The challenges imposed on Voltage Regulator Modules (VRM) become difficult to be achieved with the conventional multiphase buck converter commonly used on PC motherboards. For faster data transfer, a decrease in the output voltage is needed. This decrease causes small duty cycle that is accompanied by critical problems which impairs the efficiency. Therefore, these problems need to be addressed. Transformer-based non-isolated topologies are not new approaches to extend the duty cycle and avoid the associated drawbacks. High leakage, several added components and complicated driving and control schemes are some of the trade-offs to expand the duty cycle. The objective of this work is to present a new dc-dc buck-based topology, which extends the duty cycle with minimum drawbacks by adding two transformers that can be integrated to decrease the size and two switches with zero voltage switching (ZVS). Another issue addressed in this thesis is deriving a small signal model for a two-input two-phase buck converter as an introduction to a new evolving field of multi-input converters.

With the increasing demand for small and cost efficient DC/DC converters, the power converters are expected to operate with high efficiency. Magnetics components design is one of the biggest challenges in achieving the higher power density and higher efficiency due to the significant portion of magnetics components volume in the whole power system. At the same time, most of the experimental phenomena are related to the magnetics components. So, good magnetics components design is one of the key issues to implement low voltage high current DC/DC converter. Planar technology has many advantages. It has low profile construction, low leakage inductance and inter-winding capacitance, excellent repeatability of parasitic properties, cost efficiency, great reliability, and excellent thermal characteristics. On the other side, however, planar technology also has some disadvantages. Although it improves thermal performance, the planar format increases footprint area. The fact that windings can be placed closer in planar technology to reduce leakage inductance also often has an unwanted effect of increasing parasitic capacitances. In this dissertation, the planar magnetics designs for high current low voltage applications are thoroughly investigated and one CAD design methodology based on FEA numerical analysis is proposed. Because the frequency dependant parasitic parameters of magnetics components are included in the circuit model, the whole circuit analysis is more accurate. When it is implemented correctly, integrated magnetics technique can produce a significant reduction in the magnetic core content number and it can also result in cost efficient designs with less weight and smaller volume. These will increase the whole converter's power density and power efficiency. For high output current and low output voltage applications, half bridge in primary and current doublers in secondary are proved to be a very good solution. Based on this topology, four different integrated magnetics structures are analyzed and compared with each other. One unified model is introduced and implemented in the circuit analysis. A new integrated magnetics component core shape is proposed. All simulation and experimental results verify the integrated magnetics design. There are several new magnetics components applications shown in the dissertation. Active transient voltage compensator is a good solution to the challenging high slew rate load current transient requirement of VRM. The transformer works as an extra voltage source. During the transient periods, the transformer injects or absorbs the extra transient to or from the circuit. A peak current mode controlled integrated magnetics structure is proposed in the dissertation. Two transformers and two inductors are integrated in one core. It can force the two input capacitors of half bridge topology to have the same voltage potential and solve the voltage unbalance issue. The proposed integrated magnetics structure is simple compared with other methods implementing the current mode control to half bridge topology. Circuit analysis, simulation and experimental results verify the feasibility of these applications.

A novel concept of on-chip bondwire inductors and transformers with ferrite epoxy glob coating is proposed, offering a cost effective approach to realize power systems on chip (PSoC) or System-in-Package (PSiP). The concept has been investigated both experimentally and with finite element modeling. Improvement in total inductance is demonstrated for multi-turn bondwire inductors over single bondwire inductors. The inductance and Q factor can be further boosted with coupled multi-turn inductor concept. Transformer parameters including self- and mutual inductance, and coupling factors are extracted from both modeled and measured S-parameters. More importantly, the bondwire magnetic components can be easily integrated into SoC manufacturing processes with minimal changes to the layout, and open enormous possibilities for realizing cost-effective, high current, high efficiency PSoC's or PSiP's. The design guidelines for single bondwire inductors as well as multi-turn inductors are discussed step by step in several chapters. Not only is the innovated concept for bondwire inductor with ferrite ink presented, but also the practical implementation and design rules are given. With all the well defined steps, people who want to use these bondwire inductors with ferrite ink in their PSoC research or products will find it as simple as using commercial inductors. Last but not least, the PSoC concept using a bondwire inductor is demonstrated by building the prototype of dc-dc buck converter IC as well as the whole package. IC and the whole function block are tested and presented in this work.

The goal of this thesis is to design and fabricate a DC-to-DC converter for use in high-voltage capacitor charging applications. The primary objectives include increasing the efficiency and reducing the cost of traditional methods used for this application. Traditional methods were not designed specifically for high-voltage capacitor charging and were thus very primitive and exhibited lower efficiency. Prior methods made use of a high voltage power supply and a current limiting resistor or control scheme. The power supply would often only operate efficiently at a single voltage value and would thus function poorly over a range used in charging a capacitor. The resistor would also dissipate a fair amount of power, also limiting efficiency. This design makes use of a traditional flyback topology utilizing a controller developed specifically for this application, centering the design approach on the LT3750. Hence, taking full advantage of the efficiency improving control scheme it provides. Additionally, through the use of advanced techniques to eliminate noise and power losses, the efficiency may be significantly improved. A detailed theoretical analysis of the charger is also presented. The analysis will then be applied to optimization techniques to select ideal component values to meet specific design specifications. In this research, a specifically designed and developed prototype will be used to experimentally verify the theoretical work and optimization techniques.

Convolutional neural networks (CNNs) have dominated the computer vision field since the early 2010s, when deep learning largely replaced previous approaches like hand-crafted feature engineering and hierarchical image parsing. Meanwhile transformer architectures have attained preeminence in natural language processing, and have even begun to supplant CNNs as the state of the art for some computer vision tasks. This study proposes a novel transformer-based architecture, the attentional parsing network, that reconciles the deep learning and hierarchical image parsing approaches to computer vision. We recast unsupervised image representation as a sequence-to-sequence translation problem where image patches are mapped to successive layers of latent variables; and we enforce symmetry and sparsity constraints to encourage these mappings take the form of a parse tree. We measure the quality of learned representations by passing them to a classifier and find high accuracy (> 90%) for even small models. We also demonstrate controllable image generation: first by “back translating” from latent variables to pixels, and then by selecting subsets of those variables with attention masks. Finally we discuss our design choices and compare them with alternatives, suggesting best practices and possible areas of improvement.

High voltage power supply design presents unique requirements, combining safety, controllability, high performance, and high efficiencies. A new Resonant Self-Tracking Current-Fed Converter (RST-CFC) is investigated as a proof-of-concept of a high voltage power supply particularly for an X-ray system. These systems require fast voltage rise times and low ripple to yield a clear image. The proposed converter implements high-frequency resonance among discrete components and transformer parasitics to achieve high voltage gain, and the self-tracking nature ensures operation at maximum gain while power switches achieve zero-voltage switching across the full load range. This converter exhibits an inherent indefinite short-circuit capability. Theoretical results were obtained through simulations and verified by experimental results through a complete test configuration. Converter topology viability was confirmed through hardware testing and characterization.

Energy distribution in a conventional magnetic component is generally not at a designer's disposal. In a conventional toroidal inductor, the energy density is inversely proportional to the square of the radius. Thus, a designer would be unable to prescribe uniform field distribution to fully utilize the inductor volume for storing magnetic energy. To address this problem a new inductor design, called a "constant-flux" inductor, is introduced in this thesis. This new inductor has the core and windings configured to distribute the magnetic flux and energy relatively uniformly throughout the core volume to achieve power density higher than that of a conventional toroidal inductor. The core of this new inductor design is made of concentric cells of magnetic material, and the windings are wound in the gaps between the cells. This structure is designed to avoid crowding of the flux, thus ensuring lower core energy losses. In addition, the windings are patterned for shorter length and larger area of cross-section to facilitate lower winding energy losses. Based on this approach, a set of new, constant flux inductor/transformer designs has been developed. This design set is based on specific input parameters are presented in this thesis. These parameters include the required inductance, peak and rms current, frequency of operation, permissible dc resistance, material properties of the core such as relative permeability, maximum permissible magnetic flux density for the allowed core loss, and Steinmetz parameters to compute the core loss. For each constant flux inductor/transformer design, the winding loss and core loss of the magnetic components are computed. In addition, the quality factor is used as the deciding criterion for selection of an optimized inductor/transformer design. The first design presented in this thesis shows that for the same maximum magnetic field intensity, height, total stored energy, and material, the footprint area of the new five-cell constant-flux inductor is 1.65 times less than that of an equivalent conventional toroidal inductor. The winding loss for the new inductor is at least 10% smaller, and core loss is at least 1% smaller than that in conventional inductors. For higher energy densities and taller inductors, an optimal field ratio of the dimensions of each cell (α = Rimin/Rimax) and a larger number of cells is desired. However, there is a practical difficulty in realizing this structure with a larger number of cells and higher field ratio α. To address this problem, an inductor design is presented that has a footprint area of a three-cell constant-flux inductor (α = 0.6) that is 1.48 times smaller in comparison to an equivalent conventional toroidal inductor. For the same maximum magnetic flux density, height, material, and winding loss, the energy stored in this new three-cell constant-flux inductor (α = 0.6) is four times larger than that of an equivalent conventional toroidal inductor. Finally, new designs for application-specific toroidal inductors are presented in this thesis. First, a constant-flux inductor is designed for high-current, high-power applications. An equivalent constant-flux inductor to a commercially available inductor (E70340-010) was designed. The height of this equivalent inductor is 20% less than the commercial product with the same inductance and dc resistance. Second, a constant-flux inductor design of inductance 1.2 µH was fabricated using Micrometal-8 for the core and flat wire of 0.97 mm x 0.25 mm for the conductor. The core material of this inductor has relative permeability < 28 and maximum allowed flux density of 3600 Gauss. The dc resistance of this new, constant flux inductor was measured to be 14.4 mΩ. / Master of Science

The power transformer is a critical equipment in which the protection process is essential for modern societies where continuous electric power supplies are required. Copper corrosion problems due to the formation of sulfur deposits on the copper windings of mineral oil-filled power transformers are considered a major issue that can lead to sudden failures, and in some cases, to costly fire and explosion accidents in the power plants. These kinds of problems are still being reported regardless of available condition monitoring (CM) parameters applied in power transformers' maintenance strategy. The currently applied CM parameters are based on three different types of technologies. The first is oil analysis focuses more on measurable variables such as measuring the concentration of the corrosive sulfur compounds in the insulating oil, evaluating the oil's capability to form sulfur deposits, and measuring an increase in the concentration of specific gases. The second is on-site electrical testing focuses on the variation of the transformer's electrical properties due to the sulfur deposits. The measurable variables used in the electrical testing are Frequency Domain Spectroscopy test and Polarization/Depolarization Current test. The last is online sensor technology using Corrosive Sulfur Sensor, where the sensor's outcome data provide information about the oil's capability to form sulfur deposits. The research problem addressed is how to establish more effective CM parameters for early detection of copper corrosion problems.  The research problem is divided into three concretized research problems: What are the strengths and weaknesses of the currently applied condition monitoring parameters? Which measurable variables could be utilized to improve the currently applied condition monitoring parameters to be more effective for early detection of copper corrosion problems? And how to establish a procedure for the condition monitoring for detecting copper corrosion? Two research methodologies were applied to answer these questions, literature review and experimental work. The literature review showed significant gaps in the currently applied CM parameters for early detection of copper corrosion problems due to incomplete data of the corrosion reaction mechanism. Therefore, qualitative and quantitative investigations in the experimental work were carried out. The most important result was finding new relevant measurable variables, i.e. hydrogen sulfide gas and toluene compound, which are by-products of corrosion reaction. These measurable variables are utilized to establish more effective CM parameters for early detection of copper corrosion problems. The main conclusion of this thesis is the importance of detection corrosion problems in the initial stage by implementing more effective CM parameters to prevent catastrophic and costly failures, reduce the negative impacts on human life and the environment, and save the economic losses. Another conclusion is the importance of regularly following the measurable variables' uptrend during transformer useful life to avoid incorrect evaluation of corrosion conditions.

Yes / The single-stage flyback Photovoltaic (PV) micro-inverter is considered as a simple and small in size topology but requires expensive digital microcontrollers such as Field-Programmable Gate Array (FPGA) or Digital Signal Processor (DSP) to increase the system efficiency, this would increase the cost of the overall system. To solve this problem, based on a single-stage flyback structure, this paper proposed a low cost and simple analog-digital control scheme. This control scheme is implemented using a low cost ATMega microcontroller built in the Arduino Uno board and some analog operational amplifiers. First, the single-stage flyback topology is analyzed theoretically and then the design consideration is obtained. Second, a 120 W prototype was developed in the laboratory to validate the proposed control. To prove the effectiveness of this control, we compared the cost price, overall system efficiency, and THD values of the proposed results with the results obtained by the literature. So, a low system component, single power stage, cheap control scheme, and decent efficiency are achieved by the proposed system. Finally, the experimental results present that the proposed system has a maximum efficiency of 91%, with good values of the total harmonic distortion (THD) compared to the results of other authors / This work was supported in-part by Innovate UK GCRF Energy Catalyst PiCREST project under Grant number 41358, in-part by British Academy GCRF COMPENSE project under Grant GCRFNGR3\1541

Transformer neural networks have made a significant impact on natural language processing. The Transformer network self-attention mechanism effectively addresses the vanishing gradient problem that limits a network learning capability, especially when the time series gets longer or the size of the network gets deeper. This dissertation examines the usage of the Transformer model for time-series forecasting and customizes it for a simultaneous multistep-ahead prediction model in a surrogate model predictive control (MPC) application. The proposed method demonstrates enhanced control performance and computation efficiency compared to the Long-short term memory (LSTM)-based MPC and one-step-ahead prediction model structures for both LSTM and Transformer networks. In addition to the Transformer, this research investigates hybrid machine-learning modeling. The machine learning models are known for superior function approximation capability with sufficient data. However, the quantity and quality of data to ensure the prediction precision are usually not readily available. The physics-informed neural network (PINN) is a type of hybrid modeling method using dynamic physics-based equations in training a standard machine learning model as a form of multi-objective optimization. The PINN approach with the state-of-the-art time-series neural networks Transformer is studied in this research providing the standard procedure to develop the Physics-Informed Transformer (PIT) and validating with various case studies. This research also investigates the benefit of nonlinear model-based control and estimation algorithms for managed pressure drilling (MPD). This work presents a new real-time high-fidelity flow model (RT-HFM) for bottom-hole pressure (BHP) regulation in MPD operations. Lastly, this paper presents details of an Arduino microcontroller temperature control lab as a benchmark for modeling and control methods. Standard benchmarks are essential for comparing competing models and control methods, especially when a new method is proposed. A physical benchmark considers real process characteristics such as the requirement to meet a cycle time, discrete sampling intervals, communication overhead with the process, and model mismatch. Novel contributions of this work are (1) a new MPC system built upon a Transformer time-series architecture, (2) a training method for time-series machine learning models that enables multistep-ahead prediction, (3) verification of Transformer MPC solution time performance improvement (15 times) over LSTM networks, (4) physics-informed machine learning to improve extrapolation potential, and (5) two case studies that demonstrate hybrid modeling and benchmark performance criteria.