Search

Rapid advances in machine learning have made information analysis more efficient than ever before. However, to extract valuable information from trillion bytes of data for learning and decision-making, general-purpose computing systems or cloud infrastructures are often deployed to train a large-scale neural network, resulting in a colossal amount of resources in use while themselves exposing other significant security issues. Among potential approaches, the neuromorphic architecture, which is not only amenable to low-cost implementation, but can also deployed with in-memory computing strategy, has been recognized as important methods to accelerate machine intelligence applications. In this dissertation, theoretical and practical properties of a hybrid neural computing architecture are introduced, which utilizes a dynamic reservoir having the short-term memory to enable the historical learning capability with the potential to classify non-separable functions. The hybrid neural computing architecture integrates both spatial and temporal processing structures, sidestepping the limitations introduced by the vanishing gradient. To be specific, this is made possible through four critical features: (i) a feature extractor built based upon the in-memory computing strategy, (ii) a high-dimensional mapping with the Mackey-Glass neural activation, (iii) a delay-dynamic system with historical learning capability, and (iv) a unique learning mechanism by only updating readout weights. To support the integration of neuromorphic architecture and deep learning strategies, the first generation of delay-feedback reservoir network has been successfully fabricated in 2017, better yet, the spatial-temporal hybrid neural network with an improved delay-feedback reservoir network has been successfully fabricated in 2020. To demonstrate the effectiveness and performance across diverse machine intelligence applications, the introduced network structures are evaluated through (i) time series prediction, (ii) image classification, (iii) speech recognition, (iv) modulation symbol detection, (v) radio fingerprint identification, and (vi) clinical disease identification.

This thesis consists of two parts: (1) graft copolymer stabilized gold nanoparticles (AuNPs) and their biological application and (2) chemically cleavable -azido ether and its biological application. In the first part, poly(L-lysine)-graft-poly(ethylene glycol) (PLL-g-PEG) copolymers that bear multiple thiol groups on the polymer backbone are used for exceptional ligands to stabilize AuNPs. To characterize the effect of copolymer structure on AuNP stability, we synthesized PLL-g-PEGs with different backbone lengths, PEG grafting densities, and number of thiols per polymer chain. AuNPs were then combined with these polymer ligands, and the stabilities of the resulting AuNP@PLL-g-PEG particles against high temperature, oxidants, and competing thiol ligands were characterized using dynamic light scattering (DLS), visible absorption spectroscopy, and fluorescence spectrophotometry. Our observations indicate that thiolated PLL-g-PEG ligands (PLL-g-[PEG:SH]) combine thermodynamic stabilization via multiple Au-S bonds and steric stabilization by PEG grafts, and the best graft copolymer ligands balance these two effects. This new ligand system enables AuNPs to be used for solid phase polymerase chain reaction (SP-PCR) that requires harsh reaction conditions, such as, elevated temperature and competing thiol molecules. Azide functionalized PLL-g-[PEG:SH] were conjugated to oligodeoxy-nucleotide (ODN) primers via click chemistry and bound to AuNPs to yield AuNP-primers that successfully primed target DNA synthesis on the surface of the AuNPs through PCR, as demonstrated by gel electrophoresis, DLS, and fluorescent analysis. Moreover, the graft copolymer stabilized AuNPs were applied to rapid DNA diagnostics in a single PCR tube with magnetic particles through color change without any instrumental analysis. In the second part, bioorthogonal, chemically cleavable -azido ether has been studied and used to develop novel degradable materials. In order to understand the chemistry of the -azido ether, model molecules bearing the -azido ether were prepared. Hydrolytic stability of the model molecules was investigated by measuring their degradation rate using NMR, which leads to the relationship between the stability and chemical structures. Additionally, the cleavage kinetics of the model molecule, which was triggered by a couple of azide reducing reagents, was studied by NMR and UV-Vis absorption spectroscopy. The kinetic studies enable us to develop mechanistic investigation of the chemical cleavage as well as optimal cleavage conditions. Furthermore, the products after the chemical cleavage of the -azido ether were characterized using NMR. The novel -azido ether was then incorporated into degradable polyacrylamide gel electrophoresis (PAGE), in which biological macromolecules, including plasmid, microRNA, and proteins, were separated electrophoretically and recovered from the gel matrix with the optimal cleavage conditions. The kinetics of the recovery was quantitatively studied using UV-Vis absorption spectroscopy and fluorescence spectrophotometry. Furthermore, the recovered biological macromolecules were analyzed to investigate biocompatibility of our system. We anticipate further expansion of the -azido ether to a broad range of biological applications based on the fundamental studies and the representative example in PAGE.

We discuss a full-waveform based material profile reconstruction in two-dimensional heterogeneous semi-infinite domains. In particular, we try to image the spatial variation of shear moduli/wave velocities, directly in the time-domain, from scant surficial measurements of the domain's response to prescribed dynamic excitation. In addition, in one-dimensional media, we try to image the spatial variability of elastic and attenuation properties simultaneously. To deal with the semi-infinite extent of the physical domains, we introduce truncation boundaries, and adopt perfectly-matched-layers (PMLs) as the boundary wave absorbers. Within this framework we develop a new mixed displacement-stress (or stress memory) finite element formulation based on unsplit-field PMLs for transient scalar wave simulations in heterogeneous semi-infinite domains. We use, as is typically done, complex-coordinate stretching transformations in the frequency-domain, and recover the governing PDEs in the time-domain through the inverse Fourier transform. Upon spatial discretization, the resulting equations lead to a mixed semi-discrete form, where both displacements and stresses (or stress histories/memories) are treated as independent unknowns. We propose approximant pairs, which numerically, are shown to be stable. The resulting mixed finite element scheme is relatively simple and straightforward to implement, when compared against split-field PML techniques. It also bypasses the need for complicated time integration schemes that arise when recent displacement-based formulations are used. We report numerical results for 1D and 2D scalar wave propagation in semi-infinite domains truncated by PMLs. We also conduct parametric studies and report on the effect the various PML parameter choices have on the simulation error. To tackle the inversion, we adopt a PDE-constrained optimization approach, that formally leads to a classic KKT (Karush-Kuhn-Tucker) system comprising an initial-value state, a final-value adjoint, and a time-invariant control problem. We iteratively update the velocity profile by solving the KKT system via a reduced space approach. To narrow the feasibility space and alleviate the inherent solution multiplicity of the inverse problem, Tikhonov and Total Variation (TV) regularization schemes are used, endowed with a regularization factor continuation algorithm. We use a source frequency continuation scheme to make successive iterates remain within the basin of attraction of the global minimum. We also limit the total observation time to optimally account for the domain's heterogeneity during inversion iterations. We report on both one- and two-dimensional examples, including the Marmousi benchmark problem, that lead efficiently to the reconstruction of heterogeneous profiles involving both horizontal and inclined layers, as well as of inclusions within layered systems.

The purpose of this work is to investigate thin film CdTe as a possible large area detector system to be implemented for electronic portal imaging devices (EPID) in radiation therapy application. The material properties, average atomic number of 50, high electron density, remarkable radiation hardness and high band gap, already made them popular as small-area room-temperature crystalline x-ray detectors. The objective of our project is to define the range of the parameters applicable to a thin-film polycrystalline CdTe x-ray detector for mega-voltage energy beam. Due to small thickness of the film, we plan on using photovoltaic CdTe detector in combination with a metal plate converter in order to enhance high-energy photon conversion into secondary electrons, injected from the plate into CdTe layer and resulting in higher output signal. This approach increases efficiency of the device, while minimizing signal spreading due to reduced CdTe sensor thickness. In addition, the presence of the plate helps to reduce input signal-to-noise ratio by shielding the sensor layer from the unavoidable scatter from the patient. The optimal CdTe thickness was found based on both device modeling and experimental verifications. A set of geometrical configurations appropriate for EPID application, was determined along with an approximate compromise between the sensor and metal plate thickness, detector efficiency and spatial resolution, and electrical properties of the material.Our proposed detector system operates on the same basic principle as a regular CdTe based solar cell. When irradiated with x- or gamma-rays, electron hole pairs are produced in response to energy deposition by electrons injected from the metal layer. The device will rely on semiconductor junction as the source of electric field for charge carrier separation, thus eliminating the need for external high-voltage biasing, typically used with x-ray solid state detectors. The generated photocurrent or the developed photo-voltage between the contacts will be used to measure the intensity of incident radiation. Despite the similarities to solar cells, there is a number of important differences discussed in the dissertation that have to be taken into account in the detector design.In this work we 1) present theoretical optimization of novel thin-film based CdTe x-ray detector with a metal plate converter by employing Monte Carlo software package MCNP5, 2) estimate the detector output for the optimized configuration with a thin film device modeling tool SCAPS, and 3) verify the device performance with experiments with a 6MeV clinical x-ray beam. Based on our successful theoretical modeling of an optimized detector and measured device responsivity to radiation, we believe the thin-film CdTe-based detector is well suited for imaging with high energy x-rays used in clinical radiation therapy.

The imperfect trench installation method is used to reduce earth pressure on deeply buried conduits. Few quantitative refinements to the imperfect trench installation method, however, have been added since the fundamental mechanics of the reverse arching action was proposed by Marston and Spangler. There have been limited research results published regarding primarily qualitative aspects of earth load reduction for imperfect trench conditions. It was found during the course of this study that significant frictional forces develop along the sidewalls of buried conduits and adjacent sidefills in imperfect trench installations. Current American Association of State Highway and Transportation Officials provisions do not consider these frictional forces, but they cannot be neglected in imperfect trench installations as their effect is dominant. The objectives of this study were to study the soil-structure interaction for deeply buried roadway conduits (concrete pipes, corrugated polyvinyl chloride (PVC) pipes, corrugated steel pipes, and box culverts) and the efficiency of the imperfect trench method for their installations. This research identifies variables that significantly affect earth loads, as well as the effects of bedding and sidefill treatments. The soil-structure interaction was computed using the finite element method with soil response simulated with the Duncan soil model and Selig soil parameters. The geometry of the soft material zone that induces the reverse arching action was optimized to maximize the earth load reduction for imperfect trench installations. The optimization process was based on parametric studies of the geometry and location of the soft material zone, combined with bedding and sidefill treatments. Predictor equations for earth load, maximum wall stress, and deflection of the conduits were formulated that incorporate the proposed optimum soft material zone geometry and installation techniques. Parametric studies revealed that the optimum geometry of the soft material zone in the proposed imperfect trench installations could reduce the maximum wall stress or vertical earth load by 69-85%.

Fuel economy is one of the major objectives of modern engine control. In addition, three major pollutants from the vehicle (CO , HC and NOx) have been stringently regulated both in the U.S. and world wide, and these emissions mandates as well as fuel economy have engendered a great deal of research on engine technology. The recently developed engines are fuel-injected and host a wide-array of new aftertreatment devices, sensors, actuators, and recent innovative engine operations such as lean burn and gasoline direct injection. Part of this dissertation exploits the potential benefits of air flow actuation in a SI engine. For this purpose, a mostly classical, and modern nonlinear controllers are designed, and significant engine performance measures such as average manifold pressure, emissions, and drivability are evaluated. In particular, the nonlinear controller has two novel features. Firstly, a recovery scheme for integrator wind-up due to input constraints is directly integrated into the control design. The second novel feature is that the positive semi-definite control Lyapunov function methodology is applied to a discrete-time model. This dissertation also investigates the benefits of a gasoline direct injection engine with novel aftertreatment. The lean burn and gasoline direct injection engine can consume a smaller amount of fuel in part load. However, the major drawback of these engines is that the tailpipe NOx emissions level is not acceptable. A promising solution that has emerged in the area is the implementation of the lean NOx trap for aftertreatment. The lean NOx trap is a catalytic device capable of capturing NOx in fuel-lean exhaust gas, which must be periodically purged by fuel-rich exhaust gas. Thus, the purging strategy, which includes the purging duration, purging frequency, and the amount of rich fuel during purging, strongly affects the fuel economy. In this dissertation, the optimal purging strategy to achieve fuel economy and low NOx emissions is studied.