Font Size: a A A

Research And Applications Of Global Optimization Algorithm In HPM Sources Design

Posted on:2017-02-04Degree:MasterType:Thesis
Country:ChinaCandidate:H ZhangFull Text:PDF
GTID:2308330488483501Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
High Power Microwave (HPM) has potential and enormous prospect in numerous domains of science、technology and security, such as electronic countermeasures、radar communication particle acceleration and wireless power transmission, so it attracts much attention definitely. However, generation of HPM involves interaction processes that are nonlinear transient and nonequilibrium, thus it’s difficult to make theoretical analysis for the process of generation. Meanwhile, experiments are costly. At present, it’s the general method to use particle-in-cell (PIC) simulation software to optimize HPM sources by sequential scanning parameters based on their structures defined by theoretical analysis. This method is inefficient, and because of parameters’nonlinear effects on performance, it’s almost impossible to get global best solutions. Thus, global optimization algorithms will improve efficiency of HPM sources’design process and make results closer to the global best solutions. Besides, this paper takes into account both global optimization algorithms and parallel strategy, thus the level of optimization design is enhanced again.Firstly, this paper develops a 2.5-dimensional parallel PIC code called NEPTUNE2D, which is capable for fast simulation and design of HPM devices and electron-vacuum tubes in r-z coordinate systems. Instead of traditionally developing from the serial program, this paper develops parallel code directly based on the J parallel adaptive structured mesh application infrastructure (JASMIN), which has high parallel efficiency and strong expansibility. Physical figures are displayed by Teravap that has powerful post-processing capabilities. Electromagnetic fields are updated by using finite-difference-time-domain (FDTD) to solve Maxwell equations and particles are advanced by using Buneman-Boris method to solve the Newton-Lorentz equation. In order to make the code fitter to massively parallel simulation and effectively improve the codes’computing efficiency, a charge conservation method in PIC simulation called zigzag is adopted instead of the classical one to analyze interaction between particles and fields, avoiding to solve complex Poisson equation. To apply to the actual HPM device design, basic physical function modules are accomplished, such as external magnetic field load, electromagnetic wave input/output, and particle injection/absorption. These modules are verified and validated by simulations of a coaxial-line, a cylindrical waveguide, a coaxial diode and a foilless diode. The code is integrally verified by simulating a MILO device. By using this code, a new X-band Coaxial Relativistic Backward Wave Oscillator (CRBWO) with high efficiency which has a metal foil as current reduction structure is researched, its output power is 3GW and its efficiency is 21%. The parallel efficiency is given, and this code’s parallel capability and expansibility is also tested and verified.And then, global intelligent optimization algorithms in HPM devices are studied. Genetic algorithm is adopted to design HPM device, and this paper puts forward the idea that two layers are both parallel, which are optimization algorithm and PIC simulation. Parallel genetic algorithm code is developed based on Message Passing Interface (MPI). Syncretized with NEPTUNE2D, the fully electromagnetic code is developed to be capable of global optimization and two-layer parallel, using computer hardware more efficiently. The code is used to optimize a MILO device to make it compact and improve its performance. Compared with the old device, the weight and volume are reduced, and output power rises from 6.0GW to 9.3GW while efficiency rises from 13.3% to 19.8%. A CRBWO is optimized, output power rises from 3GW to 4GW while efficiency rises from 21.4% to 28.6%.
Keywords/Search Tags:particle-in-cell simulation, HPM sources, Genetic Algorithm, parallel computing, optimization algorithm
PDF Full Text Request
Related items