發(fā)布時(shí)間：2018-02-14 14:28

  本文關(guān)鍵詞： 微波腔體帶通濾波器 優(yōu)化算法 神經(jīng)網(wǎng)絡(luò) 遺傳算法 擬牛頓算法 極限學(xué)習(xí)機(jī)　出處：《華北電力大學(xué)》2015年碩士論文　論文類型：學(xué)位論文

【摘要】：信號(hào)間的頻帶隨著通信技術(shù)的發(fā)展變得越來越窄,為了抑制各種干擾信號(hào)和降低系統(tǒng)對(duì)信號(hào)的衰減,國內(nèi)外學(xué)者對(duì)微波濾波器的設(shè)計(jì)已經(jīng)做了大量的研究工作。微波腔體濾波器由于其具有大功率、低損耗、高抑制、結(jié)構(gòu)緊湊、性能穩(wěn)定等優(yōu)點(diǎn),近些年被廣泛地應(yīng)用在通信系統(tǒng)中。同時(shí),隨著計(jì)算機(jī)技術(shù)的高速發(fā)展,高頻電磁仿真軟件與濾波器設(shè)計(jì)的結(jié)合使得濾波器的設(shè)計(jì)精度和速度都得到了很大的改善,大大縮短了設(shè)計(jì)周期。但是對(duì)于結(jié)構(gòu)復(fù)雜的微波濾波器來說,由于其設(shè)計(jì)的參數(shù)較多,電磁仿真軟件很難在預(yù)期的時(shí)間內(nèi)達(dá)到設(shè)計(jì)所要求的精度,有時(shí)甚至還會(huì)出現(xiàn)難以收斂的現(xiàn)象。為解決上述問題,本文將改進(jìn)的神經(jīng)網(wǎng)絡(luò)算法和極限學(xué)習(xí)機(jī)算法分別引入到微波腔體濾波器的設(shè)計(jì)當(dāng)中,在Matlab環(huán)境下實(shí)現(xiàn)了微波腔體濾波器HFSS(High Frequency Structure Simulator)的快速建模,一定程度上提高了濾波器的設(shè)計(jì)效率,縮短了設(shè)計(jì)周期。首先,研究了帶通濾波器的基本理論,以及同軸腔體帶通濾波器的諧振腔之間的耦合結(jié)構(gòu)與耦合端口的設(shè)計(jì)；隨后分析并研究了用神經(jīng)網(wǎng)絡(luò)優(yōu)化微波腔體帶通濾波器結(jié)構(gòu)參數(shù)的方法,提出了遺傳神經(jīng)網(wǎng)絡(luò)(Genetic Neural Networks, GA-BP)和并行擬牛頓神經(jīng)網(wǎng)絡(luò)(Parallel Quasi-Newton Neural Networks, PQN-BP)算法,并將這兩種優(yōu)化算法分別用在微波腔體濾波器的結(jié)構(gòu)參數(shù)優(yōu)化中；接著分析介紹了極限學(xué)習(xí)機(jī)(Extreme Learning Machine, ELM)、核函數(shù)、量子粒子群(Quantum particle Swarm Optimization Algorithm, QPSO)的相關(guān)理論,首次將小波核極限學(xué)習(xí)機(jī)(Wavelet Kernel Extreme Learning Machine, WKELM)方法引入到微波濾波器的結(jié)構(gòu)參數(shù)優(yōu)化領(lǐng)域中,對(duì)其進(jìn)行改進(jìn)后,給出了一種基于量子粒子群的小波核極限學(xué)習(xí)機(jī)的優(yōu)化方案。最后,將三種算法進(jìn)行對(duì)比分析后,得出本文研究的三種方法均能有效的提高收斂速度,克服傳統(tǒng)方法易陷入局部最小等缺點(diǎn),但量子粒子群小波核極限學(xué)習(xí)機(jī)方法(QPSO-WKELM)效果最好。因此,運(yùn)用量子粒子群小波核極限學(xué)習(xí)機(jī)方法設(shè)計(jì)了一款七腔帶通濾波器,仿真結(jié)果表明,該算法能夠?qū)崿F(xiàn)對(duì)微波濾波器結(jié)構(gòu)參數(shù)的快速精準(zhǔn)的設(shè)計(jì),與傳統(tǒng)的濾波器建模技術(shù)相比在精度和速度上都有明顯的改善。
[Abstract]:The signal between the band with the development of communication technology has become more and more narrow, in order to suppress the interference signals and reduce the attenuation of the signal system, the design of microwave filters by domestic and foreign scholars have done a lot of research work. The microwave cavity filter because of its high power, low loss, high rejection, compact structure, stable performance in recent years, is widely used in the communication system. At the same time, with the rapid development of computer technology, combined with the high frequency electromagnetic simulation software and the filter design has greatly improved the accuracy and speed of filter design, shorten the design cycle. But for the complicated structure of the microwave filter, due to its design parameters the electromagnetic simulation software is difficult to meet the design requirements of the precision within the expected time, sometimes even difficult for the solution of convergence phenomenon. To solve these problems, the improved neural network algorithm and extreme learning machine algorithm was introduced into the design of microwave cavity filter, Matlab environment in the realization of a microwave cavity filter (High HFSS Frequency Structure Simulator) the rapid modeling, improves the efficiency of filter design, shorten the design cycle. First of all, research the basic theory of bandpass filter, and the design of the coupling structure and the coupling between the coaxial cavity resonator port of the band-pass filter; then analyzed and studied with method of structural parameters of filter using neural network to optimize the micro wave cavity, proposed a genetic neural network (Genetic Neural Networks, GA-BP) and parallel quasi Newton neural network (Parallel Quasi-Newton Neural Networks PQN-BP) algorithm, and the structure of the two kinds of optimization algorithms were used in microwave cavity filter Parameter optimization analysis; then introduces the extreme learning machine (Extreme Learning Machine, ELM), kernel function, quantum particle swarm (Quantum particle Swarm Optimization Algorithm, QPSO) related to the theory, for the first time the wavelet kernel extreme learning machine (Wavelet Kernel Extreme Learning Machine, WKELM) method is introduced to optimize the field structure parameters of microwave filter in the improved, a wavelet kernel extreme learning machine based on quantum particle swarm optimization scheme is given. Finally, the three algorithms are compared. The three methods in this paper can improve the convergence speed effectively, overcome the disadvantages of traditional method is easy to fall into local minimum points. But quantum particle swarm wavelet kernel extreme learning machine method (QPSO-WKELM) had the best effect. Therefore, the design of a seven cavity band-pass filter quantum particle swarm wavelet kernel extreme learning machine method The simulation results show that the algorithm can achieve the fast and accurate design of the microwave filter's structural parameters, and has obvious improvement in accuracy and speed compared with the traditional filter modeling technology.

【學(xué)位授予單位】：華北電力大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：TN713.5

【相似文獻(xiàn)】

中國期刊全文數(shù)據(jù)庫 前10條

1 趙昭;劉利林;張承學(xué);李海濤;汪建敏;;形態(tài)學(xué)濾波器結(jié)構(gòu)元素選取原則研究與分析[J];電力系統(tǒng)保護(hù)與控制;2009年14期

2 羅勇;何曉峰;吳文啟;羅兵;;基于雙濾波器結(jié)構(gòu)的基帶信號(hào)預(yù)處理濾波模型[J];中國慣性技術(shù)學(xué)報(bào);2011年05期

3 李普,楊華,方玉明;電信系統(tǒng)中的硅微機(jī)械濾波器結(jié)構(gòu)魯棒設(shè)計(jì)[J];機(jī)械工程學(xué)報(bào);2005年05期

4 王繼山,李揮;一種快速去塊濾波器結(jié)構(gòu)[J];微電子學(xué)與計(jì)算機(jī);2005年11期

5 周金元;王琳凱;汪文義;周曉方;;小數(shù)采樣率轉(zhuǎn)換中的μ-發(fā)生器設(shè)計(jì)[J];計(jì)算機(jī)工程;2011年19期

6 李愛社;油田注水系統(tǒng)壓力脈動(dòng)濾波器結(jié)構(gòu)參數(shù)仿真分析[J];機(jī)床與液壓;1998年02期

7 肖高標(biāo)，吳杰，洪瀾;一種電流模式多功能濾波器結(jié)構(gòu)[J];湖南大學(xué)學(xué)報(bào)(自然科學(xué)版);1996年04期

8 黃海泉;許愛國;謝少軍;;一種級(jí)聯(lián)逆變器的分散式濾波器結(jié)構(gòu)研究[J];電源世界;2008年09期

9 趙春暉,孫銳,孫圣和;形態(tài)濾波器結(jié)構(gòu)元的自適應(yīng)優(yōu)化算法[J];電子與信息學(xué)報(bào);2002年03期

10 陳斌,周震,黃永清,任曉敏;InP基微機(jī)械可調(diào)濾波器結(jié)構(gòu)特性分析[J];半導(dǎo)體光電;2004年05期

中國重要會(huì)議論文全文數(shù)據(jù)庫 前3條

1 趙昭;陳勇;汪海燕;;形態(tài)學(xué)濾波器結(jié)構(gòu)元素選取原則研究[A];中國高等學(xué)校電力系統(tǒng)及其自動(dòng)化專業(yè)第二十四屆學(xué)術(shù)年會(huì)論文集（中冊(cè)）[C];2008年

2 張敬彬;戚金清;王兢;吳浩;;一種高吞吐量H.264/AVC去方塊濾波器結(jié)構(gòu)設(shè)計(jì)[A];第八屆全國信息獲取與處理學(xué)術(shù)會(huì)議論文集[C];2010年

3 姚海瑞;鄒積勇;吳新兵;楊志;;大容量并聯(lián)有源電力濾波器結(jié)構(gòu)對(duì)比分析[A];分布式發(fā)電、智能微電網(wǎng)與電能質(zhì)量——第三屆全國電能質(zhì)量學(xué)術(shù)會(huì)議暨電能質(zhì)量行業(yè)發(fā)展論壇論文集[C];2013年

中國博士學(xué)位論文全文數(shù)據(jù)庫 前1條

1 殷瑞祥;DCT快速新算法及濾波器結(jié)構(gòu)研究與子波變換域圖像降噪研究[D];華南理工大學(xué);2000年

中國碩士學(xué)位論文全文數(shù)據(jù)庫 前2條

1 王雪;優(yōu)化算法在微波腔體濾波器結(jié)構(gòu)參數(shù)中的應(yīng)用研究[D];華北電力大學(xué);2015年

2 劉輝;微環(huán)諧振濾波器結(jié)構(gòu)設(shè)計(jì)與傳輸特性研究[D];南京郵電大學(xué);2013年

，

本文編號(hào)：1510914