Trident Scholar Abstracts 2014
Colin E. Bodgan
Midshipman First Class
United States Navy
Applications of Graph-Theoretic Tests to Online Change
Given a sequence of observations, has a change occurred in the underlying probability distribution with respect to observation order? How well can such a change be detected if the sequence is being monitored in real-time? The problem of detecting change, and detecting it with minimal delay, is an important one in a wide variety of real-world situations. For example, one might monitor a complicated multivariate system (such as a military helicopter or a human being) with the goal of detecting subtle change in order to provide advance warning of system failure.
Change-point problems may be classified as “online” or “offline.” In offline problems, all data under consideration are on-hand at the time of analysis and the goal is to determine if, and perhaps when, a change occurred in the observation sequence. This leads to problems if a fatal result is encountered in the middle of the process from which data is being collected because it is too late for detection to do any good. In online problems, data are collected in real-time with the goal of identifying a change as soon as possible after it occurs and thus as far as possible in advance of death.
This project explores nonparametric graph-theoretic approaches to solving online change-point problems. The foundation for our methodology is the Ensemble Sum of Pair-Maxima (ESPM) Test, a powerful offline test developed by Ruth and Koyak (2011). Our work investigates the efficacy of the ESPM Test in a variety of offline settings, and ultimately extends that test to online settings through a novel modification to recently developed multiple testing procedures designed to control false discovery rate. The effectiveness of this modified procedure is tested against simulated and real-world data.
FACULTY ADVISOR
CDR David M. Ruth, USN, Permanent Military Professor
Mathematics Department
Brendan C. Egan
Midshipman First Class
United States Navy
Pitching Flexible Propulsors: Experimental Assessment of Performance Characteristics
Inspiration from nature can provide insight for even the most longstanding engineering challenges. When comparing marine propulsion systems, traditional propellers and rudders lack the acceleration and maneuverability provided by the organic movement of fish. The fish’s unsteady dynamics make it a relatively complex system, but a reduced order system may reveal the most important dynamical mechanisms. Using a flexible hydrodynamic foil that pitches to produce thrust, the most pertinent aspects of a fish-like propulsion system may be replicated in a laboratory environment. The pitching and flexing combination creates a hydroelastic coupling in which the fluid and flexible foil simultaneously affect each other’s behavior.
An understanding of the flexible propulsor’s performance characteristics is required for its application in marine systems. Although the system can be modeled in computer simulations, these may disregard several significant factors. Thus, the experimental campaign seeks to fully grasp the influence of the fluid structure coupling and viscous effects on performance.
The experiments take place in USNA’s large water channel, using full-span flexible propulsor models. The propulsor pitches about a fixed axis at its quarter chord, with a six-axis load cell measuring the forces and torques on the shaft. The non-dimensional parameters, the Reynolds Number, Strouhal Number, and Stiffness Constant, act as a reference for the efficiency of the propulsor and the coefficients of thrust and lift. The project investigates correlations between the propulsors’ instantaneous shape, performance, and interaction with the surrounding fluid. The analysis is conducted through reduction of the measured force and torque data, multiple wake flow analysis techniques, and high speed imaging that captures the instantaneous shape of the foil throughout its oscillations. The ensemble data will in turn facilitate the engineering of fish-like propulsion systems and make it easier and cheaper for future application of this technology.
FACULTY ADVISORS
Associate Professor Mark M. Murray
Mechanical Engineering Department
Associate Professor Cody J. Brownell
Mechanical Engineering Department
Eric N. Fugleberg
Midshipman First Class
United States Navy
System Identification and Control of a Joint-Actuated Buoy
Consider the example of a small, free-floating buoy using a directional antenna to communicate with a satellite. In order to communicate effectively, the buoy’s position must remain stable and directed towards the satellite at all times. With a joint-actuated buoy, it is mechanically possible to stabilize the free-floating buoy and its payload, the antenna. The goal of this Trident project was to perform system identification of the joint-actuated buoy; prepare a simulation of the buoy’s response to different system disturbances; and create single loop control laws for the elevation and azimuth of the buoy to keep it vertical at all times.
To compliment theoretical models derived from Newtonian physics, an experimentally derived model allows non-linear effects such as drag, added mass effects, and hydrodynamics to be inherently accounted for within the experimental data. Testing of the buoy’s response to the movement of the buoy housing and response to wave disturbances were the two primary methods used to formulate the experimental model. From the experimental model, a simulation was created using a new set of experimental data to validate the model. Next, single loop control laws were developed via the simulated model and implemented to control the buoy’s elevation and azimuth to keep the buoy vertical at all times; keeping the buoy stable while at a specified angle off of vertical was not attempted in this project due to the complexities of three dimension control with only two control surfaces.
The system identification model, simulation, and single loop control laws for buoy could be utilized by the Navy to monitor meteorological measurements of the air column above the ocean surface, directional communications with a satellite, and support many other diverse operations.
FACULTY ADVISORS
CAPT Owen G. Thorp III, USN, Permanent Military Professor
Weapons and Systems Engineering Department
Professor George E. Piper
Weapons and Systems Engineering Department
Grant N. Genzman
Midshipman First Class
United States Navy
Proposed offshore structure designs for algal production using wastewater may incorporate floating flexible tubes. This study includes an extensive set of physical modeling experiments to investigate the loads on and dynamic response of these tubes in waves and currents. The physical modeling approach involved designing and building an approximately 1:4 scaled representation of a potential design. Experiments were conducted with the model in the 37-meter tow/wave tank in Hydromechanics Laboratory at the United States Naval Academy. Several combinations of model tests were performed in scaled waves and currents at tube fill levels of 50 and 95%. Several of the tow tests done in the tank, representing oceanic currents, were validated with computational fluid dynamics.
Along with the time series drag results, the physical modeling experiments that included both regular and random waves were analyzed to produce linear transfer functions for both heave and force. An assessment of these data sets indicated that the flexible floating tube response generally followed the wave forcing. The response lessened, however, in low frequency waves and at a lower fill level. Additionally, the forward end of the model typically exhibited a larger response. The results did not indicate a significant difference in dynamic response when the model was tested in both waves and current. The average attachment loads, however, were typically higher when the model was tested in waves and a current than in waves only. The attachment loads were also higher in testing conditions with faster current or lower-frequency waves.
Associate Professor David W. Fredriksson
Professor Michael P. Schultz
Naval Architecture and Ocean Engineering Department
Midshipman First Class
United States Navy
The field of microrobotics has vast applications including non-invasive surgery, targeted drug delivery, and telemetry. Some challenges with current systems include achieving propulsion and control. Many groups are developing or have developed magnetically based actuation methods. These magnetically based systems can potentially lead to undesirable effects on the human body. Acoustic control provides an interesting alternative to existing magnetic or electrostatic actuation in that acoustic signals include few harmful effects on the human subject. Furthermore, the use of acoustic signals allow for the possibility to leverage existing medical imaging technology.
This project focused on an alternative method of actuation which utilizes double-jointed, flagella-like, flappers designed for whip-like, “non-reciprocal” motion. Unlike reciprocal motion, non-reciprocal motion has a time irreversible nature, causing displacement regardless of the surrounding environment. Such motion is essential for microscale propulsion, where surface forces dominate inertial forces. Milli-scale flappers were modeled using a multiphysics finite element modeling program, fabricated, and subsequently tested to confirm agreement between the modeling software and expected behavior, as well as the presence of non-reciprocal modes at resonance. Flappers were then attached to a robot body, and the structure was tested for displacement while submerged in fluid. The body was designed to respond to an acoustic source, driving the structure to resonance and resulting in forward propulsion at the expected resonance frequency. The milli-scale robot was also tested in a high viscosity fluid to reproduce the low Reynolds number environment of microrobotics in water. Steering could be achieved in these structures by attaching multiple “flagella” tuned to different resonance frequencies. This project lays the foundation for the development of an acoustically actuated microscale robot.
FACULTY ADVISORS
Professor Samara L. Firebaugh
Electrical and Computer Engineering Department
Associate Professor Jenelle A. Piepmeier
Weapons and Systems Engineering Department
Associate Professor John A. Burkhardt
Mechanical Engineering Department
Matthew J. Lanoue
Midshipman First Class
United States Navy
Next Generation Satellite Communications: Automated Doppler Shift Compensation of PSK-31 Via Software-Defined Radio
Satellite communication systems fall into two broad categories: Amplify and Forward (AF), and Regenerative. AF systems operate as a “bent-pipe” where the information received at the satellite is simply amplified and retransmitted with no alteration of the original signal. In a regenerative repeater, circuitry is designed to demodulate the signal, recover the original information, re-modulate the signal and transmit a new version. Limitations exist in both of these categories: AF systems are unable to compensate for distortion and hardware-defined regenerative repeaters cannot be updated over time.
Software-defined radio (SDR) leverages the processing power of computers to create communications systems that run as flexible software applications, where the radio operating parameters can be set or altered by software. SDR satellite communications systems developed today can run on future hardware platforms and update the applications already running on the satellite. GNU Radio, a framework for creating SDR applications, has recently become capable of developing satellite communications systems.
This project implemented PSK-31, a terrestrial narrowband form of multi-user amateur radio communications for text and simple data messaging, in GNU Radio as part of a as a regenerative satellite repeater. One of the major issues with satellite communications is the Doppler shift experienced as the satellite passes overhead. Demodulating signals affected by Doppler shift requires ground stations with circuits dedicated to track and synchronize with the satellite in order to compensate for the Doppler shift. For the PSK-31 waveform, the amount of Doppler shift exhibited by the satellite would prevent communications using standard receivers.
FACULTY ADVISORS
Associate Professor Christopher R. Anderson
Electrical and Computer Engineering Department
LCDR Jennie H.G. Wood, USN, Junior Permanent Military Professor
Electrical and Computer Engineering Department
Midshipman First Class
United States Navy
Recent studies have shown that graph theory is a useful tool in studying changes in brain connectivity resulting from degenerative conditions such as Alzheimer’s disease (AD). The human brain can be naturally modeled as a network and graph theory measures enable the connectivity properties of these models to be quantified. These measures allow differences in connectivity between brains with and without signs of dementia to be identified.
This study is an investigation of methods used to create network models from magnetic resonance imaging (MRI) data and the impact of these methods have on connectivity measures. We tested previous network creation methods and newly developed methods, in combination with connectivity measures to determine which combinations yielded the most reliable identification of dementia severity. We categorized dementia severity using four diagnostic groups: healthy older adults who maintained normal cognition for 36 months, individuals with Mild Cognitive impairment (MCI) who remained MCI for 36 months, individuals who started the study with MCI but developed AD within 36 months (MCI-AD), and individuals with AD. We modeled connectivity between brain regions using correlations between regional cortical thickness measurements obtained using MRI.
The network creation method that produced the most reliable differences between diagnostic groups across connectivity measures was using partial correlations to determine which edges to include in the network, and then weighting these edges by the normalized product of the mean cortical thickness of the two adjacent regions. Our results also suggest that different graph measures change in an ordered fashion for the structural brain network as an individual develops AD and may be useful as early-diagnosis tools.
FACULTY ADVISOR
Assistant Professor David J. Phillips
Mathematics Department
EXTERNAL COLLABORATOR
Assistant Professor Anja Soldan, Department of Neurology
The Johns Hopkins University
Brynn E. Umbach
Midshipman First Class
United States Navy
Characterization of Microalgal Lipids for Optimization of Biofuels
This research project investigates the lipid content and composition of extremophilic and estuarine microalgae under different growth conditions for suitability as biofuel feedstocks. Fatty acid content and lipid composition are major considerations in optimizing algal biofuel feedstocks for fuel production, as they determine important characteristics including melting point, flashpoint, cetane number, and fuel value. In this project I used fatty acid-methyl ester analysis (FAME GC-MS) to characterize the acyl content of the well-known alga Chlorella, the unusual acidothermophilic alga Galdieria sulphuraria, and cold-tolerant strains extracted from the Severn River through winter bioprospecting. The cold tolerant species’ acyl content included a greater percentage of C16 and highly unsaturated hydrocarbon chains in comparison to Chlorella.
In addition to these investigations of cold-tolerant algal lipids, I used a new, high-recovery method of direct FAME extraction from algal biomass to examine lipid content and composition from Galdieria and Chlorella cultures grown under a variety of conditions, including mixotrophic and autotrophic growth in media with varying amounts of nitrogen. Chlorella demonstrated the expected effect of nutrient variation on algal lipid production: both sugar and nitrogen significantly decreased the percent by mass fatty acid yield. In contrast, preliminary results suggest that the hyperthermophilic alga Galdieria organism produced a greater amount of fatty acids when grown with sugar and nitrogen. This unique characteristic of the Galdieria strain suggests it has potential to be beneficial as a biofuels feedstock.
I have also performed pilot experiments in the simultaneous total molecular characterization of the structure of mixed commercial lipids standards using MALDI-TOF mass spectrometry (lipidomics). Applying this approach to algal lipid extracts will allow a more rapid and concrete structural characterization of these potential lipid feedstocks and provide further understanding of the biological context of these lipids and their potential optimization.
FACULTY ADVISOR
Assistant Professor Charles R. Sweet
Chemistry Department
