Search

The present Thesis investigates the quasi-static and cyclic-fatigue properties of nano-modified epoxy polymers. A diglycidyl ether of bisphenol-A was used as the base epoxy and polymer-based materials, block copolymers and core-shell rubber particles were used as the toughening agents. The morphologies formed by the tougheners were determined and the associated fracture and fatigue properties, including the toughening mechanisms, were investigated. The modified epoxy polymers with some of the best properties are described below. The unmodified epoxy polymer had a fracture energy, GIc, of 494 J/m2 and this was increased to 2380 J/m2, when 10 wt.% of a polyurethane toughener was employed. This modified epoxy polymer possessed a second-phase morphology, which was a nano-sized 'worm-like' structure. This nano-phase debonded from the epoxy, which enabled plastic void growth to occur in the epoxy polymer. This mechanism also led to an increase in the cyclic-fatigue threshold fracture energy, Gth, from 153 J/m2 of the unmodified epoxy to 444 J/m2. Nano-sized core-shell rubber particles (either 100 nm or 200 nm nominal diameters) were employed and they formed a well dispersed rubbery particulate phase. A maximum fracture energy of 2540 J/m2 was observed and the particles were found to undergo internal cavitation that induced plastic void growth. However, only the relatively large particles were able to cavitate under the cyclic-fatigue loading. Hence only these type of particles led to a meaningful increase in the fatigue threshold, Gth. 'Hybrid' epoxy polymers contained a polymer based toughener and core-shell rubber particles. Several of these hybrid formulations gave very significant improvements in both values of GIc and Gth. For example, one based upon a polyurethane and styrenebutadiene-core/functionalised polymethylmethacrylate-shell particles gave a value of 2570 J/m2 and 438 J/m2 for GIc and Gth, respectively.

Multiple input multiple output (MIMO) techniques are the most promising technologies for next-generation wireless systems to achieve improved channel reliability as well as high spectral efficiencies. In real MIMO downlink scenarios, the numbers of users are usually greater than that of transmit antennas at a base station and the base station is likely to provide a variety of services to different quality-of-service (QoS) users. Therefore, a multiuser scheduling algorithm for the real MIMO BC scenarios has to support a mixture of QoS users simultaneously by exploiting the performance gains of multiple antennas whilst maximizing the sum-rate capacity by selecting a user set for transmission according to performance criteria. The main topic of this thesis is the design of QoS-guaranteed multiuser schedulers for MIMO systems. This should provide different QoS services to different users whilst satisfying the system-level requirements such as fairness among users, minimum data rate and delay constraints as well as trying to maximize the sum-rate capacity of MIMO channels. For this, this thesis first investigates the performance of MIMO transceiver techniques in terms of error rates and the sum-rate capacity with practical considerations to select a practically appropriate MIMO precoding technique. Then a QoS-aware sequential multiuser selection algorithm is proposed, which selects a user set sequentially from each QoS group in order to satisfy QoS requirements by trading off the transmit antennas between different QoS groups. Using a temporally-correlated MIMO channel model validated by channel measurements, a statistical channel state information (SCSI)-assisted multiuser scheduling algorithm is also proposed, which can minimize the effect of the temporal correlation on the sum-rate capacity. Finally, new metrics are proposed to support fairness among users in terms of throughput or delay whilst maximising the sum-rate capacity. With these proposed algorithms, the objective of this thesis, to support a mixture of different QoS users simultaneously with fairness considerations whilst maximising the sum-rate capacity by exploiting the advantages of MIMO techniques with practical implementation in mind, can be achieved.

Neuromorphic computing systems, which aim to mimic the function and structure of the human brain, is a promising approach to overcome the limitations of conventional computing systems such as the von-Neumann bottleneck. Recently, memristors and memristor crossbars have been extensively studied for neuromorphic system implementations due to the ability of memristor devices to emulate biological synapses, thus providing benefits such as co-located memory/logic operations and massive parallelism. A memristor is a two-terminal device whose resistance is modulated by the history of external stimulation. The principle of the resistance modulation, or resistance switching, for a typical oxide-based memristor, is based on oxygen vacancy migration in the oxide layer through ion drift and diffusion. When applied in computing systems, the memristor is often formed in a crossbar structure and used to perform vector-matrix multiplication operations. Since the values in the matrix can be stored as the device conductance values of the crossbar array, when an input vector is applied as voltage pulses with different pulse amplitudes or different pulse widths to the rows of the crossbar, the currents or charges collected at the columns of the crossbar correspond to the resulting VMM outputs, following Ohm’s law and Kirchhoff’s current law. This approach makes it possible to use physics to execute direct computing of this data-intensive task, both in-memory and in parallel in a single step. First of all, I will present a comprehensive physical model of the TaOx-based memristor device where the internal parameters including electric field, temperature, and VO concentration are self-consistently solved to accurately describe the device operation. Starting from the initial Forming process, the model quantitatively captures the dynamic RS behavior, and can reliably reproduce Set/Reset cycling in a self-consistent manner. Beyond clarifying the nature of the Forming and Set/Reset processes, a bulk-like doping effect was revealed by the model during Set and supported by experimental results. This phenomenon can lead to linear analog conductance modulation with a large dynamic range, which is very beneficial for low-power neuromorphic computing applications. Second, an integrated memristor/CMOS system consisting of a 54×108 passive memristor crossbar array directly fabricated on a CMOS chip is presented. The system includes all necessary analog/digital circuitry (including analog-digital converters and digital-analog converters), digital buses, and a programmable processor to control the digital and analog components to form a complete hardware system for neuromorphic computing applications. With the fully-integrated and reprogrammable chip, we experimentally demonstrated three popular models – a perceptron network, a sparse coding network, and a bilayer principal component analysis system with an unsupervised feature extraction layer and a supervised classification layer – all on the same chip. Beyond VMM operations, the internal dynamics of memristors allow the system to natively process temporal features in the input data. Specifically, a WOx-based memristor with short-term memory effect caused by spontaneous oxygen vacancy diffusion was utilized to implement a reservoir computing system to process temporal information. The spatial information of a digit image can be converted into streaming inputs fed into the memristor reservoir, leading to 100% accuracy for simple 4×5 digit recognition and 88.1% accuracy for the MNIST data set. The system was also employed for solving other nonlinear tasks such as emulating a second-order nonlinear system.

This research investigates the fundamental physics of laser processing of multi-coated materials, through spectroscopic studies and a mathematical modeling of laser material interaction. This work focuses particularly on developing an in-situ quality monitoring system by detecting defects generated in the processing, understanding the effect of coated materials on defects formation, and further characterizing differences between newly developed lasers in regard to energy transfer. First, several welding defects generated in CO2 laser processing of a multi-coated material are monitored using Optical Emission Spectroscopy (OES). Tracking the specific constituent behaviors that induce the defects is proposed as a novel way to monitor the process. Second, in order to obtain promising results in both defect detection and defect classification, a machine learning algorithm, a Support Vector Machine (SVM), is adopted for the spectral data analysis using the richness of the available data. The richness is a major benefit in the use of the optical emission spectroscopy because the spectrometer can resolve and distinguish each spectral line of the constituents of the target materials. Third, a numerical simulation study is presented to investigate the effect of the coating material for understanding the interfacial phenomena in the laser processing of the multi-coated material. These interfacial phenomena are important because they determine the processed qualities of the target samples in the laser material interaction. The interfacial phenomena such as recoil pressure, capillary and thermo capillary force are investigated by comparing a coating free material with a coated material. Finally, characteristics of the energy transfer of the disk laser and the fiber laser are identified to provide users with insight into which laser might be more suitable for a given application. To assess the laser systems, two factors are considered: energy absorption by the laser induced plasma, which is an inevitable phenomenon in laser material interactions, and the penetration features of the samples irradiated by the attenuated laser beam after absorption by the plasma. The work presented in this study can be utilized to achieve the quality assurance, to understand energy transfer in the laser material processing, and thus eventually to control the process.

This dissertation develops two new parametric and nonparametric methods for estimating risk-neutral measures (RNM) which embody important information about market participants’ sentiments concerning prices of the underlying asset in the future, and investigates empirical performance of parametric RNM estimation methods. The first essay, “Estimation of Risk Neutral Measures using the Generalized Two-Factor Log-Stable Option Pricing Model”, constructs a simple representative agent model to provide a theoretical framework for the log-stable option pricing model and then implements a new parametric method for estimating the RNM using a generalized two-factor log-stable option pricing model. Under the generalized two-factor log-stable uncertainty assumption, the RNM for the log of price is a convolution of two exponentially tilted stable distributions. The generalized two-factor log-stable RNM provides a sufficiently accurate tool for estimating the RNM from observed option prices even if the log-stable assumption might not be satisfied. I estimate the RNM using the S&P 500 index options and find that the generalized two-factor log-stable model gives better performance than alternative models in fitting the observed option prices.The second essay, “Parametric Risk Neutral Measure Estimation Methods: A Horse Race”, implements 12 parametric RNM estimation methods by means of the closed-form or characteristic function of RNM distributions and then compares the empirical performance under three criteria – the root mean squared error (RMSE), likelihood ratio (LR), and the root mean integrated squared error (RMISE). The empirical results show that the generalized two-factor log-stable model outperforms other alternative parametric RNM estimation methods. The third essay, “Nonparametric Estimation of Risk-Neutral Measures using Quartic B-Spline CDFs with Power Tails”, proposes a new nonparametric (BSP) method. I model a RNM cumulative distribution function (CDF) using quartic B-splines with power tails so that the resulting RNM probability density function (PDF) has continuity C2 and arbitrage-free properties. Since the number of knots is selected optimally in constructing the quartic B-spline RNM CDF, my method avoids both overfitting and oversmoothing. To improve computational efficiency and accuracy I introduce a 3-step RNM estimation procedure that transforms a nonlinear optimization problem into a convex quadratic program, which is efficiently solved by numerical optimization software.