2020 UMKC Dissertations - Freely Available Online
Permanent URI for this collection
Browse
Recent Submissions
Item The Nursing Workaround: Development of a Pyschometric Instrument(2021) McCord, Jennifer; Abreu, Eduardo (Professor)The development of nursing workarounds is a significant concern for healthcare organizations with the potential to have longstanding consequences to patient safety. Although numerous studies have been published about workarounds in general, little is known about influential factors resulting in workaround development by nurses. At this time, only one valid and reliable instrument is available to measure nursing workarounds, the Workaround Instrument (Halbesleben et. al., 2013), however, it does not examine the potential relationships between RN’s decision to employ a workaround and other demographic variables or other personal influences. The objective of this research was to psychometrically test the Workaround Motivation Survey (WMS). The WMS was designed specifically to distinguish between personal and professional motivational influences resulting in nursing workarounds as a method to predict those at greater risk for workaround development. The sample included nurses from four Mid-Western hospitals. Data were collected using RedCap, a web-based format that provides respondent confidentiality. Results indicate that the newly developed instrument is a reliable tool to identify nurses at greater risk for workaround development. Findings indicate the need for a larger sample size to accurately conduct factor analysis and to increase generalizability.Item Identification and Development of a Reliable Framework to Predict Passive Scalar Transport for Turbulent Bounded Shear Flows(2020) Ziefuss, Matthias; Mehdizadeh, Amirfarhang; Bani-Yaghoub, MajidHeat transfer modeling plays an integral role in optimization and development of highly efficient modern thermal-fluid systems. However, currently available heat flux models suffer from fundamental shortcomings. For example, their development is based on the general notion that an accurate prediction of the flow field will guarantee an appropriate prediction of the thermal field, as the Reynolds Analogy does. Furthermore, literature about advanced models that aim to overcome this notion, does not provide reliable information about prediction capabilities. These advanced models can be separated into two distinct heat flux model categories, namely the implicit and explicit models. Both model categories differ fundamentally in their mathematical and physical formulation. Hence, this dissertation presents a comprehensive assessment of the Reynolds Analogy regarding steady and unsteady calculations. It further analyses the entropy generation capability in detail and evaluates the prediction accuracy of implicit and explicit models when applied to turbulent shear flows of fluids with different Prandtl numbers. Moreover, the implicit and explicit models are modified such that important thermal second order statistics are included. This enables deeper insight into the mechanics of thermal dissipation and delivers a better understanding towards the sensitivity and reliability of predictions using heat flux models. Finally, to overcome the shortcomings of the Reynolds Analogy in unsteady calculations, an anisotropic extension is proposed. This dissertation shows that even for first order statistics within steady state calculations, the Reynolds Analogy is only appropriate for fluids with Prandtl numbers around unity. For second order statistics within unsteady simulations, the Reynolds Analogy could provide acceptable results only if an appropriate grid design/resolution is provided that allows resolving essential dynamics of the thermal field. Concerning entropy generation, the Reynolds Analogy provides acceptable results only for mean entropy generation, while it fails to predict entropy generation at small/sub-grid scales. The anisotropic extension of the Reynolds Analogy is a promising approach to overcome these shortcomings. Furthermore and concerning the implicit and explicit heat flux models, this work shows that only the explicit framework is potentially capable of dealing with complex turbulent thermal fields and to address longstanding shortcomings of currently available models, if the flow field is predicted accurately. Moreover, it has been shown that thermal time scale plays an integral role to predict thermal phenomena, particularly those of fluids with low/high Pr numbers.Item Crosstalk computing: circuit techniques, implementation and potential applications(2020) Macha, Naveen Kumar; Rahman, MostafizurThis work presents a radically new computing concept for digital Integrated Circuits (ICs), called Crosstalk Computing. The conventional CMOS scaling trend is facing device scaling limitations and interconnect bottleneck. The other primary concern of miniaturization of ICs is the signal-integrity issue due to Crosstalk, which is the unwanted interference of signals between neighboring metal lines. The Crosstalk is becoming inexorable with advancing technology nodes. Traditional computing circuits always tries to reduce this Crosstalk by applying various circuit and layout techniques. In contrast, this research develops novel circuit techniques that can leverage this detrimental effect and convert it astutely to a useful feature. The Crosstalk is engineered into a logic computation principle by leveraging deterministic signal interference for innovative circuit implementation. This research work presents a comprehensive circuit framework for Crosstalk Computing and derives all the key circuit elements that can enable this computing model. Along with regular digital logic circuits, it also presents a novel Polymorphic circuit approach unique to Crosstalk Computing. In Polymorphic circuits, the functionality of a circuit can be altered using a control variable. Owing to the multi-functional embodiment in polymorphic-circuits, they find many useful applications such as reconfigurable system design, resource sharing, hardware security, and fault-tolerant circuit design, etc. This dissertation shows a comprehensive list of polymorphic logic gate implementations, which were not reported previously in any other work. It also performs a comparison study between Crosstalk polymorphic circuits and existing polymorphic approaches, which are either inefficient due to custom non-linear circuit styles or propose exotic devices. The ability to design a wide range of polymorphic logic circuits (basic and complex logics) compact in design and minimal in transistor count is unique to Crosstalk Computing, which leads to benefits in the circuit density, power, and performance. The circuit simulation and characterization results show a 6x improvement in transistor count, 2x improvement in switching energy, and 1.5x improvement in performance compared to counterpart implementation in CMOS circuit style. Nevertheless, the Crosstalk circuits also face issues while cascading the circuits; this research analyzes all the problems and develops auxiliary circuit techniques to fix the problems. Moreover, it shows a module-level cascaded polymorphic circuit example, which also employs the auxiliary circuit techniques developed. For the very first time, it implements a proof-of-concept prototype Chip for Crosstalk Computing at TSMC 65nm technology and demonstrates experimental evidence for runtime reconfiguration of the polymorphic circuit. The dissertation also explores the application potentials for Crosstalk Computing circuits. Finally, the future work section discusses the Electronic Design Automation (EDA) challenges and proposes an appropriate design flow; besides, it also discusses ideas for the efficient implementation of Crosstalk Computing structures. Thus, further research and development to realize efficient Crosstalk Computing structures can leverage the comprehensive circuit framework developed in this research and offer transformative benefits for the semiconductor industry.Item Ocular motion classification for mobile device presentation attack detection(2020) Lowe, Jesse; Derakhshani, RezaAs a practical pursuit of quantified uniqueness, biometrics explores the parameters that make us who we are and provides the tools we need to secure the integrity of that identity. In our culture of constant connectivity, an increasing reliance on biometrically secured mobile devices is transforming them into a target for bad actors. While no system will ever prevent all forms of intrusion, even state of the art biometric methods remain vulnerable to spoof attacks. As these attacks become more sophisticated, ocular motion based presentation attack detection (PAD) methods provide a potential deterrent. This dissertation presents the methods and evaluation of a novel optokinetic nystagmus (OKN) based PAD system for mobile device applications which leverages phase-locked temporal features of a unique reflexive behavioral response. Background is provided for historical and literary context of eye motion and ocular tracking to provide context to the objectives and accomplishments of this work. An evaluation of the improved methods for sample processing and sequential stability is provided with highlights for the presented improvements to the stability of convolutional facial landmark localization, and automated spatiotemporal feature extraction and classification models. Insights gleaned from this work are provided to elucidate some of the major challenges of mobile ocular motion feature extraction, as well as additional future considerations for the refinement and application of OKN motion signatures as a novel mobile device based PAD method.Item Automated End-to-End Management of the Deep Learning Lifecycle(2020) Gharibi, Gharib; Lee, Yugyung, 1960-Deep learning has improved the state-of-the-art results in an ever-growing number of domains. This success heavily relies on the development of deep learning models--an experimental, iterative process that produces tens to hundreds of models before arriving at a satisfactory result. While there has been a surge in the number of tools and frameworks that aim at facilitating deep learning, the process of managing the models and their artifacts is still surprisingly challenging and time-consuming. Existing model-management solutions are either tailored for commercial platforms or require significant code changes. Moreover, most of the existing solutions address a single phase of the modeling lifecycle, such as experiment monitoring, while ignoring other essential tasks, such as model sharing and deployment. In this dissertation, we present a software system to facilitate and accelerate the deep learning lifecycle, named ModelKB. ModelKB can \textit{automatically} manage the modeling lifecycle end-to-end, including (1) monitoring and tracking experiments; (2) visualizing, searching for, and comparing models and experiments; (3) deploying models locally and on the cloud; and (4) sharing and publishing trained models. Our system also provides a stepping-stone for enhanced reproducibility. ModelKB currently supports TensorFlow 2.0, Keras, and PyTorch, and it can be extended to other deep learning frameworks easily. A video demo is available at https://youtu.be/XWiJpSM_jvA. Moreover, we study static call graphs to form a stepping-stone to facilitate the \textit{comprehension} of the overall lifecycle implementation (i.e., source code). Specifically, we introduce Code2Graph to facilitate the exploration and tracking of the implementation and its changes over time. Code2Graph is used to construct and visualize the call graph of a software codebase. We evaluate the functionality by analyzing and studying real software systems throughout their entire lifespan. The tool, evaluation results, and a video demo are available at https://goo.gl/8edZ64. Finally, we demonstrate a software system that brings together the contributions mentioned above to build a robust, open-collaborative platform for deep learning applications in the health domain, named Medl.AI. Medl.AI enables researchers and healthcare professionals to easily and efficiently: explore, share, reuse, and discuss deep learning models specific to the medical domain. We present six illustrative deep learning medical applications using Medl.AI. We conduct an online survey to assess the feasibility and benefits of Medl.AI. The user study suggests that Medl.AI provides a promising solution to open collaborative research and applications. Our live website is currently available at http://medl.ai.