<p>The traditional approach for estimating sensitivities utilizes finite difference (FD) approximations which are time-intensive even for simple problems. This involves perturbing each design parameter in the forward and/or backward directions and simulating the perturbed structures. For a problem with <em>N</em> optimizable parameters, this approach requires at least <em>N</em>+1 simulations. If<em> N</em> is large, this approach may easily become impractical due to the intensive simulation time.</p> <p>The Adjoint Variable Method (A VM) aims at obtaining the response sensitivities using at most one extra simulation regardless of the number of designable parameters. The field information is stored at specific points related to each parameter in both the original and adjoint simulations. This approach was applied to sensitivity analysis of scalar objective functions and frequency domain responses.</p> <p>This thesis addresses a new A VM algorithm for estimating time domain response sensitivities using time domain transmission-line modeling (TD-TLM) method. Our algorithm obtains the sensitivities of any electromagnetic time domain response over the whole simulation time with respect to all parameters using only one extra time domain simulation. A very good match is obtained between our sensitivity estimates and those obtained through the accurate and time-intensive central difference approximation.</p> <p>One of the motivations for sensitivity analysis is gradient-based optimization. The optimization process speeds up by using our A VM algorithm for gradient estimation. In this literature, we will implement gradient-based optimization using our AVM through different applications including microwave imaging problem. The results show good match between the sensitivities obtained using our A VM approach and those obtained using the more expensive finite difference approximation.</p> / Master of Applied Science (MASc)

<p>The traditional approach of Magnetic Tracking Systems (MTS) utilizes approximate models and Parameter Extraction (PE) for Position and Orientation (P&O) determination. The approximate models give inaccurate P&O information outside the "constrained region". PE is an iterative, intensive process for P&O calculations, which limits the speed of the tracking process.</p> <p>Our MTS approach aims at accurate real-time P&O tracking. We utilize Artificial Neural Networks (ANN) with PE functionality to carry out the computational task for real-time P&O tracking. We apply Space Mapping (SM) modeling afterwards for system calibration to improve the accuracy of P&0 determination</p> <p>This thesis addresses a different approach for P&O determination. The main motivation of this work is to determine the P&O in a fast and accurate manner. It this work, we mathematically develop and experimentally implement our MTS for both 2-D and 3-D examples. The results show good match between our extracted P&O based on our MTS approach and the actual P&0 measured values.</p> / Master of Applied Science (MASc)

<p>This dissertation studies the problem of identification of the emergency locator transmitter (ELT) signals as related to the Search and Rescue Satellite Aided Tracking (SARSAT) system. The ELT identification is particularly important in order to increase the probability of detection and eliminate sources of interference from the data set.</p> <p>A set of parameters that uniquely characterizes the ELT signals is selected, namely, the width of the average spectrum sidebands, the ratio of the sideband plateaus of the average spectrum with respect to the carrier peak and the sweep period of the signals. Two methods for estimating the sweep period are developed theoretically. These are the sawtooth and the crosscorrelation methods.</p> <p>The identification techniques are tested using computer generated signals and real testbed ELT signals recorded by the Communications Research Centre (CRC) in Ottawa.</p> <p>A study of the different interference sources in the 121.5/243 MHz SARSAT frequency bands is provided. Three different sources of interference are generated and tested using the identification techniques.</p> <p>The performance of the proposed techniques is further investigated using real data from two different passes of COSPAS satellite CI.</p> <p>Using the identification techniques presented in this thesis, it is possible to consolidate the number of detections per day from multiple satellites and multiple satellite passes.</p> / Doctor of Philosophy (PhD)

<p>This work describes two network designs--Photonic Bus Networks and Photonic Mesh Networks. These two designs were developed in order to provide the basis for exploring the possibilities of novel design techniques in future fibre networks. In both cases, attempts are made to take advantage of emerging technologies in the area of fibre optic. The networks were not created to compete with other network designs, athough obvious comparisons with other netorks are possible. Rather, these networks were conceived to create a framework within which to study the behaviour of certain phenomenon that may be present in some future fibre networks. Both simulation studies and analytical modelling were used to improve understanding of the behaviour of the network design techniques. Photonic Bus Networks build upon the emerging IEEE 802.6 network standard by considering a multihop implementation on a physical transmissive optical star topology. Media access techniques are designed to allow for slot reuse at each node, to improve network performance while ensuring fair network access. The combination of the new media access with the broadcast nature of the physical optical network allows for the use of new and novel bandwidth allocation design techniques that would not be possible in IEEE 802.6 networks. Photonic Mesh Networks can be considered a two-dimensional generalization of Photonic Bus Networks. They are topologically regular two-dimensional grid networks that use a novel routing technique known as deflection routing and allow for easy bandwidth allocation. A framework is created for the consideration of traffic processing at a node in a Photonic Mesh Network. Algorithms are defined which consist of three separate but related strategies - the access strategy, the routing strategy and the allocation strategy. Simulation studies indicate that the interaction of these separate strategies affect the performane of the network.</p> / Doctor of Philosophy (PhD)

<p>The von Neumann computer model, upon which most of our modern computers are based, is lacking in both parellelism and support for multitasking. One convenient measure of processor-parallelism is the degree to which a processor utilizes its available memory bandwidth for useful work: its Memory Bandwidth Efficiency (MBE). A processor which exhibits an MBE of 100% is operating as fast as possible for its given memory speed.</p> <p>Conventional processor accelerators like prefetching and pipelining increase parallelism but suffer from the jump problem, in which a taken jump may cause incorrect prefetching. These accelerators are not well adapted to multitasking, although enhancements like selectable register files may be used to reduce context switching overhead. Pipelined multistreaming achieves a high degree of parallelism, avoids the jump problem and supports efficient context switching, but its performance is load dependent and it is awkward to implement. Furthermore, none of these architectures support efficient process resource sharing.</p> <p>Microcode Level Timeslicing (MLT) is a multistream processor architecture that achieves very high processer MBE, has no-overhead context switching and provides support for resource sharing. Within the processor, process state information is replicated N-fold. Prefetching occurs horizontally across streams, allowing the jump problem to be circumvented. Context switching occurs at microinstruction boundaries, giving no overhead for up to N streams. The fetching and executing mechanisms are controlled by a Stream Control Unit (SCU), which contains task status information for each stream. Efficient process control and resource sharing operations are readily supported in the SCU.</p> <p>The hardware and software design of a prototype processor demonstrating the MLT principle for up to 16 streams is presented. Process control and syncronization operations are implemented at the microcode level. Two high-level language benchmarks, the Sieve of Eratosthenes and they Dhrystone, are used to evaluate the prototype's performance.</p> / Doctor of Philosophy (PhD)

<p>This thesis investigates the carrier synchronization problem due to Doppler shift encountered in low data rate, low earth orbit (LEO) satellite communication systems. In particular, two new techniques are developed. The first technique which is referred to as the novel technique, is based on the idea of applying a linearly decreasing frequency sweep to the received signal as a method to reduce the Doppler shift to tolerable levels, whereby conventional carrier synchronization may be employed. This method has shown to be better suited for satellites at larger cross track angles. However, for direct overhead satellite pass, improved Doppler compensation is achieved by utilizing the modified novel technique. Simulations have been carried out to explore the performance of the novel technique. Simulations have been carried out to explore the performance of the novel technique and results have shown that it is capable of reducing Doppler shift by at least a factor of two and performs better for satellites at large cross track angles. The second technique of Doppler compensation is based on Maximum Likelihood (ML) estimation of multiple parameters. This algorithm makes use of a model developed for the phase error introduced due to Doppler frequency shift and Doppler rate, and the two implementable solutions investigated are; (1) the grid search, and (2) the ML quadriphase shift keying (QPSK) Doppler compensator. These solutions have been realized by utilizing approximations for low signal to noise ratio (SNR) conditions. The grid search is a non tracking technique and may be used for coarse Doppler compensation. Unfortunately, simulation results have shown that it is computationally intensive. On another note, a block diagram consisting of a phase rotator, matched filtering and two feedback loops has been developed to represent the ML QPSK Doppler compensator. Three digital implementations of this structure have been analyzed, namely; steepest descent, optimized steepest descent and Newton Raphson method. The first two, have two modes of operation; conventional feedback and adaptive feedback. Simulations have shown that the adaptive optimized steepest descent has the best performance, that is, the shortest lockup time. Comparisons to previous work in this area, have shown that the proposed Doppler compensator has superior performance.</p> / Doctor of Philosophy (PhD)

<p>On August 1, 1983, a sudden failure of a coolant tube, known as a pressure tube, was experienced in one of the CANDU reactors at Pickering 'A' Nuclear Generating Station of Ontario Hydro. A contributing factor to the rupture of the tube was a contact for a period of several years between the tube and an outer coaxial tube, called the calandria tube. The contact between the pressure tube and calandria tube occurred because the annular spacers (garter springs), used to maintain the coaxial configuration when the uranium fuel bundles are loaded inside the pressure tube, had shifted out of their design locations during the construction stage and subsequent hot conditioning of the reactor. The displacement of the spacers was attributed to vibration induced by various sources and was determined to have taken place prior to loading the fuel bundles in the reactor.</p> <p>Since the garter springs are not directly accessible by mechanical means, extensive dismantling of the fuel channels would have been necessary to reposition the springs in their designated locations. This Thesis describes a novel electromagnetic method to reposition without the garter springs without dismantling the fuel channels. The method consists of inserting an electromagnetic coil into the pressure tube to a location adjacent to the spacer and passing a time-varying electric current in the coil to induce secondary currents in the pressure tube and the spacer. The interaction between the induced current in the spacer and net magnetic field in the spacer region results in an electro-magnetic force having an axial component large enough to displace the spacer in the required direction. In practice, current impulses generated by discharging a large capacitor bank were used to achieve the required garter: spring displacement.</p> <p>The new method was successfully developed and impIemented to reposition the displaced garter springs in five new CANDU reactors in Ontario. The savings in the reactor repair cost, interest charges and replacement energy costs were in the order of hundreds of millions of dollars. Equally large benefits and savings will be realized if the need to use this technique in commissioned reactors becomes mandatory. The practical aspects of the work including the design of pulse power cables and a coil sufficiently compact to fit in the pressure tube, but strong enough to withstand the stress, were developed by the author at the Ontario Hydro laboratories. The Thesis describes not only this aspect but also develops a model for analysis of the operation and design of the device.</p> <p>There are a number of other applications of the electromagnetic technique which may also benefit from the analysis, especially for electromagnetic metal forming and fusion technologies.</p> / Doctor of Philosophy (PhD)

<p>In this thesis, we study both theoretical and practical aspects of the regularized strict interpolation radial basis function (SIRBFN) estimate or neural network. From a theoretical perspective, we show that the regularized SIRBFN can be globally mean-square (m.s.) consistent whenever the Nadaraya-Watson regression estimate is and the regularization parameter sequence for the SIRBFN is chosen to be asymptotically optimal in the mean-squared fitting error. Hence we prove the Bayes risk consistency of the approximate Bayes decision rules formed from (m.s.-consistent) regularized SIRBFN posterior probability estimates. Similarly, we prove the m.s.-consistency of the regularized SIRBFN predictor for the class of Markovian nonlinear autoregressive time series generated by an i.i.d. noise process. In a one-step-ahead prediction experiment with a phonetically-balanced suite of male and female speech waveforms, the proposed predictor offers an average 2.2dB improvement in predction SNR over corresponding expponentially-weighted RLS predictors. We also show that linearly combining an ensemble of three such proposed predictors via RLS filtering can yield an average 4.2dB improvement over the previous standard RLS predictors, and develop recursive algorithms to update the proposed predictor on-line with reduced computational complexity for certain situations. Two emerging applications areas are then considered. The first is the regression-based approach to nonlinear filtering or state estimation, where the proposed network provides comparable performance to a recurrent MLP-based solution. The second is the dynamic reconstruction of chaotic systems from noisy observational data, where the reconstructed system is shown to generate sequences whose estimated long and short-term dynamical invariants agree closely with those of the original, noise-free system. Taken together, these theoretical and practical results point to the regularized SIRBFN as a principled design choice for RBF neural networks.</p> / Doctor of Philosophy (PhD)

<p>Search and rescue satellite aided tracking (SARSAT) is a method of employing satellites in low polar orbits to relay the emergency signals of distressed aircraft and marine vessels to an earth station. At the earth station, the signals are processed using spectral estimation techniques which permit the calculation of the location of the distressed vehicle. Of considerable importance are the characteristics of the spectrum of the emergency locator transmitter (ELT) signal since the probability of locating the downed aircraft is closely related to the quality of the ELT signal itself.</p> <p>This thesis analyses and investigates the ELT signal spectra for ranges of different parameter values. A mathematical representation for the variation of the pulse duration across the sweep period for actual real ELT signals has been developed and verified. New models are proposed with nearly ideal spectral properties.</p> <p>The matched filter performance for detection of the ELT signals has been calculated. Also, the periodogram and averaged periodogram performances for detection of the ELT signals are calculated and verified by computer simulation.</p> <p>A new processor, called the complex baseband processor, which employs both the periodogram and the maximum entropy method (MEM) is developed. Baseband processing has many advantages over bandpass processing.</p> <p>This thesis examines the processing of simulated ELT signals, real testbed ELT signals and actual data received from satellite passes using this new baseband processor. It is shown theoretically and by computer simulation that the minimum detectable carrier-to-noise density ratio (CNDR) is approximately 21 dB-Hz when the 512 complex-point FFT is applied to 50 blocks comprising 1 second of data.</p> <p>A new technique called rate reduction filtering is developed. By applying this technique to the data sequence, improved frequency resolution is obtained along with an increase in signal-to-noise ratio. Further, a new algorithm called "ELT Tracking Algorithm" based on using rate reduction filtering is developed and analysed.</p> / Doctor of Philosophy (PhD)

<p>An investigation is conducted into the practical electric vehicle implementation of a high speed flywheel energy storage system that can supply and accept peak acceleration and braking power. Electric vehicle battery life cycle can be extended considerably by supplying peak energy requirements from a secondary source. Simulations are performed to determine the peak power and energy requirements over the SAE recommended electric vehicle test procedure. A scaled prototype secondary energy storage unit is built using flywheel energy storage. Tests are conducted to determine the energy transfer capabilities of a flywheel couples high speed permanent magnet synchronous machine through the proposed system's energy storage tank. Results are presented that indicate the necessity of the energy storage tank. A system evaluation is also included which indicates its near-term practically when compared to conventional regenerative control applications. Construction techniques are discussed for the development of high speed drive circuits used in electrically noisy environments. Shielding and electromagnetic interference reduction methods are proposed for application to an actual high efficiency sinusoidal phase current, microprocessor interfaced, vector controlled, high speed permanent magnet synchronous machine drive. A discussion on the theory and construction of such a drive is provided to demonstrate the implementation of a laboratory prototype high speed flywheel energy storage system. An experimental investigation is conducted using the developed high speed drive to determine permanent magnet synchronous machine parameter behaviour in the high speed flux weakening operating range. Special computer assisted measuring techniques are used to determine the realtime operating characteristics for two commercially available machines. The repeatable experimental results show a significant variation in the machine parameters as a function of the speed and torque angle. Implications of the newly identified dynamic parameters are discussed regarding their application to, and effect on, high performance permanent magnet synchronous machine torque control.</p> / Doctor of Philosophy (PhD)