資源描述:
《基于模糊神經(jīng)網(wǎng)絡(luò)的熱風(fēng)回流爐溫度控制》由會員上傳分享,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。
1、摘要熱風(fēng)回流爐是表面貼裝工藝中的關(guān)鍵設(shè)備,回流爐溫度的穩(wěn)定性直接影響著產(chǎn)品的質(zhì)量與成品率。在熱風(fēng)回流爐的溫度控制過程中,被控參數(shù)具有時變、非線性、不確定因素等因素。為了提高系統(tǒng)的自適應(yīng)能力和抗干擾能力,本文使用了一種新型的控制器一模糊神經(jīng)網(wǎng)絡(luò)控制器,與常規(guī)PID控制器并聯(lián),形成并聯(lián)復(fù)合控制,實(shí)現(xiàn)了對熱風(fēng)回流爐溫度的精確控制,有效地克服了采用單純的PID控制帶來的響應(yīng)慢、超調(diào)大、控制穩(wěn)定性差等缺點(diǎn)。在設(shè)計的過程中,將模糊理論的知識表達(dá)容易和神經(jīng)網(wǎng)絡(luò)學(xué)習(xí)能力強(qiáng)這兩種優(yōu)勢結(jié)合起來,設(shè)計了一種具有模糊結(jié)構(gòu)的等價神經(jīng)網(wǎng)絡(luò),構(gòu)成一個新的網(wǎng)絡(luò)結(jié)構(gòu)(FuzzyNeuralNetworksControlFN
2、NC),解決了傳統(tǒng)模糊控制由于隸屬函數(shù)和模糊規(guī)則選取不當(dāng)造成的控制缺陷。它不僅具有清晰的空間結(jié)構(gòu),而且具有良好的自學(xué)能力和非線性逼近能力。考慮到熱風(fēng)回流爐溫度控制的實(shí)際特點(diǎn),使用了一種用于被控對象輸出預(yù)測的神經(jīng)網(wǎng)絡(luò)預(yù)測器(NeuralNetworksPredictorNNP),預(yù)測器通過對網(wǎng)絡(luò)的學(xué)習(xí),預(yù)測被控對象的未來輸出,使控制器預(yù)先感知系統(tǒng)輸出狀態(tài)的變化趨勢,從而做出相應(yīng)的調(diào)整。利用Matlab仿真軟件,建立Matlab仿真模型,分別對常規(guī)PID控制、模糊神經(jīng)網(wǎng)絡(luò)復(fù)合控制(PID+FNNC)、帶有神經(jīng)網(wǎng)絡(luò)預(yù)測器的模糊神經(jīng)網(wǎng)絡(luò)復(fù)合控制(PID+FNNC+NNP)進(jìn)行對比仿真實(shí)驗(yàn),仿真實(shí)驗(yàn)
3、結(jié)果表明:帶有神經(jīng)網(wǎng)絡(luò)預(yù)測器的模糊神經(jīng)網(wǎng)絡(luò)復(fù)合控制(PID+FNNC+NNP)的控制效果在三者中最優(yōu)。關(guān)鍵詞:熱風(fēng)回流爐溫度控制模糊控制模糊神經(jīng)網(wǎng)絡(luò)神經(jīng)網(wǎng)絡(luò)預(yù)測ABSTACTSirocco—circumflueneesolderingsystemistheimportantequipmentofSurfaceMountTechnology(SMT).n圯temperaturedynamic—characteristicofthesystemhasinfluencedthequalityofproductiondirectly.Inproduction,theprocedureofthesy
4、stemhasmanycharacterssuchastimevariable,nonlinearandindefinite.Inthepaper,aFuzzyNeuralNetworksController(FNNC)connectedparallelwi也generalPIDcontroHerhasbeenusedtoimprovetheadaptiveandanti-jammingability,avoidthegeneralPIDcontroldisfigurement,e.g.slowlyresponse.badcontrolstabilityandoverrn.1'l,andg
5、etgoodeffects.Fuzzylogichasvirtueofexpressingknowledgeeasily,neuralnetworkshasgoodself-learningability.Akindofnewcon血011e卜—_FuzzyNeuralNetworksController(RXrNC)hasbeendesignedbycombiningthevirtueofthesetwotechniques.Intraditionalfuzzycontrol,unsuitableselectedmembershipfunctionandfuzzyruleswillind
6、ucethecontroldisfigurement,buttheFuzzyNeuralNetworksControllerhasavoidedit.Ithasclearstructure,goodself-learningandnonlinearmapability.Consideredt11echaracteristicofSirocco—citeumfluencesolderingsystem.a(chǎn)NeuralNetworksPredictor(NNP)hasbeendesignedtopredicttheoutputofcontrolledsystem,thenthecontroll
7、erwillforecastthechange廿endofcontrolledsystem,andadjustthesysteminadvance.Atlast,establishthemodelofSirocco-cimumfluencesolderingsystemforsimulation誦t11Mauab,simulationisdoneforgeneralPIDcontrolsystem,F(xiàn)uzzyNeural