PhD, University of Cincinnati, 2024, Arts and Sciences: Psychology

Annually, over 12 million people suffer a stroke worldwide, with approximately 65% of survivors facing lasting upper limb impairments even after prolonged rehabilitation. Most standard rehabilitation interventions share a common goal: Teach stroke survivors to move their paretic upper limbs according to non-disabled norms, based on the assumption that deviations from norms reduce function and, therefore, need correction. This assumption, however, conflicts with the lived experiences of individuals with disability and findings from basic research in motor control. Findings from this research suggest that what underlies impairments in upper limb function in stroke survivors is not deviations from neurotypical norms but reduced expression of motor flexibility, a behavior that may depend on how individuals with stroke visually explore the environment while performing upper limb tasks. Therefore, the purpose of this dissertation was to investigate the interplay between motor flexibility, perceptual exploration, and upper limb performance in individuals post-stroke. It was hypothesized that motor flexibility would predict upper limb performance in individuals post-stroke and would be shaped by the pattern of visual exploration of the task environment. Individuals with chronic stroke (n = 29) participated in the study. Participants sat on a chair without back support to perform a transportation task with their paretic upper limbs in virtual reality. While performing the task, the 3D kinematics of their upper limb, trunk, and head were recorded. The task consisted of using a virtual pad—controlled by hand movements—to move virtual pucks across a bridge and into a goal container. Participants completed the task under two experimental conditions: Pushing and Hitting. In each condition, participants were required to score 20 points. The number of attempts necessary to score the required 20 points was used as a measure of performance, with lower numbers indicating better ta (open full item for complete abstract)

Committee: Paula Silva Ph.D. (Committee Chair); Anthony Chemero Ph.D. (Committee Member); Nikita Kuznetsov Ph.D. (Committee Member); Michael Riley Ph.D. (Committee Member) Subjects: Psychology

MA, University of Cincinnati, 2024, Arts and Sciences: Psychology

We must align our behaviors with others to achieve our shared goals. Understanding the factors that promote stable and accurate performance during joint actions has important theoretical and practical implications. Though quantifying the relevant predictors of joint performance remains an ongoing challenge, one way to address this issue is to view joint action as synergy that links various elements to covary to stabilize task performance. Reciprocal compensation of synergy is the critical mechanism that not only facilitates load sharing but also compensates for variations among elements to preserve performance stability. The Uncontrolled Manifold (UCM) approach is used to quantify synergy by separating the motor variance within the UCM subspace as "good variance" (VUCM) and orthogonal to the UCM subspace as "bad variance" (VORT). While the amount of compensation (VUCM) is a key component of the synergy index, there is some ambiguity about the amount of compensation and performance. Temporal structure of VUCM has been shown to provide additional insight into intrapersonal coordination, however, its role in predicting performance at the interpersonal coordination remains unexplored. Thirty dyads participated, performing a force stabilization task using pinch gauge dynamometers and compression load cells. Participants produced a fixed total force for 30-second trials under low-force and high-force conditions (10% and 30% MVC). Finger-force data were displayed in real-time on a monitor, with force deviations recorded across 10 randomized trials per dyad. The standard UCM approach was used to quantify the degree of synergy between the dyads performing steady-state isometric force stabilization task. The amount of compensation (VUCM), calculated from the UCM method, was extended to sample entropy (SampEnUCM) to analyze the regularity of the compensation. RMS error values (deviations of total force from target performance) were submitted to a mixed effects model with V (open full item for complete abstract)

Committee: Tehran Davis Ph.D. (Committee Chair); Tamara Lorenz Ph.D. (Committee Member); Paula Silva Ph.D. (Committee Member) Subjects: Experimental Psychology

Doctor of Philosophy, Case Western Reserve University, 2024, Applied Mathematics

This doctoral dissertation is concerned with solving the inverse problem of human movement in terms of muscle forces and activation. While feasible to measure some of these forces directly in the human body through invasive procedures, the inverse problem of musculoskeletal modeling can estimate these forces and activation non-invasively, thus presenting a safer and more practical alternative once estimates are thoroughly validated. In this thesis, we set up the muscle recruitment problem as a Bayesian inverse problem, and we estimate muscle forces and activations while concurrently quantifying the associated uncertainties. The abundance of muscles relative to available degrees of freedom grants the human musculoskeletal system redundancy, enabling diverse muscle activation patterns during motor tasks. This redundancy is crucial for the system's functionality across various conditions, including pathological states. A fundamental challenge in biomechanics involves understanding how the complex interaction between the central nervous system and musculoskeletal system, characterized by redundancy, governs normal activation patterns and their evolution in abnormal conditions, such as neurodegenerative diseases and aging. This dissertation presents a mathematical framework to address this challenge through Bayesian probabilistic modeling of the musculoskeletal system. Using Lagrangian dynamics, observed movements are transformed into time series of equlibria that constitute the basis of the likelihood model. Various prior models, aligned with biologically inspired assumptions regarding muscle dynamics and control, are introduced and tested. The corresponding posterior distributions of muscle activations are explored using Markov chain Monte Carlo (MCMC) sampling techniques. The different priors are evaluated by comparing the model predictions with actual observations. This thesis also proposes a model for sparse muscle recruitment that could be of use in (open full item for complete abstract)

Committee: Erkki Somersalo (Committee Chair); Daniela Calvetti (Committee Member); Jenny Brynjarsdottir (Committee Member); Kathryn Daltorio (Committee Member) Subjects: Applied Mathematics; Biomechanics; Biomedical Engineering; Biomedical Research; Mathematics; Sports Medicine; Statistics

MA, University of Cincinnati, 2022, Arts and Sciences: Psychology

Resilience is defined as an individual's ability to preserve or stabilize performance when exposed to challenges or stressors. Living systems seem to show a particular kind of resilience that allows successful completion of a task across a variety of circumstances without requiring maintenance of one specific solution but rather allow motor elements to compensate for one another. The Uncontrolled Manifold (UCM) is a method for investigating synergies by quantifying the amount of compensatory variance (VUCM) used in a motor task. Previous work has applied sample entropy to VUCM (SampEnUCM) to examine the regularity with which an individual explores compensatory patterns over time. The current study investigates how resilience is altered by an individual's history with task demands and related changes in synergy characteristics. Thirty-two college students participated in an isometric finger force task in which visual feedback was altered via eight levels of error amplification (gain). Half of the participants experienced the gains from low-to-high (L-H) precision demand and the other half from high-to-low (H-L). Ultimately the H-L group showed more resilience at higher gains than the L-H group. The VUCM differed between the groups and in its relationship to resilient performance in the face of increasing gain. Both groups showed more irregularity in compensatory patterns over trials. These results suggest that an individual's history with a task does indeed play an important role in shaping synergies and affecting resilience in a steady state task. Future work is suggested to elucidate unclear mechanisms underlying the current results.

Committee: Paula Silva Ph.D. (Committee Member); Michael Riley Ph.D. (Committee Member); Tamara Lorenz Ph.D. (Committee Member) Subjects: Experimental Psychology

Bachelor of Science, Wittenberg University, 2020, Math

The objective of this thesis is to understand the systematic analytic treatment of the model presented in Anna Ghazaryan and Vahagn Manukian's journal article, “Coherent Structures in a Population Model for Mussel-Algae Interaction," which concentrates on the formation of mussel beds on soft sediments, like those found on cobble beaches. The study will investigate how the tidal flow of the water is the main structure that creates the mussel-algae interaction observed on soft sediments. With this investigation, the idea of fast-time and slow-time systems is explicated according to Geometric Singular Perturbation Theory, how Invariant Manifold Theory proves the existence of our solutions, the process of non-dimensionalization, and the re-scaling of the model. It will apply concepts found in nonlinear dynamics to discover equilibria and nullclines of the system. Finally, the study will discuss what the findings mean in context of the population model and the implications of tidal flow on other ecological relationships.

Committee: Adam Parker (Advisor); Alyssa Hoofnagle (Committee Member); Jeremiah Williams (Committee Member) Subjects: Applied Mathematics; Aquatic Sciences; Ecology; Mathematics

Doctor of Philosophy, The Ohio State University, 2020, Electrical and Computer Engineering

Antennas mounted on aircraft, UAVs, and other platforms are used in a number of critical applications, such as navigation, communication, and situational awareness. Since the platform can heavily affect the antenna pattern, one should carry out in situ characterization of the antenna to evaluate the performance of the RF systems. It is often too expensive or impractical to measure the antenna on the intended platform, so instead, the antenna under test (AUT) is measured on a simple ground plane. The measurements are then imported into computational electromagnetics (CEM) codes to simulate platform scattering from the platform of interest. However, current approaches struggle to isolate the antenna radiation from the measurement ground plane interactions, leading to inaccuracies in the AUT representation. Furthermore, many approaches rely on near-field measurements for accuracy and use many current elements to represent the AUT leading to long simulation run-times. This dissertation presents a novel approach for in situ manifold estimation which represents measured data via a weighted sum of simple basis element far-fields. The approach, the Sparse-Constrained Equivalent Element Distribution Method (SC-EEDM), provides a more accurate representation of the AUT compared to existing techniques. The SC-EEDM accurately represents the AUT using measured far-field data only, and represents the AUT using a small number of current elements. In addition, the SC-EEDM isolates antenna radiation from antenna-ground plane interactions, leading to more accurate in situ manifold estimations. Using high-fidelity simulations, the method is shown to accurately estimate antenna far-fields on complex platforms from antenna measurements on simple structures.

Committee: Inder Gupta (Advisor); Robert Burkholder (Committee Member); Teixeira Fernando (Committee Member) Subjects: Electrical Engineering; Electromagnetics; Electromagnetism

Doctor of Philosophy, The Ohio State University, 2020, Chemical Engineering

Quantitative Structure-Activity Relationships (QSARs) are empirical or semi-empirical models which correlate the structure of chemical compounds with their biological activities. QSAR analysis frequently finds application in drug development and environmental and human health protection. It is here that these models are employed to predict pharmacological endpoints for candidate drug molecules or to assess the toxicological potential of chemical ingredients found in commercial products, respectively. Fields such as drug design and health regulation share the necessity of managing a plethora of chemicals in which sufficient experimental data as to their application-relevant profiles is often lacking; the time and resources required to conduct the necessary in vitro and in vivo tests to properly characterize these compounds make a pure experimental approach impossible. QSAR analysis successfully alleviates the problems posed by these data gaps through interpretation of the wealth of information already contained in existing databases. This research involves the development of a novel QSAR workflow utilizing a local modeling strategy. By far the most common QSAR models reported in the literature are “global” models; they use all available training molecules and a single set of chemical descriptors to learn the relationship between structure and the endpoint of interest. Additionally, accepted QSAR models frequently use linear transformations such as principal component analysis or partial least squares regression to reduce the dimensionality of complex chemical data sets. To contrast these conventional approaches, the proposed methodology uses a locality-defining radius to identify a subset of training compounds in proximity to a test query to learn an individual model for that query. Furthermore, descriptor selection is utilized to isolate the subset of available chemical descriptors tailored specifically to explain the activity of each test compound. Finally, this (open full item for complete abstract)

Committee: James Rathman (Advisor); Bhavik Bakshi (Committee Member); Jeffrey Chalmers (Committee Member) Subjects: Chemical Engineering

Doctor of Philosophy (PhD), Wright State University, 2019, Interdisciplinary Applied Science and Mathematics PhD

This work contains the derivation and type analysis of the conical Euler and Ideal Magnetohydrodynamic equations. The 3 dimensional Euler equations and the Ideal MHD equations with Powell source terms, subject to the assumption that the solution is conically invariant, are projected onto a unit sphere using tools from tensor calculus. Conical flows provide valuable insight into supersonic and hypersonic flow past bodies, but are simpler to analyze and solve numerically. Previously, work has been done on conical inviscid flows governed by the compressible Euler equations with great success. It is known that some flight regimes involve flows of ionized gases, and thus there is motivation to extend the study of conical flows to the case where the gas is electrically conducting. This thesis shows that steady conical flows for these cases do exist mathematically and that the governing systems of partial differential equations are of mixed type. Throughout the domain they can be either hyperbolic or elliptic depending on the solution. A numerical scheme is also developed to solve the conical Euler and Ideal Magnetohydrodynamic equations. Special care had to be taken in developing the method because these equations contain geometric source terms which account for the fact that they are defined on a curved surface. In order for a numerical method to accurately capture the behavior of the system it is solving, any source terms must be discretized in a way which preserves the appropriate behavior. For a partial differential equation which has been formulated on a curved manifold using tensor calculus, it is desirable for the discretization to preserve the tensorial transformation relationships. Such discretizations are presented in this work, and a numerical method involving them is developed and demonstrated.

Committee: Sivaguru Sritharan Ph.D. (Committee Co-Chair); Qun Li Ph.D. (Committee Co-Chair); Qingbo Huang Ph.D. (Committee Member); Mohammed Sulman Ph.D. (Committee Member) Subjects: Applied Mathematics

PhD, University of Cincinnati, 2019, Engineering and Applied Science: Mechanical Engineering

Approximately 1.5 million senior citizens in the US live under nursing supervision, and most require assistance with at least one or more Activities of Daily Living (ADL) one of which is sit-to-stand (STS) transitions. The STS transition includes an inherent phase of instability which introduces a danger of falling for senior citizens and requires continuous supervision by healthcare workers or caregivers. Therefore, the development of assistive technologies to support human movements (i.e., exoskeletons, prostheses) has become a topic of increasing interest and urgency. Human motion is highly variable due to the effects of interaction with the environment and the intentionality of movement. Assistive robotic devices which aim to restore human motion need to account for and incorporate this variability in their operations. The aim of this dissertation is to analyze the dynamics of healthy STS transitions, present an approach to effectively plan STS trajectories, explore the efficacy of detecting intent of subsequent activity after STS, and analyze the effects of intent on the variability in human motion. Furthermore, the aim is to use the results obtained from the analysis of healthy STS transitions to develop a control strategy for exoskeletons which can exhibit the human-like variability behavior. In this work, an analysis of STS trajectories at different velocities and chair heights is presented which shows a clear correlation between the critical events (start of knee extension and time of weight transfer) and the way the momentum in modulated during the complete STS. Based on this, a model which approximates the velocities of the center of mass (CoM) in the vertical and horizontal directions and thus the whole-body momentum is presented and validated. The advantage of this model is that all the factors can be derived as a function of the total time required for STS. To analyze the effects of intent, an experiment with four subsequent actions of STS was des (open full item for complete abstract)

Committee: Manish Kumar Ph.D. (Committee Chair); Tamara Lorenz Ph.D. (Committee Chair); Adam Kiefer Ph.D. (Committee Member); Anca Ralescu Ph.D. (Committee Member); David Thompson Ph.D. (Committee Member) Subjects: Robots

Doctor of Philosophy, The Ohio State University, 2019, Statistics

With rapid increase in the quantity and complexity of available data, we have embraced the idea of using probability models for gaining deeper insights into data generating mechanisms. Probability density functions (PDFs) form the crux of such models and provide a rigorous framework for statistical inference. However, the representation space of PDFs is infinite-dimensional and non-linear, and relevant statistical tools need to be developed for use on this restricted function space of PDFs accordingly. This dissertation focuses on building a unified geometric framework for analyzing PDFs, and subsequently defining efficient computational tools for their statistical analysis. Specifically, we use a Riemannian geometric framework based on the nonparametric Fisher–Rao metric on the manifold of PDFs. Under the square-root density representation, the manifold can be identified with the positive orthant of the unit hypersphere, and the Fisher–Rao metric reduces to the standard L^2 metric. We consider three different theoretical and applied statistical problems, all of which utilize the fundamental idea of exploiting such a Riemannian structure of PDFs to perform valid statistical analysis in an efficient manner. First, we demonstrate the utility of geometry-based approaches in a class of PDF approximation methods. We propose a geometric framework for variational inference in Bayesian models, where we formulate the task of approximating the posterior density based on the α-divergence function, instead of the classical Kullback-Leibler divergence function. We also propose a novel gradient-based algorithm for the variational problem and examine its properties. Through multiple simulations and real-data applications, we demonstrate the utility of the proposed geometric framework and algorithm on several Bayesian models. Second, we consider the idea of assessing model performance under misspecifications, which is intricately linked to the geometry of the space of PDFs, a (open full item for complete abstract)

Committee: Sebastian Kurtek (Advisor); Catherine Calder (Committee Member); Steven MacEachern (Committee Member); Jennifer Sinnott (Committee Member) Subjects: Statistics

Doctor of Philosophy, The Ohio State University, 2018, Computer Science and Engineering

Shape analysis has been applied in many applications across a broad range of domains. Among various different families of ``complex'' shapes, the ones with 1D non-linear topological structures (skeletons) are particularly interesting. These shapes are simple, as they can be decomposed into 1-d pieces, but still informative in representing the connectivity and other important information behind data. In this thesis, I focus on two objects from computational topology that have been useful for modeling the skeleton of data: the Reeb graph (and its variants) and the 1-(un)stable manifold from discrete Morse theory, and study their properties as well as applications to shape analysis. The two specific topological objects that I focus on have both already been widely used in practical applications. Further theoretical understanding and applications of these two objects in modeling and studying the skeleton of data will be provided. The first part of the dissertation work concerns the so-called Reeb graph and its loop-free variant, the contour tree, which can be used to provide a 1D tree summary of an input scalar field. It has been commonly used in computer graphics and visualization. I have investigated one problem regarding to the theoretical understanding of the contour tree, as well as developing a variant of the Reeb graph to address the issue of noise. Carr et. al. has proposed an algorithm for computing the contour tree for a piece-wise linear function defined on a simplicial complex domain. This algorithm is simple, efficient and has been widely used in practice. However, the algorithm is often applied even when the output contour tree may not exist, in which case the algorithm may not terminate or exit with only partial output. My work provides new understanding of this contour tree algorithm and characterizes the cause for such behavior. I also propose a simple variation of the contour tree (called JS-graph) to handle this situation in p (open full item for complete abstract)

Committee: Yusu Wang (Advisor); Rephael Wenger (Committee Member); Tamal Dey (Committee Member) Subjects: Computer Engineering; Computer Science; Geographic Information Science; Neurosciences

BA, Oberlin College, 2017, Mathematics

In this paper we focus on the study of smooth closed simply-connected 4-manifolds. In particular, we study such manifolds arising from Artin presentations. Artin pre- sentations are intimately related to the pure braid group, and they characterize the fundamental groups of closed orientable 3-manifolds. Each Artin presentation r gives rise to a 4-manifold W4(r), whose boundary is a closed orientable 3-manifold M3(r) with fundamental group presented by r. In case the 4-manifold has bound- ary S3, we may close up W 4(r) by attaching a 4-handle to obtain W 4(r) ¿S3 D4. This closed 4-manifold has quadratic form represented by the exponent sum matrix of r. Artin presentations arising from 2-strand pure braids are particularly tractable. Further, we show that they are naturally related to von Dyck (triangle) groups. These two facts play crucial roles in our determination of all closed 4-manifolds arising from Artin presentations on two generators. Section 2 introduces the open book decomposition, which is an important con- struction of closed orientable 3-manifolds that leads to their correspondence with Artin presentations. Section 3 explains how Artin presentations give rise to 4- manifolds, specifically how the 3-manifold boundary of a 4-manifold can be studied with an Artin presentation group as their fundamental group. Then in section 4 we find all possible closed, smooth, simply-connected 4-manifolds from Artin pre- sentations on two generators.

Committee: Jack Calcut (Advisor) Subjects: Mathematics

Doctor of Philosophy, The Ohio State University, 2016, Electrical and Computer Engineering

This work advances state-of-the-art Radio Frequency (RF) emitter geolocation from an airborne or spaceborne antenna array. With an antenna array, geolocation is based on Direction of Arrival (DOA) estimation algorithms such as MUSIC. The MUSIC algorithm applies to arbitrary arrays of polarization sensitive antennas and yields high resolution. However, MUSIC fails to obtain its theoretical resolution for simultaneous, closely spaced, co-frequency signals. We propose the novel Nullspace MUSIC algorithm, which outperforms MUSIC and its existing modifications while maintaining MUSIC's fundamental orthogonality test. Nullspace MUSIC applies a divide-and-conquer approach and estimates a single DOA at a time. Additionally, an antenna array on an aircraft cannot be perfectly calibrated. RF waves are blocked, reflected, and scattered in a time-varying fashion by the platform around the antenna array. Consequently, full-wave electromagnetics simulations or demanding measurements of the entire platform cannot eliminate the mismatch between the true, in-situ antenna patterns and the antenna patterns that are available for DOA estimation (the antenna array manifold). Platform-induced manifold mismatch severely degrades MUSIC's resolution and accuracy. We show that Nullspace MUSIC improves DOA accuracy for well separated signals that are incident on an airborne antenna array. Conventionally, geolocation from a mobile platform draws Lines of Bearing (LOB) from the antenna array along the DOAs to find the locations where the DOAs intersect with the ground. However, averaging the LOBs in the global coordinate system yields large errors due to geometric dilution of precision. Since averaging positions fails, a single emitter is typically located by finding the position on the ground that yields the Minimum Apparent Angular Error (MAAE) for the DOA estimates over a flight. We extend the MAAE approach to cluster LOBs from multiple emitters. MAAE clustering geolocates multiple sim (open full item for complete abstract)

Committee: Inder Gupta (Advisor); Joel Johnson (Committee Member); Fernando Teixeira (Committee Member); Can Koksal (Committee Member) Subjects: Aerospace Engineering; Applied Mathematics; Computer Engineering; Computer Science; Electrical Engineering; Electromagnetics; Electromagnetism; Engineering; Experiments; Mathematics; Music; Remote Sensing; Scientific Imaging; Systems Design

Doctor of Philosophy, The Ohio State University, 2016, Mathematics

In this dissertation, we study the action dimension of general Artin groups. Please review the abstract in the dissertation to see a complete abstract.

Committee: Michael Davis (Advisor) Subjects: Mathematics

Doctor of Philosophy (Ph.D.), University of Dayton, 2015, Electrical Engineering

Recognizing multiple types of actions appearing in a continuous temporal order from a streaming video is the key to many possible applications ranging from real-time surveillance to egocentric motion for human computer interaction. Current state of the art algorithms are more focused either on holistic video representation or on finding a specific activity in video sequences. But the major drawback is that these algorithms work only on applications pertaining to unconstrained video search from the web and requires the complete sequence for reporting what kind of actions are present. In this dissertation, we propose an algorithm to detect and recognize multiple actions in a streaming sequence at every instant. This approach was successful in recognizing the type of action being performed and also provides a percentage of completion of that action at every instant in real-time. This system is invariant to the number of frames and the speed at which the action is being performed. Apart from these benefits, the proposed model can also predict the motion descriptors at future instances corresponding to the action present. Since human motion is inherently continuous in nature, the algorithm presented in this dissertation computes novel motion descriptors based on the dense optical flow at every instant and evaluates their variations along the temporal domain using deep learning techniques. For each action type, we compute a non-linear transformation from motion descriptor space into the latent temporal space using stacked autoencoders where this transformation is learned from its training patterns. The latent features thus obtained, forms a temporal manifold where the transitions along it are modeled using the Conditional Restricted Boltzmann Machines (CRBMs). Using these trained autoencoders and CRBMs for every action type, we can make an inference into multiple latent temporal action manifolds at an instant from a set of streaming input frames. Our model achieved (open full item for complete abstract)

Committee: Kimberly Kendricks (Committee Member); Keigo Hirakawa (Committee Member); Raul Ordonez (Committee Member); Vijayan Asari (Committee Chair) Subjects: Computer Engineering; Computer Science; Electrical Engineering; Statistics

Doctor of Philosophy, The Ohio State University, 2014, Linguistics

Vowel normalization is a computation that is meant to account for the differences in the absolute direct (physical or psychophysical) representations of qualitatively equivalent vowel productions that arise due to differences in speaker properties such as body size types, age, gender, and other socially interpreted categories that are based on natural variation in vocal tract size and shape. In this dissertation, we address the metaphysical and epistemological aspects of vowel normalization pertaining to spoken language acquisition during early infancy. We begin by reviewing approaches to conceptualizing and modeling the phonetic components of early spoken language acquisition, forming a catalog of phenomena that serves as the basis for our discourse. We then establish the existence of a vowel normalization computation carried out by infants early in their spoken language acquisition, and put forward a conceptual and technical framework for its investigation which focuses attention on the generative nature of the computation. We then situate the acquisition of vowel normalization within a broader developmental framework encompassing a suite of vocal learning phenomena, including language-specific caretaker vocal exchanges, perceptual warping, and multisensory matching and narrowing. We demonstrate the applicability of the technical formulation through the creation of a virtual environment for vocal learning which provides the means to model the acquisition of vowel normalization, along with other aspects of vocal learning. We conclude with a discussion of the broader implications of the conceptual and technical formulation.

Committee: Mary Beckman (Advisor); Eric Fosler-Lussier (Committee Member); William Schuler (Committee Member) Subjects: Linguistics

Master of Science, The Ohio State University, 2013, Mechanical Engineering

The performance of a liquid cooled exhaust manifold installed on a Ford 2011 3.5L V6 EcoBoost engine is quantified experimentally. An external cooling circuit was constructed to control the manifold coolant flow rate V_WCM, inlet temperature, and outlet pressure. The manifold has a jacket that is located near the exhaust outlet and cooled with a water/ethylene-glycol mixture from the external cooling circuit to reduce skin temperatures. The manifold is characterized in terms of: heat rejection rates, skin temperatures, and cooling jacket boiling. Results were obtained by sweeping V_WCM from 0.2 to 2.0 gpm while maintaining a nominal manifold inlet temperature and outlet pressure of 85 degC and 13 psig, respectively. For each V_WCM sweep, the engine operation was held constant. These sweeps were completed at a total of 12 engine operating points. Nine of these points were from experiments at a constant speed of 2000 rpm with BMEP ranging from 1 to 16 bar. The remaining 3 operating points were acquired at a constant nominal load of 16 bar and speeds of 2500, 3000, and 3500 rpm. The measurements include engine exhaust flow rate and exhaust gas temperature at the outlet of the manifold. Manifold coolant temperatures were measured at the jacket inlet and outlet. Manifold coolant flow rate was measured with a turbine flow meter installed upstream. Moreover, a dynamic pressure transducer was installed inside the cooling jacket to capture hydraulically and thermally induced pressure fluctuations. Finally, manifold outer skin temperatures were obtained at 12 different locations. The manifold coolant heat rejection rate was found to depend primarily on engine operation. Due to high heat transfer resistance on the exhaust gas side, only slight sensitivity to V_WCM was observed for power levels above 122 hp. The peak manifold heat rejection rate found was 3.2 kW for engine operation at 3500 rpm and 16 bar. When normalized by engine brake power, the manifold heat rejection r (open full item for complete abstract)

Committee: Ahmet Selamet (Advisor); Xiaodong Sun (Committee Member) Subjects: Automotive Engineering; Mechanical Engineering

PhD, University of Cincinnati, 2011, Engineering and Applied Science: Computer Science and Engineering

Graphs serve as an important tool for discrete data representation. Recently, graph representations have made possible very powerful machine learning algorithms, such as manifold learning, kernel methods, semi-supervised learning. With the advent of large-scale real world networks, such as biological networks (disease network, drug target network, etc.), social networks (DBLP Co-authorship network, Facebook friendship, etc.), machine learning and data mining algorithms have found new application areas and have contributed to advance our understanding of properties, and phenomena governing real world networks. When dealing with real world data represented as networks, two problems arise quite naturally: I) How to integrate and align the knowledge encoded in multiple and heterogeneous networks? For instance, how to find out the similar genes in co-disease and protein-protein interaction networks? II) How to model and predict the evolution of a dynamic network? A real world example is, given N years snapshots of an evolving social network, how to build a model that can capture the temporal evolution and make reliable prediction? In this dissertation, we present an innovative graph embedding framework, which identifies the key components of modeling the evolution in time of a dynamic graph. Different from the many state-of-the-art graph link prediction and modeling algorithms, it formulates the link prediction problem from a geometric perspective that can capture the dynamics of the intrinsic continuous graph manifold evolution. It is attractive due to its simplicity and the potential to relax the mining problem into a feasible domain which enables standard machine learning and regression models to utilize historical graph time series data. To address the first problem, we first propose a novel probability-based similarity measure which led to promising applications in content based image retrieval and image annotation, followed by a manifold alignment framework t (open full item for complete abstract)

Committee: Anca Ralescu PhD (Committee Chair); Anil Jegga DVMMRes (Committee Member); Fred Annexstein PhD (Committee Member); Kenneth Berman PhD (Committee Member); Yizong Cheng PhD (Committee Member); Dan Ralescu PhD (Committee Member) Subjects: Computer Science

PhD, University of Cincinnati, 2005, Arts and Sciences : Psychology

Synergies have been widely recognized as a way the CNS functionally deals with the degrees of freedom problem (Bernstein, 1967). Results of previous research have suggested the CNS organizes groups of motor elements for context-specific tasks. Recently, the uncontrolled manifold (UCM) hypothesis has been developed to quantitatively detect the existence of synergies and asses the strength of synergies (Scholz & Schoner, 1999). According to this hypothesis a synergy exists when a desired value of a performance variable is preserved by structuring variability within a subspace (the UCM) of the multidimensional space composed of the motor elements (Varcomp) and minimizing variability in the subspace orthogonal to the UCM (Varuncomp). The UCM approach was applied to interlimb rhythmic coordination to determine if relative phase is stabilized as a control variable in both intrapersonal and interpersonal tasks and if the ratio of Varcomp to Varuncomp is affected by parameters associated with reduced coordination stability. Participants oscillated detuned or non-detuned pendulum pairs in either inphase or antiphase coordination modes at or above the coupled wrist-pendulum system's eigenfrequency. Experiment 1 employed an intrapersonal coordination task while Experiment 2 employed an interpersonal coordination task. Experiment 3 explored the effects of anchoring in within-person coordination. The results of all three experiments were consistent with the HKB model predictions (e.g., greater variability in antiphase, with the detuned pendulums, and when the metronome frequency was greater than the coupled wrist-pendulum system's eigenfrequency). The UCM analysis quantitatively verified synergies exist during a rhythmic motor task within a single person and between two people. Greater stabilization of relative phase was observed during inphase than antiphase coordination (Varcomp > Varuncomp) and, in Experiments 1 and 2, at the endpoints of the movement cycles. The latter effec (open full item for complete abstract)

Committee: Michael Riley (Advisor) Subjects: Psychology, Experimental

Doctor of Philosophy, The Ohio State University, 2008, Computer Science and Engineering

Surveillance and monitoring are two important applications of computer vision research. Many computer vision algorithms have been proposed to monitor both indoor and outdoor spaces using one or more visual sensors. Finding the location of objects (detection) in complex real-world scenes and following their position over time (tracking) are two principal tasks for these systems. In this thesis, we develop a computational framework for object tracking that layers multiple sources of information (such as position, velocity, and appearance) by automatically selecting the most appropriate tracking algorithm based on the given scene context. The proposed framework is independent of the choice of tracking algorithms used and is capable of automatically evaluating the spatial context based on object interactions in the scene. Our approach is to employ multiple trackers that have complimentary modes of success and failures. Also, these tracking algorithms range from simple to complex in terms of computation. We do not always need to deploy the most expensive tracker in all cases. Furthermore, we also propose a set of novel computer vision algorithms for tracking objects in both 2D and 3D spaces. These algorithms address various shortcomings of the existing tracking approaches such as occlusion reasoning, information fusion, model update strategies, execution speeds, etc. The proposed algorithms are deployed in the aforementioned layered tracker switching framework, resulting in robust and efficient paradigm for tracking objects in complex urban environments. First, we present an online, recursive filtering technique to model linear dynamical systems that operate on the state space of symmetric positive definite (SPD) matrices that lie on a Riemannian manifold. This filtering approach is applied to the problem of object tracking by recursively estimating and updating the SPD covariance feature matrices representing objects in the scene. The online filtering process on the R (open full item for complete abstract)

Committee: James W. Davis PhD (Advisor); Richard Parent PhD (Committee Member); James Todd PhD (Committee Member) Subjects: Computer Science