DissertationsCopyright (c) 2016 Carnegie Mellon University All rights reserved.
http://repository.cmu.edu/dissertations
Recent documents in Dissertationsen-usMon, 01 Aug 2016 14:24:35 PDT3600System-Level Adaptive Monitoring and Control of Infrastructures: A POMDP-Based Framework
http://repository.cmu.edu/dissertations/664
http://repository.cmu.edu/dissertations/664Thu, 28 Jul 2016 07:49:38 PDT
Many infrastructure systems in the US such as road networks, bridges, water and wastewater pipelines, and wind farms are aging and their condition are deteriorating. Accurate risk analysis is crucial to extend the life span of these systems, and to guide decision making towards a sustainable use of resources. These systems are subjected to fatigue-induced degradation and need periodic inspections and repairs, which are usually performed through semi-annual, annual, or bi-annual scheduled maintenance. However, better maintenance can be achieved by flexible policies based on prior knowledge of the degradation process and on data collected in the field by sensors and visual inspections. Traditional methods to model the operation and maintenance (O&M) process, such as Markov decision processes (MDP) and partially observable MDP (POMDP) have limitations that do not allow the model to properly include the knowledge available and that may result in nonoptimal strategies for management of infrastructure systems. Specifically, the conditional probabilities for modeling the degradation process and the precision of the observations are usually affected by epistemic uncertainty: this cannot be captured by traditional methods. The goal of this dissertation is to propose a computational framework for adaptive monitoring and control of infrastructures at the system-level and to connect different aspects of the management process together. The first research question we address is how to take optimal sequential decisions under model uncertainty. Second, we propose how to combine decision optimization with learning of the degradation of components and the precision of monitoring system. Specifically, we address the issue of systems made by similar components, where iv transfer of knowledge across components is relevant. Finally, we propose how to assess the value of information in sequential decision making and whether it can be used as a heuristic for system-level inspection scheduling. In this dissertation, first a novel learning and planning method is proposed, called “Planning and Learning for Uncertain dynamic Systems” (PLUS), that can learn from the environment, update the distributions of parameters, and select the optimal strategy considering the uncertainty related to the model. Validating with synthetic data, the total management cost of operating a wind farm using PLUS is shown to be significantly less than costs achieved by a fixed policy or though the POMDP framework. Moreover, when the system is made up by similar components, data collected on one is also relevant in the management of others. This is typically the case of wind farms, which are made up by similar turbines. PLUS models the components as independent or identical and eithers learn the model for each component independently or learn a global model for all components. We extend that formulation, allowing for a weaker similarity among components. The proposed approach, called “Multiple Uncertain POMDP” (MU-POMDP), models the components as POMDPs, and assumes the corresponding model parameters as dependent random variables. By using this framework, we can calibrate specific degradation and emission models for each component while, at the same time, processing observations at the level of the entire system. We evaluate the performance of MU-POMDP compared to PLUS and discuss its potentials and computational complexity. Lastly, operation and maintenance of an infrastructure system rely on information collected on its components, which can provide the decision maker with an accurate assessment of their condition states. However, resources to be invested in data gathering are usually limited and v observations should be collected based on their value of information (VoI). VoI is a key concept for directing explorative actions, and in the context of infrastructure operation and maintenance, it has application to decisions about inspecting and monitoring the condition states of the components. Assessing the VoI is computationally intractable for most applications involving sequential decisions, such as long-term infrastructure maintenance. The component-level VoI can be used as a heuristic for assigning priorities to system-level inspection scheduling. In this research, we propose two alternative models for integrating adaptive maintenance planning based on POMDP and inspection scheduling based on a tractable approximation of VoI: the stochastic allocation model (and its two limiting scenarios called pessimistic and optimistic) that assumes observations are collected with a given probability, and the fee-based allocation model that assumes observations are available at a given cost. We illustrate how these models can be used at component-level and for system-level inspection scheduling. Furthermore, we evaluate the quality of solution provided by pessimistic and optimistic approaches. Finally, we introduce analytical formulas based on the stochastic and fee-based allocation models to predict the impact of a monitoring system (or a piece of information) on the operation and maintenance cost of infrastructure systems.
]]>
Milad MemarzadehAn Embryonic-Inspired Approach to Engineer Functional Human Cardiac Tissue
http://repository.cmu.edu/dissertations/663
http://repository.cmu.edu/dissertations/663Thu, 28 Jul 2016 06:17:49 PDTQuentin JalleratSetting Drinking Water Standards: Historical Perspective and Simulation Modeling
http://repository.cmu.edu/dissertations/662
http://repository.cmu.edu/dissertations/662Tue, 21 Jun 2016 07:09:36 PDT
Setting drinking water standards involves both technical knowledge and an understanding of societal values and institutions. To provide perspective and tools for evaluating these issues a set of historical and current regulatory assessments are presented here. The first of these case studies considers the history of the 1914 Public Health Service drinking water standards and is based on information in archival materials and journal articles of the time period. A simulation model to estimate the costs and benefits of proposed drinking water regulations on U.S. community water systems is then developed. The model simulates current contaminant concentrations and existing treatment types based on fitted statistical models. For systems that exceed any of the drinking water standards included in the model, the costs and effectiveness of alternative compliance strategies are simulated, and the system is assumed to select the least costly strategy capable of achieving compliance with the standards. This modeling approach allows for quantitative estimates of the uncertainty in regulatory impacts, geographic and size class specificity, and the consideration of multiple standards simultaneously. The model is applied first to the case of a lower drinking water standard for arsenic. The marginal cost-effectiveness of different standards and the impacts of several alternative regulatory approaches are considered. Discrepancies in previous estimates of compliance costs are shown to result primarily from differences in the treatment process cost estimates used by the different studies. An evaluation of alternative regulatory approaches for arsenic indicates that point-of-use treatment has the potential to be a lowcost means of compliance for smaller water systems but would most likely provide less uniform water quality than centralized treatment, with costs and performance highly dependent on the frequency of monitoring and service. The simulation model is then applied to consider jointly standards for three contaminants: arsenic, nitrate, and uranium. The costs and benefits of imposing the three standards simultaneously are smaller than the sum of the costs and benefits of the individual standards. For these contaminants the difference between the sum of the individual analyses and the integrated analysis is fairly small, but the effects of joint regulation may be larger for contaminants with more highly correlated occurrence distributions.
]]>
Patrick Lee GurianEnhanced Digital Design Exploration and Recovery through 3D Wireframe Assistance
http://repository.cmu.edu/dissertations/661
http://repository.cmu.edu/dissertations/661Thu, 16 Jun 2016 09:46:57 PDT
Almost seventy percent of final product forms are determined at conceptual shape modeling and exploration stages which involve generation of a rich set of geometric proxies with the assistance of various computer-aided design (CAD) and engineering (CAE) tools [1]. This practice facilitates existing computational tools in congruent with visual and physical references such as rough sketches, 2D line drawings, prototypes or 3D wireframe drawings. Although these initial proxies can be ample to capture the intended forms with sufficient accuracy, design practice is impeded by current CAD tools as they require careful and laborious dictation and control of the geometric data for digital content generation and recovery. The root cause is lack of computational support for utilization of design content from primitive proxies that moves designers’ focus from aesthetic shape ideation to laborious digital model construction or modification. In this research, we propose new computational tools to remedy technical challenges that prevent utilization of geometric content from primitive proxies in construction or recovery of approximate geometrical models for rapid exploration and reverse engineering purposes. The overarching objective of this Ph.D. research is identification and utilization of geometric design content that is common among alternative form candidates in a conceptual design activity. If this can be achieved, novel computational tools that will enable rapid generation and modifications of digital forms by alleviating redundant and laborious work flows required by existing CAD tools can be developed. To achieve this objective we identified three main technical goals: (1) rapid conversion of design information contained in primitive shape proxies without any topological or geometric constraints into a digital data that is suitable for further beatification and refinement using conventional CAD tools, (2) an automatic computational tool for deformation transfer from physical prototypes to expedite digital shape exploration and design recovery from physical prototypes (3) sketch-based computational techniques that allow rapid topology insertions and modifications to provide effortless transitions between physical and digital media for final product form quest.
]]>
Erhan Batuhan ArisoyInforming Ankle-Foot Prosthesis Design and Prescription Through Systematic Experimentation with a Tethered Robotic Prosthesis
http://repository.cmu.edu/dissertations/660
http://repository.cmu.edu/dissertations/660Thu, 16 Jun 2016 09:26:05 PDT
Recently-developed robotic prostheses have demonstrated that it is possible to design a prosthesis which makes it easier for unilateral transtibial amputees to walk. Unfortunately, it is unclear which design features are most important and which users will benefit most from these advanced technologies that increase prosthesis cost by an order of magnitude. I developed a novel experimental approach to resolving these design and prescription uncertainties. Candidate prosthetic feet are emulated during treadmill walking experiments using a high-performance o-board actuated and controlled lightweight robotic prosthesis. Prosthesis behavior is systematically varied while users' walking economy, performance, and satisfaction are measured. This process thereby determines unambiguous relationships between device behavior and outcomes of interest. In Chapter 1 of this thesis I motivate the approach. In Chapter 2 I detail the design and evaluation of the novel prosthesis emulator system. Then, in Chapter 3, I detail an experiment in which I test the simple walking model prediction that increasing prosthetic ankle push-o work will lessen leading limb collision, thereby reducing users' metabolic energy consumption. I demonstrate that increased push-o instead seems to primarily reduce energy consumption by aiding in the acceleration of the swing leg. In Chapter 4, I emulate the behavior of o-the-shelf prostheses, giving patients the opportunity to test-drive candidate devices prior to purchase, and enabling prescriptions to be justified by predictive experimental data. Finally, in Chapter 5, I demonstrate a human-in-the-loop prosthesis design optimization scheme that enables the manufacture of user-customized prostheses, which could ultimately supersede the need for prosthesis selection. vi vii This thesis lends insight into how powered ankle-foot prostheses can make walking easier and demonstrates that the degree to which individual users will benefit is highly dependent on their specific needs, expected walking conditions, and the choice of outcome measures. I hope that this work serves to improve mathematical models of human walking and contributes to a shift towards individualized prosthesis design and prescription.
]]>
Joshua M. Caputo3D Manipulation of Objects in Photographs
http://repository.cmu.edu/dissertations/659
http://repository.cmu.edu/dissertations/659Thu, 16 Jun 2016 08:42:25 PDT
This thesis describes a system that allows users to to perform full three-dimensional manipulations to objects in photographs. Cameras and photo-editing tools have contributed to the explosion in creative content by democratizing the process of creating visual realizations of users’ imaginations. However, shooting photographs using a camera is constrained by real-world physics, while existing photo-editing software is largely restricted to the 2D plane of the image. 3D object edits, intuitive to humans, are simply not possible in photo-editing software. The fundamental challenge in providing 3D object manipulation is that estimating the 3D structure of the object, including the geometry and appearance of object parts hidden from the viewpoint of the camera is ill-posed. 3D object manipulations reveal hidden parts of objects that were not previously seen from the viewpoint of the camera. The key contributions of this thesis are algorithms that leverage 3D models from public repositories to obtain a three-dimensional representation of objects in photographs for 3D manipulation with seamless transition in appearance of the object from the original photograph. 3D models of objects in online repositories cannot be directly used to manipulate photographed objects, as they show mismatches in geometry and appearance, and do not contain three-dimensional illumination representing the scene where the photograph was captured. The work in this thesis provides a system that align the 3D model geometry, estimates three-dimensional illumination, and completes the appearance over the object in three dimensions to provide full 3D manipulation. To correct the mismatch between the geometry of the 3D model and the photographed object, the thesis presents an automatic model alignment technique that performs an exhaustive search in the space of viewpoint, object location, scale, and non-rigid deformation. We also provide a manual geometry adjustment tool that allows users to perform final corrections while imposing smoothness and symmetry constraints. Given the matched geometry, we present an illumination estimation approach that uses the visible pixels to obtain three-dimensional environment illumination that produces plausible effects such as cast shadows and smooth surface shading. Our appearance completion approach relates visible parts of the object to hidden parts using symmetries over the publicly available 3D model. Our interactive system for editing photographs re-imagines typical photo-editing operations such as rotation, translation, copy-paste, scaling, and deformation as 3D manipulations to objects. Using our system, users have created a variety of manipulations to photographs, such as flipping cars, making dynamic compositions of multiple objects suspended in the air, performing animations, and altering the stories of historical images and personal photographs.
]]>
Natasha Kholgade BanerjeeBayesian Aggregation of Evidence for Detection and Characterization of Patterns in Multiple Noisy Observations
http://repository.cmu.edu/dissertations/658
http://repository.cmu.edu/dissertations/658Thu, 16 Jun 2016 07:35:50 PDT
Effective use of Machine Learning to support extracting maximal information from limited sensor data is one of the important research challenges in robotic sensing. This thesis develops techniques for detecting and characterizing patterns in noisy sensor data. Our Bayesian Aggregation (BA) algorithmic framework can leverage data fusion from multiple low Signal-To-Noise Ratio (SNR) sensor observations to boost the capability to detect and characterize the properties of a signal generating source or process of interest. We illustrate our research with application to the nuclear threat detection domain. Developed algorithms are applied to the problem of processing the large amounts of gamma ray spectroscopy data that can be produced in real-time by mobile radiation sensors. The thesis experimentally shows BA’s capability to boost sensor performance in detecting radiation sources of interest, even if the source is faint, partiallyoccluded, or enveloped in the noisy and variable radiation background characteristic of urban scenes. In addition, BA provides simultaneous inference of source parameters such as the source intensity or source type while detecting it. The thesis demonstrates this capability and also develops techniques to efficiently optimize these parameters over large possible setting spaces. Methods developed in this thesis are demonstrated both in simulation and in a radiation-sensing backpack that applies robotic localization techniques to enable indoor surveillance of radiation sources. The thesis further improves the BA algorithm’s capability to be robust under various detection scenarios. First, we augment BA with appropriate statistical models to improve estimation of signal components in low photon count detection, where the sensor may receive limited photon counts from either source and/or background. Second, we develop methods for online sensor reliability monitoring to create algorithms that are resilient to possible sensor faults in a data pipeline containing one or multiple sensors. Finally, we develop Retrospective BA, a variant of BA that allows reinterpretation of past sensor data in light of new information about percepts. These Retrospective capabilities include the use of Hidden Markov Models in BA to allow automatic correction of a sensor pipeline when sensor malfunction may be occur, an Anomaly- Match search strategy to efficiently optimize source hypotheses, and prototyping of a Multi-Modal Augmented PCA to more flexibly model background and nuisance source fluctuations in a dynamic environment.
]]>
Prateek TandonA Macro-scale, Tribological Modeling Framework for Simulating Multiple Lubrication Regimes and Engineering Applications
http://repository.cmu.edu/dissertations/657
http://repository.cmu.edu/dissertations/657Thu, 16 Jun 2016 07:05:24 PDT
Tribology is the science of interacting surfaces and the associated study of friction lubrication and wear. High friction and wear cause energy loss and deterioration of interacting surfaces. Lubrication, using hydrodynamic liquids is the primary mechanism to reduce friction and wear. Unfortunately, not all applications can be ideally lubricated to operate in a low friction zone. In the majority of cases, the relative velocities between the moving components is either too low, or the transferred force is too high for them to get perfectly lubricated, with minimal solid to solid contact. In such conditions they operate in the boundary or mixed lubrication regime, where there is significant solid-solid contact. Examples of such conditions are commonplace in our daily lives. From the food in our mouth to a floating hard disk drive read/write heads, or artificial hip joints to a polishing process, all operate in the mixed lubrication regime. In this thesis, a generalized numerical modeling framework has been developed that can be applied to simulate the operation of a large variety tribological applications that operate in any of the three lubrication regimes. The framework called the Particle Augmented Mixed Lubrication - Plus (PAML+), accounts for all the major mechanical interactions encountered in any tribosystem. It involves coupled iii iv ABSTRACT modules for solid mechanics and fluid mechanics. Depending on the application, additional fidelity has been added in the form of modules relevant to the physical interactions unique to the application. For example, modeling of the chemical mechanical polishing process requires treatment of particle dynamics and wear to be able to generate predictions of meaningful quantities such as the material removal rate. Similarly, modeling of artificial hip joints requires additional treatment of mass transport and wear to simulate contamination with debris particles. The fluid mechanics have been modeled through the thin film approximation of Navier Stokes equations, known as the Reynolds Equation. The solid mechanics have been modeled using analytical or semi-analytical techniques. Statistical treatments have been applied to model particle dynamics wherever required to avoid huge computational requirements associated with deterministic methods such as the discrete element method. To demonstrate the strengths and general applicability of the modeling approach, four major tribological applications have been modeled using the new modeling approach in order to broadly impact key industries. The four tribological applications are (i) Pin-on-disk tribosystems (ii) Chemical mechanical polishing (CMP) (iii) Artificial hip joints, and (iv) Mechanical seals. First the model was employed to simulate pin-on-disk interfaces to evaluate different surface texture designs. It also served as a platform to test the model’s ability to capture, and seamlessly traverse through different lubrication regimes. The model predicted that an intermediate texture dimension of 200mm resulted in 80% lesser wear than a larger texture of 200mm, and up to 90% lesser wear than an untextured sample. Second, the framework was employed to study the CMP process. Overall, the model was found to be at least 50% more accurate than the previous generation model. Third, the model v was tailored to study the artificial joints. Wear predictions from the model remained within 5% error upon comparing against the experiments, while studying different “head” sizes. It was discovered that textured joints can reduce the concentration of the wear debris by at least 2:5% per cycle. For an expected lifetime of 12 years, that translates to lifetime enhancement of 3 months. Lastly, the model was employed to study the performance of mechanical seals. Even though the model was much more computationally efficient, it remained within 5% of much more detailed and computationally expensive FEA models. The model also predicted that the seals allow the highest leakage at shaft speeds of about 950 RPM.
]]>
Gagan SrivastavaTilt-Dependent Analysis of Diffuse X-ray Scattering from Oriented Stacks of Fluid Phase Lipid Bilayers
http://repository.cmu.edu/dissertations/656
http://repository.cmu.edu/dissertations/656Thu, 16 Jun 2016 05:36:31 PDT
Recent simulations have indicated that the traditional Helfrich-Canham model for topographical fluctuations in fluid phase biomembranes should be enriched to include molecular tilt. Experimental evidence supporting the aforementioned enrichment is reported.
]]>
Michael JablinGeometric Modeling and Shape Analysis for Biomolecular Complexes Based on Eigenfunctions
http://repository.cmu.edu/dissertations/655
http://repository.cmu.edu/dissertations/655Wed, 15 Jun 2016 12:00:29 PDT
Geometric modeling of biomolecules plays an important role in the study of biochemical processes. Many simulation methods depend heavily on the geometric models of biomolecules. Among various studies, shape analysis is one of the most important topics, which reveals the functionalities of biomolecules.
]]>
Tao LiaoQuantification of Temperature Implications and Investigation of Battery Design Options for Electrified Vehicles
http://repository.cmu.edu/dissertations/654
http://repository.cmu.edu/dissertations/654Sat, 11 Jun 2016 06:58:09 PDT
Battery cost, limited battery life and range anxiety are some barriers to widespread adoption of electrified vehicles. This thesis examines the implications of these issues with a particular focus of analyzing the effect of temperature by addressing several questions: How do range, emissions, and battery life vary with regional climate and driving patterns? How much does thermal management affect these outcomes? How does the cost-minimizing battery design change with chemistry? A modeling and simulation approach is followed throughout the thesis, where physics based models, as well as models based on real world and experimental data are developed to address the aforementioned questions. Battery electrical, thermal and life models are created to estimate battery degradation under various different usage scenarios, and the effect of air-cooling on improving battery life is investigated. Real world driving data and dynamometer test data are used to estimate driving behavior, and are combined with regional effects of climate and electrical grid mix to evaluate emissions benefits of vehicle electrification across different regions. A battery cost model is used as an objective function in a mixed integer nonlinear program to find the battery design that minimizes the purchase cost for different battery chemistries. Sensitivity analyses are performed to understand the effect of modeling assumptions and design decisions on the results. Results indicate that battery degradation is particularly sensitive to battery and vehicle design characteristics, such as battery size and powertrain control strategies. In addition, operational factors that change regionally, such as driving cycle and climate, can have significant implications. Aggressive driving can decrease battery life by 67% compared to average driving conditions, and battery life is about 46% shorter in Phoenix than in San Francisco. However battery life can be doubled if battery is thermally conditioned by air-cooling. Regional climate has also significant implications on battery electric vehicle range and energy consumption. Annual energy consumption of battery electric vehicles can increase by an average of 15% in the Upper Midwest or in the Southwest compared to the Pacific Coast due to temperature differences, and cold climate regions can encounter days with substantial reduction in EV range. vii Environmental benefits of electrified vehicles vary substantially by vehicle model and region: The Nissan Leaf battery electric vehicle creates lower GHG emissions than the most efficient gasoline vehicle (Toyota Prius) in most of the country except in the Midwest and the South. The Chevrolet Volt plug-in hybrid electric vehicle has higher emissions than the Prius everywhere. Regional grid mix, temperature, driving patterns, and vehicle model all have significant implications on the relative benefits of PEVs versus gasoline vehicles. Similar to degradation profile and environmental benefits, the cost minimizing design depends on battery energy requirement as well. As the energy requirement from the battery pack increases and the pack gets bigger, optimum design uses the maximum allowable cathode thickness. Among the chemistries explored, Lithium Manganese Oxide (LMO) provides the battery design with the least expensive production cost for vehicles with small size batteries; however as battery size increases it becomes comparable with other chemistries. Lithium Iron Phosphate (LFP) based batteries lead to most expensive design.
]]>
Tugre YukselSupporting Learner-Controlled Problem Selection in Intelligent Tutoring Systems
http://repository.cmu.edu/dissertations/653
http://repository.cmu.edu/dissertations/653Sat, 11 Jun 2016 06:34:49 PDT
Many online learning technologies grant students great autonomy and control, which imposes high demands for self-regulated learning (SRL) skills. With the fast development of online learning technologies, helping students acquire SRL skills becomes critical to student learning. Theories of SRL emphasize that making problem selection decisions is a critical SRL skill. Research has shown that appropriate problem selection that fit with students’ knowledge level will lead to effective and efficient learning. However, it has also been found that students are not good at making problem selection decisions, especially young learners. It is critical to help students become skilled in selecting appropriate problems in different learning technologies that offer learner control. I studied this question using, as platform, a technology called Intelligent Tutoring Systems (ITSs), a type of advanced learning technology that has proven to be effective in supporting students’ domain level learning. It has also been used to help students learn SRL skills such as help-seeking and self-assessment. However, it is an open question whether ITS can be designed to support students’ learning of problem selection skills that will have lasting effects on their problem selection decisions and future learning when the tutor support is not in effect. ITSs are good at adaptively selecting problems for students based on algorithms like Cognitive Mastery. It is likely, but unproven, that ITS problem selection algorithms could be used to provide tutoring on students’ problem selection skills through features like explicit instructions and instant feedback. Furthermore, theories of SRL emphasize the important role of motivations in facilitating effective SRL processes, but not much prior work in ITS has integrated designs that could foster the motivations (i.e., motivational design) to stimulate and sustain effective problem selection behaviors. Lastly, although students generally appreciate having learner control, prior research has found mixed results concerning the effects of learner control on students’ domain level learning outcomes and motivation. There is need to investigate how learner control over problem selection can be designed in learning technologies to enhance students’ learning and motivation. My dissertation work consists of two parts. The first part focuses on creating and scaffolding shared student/system control over problem selection in ITSs by redesigning an Open Learner Model (OLM, visualizations of learning analytics that show students’ learning progress) and integrating gamification features to enhance students’ domain level learning and enjoyment. I conducted three classroom experiments with a total of 566 7th and 8th grade students to investigate the effectiveness of these new designs. The results of the experiments show that an OLM can be designed to support students’ self-assessment and problem selection, resulting in greater learning gains in an ITS when shared control over problem selection is enabled. The experiments also showed that a combination of gamification features (rewards plus allowing re-practice of completed problems, a common game design pattern) integrated with shared control was detrimental to student learning. In the second part of my dissertation, I apply motivational design and user-centered design techniques to extend an ITS with shared control over problem selection so that it helps students learn problem selection skills, with a lasting effect on their problem selection decisions and future learning. I designed a set iv of tutor features that aim at fostering a mastery-approach orientation and learning of a specific problem selection rule, the Mastery Rule. (I will refer to these features as the mastery-oriented features.) I conducted a fourth classroom experiment with 200 6th – 8th grade students to investigate the effectiveness of shared control with mastery-oriented features on students’ domain level learning outcomes, problem selection skills and enjoyment. This experiment also measured whether there were lasting effects of the mastery-oriented shared control on students’ problem selection decisions and learning in new tutor units. The results of the experiment show that shared control over problem selection accompanied by the mastery-oriented features leads to significantly better learning outcomes, as compared to full system-controlled problem selection in the ITS. Furthermore, the mastery-oriented shared control has lasting effects on students’ declarative knowledge of problem selection skills. Nevertheless, there was no effect on future problem selection and future learning, possibly because the tutor greatly facilitated problem selection (through its OLM and badges). My dissertation contributes to the literatures on the effects of learner control on students’ domain level learning outcomes in learning technologies. Specifically, I have shown that a form of learner control (i.e., shared control over problem selection, with mastery-oriented features) can lead to superior learning outcomes than system-controlled problem selection, whereas most prior work has found results in favor of system control. I have also demonstrated that Open Learner Models can be designed to enhance student learning when shared control over problem selection is provided. Further, I have identified a specific combination of gamification features integrated with shared control that may be detrimental to student learning. A second line of contributions of my dissertation concerns research on supporting SRL in ITSs. My work demonstrates that supporting SRL processes in ITSs can lead to improved domain level learning outcomes. It also shows that the shared control with mastery-oriented features have lasting effects on improving students’ declarative knowledge of problem selection skills. Regarding using ITSs to help students learn problem selection skill, the user-centered motivational design identifies mastery-approach orientation as important design focus plus tutor features that can support problem selection in a mastery-oriented way. Lastly, the dissertation contributes to human-computer interaction by generating design recommendations for how to design learner control over problem selection in learning technologies that can support students’ domain level learning, motivation and SRL.
]]>
Yanjin LongSupervised Descent Method
http://repository.cmu.edu/dissertations/652
http://repository.cmu.edu/dissertations/652Sat, 11 Jun 2016 06:16:59 PDT
In this dissertation, we focus on solving Nonlinear Least Squares problems using a supervised approach. In particular, we developed a Supervised Descent Method (SDM), performed thorough theoretical analysis, and demonstrated its effectiveness on optimizing analytic functions, and four other real-world applications: Inverse Kinematics, Rigid Tracking, Face Alignment (frontal and multi-view), and 3D Object Pose Estimation. In Rigid Tracking, SDM was able to take advantage of more robust features, such as, HoG and SIFT. Those non-differentiable image features were out of consideration of previous work because they relied on gradient-based methods for optimization. In Inverse Kinematics where we minimize a non-convex function, SDM achieved significantly better convergence than gradient-based approaches. In Face Alignment, SDM achieved state-of-the-arts results. Moreover, it was extremely computationally efficient, which makes it applicable for many mobile applications. In addition, we provided a unified view of several popular methods including SDM on sequential prediction, and reformulated them as a sequence of function compositions. Finally, we suggested some future research directions on SDM and sequential prediction.
]]>
Xuehan XiongImproved Formulations and Computational Strategies for the Solution and Nonconvex Generalized Disjunctive Programs
http://repository.cmu.edu/dissertations/651
http://repository.cmu.edu/dissertations/651Sat, 11 Jun 2016 05:49:23 PDT
Many optimization problems require the modelling of discrete and continuous variables, giving rise to mixed-integer linear and mixed-integer nonlinear programming (MILP / MINLP). An alternative representation of MINLP is Generalized Disjunctive Programming (GDP)1. GDP models are represented through continuous and Boolean variables, and involve algebraic equations, disjunctions, and logic propositions. This higher level representation facilitates the modelling process while keeping the logic structure of the problem. GDP models are typically reformulated as MINLP problems to exploit the developments in these solvers. The two traditional GDP-to-MINLP reformulations are the Big-M (BM) and Hull-reformulation (HR). Alternatively to direct MINLP reformulations, special techniques can help to improve the performance in solving GDP problems. There are two main contributions in this thesis. The first contribution involves the development of reformulations and methods that generate improved MINLP models form GDP problems. This development is achieved by exploiting the logic-nature of GDP, as well as alternative GDP-to-MINLP reformulations, to obtain relatively small MINLP models with tight continuous relaxations. The second contribution of this thesis is the improvement of existing GDP solution methods by the use of novel concepts. In particular, we improve the linear disjunctive branch and bound through the use of a Lagrangean relaxation of the HR. Also, we extend the logic-based outer-approximation to nonconvex problems, and develop a novel method to obtain cutting planes that improves the linear relaxation of the nonconvex problem. In the thesis, we first present a new Big-M reformulation of GDPs. Unlike the traditional iii Big-M reformulation that uses one M-parameter for each constraint, the new approach uses multiple M-parameters for each constraint. The multiple-parameter Big-M (MBM) reformulation is at least as tight as the traditional BM. Furthermore, it does not require additional variables or constraints. We present the new MBM and analyze the strength in its continuous relaxation compared to that of the traditional Big-M. We then present two algorithmic approaches to improve mixed-integer models that are originally formulated as convex GDPs. The algorithms seek to obtain an improved continuous relaxation of the MINLP reformulation of the GDP, while limiting the growth in the problem size. Both algorithms make use of the logic operation called basic step. This operation allows the derivation of formulations with continuous relaxations that are stronger than the direct BM and HR reformulations. The two algorithms differ in the method to exploit the advantages of the small problem size of the BM, and the tight continuous relaxation of the HR after the application of basic steps. The first algorithm uses a hybrid reformulation of GDP that seeks to exploit both advantages of the BM and HR. The second algorithm uses the strong formulation to derive cuts for the BM, generating a stronger formulation with small growth in problem size. In terms of GDP solution methods, we first present an enhancement to the disjunctive branch and bound for linear GDPs. In particular, we present a Lagrangean relaxation of the HR. The proposed Lagrangean relaxation can be applied to any linear GDP, and it always assigns 0-1 values to the binary variables of the HR. Furthermore, this relaxation is much simpler to solve than the continuous relaxation of the HR. The Lagrangean relaxation can be used in different manners to improve GDP solution methods. In this thesis, we explore the use of the Lagrangean relaxation as a primal heuristic to find feasible solutions in a disjunctive branch and bound. We note that the proposed Lagrangean relaxation, and its use in the disjunctive branch and bound, can be extended to nonlinear convex problems. We then extend the logic-based outer-approximation to the global solution of non-convex GDPs. The general idea of the algorithm is to have a linear master GDP that overestimates the feasible region of the GDP. This master problem provides a valid lower bound (in a minimization problem), and the selection of only one disjunctive term in each of the disjunctions. With the alternative provided by the master problem, an NLP subprobiv lem is solved to global optimality. This NLP subproblem is smaller and simpler than the continuous relaxation of the MINLP reformulation of the original GDP. After solving the subproblem, infeasibility or optimality integer cuts can be added to the master problem. This basic algorithm has the advantage of solving only small NLP problems to global optimality, instead of solving a larger MINLP to global optimality from the beginning. Furthermore, by using GDP as framework the NLP subproblem is smaller and simpler than an equivalent method directly applied to the MINLP reformulation. In order to further improve the performance of this logic-based outer approximation, two main features were implemented: derivation of additional cuts and partition of the algorithm in two stages. Finally, we apply a modified version of the global logic-based outer-approximation to the multiperiod blending problem. In addition to the proposed solution method, we present an improved problem formulation that makes use of redundant constraints. In order to generate such constraints, an alternative formulation was derived. The main idea in the new formulation is to track sources or commodities in the system, instead of tracking compositions. The main advantage is that it is possible to create redundant constraints in which the sum of individual source flows adds up to the total flow. Similarly, the sum of individual source inventories adds up to the total inventory. These redundant constraints considerably improve the relaxation of the model when linear approximations are used for the bilinear terms. Furthermore, the additional constraints can be included in the original model, strengthening its linear relaxation. This thesis makes several important contributions. From an aggregated perspective, our most significant contribution is the use of GDP and its logic structure to obtain improved models and develop solution methods. In this thesis we show that GDP is not only an intuitive and structured modeling framework, but it also opens a set of tools that are not accessible when modeling problems using mixed-integer programming. The tools we have developed can help to solve some problems in Process Systems Engineering (PSE). Furthermore, we hope that the advantages of formulating some problems using GDP become apparent. As the PSE community continues to increasingly use GDP as modeling framework, we hope it brings greater attention to the OR community.
]]>
Francisco TrespalaciosDefect Analysis and Microstructural Effects on the Surface Exchange Properties of La0.7Sr0.3MnO3(LSM) Epitaxial Thin Films
http://repository.cmu.edu/dissertations/650
http://repository.cmu.edu/dissertations/650Sat, 11 Jun 2016 05:20:26 PDT
La0.7Sr0.3MnO3 (LSM) is a perovskite oxide material that possesses many interesting electromagnetic and electrochemical properties, making it desirable as magnetic tunnel junction (MTJ) and solid oxide fuel cell (SOFC) electrodes.
]]>
Miaolei YanField Dislocation Mechanics with Applications in Atomic, Mesoscopic and Tectonic Scale Problems
http://repository.cmu.edu/dissertations/649
http://repository.cmu.edu/dissertations/649Sat, 11 Jun 2016 04:55:30 PDT
This thesis consists of two parts. The first part explores a 2-d edge dislocation model to demonstrate characteristics of Field Dislocation Mechanics (FDM) in modeling single and collective behavior of individual dislocations. The second work explores the possibility of modelling adiabatic shear bands propagation within the timespace averaged framework of Mesoscopic Field Dislocation Mechanics (MFDM). It is demonstrated that FDM reduces the study of a significant class of problems of discrete dislocation dynamics to questions of the modern theory of continuum plasticity. The explored questions include the existence of a Peierls stress in translationally-invariant media, dislocation annihilation, dislocation dissociation, finite-speed-of-propagation effects of elastic waves vis-a-vis dynamic dislocation fields, supersonic dislocation motion, and short-slip duration in rupture dynamics. A variety of dislocation pile-up problems are studied, primarily complementary to what can be dealt by existing classical pile-up models. In addition, the model suggests the possibility that the tip of a shear band can be modelled as a localized spatial gradient of elastic distortion with the dislocation density tensor in continuum dislocation mechanics; It is demonstrated that the localization can be moved by its theoretical driving force and forms a diffuse traveling band tip, thereby extending the thin layer of the deformation band. A 3-d, parallel finite element framework of MFDM is developed in a geometrically nonlinear context for the purpose of modelling shear bands. The numerical formulations and algorithm are presented in detail. Constitutive models appropriate for single crystal plasticity response and J2 plasticity with thermal softening are implemented.
]]>
Xiaohan ZhangFunctions and Kinetics of Mitochondrial Fusion and Fission in the Axon: a Quantitative Study
http://repository.cmu.edu/dissertations/648
http://repository.cmu.edu/dissertations/648Thu, 02 Jun 2016 09:24:24 PDT
In eukaryotic cells, mitochondria form a dynamic interconnected network to respond to changing needs at different subcellular locations. A fundamental yet unanswered question regarding this network is whether, and if so how, local fusion and fission of individual mitochondria affect their global distribution. To address this question, we developed high-resolution computational image analysis techniques to examine the relations between mitochondrial fusion/fission and spatial distribution within the axon of Drosophila larval neurons. We found that stationary and moving mitochondria underwent fusion and fission regularly but followed different spatial distribution patterns and exhibited different morphology. Disruption of inner membrane fusion by dOpa1 knockdown not only increased the spatial density of stationary and moving mitochondria but also changed their spatial distribution and morphology differentially. We found that changes to the spatial distribution of axonal mitochondria under dOpa1 knockdown could not be fully accounted for by changes to their motility but, instead, resulted from the disruption of inner membrane fusion. To understand the complex dynamic behavior of axonal mitochondria observed in our experimental studies quantitatively and at the mechanistic level, we built experimental data driven computational models. We found that the stationary mitochondria were composed of two morphologically different populations, which were generated by fusion/fission and long pause, respectively. Furthermore, computational modeling confirmed our experimental findings that motility and morphological dynamics of mitochondria synergistically regulated their spatial distribution in the axon. Together, our data revealed that stationary mitochondria within the axon interconnected with moving mitochondria through fusion and fission and that fusion between individual mitochondria mediated their global distribution.
]]>
Yiyi YuData-Driven Modeling of Morphological Dynamics and Intracellular Transport of Organelles
http://repository.cmu.edu/dissertations/647
http://repository.cmu.edu/dissertations/647Thu, 02 Jun 2016 09:00:10 PDT
Data-driven modeling is essential to understanding complex cellular processes. In this thesis, we present a series of studies of analyzing morphological dynamics and intracellular transport of organelles using techniques of mathematical modeling, image processing and machine learning. We first characterized the morphology of organelles, focusing specifically on mitochondria. We developed a morphological data processing pipeline. Using this pipeline, we discovered a bi-modal distribution of mitochondrial sizes, with a stable mean value in each mode. We then developed a data-driven model to investigate how fusion / fission of mitochondria modulates their sizes. For further analysis of morphology of mitochondria as well as other cellular components, we developed a general purpose machine learning algorithm, which we refer to as shape component analysis (SCA). We used it for dimension reduction and classification of mitochondrial morphology and protein geometry. In addition to studying the morphological dynamics of cellular organelles using data-driven modeling, we investigated the intracellular transport of organelles. We first proposed a probabilistic model for studying the relation between mitochondrial size and the velocity of their active transport. The proposed model not only explained the relation between mitochondrial size and velocity observed in experiments under normal conditions but also suggested a novel relation under changed conditions. Further analysis of the proposed model also suggested a way to evaluate the binding/unbinding rates of motors carrying the mitochondria. We further studied the global organization of organelle transport. We proposed an image processing framework to characterize the spatiotemporal dynamics of intracellular transport in terms of the spatial localization of stationary organelles and the spatiotemporal patterns of organelle movement, respectively. We used this framework to analyze time-lapse images of Lamp1 transport and found different global transport patterns. Overall, our studies produced both computational modeling methods and specific biological results for quantitative and systems-level understanding the complex behavior of intracellular organelles.
]]>
Hao-Chih LeeMechanosensing of Substrate Dimension and Migration State in Adherent Cells
http://repository.cmu.edu/dissertations/646
http://repository.cmu.edu/dissertations/646Thu, 02 Jun 2016 07:29:38 PDT
The behavior of adherent cells is known to be affected by both chemical signals and physical cues in the extracellular environment, including substrate topography and rigidity. The process of sensing physical features and converting them to intracellular signals is believed to rely on the formation of adhesion structures and the generation of actomyosin-based traction forces. Equally important is signaling in the reverse direction, as internal cellular activities regulate mechanical output to the extracellular environment. This thesis explores how substrate dimension and migration state are monitored by adherent cells and how they affect cell behavior.
]]>
Stephanie S. ChangSome Results on Classical Semantics and Polymorphic Types
http://repository.cmu.edu/dissertations/645
http://repository.cmu.edu/dissertations/645Thu, 02 Jun 2016 06:40:43 PDT
In the first chapter we consider the simply typed-calculus over one ground type with a discriminator which distinguishes terms, augmented additionally with an existential quantifier and a description operator, all of lowest type.
]]>
William J. Gunther