畢業(yè)論文外文翻譯-并條機自調勻整利用人工神經(jīng)網(wǎng)絡確定在自調勻整作用點(外文原文+中文翻譯)

ResearchseofArtificialNeuralNetworkstheActionPointatAuto-levelingDrawAssadFarooq1andChokriofTextileClothingTechnology,TechnischeUniversit?tDresden,Abstract

ArticleArtificialneuralnetworkswithabilityoffromhavebeenappliedinthetextileThepointisofimportantparametersofthedrawingqualityofreportsamethodoftheactionusingartificialnetworks.actionvariableswereselectedasfortrainingtheartificialneuralnetworkswiththeaimtooptimizeauto-levelingbylimitingthelevelingactionpointsearchrange.TheMarquardtalgorithmisincorporatedintotheback-propagationtoacceleratetheBayesianregularizationistheofnetworks.arequitepromising.Keyartificialauto-lev-eling,frame,levelingactionpoint。Theoftheanroleinwhilesliverisofthecriticalwhenqualityyarn.ismajorcriteriaforassessmentofoperationofInprinciple,twotoirregularities.istothedraftingandrecognizetheforirregularities,thatmeansbethem.Thevaluableistouseauto-levelersintheinadequatetocorrectthevariationsinsliver.Thecontrolofirregularitiesloweronsliveruniformity,ambientframeparameters.Atthedrawframevariationsinarecontinuallybyamechanicaldevice(aroll)convertedintoelectricalThetoanelectronicwithvariable,timedelayedThetimeallowsdraftbetweenthemid-rollthedeliveryrollofdrawtoadjustthatmomentwhendefectivesliverpiece,whichhadmeasuredbyapairofscanningfindsitselfatapointofAtthispoint,dependingupontheamountofinsliverpiece.ThedistancethatscanningpairofdraftcalledthepointofregulationthelevelingactionpointinFigure1.Thisleadstocalculatedthecorrespondingdefective[2,3].Inframes,especiallyintheofoffibermaterial,orbatchesmachineparametersmustbeTheLAPismostauto-levelingparameterwhichisinfluencedbyparametersspeed,material,draftgauge,draftgauge,feedingtension,draft,andsettingtheetc.UseofArtificialNeuralfortheActionPointA.C.1ofanPreviously,sampleshadbewithsettings,tothelaboratory,onthetesteroptimumLAPwasdrawautomaticsearchfunctionforoptimumofLAP.Duringthisfunction,isscannedbyLAPcess,thequalityparametersareconstantlyalgorithmautomaticallycalculatestheoptimumLAPbythepointwiththeminimumsliverCV%.Atpresentasearchof120mmscanned,21examinedusing100ofineach2100ofisnecessarysearchatime-consumingmethodaccompaniedbythehencedirectlyaffectingcostparameters.Inthiswork,wefindthepossibilityofLAP,artificialneurallimitautomaticsearchandtheabove-mentioneddisadvantages.ArtificialNeuralNetworksThemotivationofusingartificialneuralnetworksliesintheirflexibilityofinformationnothave.Theneuralnetworksystemcansolveaproblem“byexperienceandlearning”theinputoutputpatternsprovidedbytheuser.Inthefieldofartificialneuralnetworks(mostlyhavebeenextensivelystudiedthelasttwodecades[4–6].thefieldofpreviousonpredictingpropertiesandspinningprocessusingthefiberpropertiesacombinationoffiberpropertiesandmachinesettingsasofneuralnetworks[7–12].Back-propagationissupervisedmostfrequentlyforneuraltraining.back-propagationalgorithmbasedtheWidrow-Hoffdeltaruleinwhichweightoutthroughthemeanofoutputtoinput[13].ThesetpatternstheuntiltheTheusessteepestdescentamethodtodetermineofOverfittingTheofneuralnetworktrainingproduceanetworkproducesontrainingwhichalsorespondsproperlynovelinputs.WhennetworkwelloninputsasonsetthesaidtobeThecapacitytheislargelybythe(numberofhiddenneurons)thisvitalrolethetraining.Anetworkwhichnottoalltheinformationintheisbeunderfitted,whileathatiscomplextofitthe“noise”inthedataleadstooverfitting.“Noise”meansvariattheunpredictablefrominputsofspecificAllstandardneuralarchitecturesasfullytooverfitting.itisdifficultacquirethenoisefromspinningdependenceofonthevariationsandenvironmentalconditions,etc.stoppingcommonlyusedtotacklethisproblem.Thisinvolvesofdataintothreei.e.atrainingavalidationsetandwiththatlargepartofthedata(validationset)canneverbetheofTheotherofoverfittingwhichthemethodofimprovingconstrainingofnetworkweights.[14]discussedBayesianforback-propagationnetworks,whichconsistentlyproducedwithgoodTheobjectiveofprocessismini-mizetheoferrors:n(t)Diii

2

（1）Wheretthetargetsandaresponsestheiitoreduceofsquarederrorsregularizationadditionalterm,theD

（2）Inequation

E

isthesumofsquaresofthenetworkweights,andαandwfunctionparameters.relativeofobjectivefunctionparametersdictatestheemphasisfortraining.Ifα<<β,thenthetrainingwilldrivetheerrorssmaller.α>>β,trainingwillemphasizeweightsizereductionattheexpenseofnetworksmootherresponse[15].Theofisonviewwhatittofromdata,inwhichusedtorepresenttheuncertaintythebeinglearned.anythepriorwhatmightbecanbeindistributionthatdefineAfterthedata,therevisedarebyaposteriordistributionoverweights.Networkthatplausiblebutwhichthewillbebeingmuchwhiletheforvaluesofweightsdofitthedatawillhaveincreased[16].InframeworktheweightsofareconsideredAfterthetaken,thefunctionfortheweightsbeupdatedaccordingtoBayes’rule:(w/D,

M)

(D/,)(w/(D/M)

M)

(3)InequationDdataset,Mneuralnetworkwistheofweights.P(/,M)isthepriorprobability,whichrepresentsourofbeforeanydatais(D/,

M)whichprobabilityofdataoccurring,giventheweightsw.Pw

)isathatthetotal1[15].Inthisstudy,employedMATLABNeuralNet-worksToolbox“trainbr”whichisanincorporationoftheLevenbergMarqaurdtregularizationtheorem(orBayesianintotrainthenetworktoreduceofapproximationofHessianmatrixproducegoodcapabilities.Thisaofparametersbiases)effectivelyusedbytheofshouldthesame,irrespectiveoftotalofparametersinThiseliminatesrequireddeterminingthenetworkExperimentalTheexperimentaldatawasobtainedfromRieter,ofdrawframeRSB-D40[17].Forthematerialexperimentalwasbasedofuseinspinningindustry.Forexample,isthefrequentlyusedmaterial,soitwereperformedoncottonwithallsettings,wasnotcasewithotherAlso,owingtothefactthatallcouldnotbeprocessedwithsamepressuredraftdifferentweredesigned.Thewiththeirplansgivenin1.Theprocedurewasappliedtomaterialsandstandardforsettings(sliverlineardensity,LAP,wasadopted.AcomparisonofautomaticwasperformedandresultswerethesearchfunctionfromRSBD-40.ThereforetheLAPsearcheswereaccomplishedbyQualityMonitorabstractoftheinFigure2.UseofArtificialNeuralfortheActionPointA.C.2Abstractnetworkmodel.HeretobeconsideredisthatisnoinmachinetheLAPinfluencingparameter,i.e.feedingSofeedingspeedconsideredberelatedspeedofdoublings(4).Thedeliveryspeedvariedbetween3001100thedoublings7,toachieveofCount×speed/（Doublings×Feed）（）TrainingandTestingSetsFortrainingtheneuralnetwork,thedataintofirstphaseincludedexperimentaldataforcompilationthedatasubsequentThepriortheinfluencingi.e.speed,deliveryspeed,gaugesofdraft,andofthesliverwastoselecttheSofirstphasecontainedtheinwhichstandardweretakenasafoundationandthenLAPinfluencingineachInsecondphase,experimentswereselectedinwhichinfluencingparameterandnetworkwasallowedtotheThiswasontheofinfluencingparameters,withtheaimtoLAPlength.thirdphaseinvolvedtheexperimentsconductedonpilotscaleThesepilotwerecarriedbymachinefordifferentsettings.theseresultsweretoassesstheoftheneuraldataNormalizingtargetvariablesmaketrainingimprovingnumericalconditiontheproblem.Alsoitcanreducechancesofinforneuralnetworkofthelargespansofnetwork-inputdata,targetsforbetterperformance.Atfirsttheinputswerebetweeninterval[whichnotanypromisingresults.databetweentheinterval1]networkswithsuccess.NeuralNetworkTrainingWefiveneuralnetworkstopredictLAPbyofdataThesetsdividedintotrainingasinwasperformedtrainingweretotheofneuralinformFigure3thetrainingperformanceofthenetwork5.3TrainingperformanceofNNResultsandDiscussionalreadyinTable2,differentofdatawereusedtotrainfivenetworksatheirtestonunseendata.Asistoeliminatetheguessfortheappropriateofhiddentheofhiddenrangingfrom1422,wereintwolayerstrainnetworks.theselectionoftopologythatshouldbecomplextolearnalltheintheaspossibilityofwastackledwithduringtrainingthetestsetsbeseenfromFigures44ofNN1.UseofArtificialNeuralfortheActionPointC.5ofNN2.6ofNN3.7ofNN4.8ofNN5.wascalculatedtopredictedactualvalues,bothduringtrainingandtesting,andthearepresentedinFigure9.overallresultscanbeexplainedthepresenceorof–outputinteractionsindataandtheincreaseinthewiththeoftrainingdatasets.Figure9clarifiestheincreaseintestingerror,evenwithanincreaseintheoftraining,sets,whentrainingandperformedonfromdifferentphases(NN2&2).theerroraofphasewastrainremainderusedtesting,inNN3and5comparisonwithNN2and4,Thepresenceofdifferentinputoutputinteractionsindifferentphasesthistrend.behaviorofNN1withtoabove-mentionedfactattributedtosetsinphase1.9betweentrainingtestsets.InordertoassesstheoffitofNN5,10-foldcross-validationpreformed,offorandfortesting,timestestingnetworktimeon10%ofunseenAnaverageR

2

reported.wasadoptedfor80%andtestthecalculatedof

2

wasintheperformancethe80%datasetsforconfirmaveryfitofNN5ConclusionTheartificialnetworkmodelwasthenetworkstrainedatofTextileClothingTechnology,TechnischeUniversitatDresden.TheuseofBayesianregularizationtestingforpracticalapplicationspromisingresults.testingofNN5asin3,aofabout2mmisobservable,whichfallswellwithinmmnegligiblerangedeterminationofLAP.Thisthatcanbeinthefutureforquickcomputationofthewiththeadvantagesofofmaterialandtime.TheaccuracyincomputationcanleadtoCV%quality.TestofNN紡織研究期刊并條機自調勻整利用人工神經(jīng)網(wǎng)絡確定在自調勻整作用點摘要用他們的人工神經(jīng)網(wǎng)絡從數(shù)據(jù)中學習的能力已經(jīng)成功的應用在紡織行業(yè)，在并條機自調勻整參數(shù)中勻整作用點是一個重要的參數(shù)且強烈的影響到紗線的質量本文主要是講述的是用人工神經(jīng)網(wǎng)絡來檢測勻整作用點各種各樣的勻整作用點會影響導致不同的變量，我們把這些變量輸入到人工神經(jīng)網(wǎng)絡進行測試，目的是為了通過限制勻整作用點改變范圍來優(yōu)化并條機的自調勻整，Levenberg-Marquardt算法是納入反向傳播加速系和貝葉斯正則化方法,進了網(wǎng)絡的泛化。得到的結果是非常的精確的。關鍵詞

人工神經(jīng)網(wǎng)絡自調勻整并條機勻整作用點紗線的均勻度在紡織行業(yè)中發(fā)揮著越來越重要的角,

而棉條均勻度是產(chǎn)生高質量紗線的重要因素之一。棉條均勻度也是并條機運轉的主要評價標準。在原則上,有兩種方法可以減少棉條不均勻。一種途徑是研究牽伸機構和認識棉條不均的原因,以這意味著可能會發(fā)現(xiàn)可以降低不均勻度的方法。其他更有有效的方法是使用自調勻整,因為在大多數(shù)情況下并條機是不足以正確的改變紗線的均勻度。長片段不勻控制在一個較低水平，取決于梳棉棉條均勻性、環(huán)境條件、和幀參數(shù)。在自調勻整并條機（RSB-40喂入的棉條厚度的不斷變化變化可以通過一種機械檢測設備來監(jiān)測其變化并且將監(jiān)測到的信號轉換為電子信號量數(shù)據(jù)被傳輸?shù)揭粋€可變電子記憶器里時間延遲響應當棉條片段有缺陷的時候可以在一定的時間延遲允許并條機牽伸的中間羅拉和傳送羅拉之間來進行棉條的不均勻度精確調整當已被一對檢測羅拉測量后發(fā)現(xiàn)在本身的一個牽伸點在這一點上伺服電機運行取決于在棉條部分被檢測的數(shù)量的變化一對分離的檢測羅拉和牽伸點之間的距離，稱為零點調節(jié)或勻整作用點LAP），如下1所示，這在計算相應有缺陷的材料應該進行相應的修正2、3）,在并條機的自調勻整裝置上,特別是在纖維材料改變的情況下或者是各種不同型號機器選擇和過程控制參數(shù)必須優(yōu)化整作用點是自調勻整中最重要的參數(shù)許多的因素影響，比喻喂條速度，棉條材料，前牽伸測量計，喂入張緊力，前牽伸和棉條導條羅拉設置等。insliver:均勻度變化的棉條

Draft主要牽伸區(qū)Rollers:檢測羅拉BreakDraft前牽伸區(qū)PressuereBar:壓力棒actionpoint勻整作用點此前，棉條樣品必須在實驗室在不同的條件下進行，在測試儀上進行檢測棉條均勻度，直到找到最好的勻整作用點（手動進行檢測RSB-40并條機的自調勻整具有自動搜索確定最好的勻整作用點這一功能，棉條在進行此功能期間，棉條被自動檢測是通過調整臨時的勻整作用點和被記錄的有用的數(shù)據(jù)此過程中，不斷的監(jiān)測的質量參數(shù)和通過選擇棉條的最低點的棉條的V%，通過一種算法自動計算出最佳的自調勻整作用點

，

目前的一個搜索檢測范圍是120米，例如在每檢查100m的棉條是21點，因此2100m的棉條必須進行搜索功能，這是一種比較費時的方法材料和生產(chǎn)的損失會直接影響成本在這項工作中我們試圖找出一種能夠預測出勻整的作用點利用人工神經(jīng)網(wǎng)絡限制自動搜索范圍和減少上述缺點。人工神經(jīng)網(wǎng)絡利用人工神經(jīng)網(wǎng)絡的目的在于他們的靈活性和信息處理能力，傳統(tǒng)的計算方法不具備。神經(jīng)網(wǎng)絡系統(tǒng)可以由用戶提供的輸入輸出模式通過“經(jīng)驗和學習”這種方式來解決問題，在紡織品領域，人工神經(jīng)網(wǎng)絡主要使用反向傳播)在過去20年已經(jīng)在廣泛的研究紡織領域之前的研究領域一直專注于預測紗線性能和紡紗纖維工藝使用性能或結合纖維的性能和機器型號作為的人工神經(jīng)網(wǎng)絡的輸入。反向傳播是一種監(jiān)督學習技術，頻繁用于人工神經(jīng)網(wǎng)絡的訓練。反向傳播算法是基于Widrow-的霍夫三角洲學習法則，其中權重是通過樣輸出到樣品輸入之間的響應的平方差進行調整。這系列單一模式反復提供給網(wǎng)絡,直到誤差值最小化差反向傳播算法采用最快下降法這實質上是一個一階的方法確定一個合適的梯度運動方向。過度擬合神經(jīng)網(wǎng)絡的訓練的目標是產(chǎn)生一個網(wǎng)絡這個開發(fā)的網(wǎng)絡上產(chǎn)生較小的錯誤,并且這種情況依然可能會反映出的是一個異常的輸入。當一種網(wǎng)絡模式異常的輸入作為研發(fā)系統(tǒng)的輸入,這種網(wǎng)絡被認為是廣義的網(wǎng)絡。很大程度上是受到網(wǎng)絡的泛化能力的網(wǎng)絡架構(一些隱藏的神經(jīng)元這在開發(fā)過程中起著至關重要的作用,而網(wǎng)絡是太復雜的，不足以來收集的所有信息數(shù)據(jù)，這被認為是不充分擬合,而當一個網(wǎng)絡是過于復雜,那么很難適應與紊亂的數(shù)據(jù)將會導致過度擬合噪音”的意思是從一個特定的網(wǎng)絡中的輸入值中變化的目標值是預測不到的所有的標準例如完全連接的多層感應器神經(jīng)網(wǎng)絡結構容易出現(xiàn)過擬合此外從紡織行業(yè)它是很難獲得沒有紊亂的數(shù)據(jù)由于最終產(chǎn)品依賴于固有材料變化和環(huán)境條件等問題解決這個問題最普遍的方法是阻止這些問題的出現(xiàn)這包括訓練的數(shù)據(jù)劃分成的三個組,即一個訓練集驗證集和測試集,這其中有個缺點是一個大量的數(shù)據(jù)部分(驗證系統(tǒng))不可能是訓練的部分。正規(guī)化另一個過擬合解決方案是正規(guī)化這是提高泛化通過約束網(wǎng)絡的權值的大小的方法。麥凱討論了一個實用的貝葉斯框架網(wǎng)——反向傳播網(wǎng)絡連續(xù)生產(chǎn)的網(wǎng)絡具有良好的泛化。最初的開發(fā)的過程中目標是是求最小平方誤差的總和n(t)Diii

2

（1）其中t是目標函數(shù)和a是神經(jīng)網(wǎng)絡反應各自的目標通常情況下開發(fā)的目標是減少平方差和F=Ed然而正規(guī)化增加了一個額外的長期目標函數(shù)這是由以下函數(shù)得到的：D

（2）在方程（2）中，是網(wǎng)絡權重的平方的總和是目標函數(shù)的參數(shù)，目標函數(shù)參數(shù)的相對大小決定了訓練的重點，如然后訓練的算法就會產(chǎn)生較小的誤差練的神經(jīng)網(wǎng)絡就會在網(wǎng)絡錯誤代價強調權重的減小，從而產(chǎn)生一個流暢的網(wǎng)絡響應。貝葉斯學校的統(tǒng)計數(shù)據(jù)是基于一個完全不同的角度來獲取數(shù)據(jù)其中的概率是用來表示關系的不確定性有學問看到任何數(shù)據(jù)之前先前的看法可能是真正的關系可以用概率分布來表示通過網(wǎng)絡權值來定義這種關系方案構思后的數(shù)據(jù)訂后的觀點通過之后的絡權值分布來捕獲以前網(wǎng)絡權重似乎有道理，但是與數(shù)據(jù)不相符現(xiàn)在被看做是不太可能而對不適合數(shù)據(jù)的權重值的概率將增加。在貝葉斯框架的權重網(wǎng)絡被認為是隨機變量。數(shù)據(jù)被采取后后驗概率密度函數(shù)的權重可以根據(jù)貝葉斯規(guī)則更新:(w/D

M)

(D/,)(w/(D/M)

M)

（3）在方程(3),D代表數(shù)據(jù)集,M是使用于特定的神經(jīng)網(wǎng)絡模型和w是網(wǎng)絡權值向量。(w/

,)是先驗率,代表我們的知識收集數(shù)據(jù)之前權重P(D/w

M)是似然函數(shù),這是數(shù)據(jù)的可能性發(fā)生,由于重量w.(/它可以保證總概率是1。

)一個標準化的因素,在這項研究中,我們使用矩陣神經(jīng)微機網(wǎng)絡化工具箱函數(shù)”trainbr”,這是一個公司麥夸特法和貝葉斯正則化定理(或貝葉斯學習融入到反向傳播來訓練神經(jīng)網(wǎng)絡來減少計算超過海森矩陣的近似值和產(chǎn)生好的的泛化能力種算法提供了一個網(wǎng)絡參數(shù)量度(權值和偏差)被網(wǎng)絡有效的利用不考慮在網(wǎng)絡中參數(shù)的總數(shù)，有效的參數(shù)的數(shù)量應該保持相同的。這就消除了猜測需要在確定最優(yōu)網(wǎng)絡大小。測試實驗數(shù)據(jù)可從立達,戈爾的制造商（并條RSB-D40）得到。對于這些實,在紡織行業(yè)材料的選擇和實驗設計是基于特定的材料使用的頻率。例如粗梳棉紗是使用最為頻繁的材料,所以它被作為標準和實驗研究在棉花的所有可能的設置,它不同于其他材料時,認清這個事實,所有的材料無法進行輥壓力和牽伸裝置,不同紡紗是事先是設置好的。材料的處理方案計在表中給出。標準的程序適用于所有的材料和標準程序對于自調勻整裝置棉條線密度、強度)也被采納手動和自動搜索進行比較從SBD-40自動搜索搜索功能獲得了了更好的CV%結果。因此，勻整作用點搜索時候由立達質量監(jiān)控器（）完成的。一個實驗模型的抽象描述如圖2所示。這里主要考慮的一點是在機械上調整勻整作用點主要影響因素是不可能的，例如，喂棉速度，因此給棉速度被認