Multiple-input multiple-output (MIMO) systems are a mature technology that has been incorporated into current wireless broadband standards to improve the channel capacity and link reliability. Nevertheless, due to the continuous increasing demand for wireless data traffic new strategies are to be adopted. Very large MIMO antenna arrays represents a paradigm shift in terms of theory and implementation, where the use of tens or hundreds of antennas provides significant improvements in throughput and radiated energy efficiency compared to single antennas setups. Since design constraints limit the number of usable antennas, virtual systems can be seen as a promising technique due to their ability to mimic and exploit the gains of multi-antenna systems by means of wireless cooperation. Considering these arguments, in this work, energy efficient coding and network design for large virtual MIMO systems are presented. Firstly, a cooperative virtual MIMO (V-MIMO) system that uses a large multi-antenna transmitter and implements compress-and-forward (CF) relay cooperation is investigated. Since constructing a reliable codebook is the most computationally complex task performed by the relay nodes in CF cooperation, reduced complexity quantisation techniques are introduced. The analysis is focused on the block error probability (BLER) and the computational complexity for the uniform scalar quantiser (U-SQ) and the Lloyd-Max algorithm (LM-SQ). Numerical results show that the LM-SQ is simpler to design and can achieve a BLER performance comparable to the optimal vector quantiser. Furthermore, due to its low complexity, U-SQ could be consider particularly suitable for very large wireless systems. Even though very large MIMO systems enhance the spectral efficiency of wireless networks, this comes at the expense of linearly increasing the power consumption due to the use of multiple radio frequency chains to support the antennas. Thus, the energy efficiency and throughput of the cooperative V-MIMO system are analysed and the impact of the imperfect channel state information (CSI) on the system's performance is studied. Finally, a power allocation algorithm is implemented to reduce the total power consumption. Simulation results show that wireless cooperation between users is more energy efficient than using a high modulation order transmission and that the larger the number of transmit antennas the lower the impact of the imperfect CSI on the system's performance. Finally, the application of cooperative systems is extended to wireless self-backhauling heterogeneous networks, where the decode-and-forward (DF) protocol is employed to provide a cost-effective and reliable backhaul. The associated trade-offs for a heterogeneous network with inhomogeneous user distributions are investigated through the use of sleeping strategies. Three different policies for switching-off base stations are considered: random, load-based and greedy algorithms. The probability of coverage for the random and load-based sleeping policies is derived. Moreover, an energy efficient base station deployment and operation approach is presented. Numerical results show that the average number of base stations required to support the traffic load at peak-time can be reduced by using the greedy algorithm for base station deployment and that highly clustered networks exhibit a smaller average serving distance and thus, a better probability of coverage.

Systems that exhibit the extraordinary magnetoresistance (EMR) effect and other more disordered semiconductor-metal hybrid structures have been investigated numerically with the use of the finite element method (FEM). Initially, modelling focused on circular geometry EMR devices where a single metallic droplet is embedded concentrically into a larger semiconducting disk. The dependence of the magnetoresistance of such systems on the transverse magnetic field (0 5T) and filling factor (1/16 15/16) are reported and generally show a very good agreement with existing experimental data. The influence of the geometry of the conducting region of these EMR systems was then investigated. The EMR effect was found to be highly sensitive to the shape of the conducting region with a multi-branched geometry producing a four order of magnitude enhancement of the magnetoresistance over a circular geometry device of the same filling factor. Conformal mapping has previously been shown to transform a circular EMR device into an equivalent linear geometry. Such a linear EMR device has been modelled with the EMR mechanism clearly observed. The magnetoresistive response of a circular EMR device upon changes to: the mobility of the semiconducting region; the ratio of metal to semiconductor conductivity; and the introduction of a finite resistance at the semiconductor-metal interface, have also been investigated. In order for a large EMR effect to be observed the system requires: the semiconductor mobility to be large; the conductivity of the metal to be greater than two orders of magnitude larger than that of the semiconductor; and a very low interface resistance. This modelling procedure has been extended to include inhomogeneous semiconductor-metal hybrids with a more complex and disordered structure. Two models are presented, both based upon the random distribution of a small proportion of metal inside a semiconducting material. The resultant magnetoresistance in each case is found to have a quasi-linear dependence on magnetic field, similar to that observed in the silver chalcogenides.

<p>The numerical simulation of physical systems plays a key role in different fields of science and engineering. The popularity of numerical methods stems from their ability to simulate complex physical phenomena for which analytical solutions are only possible for limited combinations of geometry, boundary, and initial conditions. Despite their flexibility, the computational demand of classical numerical methods quickly escalates as the size and complexity of the model increase. To address this limitation, and motivated by the unprecedented success of Deep Learning (DL) in computer vision, researchers started exploring the possibility of developing computationally efficient DL-based algorithms to simulate the response of complex systems. To date, DL techniques have been shown to be effective in simulating certain physical systems. However, their practical application faces an important common constraint: trained DL models are limited to a predefined set of configurations. Any change to the system configuration (e.g., changes to the domain size or boundary conditions) entails updating the underlying architecture and retraining the model. It follows that existing DL-based simulation approaches lack the flexibility offered by classical numerical methods. An important constraint that severely hinders the widespread application of these approaches to the simulation of physical systems.</p> <p><br></p> <p>In an effort to address this limitation, this dissertation explores DL models capable of combining the conceptual flexibility typical of a numerical approach for structural analysis, the finite element method, with the remarkable computational efficiency of trained neural networks. Specifically, this dissertation introduces the novel concept of <em>“Finite Element Network Analysis”</em> (FENA), a physics-informed, DL-based computational framework for the simulation of physical systems. FENA leverages the unique transfer knowledge property of bidirectional recurrent neural networks to provide a uniquely powerful and flexible computing platform. In FENA, each class of physical systems (for example, structural elements such as beams and plates) is represented by a set of surrogate DL-based models. All classes of surrogate models are pre-trained and available in a library, analogous to the finite element method, alleviating the need for repeated retraining. Another remarkable characteristic of FENA is the ability to simulate assemblies built by combining pre-trained networks that serve as surrogate models of different components of physical systems, a functionality that is key to modeling multicomponent physical systems. The ability to assemble pre-trained network models, dubbed <em>network concatenation</em>, places FENA in a new category of DL-based computational platforms because, unlike existing DL-based techniques, FENA does not require <em>ad hoc</em> training for problem-specific conditions.</p> <p><br></p> <p>While FENA is highly general in nature, this work focuses primarily on the development of linear and nonlinear static simulation capabilities of a variety of fundamental structural elements as a benchmark to demonstrate FENA's capabilities. Specifically, FENA is applied to linear elastic rods, slender beams, and thin plates. Then, the concept of concatenation is utilized to simulate multicomponent structures composed of beams and plate assemblies (stiffened panels). The capacity of FENA to model nonlinear systems is also shown by further applying it to nonlinear problems consisting in the simulation of geometrically nonlinear elastic beams and plastic deformation of aluminum beams, an extension that became possible thanks to the flexibility of FENA and the intrinsic nonlinearity of neural networks. The application of FENA to time-transient simulations is also presented, providing the foundation for linear time-transient simulations of homogeneous and inhomogeneous systems. Specifically, the concepts of Super Finite Network Element (SFNE) and network concatenation in time are introduced. The proposed concepts enable training SFNEs based on data available in a limited time frame and then using the trained SFNEs to simulate the system evolution beyond the initial time window characteristic of the training dataset. To showcase the effectiveness and versatility of the introduced concepts, they are applied to the transient simulation of homogeneous rods and inhomogeneous beams. In each case, the framework is validated by direct comparison against the solutions available from analytical methods or traditional finite element analysis. Results indicate that FENA can provide highly accurate solutions, with relative errors below 2 % for the cases presented in this work and a clear computational advantage over traditional numerical solution methods. </p> <p><br></p> <p>The consistency of the performance across diverse problem settings substantiates the adaptability and versatility of FENA. It is expected that, although the framework is illustrated and numerically validated only for selected classes of structures, the framework could potentially be extended to a broad spectrum of structural and multiphysics applications relevant to computational science.</p>