本文選題：智能優(yōu)化算法　切入點：磁滯回線　出處：《浙江師范大學》2015年碩士論文　論文類型：學位論文

【摘要】：磁滯非線性現(xiàn)象常見于物理系統(tǒng)和電磁裝置中,對設(shè)備的安全運行及系統(tǒng)能否穩(wěn)定運行起著重要的作用。近年來,隨著新型材料的發(fā)展,例如磁電復合材料,由于其獨特的物理性質(zhì),在各種各樣的微型器件及整體化裝置,如微傳感器、微電子機械系統(tǒng)設(shè)備以及高密度信息存儲器中都有潛在的應(yīng)用；磁性形狀記憶合金(MSMA)作為一種新型的智能材料,在驅(qū)動器制造方面具有良好的應(yīng)用前景；另外,磁熱療是一種很有前途的癌癥治療技術(shù),為癌癥患者帶來了新的希望。上述這些新型材料及技術(shù)的應(yīng)用都涉及到了磁滯非線性現(xiàn)象,因此新型智能材料的設(shè)計和分析與眾多高新技術(shù)的研發(fā)在很大程度上依賴于磁滯建模及其參數(shù)計算的準確性,了解和分析磁滯特性具有重要意義。建立磁滯模型后,一般會包含很多待確定的參數(shù),參數(shù)的取值不同代表著不同的物理狀態(tài)。模型設(shè)計的有效性在很大程度上取決于參數(shù)提取的準確性,無論一個磁滯模型的理論有多完美,如果沒有行之有效的參數(shù)計算方法,其可操作性也就無法得到保證。目前從物理和數(shù)學兩種角度出發(fā),研究者已提出了多種描述磁滯現(xiàn)象的模型,較常見的有：Bouc-Wen磁滯模型、Preisach磁滯模型、Jiles-Atherton(JA)磁滯模型。在眾多的磁滯模型中,JA模型的物理意義清晰、參數(shù)較少、僅包含一個一階常微分方程,但是模型參數(shù)識別的復雜和困難性一直以來困擾著人們。許多傳統(tǒng)的優(yōu)化方法曾用來計算JA模型參數(shù),但是這些方法容易受初始值選擇的影響,其結(jié)果是算法的收斂性往往得不到保證、易陷入局部極小且計算量通常也會很大。更令人擔憂的是,很多優(yōu)化方法,尤其是基于求導尋優(yōu)的優(yōu)化方法,面對模型方程的不連續(xù)、離散、單峰與多峰等數(shù)學性質(zhì)也越來越顯得“力不從心”。最近發(fā)展起來的基于群體智能的新型優(yōu)化技術(shù)在各種領(lǐng)域的參數(shù)計算中受到了很大的關(guān)注。智能優(yōu)化算法是基于仿生學的隨機優(yōu)化算法,典型的方法有Eberhart與Kennedy提出的粒子群優(yōu)化(Particle Swarm Optimization, PSO)算法、遺傳算法(Genetic Algorithm, GA)、差分進化算法(Differential Evolutionary,DE)和Dorigo提出的蟻群算法等。這些方法被廣泛應(yīng)用于科學研究和實際問題求解中,并取得了傳統(tǒng)優(yōu)化方法無法取代的成效。智能優(yōu)化類算法本身具有很強的適用性,且對目標函數(shù)的連續(xù)性無任何要求、對初始解的選取不敏感,因此把智能優(yōu)化算法作為求解復雜優(yōu)化問題的候選算法是非常具有現(xiàn)實意義的。本文所要解決的主要問題包括：(1)、在充分熟悉國內(nèi)外研究現(xiàn)狀、深刻理解并掌握智能優(yōu)化算法及MATLAB/Simulink動態(tài)仿真集成環(huán)境的基礎(chǔ)上,基于物質(zhì)磁化機理從物理角度出發(fā)提出了一種粒子群優(yōu)化算法(PSO)結(jié)合MATLAB/Simulink動態(tài)仿真集成環(huán)境的Jiles-Atherton(JA)磁滯回線模型參數(shù)計算方法。并分別以無噪及加噪的仿真數(shù)據(jù),對三組參數(shù)值不同的JA模型進行數(shù)值實驗。(2)、另外本文構(gòu)建了基于Simulink模塊的JA模型方程,通過求解模型方程得到了準確的B-H磁滯回線。并將JA模型的Simulink模塊方程與算法實現(xiàn)了無縫融合,為算法優(yōu)化的順利運行提供了保障。(3)、算法中控制變量的取值不同會對優(yōu)化精度、計算時間及收斂性有很大的影響,因此針對不同的問題,如何選擇最優(yōu)的參數(shù)配置是本文需要關(guān)注的一個問題。(4)、最后將計算結(jié)果與遺傳算法做了比較,經(jīng)分析發(fā)現(xiàn)對于復雜的Jiles-Atherton非線性磁滯模型參數(shù)計算,PSO算法表現(xiàn)出很好的魯棒性,而且無論是計算精度、還是收斂性都要好于遺傳算法�？梢�,在材料磁滯特性的研究上,針對模型參數(shù)識別的復雜和困難性,粒子群優(yōu)化算法結(jié)合MATLAB/Simulink動態(tài)仿真集成環(huán)境是一種有效可行的研究方法。
[Abstract]:The nonlinear hysteresis phenomenon is common in physical systems and electromagnetic devices, for the safe operation of equipment and system plays an important role in the stable operation. In recent years, with the development of new materials, such as magnetoelectric composite materials, because of its unique physical properties, micro devices and integrated in a variety of devices, such as micro sensors. The potential applications of micro electro mechanical system equipment and high density information storage; magnetic shape memory alloy (MSMA) as a new type of smart materials, has a good application prospect in the drive manufacturing; in addition, magnetic hyperthermia is a promising cancer treatment technology, has brought new hope for cancer patients. The application of these new materials and techniques are related to the phenomenon of nonlinear hysteresis, so the design and analysis of a new type of intelligent material with many high-tech R & D in it A large extent dependent on the accuracy of Hysteresis Modeling and parameter calculation, is of great significance to understand and analyze the hysteresis characteristics. A hysteresis model, usually contain many parameters to be determined, different parameters represent different physical states. The validity of the model design depends largely on the accuracy of parameter extraction, no matter a hysteresis model theory is perfect, if there is no effective parameter calculation, the operability is not guaranteed. Starting from the two perspectives of physics and mathematics at present, researchers have proposed a variety of models to describe the hysteresis phenomenon, the more common are: Bouc-Wen hysteresis model, Preisach hysteresis model, Jiles-Atherton (JA) model. In many of the hysteresis hysteresis model, the JA model with clear physical meaning, less parameters, contains only a first-order differential equation, but the model parameters The identification of complex and difficult has been plaguing people. Many of the traditional optimization method was used to calculate the parameters of the JA model, but these methods are easily affected by the initial value of the impact of the choice, the result is the convergence of the algorithm are not guaranteed, easy to fall into the local minimum and the amount of calculation is usually large. More worrying yes, many optimization methods, especially the optimization method of derivation optimization based on face model equation is not continuous, discrete, unimodal and multimodal mathematical properties is becoming more and more powerless. Recently developed new optimization technique based on swarm intelligence computation parameters in various fields have been paid great attention in the intelligent optimization algorithm is a stochastic optimization algorithm based on bionics, the typical method of particle swarm optimization and Eberhart proposed by Kennedy (Particle Swarm Optimization PSO) algorithm, genetic algorithm (Ge Netic Algorithm, GA), differential evolution algorithm (Differential, Evolutionary, DE and Dorigo) of the ant colony algorithm. These methods are widely used in scientific research and solving practical problems, and made the traditional optimization methods can not replace the effectiveness of intelligent optimization algorithm. The class itself has very strong applicability, and continuous the objective function without any requirement, is not sensitive to the initial solution selection, so the intelligent optimization algorithm as a candidate algorithm for solving complex optimization problems is very realistic. Including the main problem to be solved in the paper: (1), in the fully familiar with the domestic and foreign research present situation, understand and master the basis of intelligence optimization and dynamic simulation of MATLAB/Simulink integrated environment, based on the material magnetization mechanism this paper puts forward a particle swarm optimization algorithm from the angle of Physics (PSO) combined with MATLAB/Simulink dynamic imitation It integrated environment Jiles-Atherton (JA) parameters of hysteresis loop model method. And the results of simulated data without noise and with noise, the three group of parameters in the JA model for numerical experiments. (2), this paper constructs a JA model equation based on Simulink module, by solving the model equations obtained B-H hysteresis loop accurately. And the Simulink module equation and the model of the JA algorithm to achieve a seamless integration, provide a guarantee for the smooth operation of the algorithm. (3), values of control variables in different algorithms to optimize accuracy, computing time and has great impact on convergence, therefore, for different problems, how to choose the optimal parameter configuration is a problem needing attention in this paper. (4), the results are compared with the genetic algorithm, the analysis shows that for Jiles-Atherton nonlinear hysteresis model of complicated calculation, PSO algorithm 鐜板嚭寰堝ソ鐨勯瞾媯掓，

本文編號：1610474