采煤机相关英文文献翻译.docx

采煤机相关英文文献翻译.docx

《采煤机相关英文文献翻译.docx》由会员分享，可在线阅读，更多相关《采煤机相关英文文献翻译.docx（24页珍藏版）》请在冰豆网上搜索。

采煤机相关英文文献翻译

英文原文：

Controlstrategyforanintelligentshearerheightadjustingsystem

FANQigao*,LIWei,WANGYuqiao,ZHOULijuan,YANGXuefeng,YEGuoSchoolofMechanical&ElectricalEngineering,ChinaUniversityofMining&Technology,Xuzhou221008,China

Abstract:

Anintelligentshearerheightadjustingsystemisakeytechnologyforminingataman-lessworkingface.Acontrolstrategyforashearerheightadjustingsystembasedonamathematicalmodeloftheheightadjustingmechanismisproposed.Itconsidersthenon-linearityandtimevariationsinthecontrolprocessandusesDynamicFuzzyNeuralNetworks（D-FNN）.Theinversecharacteristicsofthesystemarestudied.Anadaptiveon-linelearninganderrorcompensationmechanismguaranteessystemreal-timeperformanceandreliability.ParametersfromaGermanEickhoffSL500shearerwereusedwithMatlab/Simulinktosimulateaheightadjustingcontrolsystem.SimulationshowsthatthetraceerrorofaD-FNNcontrollerissmallerthanthatofaPIDcontroller.Also,theD-FNNcontrolschemehasgoodgeneralizationandtrackingperformance,whichallowittosatisfytheneedsofashearerheightadjustingsystem.

Keywords:

shearer;heightadjustingsystem;dynamicfuzzyneuralnetwork

1Introduction

Thesheareranditscontrolsystemaremaincomponentsforcoalmining.Theshearingprocessincludesdrumliftingandtractioncontrol.Domesticsheardrumliftingnowusesmanualadjustmentsafterartificialobservationorageometrictrackcutting-memorymethodaftertrialmanualadjustmentsfromtestcuttings.Theinstallationofsensorsontheshearerthatcouldidentifycoal-rockhasbeenproposed.Informationfromthesensorswouldbeusedtoachievedrumheightcontroldirectlybyautomaticallyliftingtheshearer

.Thistechnology,whichisbasedonsimpledrumheightfeedback,hasnotbeenwidelyappliedduetothestructuralcomplexityofthecoalseam,technicalproblemsrelatedtoidentificationofthecoal-rockinterfaceaswellasroof,andfloor,requirementsforsuchcomprehensivecoalminingmechanization.Othershaveproposedanintelligentshearerheightadjustingsystembasedonaself-adaptivePIDneuralnetworkcontrolmethod

.Thisrequiresdatasamplesfromanoperatingshearerheightadjustingsystemfollowedbycarefulchoiceoftheneuralnetworkandadjustmentofthealgorithmicparameters.Thesuitabilityofthesystemwouldthenbedeterminedbycheckingperformanceagainsttestsamples.Afterthestructureandparametersweredeterminedthetrainedneuralnetworkcouldbeappliedtopracticalsystems.Theparameterscouldbead-justedfurtherwhilethesystemwasrunningto

achieveself-adaptivelearningandcontrol.Settingupsuchasysteminvolvesconsiderableuncertaintyandagreatdealoftime.

Consideringthefactorsandtheneedforimprovingproductqualityandresourcerecoverybyautomaticcontrolofthedrumheightweproposeanewmethodcalledtheshearerintelligentheightadjustingsystemcontrolmethod.ItisbasedonDynamicFuzzyNeuralNetworks（D-FNN）.D-FNNareneuralnetworksthathavethecharacteristicsofpowerfulon-linelearning,fastlearningandgoodgeneralization.D-FNNgivereal-timecontrolandimprovedynamiccharacteristicsofashearerheightadjustingsystemandprovideatheoreticalbasisfordesigninganintelligentheightadjustingcontrolsystemfortheshearer.

2Analysisofashearerheightadjustingsystem

2.1Structureoftheshearerheightadjustingsystem

Theshearerheightadjustingmechanismusesahydraulicservosystemhavinggooddynamicperformance.Fig.1diagramsadrumshearer.Theelectro-hydraulicservosystemcontrolsextensionofthehydrauliccylinderandmovestherockerarmtosettheheight.Theadjustingmechanismisaplanaropenchainconsistingofaseriesofconnectedrodstructuresandcorrespondingkinematicpairs.Adescripionoftherelativemotionofthepartsshowshowheightadjustmentoccurs.Adetailedmotionanalysisfollows.Suppose:

1）Allcomponentsarerigidandelasticdeformationisignored;

2）Gapsbetweenallmechanismsareignored.

2.2Mathematicalanalysisoftheshearerheightadjustmentsystem

Fig.2showstheinitialpositionofthehydrauliccylinderas

theendpositionas

thelongarmoftherockerarmisL,shortarmis

thedrawbarbetweentheheightadjustmentcylinderandtherockerarmis

thedistancebetweentheheightadjustmentcylinderandtherockerpivotisDandtheanglebetweenthelongarmandtheshortarmis

.

Definition1.ShearerminingheightH:

H=L

（1）

Endposition

isgivenby

allowingthedisplacementofthehydrauliccylinder,

tobeestablished.

Definition2.Displacementofthehydrauliccylinder,

is:

（2）

where

Wewrite:

（3）

where

Substitutiongives

as:

（4）

Sincebisgivenby

canbeexpressedasafunctionofrocker-heighttoangle:

（5）

Kineticanalysisofthemodelshearerheightadjustingsystemshowsitisathirdordersystem.Thesystemtransferfunctionis

:

（6）

whereKisthesystemgain,ζisthesystemdampingratio,wisthenaturalfrequencyofthesystem,F（s）theLaplacetransformoftheservomechanism,

theLaplacetransformof

（inEq.（5））,

isderivedfromEq.（6）,theswingangle,θ,oftherockerarmisfromEq.（5）andθcontrolsthefeedback.

Sincetheheightadjustingsystemisnon-linearandatime-varyingdynamicsystematraditionalPIDcontrollercannotprovidesatisfactorycontrol.D-FNNareproposedasmeetingtherequirementsofreliabilityandrealtimeperformance.

3Dynamicfuzzyneuralnetworks

D-FNNarebasedontheexpansionofRadialBasisFunction（RBF）neuralnetworks.Theprominentcharacteristicsofthislearningalgorithmarethesimultaneousadjustmentofparametersandtheidentificationofanappropriatestructure.Thisprovidesrapidlearningsuitableforreal-timecontrolandformodelingoftheshearerheightadjustingsystem

ThestructureofadynamicfuzzyneuralnetworkisshowninFig.3.

InFig.3

…,

arethesysteminputvariables,yisthesystemoutput,

isthemembershipfunction,j,oftheinputvariable,i,

isthefuzzyruleofmembershipfunctionj,

isthenormalizednodeofj,

istheconnectionweightofrulejanduisthewholesystemrulenumber.

Theswingangle,θ,oftherockerarmwaschosenasthesysteminputvariablethatcontrolsexpansionofthehydrauliccylinder.AGaussianfunction,Eq.（7）,isusedforthemembershipfunction.

（7）

whereirangesfrom1tor,jrangesfrom1tou,

isthemembershipfunction,j,of

isthecenteroftheGaussianmembershipfunction,j,of

isthewidthoftheGaussianmembershipfunction,j,of

ristheinputvariablenumberanduisthenumberofthemembershipfunctionaswellasthewholesystemrulenumber.

Theoutputof

rulej,isobtainedfrom:

（8）

whereXisgivenby:

andthecenterofRBFneuralnetworkjisgivenby:

ThisgivestheD-FNNmodelas:

（9）

whereαistheconnectionweightofrulei.

4D-FNNcontrolstrategy

TheD-FNNcontrolschemeisshowninFig.4.Thebasicideaisobtainingtheinversecharacteristicoftheshearerheightadjustingsystemandthenproducingacompensationsignalfromthisinversedynamicmodel.Therearetwodynamicfuzzyneuralnetworkshere:

AandB.NetworkAisforsystemweighttrainingwhilenetworkBisacopyofthetrainedAnetworkthatisusedforproducingthecontrolsignal.

Thecontrolalgorithmis:

（10）

wherexΔistheexpecteddisplacementoftheheightadjustinghydrauliccylinder;PDΔxtheactualdisplacementofthecylinderproducedbythePDcontrollerandDFNNBΔxtheactualdisplacementofthecylinderproducedbynetworkB.

ThePDcontrollerisforfasterandmoreaccuratetrackingperformance.ThekeytotheD-FNNcontrol

systemisthetrainingofD-FNNBtominimizethesquarederrorbetweenexpectedandactualdisplacementsproducedbynetworkB

:

（11）

Agradientdescentmethodisusedfortheweightadjustingalgorithm

:

（12）

whereλisthelearningrateandλ>0.λhasalargeinfluenceontheconvergencerate.Increasingofλcanspeeduptheconvergencerate,whichismoresuitablefortime-varyingsystemmodelingandcontrol.Atthesametimetheanti-interferenceperformanceofthesystemdeclines.Adecreaseinλslowsdownconvergencebutproducesasystemlesssensitivetointerference.Aself-adjustinglearningratemethodisproposedherein,theprinciplebeingthatwhenthenewerrorexceedsthelasterrorovershootinghasoccurredandλshouldbereduced.Ifthenewerrorissmallerthanthelasterrortheweightadjustmentsareeffectiveandλshouldbeincreased.Iftheerrorisconstantthenλiskeptthesame.Thismaybewrittenas:

（13）

TestsshowthatD-FNNusingtheself-adjustinglearningratemethodrequiresmuchlesstrainingtimethansystemsusingafixedlearningrate.

5Systemsimulation

ThemathematicalmodelandaD-FNNcontrolalgorithmmaybeusedinamodelshearerheightad-

justingsystembuiltusingMatlab/Simulink[

.TheactualparametersarefromaGermanEickhoffSL500machine.Theshearermaximumcuttingheightis5.50mandthefootwallis1.08m.Theangleoftherockerarmis–21.3°~+55°.Thedrawbar,LG,is2.05m,theshortarm,LR,is1.20m,Dis0.9mandtheangle

5.1SimulationofaD-FNNcontroller

Supposetherockerarmmoveswithinarangeof–21.3°~+55°.TheD-FNNcontrolstrategytracesthetrajectoryoftherockerarmandthetrajectorytracingerrorareshowninFig.5.InFig.5bthemaximumtrajectorytracingerroroftherockerarmis0.65°,whichoccursearlyinthetrainingstage.AtthispointtheD