linear equations used to describe it. No matter how meticulous a model purported to be, “intangibles” such as situational awareness, fatigue, and morale were discarded since they played no part in an equation. The result of this top-down modeling approach was that simulations that were developed became deterministic with agents the behaved unnaturally. In contrast, MANA utilizes a bottom-up approach by focusing on the interactions between agents while manipulating agent properties in an attempt to capture real-world nonlinearity.
This is not to say that MANA agent-based distillations are without their own set of problems. Situations modeled in such an environment are highly specific and are not intended to be a highly detailed analysis of every phase of military operations. Additionally, agents often make mistakes and are input into the simulation without any sort of intelligence to discover the optimal plan. However, MANA can still be used to determine the effectiveness of given plans since it is the effects of the “mistakes” that are being modeled. With this in mind, this research project will use MANA to evaluate a variety of tactics, techniques, and procedures (TTP’s) using statistical analysis to determine the positive and negative aspects of several approaches to a humanitarian aid mission with the intent of discovering an optimal set of TTP’s. By modeling a platoon sized element in a real-world setting, the analysis of the outcomes would prove invaluable in future operations.
- 1/C Anthony R. Bracalente
- Major: Ocean Engineering
- Title: Diver Thermal Protection via a Metal Hydride Reaction
- Advisor: LCDR David C. Robinson, USN, Naval Architecture and Engineering Department
- Advisor: Professor Keith W. Lindler, Mechanical Engineering Department Abstract
Diver Thermal Protection via a Metal Hydride Reaction
This research project involves an investigation of the use of metal hydrides for use in Diver Thermal Protection. Navy Divers such as SEALS and EOD often have to dive in very cold water or very warm water and must wear protection. Currently, on station time is reduced due to the extreme temperatures in which the divers are subject. The goal of this project is to show that metal hydrides can provide enough heating and or cooling power to offset the negative effects divers experience while diving in cold and or warm water. These negative effects include fatigue both mentally and physically as well as low productivity and safety concerns.
This project investigates the use of metal hydride thermal batteries as the heat source and/or sink for water that will be pumped through a suit worn by the diver. A metal hydride thermal battery is a pressurized system that contains two dissimilar metals both of which are able to absorb hydrogen and store it efficiently. The battery contains a high pressure and low pressure side separated by a valve. Once the valve is opened and the hydrogen begins to flow from one side to the other, the heat of absorption and of desorption of the metals cause the high pressure hydride to cool rapidly while the low pressure side increases in temperature. Preliminary studies have shown that a temperature difference of 15°C will be sufficient to negate the environmental effects placed on the diver.
The efficiency and usability of the thermal battery will be determined by running experiments on metal hydride thermal batteries. These experiments will take place in a constant temperature water-bath where the temperatures and pressures of the hydrides can be accurately monitored and recorded throughout the course of the experiment. From this information, the rate of absorption and desorption can be calculated and the feasibility of a prototype can be determined.
- 1/C Mark E. Daniel
- Major: Electrical Engineering
- Title: Universal Wavelength Multiplexed Optical Control Node
- Advisor: CAPT Robert J. Voigt, USN, Electrical Engineering Department
- Advisor: Associate Professor R. Brian Jenkins, Electrical Engineering Department Abstract
Universal Wavelength Multiplexed Optical Control Node
The goal of this project is the development of a Universal Wavelength Multiplexed Optical Control Node for use in shipboard and avionic fiber optic networks. This network provides a protocol independent path between sensors, networks and systems onboard naval ships and aircraft. The control node would allow for distributed, hierarchical, out-of-band control of the data traveling between potentially heterogeneous local area networks (LANs). Data traffic remains local within its LAN but has the ability to be transported across the hierarchical access network to a destination LAN. The control node also allows for the transmission of mixed signal data, enabling digital and analog data to propagate within the same fiber simultaneously. This is of great interest to the Navy because it would allow for networks, such as a fire control network, to have analog radar data propagating in the same fiber as the digital missile guidance data. The control signal for the data, used for routing the information to its destination, will be transmitted on the same fiber as the data itself. However, the control signal will be transmitted on a distinct wavelength band and remain completely separate from the data, implementing out-of-band control. The control node uses a wavelength converter to achieve reconfigurability within the backbone of the network.
The project begins with the construction of a single control node which involves two wavelength converters and two Bidirectional Add/Drop Multiplexers. Four of these control nodes would be linked together in a crossbar topology to create the backbone of the avionics network. The crossbar allows for a direct connection between any two nodes on the backbone. Each of the four backbone nodes would be connected a LAN consisting of some number of destination nodes. A control algorithm will then be programmed to route the data from source to destination as it travels between different levels in the hierarchy.
The node, when implemented in a fiber optic local area access network, would have many advantages over networks that are in use today. Traffic within the node would be bidirectional, increasing the reliability, reconfigurability and fault tolerance of the network. The increased bandwidth and ability to transmit mixed signal data would allow existing networks to be reduced to a single fiber, greatly reducing the number of connections and weight of the network infrastructure.
- 1/C Philip A. Johnson
- Major: Systems Engineering
- Title: An Investigation of Interceptor Guidance Methods through Modeling and Simulation
- Advisor: Associate Professor Richard T. O'Brien, Jr., Weapons and Systems Engineering Department Abstract
An Investigation of Interceptor Guidance Methods through Modeling and Simulation
The purpose of this project is to investigate a new method of missile interceptor guidance that may be more effective against a maneuvering target. The U.S. Department of Defense is committed to the creation of a multilayered net of protection against all types of ballistic missile threats. New developments in missile technology, such as the ability to perform evasive maneuvers, may cause current interception guidance techniques to become ineffective.
The present work is an initial stage in the evaluation of a relatively new technique of interceptor guidance called Angular Acceleration Guidance (AAG). A kinematic model has been developed that allows for the evaluation of various guidance methods including AAG against targets with non-constant acceleration. The kinematic model was first verified through its behavior during a constant velocity intercept.
Current work on this project involves testing the limits of proportional navigation (PN), the most common method for missile guidance currently in operational use. PN is adequate for conventional interception of slow moving targets such as airplanes but is limited in its ability to track targets with large variations in acceleration. The simulations used to verify the kinematic model will be adapted to include a non-constant target acceleration term. A set of metrics, such as divert velocity vs. velocity error and divert velocity vs. acceleration error, will be used to make a quantitative comparison of PN and AAG models. This will provide insights into the limits for PN and possible applications of AAG. Future work will focus on the performance and limits of AAG and the calculation of its variable guidance gain term. An investigation into the methods for calculating that term will be conducted and possible new approaches for calculating the term will be considered.
- 1/C Scott F. Lord
- Major: Electrical Engineering
- Title: Remote Measurement of High Temperatures in the Presence of a Strong Magnetic Field
- Advisor: Associate Professor Samara L. Firebaugh, Electrical Engineering Department
- Advisor: Associate Professor Andrew N. Smith, Mechanical Engineering Department Abstract
Remote Measurement of High Temperatures in the Presence of a Strong Magnetic Field
This project involves an investigation of measuring temperatures in an environment similar to that of an electromagnetic railgun. The goals of the project are to build, calibrate, and finally test an interferometric temperature sensor.
The sensor is to be constructed from a thin sapphire wafer which will be monitored by a low power laser. The laser is used to measure changes in the wafer’s optical properties which correspond to changes in the wafer’s temperature. To craft the sensor wafer, a 100 micron sapphire wafer will be purchased from a standard vendor. Next, the micro-fabrication facility will be used to deposit reflective metal onto the surfaces of the wafer via metal vapor deposition. The wafer will then be attached to a heating block adjacent to a thermocouple to evaluate the reflected light intensity at various known temperatures. From that data a function correlating light intensity to temperature will be constructed. Finally, the sensor will be taken to Naval Research Laboratories (NRL) where it will be tested under pulsed power conditions comparable to an electromagnetic railgun’s environment.
The principle behind the sensor is that some laser light will reflect off the top surface of the wafer while some of the light will pass into the wafer, reflect off the bottom side of the wafer, and rejoin the initially reflected light. The laser light that rejoins the initially reflected light will have traveled a longer path length and as a result be out of phase with the initially reflected light. Being out of phase, the rejoining light will generate interference which alters the total reflected light intensity. As the wafer changes temperature, the thickness and refractive index of the wafer change, changing the path length traveled by the light inside the wafer and thus the amount of interference it causes. The change in reflected light intensity caused by changes in interference is what is ultimately measured and then correlated to a change in temperature.
- 1/C William P. Murphy
- Major: Aerospace Engineering
- Title: Engine Mapping and Water Ingestion in the SR-30 Turbojet
- Advisor: CDR David D. Myre, USN, Aerospace Engineering Department
- Advisor: Professor Martin R. Cerza, Mechanical Engineering Department Abstract
Engine Mapping and Water Ingestion in the SR-30 Turbojet
This project involves an investigation of the effects of water ingestion through a small turbojet engine. Engine performance of the SR-30 turbojet running dry will be compared to the engine's performance with various amounts of water being atomized and injected into the inlet. A metal mesh will be placed at the exit plane of the engine and the thermal plume will be observed using an infrared camera. Observations of this plume will be used to analyze the performance of the engine.
The SR-30 turbojet engine is a small jet engine used in a classroom laboratory environment. It is located in the United States Naval Academy Propulsion Laboratory, and has not been evaluated in detail or performance mapped. Data indicates that the exit temperature probe of the engine is currently misaligned, thus generating erroneous data during laboratory exercises. As a result, prior to mapping the efficiency of the SR-30 or conducting any experimentation with atomized water, the temperature probe at the exit of the turbine will be repositioned. The position of this probe, which is used to measure the exit temperature of the SR-30 engine, must be corrected experimentally. This will involve a hands-on approach to determine how the probe is misaligned or is otherwise inaccurate. Once the probe has been repositioned, the engine temperature and pressure data will be recorded, the component efficiencies will be calculated, and the water ingestion experimentation will be conducted.
Thermodynamic analysis of SR-30 will be conducted to estimate what the optimum temperature should be, and to provide preliminary component and cycle efficiency estimates. Thermodynamic analysis will provide a calibration against the probes' repositioned temperature readings.
Initially, data will be taken from the SR-30 without any water injection system or mesh grid. This data and the method in which it is taken will be used for familiarization with the components of the system and performance of the engine. The first portion of this project will be devoted to the physical development of the engine, and the completion of the initial calculations needed in order to begin experimentation. The physical construction of the system includes building a water injection system, testing nozzles for atomization, and adding the metal mesh. After the system is constructed, testing will begin and the analysis of the system to explore the effects of atomized water on a turbojet will be investigated.
- 1/C Jeremie A. Papon
- Major: Electrical Engineering
- Title: Development of a High Recognition Security Camera System
- Advisor: Associate Professor Robert W. Ives, Electrical Engineering Department
- Advisor: Assistant Professor Randy P. Broussard, Weapons and Systems Engineering Department Abstract
Development of a High Recognition Security Camera System
After the London subway bombings of 7 July 2005, British authorities attempted to use the extensive security camera network located in the subway system to identify suspects. Unfortunately, the cameras were static and provided only very low resolution images in which subjects’ faces were practically indistinguishable. There is a distinct need for the development of a security system that is able to actively locate individuals in a busy scene and to acquire high resolution images of their faces. Such a system would allow authorities to essentially have a mug-shot of every individual who passed by the camera. These images could be used for identification by authorities or possibly integrated into a facial identification algorithm to automatically identify them.
The primary goal of this project is to develop a complete surveillance system that integrates a low cost, wide angle, low resolution camera with a high resolution digital camera in a complete, self-contained package. The system will use a high speed pan tilt mount to move the high resolution camera, to focus on any point in the low resolution camera’s field of view. The low resolution camera will input into a program on a computer that uses pattern recognition techniques to distinguish where individual faces are in the field of view. When a face is detected, the program will command the pan tilt mount to move so that the high resolution camera can take a close up image of the subjects’ faces.
Once a high resolution image is taken, it will be sent over a network to a central database which will keep a log of all individuals seen. Planned testing for future work involves the set up of one or several of the surveillance cameras in a building of the Naval Academy with high pedestrian traffic. The system will run continuously with data being sent constantly back to a central server in the laboratory. This data acquisition will provide real world data to assess the system’s practicality in a high traffic area. Future testing will also encompass a multitude of experiments which will have various predetermined numbers of people pass by the system at various speeds in order to find successful capture rates, acquisition speeds, and the effect of movement speeds, number of subjects in scene, separation of subjects and facial obstructions such as beards or sunglasses.
- 1/C Matthew R. Surprenant
- Major: Systems Engineering
- Title: Reconfigurable Control of an Electric Warship Thermal Cooling System
- Advisor: Associate Professor Edwin L. Zivi, Systems Engineering Department
- Advisor: Assistant Research Professor Yonggon Lee, Systems Engineering Department Abstract
Reconfigurable Control of an Electric Warship Thermal Cooling System
The focus of this project is to develop the control algorithms of a shipboard thermal cooling system. In particular, the design of one of several dynamically interdependent electric warship systems will be investigated. Moreover, the research will explore the use of the genetic algorithm as a tool to design and refine control systems.
One of the Navy’s major research initiatives is finding ways to reduce the crew size of ships. One particular way to accomplish this goal is through the automation of shipboard systems. This research will focus on the thermal cooling system of a ship. The primary research challenge is the development of survivable automation after a single weapon hit.
The goal of this research is to develop a dynamically reconfigurable control strategy that maximizes cooling fluid delivery to critical ship systems despite combat induced damage to the ship. The research will use an electric warship model already developed through an Office of Naval Research sponsored collaboration between the Naval Academy, Purdue University and the Massachusetts Institute of Technology. Specifically, the notional ship model created by the Integrated Reconfigurable Intelligent Systems (IRIS) research project will be the basis of the research. This model contains a three zone ship including spatial arrangements and dynamically interdependent electrical and thermal systems. The thermal layer contains a saltwater distribution network, fresh water cooling loops as well as the power components and heat exchangers. Because of the coupling between these systems, the model supports the exploration of the design of dynamically interdependent systems.
The IRIS research also provides innovative operability and dependability performance metrics that quantify the continuity of vital services under disruptive conditions. These performance metrics will be used in this research to guide the genetic algorithm optimization of the control system. Genetic algorithm optimization will be used to find solutions which are robust with respect to combat induced disruption. In particular, the MATLAB® based Genetic algorithm Optimization System Tool (GOSET) will be used to optimize thermal cooling system performance metrics.
The thermal cooling system control algorithm must be dynamically reconfigurable to react to changing damage and mission scenarios. Also, the controller will be hierarchical with graceful degradation, which means that the system will continue to function even with the loss of the main controller, because zonal controllers will be capable of operating their individual zones. Lastly, the controller must be able to allow the ship to carry out its mission to the maximum extent possible.
- 1/C Nathan S. Tyler
- Major: Systems Engineering
- Title: Electric Warship Thermal Cooling Prototype
- Advisor: Associate Professor Edwin L. Zivi, Weapons and Systems Engineering Department
- Advisor: Assistant Research Professor Yonggon Lee, Weapons and Systems Engineering Department Abstract
Electric Warship Thermal Cooling Prototype
With the increase of automated systems onboard warships, the need for these systems to remain operable in the event of a casualty also increases. This project involves building a bench-top model of a cooling system for an automated, survivable power system for a notional electric warship. The model will demonstrate the interdependence of the power and the fluid cooling systems for survivability with respect to continuity and recoverability of service. Current shipboard casualty response practices emphasize redundancy and manual intervention. Given the need to accomplish more with less, automated casualty reconfiguration must proceed despite casualties and in the absence of manual intervention. This level of dependability requires that the elements are available, robust and reconfigurable under abnormal conditions.
The bench-top model will exhibit the removal of heat through two separate networks: a primary seawater network and a secondary freshwater network cooled via the seawater network. These two networks will be configured into three machinery zones. Each zone will include: (a) heat generation power components, (b) fresh water distribution and heat exchangers; and (c) seawater distribution and heat exchangers.
This scaled model of a notional electric warship will use scaled heat loads from the Naval Combat Survivability (NCS) test bed located at Purdue University. In the research, the cooling system is viewed as service provider with an emphasis on continuity of service despite battle damage disruption. The ultimate objective is to contribute the development of survivable electric warships.
Division of Math and Science
- 1/C Alexander R. Ludington
- Major: Physics
- Title: Search for Triaxial Deformation in Tantalum-167
- Advisor: Assistant Professor Daryl J. Hartley, Physics Department Abstract
Search for Triaxial Deformation in Tantalum-167
By observing the gamma rays emitted by the excited 167Ta nucleus, a “wobbling mode” may be observed, which indicates that the nucleus has a very unusual, asymmetric shape.
Not all nuclei are spherical; indeed, some are prolate (shaped like a football) and others are octupole (shaped like a pear). Spherical, prolate and octupole nuclei all have a symmetry axis. However, one rare shape has no symmetry axis whatsoever, and it is called triaxially deformed. When rotated, these nuclei are said to wobble in a manner similar to the spinning of an asymmetric top. The top will spin, but due to its asymmetry, the axis of its spin continually changes. Only lutetium nuclei, with 92, 94 or 96 neutrons, have shown this behavior. Theory suggests that 94 neutrons could play a key role in creating this rare shape of the nucleus. No wobbling has been observed outside lutetium. Other nuclei have been investigated but have not shown any evidence of triaxial deformation. This experiment will produce the 167Ta nucleus, which has not been chosen as an experimental subject for about 15 years.
For the experiment, 51V was accelerated to an energy of ~230 MeV and bombarded onto a 120Sn target to produce four neutrons and 167Ta in a highly excited state. This means that the 167Ta nucleus gives off energy in the form of gamma rays as it decays to its ground state. Because the nucleus is a quantum system, only certain energy levels are allowed. Those levels can be identified from the frequencies of the gamma rays that are detected. Analysis of the various states of the nucleus, which will be constructed from the observed gamma rays, will reveal information about the shape of the nucleus. The experiment just described was conducted at the Wright Nuclear Structure Laboratory at Yale University in August 2006, where they have nine detectors. All of the data analysis will be conducted during this project at the Naval Academy using existing equipment and programs. In the future, a further experiment may be conducted at Gammasphere, an array of 100 detectors.
- 1/C Josh L. Nickerson
- Major: Applied Mathematics
- Title: Modeling Bubble Growth in Micro-channel Cooling Systems
- Advisor: Associate Professor Sonia M. Garcia, Mathematics Department
- Advisor: Professor Martin R. Cerza, Mechanical Engineering Department Abstract
Modeling Bubble Growth in Micro-channel Cooling Systems
The goal of the project is to develop a mathematical model for bubble growth in micro-channel cooling systems employed in advanced microchip designs. With increased computing power, microprocessors produce more heat which must be carried away to prevent the circuit-board from melting. Conventional systems employ simple air-convection fans to accomplish this; however, cooling via micro-channels embedded in the surface of a microchip could allow for much more computing power. Heat is transferred from the microchip into the flowing micro-channel fluid. Understanding the formation and growth of bubbles within micro-channels is crucial for the development of this cooling system.
The project involves solving several partial differential equations, including the heat conduction equation in the liquid layer under a bubble’s base, as well as solving for the temperature profile beneath the bubble’s base. Using the solutions of these equations, a MATLAB® program will be constructed to generate data which can be analyzed and compared to experimental results. The final product will be a mathematical model for bubble growth in a micro-channel, which would aid in the development of this new technology. If microprocessors employing micro-channel cooling systems perform according to theory, the field of high-performance computing would be revolutionized. Possible Naval applications of such technology include advanced RADAR systems.
- 1/C David G. Underhill
- Major: Computer Science
- Title: Exploring Dimensionality Reduction for Text Mining and Literature-based Discovery
- Advisor: Assistant Professor Lucas K. McDowell, Computer Science Department Abstract
Exploring Dimensionality Reduction for Text Mining and Literature-based Discovery
The goals of this project are to determine how various data dimensionality reduction techniques affect the quality of conclusions which can be drawn from text mining. This will be achieved by comparing the results of a variety of classification and clustering 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 analyze very large quantities of data.
Individuals, companies, and governments are surrounded by millions of potentially relevant database records, web pages, communications, and other documents that are in danger of being overlooked. In particular, a number of modern Naval missions require the prompt fusion and accurate interpretation of large amounts 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 help 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, has been studied in many contexts. Recently, several promising new techniques, such as Lafon’s spectral method, 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 work will evaluate the effectiveness of several methods for dimensionality reduction as they relate to various text mining processes. Experiments will be conducted to determine how document conditions affect each of these approaches to dimensionality reduction. How the dimensionality reduction impacts the ability of standard algorithms to effectively classify documents among known categories and to cluster such documents when categories are unknown will also be studied. It is expected that some newer dimensionality reduction methods which stress local relationships will perform best. These results will facilitate future evaluation of dimensionality reduction on LBD, and may also provide insights into the best ways to use dimensionality reduction for other text mining processes.