Annotating large unlabeled datasets can be a major bottleneck for machine learning applications. We introduce a scheme for inferring labels of unlabeled data at a fraction of the cost of labeling the entire dataset. We refer to the scheme as Bounded Expectation of Label Assignment (BELA). BELA greedily queries an oracle (or human labeler) and partitions a dataset to find data subsets that have mostly the same label. BELA can then infer labels by majority vote of the known labels in each subset. BELA makes the decision to split or label from a subset by maximizing a lower bound on the expected number of correctly labeled examples. BELA improves upon existing hierarchical labeling schemes by using supervised models to partition the data, therefore avoiding reliance on unsupervised clustering methods that may not accurately group data by label. We design BELA with strategies to avoid bias that could be introduced through this adaptive partitioning. We evaluate BELA on labeling of four datasets and find that it outperforms existing strategies for adaptive labeling. / Master of Science / Most machine learning classifiers require data with both features and labels. The features of the data may be the pixel values for an image, the words in a text sample, the audio of a voice clip, and more. The labels of a dataset define the data. They place the data into one of several categories, such as determining whether a image is of a cat or dog, or adding subtitles to Youtube videos. The labeling of a dataset can be expensive, and usually requires a human to annotate. Human labeled data can be moreso expensive if the data requires an expert labeler, as in the labeling of medical images, or when labeling data is particularly time consuming. We introduce a scheme for labeling data that aims to lessen the cost of human labeled data by labeling a subset of an entire dataset and making an educated guess on the labels of the remaining unlabeled data. The labeled data generated from our approach may be then used towards the training of a classifier, or an algorithm that maps the features of data to a guessed label. This is based off of the intuition that data with similar features will also have similar labels. Our approach uses a game-like process of, at any point, choosing between one of two possible actions: we may either label a new data point, thus learning more about the dataset, or we may split apart the dataset into multiple subsets of data. We will eventually guess the labels of the unlabeled data by assigning each unlabeled data point the majority label of the data subset that it belongs to. The novelty in our approach is that we use supervised classifiers, or splitting techniques that use both the features and the labels of data, to split a dataset into new subsets. We use bias reduction techniques that enable us to use supervised splitting.

This work investigates the use of multi-degree of freedom (MDOF) passive and active vibration absorbers for the control of structural vibration as an improvement to conventional single degree of freedom (SDOF) vibration absorbers. An analytical model is first used to compare passive two degree of freedom (2DOF) absorbers to SDOF absorbers using point impedance as the performance criterion. The results show that one 2DOF absorber can provide the same impedance at two resonance frequencies as two SDOF absorbers for equal amounts of total mass. Experimental testing on a composite cylindrical shell supports the assertion that a 2DOF absorber can attenuate two resonance frequencies. Further modeling shows that MDOF absorbers can utilize the multiple mode shapes that correspond to their multiple resonance frequencies to couple into modes of a distributed primary system to improve the attenuation of structural resonance. By choosing the coupling positions of the MDOF absorber such that its mode shape mirrors that of the primary system, the mass of the absorber can be utilized at multiple resonance frequencies. For limited ranges of targeted resonance frequencies, this technique can result in MDOF absorbers providing attenuation equivalent to SDOF absorbers while using less mass. The advantage gained with the MDOF absorbers is dependent on the primary system. This work compares the advantage gained using the MDOF absorbers for three primary systems: MDOF lumped parameter systems, a pinned-pinned plate, and a cylindrical shell. The active vibration absorber study in this work is highly motivated by the desire to reduce structural vibration in a rocket payload fairing. Since the efficiency of acoustic foam is very poor at low frequencies, the target bandwidth was 50 to 200 Hz. A 2DOF active vibration absorber was desired to exhibit broad resonance characteristics over this frequency band. An analytical model was developed to facilitate the design of the mechanical and electrical properties of the 2DOF active vibration absorber, and is supported by experimental data. Eight active vibration absorbers were then constructed and used in a multiple-input multiple-output (MIMO) feed-forward control system on a mock payload fairing under high level acoustic excitation. The results show significant levels of global attenuation within the targeted frequency band. / Master of Science

Electronic textbooks are different from ebooks (electronic books) in that they allow the user to go beyond just reading material on a computer screen. Electronic textbooks encourage the user to accomplish all of the operations typically performed with a hardcopy text in addition to some functions not possible with paper books. With electronic textbooks users can make annotations in the textbook with e-ink; mark important sections; search over the ink, the text, or even the scanned images; look up items in online dictionaries or encyclopedias; perform interactive reinforcement drills; view simulations; and many other operations afforded by the computing power of the underlying computer and the reach of the Internet connection. These operations encourage students to engage in active reading. The VText framework is designed to provide many of the desired features of an e-textbook in such a way that it produces pedagogical value rather than just convenience for students. Many so-called e-textbook solutions available today provide few features beyond those possible with hardcopy textbooks. The VText framework is built as an add-in to Microsoft's note-taking program, OneNote. The add-in provides features that facilitate the use of OneNote as a reader and as an educational tool while leaving in place OneNote's strengths in note-taking, collaboration, and search. / Master of Science

In recent years there has been a growing interest in moving away from large centralized power generation toward distributed energy resources. Solar energy generation presents several benefits for use as a distributed energy resource, especially as a peaking power source. One drawback of solar energy sources is the need for energy storage for the system to be utilized for a significant percentage of the day. One way of avoiding adding energy storage to a solar generation system while still maintaining high system utilization is to design the power conversion subsystem to also provide harmonic and reactive compensation. When the sun is unavailable for generation, the system hardware can still be utilized to correct for harmonic and reactive currents on the distribution system. This system's dual-purpose operation solves both the power generation need, and helps to improve the growing problem of harmonic and reactive pollution of the distribution system. A control method is proposed for a system that provides approximately 1 kW of solar generation as well as up to 10 kVA of harmonic and reactive compensation simultaneously. The current control for the active was implemented with the synchronous reference frame method. The system and controller was designed and simulated. The harmonic and reactive compensation part of the system was built and tested experimentally. Due to the delay inherent in the control system from the sensors, calculation time, and power stage dynamics, the system was unable to correct for higher order harmonics. To allow the system to correct for all of the harmonics of concern, a hybrid passive - active approach was investigated by placing a set of inductors in series with the AC side of the load. A procedure was developed for properly sizing the inductors based on the harmonic residuals with the compensator in operation. / Master of Science

Power harvesting, energy harvesting, power scavenging, and energy scavenging are four terms commonly used to describe the process of extracting useful electrical energy from other ambient energy sources using special materials called transducers that have the ability to convert one form of energy into another. While the words power and energy have vastly different definitions, the terms "power harvesting" and "energy harvesting" are used interchangeably throughout much of the literature to describe the same process of extracting electrical energy from ambient sources. Even though most of the energy coupling materials currently available have been around for decades, their use for the specific purpose of power harvesting has not been thoroughly examined until recently, when the power requirements of many electronic devices has reduced drastically. The overall objective of this research is to typify the power source characteristics of various transducer devices in order to find some basic way to compare the relative energy densities of each type of device and, where possible, the comparative energy densities within subcategories of harvesting techniques. Included in this research is also a comparison of power storage techniques, which is often neglected in other literature sources. An initial analysis of power storage devices explores the background of secondary (rechargeable) batteries and supercapacitors, the advantages and disadvantages of each, as well as the promising characteristics of recent supercapacitor technology developments. Also explored is research into the effectiveness of piezoelectric energy harvesting for the purpose of battery charging, with particular focus on the current output of piezoelectric harvesters. The first objective involved presenting and verifying a model for a cantilever piezoelectric bimorph. Next, an investigation into new active fiber composite materials and macro fiber composite devices utilizing the d31 coefficient is performed in comparison to a monolithic piezoelectric bimorph. The information gathered here was used to design a two bimorph device termed the mobile energy harvester (MEH). Worn by a human being at the waste level, the MEH harvests energy from each footfall during walking or running. The next objective involved characterizing small temperature gradient (less than 200 oC) thermoelectric generators (TEGs). Four TEGs were linked in series and joined with a specially made aluminum base and fin heat sink. This device was then mounted to the exhaust system of an automobile and proved capable of recharging both an 80 and a 300 milliamp-hour battery. A switching circuit concept to step up the output voltage is also presented. However, the circuit proves somewhat difficult to implement, so an alternative DC/DC device is proposed as a possible solution. With the advent of highly efficient, low voltage DC to DC converters, it is shown that their high current, low voltage output can be converted to a higher voltage source that is suitable for many electronic and recharging applications. As extensive literature exists on the capabilities of photovoltaic and electromagnetic energy harvesting, no original experimentation is presented. Instead, only a brief overview of the pertinent technological advances is provided in this document for the purpose of comparison to piezoelectric and thermoelectric energy harvesting. The main research focus, as described above, is dedicated to designing and performing original experiments to characterize cutting edge piezoelectric and thermoelectric transducer materials. To conclude and unify the document, the final section compares the power harvesting techniques with one another and introduces methods of combining them to produce a hybrid, multiple energy domain harvesting device. A piezoelectric-electromagnetic harvesting combination device is presented and scrutinized, revealing that such a device could improve the amount of energy extracted from a single harvesting unit. The research presented here not only expands on the present understanding of these materials, but also proposes a new method of creating a hybrid power harvesting device utilizing two of the energy coupling domains, electromechanical and piezoelectric. The goal is to maximize the harvested energy by tapping into as many ambient sources as are available and practical. / Master of Science

Magnetorheolgical (MR) fluid dampers have the capability of changing their effective damping force depending on the current input to the damper. A number of factors in the construction of the damper, as well as the properties of the fluid and the electromagnet, create a dynamic response of the damper that cannot be fully described with a static model dependent on current and velocity. This study will compare different techniques for modeling the force response of the damper in the current-velocity space. To ensure that all the dynamic response characteristics of the damper are captured in data collection, random input signals were used for velocity and current inputs. By providing a normally distributed random signal for velocity to a shock dynamometer and a uniformly distributed random signal for current to a Lord rheonetic seat damper, the force response could be measured. The data from this test is analyzed as a two dimensional signal, a three dimensional force plot in the current velocity plane, and as a probability density function. Four models are created to fit the data. The first is a linear model dependent solely on current. The second is a nonlinear model dependent on both current and velocity. The third model takes the nonlinear model and includes a filter that affects the force response of the model with time. Each of these three approaches are compared based on the total error in the force response and the models? ability to match the PDF of the data. Finally, a fourth model is created for the damper that improves the nonlinear model by making one parameter a probability parameter defined by a PDF calculated from the data. However, because it is a probability model, the error cannot be found through comparison to the data. / Master of Science

The Pedestrian is compromised in the majority of our built landscapes. Today's dominant fixture is the automobile. Pedestrian and automobile efficiency are in direct competition with each other; to facilitate one is to inhibit the other. Pedestrian functionality depends on the presence of walkable destinations, commonly referred to as multi-use areas. Pedestrian functionality is an important issue because sprawl, the current development norm, is reaching the physical limits of the countryside. Density is the positive alternative to issues we encounter as a result of low density such as increased runoff, pollution, congestion, obesity, physical inactivity, and road rage. â The alternative to sprawl is simple and timely: neighborhoods of housing, parks and schools placed within walking distance of shops, civic services, jobs and transit — a modern version of the traditional town.â 1 Improving pedestrian functionality has the ability to impact multiple aspects of our lives and improve the quality of life we experience. â We need communities that are occupied full time and that provide a world of opportunity for kids, communities that support women and men in their efforts to weave together an ever more complex life of home and work.â 2 1 Calthorpe, Peter, p. 16 2 Duany, Andres, p. 25 / Master of Landscape Architecture

Active magnetic bearings (AMBs) have the ability to act concurrently as support bearings and as load sensing measurement tools. Previous work in the area of AMB force measurement has relied upon basic magnetic equations requiring knowledge of coil currents and air gap lengths. Some researchers have utilized magnetic flux probes to eliminate the need for air gap measurements, but these are limited by physical size constraints and require complex hardware. This thesis presents a new method for measuring forces with AMBs that utilizes multiple current pairs with no gap measurement to provide accurate and precise force predictions. Previous methods for force measurement with AMBs rely on the controlled environment of a laboratory setting for accurate measurements. The goal of this work is to develop a robust force measurement procedure for use in industrial and field applications, as well as the laboratory. The harsh environment of a factory floor makes air gap measurements difficult, which limits the use of current-based force equations. Additionally, the flexibility of AMB-equipped thrust measurement systems (TMSs) to measure many types of forces with little to no reconfiguration or calibration makes them appealing. The multi-point method provides predictions of both shaft force and rotor position using only current pairs without air gap measurements. Static and dynamic load scenarios were investigated to determine the feasibility of this new approach to force measurement. For both, the effects of bearing load and rotor position within the bearing were analyzed. Under dynamic loading, different amounts of unbalance as well as various rotor speeds were used to provide multiple test cases. The multi-point predictions of rotor position were analyzed and compared with the measured rotor positions. It was shown that this new multiple-point method for measuring bearing loads with AMBs provides equivalent or better force predictions to analogous single-point methods for static loads while eliminating the need for measuring rotor position. / Master of Science

Intrusion detection systems are an integral part of web application security. As Internet use continues to increase, the demand for fast, accurate intrusion detection systems has grown. Various IDSs like Snort, Zeek, Solarwinds SEM, and Sleuth9, detect malicious intent based on existing patterns of attack. While these systems are widely deployed, there are limitations with their approach, and anomaly-based IDSs that classify baseline behavior and trigger on deviations were developed to address their shortcomings. Existing anomaly-based IDSs have limitations that are typical of any machine learning system, including high false-positive rates, a lack of clear infrastructure for deployment, the requirement for data to be centralized, and an inability to add modules tailored to specific organizational threats. To address these shortcomings, our work proposes a system that is distributed in nature, can actively learn and uses experts to improve accuracy. Our results indicate that the integrated system can operate independently as a holistic system while maintaining an accuracy of 99.03%, a false positive rate of 0.5%, and speed of processing 160,000 packets per second for an average system. / Master of Science / Intrusion detection systems are an integral part of web application security. The task of an intrusion detection system is to identify attacks on web applications. As Internet use continues to increase, the demand for fast, accurate intrusion detection systems has grown. Various IDSs like Snort, Zeek, Solarwinds SEM, and Sleuth9, detect malicious intent based on existing attack patterns. While these systems are widely deployed, there are limitations with their approach, and anomaly-based IDSs that learn a system's baseline behavior and trigger on deviations were developed to address their shortcomings. Existing anomaly-based IDSs have limitations that are typical of any machine learning system, including high false-positive rates, a lack of clear infrastructure for deployment, the requirement for data to be centralized, and an inability to add modules tailored to specific organizational threats. To address these shortcomings, our work proposes a system that is distributed in nature, can actively learn and uses experts to improve accuracy. Our results indicate that the integrated system can operate independently as a holistic system while maintaining an accuracy of 99.03%, a false positive rate of 0.5%, and speed of processing 160,000 packets per second for an average system.

Vehicle suspensions have been a major focus of research and design since the introduction of the automobile. The two major characteristics that define the performance of a specific suspension design are the vehicle ride and handling. A simplified view of chassis development is that a desirable vehicle ride is provided by a "soft" suspension, and desirable vehicle handling is provided by "stiff" suspensions, it is impossible to simultaneously maximize both vehicle ride and handling in a passive suspension design. However, with the introduction of active components, such as actuators and semiactive dampers, coupled with dynamic control strategies, a more desirable compromise between the benefits of soft and stiff suspensions can be achieved. The purpose of this research is not to create a better suspension, but to investigate how active control can improve vehicle ride. As with any research area, a method for evaluating new ideas is required. Many computational methods exist for determining the ride performance of various suspension designs, however computational results alone can be hard to interpret and often deceiving. The following thesis outlines the design and validation of an experimental test rig for evaluating fully active or semi-active suspension algorithms. The test rig utilizes a fully active electromagnetic actuator which can simulate the performance characteristics of fully active as well as semi-active components which a suspension design may utilize. The demonstration rig couples visual inspection with computational analysis and provides the tools necessary for the designer to accurately interpret the ride performance of a new design and simultaneously compare it with a passive design. The goal is to provide a visual development platform in which new algorithms can be quickly and easily implemented and compared against existing algorithms to determine the performance characteristics of each on a physical system. The test rig demonstrated its ability to test both fully-active and semi-active skyhook control algorithms, and accurately simulated both fully-active and semi-active suspension components. The test rig provides a simple and cost effective way to evaluate new algorithms both computationally and visually on a physical system. / Master of Science