自动化专业外文翻译alicia3爬壁机器人的粘着控制本科论文.docx

自动化专业外文翻译alicia3爬壁机器人的粘着控制本科论文.docx

《自动化专业外文翻译alicia3爬壁机器人的粘着控制本科论文.docx》由会员分享，可在线阅读，更多相关《自动化专业外文翻译alicia3爬壁机器人的粘着控制本科论文.docx（15页珍藏版）》请在冰点文库上搜索。

自动化专业外文翻译alicia3爬壁机器人的粘着控制本科论文

英语原文：

AdhesionControlfortheAlicia3ClimbingRobot

D.LongoandG.Muscato

DipartimentodiIngegneriaElettricaElettronicaedeiSistemi,Universit`adegli

StudidiCatania,vialeA.Doria6,95125CataniaItaly

Abstract.Climbingrobotsareusefuldevicesthatcanbeadoptedinavarietyof

applicationslikemaintenance,building,inspectionandsafetyintheprocessand

constructionindustries.

ThemaintargetoftheAlicia3robotistoinspectnonporousverticalwallwithanyregardforthematerialofthewall.Tomeetthistarget,apneumatic-likeadhesionforthesystemhasbeenselected.AlsothesystemcanmoveoverthesurfacewithasuitablevelocitybymeansoftwoDCmotorsandovercomesomeobstaclethankstoaspecialcupsealing.

Thisadhesiontechnologyrequiresasuitablecontrollertoimprovesystemreliability.Thisisbecausesmallobstaclespassingunderthecupandwallirregularitycanvarythevalueoftheinternalpressureofthecupputtingtherobotinsomeanomalousworkingconditions.Themethodologiesusedforderivinganaccuratesystemmodelandcontrollerwillbeexplainedandsomeresultwillbepresentedinthiswork.

1Introduction

Climbingrobotscanbeusedtoinspectverticalwallstosearchforpotentialdamageorproblemsonexternalorinternalsurfaceofaboveground/undergroundetrochemicalstoragetanks,concretewallsandmetallicstructures[1–4].Byusingthissystemascarrier,itwillbepossibletoconductanumberofNDIoverthewallbycarryingsuitableinstrumentation[5,6].Themainapplicationoftheproposedsystemistheautomaticinspectionoftheexternalsurfaceofabovegroundpetrochemicalstoragetankswhereitisveryimportanttoperformperiodicinspections（rateofcorrosion,riskofairorwaterpollution）atdifferentrates,asstandardizedbytheAmericanPetroleumInstitute[7].Thesystemcanbealsoadoptedtoinspectconcretedams.

Whilethesekindsofinspectionsareimportanttopreventecological

disastersandrisksforthepeopleworkingaroundtheplant,thesearevery

expensivebecausescaffoldingisoftenrequiredandcanbeverydangerous

Fig.1.TypicaloperatingenvironmentandtheAlicia3robot

fortechniciansthathavetoperformtheseinspections.Moreover,forsafety

reasons,plantoperationsmustbestoppedandthetankmustbeemptied,

cleanedandventilatedwhenhumanoperatorsareconductinginspections.In

Fig.1（a）and1（b）typicalenvironmentsforclimbingrobotsareshown.Figure

1cshowstheAlicia3robotprototypewhileattachedtoaconcretewallduring

asystemtest.

2SystemDescription

TheAliciaIIsystem（thebasicmodulefortheAlicia3system）ismainly

composedbyacup,anaspirator,twoactuatedwheelsthatusetwoDCmotors

withencodersandgearboxesandfourpassivesteelballswithclearanceto

guaranteeplaincontactofthecuptothewall.Thecupcanslideoverawall

bymeansofaspecialsealingthatallowsmaintainingasuitablevacuuminside

thecupandatthesametimecreatingtherightamountoffrictionwithrespect

systemweightandarangeofatargetwallkind.

ThestructureoftheAliciaIImodule,showninFig.2,currentlycomprises

threeconcentricPVCringsheldtogetherbyanaluminumsdisc.Thebigger

ringandthealuminumsdischaveadiameterof30cm.Thesealingsystemis

allocatedinthefirsttwoexternalrings.Boththetworingsandthesealingare

Fig.2.StructureoftheAliciaIImodule

designedtobeeasilyreplaceable,astheywearoffwhiletherobotisrunning.

Moreoverthesealingallowstherobotpassingoversmallobstacles（about1cm

height）likescrewsorweldingtraces.Thethirdring（thesmallestone）isused

asabaseforacylinderinwhichacentrifugalairaspiratoranditselectrical

motoraremounted.Theaspiratorisusedtodepressurizethecupformedby

theringsandthesealing,sothewholerobotcanadheretothewalllikea

standardsuctioncup.

Themotor/aspiratorsetisveryrobustandiscapableofworkinginharsh

environments.Thetotalweightofthemoduleis4Kg.

TheAlicia3robotismadewiththethreemoduleslinkedtogetherbymeans

oftworodsandaspecialrotationaljoint.Byusingtwopneumaticpistonsitis

possibletoriseandtolowereachmoduletoovercomeobstacles.Eachmodule

canberaised15cmwithrespecttothewall,soobstaclesthatare10–12cm

height,canbeeasilyovercame.Thesystemisdesignedtobeabletostay

attachedusingonlytwocupswhilethethird,anyofthethree,israisedup.

Thetotalweightofthesystemisabout20Kg.

3Electro-PneumaticSystemModel

Byusingthiskindofmovementandsealingmethod,itispossible,dueto

unexpectedsmallobstaclesonthesurface,tohavesomeairleakageinthe

cup.Thisleakagecancausetheinternalnegativepressuretoriseupandin

thissituationtherobotcouldfalldown.Ontheothersideiftheinternal

pressureistoolow（highΔp）,averybignormalforceisappliedtothe

system.Asaconsequence,thefrictioncanincreaseinsuchawaytonot

allowrobotmovements.Thisproblemcanbesolvedbyintroducingacontrol

looptoregulatethepressureinsidethechambertoasuitablevaluetosustain

thesystem.Theconsideredopenloopsystemandthemosteasilyaccessible

systemvariableshasbeenschematizedinFig.3;inthisschemethefirstblock

includestheelectricalandthemechanicalsubsystemandthesecondblock

includesthepneumaticsubsystem.TheusedvariablesaretheMotorvoltage

reference（theinputsignalthatfixesthemotorpower）andtheVacuumlevel

（thenegativepressureinsidethechamber）.

Fig.3.Theopenloopsystemconsidered

Fig.4.I/Ovariableacquisitionscheme

Sinceitisverydifficulttohaveareliableanalyticalmodelofthatsystem,

becauseofthebignumberofparametersinvolved,ithasbeendecidedto

identifyablackboxdynamicmodelofthesystembyusinginput/output

measurements.Thismodelwasdesignedwithtwopurposes:

tocomputea

suitablecontrolstrategyandtoimplementasimulatorfortuningthecontrol

parameters.

Anexperimentalsetupwasrealized,asrepresentedinFig.4,byusing

theDS1102DSPboardfromDspaceinordertogenerateandacquirethe

input/outputvariables.SincetheaspiratorisactuatedbyanACmotor,a

powerinterfacehasbeenrealizedinordertotranslateinpowerthereference

signalforthemotorcomingfromaDACchanneloftheDS1102board.The

outputsystemvariablehasbeenmeasuredwithapiezoresistivepressure

sensorwithasuitableelectronicconditioningblockandacquiredwithone

analoginputoftheDS1102.ThesoftwarerunningontheDSpaceDSPboard,

inthisfirstphasesimplygeneratesanexcitingmotorvoltagereferencesignal

（pseudorandom,ramporstepsignals）andacquiresthetwoanaloginputs

withasamplingtimeof0.1s,storingthedatainitsinternalSRAM.

TypicalInput/OutputmeasurementsarerepresentedinFig.5andFig.6.

Inordertoobtainbetterresultsinsystemmodeling,therelationshipbetween

InputandOutputneedstobeconsideredasnon-linear.ANARXmodelhas

beenusedisintheformof

（1）,wherefisanonlinearfunction[8,9].

y（k）=f（u（k）,u（k−1）,...;y（k−1）,y（k−2）,...）

（1）

Toimplementthiskindofnon-linearity,sometrialshavebeendoneusing

Neuro-FuzzyandArtificialNeuralNetwork（ANN）methodologies.Oncethat

modelhasbeentrainedtoasuitablemeansquareerror,ithasbeensimulated

givingitasinputtherealinputmeasurementonly（infinitesteppredictor）[8].

So

（1）canbemodifiedinordertoobtain

（2）.

˜y（k）=f（u（k）,u（k−1）,...;˜y（k−1）,˜y（k−2）,...）

（2）

In

（2）,4yistheestimatedsystemoutput.Inordertocomparethesimulation

results,anumberofdescriptorhasbeendefinedandused.Amongtheseare

meanerror,quadraticmeanerrorandsomecorrelationindexes.Afirstset

ofsimulationforbothmethodologieshasbeendonetofindoutthebestI/O

regressiontermschoice.

3.1Neuro-FuzzyIdentification

Usingthiskindofmethodology,thebestmodelstructurewasfoundtobein

theformof（3）.

y（t）=f（u（t）,y（t−1））（3）

Oncethebestmodelstructurehasbeenfound,sometrialshavebeen

performedmodifyingthenumberofmembershipfunctions.Thebestresults,

comparingtheindexesdescribedabove,havebeenobtainedwith3functions

andinFig.7thesimulationresultshasbeenreported.Thestructureofthe

Neuro-FuzzymodelistheANFIS-Sugeno[10].

3.2ANNIdentification

Usingthiskindofmethodology,thebestmodelstructurewasfoundtobein

theformof（4）.

y（t）=f（u（t）,u（t−1）,u（t−2）,y（t−1）,y（t−2））（4）

Asinglelayerperceptronnetworkhasbeenused.Thetrainingalgorithm

isthestandardLevenberg–Marquardt.

Oncethebestmodelstructurehasbeenfound,sometrialshavebeenperformed

modifyingthenumberofhiddenneurons.Thebestresults,comparing

theindexesdescribedabove,havebeenobtainedwith7hiddenneuronsand

inFig.8thesimulationresultshasbeenreported.

Fromacomparisonbetweenthetwomodelsandtheirrelatedindexes,it

canbeseenthattheNeuro-Fuzzymodelhasbestapproximationperformance

anduselessinputinformation.Inthenextsection,thismodelwillbeusedas

systememulatortotuneandtesttherequiredregulator.

4PressureControlAlgorithm

Onceasystemmodelhasbeenobtained,aclosedloopconfigurationlikethat

inFig.9,hasbeenconsidered.

Thetargetofthecontrolalgorithmistoregulatetheinternalvacuumlevel

toasuitablevalue（fromsometrials,itwasfixedtoabout10kPa）tosustain

thewholesystemanditspayload;themaximumsteadystateerrorallowed

wasfixed