| This paper studied two research topics related to High Power Microwave (HPM) . The first is on microwave imaging as an evaluation method for the biological effects of HPM. The second is on automated HPM antennas design. Research works on both topics are based on network parallel computation and Genetic Algorithms.Microwave biological effects are usually divided into thermal biological effects and nonthermal biological effects. When a biologic tissue is exposed into microwave radiation, its temperature will rise, and it may be caused some pathological changes. Due to its huge radiating power, HPM can cause serious biological effects.Microwave imaging is a kind of contactless imaging technologys based on complex permittivity of biological tissues. Because pathological changes will change the complex permittivity of a tissue, as well as microwave imaging can obtain temperature distribution inside biological tissues because complex permittivity is sensitive to temperature, microwave imaging is a feasible evaluating method of HPM biological effects.An antenna is a key component of a HPM system. It directly affects the biological effects cuased by HPM radiation. An antenna in a HPM system not only should be end-fire, high gain, and high power capacity so that the radiation is easy to be directed and can concentrate powerful electromagnetic energy in an object, but also it should satisfy requirements such as low sidelobe level (SLL) and high the front to back ratio (FBR) so as to a HPM radiation will not cause bad biological effects to it operators. Designing a HPM antenna is a real challenge. To meet the challenge, this paper exploredautomated antenna design.Main research works and innovations in this paper are introduced as followings: 1. Modifying the Simple Genetic Algorithm.Genetic Algorithms (GAs) are widely used as a kind of optimization methods. In this paper, a modified GA, which modified select and mutation operators of a Simple GA in order to improve its convergence, was brought forward.Elite technique is used in the select operator of the modified GA. In the modified GA, the select operator copies the best individual directly to next generation instead randomly selects individuals to be coped as it does in a simple GA. So the phenomena of losing best individual during optimization process can be avoided and the speed of convergence can be improved.Dynamic mutation probability is used in the mutation operation of the modified GA. When individuals in a GA group become similar, the optimization process stagnates because of the shortage of individual diversity. The modified GA will increase its mutation probability to enhance individual diversity, break down optimization stagnation and accelerate the optimization process.Results of a numerical example of curve fitting a multi-exponent function show the modified GA can obtain optimum more quickly than a simple GA.2. Constructing the "YuanMou II" Beowulf Parallel Computing SystemParallel computation can greatly improve computing efficiency and decrease computing time. Beowulf system is a kind of network parallel computing systems, which mentions very high performance-price ratio and easy scalability. A Beowulf system comprised of 16 PCs, namely "YuanMou II", was constructed. The system runs Windows 2000 operating system and is controlled and supported by MPICH.NT 1.2.3 software, a latest version of MPI (Message-Passing-Interface) implementation. A high-speed 100 Mb/s Ethernet network interconnected all PCs. Results of parallel tests showed this Beowulf system could run high-performance computing tasks.3. A Novel Dynamic Task Scheduling AlgorithmTask scheduling plays a key role in the operation of cluster systems. For general cases, task scheduling has been proved as NP-complete. This paper introduces a novel dynamic task scheduling algorithm. Without executing a computationally complex scheduling function and knowing some auxiliary informations such as the execution time of tasks in each processor, the proposed algorithm uses heuristics to maximize the utilizati...
|
PDF Full Text Request