小波神經(jīng)網(wǎng)絡(luò)在系統(tǒng)辨識(shí)應(yīng)用中的結(jié)構(gòu)與算法研究

《小波神經(jīng)網(wǎng)絡(luò)在系統(tǒng)辨識(shí)應(yīng)用中的結(jié)構(gòu)與算法研究》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。

1、摘要作為久工餐娩靛羹簧緞戲部分，入工耱經(jīng)掰絡(luò)餐較大躺瘦箱潛鴦。本文在闡述^『耱經(jīng)黼終靜發(fā)震、褒狹、蒸本原理及其猩蠡渤羧鑭巾瓣斑嗣瓣蒸穡主，主要騷究了夸浚神綴瓣絡(luò)在寨統(tǒng)辨談盛愆孛豹縫穩(wěn)≮葵法。奪波神經(jīng)麗絡(luò)燕在奪波勢(shì)轎理論靜萋磁上撬毒瓣，是奪渡變換秘襻經(jīng)閼終鴦穰結(jié)會(huì)豹產(chǎn)秘。巍予夸沒章棗綴弼終每警遴熬藩矮闞絡(luò)穗魄，暴露瓣絡(luò)縫擒秘參數(shù)豹確定番臻埝祓據(jù)獲駿斂蘸度捩、耱度離譬筑贏，囂魏零波褲經(jīng)瓣絡(luò)凌逶援予系統(tǒng)辨談瓣其露獨(dú)特魏貔勢(shì)。本文詳鑭奔綏了小波分掇瓔論翻，l、波專棗艇瘸終鶼基破知識(shí)，勢(shì)褥?！畚厚蛄髀秺Z波耱經(jīng)網(wǎng)絡(luò)攀霹霪滾豹傀缺熹，系絞縫分綏了系絞辨談爨孽一

2、般聯(lián)浚驤及糖變系綾辨識(shí)煞轉(zhuǎn)統(tǒng)方法疑職究瑗捩，靜農(nóng)魏藻皴主捉爨了基爭(zhēng)夸渡孛牽縫鬻絡(luò)豹鬟統(tǒng)辮識(shí)囂豹一般績(jī)棱。小波享睪緞翅終縫糗貔確定瓣蘧，即如鍵疆定隱鼷攤經(jīng)囂懿個(gè)數(shù)，一直楚冀疆究串熬一今臻點(diǎn)

3、穰美蕤。零文贊瓣疆土瓣趣，在分掇了懿人研究皴袋熬罄皴上，受Pati秘Krishnaprcsad關(guān)于三個(gè)Sigmoid函數(shù)憋絞性綴會(huì)可以作淹小波纂函數(shù)豹聯(lián)論豹瘺發(fā)，捷逡了一耪確定夸渡聯(lián)絡(luò)績(jī)擒越瓤方法，麓小波翹絡(luò)豫聯(lián)毒牽綴元數(shù)露憋確定提供了～個(gè)攢途徑。其霖理是合理選擇小波基函數(shù)序列中的能夠覆箍被遙近函數(shù)整個(gè)避頻囂域馳一個(gè)元素域幾個(gè)囂素辯魏合豫為瓣終豹纂函數(shù)，以姥米確

4、定隱層章唪經(jīng)囂的個(gè)數(shù)，從兩確定小波嘲絡(luò)的結(jié)梅。本文從理論方瑟鼯冀法避褥了潦入豹剿爨，介纓了算法酌詳纓憋臻秘冀髂過載，勢(shì)將簿法訓(xùn)練最的小波神經(jīng)嘲絡(luò)運(yùn)用剃災(zāi)瓣的囂線憾系統(tǒng)的辨識(shí)過程中去。幾個(gè)典濺鰓系統(tǒng)瓣識(shí)彷糞災(zāi)陵襲明該算法輿有羰踩精度箍鞠計(jì)舞越便靜良好性能。關(guān)鍵詞：人工搿髓、小波神經(jīng)鬻絡(luò)、系統(tǒng)辨識(shí)、聰數(shù)邋旋、辯頻囂域、非線性時(shí)變系統(tǒng)AbstractArtificialneuralnetwork(ANN)，asallimportantpartofartificialintelligence，hasgreatpotenceinapplication．Af

5、terintroducingthedevelopment，statusquo，basictheoryofneuralnetworkanditsapplicationtoautomaticcontrol，thisthesismainlystudiesthestructuresandalgorithmsofwaveletneuralnetworks(wN駒anditsapplicationtosystemidentification．WNN，basedonthetheoryofwaveletanalyses，istheproductionofcomb

6、iningwavelettransformswithneuralnetworks．Boc遍nseoftheiradvantagesofrapidconvergenceandhightrackingaccuracy,WNNhavemuchparticularadvantageswhenappliedinsystemidentification．ThispaperstudiesthetheoriesofWNNandwavelettransforms，analyzestheadvantagesanddisadvantagesofseveralpopul

7、artrainingalgorithms，introducesthemainprincipleofsystemidentificationandtraditionalmethodsandstatusquooftheidentificationoftime-varyingsystem，andthenbringforwardthestructureofsystemidentificationwhichisbasedonWNN．Howtodeterminethereasonableamountofneuronsinhiddenlayerisacruci

8、alanddifficultprobleminthestudiesofWNN．AfteranalyzingthetheoryofPatiandKrishnapresad，thatthelinearcombinationofthreesigmoidfunctionscanbetheradicalfunctionofwavelet，thepaper嘶ngforwardanewalgorithmtOsolvetheproblem．Theprincipleofthisalgorithmisthatinthesequencesofwaveletfuncti

9、onsoneOrsomeelements，whichcancoverthewholetime-frequencyfieldofthegi