Global energy consumption continues to grow, driving the need for cheap, power-dense power electronics. Replacing the incumbent silicon insulated gate bipolar transistors with silicon carbide (SiC) metal oxide semiconductor field effect transistors (MOSFETs) has been proposed as a solution to increase the power densities of power converters in some applications. One such application is electric vehicles (EVs) where the efficiency and weight of the power electronics are critical; however, modern packaging technologies are still limiting the performance of SiC MOSFETs. One promising trend in power semiconductor packaging technologies is the use of printed circuit boards (PCBs) because the technology is mature—resulting in low costs—and the allowable stackups are ideal for integrating driving circuitry and power loop components—resulting in reduced manufacturing complexity. This thesis presents the design and analysis of two PCB-embedded 1.2 kV SiC MOSFET half-bridge packages and a hybrid PCB/DBC-based 1.2 kV SiC MOSFET full-bridge package for EV on-board charger applications. The first of the two PCB-embedded packages has integrated gate drive circuitry, less than 2.3 nH loop inductances, and dual-sided cooling with a total junction-to-case thermal resistance (RTH,JC) of 0.12 K/W. The second PCB-embedded package has only drain-side cooling to allow for surface mount terminals, has an area of 37.1 mm x 18.5 mm due to the removal of the gate drive circuitry, and has less than 2.4 nH loop inductances. The PCB/DBC-based full-bridge package has an RTH,JC of 0.65 K/W, less than 4.5 nH, and integrated gate drive circuitry. / M.S. / The continued increase in global energy consumption has led to concerns about sustainability, and as renewable energy generation is adopted more broadly, more efficient means of converting electrical energy from one form to another are required. Some applications, such as electric vehicles (EVs), also require a lightweight and a low volume from their converters in addition to high efficiency. The packaging of the semiconductors used in converters is important to the overall electrical efficiency of the converter and can also have an impact on the size of the converter as well. This thesis explores the design and analysis of three package structures for the semiconductors used in the on-board charger of an EV. These package structures are unified under the common theme of using printed circuit boards (PCBs) in the package itself. PCBs are commonly used to route the electrical connections between packaged semiconductors and other components in the converter, but they are not usually integrated into the package itself. The hope is that by integrating the PCB into the semiconductor package, higher-efficiency, lighter-weight, and smaller-volume converters will be possible.

In today’s digital era, establishing an online presence and maintaining a well-structured website is vitalfor companies to remain competitive in their respective markets. A crucial aspect of online success liesin strategically selecting the right words to optimize customer engagement and search engine visibility.However, this process is often time-consuming, involving extensive analysis of a company’s website aswell as its competitors’. This thesis focuses on developing an efficient binary classification approachto identify key words and phrases extracted from multiple company websites. The data set used forthis solution consists of approximately 92,000 scraped samples, primarily comprising non-key samples.Various features were extracted, and a word embedding model was employed to assess each sample’srelevance to its specific industry and topic. The logistic regression, decision tree and random forestalgorithms were all explored and implemented as different solutions to the classification problem. Theresults indicated that the logistic regression model excelled in retaining keywords but was less effectivein eliminating non-keywords. Conversely, the tree-based methods demonstrated superior classificationof keywords, albeit at the cost of misclassifying a few keywords. Overall, the random forest approachoutperformed the others, achieving a result of 76 percent in recall and 20 percent in precision whenpredicting key samples. In summary, this thesis presents a solution for classifying words and phrasesfrom company websites into key and non-key categories, and the developed methodology could offervaluable insights for companies seeking to enhance their website optimization strategies.

<p>Latent variable models are widely used in modern data science for both statistic and dynamic data. This thesis focuses on large-scale latent variable models formulated for time series data and static network data. The former refers to the state space model for dynamic systems, which models the evolution of latent state variables and the relationship between the latent state variables and observations. The latter refers to a network decoder model, which map a large network into a low-dimensional space of latent embedding vectors. Both problems can be solved by adaptive stochastic gradient Markov chain Monte Carlo (MCMC), which allows us to simulate the latent variables and estimate the model parameters in a simultaneous manner and thus facilitates the down-stream statistical inference from the data. </p> <p><br></p> <p>For the state space model, its challenge is on inference for high-dimensional, large scale and long series data. The existing algorithms, such as particle filter or sequential importance sampler, do not scale well to the dimension of the system and the sample size of the dataset, and often suffers from the sample degeneracy issue for long series data. To address the issue, the thesis proposes the stochastic approximation Langevinized ensemble Kalman filter (SA-LEnKF) for jointly estimating the states and unknown parameters of the dynamic system, where the parameters are estimated on the fly based on the state variables simulated by the LEnKF under the framework of stochastic approximation MCMC. Under mild conditions, we prove its consistency in parameter estimation and ergodicity in state variable simulations. The proposed algorithm can be used in uncertainty quantification for long series, large scale, and high-dimensional dynamic systems. Numerical results on simulated datasets and large real-world datasets indicate its superiority over the existing algorithms, and its great potential in statistical analysis of complex dynamic systems encountered in modern data science. </p> <p><br></p> <p>For the network embedding problem, an appropriate embedding dimension is hard to determine under the theoretical framework of the existing methods, where the embedding dimension is often considered as a tunable hyperparameter or a choice of common practice. The thesis proposes a novel network embedding method with a built-in mechanism for embedding dimension selection. The basic idea is to treat the embedding vectors as the latent inputs for a deep neural network (DNN) model. Then by an adaptive stochastic gradient MCMC algorithm, we can simulate of the embedding vectors and estimate the parameters of the DNN model in a simultaneous manner. By the theory of sparse deep learning, the embedding dimension can be determined via imposing an appropriate sparsity penalty on the DNN model. Experiments on real-world networks show that our method can perform dimension selection in network embedding and meanwhile preserve network structures. </p> <p><br></p>

No description available.

Many studies show that manual code generation is error-prone and results in vulnerabilities. Vulnerability fixing has been shown as the most time-consuming process among multiple steps of code repair. To help developers repair these security vulnerabilities, my dissertation aims to develop an automatic or semi-automatic secure code generation system with neural network based approaches. Trained with huge amounts of good-quality code, I expect the neural network to learn the secure usage and produce the correct code suggestions. Despite the great success of neural networks, the vision of comprehending and generating programming languages through neural networks has not been fully realized. There are many fundamental questions that need to be answered. These questions include 1) what are the accuracy impacts of the various choices in code embedding? 2) How to address the accuracy challenges caused by the programming language specific properties in the task of secure code suggestion? My dissertation work answers the two questions with a systematical measurement study and specialized neural network designs. My experiments show that program analysis is a necessary preprocessing step to guide the code embedding – resulting in a 36.1% accuracy improvement. Furthermore, I identify two previously unreported deficiencies in the cryptographic API suggestion task. To close the gap, I invent a highly accurate API method suggestion solution, referred to as Multi-HyLSTM, with specialized neural network designs to recognize unique programming language characteristics. My work points out the important differences between natural languages and programming languages, which pure data-driven learning approaches may not recognize. / Doctor of Philosophy / Neural network techniques that automatically learn rules from data show great potential to provide vulnerability-agnostic solutions for securing code. Recent research community has witnessed the rapid progress of neural network techniques in various application domains, such as computer vision, natural language processing, etc. However, how to harness the success of neural network based approaches for dealing with programs is still largely unknown. Many fundamental questions are required to be answered. This dissertation aims to provide neural network based solutions to help developers write secure code, as well as answer several important but unknown research questions about promoting neural network based approaches specialized for the programming language domain. Learning from Java cryptographic code, I explore the accuracy challenges for neural networks to understand the secure API usage rules and generate appropriate suggestions based on them. One of my research focuses is on how to express code in a way that neural networks can comprehend, aka code embedding. Code embedding is the process of transforming code into numeric vectors. It is important for accuracy as all the subsequent neural network calculation is performed on it. I conduct a systematic comparison to evaluate several key embedding design choices and reveal their impacts on accuracy improvements. To further improve the accuracy, I focus on the accuracy challenges in the specific task, generating API suggestions by neural networks. I identify the unreported program dependency specific challenges and present several specialized neural network designs to address them.

Traditional generators have been decommissioned or replaced by renewable energy generation due to utility long-standing goals. However, instead of flattening the entire plant, the rotating mass of generator can be utilized as a storage unit (inertia resource) to mitigate the frequency swings during transient caused by the renewables. The goal of this work is to design a control strategy utilizing the decommissioned generator interfaced with power grid via a back-to-back converter to provide inertia support. This is referred to as decoupled synchronous machine system (DSMS). On top of that, the grid-side converter is capable of providing reactive power as an auxiliary voltage controller. However, in a practical setting, for power utilities, the detailed state equations of such device as well as the complicated nonlinear power system are usually unobtainable making the controller design a challenging problem. Therefore, a model free, purely data-driven strategy for the nonlinear controller design using Koopman operator-based framework is proposed. Besides, the time delay embedding technique is adopted together with Koopman operator theory for the nonlinear system identification. Koopman operator provides a linear representation of the system and thereby the classical linear control algorithms can be applied. In this work, model predictive control is adopted to cope with the constraints of the control signals. The effectiveness and robustness of the proposed system are demonstrated in Kundur two-area system and IEEE 39-bus system. / Doctor of Philosophy / Power system is facing an energy transformation from the traditional fuel to sustainable renewable such as solar, wind and so on. Unlike the traditional fuel energized generators, the renewable has very little inertia to maintain frequency stability. Therefore, this work proposes a new system referred to as decoupled synchronous machine system (DSMS) to support the grid frequency. DSMS consists of the rotating mass of generator and a back-to-back converter which can be utilized as an inertia resource to mitigate the frequency oscillations. In addition, the grid-side converter can provide reactive power to improve voltage performance during faults. This work aims to design a control strategy utilizing DSMS to support grid frequency and voltage. However, an explicit mathematical model of such device is unobtainable in a practical setting making data-driven control the only option. A data-driven technique which is Koopman operator-based framework together with time delay embedding algorithm is proposed to obtain a linear representation of the system. The effectiveness and robustness of the proposed system are demonstrated in Kundur two-area system and IEEE 39-bus system.

The layer-by-layer nature of additive manufacturing (AM) allows for access to the entire build volume of a component during manufacture including the internal structure. Voids are accessible during the build process and allow for components to be embedded and sealed with subsequently printed layers. This process, in conjunction with direct write (DW) of conductive materials, enables the direct manufacture of parts featuring embedded electronics, including interconnects and sensors. The scope of previous works in which DW and AM processes are combined has been limited to single material AM processes. The PolyJet process is assessed for hybridization with DW because of its multi-material capabilities. The PolyJet process is capable of simultaneously depositing different materials, including rigid and elastomeric photopolymers, which enables the design of flexible features such as membranes and joints. In this work, extrusion-based DW is integrated with PolyJet AM technology to explore opportunities for embedding conductive materials on rigid and elastomeric polymer substrates. Experiments are conducted to broaden the understanding of how silver-loaded conductive inks behave on PolyJet material surfaces. Traces of DuPont 5021 conductive ink as small as 750?m wide and 28?m tall are deposited on VeroWhite+ and TangoBlack+ PolyJet material using a Nordson EFD high-precision fluid dispenser. Heated drying at 55°C is found to accelerate material drying with no significant effect on the conductor's geometry or conductivity. Contact angles of the conductive ink on PolyJet substrates are measured and exhibit a hydrophilic interaction, indicating good adhesion. Encapsulation is found to negatively impact conductivity of directly written conductors when compared to traces deposited on the surface. Strain sensing components are designed to demonstrate potential and future applications. / Master of Science

A challenging goal for both the game industry and the research community of computer graphics is the generation of 3D virtual avatars that automatically perform realistic human motions with high speed at low monetary cost. So far, full body motion estimation of human complexity remains an important open problem. We propose a realistic motion estimation framework to control the animation of 3D avatars. Instead of relying on a motion capture device as the control signal, we use low-cost and ubiquitously available 3D accelerometer sensors. The framework is developed in a data-driven fashion, which includes two phases: model learning from an existing high quality motion database, and motion synthesis from the control signal. In the phase of model learning, we built a high quality motion model of less complexity that learned from a large motion capture database. Then, by taking the 3D accelerometer sensor signal as input, we were able to synthesize high-quality motion from the motion model we learned. In this thesis, we present two different techniques for model learning and motion synthesis, respectively. Linear and nonlinear reduction techniques for data dimensionality are applied to search for the proper low dimensional representation of motion data. Two motion synthesis methods, interpolation and optimization, are compared using the 3D acceleration signals with high noise. We evaluate the result visually compared to the real video and quantitatively compared to the ground truth motion. The system performs well, which makes it available to a wide range of interactive applications, such as character control in 3D virtual environments and occupational training. / Master of Science

Probability-one homotopy algorithms have strong convergence characteristics under mild assumptions. Such algorithms for mixed complementarity problems (MCPs) have potentially wide impact because MCPs are pervasive in science and engineering. A probability-one homotopy algorithm for MCPs was developed earlier by Billups and Watson based on the default homotopy mapping. This algorithm had guaranteed global convergence under some mild conditions, and was able to solve most of the MCPs from the MCPLIB test library. This thesis extends that work by presenting some other homotopy mappings, enabling the solution of all the remaining problems from MCPLIB. The homotopy maps employed are the Newton homotopy and homotopy parameter embeddings. / Master of Science

Many concrete problems, ranging from Portfolio selection to Water resource management, may be cast into a multiobjective programming framework. The simplistic way of superseding blindly conflictual goals by one objective function let no chance to the model but to churn out meaningless outcomes. Hence interest of discussing ways for tackling Multiobjective Programming Problems. More than this, in many real-life situations, uncertainty and imprecision are in the state of affairs. In this dissertation we discuss ways for solving Multiobjective Programming Problems with fixed and fuzzy coefficients. No preference, a priori, a posteriori, interactive and metaheuristic methods are discussed for the deterministic case. As far as the fuzzy case is concerned, two approaches based respectively on possibility measures and on Embedding Theorem for fuzzy numbers are described. A case study is also carried out for the sake of illustration. We end up with some concluding remarks along with lines for further development, in this field. / Operations Research / M. Sc. (Operations Research)