Trident Scholar Abstracts 2015

A. Eileen Dilks
Midshipman First Class
United States Navy

Towards a Personalized Prescription Tool for Diabetes Treatment

For diabetic patients, insulin is unable to effectively assist in transporting glucose into cells to be used for energy.  Type I diabetes arises when the pancreas does not produce enough insulin, and type II diabetes develops when cell receptors become insensitive to insulin. Both conditions present the danger of causing unhealthy glucose levels in the blood stream and are treated with insulin injection therapy to trigger glucose uptake in the cells.

A mathematical model of natural glucose and insulin control allows for a quantitative understanding of the internal glucose-insulin dynamics of healthy and diabetic patients. Cobelli et. al. presented a simulation model composed of glucose and insulin subsystems and four unit process models which use differential equations to describe the kinetics of glucose digestion and absorption that occur after a meal. This model has achieved the greatest physiological accuracy to date and was used as a basis for a computer simulator for type 1 diabetes that received FDA approval as a substitute to animal trials for preclinical testing. This model provides the means for developing a systematic approach to prescribing insulin injection therapy for diabetic patients in order to maintain healthy glucose levels.

This research extends the Cobelli model of glucose and insulin dynamics to include both long and short-acting insulin inputs currently used to treat diabetic patients. The project will introduce a personalized approach to treatment by adapting the set of average diabetic Cobelli model parameters over time in response to observed patient feedback data. The personalized model will be combined with a nonlinear model predictive control strategy to determine the best insulin injection routine to achieve healthy glucose levels in diabetic patients. This work will contribute to the development of an individualized prescription tool which physicians can use to more effectively treat diabetic patients. 

In this project, we will investigate the impact of the maritime environment on the propagation of laser beams. This study will primarily use data collected at the Naval Academy with the goal of quantifying the correlation between the statistics of the environmental parameters and the statistics of laser beam intensity at the target. The project will have two parts to it: First, we present a computational analysis of different probability density function approximation techniques; and 2) we introduce preliminary steps towards developing a stochastic model for the maritime laser beam propagation. In the first part of this work we will apply three mathematical methods to construct the probability density function of the data: i) the Kernel Density Estimator (KDE) method, ii) the Barakat Method using lower-order moments, and iii) the Bayesian Mixture Model. We will compare and contrast the features of the three approximation techniques, first in the context of a synthetic data whose true pdf is known, and next in the context of the laser data. In the second task, we will analyze how a complex medium causes the photons of the laser light to behave differently than if they were acting in freespace, by focusing on the stochastic behavior that our data exhibits. We will develop a stochastic paraxial wave equation in order to have a mathematical model capable of accepting statistical parameters from the atmosphere as input to allow us to investigate the statistical properties of light intensity at a specified target.  

Contrary to popular belief, not all nuclei are spherical. Nuclei tend to be spherical when they have close to a “full shell” of nucleons. This concept is similar to the stability of noble gases caused by full shells of electrons. Past studies have determined that shells fill at certain “magic numbers” of nucleons when either the proton number ( Z) or the neutron number ( N) of the nucleus is, among other values, 82. The farther away a nucleus’ Z or N is from these magic numbers, the more deformed it will be.

Lutetium has 71 protons ( Z=71). A wide variety of isotopes of lutetium exist, which have the same Z but different N. It is possible to watch how the nucleus deforms by observing isotopes with more and more neutrons. ¹⁵⁷Lu is still nearly spherical even with 86 neutrons, 4 beyond the magic number of 82. However, at N = 90 ( ¹⁶¹Lu), the nucleus becomes well-deformed and exhibits very few properties of a spherical nucleus. As such, the transition between generally spherical lutetium nuclei and generally non-spherical nuclei occurs between N = 86 and N = 90, where ¹⁶⁰Lu ( N = 89) lies.

This project set out to investigate the structure of such a transitional nucleus and identify the predominant factors influencing its shape. The study required creating rapidly-spinning ¹⁶⁰Lu nuclei by means of a particle accelerator at Argonne National Laboratory. Once excited, an instrument called Gammasphere was employed to capture information about the high-energy photons (gamma rays) that these excited states emit as they cool. Our results indicate the possibility that, at lower spins, ¹⁶⁰Lu is asymmetric in shape, being deformed along all three dimensions. At higher spins, our data suggests that the nucleus returns to the one-dimensional deformation typically associated with such rare-earth nuclei. This phenomenon is not observed in neighboring nuclei (those with similar Z or N). We propose that the behavior observed in ¹⁶⁰Lu is a result of increased malleability of transitional nuclei when compared to near-magic number nuclei and an extra unpairing of the nucleons that is not typically observed in nearby nuclei.

FACULTY ADVISOR
Professor Daryl J. Hartley
Physics Department

Brian R. He
Midshipman First Class
United States Navy

Insecurity of food and water supplies in the Arabian Gulf is an important concern for stability in the region, where national security policy and food security policy interrelate. Even with three wars in Libya, Yemen, and Syria and several government overthrows in 2011—a year marked by doubled world grain prices — Arabian Gulf nations, other than Qatar, appear hesitant to publicly declare the severity of impending food and water insecurity. In Oman, population growth at 4.98% between 2003 and 2013, an expatriate community comprising 44% of the total population, salinization issues and sinking groundwater tables, rising obesity, a culture of overindulgence, an overreliance on imported food, and instability in the international marketplace threaten the adequacy of the food and water supply.

This project endeavors to quantify the sensitivity of Oman’s food security strategy to various shocks with a Bayesian belief network (BBN).  A BBN is a model that estimates changes in conditional probability, given assumptions about the causal relationships between variables. In this present study, the probability that the daily energy supply (DES) exceeds a healthy lower bound, estimated at 2100 kilocalories/person/day, serves as the primary output of the BBN. The inputs to the BBN are eighteen variables organized into four categories: energy, trade, domestic agriculture, and human factors. Statistical analyses connect each of these input variables to historical effects on the output variable, DES.

The BBN is then used to test the sensitivity of DES in possible future scenarios. Example scenarios include (1) an international refusal to sell cereals to Oman, (2) a plummet in the price of oil, and (3) the mass emigration of the expatriate workforce from Oman. By focusing on DES, the model meets the standard international definition of a food secure nation and provides an indication of how possible future events could affect the food security of Oman. Beyond the specific model results, this effort also serves as a template and model for building future studies that could help identify—and avert—crises before they happen. 

FACULTY ADVISORS
Associate Professor Deborah L. Wheeler
Political Science Department

Professor Frederick L. Crabbe
Computer Science Department

Associate Professor Patrick A. Caton
Mechanical Engineering Department

Assistant Professor Gina R. Henderson
Oceanography Department

Assistant Professor Michael R. Kellerman
Political Science Department

Michael K. Johnson
Midshipman First Class
United States Navy

Probe-Independent EEG Assessment of Mental Workload in Pilots

Mental workload can be described as a ratio between task-complexity and a person’s cognitive capacity to meet task demands. This description captures the intuitive idea that mental workload depends both on external factors such as the objective difficulty of required tasks, and internal factors such as a person’s past experiences and skill set. There is a growing body of research focused on developing quantitative methods to assess mental workload in order to improve the mental resiliency of people in high stress environments. Various metrics derived from physiological signals such as heart rate, blood pressure, galvanic skin response, and eye-gaze have been investigated as biomarkers of mental workload. These signals have been used to distinguish mental workload levels with accuracies significantly better than chance, but there are still no widely accepted standards or commercial products for mental workload monitoring. 

With recent improvements in the ease-of-use, reliability, and costs of portable electroencephalography (EEG) systems, there has been increasing interest in using brain signals to measure mental workload. It is hypothesized that EEG offers a more direct assay of mental workload than other physiological biomarkers because of the proximity of EEG sensors to the neural substrates of cognitive stress.

Existing approaches for quantifying mental workload using electroencephalography often rely on probe stimuli to elicit stereotyped neural responses such as the P300 wave.  The goal of this research was to develop an EEG-based algorithm to classify different levels of task complexity that does not rely upon an auditory probe.  By choosing subjects with a similar level of task-experience, we partially control for differences in the capacity to perform the experimental task and therefore use task-complexity as a surrogate for mental workload.  As we were particularly interested in understanding the response of aircraft pilots to the cognitive demands imposed by their flight-missions, we used flight simulator tasks of varying challenge-level as our experimental paradigm.  Furthermore, since pilots are typically in persistent radio or intercom communications via headset during flight, this also represents a scenario that would be particularly well-suited to a probe-independent index of cognitive workload.

FACULTY ADVISOR
Assistant Professor Justin A. Blanco
Electrical and Computer Engineering Department

Brownout is a phenomenon encountered when a rotorcraft hovers over an unprepared surface and becomes engulfed in a cloud of sediment. The generated brownout cloud obscures a pilot’s vision, greatly increasing flight risks. Brownout also reduces the service life of mechanical components (i.e., rotor blades, engines, etc.), significantly increasing maintenance costs and reducing operational readiness. The problem of brownout therefore poses a significant hazard to naval rotorcraft operations.

Brownout is caused by the interaction between the rotor wake and loose sediment on the surface. The trailed tip vortices are the primary means by which sediment is entrained into the airflow. Therefore, faster diffusing tip vortices would be expected to reduce brownout intensity. However, when a rotorcraft operates near the ground, the tip vortex filaments are stretched, thereby reintensifying their vorticity and arresting their diffusion rate. Rotor blade design can significantly affect tip vortices. A slotted tip design, with intakes on the leading edge near the tip, was shown to diffuse tip vortices. However, it also incurred a 2% power penalty, due to increased profile drag on the slots.

It was hypothesized that moving the intake slots to the hub, where dynamic pressure is lower, would reduce the profile power penalty while still effectively diffusing tip vortices. With this in mind, the present study investigated a pumping blade design with an intake slot at the hub and various upward orientations of the exit slot at the blade tip.

Rotor performance measurements showed that at lower thrust conditions, where profile losses dominate, the baseline (i.e. non-pumping) blade required less power. However, at higher thrust conditions, where induced losses dominate, the power required for the baseline and pumping blade designs began to converge.

Flowfield measurements were taken using flow visualization and particle image velocimetry.  It was found that the pumping blade designs initially produced more diffused tip vortices as compared the baseline blade. However, at the higher thrust condition, this initial diffusion was not sufficient to overcome the reintensification process resulting from the ground. At the lower thrust condition, the pumping blade effectively diffused the tip vortices.

FACULTY ADVISORS
Assistant Professor Joseph I. Milluzzo
Aerospace Engineering Department

Associate Professor David S. Miklosovic
Aerospace Engineering Department

Samuel S. Lacinski
Midshipman First Class
United States Navy

Multiple Sensor Discrimination of Closely-Spaced Objects on a Ballistic Trajectory

One of the challenges associated with defending against ballistic missiles is to identify and track the object of interest among multiple closely-spaced objects (CSOs) that travel on a ballistic trajectory. One approach that will improve discrimination performance is to combine data from multiple sensors. Multiple sensor correlation and discrimination involves the integration of several sensor returns that are often collected by dissimilar sensors placed on the ground and on-orbit to improve the likelihood of identifying and tracking an object of interest within the CSOs. This project investigates the development of the algorithms necessary for fusing data obtained from multiple, dissimilar sensors. The algorithms employ a target object map (TOM) created using multiple sensor measurements for correlation. The object of interest is then selected using a probability-based Dempster-Shafer discrimination algorithm combined with the TOM correlation probability. A simulation environment was developed to examine the performance of these algorithms. The environment includes relevant characteristics of the sensors in the discrimination algorithm, a modeling process for the ballistic trajectories of the CSOs, and a decision-making process for handling the multi-sensor data to correlate and discriminate the object of interest. This simulation environment was utilized to assess system performance characteristics using Monte Carlo simulation by changing system parameters such as sensor measurement accuracy, sensor locations, Kalman filtering approaches, state propagation algorithms, TOM correlation approaches, probability distribution of characteristics and the number of CSOs.

FACULTY ADVISORS
Associate Professor Tae W. Lim
Aerospace Engineering Department

CDR Tracie A. Severson, USN
Aerospace Engineering Department

Zane A. Markel
Midshipman First Class
United States Navy

Machine Learning Based Malware Detection

Current antivirus software is effective at detecting well known threats, but cannot keep up with the rate at which new malware is authored or modern antivirus avoidance techniques, such as using polymorphic code. Some studies have investigated augmenting current antivirus techniques with machine learning, which could potentially detect some previously unknown malware. However, previously proposed methods either do not detect malware with satisfactory performance, or they have only been tested on laboratory software databases that cannot suitably be projected into realistic performance. This work explores several aspects of machine learning based malware detection. First, we propose an approach to learn primarily from program metadata, particularly header data in the 32-bit Windows Portable Executable (PE32) file format. We identify learning methods that learn effectively from this metadata, explore which metadata features can be trivially modified and are not appropriate for malware detection, test it on approximately realistic datasets, and find that it performs favorably compared to Windows API imports, another category of file characteristic that shows promise for machine learning based malware detection. Additionally, we find and explore the drastic performance drop which occurs when using a realistically low proportion of malware in test datasets instead of datasets split evenly between malware and benign software. Ensemble learning, which commonly alleviates this problem in other similar machine learning applications, does not appreciably help in this context. Training with datasets that have the same proportion of malware as the test datasets optimizes performance, yet the file characteristics that are informative for malware detection change with the proportion of malware in the training dataset. We conclude that file characteristics must be trained on and tested in approximately realistic settings in order to demonstrate their robustness in operational malware detection, and we propose a test procedure which meets these standards. 

FACULTY ADVISOR
Permanent Military Professor CDR Michael B. Bilzor
Computer Science Department

Fletcher D. Rydalch
Midshipman First Class
United States Navy

A Characterization of the Ship-Effect in a Maritime Environment and Special Nuclear Material Detection

The interdiction of nuclear and radiological materials out of regulatory control is of utmost importance in national security. In the maritime environment where such material may be moved on large vessels, detection is complicated due to the environment, the ship’s motion, time constraints, and the amount of potential shielding; either incidental or purposefully placed. Additionally, the level of the radiation background on and in the immediate vicinity of the ship (where an illicit source might be detected) is affected by the “ship effect.” The neutron and gamma radiation ship effect is a phenomenon involving high energy physics where cosmic radiation interacts preferentially with high atomic number materials to produce additional background radiation. A classic example (for which the effect is named) is that of a ship afloat. The objectives of this research project were to spatially characterize the ship effect in the vicinity of a naval warship and to gage the impact of the characterized environment on detection feasibility for onboard radioactive sources. The results of this characterization will inform the development of survey protocols and equipment; allowing improved identification of nuclear material aboard a maritime vessel by stand-off radiation detection. The project included completion of the following tasks: (1) Integrating a suitable radiation detection system, (2) conducting measurements of the background neutron and gamma radiation both on land and surrounding a ship on the water, (3) simulation of a radiation signature emitted from nuclear material aboard a ship using radiation transport software, and (4) comparing the measured radiation signatures and modeled source signatures to show the impact on detection feasibility of nuclear material in a ship effect environment. The objectives of this research project have been met.  Results show an expected increase in radiation background while approaching a ship; with the greatest increase measured to be greater than 40% near the ship’s center of mass. The effect of this increased background radiation on detection feasibility has been estimated. For one configuration, a modeled detector’s response to a notional on-board source has been combined with the characterized background to show the decrease in detectable range due to the ship effect.

FACULTY ADVISORS
Assistant Professor Marshall G. Millett
Mechanical Engineering Department

Professor Martin E. Nelson
Mechanical Engineering Department

VADM Charles I. Leidig, USN (Retired)
Mechanical Engineering Department

David A. Stevens
Midshipman First Class
United States Navy

Computational Sensitivity Analysis for the Aerodynamic Design of Supersonic and Hypersonic Air Vehicles

The conceptual design of hypersonic vehicles relies on computational methods to produce estimates of aerodynamic, structural, thermal protection, and propulsion design requirements. Additionally, conceptual vehicle designs must begin to solve the multidisciplinary optimization problem presented by these competing factors.  Solving the design optimization problem is not possible using traditional low-fidelity vehicle models and flow simulations because these models represent the underlying physics of the combined disciplines too poorly to be dependable. Moreover, it may not even be possible to solve the design optimization problem because these methods may preclude unconventional vehicle designs that do not fall within historical design paradigms. Additionally, designs based on these undependable methods may suffer from unanticipated phenomenon during testing that lead to costly vehicle redesigns and program delays.

The sensitivities of a hypersonic air vehicle’s aerodynamics to geometric variations were calculated using a computational framework developed for this project.  This framework automates the process of design space sampling, vehicle model generation, flow solution, response surface generation, and global sensitivity analysis. It is unique in its integration of modern design tools such as parametric vehicle models, automated volumetric mesh generation and refinement, Eulerian flow simulations, Kriging response surface generation, and global sensitivity analysis.

As a stepping stone to solving the multidisciplinary design optimization problem, this work focuses on quantifying the sensitivity of the vehicle’s lift-to-drag ratio and aerodynamic moments to changes in the planform of the vehicle’s lifting surfaces. The analysis of these aerodynamic forces and moments provides insight into the vehicle’s range and provides a baseline for future structural and control design analyses.

The framework was verified by computing the sensitivity of the vehicle’s lift-to-drag ratio to changes in the wing’s dihedral angle and span. This study demonstrated that the hypersonic vehicle’s lift-to-drag ratio was approximately three times more sensitive to variations in the wing’s span compared to variations in the wing’s dihedral. Future work will expand the framework to include a higher dimensional design space and additional aerodynamic forces and moments. 

FACULTY ADVISOR
Associate Professor Chris L. Pettit
Aerospace Engineering Department

Eric A. Swanson
Midshipman First Class
United States Navy

Black Hole Entropy: Quantum Mechanics in Relativistic Spacetime

The thermodynamic properties of any system can be calculated from knowing the quantum states of that system.  The connection between thermal physics and quantum mechanics is well understood.  Near a black hole, the effects of gravity, understood through general relativity, affect the particle states. In general, quantum mechanics and general relativity are not compatible in describing a system but the extreme environment of a black hole demands the use of both theories.

The starting point for this project is calculating the thermodynamics of system in a Minkowski spacetime. This is the general “flat” spacetime geometry that we experience in our day to day lives.  This  should reproduce the same thermodynamic properties that we are familiar with in classical physics.  The benefit is that we have a framework to introduce the effects of the black hole. The Minkowski  spacetime is a relativistically correct spacetime in the absence of gravity. The form of the equation allows for the effect of the black hole to be input directly and calculated as a perturbation of the flat spacetime solution.

The black hole spacetime introduces mathematical difficulties that are absent in the more simple case. In taking a limiting case of a particle constrained to move in a radial motion (not orbiting), the Hawking Temperature of a black hole can be reproduced. Allowing for general motion of the particle introduces angular momentum  to the problem. The challenges of this extra term can be dealt with using the group theory of representation.

After calculating the wave-function of a particle on the black hole spacetime, thermodynamic quantities such as temperature and entropy can be calculated trivially.

FACULTY ADVISORS
Assistant Professor Eyo E. Ita
Physics Department

CDR Richard H. Downey, USN
Physics Department

Associate Professor Carl E. Mungan
Physics Department

Gabriel Tang Ying Kit
Midshipman First Class
United States Navy

Cooperative Control of Unmanned Surface Vessels and Unmanned Underwater Vessels for Asset Protection

The field of cooperative autonomous control has traditionally focused on a swarm of homogeneous vehicles working together to fulfill a task. However, heterogeneous swarms working cooperatively in a multi-modal manner has the potential to synergize the disparate functional capabilities in order to better fulfill mission requirements.

This project focuses on the development of a cooperative control system for a heterogeneous swarm of unmanned surface vessels (USVs) and unmanned underwater vessels (UUVs) specifically utilized for the task of asset protection. Relying on a hybrid control scheme that combines both behavior-based and systems-theoretic concepts, the swarm provides better adaptability, robustness and overall performance then traditional control methods. Instead of simply defining the unit positions or the shape of the unit distribution desired for the swarm state, a novel capability function is used as the driver for the swarm.  This capability function uses real time data in order to define the actual mission parameters – for example, the probability of detection of a patrol vessel. Based on the capability desired, the swarm then maneuvers itself to generate the required capability.

A well-recognized difficulty with UUVs is the persistent localization problem. Typical methods for localization underwater suffer from buildup of uncertainty over time, reducing the efficacy of the UUV units due to positioning errors.  Surfacing in order to get a GPS fix causes a temporary reduction in the quality of sensing by moving the underwater units to the same plane as the surface vessels, and may also reduce stealth for the UUVs and jeopardize mission fulfillment.  As such, long baseline techniques were implemented in a secondary controller in order to incorporate cooperation between the USV and UUV units as a tool for improving localization. Using the USVs as navigation beacons, the UUVs were able to ascertain their position, mitigating the localization uncertainty while still ensuring that the full heterogeneous swarm provides the desired asset protection capability.

Overall, this model of a hybrid cooperative control has proven itself to be effective, robust and easily manipulated to suit different secondary objectives – setting the foundation for future models of control systems for multi-modal swarms.

FACULTY ADVISOR
Professor Bradley E. Bishop
Weapons and Systems Engineering Department