資源描述:
《基于神經(jīng)網(wǎng)絡(luò)的自適應(yīng)逆控制方法研究》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、摘要由B.Widrow教授提出的自適應(yīng)逆控制方法經(jīng)過十多年的發(fā)展,已經(jīng)取得了很多成果。然而,目前針對(duì)自適應(yīng)逆控制的研究,還主要以線性系統(tǒng)為主,針對(duì)非線性系統(tǒng)的研究成果還不多見,迫切需要研究者對(duì)非線性系統(tǒng)的自適應(yīng)逆控制加以研究。當(dāng)前神經(jīng)網(wǎng)絡(luò)、模糊和神經(jīng)一模糊融合理論系統(tǒng)等的迅猛發(fā)展,給各種非線性系統(tǒng)的研究提供了有力的工具,本課題主要研究利用神經(jīng)網(wǎng)絡(luò)實(shí)現(xiàn)非線性自適應(yīng)逆控制的問題,具體工作如下:首先,針對(duì)多層前向網(wǎng)絡(luò)作為逆控制器學(xué)習(xí)收斂速度慢,易陷入局部極小的問題,研究了基于正交神經(jīng)網(wǎng)絡(luò)的直接非線性自適應(yīng)
2、逆控制,通過修改網(wǎng)絡(luò)隱層s函數(shù)的傾斜度,加快了神經(jīng)網(wǎng)絡(luò)的學(xué)習(xí)過程,并設(shè)計(jì)了無(wú)刷直流電機(jī)直接自適應(yīng)逆控制系統(tǒng),實(shí)現(xiàn)了高性能速度控制。其次,論文研究了基于模糊RBF神經(jīng)網(wǎng)絡(luò)的系統(tǒng)建模、逆建模方法,并且針對(duì)熱工系統(tǒng)純延遲對(duì)象,設(shè)計(jì)了相應(yīng)的模型參考自適應(yīng)逆控制系統(tǒng)。被控對(duì)象能快速跟蹤參考模型的輸出,有效的克服擾動(dòng),適應(yīng)環(huán)境及參數(shù)的變化。再次,由于動(dòng)態(tài)遞歸神經(jīng)網(wǎng)絡(luò)相比與靜態(tài)神經(jīng)網(wǎng)絡(luò)的優(yōu)越性,論文討論了輸出—輸入反饋Elmaa網(wǎng)絡(luò)應(yīng)用于非線性自適應(yīng)逆控制的可行性,并設(shè)計(jì)了隱層為徑向基函數(shù)的Elman網(wǎng)絡(luò)自適應(yīng)逆
3、控制系統(tǒng)。針對(duì)不同的非線性對(duì)象,研究了直接逆控制及存在擾動(dòng)情況下的自適應(yīng)擾動(dòng)消除問題。最后,針對(duì)船舶可調(diào)螺距螺旋槳控制系統(tǒng)中,傳統(tǒng)PID抗干擾能力差、難以獲得最佳整定效果的缺點(diǎn),論文研究了基于遺傳優(yōu)化的非線性PID方法進(jìn)行控制,引入遺傳算法對(duì)系統(tǒng)控制器各部分參數(shù)進(jìn)行優(yōu)化,取得了滿意的效果。關(guān)鍵詞自適應(yīng)逆控制;非線性;正交網(wǎng)絡(luò);RBF模糊神經(jīng)網(wǎng)絡(luò);Elman網(wǎng)絡(luò):船舶調(diào)距槳燕山大學(xué)_T學(xué)碩士學(xué)位論文AbstractAdaptiveinversecontrol,whichwasfirstpresente
4、dbyprofessorWidrow,hasgotlargedevelopmentinthepastdecade,manyresuhshavebeenachievedinthestudyofadaptiveinversecontr01.However,uptonow,thestudyofadaptiveinversecontrolfocusesonlinearsystems.Relatively,studyandresultsofnonlinearsystemsfirerare.Thus,it’Sn
5、ecessaryfortheresearcherstodosomefurtherworkinthestudyofnonlinearadaptiveinversecontr01.Theadvancesinneuralnetworksandfuzzysystemsprovidetheresearcherswithpowerfultoolsfornonlinearsystems.Thisthesisaimstoachievenonlinearadaptiveinversecontrolwiththeset
6、ools.IdidthefollowingWOrk:Firstofall,becauseslowlyconvergenceandeasilystopatthelocalminimumoftheinversecontrollerconstructedbymulti-layerneuralnetwork,thepaperconstructanadaptiveinversecontrollerbyorthogonalneuralnetwork.ThroughmodifytheparametersofS-f
7、unctioninthehiddenlayer,whichacceleratethespeedofthelearning.AndthendesignthedirectinversecontrolsystemforBDCM,iteasilygetthehighperformanceinspeedcontr01.Secondly,thepaperstudiesthemodelingandinversemodelingbyfuzz)?RBFneuralnetworks;Anddesignmodelrefe
8、renceadaptiveinversecontrolsystemforthermotechnicalautomaticcontrol,thecontroUerperformswellandCanheeasilyaccomplishedon-line.Thirdly,becausetheadvantageofdynammrecurrentNN,thepaperdiscussedthefeasibilityforOIFElmanneuralnetworkusedinad