Search

In this dissertation, we explore methods to combine multiple frequency Ground Penetrating Radar (GPR) signals in a manner to improve the resolution of images of deeply buried targets. We propose using an optimization problem solver to combine multiple GPR frequency scans over the same area to improve image resolution. First, we discuss GPR basics. Second, we report on a method to simulate GPR radar scans over any type of terrain, any frequency, and any target depth for use in our study of GPR compositing methods. Third, we define an optimization problem solver, exploring its capability to achieve reasonable results as well as propose a figure of merit for the best solution measurement tool. In comparing the optimization problem solver result to methods previously explored in the literature, detailed by Dougherty [8], then Booth [5][48] and finally Bancroft [3], we found our algorithm exhibits a meaningful improvement compared to the named methods. As an extension, we explored comparing scans from various heights using the optimization problem solver method with a Chirp excitation function at the same heights, finding edge detection improved with the response from the Chirp excitation function, but depth detection poorer than the optimization problem solver.

The development of a large-scale phased array radar system such as the future MPAR will need a cost-effective tool for predicting electromagnetic characteristics of antennas. Simulating and optimizing of large finite phased array antennas using commercially available solvers are time-consuming and memory-extensive even though they are highly capable of solving general electromagnetic problems with acceptable accuracy. In this work, a full-wave electromagnetic solver based on finite-difference time-domain (FDTD) method has been developed for simulating phased array antennas. The planar array or array element can be simulated, optimized, or analyzed using FDTD theory based on an orthogonal, regular Cartesian lattice. The FDTD updating equation for diagonally anisotropic material was obtained for periodic structure based on the cylindrical coordinate system. This FDTD algorithm can be used to simulate active element patterns of conformally cylindrical array antennas. The simulation of active element patterns in an infinite faceted-cylindrical array was accomplished with a nonorthogonal and unstructured grid. The derivation of FDTD theory and periodic boundary condition for a structure based on the nonorthogonal and unstructured grid is presented. In this work, two simulation schemes, which are based on computed near-field current density information and the physical knowledge of finite array antennas, were presented for predicting broadside array radiation characteristics with the consumption of relatively low computational resources. The validation of the simulation program and schemes was fulfilled by comparing simulation results with measurements taken by near-field and far-field techniques.

In the first part of this dissertation, we explore the metric invariance of Maxwell's equations to design metamaterial blueprints for three novel electromagnetic devices. The metric invariance of Maxwell's equations here means that the effects of an (hypothetical) distortion of the background spatial domain on the electromagnetic fields can be mimicked by properly chosen material constitutive tensors. The exploitation of such feature of Maxwell's equations to derive metamaterial devices has been denoted as `transformation optics' (TO). The first device proposed here consists of metamaterial blueprints of waveguide claddings for (waveguide) miniaturization. These claddings provide a precise control of mode distribution and frequency cut-off. The proposed claddings are distinct from conventional dielectric loadings as the former do not support hybrid modes and are impedance-matched to free-space. We next derive a class of metamaterial blueprints designed for low-profile antenna applications, whereby a simple spatial transformation is used to yield uniaxial metamaterial substrate with electrical height higher than its physical height and surface waves are not supported, which is an advantage for patch antenna applications. We consider the radiation from horizontal wire and patch antennas in the presence of such substrates. Fundamental characteristics such as return loss and radiation pattern of the antennas are investigated in detail. Finally, transformation optics is also applied to design cylindrical impedance-matched absorbers. In this case, we employ a complex-valued transformation optics approach (in the Fourier domain) as opposed to the conventional real-valued approach. A connection of such structures with perfectly matched layers and recently proposed optical pseudo black-hole devices is made.In the second part of this dissertation, we move from the derivation of metamaterial blueprints to the application of pre-defined unit-cell metamaterial structures for miniaturization purposes. We first employ electric-field-coupled (ELC) resonators and complementary electric-field-coupled (CELC) resonators to design a new class of electrically small antennas. Since electric-field coupled resonators were recently proposed in the literature to obtain negative permittivity response, we next propose ELC resonators as a new type of waveguide loadings to provide mode control and waveguide miniaturization.

Antennas utilising a dielectric medium are technologies that have become popular in modern wireless platforms. They offer several desirable features such as high efficiency, electrically small and resistance to proximity detuning. Being a volumetric radiator however, realising a final, commercially competitive solution, often requires the use of a computational optimisation algorithm. In the realm of antenna design the practice of optimisation typically involves an automated routine consisting of a heuristic algorithm and a forward solving engine such as the finite element method (FEM) or finite difference time domain (FDTD) method. The solving engine is used to derive a post-processed performance value typically referred to as an objective or fitness function, while the heuristic method uses the objective function data to determine the next trial solution or solutions that approach a design goal. Nowadays, commercially viable antenna platforms are not characterised by a single performance value, but rather, a series of objective functions that are often inherently conflicting. Thus, an increase in one objective function results in a decrease in another. The optimisation algorithm is therefore required to seek a solution dictated by the preferences of the designer. Classical literature dominantly featured preference articulation, a priori, where the set of objectives are transformed into a scalar using a predefined preference arrangement. Contemporary theory implements the articulation a posteriori, where the complete set of compromise solutions are sought by the optimisation algorithm. It is hypothesised that modern multi-objective optimisation (MOO) theory, using a posteriori preference articulation, can be more useful for contemporary antenna design. By treating the objectives as individual dimensions in a mathematical space, it allows for independent, simultaneous optimisation. At the time of writing this dissertation, all commercial simulation software that include an optimisation algorithm use a predefined preference to the performance criteria. Thus, where a large set of equally potential solutions exist, only one final solution is delivered. This thesis examines two novel dielectric antenna technologies and uses modern MOO theory to obtain new solutions that supersede their prototypes. Taking a commercial perspective by optimising the electromagnetic performance and the physical size of the antenna simultaneously, it is hypothesised this allows an unprecedented insight into the inherent tradeoffs of practical antenna configurations.

A time reversal mirror is a device capable of receiving a signal in time, keeping it in memory, and sending it back into the same medium in the reversed direction of time. The main effect is the fascinating refocusing of the scattered signal, which is formed by sending the pulse into a complex medium, after time reversal through the same medium. The refocused signal is a pulse with shape similar to the initial pulse along with some low amplitude noise. This surprising effect has a great potential for application in domains such as medical imaging, underwater acoustics and wireless communication. Time reversal is studied in reflection and transmission. In both cases, we demonstrate the self-averaging properties of the time reversed refocused pulse. An accurate numerical method for simulating waves propagating in complex one-dimensional media is employed. Numerical simulations are used to reproduce the time-reversal self-averaging effect which takes place in randomly layered media. The effect of refocusing is enhanced in a regime where the inhomogenities are smaller than the pulse, which propagates over long distances compared to its width. Time Reversal can be implemented using several numerical methods including Transfer Matrix, Finite Difference Time Domain (FDTD) & Boundary Integral Methods. This thesis includes a comprehensive comparison of the methods in terms of speed and accuracy. The most efficient method for implementing time reversal is then used to obtain numerical evidence for potential use of a sliding window time-reversal technique for detecting a buried cavity/object inside the medium. The numerical methods are well adapted for generalization to the multi-dimensional case.

The advancement of MRI as a radiological instrument has been associated with a constant drive towards higher magnetic field strengths resulting in higher operational frequencies. More powerful magnets bring the promise of enhanced signal to noise ratio, exquisite resolution, and reduced scan times. At the same time however, MRI at higher frequencies adds significant engineering complexities to the MRI experiment, most notably in designing safe, versatile, and high-performance radio frequency (RF) coils. In this work, computational and theoretical electromagnetic analysis of several RF coils used in MRI are presented at Larmor frequencies that range between 64 and 470 MHz representing clinical imaging at 1.5:11 Tesla. The electromagnetic interactions with phantoms and anatomically detailed head models, including a developed high-resolution human head mesh, are studied at different field strengths. The computational tool of choice here was the finite difference time domain (FDTD) method. Combined with measurements using an 8 Tesla MRI system, currently the most powerful clinical magnet in the world and a 1.5 Tesla system, the FDTD method is utilized to study, analyze, and eventually design RF coils. Innovative Engineering approaches using phased array techniques are presented to improve the performance of RF head coils in terms of transverse magnetic field uniformity and reduction of specific absorption rate for operation at 4.7 and 8 Tesla. Novel analytical derivations are presented to explain the source of the MR signal. The combination of the analytical derivations, FDTD modeling, experiments. and infrared imaging gives a new prospective onto the electromagnetics associated with low and high field clinical imaging.