Robot Hydraulic Control

download Robot Hydraulic Control

of 108

Transcript of Robot Hydraulic Control

  • 7/29/2019 Robot Hydraulic Control

    1/108

    AHYDRAULICFLEXIBLEJOINTROBOTSIMULATOR

    AThesisSubmittedtothe

    CollegeofGraduateStudiesandResearch

    inPartialFulfillmentoftheRequirements

    fortheDegreeofMasterofScience

    intheDepartmentofMechanicalEngineering

    UniversityofSaskatchewan

    Saskatoon,Canada

    By

    ShahramDezfulian

    CopyrightShahramDezfulianJune2007.Allrightsreserved.

  • 7/29/2019 Robot Hydraulic Control

    2/108

    i

    PERMISSIONTOUSE

    Inpresentingthisthesisinpartialfulfillmentoftherequirementsforapostgraduatedegreefrom

    theUniversityofSaskatchewan,IagreethattheLibrariesofthisUniversitymaymakeitfreely

    availableforinspection.Ifurtheragreethatpermissionforcopyingofthisthesisinanymanner,

    inwholeorinpart,forscholarlypurposesmaybegrantedbytheprofessororprofessorswho

    supervisedmythesisworkor,intheirabsence,bytheHeadoftheDepartmentortheDeanof

    theCollegeinwhichmythesisworkwasdone.Itisunderstoodthatanycopyingorpublication

    oruseofthisthesisorpartsthereofforfinancialgainshallnotbeallowedwithoutmywritten

    permission.ItisalsounderstoodthatduerecognitionshallbegiventomeandtotheUniversity

    ofSaskatchewaninanyscholarlyusewhichmaybemadeofanymaterialinmythesis.

    Requestsforpermissiontocopyortomakeotheruseofmaterialinthisthesisinwholeorpart

    shouldbeaddressedto:

    Head,DepartmentofMechanicalEngineering

    CollegeofEngineering

    UniversityofSaskatchewan

    57CampusDrive

    Saskatoon,Saskatchewan

    Canada,S7N5A9

  • 7/29/2019 Robot Hydraulic Control

    3/108

    ii

    ABSTRACT

    Theobjectiveofthisprojectwastodesignandimplementanexperimentalhydraulic

    systemthatsimulatesjointflexibilityofasinglerigidlinkflexiblejointrobotmanipulator,withtheabilityofchangingthejointflexibilitysparameters.Sucha

    systemcouldfacilitatefuturecontrolstudiesofrobotmanipulatorsbyreducing

    investigationtimeandimplementationcostofresearch.Itcouldalsobeusedtotestthe

    performanceofdifferentstrategiestocontrolthemovementofflexiblejoint

    manipulators.

    Ahydraulicrotaryservomotorwasusedtosimulatetheactionofaflexiblejointrobot

    manipulator.Itwasachallengingtask,sincethecontrolofangularaccelerationwas

    required.

    Asingle-rigid-link,elastic-jointrobotmanipulatorwasmathematicallymodeledand

    implementedusingMatlab.Jointflexibilityparameterssuchasstiffnessanddamping,

    couldbeeasilychanged.Thissimulationwasconsideredasafunctiongeneratorto

    drivethehydraulicflexiblejointrobot.Inthisstudythedesiredangularaccelerationof

    themanipulatorwasusedastheinputtothehydraulicrotarymotorandtheobjective

    wastomakethehydraulicsystemfollowthedesiredaccelerationinthefrequencyrange

    specified.Thehydraulicsystemconsistedofaservovalveandrotarymotor.

    Ahydraulicactuatorrobotwasbuiltandtested.Theresultsindicatedthatiftheinput

    signalhadafrequencyintherangeof5to15 Hzanddampingratioof0.1,the

    experimentalsetupwasabletoreproducetheinputsignalwithacceptableaccuracy.

    Becauseoftheinherentnoiseassociatedwiththemeasurementofaccelerationandsome

    severenon-linearitiesintherotarymotor,controloftheexperimentaltestsystemusing

    classicalmethodswasnotassuccessfulashadbeenanticipated.Thiswasafirststagein

    aseriesofstudiesandtheresultsprovideinsightforthefutureapplicationofmore

    sophisticatedcontrolschemes.

  • 7/29/2019 Robot Hydraulic Control

    4/108

    iii

    ACKNOWLEDGEMENTS

    IwouldliketoexpressmydeepestappreciationtoProfessorsR.T.Burton

    andR.Fotouhifortheiradvices,encouragementandpatience.Myspecial

    thanksgotoMr.D.V.Bitnerforhistechnicalsupportandnever-ending

    helpandarrangements.

    Iwouldalsoliketoacknowledgethelong-lastingsupportandloveoffered

    bymyparents,FraidoonDezfulianandAfaghDelfani,mywife,KathyAbrishamiand

    mydaughter,SheryDezfulian.

  • 7/29/2019 Robot Hydraulic Control

    5/108

    iv

    CONTENTS

    PermissiontoUse...................................................... i

    Abstract..................................................... ........ii

    Acknowledgements.................................................... iii

    Contents............................................................. iv

    ListofTables........................................................ vi

    ListofFigures....................................................... vii

    Nomenclature........................................................ xii

    1Introduction........................................................ 11.1Background.....................................................1

    1.2LiteratureReview.................................................2

    1.3Objectives......................................................51.4ThesisOutline...................................................6

    2FunctionGenerator................................................. 7

    2.1Introduction.....................................................7

    2.2GoverningEquations.............................................7

    3HydraulicSimulator................................................ 193.1Introduction....................................................19

    3.2HydraulicSystemAnalysis........................................21

    3.3ControllerDesign(Theoretical).....................................303.3.1Closed-LoopController.....................................30

    3.3.2Open-LoopInverseCompensationController....................37

    4ExperimentalSystemSetup.......................................... 39

    4.1Introduction....................................................39

    4.2PowerSupply...................................................40

    4.3PressureControlServovalve.......................................404.4HydraulicRotaryMotor..........................................44

    4.5RobotManipulator..............................................454.6PersonalComputer..............................................48

    4.7SignalAnalyzer.................................................49

    4.8Accelerometer..................................................504.9AccelerometerCompensator.......................................50

  • 7/29/2019 Robot Hydraulic Control

    6/108

    v

    5ExperimentalResults............................................... 555.1Introduction....................................................55

    5.2PressureControlServovalve.......................................55

    5.3Accelerometer..................................................59

    5.4HydraulicSimulator.............................................63

    5.4.1Open-LoopFrequencyResponses.............................635.4.2Closed-LoopFrequencyResponses............................67

    5.4.3Open-LoopCompensation...................................695.4.4TransientAccelerationResponses.............................74

    5.4.4.1Frequency5HzandDampingRatio0.1.................75

    5.4.4.2Frequency10HzandDampingRatio0.1................785.4.4.3Frequency15HzandDampingRatio0.1................80

    5.4.5TransientResponseErrors...................................83

    6ConclusionsandRecommendations...................................89

    6.1Conclusions....................................................89

    6.2FutureWork...................................................91

    References........................................................... 92

  • 7/29/2019 Robot Hydraulic Control

    7/108

    vi

    LISTOFTABLES

    2.1Simulationparametersadoptedfrom[9]................................12

    3.1Parameterssummaryofopen-loopandclosed-looptransferfunctions.........32

    4.1SpecificationofMoogservovalvemodel15-010.........................43

    4.2SpecificationsofMicromaticrotarymotorMPJ-22-1V....................44

    4.3Importantmanipulatormeasurements..................................48

    4.4Matlab-Simulinksetupforexperimentalsystemcontrol....................49

    5.1Meansquarederrorvaluesoftransientresponseerrorsindifferentcases......87

  • 7/29/2019 Robot Hydraulic Control

    8/108

    vii

    LISTOFFIGURES

    2.1Schematicofasingle-linkmanipulatorwithflexiblejoint...................8

    2.2Theopen-loopblockdiagramofsinglelinkmanipulatorwithflexiblejoint....112.3Simulationresultsforimpulseinputstothesinglelinkmanipulator.........14

    2.4Plotofequation2.29...............................................17

    3.1Schematicofthehydraulicsimulator..................................20

    3.2Blockdiagramofequation3.11......................................25

    3.3Reducedblockdiagramofequation3.11...............................25

    3.4TypicalmagnitudeBodeplotofequation3.11when nv

    1

    1.............26

    3.5TypicalmagnitudeBodeplotofequation3.11when nv

    1

    1.............26

    3.6Open-loopmagnitudefrequencyresponseofthehydraulic

    simulator(points)withsuperimposedasymptoticlines(experimental:solidlinebasedon40dBperdecadeand

    theoretical:dashedlinebasedon20dBperdecade).......................27

    3.7Simplifiedopen-loopblockdiagramofthesystem.......................28

    3.8Blockdiagramofthesimplifiedtransferfunctionwithinfluenceof

    theaccelerometersensitivity.........................................29

    3.9Blockdiagramofclosed-loopcontrolsystemwithproportionalgain.........31

    3.10Magnitudefrequencyresponse)(

    )(

    sV

    s

    e

    v oftheopen-loopand

    closed-loopsystemswithproportionalcontroller........................32

    3.11Transientresponseof v intheopen-loopandclosed-loopmodels

    with 2=PK tothefunctiongenerator d of10Hzsignal..................33

    3.12Blockdiagramofclosed-loopcontrolsystemwithproportional

    andintegralgains.................................................34

  • 7/29/2019 Robot Hydraulic Control

    9/108

    viii

    3.13Magnitudefrequencyresponse)(

    )(

    sV

    s

    e

    v ofopen-loopand

    closed-loopsystemswithPIcontroller................................35

    3.14Transientresponseofv

    intheopen-loopandclosed-loopmodels

    with 2=PK and 10=IK tothefunctiongenerator d of10Hzsignal.......36

    3.15Blockdiagramofopen-loopcompensatorcontroller.....................37

    3.16Transientresponseof v intheopen-loopcompensatedmodeltothe

    functiongenerator d of10Hzsignal.................................38

    4.1Schematicofthehydraulicsimulator.................................39

    4.2Schematicofthepressurecontrolservovalve............................41

    4.3Schematicofthetwo-linkrigidmanipulatorinwhichtheshoulderlinkwasfixed............................................46

    4.4Photoofthemanipulatorsetupusedintheexperiment....................47

    4.5SchematicofShakerTabletest.....................................51

    4.6MagnitudeandphaseBodeplotsoftheshakertabletest(pointsaredatafromtheanalyzerandthesolidlineisthestraightline

    approximationbasedona20dBperdecadeslope).......................52

    4.7Blockdiagramofthesystemwiththeaccelerometercompensator...........54

    5.1aMagnitudefrequencyresponseofthepressurecontrolservovalvewithablockedload(pointsaredatafromtheanalyzerandthesolid

    lineisthestraightlineapproximation)................................56

    5.1bPhasefrequencyresponseofthepressurecontrolservovalve

    withablockedload................................................57

    5.2aMagnitudefrequencyresponseofpressurecontrolservovalvewith

    actualhydraulicsimulatorload(pointsaredatafromtheanalyzer

    andthesolidlineisthestraightlineapproximation).....................58

    5.2bPhasefrequencyresponseofpressurecontrolservovalvewith

    actualhydraulicsimulatorload.......................................58

  • 7/29/2019 Robot Hydraulic Control

    10/108

    ix

    5.3aMagnituderatio

    ja

    ja

    po

    ac

    (

    )( versusfrequencyoftheaccelerometerwith

    proposedaccelerometercompensator(pointsaredatafromthe

    analyzerandthesolidlineisthestraightlineapproximation)...............61

    5.3bPhase)()(

    jaja

    po

    ac versusfrequencyoftheaccelerometerwith

    proposedaccelerometercompensator.................................61

    5.4Open-loopmagnitudefrequencyresponseofthehydraulicsimulator

    beforeapplyingtheaccelerometercompensator(pointsaredatafrom

    theanalyzerandthesolidlineisthebestfitstraightlineapproximation;40dBperdecadeslope)............................................62

    5.5Open-loopmagnitudefrequencyresponseofthehydraulicsimulatorafterapplyingaccelerometercompensator(pointsaredatafromthe

    analyzerandthesolidlineistheexpectedstraightlineapproximationoftheplant;20dBperdecadeslope)..................................63

    5.6aOpen-loopmagnitudefrequencyresponseofthehydraulicsimulator

    withinputamplitudeof0.5vor 2/5.11 srad (solidlineistheideal

    theoreticalfrequencyresponse)....................................65

    5.6bOpen-loopphasefrequencyresponseofthehydraulicsimulator

    withinputamplitudeof0.5v( 2/5.11 srad )...........................65

    5.7aOpen-loopmagnitudefrequencyresponseofthehydraulicsimulator

    withinputamplitudeof0.1v( 2/3.2 srad )............................66

    5.7bOpen-loopphasefrequencyresponseofthehydraulicsimulator

    withinputamplitudeof0.1v( 2/3.2 srad )...........................66

    5.8Closed-loopmagnitudefrequencyresponseofthehydraulicsimulator

    withinputamplitude1.0vor 2/0.23 srad (pointsaredatafromthe

    analyzerandthesolidlineisthestraightlineapproximation,

    20dBperdecadeslope)............................................67

    5.9Closed-loopmagnitudefrequencyresponseofthehydraulic

    simulator(inputamplitude0.5vor 2/5.11 srad )........................68

    5.10Closed-loopmagnitudefrequencyresponseofthehydraulic

    simulator(inputamplitude0.1vor 2/3.2 srad ).........................68

  • 7/29/2019 Robot Hydraulic Control

    11/108

    x

    5.11Open-looptransientresponseofthehydraulicsimulatortoa

    unitstepinputwithamplitudeof1v( 2/0.23 srad ).......................69

    5.12Transientresponseofthemodel(71.15

    4.0

    +s

    s)tothestepinput

    signal(redlineistheinputandbluelineistheoutput)....................70

    5.13aOpen-looptransientresponseofthehydraulicsimulatorto5Hz

    sinewavewithamplitudeof1v(compensations

    s

    8.0

    71.15+).................72

    5.13bOpen-looptransientresponseofthehydraulicsimulatorto5 Hz

    sinewavewithamplitudeof2v(compensations

    s

    8.0

    71.15+).................72

    5.14aOpen-looptransientresponseofthehydraulicsimulatorto15 Hz

    sinewavewithamplitudeof1v(compensations

    s

    8.0

    71.15+).................73

    5.14bOpen-looptransientresponseofthehydraulicsimulatorto15 Hz

    sinewavewithamplitudeof2v(compensations

    s

    8.0

    71.15+).................74

    5.15Open-looptransientresponseofthehydraulicsimulatortothefunctiongenerator5Hzand0.1dampingratiosignal(notcompensated).............76

    5.16Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator5Hzand0.1dampingratiosignal(compensatedbys

    s8.0

    4+ )........76

    5.17Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator5Hzand0.1dampingratiosignal(compensatedbys

    s

    8.0

    8+)........77

    5.18Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator5Hzand0.1dampingratiosignal(compensatedbys

    s

    8.0

    71.15+).....77

    5.19Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator10Hzand0.1dampingratiosignal(notcompensated)............78

    5.20Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator10Hzand0.1dampingratiosignal(compensatedbys

    s

    8.0

    4+).......79

  • 7/29/2019 Robot Hydraulic Control

    12/108

    xi

    5.21Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator10Hzand0.1dampingsignal(compensatedbys

    s

    8.0

    8+)............79

    5.22Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator10Hzand0.1dampingratiosignal(compensatedbys

    s8.0

    71.15+ )....80

    5.23Open-looptransientresponseofthehydraulicsimulatortothefunctiongenerator15Hzand0.1dampingratiosignal(notcompensated)............81

    5.24Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator15Hzand0.1dampingsignal(compensatedbys

    s

    8.0

    4+)............81

    5.25Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator15Hzand0.1dampingsignal(compensatedbys

    s

    8.0

    8+)............82

    5.26Open-looptransientresponseofthehydraulicsimulatortothefunction

    generator15Hzand0.1dampingratiosignal(compensatedbys

    s

    8.0

    71.15+)....82

    5.27TrackingerrorofFigure5.15(frequency5 Hzanddampingratio0.1,

    notcompensated).................................................83

    5.28TrackingerrorofFigure5.18(frequency5 Hzanddampingratio0.1,

    compensatedbys

    s

    8.0

    71.15+),..........................................84

    5.29TrackingerrorofFigure5.19(frequency10 Hzanddampingratio0.1,

    notcompensated),................................................84

    5.30TrackingerrorofFigure5.22(frequency10 Hzanddampingratio0.1,

    compensatedbys

    s

    8.0

    71.15+).........................................85

    5.31TrackingerrorofFigure5.23(frequency15 Hzanddampingratio0.1,notcompensated)................................................85

    5.32TrackingerrorofFigure5.26(frequency15 Hzanddampingratio0.1,

    compensatedbys

    s

    8.0

    71.15+)..........................................86

  • 7/29/2019 Robot Hydraulic Control

    13/108

    xii

    NOMENCLATURE

    linkangularacceleration )/( 2srad

    d desiredangularacceleration(functiongeneratoroutput)(

    2

    /srad )v linkangularaccelerationinvolts(v)

    vc linkangularaccelerationinvoltwithcompensator(v)

    motorviscousdamping )/( radsmN

    12P nozzlepressuredifference(Pa)

    ABP loadpressuredifference(Pa)

    jointdampingratio

    v valvedampingratio

    linkangularposition(rad)

    &

    linkangularvelocity(rad/s)&& linkangularacceleration( 2/srad ) Laplacetransformof(rad) transferfunctiontimeconstant(s)

    1 loadtimeconstant(s)

    a servo-amplifiertimeconstant(s)

    1CL closed-looptimeconstantwithPcontroller(s)

    o open-looptimeconstant(s)

    motorangularposition(rad)

    & motorangularvelocity(rad/s)

    && motorangularacceleration( 2/srad )

    Laplacetransformof(rad)

    linkangularvelocity(rad/s)

    n jointnaturalfrequency(rad/sorHz)

    nv valveundampednaturalfrequency( rad/sorHz)

    1a partial-fractionexpansionparameter

    2a partial-fractionexpansionparameter

    3a partial-fractionexpansionparameter

    4

    a partial-fractionexpansionparameter

    aca measuredaccelerometersacceleration(v)

    poa calculatedaccelerationbydoubledifferentiationofposition(v)

    AA spoolringarea(2m )

    SA spoolendarea( 2m )

    Djointviscousdamping( radsmN / )

    mD motorvolumetricdisplacement )/(3 radm

  • 7/29/2019 Robot Hydraulic Control

    14/108

    xiii

    ie accelerationerror(v)

    1F firsttermofequation2.19

    2F secondtermofequation2.19

    ac

    G accelerometertransferfunction

    1G idealaccelerometercompensator'

    ei servo-amplifieroutputcurrent(a)

    Ilinkmomentofinertia( radsmN /2 )

    eI inputcurrent(ma)

    '

    eI Laplacetransformof'

    ei (a)

    LI linkmomentofinertia )/(2 radsmN

    Jmotormomentofinertia( radsmN /2

    )

    1k transferfunctiongain(4

    / smNrad )

    Kjointstiffness( radmN / )

    1K loadgain( sParad / )

    aK servo-amplifiergain(a/v)

    1CLK closed-loopgainwithPcontroller(s)

    conK conversiongainfactor( 2/srad

    v)

    eK servo-amplifierconversiongain(ma/a)

    IK integralgain(1/s)

    oK open-loopgain( svrad / )

    OLK open-loopoverallgain(s)

    PK proportionalgain

    vK valvegain(Pa/ma)

    lrotationarm(m)

    Lservo-amplifierinductance(H)

    MSEmeansquarederrorvalue(2v )

    LP loaddifferentialpressure(Pa)

    SP supplypressure(Pa)

    Rservo-amplifierresistance()

    sLaplaceoperator(1/s)

    lS accelerometerlinearcalibratedrating( 2/sm

    v)

    ttime(s)

    LT motortorque( mN )

    umotorinputtorque( mN )

    ULaplacetransformofu( mN )

    ev servo-amplifierinputvoltage(v)

  • 7/29/2019 Robot Hydraulic Control

    15/108

    xiv

    eV Laplacetransformof ev (v)

    pox measuredposition(v)

  • 7/29/2019 Robot Hydraulic Control

    16/108

    1

    CHAPTER1

    INTRODUCTION

    1.1Background

    Arobotmanipulatorconsistsoflinks,joints,anddrivecomponents.Inordertoachieve

    accuraterobotpositioning,thelinksandjointsareusuallymadeasrigidaspossible.

    Rigidrobotmanipulatorsarebasedontheassumptionthatthetransmissionsarestiff

    andthatthelinksarerigid.Suchmanipulatorsareheavy(e.g.,load-carryingcapacityis

    typicallyonly5%to10%oftheirownweight),consumeconsiderablepower,andare

    generallyimpracticalforhighspeedmaneuvers.Ontheotherhand,flexible

    manipulatorshaveseveraladvantagesoverrigidmanipulators,suchassmallersizeand

    mass,rapidresponse,lowerpeakpowerrequirements,andenergyuse.Lightweight

    manipulatorshavelinksthatdeflectsignificantlyinhighspeedoperationssotheir

    flexibilitycannotbeignored.Flexibility,however,canalsobepresentinthejoints.

    Jointflexibilityexistswhenthereisadifferencebetweentheangularpositionofthe

    drivingactuatorandthatofthedrivenlink.Itisknownthatthejointflexibilitycan

    causeoscillationsinrobotmanipulators.Therefore,whenanaccuratetrajectorytracking

    oftheend-effecterisneeded,jointflexibilityisconsideredasaproblem[1].Inother

    words,jointelasticityisthemainsourceofcomplianceinmostcurrentmanipulator

    design.Jointelasticityoriginatesfromseveraltransmittingcomponents,suchas

    elasticityinthegears,belts,hydrauliclines,etc.[2].

  • 7/29/2019 Robot Hydraulic Control

    17/108

    2

    Thecontroltasks,incaseoftheelasticjointmanipulators,aremorecomplicatedthan

    theequivalentrigidjointmanipulators.Thisistrueduetothefactthatintheelasticjoint

    manipulators,thenumberofcontrolinputtorquesappliedbytheactuators,islessthan

    thenumberofdegreesoffreedom(DOF),alsoknownasunder-actuatedsystems.This

    happensbecauseforeachjointthereisoneactuator,andtwodegreesoffreedom.This

    conditionimpliesthattheimplementationofafullstatefeedbackcontroltaskneeds

    additionalsensorsformeasuringthestatevariablesoftheactuatoraswellasthelinks.

    Thepurposeofcontrolistoappropriatelydealwiththeoscillationcreatedbythejoint

    elasticityinordertogetfastpositioningandaccuratetrajectorytrackingofthe

    manipulatorend-effecter.Therefore,propercontrolofelasticjointrobotmanipulators,

    withdifferentcharacteristicexpectations,andwithareasonablenumberofsensors,is

    stillachallengingproblemthatneedstobeaddressedandindeed,isoneofthe

    motivatingfactorsforthisresearchproject.

    1.2LiteratureReview

    Inthissectionseveralpapersassociatedwiththreedifferenttopicsareinvestigated.First

    dynamicmodelingofflexiblejointrobotmanipulatorsisbroughtintofocus.Then

    differentstrategiesforcontrolofflexiblejointrobotmanipulatorsarebrieflystudied.

    Finallyafewpapersregardingtheanalysisofhydraulicactuatedrobotsarereviewed.It

    shouldbenotedthatthereareotherpaperspublishedonthetopicofcontrolofflexible

    andstiffmanipulatorsbuttheseareonlymarginallyrelatedtothetopicofthisthesis.

    Thefollowingrepresentsasampleofthepublicationsinthisarea.

    MarioandSpongstatedthatrecentexperimentalworksshowedthatjointelasticitywas

    themainsourceofcomplianceinmostcurrentmanipulatordesigns[2].Thisjoint

    elasticityoriginatedfromseveralmotiontransmittingcomponents,suchaselasticityin

    thegears,belts,hydrauliclines,etc.Inthispaper,asinglerigidlinkmanipulatorwithan

    elasticjointwasmodeled.Twocontrolsolutionswerederivedandcompared:onewas

  • 7/29/2019 Robot Hydraulic Control

    18/108

    3

    basedonthefeedbacklinearizationtechnique,andtheotheronebasedonthenonlinear

    compositetechnique.Theresultsindicatedthatthefeedbacklinearizationcontrol

    neededafullstatefeedback(thepositionsandvelocitiesofthelinkandmotor);

    however,thenonlinearcompositecontrolrequiredonlythepositionandvelocityofthe

    linkasthestatefeedback.

    Nicosiaetalexaminedsomeproblemsofdynamicalcontrolofmanipulatorswithhigh

    speedcontinuousdisplacement[3].Firsttherigidmodelwasdevelopedandthenelastic

    anddissipativejointswereincluded.Acontrolstrategybasedonanonlinearfeedbackof

    amodelofthelocaljointvariableswasapplied,andthentheperformanceofthe

    controlledmanipulatorwasevaluatedbymeansofsimulation.Theresultsshowedthat

    theuseofelasticjointsinducedvibrations.Theresultswerenotverifiedexperimentally.

    Spongderivedasimplemodeltorepresentthedynamicsofelasticjointmanipulators

    [4].Twobasicassumptionsaboutthedynamiccouplingbetweenthemotorandthelink

    wereusedtoderivetheequations.First,itwasassumedthatthekineticenergyofthe

    rotorwasduetoitsownrotation;thesecondassumptionwasthattherotor/gearinertia

    wassymmetricabouttherotoraxisofrotation.Twodifferentcontrollawswerethen

    examined.Thefirstonewasthefeedbacklinearizationcontrol,andthesecondonewas

    correctivecontrol,basedontheintegralmanifoldlaw.Theresultsindicatedthatthe

    correctivecontrollawdemonstratedbettertrajectorytracking.

    Yuetalpointedoutthattheirexperimentalinvestigationsindicatedthatthejoint

    elasticityhadtobeconsideredinthedynamicsofmanipulators,especiallywhensome

    components,suchasharmonicdriveswereincludedinthejoint[5].Ontheotherhand,

    considerationofjointelasticitycomplicatedthedynamicequationsinsuchamanner

    thatthecurrentstrategiesforrigidjointrobotscontrolcouldnotbeuseddirectly.Inthis

    paper,themodeldevelopedbySpong[4]wasadopted.Anadaptiveobservercontrol

    wasthenappliedforthetrajectorytrackingoftherobotmanipulatorswithflexiblejoint.

    Thesimulationresultsshowedthattheestimatedandactuallinkpositionswerewell

    matchedwiththedesiredlinkposition.

  • 7/29/2019 Robot Hydraulic Control

    19/108

    4

    Albu-SchafferandHirzingerincludedthedampingeffectsofthetransmissiondevicesin

    thedynamicmodelproposedbySpong[6].Differentcontroltechniqueswerebriefly

    discussed.Itwaspointedoutthatsometechniques,suchasfeedbacklinearization,were

    theoreticallycomplete,buttheirimplementationwereextremelydifficultduetothe

    measuringand/orcomputingofallthestatevariables.Theproposedcontrolstrategy

    startedusingthestatefeedbackcontroller;thecontrollergraduallywasextendedby

    addingmoredetailedrobotdynamics.Theresultsprovedthatthevibrationscausedby

    jointflexibilitywereeffectivelydampedbyusingtheproposedcontroltechnique.

    Thummeletalshowedthatvibrationinducedbyelasticjointscouldbesignificantly

    reducedbyusingfeedforwardcontrol,basedoninversedynamicmodels[7].The

    modelconsistedoftheinertiaofmotorandlinkconnectingbythemeansofaspringand

    adamperthatdescribedtheelasticityanddampingofthegearbox,respectively.Also

    velocitydependent,frictiontorquesinthebearingsforbothmotorsideandlinkside

    wereincluded.Introducingamorerealisticmodel,theelasticityinthejointwas

    modeledasanonlinearspring.Somecontrolalgorithmswerethenreviewed.By

    extendingthefeedforwardpartofthecontrollerandusingthemorecompleterobot

    dynamics,smalleramplitudeofvibrationwasexperienced.

    Habibietalmodeledalargehydraulicrobottoderiveageneralmathematicalfunction

    forhydraulicactuationmanipulators[8].Thepaperfocusedonthedynamicsof

    servovalveandactuatorbywritingtheflowequationsrelatedtothesecomponents.A

    comparisonwasmadeamongthedifferentflowterms.Itwasconcludedthatthe

    compressionflowsweresignificantandsubsequently,theywereincludedinthederived

    model.

    Insummary,itcanbeseenthatsomeresearchworkhasbeendevotedtothecontrolof

    theflexiblejointmanipulators,butmoreworkneedstobedonetofullyunderstandthe

    performanceoftheflexiblejointmanipulators.Itisanticipatedthatsignificant

    advancementsinthisareamightoccurinnextdecades,inadvancingtheunderstanding

  • 7/29/2019 Robot Hydraulic Control

    20/108

    5

    ofthefundamentalsofflexiblejointsandindevelopingpracticalandcosteffective

    controllawsforsuchmanipulators.

    Oneoftheconstraintsthatresearchershaveindevelopingcontrolalgorithmsforflexible

    jointsisverifyingtheapproachesexperimentally.Atpresent,commercialflexiblejoints

    haveeithernoadjustmentcapabilityoriftheydo,requiremechanicaldeviceswhich

    mustbereplacedexternally.Itishighlydesirabletohaveaflexiblejointinwhichthe

    operatingcharacteristicsofthejointsuchasdampingratioandnaturalfrequencycanbe

    adjustedonthefly.Thiswouldallowcontrolalgorithmswhichhavebeendesignedto

    reducetheeffectsofflexibilityinjoints,tobetestedunderawiderangeofconditions

    andtoconditionswhichchangeintimecontinuously.Designingaroboticjointwhich

    wouldallowadjustableflexibilitythusbecamethefocusofthisresearch.

    1.3Objectives

    Theobjectiveofthisresearchprojectwastodesignandimplementanexperimental

    hydraulicactuatedrobotmanipulatorthatsimulatesrobotjointflexibilityinthespecific

    rangeoffrequenciesanddampingratiowiththeabilityofchangingthejointflexibilitys

    parameters.Thedampingratioforsuchmanipulatorsisnormallylessthan0.1andthe

    naturalfrequencybetween5to15Hz[9].Theapproachthatwastakeninthisresearch

    isasfollows.Amodelbasedtransferfunctionwhichrepresentsthedynamicsofflexible

    jointmanipulatorswasmodeled.Applyinginputimpulsesignalstothetransferfunction,

    anoscillatoryangularaccelerationoutputsignalwasderivedbyusingcomputerbased

    simulation.Forthisresearchproject,thismodelbasedarrangementwasdefinedasa

    functiongenerator.Theoutputofthefunctiongeneratorwasemployedtodrivean

    experimentalhydraulicsystemwhichconsistedofaservovalveandarotaryhydraulic

    actuatorconnectedtoasingle-rigid-linkmanipulator.Thisexperimentalarrangement

    wasdefinedasahydraulicsimulator.Theaimofthecontrollerwhichdrivesthe

    systemwastomakethehydraulicsimulatorreproducethefunctiongeneratorsignalin

    thespecificrangeoffrequenciesanddampingratioofthefunctiongeneratorsignal.

  • 7/29/2019 Robot Hydraulic Control

    21/108

    6

    1.4ThesisOutline

    Thissectionincludesanoutlineofthisthesis.Chapter2introducesthedynamicmodel

    andsimulationofaflexiblejointrobot.Chapter3analyzestheoreticallyanactual

    hydraulicactuatedmanipulator.InChapter4theexperimentalsetupisincluded.

    Chapter5showsanddiscussestheexperimentalresultsandfinallyChapter6concludes

    theresultsandrecommendspossiblefutureworks.

  • 7/29/2019 Robot Hydraulic Control

    22/108

    7

    CHAPTER2

    FUNCTIONGENERATOR

    2.1Introduction

    Criticaltothedevelopmentoftheflexiblejointsimulatorwasthefunctiongenerator

    whichinputtheappropriatesignaltothehydraulicjoint.Thegoverningequationswhich

    describethebehavioroftheflexiblejointareconsideredinthisChapter.Theobjective

    ofthischapter,then,istopresentdynamicmodelingofasingle-rigid-linkrobot

    manipulatorwithflexiblejoint.Themodelisthensimulatedusingtypicalparameters

    foraflexiblejointinordertodemonstratethebehaviorofthelinkangularposition,

    velocity,andaccelerationforaspecifiedtorqueinput.

    2.2GoverningEquations

    Asingle-rigid-linkmanipulatorwitharevoluteflexiblejointisshowninFigure2.1.In

    thisfigure,theflexiblejointismodeledasatorsionalspringandatorsionaldamperin

    paralleltoeachother(forsimplicity,linearequivalentsymbolsofspringsanddampers

    areadopted).IandJarethelinkandthemotormomentsofinertia(withrespecttothe

    centerofrotation),respectively,uisthemotorinputtorque,Kisthejointstiffness,and

    Disthejointviscousdamping.Theangles andaretheangularpositionsofthelink

  • 7/29/2019 Robot Hydraulic Control

    23/108

    8

    Rotational

    SpringRotational

    DamperD

    K

    MotorJ

    Link

    I

    u

    (a)Linearequivalentschematicofthesingle-linkmanipulator

    (b)Dynamicequivalentschematicofthesingle-linkmanipulator

    Figure2.1-Schematicofasingle-linkmanipulatorwithflexiblejoint

  • 7/29/2019 Robot Hydraulic Control

    24/108

    9

    andthemotor,respectively.Thelinkandmotorangularpositionsarechosenasthe

    generalizedcoordinates.Thelinkmotionisassumedtobeinthehorizontalplane.

    Jointstiffnessanddampingareincludedinthedynamicsofthemodelpresentedhere.

    Forsimplicity,gravityeffectsarenotconsideredinthemodel;thisisconsistentwiththe

    experimentalsetup(hydraulicsimulator)usedinthisstudywhichisahorizontalplanar

    robotmanipulator.

    ForthelinkshowninFigure2.1thedynamicequationsareasfollow:

    0)()( =++ KDI &&&& ,(2.1)

    uKDJ =++ )()( &&&& ,(2.2)

    or:

    =

    +

    +

    uKK

    KK

    DD

    DD

    J

    I 0

    0

    0

    &

    &

    &&

    &&

    .(2.3)

    Theinputtothesinglelinkmanipulatoristorque uwhichisassumedtobeanimpulse

    signal.TakingtheLaplacetransformofequations(2.1)and(2.2),andassumingzero

    initialconditions,yields:

    0)]()([)]()([)(2 =++ ssKssssDsIs ,(2.4)

    )()]()([)]()([)(2 sUssKssssDsJs =++ .(2.5)

    Rearrangingequations(2.4)and(2.5)gives:

  • 7/29/2019 Robot Hydraulic Control

    25/108

    10

    0)()()()( 2 =+++ sKDssKDsIs ,(2.6)

    )()()()()( 2 sUsKDssKDsJs =+++ .(2.7)

    Solvingfor )(s inequation(2.6):

    )()(

    )(2

    sKDs

    KDsIss

    +

    ++= .(2.8)

    Substitutingequation(2.8)intoequation(2.7),yields:

    )()()()())((

    22

    sUsKDssKDs

    KDsIsKDsJs=+

    +

    ++++,(2.9)

    )()()(])())([( 222 sUKDssKDsKDsJsKDsIs +=+++++ ,(2.10)

    )()()(])()([ 234 sUKDssKsJIDsJIIJs +=++++ .(2.11)

    Finally,thetransferfunctionwhichrelateslinksrotation toinputtorqueuisas

    follows:

    ])()([)(

    )(22 KJIDsJIIJss

    KDs

    sU

    s

    ++++

    +=

    ,(2.12)

    or:

  • 7/29/2019 Robot Hydraulic Control

    26/108

    11

    22

    1*

    )()()(

    )(

    s

    J

    K

    I

    Ks

    J

    D

    I

    Ds

    IJ

    Ks

    IJ

    D

    sU

    s

    ++++

    +

    =

    .(2.13)

    Thisindicatesthattheundampednaturalfrequenciesofthesystemarezeroand

    )(J

    K

    I

    K+ .Zeronaturalfrequencyrepresentstherigidbodymotionofthebaseandthe

    link.

    Theblockdiagramoftheopen-loopsystembasedonequation(2.13)isshowninFigure

    2.2.Inthisfigure:

    IJ

    Kk =1 ,

    K

    D= , )(

    J

    K

    I

    Kn += and )

    11(

    1

    2 JIK

    D+=

    Figure2.2-Theopen-loopblockdiagramofsinglelinkmanipulatorwithflexiblejoint

    Itisevidentthatthesystemhastwointegratorswhichwillsignificantlyfilteroutany

    higherfrequencyperturbationswhichmightbeintroducedbythejoint.

    TheequationswereimplementedusingMATLAB.Inthissimulation,the

    manipulatorsphysicalparametershavebeentakenfrom[9]andarelistedinTable2.1:

    22

    1

    2

    )1(

    nnss

    sk

    ++

    +

    s

    1

    s

    1

  • 7/29/2019 Robot Hydraulic Control

    27/108

    12

    Table2.1-Simulationparametersadoptedfrom[9]

    Symbol Description Quantity Unit

    I Linkmomentofinertia 5.000 radsmN /2

    J Motormomentofinertia 5.000 radsmN /2

    K Jointstiffness 1.000E4 radmN /

    D Jointviscousdamping 31.63 radsmN /

    Applyingtheseparameterstoequation2.13yields:

    22

    1*

    400065.12

    )100316.0(400

    )(

    )(

    sss

    s

    sU

    s

    ++

    +=

    .(2.14)

    Comparingwiththestandardtransferfunction

    222

    1 1*2

    )1(

    )(

    )(

    sss

    sk

    sU

    s

    nn

    ++

    +=

    ,(2.15)

    thenaturalfrequencyanddampingratioareasfollows:

    sradn /25.634000 == =10.00Hzand 1.0= 0

    Itshouldbeemphasizedthattherangeofvaluesusedinthisprojectarequitecommon

    forflexiblejointrobotmanipulators.

    Theinputtothesimulationwasanapproximateimpulse(representingatorque

    impulse).SomeofthesimulationresultsareshowninFigure2.3.Inthisparticular

    simulationtwotimedelayedpulseinputtorqueswithzeroand1.5 sphasedelay,period

  • 7/29/2019 Robot Hydraulic Control

    28/108

    13

    of3s,pulsewidthof0.1%ofperiod,andamplitudeofu=+3000 mN andu=-3000

    mN havebeenused.Theoutputresultsshowninthefigurearethelinkangular

    position,velocity,andacceleration,respectively.Itshouldbenotedthatacontinuous-

    timeintegrationsolverofRunge-Kuttamethodwithzeroinitialconditionandafixed-

    stepsizeof0.001insecondshasbeenapplied.

    0

    0.5

    1

    1.5

    2

    2.5

    3

    0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

    Time(s)

    AngularPosition(rad)

    (a)Angularposition

  • 7/29/2019 Robot Hydraulic Control

    29/108

    14

    -1

    -0.5

    0

    0.5

    1

    1.5

    2

    0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

    Time(s)

    AngularVelocity(rad/s)

    (b).Angularvelocity

    -60

    -40

    -20

    0

    20

    40

    60

    0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

    Time(s)

    AngularAcceleration(rad/s2)

    (c)Angularacceleration

    Figure2.3-Simulationresultsforimpulseinputstothesingle-linkmanipulator

  • 7/29/2019 Robot Hydraulic Control

    30/108

    15

    Theresultsshowthatbyapplyinganimpulsetorqueinput,adampedoscillationis

    observableinangularaccelerationandvelocity,butthemagnitudeishighlyfilteredin

    angularpositionofthelink.Thesevibrationsmaybefelt,butnotobservedvisually.

    Sincetorqueonthejointisdirectlyrelatedtoangularacceleration,thelowdampingcan

    havesignificantimplicationswhencontrolisimplemented.

    Aclosedformsolutionofequations2.14or2.15doesexistandcanbesolvedby

    applyingPartial-FractionExpansion.Considerequation2.14:

    400065.12

    )100316.0(400

    )(

    )(2

    2

    ++

    +=

    ss

    s

    sU

    ss,(2.16)

    or

    )(400065.12

    400265.1)(

    2

    2sU

    ss

    sss

    ++

    += .(2.17)

    Inordertocreatethesameformoftheinputsignalasthesimulation(anapproximate

    impulse),twoseparatestepinputsareused:thefirstastepof3000N.matzerosteptime

    andasecondstepat0.003s(0.1%ofthe3speriodtime)steptimeandfinalvalueof

    -3000N.m.Inthiscase, s

    e

    ssU

    s003.030003000)(

    = andequation2.17becomes:

    ++

    +=

    s

    e

    sss

    sss

    s003.0

    2

    2 30003000

    400065.12

    400265.1)( ,(2.18)

    or:

    )()()400065.12(

    )12000003795(

    )400065.12(

    12000003795)( 212

    003.0

    2

    2 sFsFsss

    se

    sss

    sss

    s

    =

    ++

    +

    ++

    +=

    .(2.19)

    Knowingthat;

    )9285.62325.6)(9285.62325.6(400065.122 jsjsss +++=++ ,(2.20)

  • 7/29/2019 Robot Hydraulic Control

    31/108

    16

    andthePartial-FractionExpansionoffirsttermofequation2.19:

    400065.12)400065.12(

    120000037952

    321

    2++

    ++=

    ++

    +

    ss

    asa

    s

    a

    sss

    s,(2.21)

    multiplyingbothsidesbythedenominatorofitslefthandsideofequation2.20yields:

    sasassas 32

    2

    2

    1 )400065.12(12000003795 ++++=+ ,(2.22)

    or

    131

    2

    21 4000)65.12()(12000003795 asaasaas ++++=+ .(2.23)

    Comparingbothsidesofequation2.23,itcanbeshownthat:

    021 =+ aa ,(2.24)

    379565.12 31 =+ aa ,(2.25)

    12000004000 1 =a ,(2.26)

    .

    Solving,theparametersarefoundtobe:

    3001 =a , 3002 =a and 03 =a .

    Therefore:

    400065.12

    300300

    )400065.12(

    12000003795

    )( 221 ++=++

    +

    = ss

    s

    ssss

    s

    sF ,(2.27)

    or:

  • 7/29/2019 Robot Hydraulic Control

    32/108

    17

    22219285.62)325.6(

    )9285.62(1.0)325.6(300)

    1(300

    )400065.12(

    12000003795)(

    ++

    +=

    ++

    +=

    s

    s

    ssss

    ssF .(2.28)

    Considering )()( 1003.

    2 sFesFs

    = ,usingequation2.28andapplyingtheinverseLaplace

    transformtoequation2.19yields:

    ( ) )003.0()]003.0(9285.621.0)003.0(9285.62[300300)()9285.621.09285.62(300300)(

    )003.0(325.6

    325.6

    =

    tutSintCose

    tutSintCoset

    t

    t&&

    (2.29)

    Figure2.4showstheplotoftheanalyticalsolutionofequation2.29.

    0 0.5 1 1.5-40

    -30

    -20

    -10

    0

    10

    20

    30

    40

    50

    Time(s)

    AngularAccele

    ration(rad/s2)

    Figure2.4-Plotofequation2.29.

  • 7/29/2019 Robot Hydraulic Control

    33/108

    18

    Thetwosolutionsforaccelerationresponseusingthecomputersolverandtheclosed

    solutionformyieldthesametheresultsaswouldbeexpected.Thefunctiongenerator

    canberealizedeitherfromthesimulationusingMatlaborbyimplementingtheclosed

    solutionformofequation2.29.Inthiswork,thesimulationformwasusedbecauseof

    theeaseofchangingtheparametervalues.

  • 7/29/2019 Robot Hydraulic Control

    34/108

    19

    CHAPTER3

    HYDRAULICSIMULATOR

    3.1Introduction

    Thehydraulicsimulatorisessentiallytheexperimentalhydraulicsystemwhichreceives

    theinputsignalfromthefunctiongeneratorandoutputsanaccelerationwhichis

    controlledtofollowthedesiredinputsignal.Themainreasonforchoosingahydraulic

    actuationsystemwasbecauseofthehighervalueofitsnaturalfrequencyincomparison

    totheusualrobotmanipulatorsnaturalfrequency.Hydraulicsystemsarealsoreadily

    foundinmanyroboticapplicationswhichmakecompatibilityissuesmarginalized.It

    wasrecognizedthatthedesignofasystemtoreproduceaccuratelysuchwaveforms

    wouldbeachallengingprocess,giventhataccelerationtendstobenoisyanddifficultto

    control.Aswillbeillustratedlater,thepresenceofaseverenonlinearity(suspectedto

    bedead-band)intherotarymotorcompoundedthecontrolproblem.Figure3.1shows

    theoverallarrangementofthesystem.

  • 7/29/2019 Robot Hydraulic Control

    35/108

    20

    -60

    -40

    -20

    0

    20

    40

    60

    0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

    Time(s)

    AngularAcceleration(

    rad/s.sq)

    Figure3.1-Schematicofthehydraulicsimulator

    ThesystemshowninFigure3.1consistedofapressurecompensatedhydraulicpumpdrivenbyanelectricmotor,apressurecontrolservovalve(MOOGmodel15-010),

    andarotarymotor(MICRO-PRECISIONROTACmodelMPJ-22-1V),whichdrovea

    single-linkmanipulator.Therotarymotorcouldrotatethelink270inthehorizontal

    plane.

    Themeasuringinstrumentsconsistedofanaccelerationsensor(Bruel&Kjaermodel

    4370)andamplifier(Bruel&Kjaerchargeamplifiertype2635).Theexperimental

    systemwascontrolledbyaPCcomputerusingMatlab-Simulinksoftware.

    Aflowcontrolservovalvewasinitiallyconsideredbutbecausetheflowrateswerevery

    small,becausethecut-offfrequencyofpressureservovalvesisknowntobeveryhigh

    (relativetoaflowcontrolservovalve),andbecausepressure(force)isdirectlyrelatedto

    acceleration,apressurecontrolservovalvewaschosenasthehydrauliccontrolling

    device.AllcomponentswereofftheshelffromtheFluidPowerLaboratoryatthe

    UniversityofSaskatchewanandhence,thisbecameaconstraintforthehydraulic

    systemdesign.

  • 7/29/2019 Robot Hydraulic Control

    36/108

    21

    3.2HydraulicSystemAnalysis

    Asstatedabove,thefunctionofthehydraulicsimulatorwastobeabletoreproducethe

    inputsignalwhichinthisparticularcasewastheangularaccelerationsignalcreatedby

    thefunctiongenerator.Tofacilitatethedesignofacontroller,ananalysisofthe

    hydraulicsystemwasperformed.

    Acommonformofatransferfunctionapproximationoftherelationshipbetweenthe

    loaddifferentialpressureandtheinputcurrenttotheservovalvehasbeenderivedtobe

    [10]:

    12)(

    )(2

    +

    +

    =

    ss

    K

    sI

    sP

    nv

    v

    nv

    v

    e

    L

    ,(3.1)

    inwhich:

    LP loaddifferentialpressure(Pa) eI inputcurrent(ma)

    vK valvegain(Pa/ma)

    nv valveundampednaturalfrequency(Hz)

    v valvedampingratio

    sLaplaceoperator(1/s).

    Thetorqueonthehydraulicrotarymotorisgivenby;

    +== LmLL IDPT ,(3.2)

  • 7/29/2019 Robot Hydraulic Control

    37/108

    22

    inwhich:

    LT motortorque( mN )

    mD

    motorvolumetricdisplacement )/(

    3 radm

    LI linkmomentofinertia )/(2 radsmN

    motorviscousdamping )/( radsmN

    linkangularvelocity(rad/s)

    linkangularacceleration )/( 2srad ,

    TakingtheLaplacetransformofequation3.2yields;

    )1()(

    )(

    +

    =

    sI

    D

    sP

    s

    L

    m

    L

    ,(3.3)

    or:

    )1()(

    )(

    +

    =

    sI

    sD

    sP

    s

    L

    m

    L

    .(3.4)

    let;

    mDK =1 and

    L

    I=1

    therefore;

    1)(

    )(

    1

    1

    +

    =

    s

    sK

    sP

    s

    L

    ,(3.5)

    inwhich;

    1K loadgain( sParad / )

  • 7/29/2019 Robot Hydraulic Control

    38/108

    23

    1 loadtimeconstant(s).

    Multiplyingequations3.1and3.5,gives;

    11

    2)(

    )(

    1

    1

    2+

    +

    +

    =

    s

    sK

    ss

    K

    sI

    s

    nv

    v

    nv

    v

    e

    .(3.6)

    Thefollowingequationisvalidfortheservo-amplifierwhichshouldprovideDCcurrent

    intotheservovalve:

    Lt

    di

    Riv

    e

    ee

    ''+=

    ,(3.7)

    inwhich:

    ev servo-amplifierinputvoltage(v)

    '

    ei servo-amplifieroutputcurrent( a)

    Rservo-amplifierresistance()

    Lservo-amplifierinductance(H).

    TakingtheLaplacetransformoftheequation3.7yields;

    1

    1

    )(

    )('

    +

    =

    sR

    LR

    sV

    sI

    e

    e ,(3.8)

    LetR

    Ka1

    = andR

    La = ,therefore;

    1)(

    )('

    +=

    s

    K

    sV

    sI

    a

    a

    e

    e

    ,(3.9)

    inwhich;

  • 7/29/2019 Robot Hydraulic Control

    39/108

    24

    aK servo-amplifiergain(a/v)

    a servo-amplifiertimeconstant(s)

    Itshouldbenotedthataccordingto[11]fortheservo-amplifier, R=2000,L=9.7H.

    Therefore vaKa /0005.0= and sa 00485.0= .

    Sincetheunitofservo-amplifieroutputcurrent( )('sIe )isampereandthatofservovalve

    inputcurrent( )(sIe )ismilliampere,aconversiongainof eK isintroducedtoconvert

    theunits.Therefore, amaKe /1000= .Applyingthegaintoequation3.9gives;

    1)(

    )(

    +=

    s

    KK

    sV

    sI

    a

    ea

    e

    e

    ,(3.10)

    inwhich;

    eK servo-amplifierconversiongain(ma/a).

    Multiplyingequations3.6and3.10yields;

    11

    21)(

    )(

    1

    1

    2+

    +

    +

    +=

    s

    sK

    ss

    K

    s

    KK

    sV

    s

    nv

    v

    nv

    v

    a

    ea

    e

    (3.11)

    Theblockdiagramofequation3.11isshowninFigure3.2.

  • 7/29/2019 Robot Hydraulic Control

    40/108

    25

    Figure3.2-Blockdiagramofequation3.11.

    Itshouldbenotedthat )(s hasunitsofrad/s2.

    Since a 1 ,theservo-amplifiertransferfunctioncanbereducedto eaKK .Thisreduced

    formisillustratedinFigure3.3:

    Figure3.3-Reducedblockdiagramofequation3.11

    Nowtwopossiblescenariosareexpectedwithrespecttothetransferfunctionsofthe

    servovalveandtheload.Thefirstscenariooccurswhen nv

    1

    1,where nv isnowin

    Hztoensureunitconsistency.Atypicalmagnitudeasymptoticapproximation

    (magnitudeBodeplot)forthisfirstcaseisshowninFigure3.4.

    LoadServovalveServoamplifier

    )(sVe

    )(sPL )(sIe

    1

    22

    +

    +

    s

    s

    K

    nv

    v

    nv

    v

    11

    1

    +s

    sK

    )(s

    1+s

    KK

    a

    ea

    LoadServovalveServoamplifier

    )(sVe )(sPL )(sIe

    12

    2

    +

    +

    s

    s

    K

    nv

    v

    nv

    v

    11

    1

    +s

    sK

    )(s

    eaKK

  • 7/29/2019 Robot Hydraulic Control

    41/108

    26

    Figure3.4-TypicalmagnitudeBodeplotofequation3.11whennv

    1

    1

    Thesecondscenariohappenswhen nv

    1

    1;atypicalBodeplotforthiscaseis

    demonstratedinFigure3.5.

    Figure3.5-TypicalmagnitudeBodeplotofequation3.11when nv

    1

    1

    1

    1

    nv

    Slope-20

    (dB/decade)

    Slope+20

    (dB/decade)

    )(Hz

    )(dB

    Ve

    Slope-40

    (dB/decade)

    )(Hz

    )(dB

    Ve

    Slope-40

    (dB/decade)

    Slope+20

    (dB/decade)

    1

    1

    nv

  • 7/29/2019 Robot Hydraulic Control

    42/108

    27

    TheactualfrequencyresponseofthehydraulicsimulatorisshowninFigure3.6,where

    theinputisthesignalfromtheanalyzerunitandtheoutputisthemeasuredsignalfrom

    theaccelerometer.DetailsonobtainingthisplotaregiveninChapter5.

    -40

    -30

    -20

    -10

    0

    10

    20

    0.1 1 10 100

    Frequency(Hz)

    Figure3.6-Open-loopmagnitudefrequencyresponseofthehydraulicsimulator

    (points)withsuperimposedasymptoticlines(experimental:solidlinebasedon40dB

    perdecadeandtheoretical:dashedlinebasedon20dBperdecade)

    ItshouldbenotedthatthetwooddpointsatlowerfrequencieswereignoredinFigure

    3.6.

    Acomparisonbetweentheexperimentalresultandthetheoreticalscenariosshowsthat

    thefirstpossiblescenario( nv

    1

    1)betterdescribestheactualexperimentalsetupfor

    frequencieslessthan20Hz.

    Itshouldbenotedthattheslopeofasymptoticline(solidline)inFigure3.6is+40dB

    perdecadewhichisdifferentfromthatofthetheoreticalmodel(dashedline)ofFigure

    3.6.Measurementoftheoutputangularaccelerationwasaccomplishedusingan

    accelerometer(tobediscussedingreaterdetailbelowandinChapter4).Itwasbelieved

    )(

    )(

    )(

    dB

    jV

    j

    e

  • 7/29/2019 Robot Hydraulic Control

    43/108

    28

    thatthepoorperformanceoftheaccelerometeratlowfrequenciescausedthe

    discrepancy.Inthenexttwochaptersitwillbeshownthatanaccelerationcompensator

    couldbeusedtocorrecttheperformanceoftheaccelerometeratlowfrequencies.

    Theseresultsimplythat,1

    1

    nv ,andthereforethemodelcouldbeeasilyreducedto

    thefirstordertermofthetransferfunctionofequation3.11Hence,

    1)(

    )(

    +=

    s

    sK

    sV

    s

    o

    o

    OLe

    ,(OLinthiscasemeansopen-loop.)(3.12)

    inwhich;

    oK open-loopgain( svrad / )

    o open-looptimeconstant(s).

    Itshouldbenotedthat 1KKKKK veao = and 1 =o .Thereforetheopen-loopblock

    diagramissimplifiedtothatshowninFigure3.7.

    Figure3.7-Simplifiedopen-loopblockdiagramofthesystem

    Intheaforementionedequations, hasaunitofrad/s2.Sincealinearaccelerometer

    wasusedtoindirectlymeasuretheangularaccelerationofthejoint,alinearsensitivity

    andcalibrationfactorwhichconvertsrad/s2tovoltagewasintroduced.Thecalibrated

    outputratingofthelinearaccelerometerwasdeterminedtobe )/

    (1.02sm

    vS l = .The

    linearaccelerometerwascapableofmeasuringtangentialacceleration.However,itwas

    theangularaccelerationthatwasdesired.Sincethedistancebetweenthecenterofthe

    installedlinearaccelerometerandtheaxisoftherotation(rotationarm)wasknownvery

    accurately(l=0.435m),itwaspossibletoconvertthelineartangentialaccelerationto

    )(sVe

    1+s

    sK

    o

    o

    )(s

  • 7/29/2019 Robot Hydraulic Control

    44/108

    29

    angularacceleration.Bymultiplyingthelinearsensitivity lS byl,theangularsensitivity

    (theconversiongain)oftheaccelerometerwasobtainedas:

    )/

    (0435.0)(435.0)./

    (1.0.22 srad

    vmsmvlSK lcon === (3.13)

    Theeffectofthe conK isshownintheblockdiagramofFigure3.8;

    Figure3.8-Blockdiagramofthesimplifiedtransferfunctionwithinfluenceofthe

    accelerometersensitivity

    IntheblockdiagramofFigure3.8;

    conK conversiongainfactor( 2/srad

    v

    )

    d desiredangularacceleration(functiongeneratoroutput)(2/srad )

    linkangularacceleration( 2/srad )

    v linkangularaccelerationinvolts(v),

    therefore;

    1)(

    )(

    +=

    s

    sK

    sV

    s

    o

    OL

    OLe

    v

    ,(3.14)

    inwhich;

    OLK open-loopoverallgain(s).

    Itshouldbenotedthat conoOL KKK = .

    )(sv )(s

    )(sVe )(sd

    conK

    1+s

    sK

    o

    o

    conK

    Function

    Generator

  • 7/29/2019 Robot Hydraulic Control

    45/108

    30

    3.3ControllerDesign(Theoretical)

    Todesignthecontrollerwhichwouldprovidetherequiredclosed-loopresponse,the

    firststepwastoconsiderthehydraulicsimulatorinamathematicalformandtodesign

    thecontrollerforthissystem.Thesecondstepwastoapplythiscontrollertothe

    experimentalhydraulicsimulatorandtotunethecontrollertooptimizethe

    performance(seeChapter5).

    Inthissection,aclosed-loopcontrolsystembasedonatheoreticalmodelisstudied.The

    closed-looptransferfunctionisderivedanditsparametersarecomparedwiththeopen-

    looptransferfunction.Anapproachofcascadeinversecompensationforimprovingthe

    open-loopsystemratherthanusingaclosed-loopcontrollerisintroduced.

    3.3.1Closed-LoopController

    Theobjectiveofanycontrolleristoforcetheoutputtofollowtheinputsuchthat

    CLe

    v

    sV

    s

    )(

    )(=1(CLinthiscasemeansclosed-loop).Thismeansthehydraulicsimulator

    wouldfollowexactlythedesiredsignalfromthefunctiongenerator.Thissection

    considerstheimplementationofseveralclassicalcontrollersinanattempttomeetthe

    aboveobjective.Forallanalysis,thesimplifiedopen-looptransferfunctionofEquation

    3.14isassumed.

    Asimpleproportionalcontrollerwasfirstexamined.Theclosed-loopblockdiagramis

    illustratedinFigure3.9.

  • 7/29/2019 Robot Hydraulic Control

    46/108

    31

    Figure3.9-Blockdiagramofclosed-loopcontrolsystemwithproportionalgain

    InFigure3.9;

    PK proportionalgain

    Theclosed-looptransferfunctionisderivedwithrespecttotheblockdiagramshownin

    Figure3.9asfollows;

    11

    1

    )(

    )(

    1

    ++

    +=

    s

    sKK

    s

    sKK

    sV

    s

    o

    OLp

    o

    OLp

    CLe

    v

    ,(3.15)

    whichsimplifiesto:

    1)(

    )(

    1

    1

    1

    +=

    s

    sK

    sV

    s

    CL

    CL

    CLe

    v

    ,(3.16)

    inwhich; OLpCL KKK =1 ,and OLpoCL KK+= 1 .

    Itisclearthattheobjectiveofhaving 1)(

    )(

    1

    =

    CLe

    v

    sV

    sisnotsatisfiedatallfrequencies

    withthiscontroller.Table3.1providesacomparisonoftheopen-loopandclosed-loop

    transferfunctionparameters.

    )(sv

    +

    _

    )(sVe

    1+s

    sK

    o

    OL

    pK

  • 7/29/2019 Robot Hydraulic Control

    47/108

    32

    Table3.1-Parameterssummaryofopen-loopandclosed-looptransferfunctions

    Open-loop Closed-loop

    Gain conoOL KKK=

    OLPCL KKK=

    1

    Timeconstanto OLPoCL KK+= 1

    TheoreticallythisTableillustratesthattheclosed-loopcontrolgain(if 1PK )andtime

    constantareincreasedincomparisontotheopen-loopsystem.Usingtypicalparameter

    values,acomparisonoftheopen-loopandclosed-loopfrequencyresponses(magnitude)

    fortwocontrollergainsisshowninFigure3.10.

    10-2

    10-1

    100

    101

    102

    -60

    -50

    -40

    -30

    -20

    -10

    0

    Magnitude(dB)

    Frequency(Hz)

    Open-Loop

    Closed-Loop

    Gain=2.0

    Gain=1.5

    Figure3.10-Magnitudefrequencyresponse)(

    )(

    sV

    s

    e

    v oftheopen-loopandclosed-loop

    systemswithproportionalcontroller

    Figure3.10indicatesthattheclosed-loopcontrolwithproportionalgainincreasedthe

    bandwidthofthesystem,butonlymarginallyimprovedthemagnituderatio.Itshould

  • 7/29/2019 Robot Hydraulic Control

    48/108

    33

    benotedthatinexperiments,therewasalimitedroomforincreasingtheproportional

    gainbecauseofservovalvesaturation.

    TypicalvaluesofKOLand o wereextractedfromtheexperimentalfrequencyresponse

    oftheactualsystem(Figure3.6): sKOL 025.0= and so 064.0= .

    Theabilityofthesystem(open-loopandclosed-loop)tofollowtheinputwas

    demonstratedinthetimedomainbyinputtinga10 Hz.jointvibration(theoutputfrom

    thefunctiongenerator)intothesystem.Figure3.11comparestheresponsesoftheopen-

    loopandclosed-loopmodels.Itisevidentthattheclosed-loopcontrollermarginally

    increasestheamplitudeoftheoutputaccelerationwhencomparedtotheopen-loop

    system.Thisisconsistentwiththegainincreaseinthefrequencyresponseshownin

    Figure3.10at10Hz.Increasingthegaineventuallysaturatestheinputsignaltothe

    servovalveandnofurtherimprovementwaspossible.Itisclearthatthisisnotan

    acceptableresponse.

    1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2-2.5

    -2

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    2

    Time(s)

    Accelerationamplitude(v)

    FunctionGenerator

    Closed-Loop

    Open-Loop

    Figure3.11-Transientresponseof v intheopen-loopandclosed-loopmodelswith

    2=PK tothefunctiongenerator d of10Hzsignal

  • 7/29/2019 Robot Hydraulic Control

    49/108

    34

    AclassicalPIcontrollerwasthenimplemented.Theclosed-loopblockdiagramis

    showninFigure3.12.

    Figure3.12-Blockdiagramofclosed-loopcontrolsystemwithproportionaland

    integralgains

    InFigure3.12;

    PK proportionalgain

    IK integralgain(1/s)

    Theclosed-looptransferfunctioninthiscaseisderivedwithrespecttotheblock

    diagramshowninFigure3.12asfollows;

    1

    )(1

    1

    )(

    )(

    )(

    2

    +

    ++

    +

    +

    =

    s

    KsKK

    s

    KsKK

    sV

    s

    o

    IPOL

    o

    IPOL

    CLe

    v

    ,(3.17)

    whichissimplifiedto;

    )1()(

    )(

    )(

    )(

    2

    +++

    +=

    OLIoOLP

    IPOL

    CLe

    v

    KKsKK

    KsKK

    sV

    s

    .(3.18)

    __

    + )(sv

    s

    KK IP +

    1+s

    sK

    o

    OL

    )(sVe

  • 7/29/2019 Robot Hydraulic Control

    50/108

    35

    Usingthesamegainandtimeconstantvaluesasbefore,

    i.e. sKOL 025.0= and so 064.0= ,acomparisonoftheopen-loopandclosed-loop

    frequencyresponses(magnitude)fortwocontrollergainsisshowninFigure3.13.

    10-2

    10-1

    100

    101

    102

    -60

    -50

    -40

    -30

    -20

    -10

    0

    Magnitude(dB)

    Frequency(Hz)

    Open-Loop

    Closed-Loop

    P=2.0,I=10

    P=1.5,I=5

    Figure3.13-Magnitudefrequencyresponse)(

    )(

    sV

    s

    e

    v ofopen-loopandclosed-loop

    systemswithPIcontroller.

    ItcanbeobservedfromFigure3.13thattheproposedPIcontrollercouldimprovethe

    responseatlowerfrequenciesbetterthantheproportionalcontroller,butthemagnitude

    ratiowasstillaproblem.

    Figure3.14comparesthetimeresponsesoftheopen-loopandclosed-loopmodelsfora

    10Hzjointfunctiongeneratorsignal.

  • 7/29/2019 Robot Hydraulic Control

    51/108

    36

    1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2-2.5

    -2

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    2

    Time(s)

    Accelerationamplitude(v)

    FunctionGenerator

    Closed-Loop

    Open-Loop

    Figure3.14-Transientresponseof v intheopen-loopandclosed-loopmodelswith

    2=PK and 10=IK tothefunctiongenerator d of10Hzsignal

    ItisevidentfromFigure3.14thatthePIcontrolleralsomarginallyimprovesthat

    amplitudeoftheoutputbutwasnotsufficientforthefrequencyrangeofinterestinthis

    study.

    InChapter5itwillbeshownthattheP(proportional)orPI(proportional-integrator)

    closed-loopcontrollerscouldnotbeimplementedphysicallybecauseofnoiseinthe

    accelerometer(sensor),significantnonlinearitiesintherotarymotorandhighsystem

    internalgains.Thereforeanalternateapproachwasproposedandisnowconsidered.

  • 7/29/2019 Robot Hydraulic Control

    52/108

    37

    3.3.2Open-LoopInverseCompensationController

    Toachievethestatedobjectiveatthebeginningofthissection,acascadecompensator

    controllerwasproposedwhoseformwassimplytheinverseoftheopen-looptransfer

    functionshowninequation3.14.Ontheotherhand,sincetheplanttransferfunction

    includedazeroattheorigin(Root-locusapproach),thecontrollerdeemedtobean

    exampleofmarginallystablepole-zerocancellationwhichmightmakethesystem

    unstableinsomeconditions(e.g.withconstantdisturbanceornoiseintheinputsignal).

    Thereforethecontrollerpolewasdesignedtobeontherealaxes,atthelefthandsideof

    imaginaryaxesandveryclosetotheorigin.Thispoleusuallyshouldbeabout

    o /1.0= awayfromtheoriginsothatthemagnifyingsignaldoesnotexceedmore

    than10timesoftheconstantinputnoise.Theblockdiagramofthecontrolleris

    illustratedinFigure3.15.

    Figure3.15-Blockdiagramofopen-loopcompensatorcontroller

    Itshouldbenotedthattheopen-looppureinversecompensatorwhichwasproposed

    earliermightsaturatethehydraulicsimulatorinfrequencieslowerthano

    1(2.5Hzor

    15.71rad/sbasedonactualsystemmagnitudeBodeplotshowninFigure3.6),however

    itwasprovedbysimulationthatthepureinversecontrollerworkedproperlyinthe

    frequencyrangeofinterestwhichwas5to15 Hz.Therefore wassettozeroandthe

    controllerwasadoptedtobeintheformofsK

    s

    OL

    o 1+ .Thetransferfunctionofthesystem

    becomes

    OLe

    v

    sV

    s

    )(

    )(=1,whichdoesmeettheoriginalobjective.Thecontrollermustnot

    PlantController

    )(

    1

    +

    +

    sK

    s

    OL

    o

    1+s

    sK

    o

    OL

    )(sVe )(sv

  • 7/29/2019 Robot Hydraulic Control

    53/108

    38

    outputasignalwhichwouldsaturatetheinputtothehydraulicsimulator.Theformof

    thecontrollerisquiteidealinthatitinherentlylimitstheinputsignaltotheplant

    (assumingofcoursethatthelimitislessthansaturationvalues).Thedisadvantageof

    thiscontrolleristhatitisopen-loopandanysmalldriftintheplantinputwouldbe

    integratedtwicebytheplantresultinginsubstantialoutputpositiondrift.To

    demonstratethetimeresponseofthesystemwiththisinversecontroller,thesamesignal

    fromthefunctiongeneratorwasinputtothecompensatedservosystemandasexpected,

    theoutputaccelerationfollowsthedesiredinputsignalperfectly.Thisisshownin

    Figure3.16.

    1.5 1.6 1.7 1.8 1.9 2 2.1 2.2-2.5

    -2

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.52

    2.5

    Time(s)

    Accelerationamplitude(v)

    FunctionGenerator

    Compensatedopen-loop

    Figure3.16-Transientresponseof v intheopen-loopcompensatedmodeltothe

    functiongenerator d of10Hzsignal

  • 7/29/2019 Robot Hydraulic Control

    54/108

    39

    CHAPTER4

    ExperimentalSystemSetup

    4.1Introduction

    InChapter2,thedevelopmentofthefunctiongeneratorwasconsidered.Inthis

    chaptertheexperimentalsystem(hydraulicsimulator)setupisexplainedindetail.

    Figure4.1illustratestheschematicoftheexperimentalsystemwhichincludesthe

    powersupply,pressureservovalve,hydraulicrotarymotor,andasinglelink

    manipulator.

    Figure4.1-Schematicofthehydraulicsimulator

  • 7/29/2019 Robot Hydraulic Control

    55/108

    40

    ThesystemwascontrolledbymeansofapersonalcomputerusingMatlab-Simulink

    softwarewhichsubmittedcontrolledinputsignalstothepressureservovalve.For

    analyzingtheexperimentalsystem,aSignalAnalyzerUnitwasused.Alsotheangular

    accelerationofthelinkwasmeasuredbyanaccelerometer.Inthefollowingsections,

    eachpartandcomponentisconsideredindetail.

    4.2PowerSupply

    Thepowersupplyconsistedofahydraulicpumpandapressurereducingvalve.The

    hydraulicpumpwasavariabledisplacement,pressurecompensatedSundstrand22

    seriespumpwhichdeliveredamaximumflowrateof121.28L/min.(32GPMUS)at

    1740RPM.Thedead-headpressureofthepumpwassetat172.5bar(2500psi).A

    DenisonmodelRR12535pressurereducingvalvelimitedthedownstreamsystem

    pressureto86.25bar(1250psi).

    4.3PressureControlServovalve

    AsdiscussedinChapter3,apressureservovalvewasusedbecauseofitshighfrequency

    response,animportantdesigncriterioninthissystem.TheElectrohydraulic,two-stage,

    four-way,nozzle-flappertypepressurecontrolservovalveofMoogmodel15-010was

    utilizedtodirecttheflowfromthepowersupplytotherotarymotor.Figure4.2showsa

    schematicoftheservovalve.

  • 7/29/2019 Robot Hydraulic Control

    56/108

    41

    Figure4.2-Schematicofpressurecontrolservovalve

    Theservovalveconsistedofthreemainparts:torquemotor,hydraulicamplifierand

    valvespoolassembly.Thistypeofservovalveisdesignedtocontrolloadpressure

    difference( ABP ).Asummaryofitsoperationisasfollows.

    Theinputcurrentcreatesmagneticforcesontheendsofthearmatureinthetorque

    motorcoils.Thiscausesthearmatureandflapperassemblytorotateaboutaflexuretube

    support(notshowninFigure4.2).Theflappermovesbetweenthenozzlesandbuildsup

    adifferentialpressure 12P whichisproportionaltotorquecreatedbytheinputcurrent.

    Theflapperrestrictsflowpassingthroughoneofthenozzles,forexamplenozzle(1)in

    Figure4.2.Becauseofthefixedorificeupstreamtothenozzle,areductioninflow

    Nozzle21

    SP

    T

    SP SP

    F

    l

    ap

    p

    e

    r

    SP

    A B

    PermanentMagnet

    RotatingArmature

    SA

    AA

  • 7/29/2019 Robot Hydraulic Control

    57/108

    42

    resultsinanincreaseinthepressurejustdownstreamofthefixedorificewhichis

    sensedattheendofthespool(lefthandside).Meanwhile,flowontheothersideofthe

    nozzleincreases(wideropeningandlessrestriction)whichmeansthatthepressureon

    theupstreamsideofthenozzle(downstreamfromthefixedorificeontherighthand

    side)decreasesandissensedbytherighthandsideofthespool.Apressuredifferential

    acrosstheendsofthespoolnowexistsandhence,thespoolmovestotheright(inthis

    case).Fluidfromthesupplypressure( SP )isnowportedtoonecontrolportandfluidin

    thesecondoutletisportedtothetank(T).Asflowisportedtotheloadfrom SP ,the

    pressurebuildsup(decreasesontheothersideoftheactuator)andisfedbacktothe

    righthandsideofthespool(thedownstreamsidepressureisfedbacktothelefthand

    sideofthespool).Theloadpressuredifference( ABP )buildsupafeedbackforcewhich

    eventuallybecomesequalto SAB AP onthespoolend.When ASAB APAP = 12 the

    spoolstopsmovingandtheoutputpressuredifferentialiscontrolled.InTable4.1some

    specificationsfortheservovalvearepresented[11].

  • 7/29/2019 Robot Hydraulic Control

    58/108

    43

    Table4.1-SpecificationofMoogservovalvemodel15-010

    Inputcurrent 10.00ma(rated)

    TorqueonArmature/Flapper

    0.0100 mN (rated)

    HydraulicAmplifierFlowtoDrivethe

    Spool

    3.770 scm /3 (max.)

    ServovalveFlow,No-Load

    902.0 scm /3 (rated)

    SpoolDisplacement

    0.5080mm(rated)

    HydraulicAmplifierDifferentialPressure

    61.41bar(rated)

    LoadDifferentialPressure

    207.0bar(rated)

    SpoolDrivingArea

    26.45 2mm

    SpoolFeedbackEndArea

    7.8712

    mm

    Apressureservovalvewaschosenbecauseofitshighcutofffrequencycharacteristics

    andthefactthatrelativelysmallflowrateswererequired.Theservovalvewasinstalled

    ascloseaspossibletotherotarymotorinordertominimizetheeffectsof

    compressibilityofthehydraulicfluidandtomaximizethestiffnessofthesystemby

    reducingthevolumeoffluid(loadside).

  • 7/29/2019 Robot Hydraulic Control

    59/108

    44

    4.4HydraulicRotaryMotor

    Asinglevanehydraulicrotarymotor(MicromaticmodelMPJ-22-1V)wasusedasthe

    actuator.Table4.2illustratessomespecificationofthismotor[12].

    Table4.2-SpecificationsofMicromaticrotarymotorMPJ-22-1V

    [email protected]

    [email protected]

    Torque

    ( mN )

    [email protected]

    62.61per270

    VolumetricDisplacement

    ( 3cm )

    13.36perradian

    Max.OperatingPressure

    (bar)

    69.00

    Max.RotatingAngle

    270.0

    AswillbepresentedinChapter5,thisrotarymotordisplayedsignificantnonlinear

    characteristics.Sinceitwastheonlyoneavailableinthelaboratory,ithadtobeused.

    Thiscertainlyposedsomeinterestingchallengesforthecontrollerdesign.

  • 7/29/2019 Robot Hydraulic Control

    60/108

    45

    4.5RobotManipulator

    Theoriginalmanipulatorusedinthisresearchprojectwasdesigned,constructedand

    reportedinathesisentitledPlanarManipulatorwithThreeRevoluteJoints[13].The

    schematicfrontviewisshowninFigure4.3andaphotoisshowninFigure4.4.Some

    modificationsweremadetotheoriginalmanipulatortoprepareitforthepurposeofthe

    project.Thiswasatwo-degree-of-freedomrigidlinkmanipulator,whichincludedtwo

    hydraulicrotarymotorsattheshoulderandelbowjoints,andtworigidshoulderand

    elbowlinks.Duetothefactthattheshoulderactuator(aChar-Lynnseries2000)was

    adiscvalvetyperotarymotorwithsignificantviscousfrictionandanarrowrangeof

    motion(180),thedecisionwasmadenottousethisactuator.Instead,theelbow

    actuatorwasusedwhichhadlessviscousfrictionandawiderrangeofmotion(270).

    Tostabilizethemanipulator,theshoulderlinkwasfixedandsecuredbyutilizinga

    numberofcablesandplateswithboltsandnuts.Thewristactuatorwassubsequently

    usedasaloadattheendoftheelbowlink.Figure4.3illustratestheschematicofthe

    manipulator.

  • 7/29/2019 Robot Hydraulic Control

    61/108

    46

    Figure4.3-Schematicofthetwo-linkrigidmanipulatorinwhichtheshoulderlinkwas

    fixed

  • 7/29/2019 Robot Hydraulic Control

    62/108

    47

    Figure4.4-Photoofthemanipulatorsetupusedintheexperiment

    Thepressurecontrolservovalvewasinstalledatthetopofthefixedshoulderlinkwhich

    wastheshortestdistancetotheoperatingactuator(elbowactuator)tominimizethe

    compressibilityeffectofthehydraulicfluid.

    Table4.3listssomeparticularsofthemanipulator[13].AsindicatedintheTable4.3,

    theconstructionoftherobotmanipulatorwasstrongenoughtobeconsideredasarigid

    manipulator.

    PressureServovalve

    Rotar Motor

    Accelerometer

  • 7/29/2019 Robot Hydraulic Control

    63/108

    48

    Table4.3-Importantmanipulatormeasurements

    Shoulderlinklength

    0.6100m

    Shoulderlinkcrosssection

    0.0760msquaresteeltube

    0.0060mwallthickness

    Elbowlinklength

    0.4600m

    Elbowlinkcrosssection

    0.0750msquaresteeltube

    0.0050mwallthickness

    Shoulderlinkweight

    (withoutservovalve)

    15.90kg

    Elbowlinkweight

    8.600kg

    4.6PersonalComputer

    InadditiontousingapersonalcomputerwithMatlab-SimulinkRealTimeWindows

    Targetforsoftwarealgorithmprogramming,aNationalInstrumentsPCI-6025edata

    acquisitioncardwitha12bitA/DandD/Awasusedtointerfacetheinputsignalfrom

    thecomputertotheservovalveandtheaccelerometersignalbacktothecomputer.The

    functiongeneratorintheSimulinkprogramwasutilizedtoprovidetheinputsignalto

    thehydraulicsimulator.Table4.4presentssomethepertinentsettingsthatwereused

    intheMatlab-Simulinkalgorithm.

  • 7/29/2019 Robot Hydraulic Control

    64/108

    49

    Table4.4-Matlab-Simulinksetupforexperimentalsystemcontrol

    SolverType

    Fixed-step

    Solver

    ode4(Runge-Kutta)

    Fixed-StepSize(FundamentalSample

    Time)

    0.001s

    Analoginput NationalInstrumentsPCI-6025E

    (auto),inputrange10v

    Analogoutput NationalInstrumentsPCI-6025E

    (auto),outputrange10v

    4.7SignalAnalyzer

    Asignalanalyzerunit(Bruel&Kjaertype2035)wasusedtoobtainthefrequency

    responsedataofthesystem.TheanalyzerprovidedtheactualBodemagnitudeand

    phaseplotsoftheexperimentalsystem.

    Theanalyzergeneratesarandomsignalofvariousfrequenciesrangingfrom0 Hzto200

    Hz(forthistest).Otherfrequencyrangeswerepossiblewiththisunit.Thesignalwas

    interfacedtotheservovalveviaonechannelasthesysteminputsignal.Themeasured

    acceleration(outputsignal)wasreturnedtotheanalyzerviaanotherchannel.

  • 7/29/2019 Robot Hydraulic Control

    65/108

    50

    4.8Accelerometer

    ApiezoelectricchargeaccelerometerofBruel&Kjaertype4370wasattachedtothe

    linkage(seeFigures4.3and4.4).Theaccelerometerwasconnectedtoacharge

    amplifierofBruel&Kjaertype2635.Thecalibratedoutputratingofthecharge

    amplifierwasadjustedto )/

    (1.02sm

    v.

    Theaccelerometerwascapableofmeasuringtangentialacceleration.However,itwas

    theangularaccelerationthatwasdesired.Sincethedistancebetweenthecenterofthe

    installedaccelerometerandtheverticalaxisoftherotarymotor(rotationarm)was

    known,itwaspossibletoconvertthelinear(tangential)accelerationtoangular

    acceleration(detailswereprovidedinChapter3).

    4.9AccelerometerCompensator

    InChapter3itwasmentionedthattheactualfrequencyresponseoftheexperimental

    system,showninFigure3.6,wasbestapproximatedbythetheoreticalfrequency

    responseillustratedinFigure3.4.Inreality,theslopeoftheactualBodeplot(Figure

    3.6)wasgreaterthanthetheoreticalplotatlowfrequencies.Itwasbelievedthatthe

    poorperformanceoftheaccelerometerandamplifieratlowfrequencieswasapossible

    reasonforthisdiscrepancy.Uponfurthertesting,(asfollows)itwasdeterminedthatthe

    accelerometer/amplifierdisplayedthecharacteristicsofahigh-passfilter.

    Anaccelerometertestwasperformedonashakertableutilizinganaccuratelinear

    positionsensor(SchaevitzDCLVDT)alongwiththetestaccelerometer(B&K4370)

    whichwerebothinstalledonthesameplaneofmotion.Therandomsignalgeneratedby

    thesignalanalyzerwasinterfacedtotheshakertable.Theoutputsignaloftheposition

    transducerwasconnectedtotheanalyzer.Theanalyzerwassettoconverttheposition

  • 7/29/2019 Robot Hydraulic Control

    66/108

    51

    signaltoanaccelerationsignalbydifferentiatingittwice.Theoutputsignalofthetest

    accelerometerwasalsointerfacedtotheanalyzer.Theanalyzerwasutilizedtocompare

    amplitudeandphaseangleofacceleration(derivedfromthepositionsignal)withthose

    ofthetestaccelerometersignal.Figure4.5illustratestheschematicofthetest

    arrangement.

    Figure4.5-SchematicofShakerTabletest

    InFigure4.5;

    aca measuredaccelerometersacceleration(v)

    pox measuredposition(v)

    poa calculatedaccelerationbydoubledifferentiationofmeasuredposition( v)

    Figure4.6showsthemagnitudeandphaseBodediagramoftwicedifferentiated

    positionsignalandthetestaccelerometersignal.

    pox

    aca

    RandomSignal

    Anal zerUnit

    poa

    ShakerTable

    Accelerometer

    2s

    Position

    (DCLVDT)

  • 7/29/2019 Robot Hydraulic Control

    67/108

    52

    -10

    -5

    0

    5

    10

    0.1 1 10 100

    Frequency(Hz)

    -180

    -135

    -90

    -45

    0

    45

    90

    135

    180

    0.1 1 10 100

    Frequency(Hz)

    Figure4.6-MagnitudeandphaseBodeplotsoftheshakertabletest(pointsaredata

    fromtheanalyzerandthesolidlineisthestraightlineapproximationbasedona20dB

    perdecadeslope).

    )(

    (

    )(

    dB

    ja

    ja

    po

    ac

    )(

    )(

    )(

    Degrees

    ja

    ja

    po

    ac

  • 7/29/2019 Robot Hydraulic Control

    68/108

    53

    Anapproximatedtransferfunctionofthetwoaccelerationsignalscanbeextractedfrom

    anasymptoticanalysisofthemagnitudeBodeplotwhichisshowninFigure4.6.Itis

    evidentthatatlowfrequencies,theaccelerometerproducedresultswhichwere

    attenuated.Thisattenuationcouldbeapproximatedbyatransferfunction.Analysisof

    Figure4.6indicatesthatthecornerfrequencywas1.25 Hz(7.85rad/s).Thereforethe

    transferfunctionwhichreflectedthislowfrequencyattenuationwasestimatedtobe;

    85.7+=s

    sGac (4.1)

    Tocompensateforthisattenuation,acompensatorwaschosentobe 00.5

    85.7

    +

    +

    s

    s

    inorderto

    expandthelowerendofbandwidthoftheaccelerometer.

    Itshouldbenotedthatanidealcompensatorshouldbes

    s 85.7+butaswillbediscussed

    inChapter5,theform00.5

    85.7

    +

    +

    s

    sprovidedbettercompensation.Moredetailswillbe

    discussedonthismatterinthefollowingchapterinwhichexperimentaldatais

    considered.

    Inimplementingthisaccelerometercompensatorexperimentally,theproposed

    compensatorwasinstalledaftertheplantandphysicalaccelerometer.Figure4.7

    illustratestheplacementofthecompensatortoatheoreticalapproximationofthe

    accelerometerfrequencycharacteristics.Chapter5willconsidertheimplementationof

    thecompensatorfromanexperimentalpointofview.

  • 7/29/2019 Robot Hydraulic Control

    69/108

    54

    Figure4.7-Blockdiagramofthesystemwiththeaccelerometercompensator

    Itshouldbenotedthatthiscompensatorwasutilizedforallexperimentaltests.

    Accelerometer

    Compensator

    eV

    00.5

    85.7

    +

    +

    s

    s

    v vc Plant

  • 7/29/2019 Robot Hydraulic Control

    70/108

    55

    CHAPTER5

    EXPERIMENTALRESULTS

    5.1Introduction

    Theobjectiveofthischapteristopresentanddiscusstheresultsofthetestsperformed

    ontheexperimentalsystemwiththesetupexplainedinChapter4.Generally,thetests

    madeuseofsweptfrequencies(aprocessinwhichtheinputfrequencywas

    systematicallyincreasedoveraselectedfrequencyrange)andfunctiongeneratorinputs

    toproducebothfrequencyandtimedomainresults.Formanyofthetests,magnitude

    andphaseBodediagramsweredevelopedfromthedatatoillustratethedynamic

    performanceofthesystemandcompensationschemes.Todemonstratethedesireduse

    ofthehydraulicsimulator,thetemporalresponsestotheoutputfromthefunction

    generatorwereconsidered.TheChapterwillthenillustratethetransientresponse

    trackingerrorsandmeansquarederrorvalues.

    5.2PressureControlServovalve

    Thepressurecontrolservovalvewasusedinthehydraulicsimulatorbecauseofits

    superiorfrequencybandwidthatlowflowrateswhencomparedtoaflowcontrol

    servovalve[14].Thepressurecontrolservovalvedynamicperformance(thefrequency

    responsedata)isgenerallypresentedintwoforms;oneinwhichtheloadportsofthe

  • 7/29/2019 Robot Hydraulic Control

    71/108

    56

    servovalveareblocked(blockedload)andtheotherwhentheportsareconnectedtoa

    load.Theblockedloadtestisnormallyconducedbypluggingtheloadports(for

    exampleportsAandBinFigure4.2).Theloadpressuredifference( ABP )andtheinput

    voltagearemeasuredforawiderangeofsinusoidalfrequencies(inthefollowingcase

    between0.5to400Hz).Intheblockedloadcasetheservovalveshoulddemonstratea

    highstiffnessbecauseoftheverysmallfluidvolumeontheloadsideresultingina

    highernaturalfrequency.

    Figures5.1aandbillustratetherelationshipofblockedloaddifferentialpressureto

    inputvoltagetotheservo-amplifier.AscanbeobservedintheBodeplot,thebandwidth

    ofthevalvewasapproximatelyzeroto60Hzwhichwassomewhatlessthanthatstated

    inthemanufacturersspecifications[11].Athigherfrequencies,aslopeof-

    60dB/decadeinthemagnitudeplotindicatedthatthesystemwasapproximatelythird-

    order.

    -40

    -30

    -20

    -10

    0

    10

    20

    0.1 1 10 100 1000

    Frequency(Hz)

    Slope-60dB/decade

    Figure5.1aMagnitudefrequencyresponseofthepressurecontrolservovalvewitha

    blockedload(pointsaredatafromtheanalyzerandthesolidlineisthestraightline

    approximation).

    )(

    )(

    )(

    dB

    jV

    jP

    e

    L

  • 7/29/2019 Robot Hydraulic Control

    72/108

    57

    -360

    -315

    -270

    -225

    -180

    -135

    -90

    -45

    0

    45

    0.1 1 10 100 1000

    Frequency(Hz)

    Figure5.1bPhasefrequencyresponseofthepressurecontrolservovalvewitha

    blockedload.

    Whenthepressurecontrolservovalvewasconnectedtotheexperimentalsystem,the

    frequencyresponsewasdeterminedandisshowninFigures5.2aandb.

    )(

    )(

    )(

    Degrees

    jV

    jP

    e

    L

  • 7/29/2019 Robot Hydraulic Control

    73/108

    58

    -40

    -30

    -20

    -10

    0

    10

    20

    0.1 1 10 100

    Frequency(Hz)

    Slope-40dB/Decade

    Figure5.2a-Magnitudefrequencyresponseofpressurecontrolservovalvewithactual

    hydraulicsimulatorload(pointsaredatafromtheanalyzerandthesolidlineisthe

    straightlineapproximation).

    -180

    -135

    -90

    -45

    0

    45

    0.1 1 10 100

    Frequency(Hz)

    Figure5.2b-Phasefrequencyresponseofpressurecontrolservovalvewithactual

    hydraulicsimulatorload.

    )()(

    )(

    dB

    jV

    jP

    e

    L

    )(

    )(

    )(

    Degrees

    jV

    jP

    e

    L

  • 7/29/2019 Robot Hydraulic Control

    74/108

    59

    Itisobservedthatthebandwidthisabout40 Hzandtheorderofthesystemhaschanged

    fromthirdtosecond.Theinfluenceoftheloadisthusapparent.Thehighfrequency

    affectsoftheservo-amplifierarenotvisiblefromtheseresultsandhencetheassumption

    thattheservo-amplifiercouldbeapproximatedbyasimplegaininChapter3was

    justified.

    Thefrequencyresponseoftheactualloadpressuretoinputvoltage(Figure5.2aandb)

    demonstratesthattheselectedpressurecontrolservovalvewascapableofanappropriate

    performanceinthefrequencyrangeofinterest(5-15 Hz).

    5.3Accelerometer

    InChapter4itwasstatedthattheaccelerometerwasdemonstratingpoorperformanceat

    thelowerfrequencies.Itwasbelievedthatsomemethodofcalibratingtheacceleratorat

    lowfrequencieswasnecessary.AsdiscussedinChapter4,ashakertabletestfacility

    wasavailableinthelaboratorywhichcouldbeusedtoprovideacalibratedposition

    measurementandhenceareliableaccelerationcalculation.Boththeaccelerometerand

    positiontransducerweremountedonthetableandtheshakertablefrequencyrange

    sweptfrom1to25Hz.Thecalibratedaccelerationwasdeterminedbyadouble

    differentiationofthepositionsignal.Themagnitudeandphaseratioofthe

    accelerometer aca outputtothecalculatedacceleration poa fromthepositiontransducer

    wereshowninFigure4.6.Ideallythemagnituderatioshouldbeunityandthephase

    shiftzero.ObservationofFigure4.6illustratesthattheperformanceofthe

    accelerometerdeterioratesatfrequencieslessthan7.85rad/s(1.25 Hz).

    AsdiscussedinChapter4,anidealcompensatoroftheforms

    sG

    85.71

    += should

    extendthelowerfrequencyrangeoftheaccelerometer.Whenappliedtotheactual

    system,theresultswerenotsatisfactoryinthatthecompensatorovercompensatedthe

    accelerometerssignalsatfrequenciesbelow7.85rad/s(1.25 Hz).Theassumptionthat

  • 7/29/2019 Robot Hydraulic Control

    75/108

    60

    theactualdatafollowedthetransferfunction 1G isonlyapproximateinthattheslopes

    arecomparablebutinthetransitionregionsbetweenthetwoslopes,theexperiential

    datawasfoundtobelargerthanthatpredictedbythetransferfunction 1G .Thusdatain

    thatregionwasamplifiedbeyondthatrequired;thusovercompensated.Inaddition,thedataforfrequencieslessthan1Hzwasveryerraticandtheuseof 1G wouldonly

    amplifythescatteratlowerfrequencies(seeFigure4.6).Thusacompensatorofthe

    formbs

    as

    +

    +whichwouldnotamplifythelowerfrequenciesbutcompensateatthebreak

    pointwasused.Thiscompensatorwastunedtobeintheformof00.5

    85.7

    +

    +

    s

    swhichwas

    thenappliedtoallmeasuredaccelerationdata.

    Figures5.3aandbindicatetherelationshipbetweentheaccelerometeroutputsignal

    anddoubledifferentiatedsignalfromthepositiontransducer,withtheproposed

    compensator.AcomparisonofFigure5.3toFigure4.6inChapter4showsthatthe

    lowerendofbandwidthoftheaccelerometerhasbeenextendedfrom3 Hztoabout1

    Hz.However,thereisstillaslightaffectontheamplituderatiointherangefrom1to10

    Hz,whichisnotdesirable.Theoverallimprovementusingthiscompensatorwas

    consideredmarginalbutwasstillimplementedinallsubsequentstudies.

  • 7/29/2019 Robot Hydraulic Control

    76/108

    61

    -10

    -5

    0

    5

    10

    0.1 1 10 100

    Frequency(Hz)

    Figure5.3a-Magnituderatio

    ja

    ja

    po

    ac

    (

    )( versusfrequencyoftheaccelerometerwith

    proposedaccelerometercompensator(pointsaredatafromtheanalyzerandthesolid

    lineisthestraightlineapproximation).

    -180

    -135

    -90

    -45

    0

    45

    90

    135

    180

    0.1 1 10 100

    Frequency(Hz)

    Figure5.3b-Phase)(

    )(

    ja

    ja

    po

    ac versusfrequencyoftheaccelerometerwithproposed

    accelerometercompensator.

    )(

    )(

    )(

    Degrees

    ja

    ja

    po

    ac

    )(

    (

    )(

    dB

    ja

    ja

    po

    ac

  • 7/29/2019 Robot Hydraulic Control

    77/108

    62

    Thecompensatorwasthenappliedtotheaccelerationdatafromthehydraulicsimulator

    andthedynamicperformance(magnitudewiseonly)beforeandaftercompensationis

    comparedinFigures5.4and5.5.Itisapparentthatinthelowerfrequencyrange(less

    than2Hz),theslopeofthefrequencyresponseafterapplyingthecompensatorisnow

    20dbperdecadewhichisconsistentwithwhatwaspredictedfromthetheoretical

    modelofthesystem(Chapter3).However,consistentwiththeovercompensating

    natureoftheaccelerometerinthefrequencyrangeof1to10 Hz,thereappearssome

    distortiontothefrequencyresponseintheplantdataaswell.However,becauseinthe

    lowerfrequencyrange,theslopewiththecompensatedaccelerometerapproachedthat

    whichwaspredictedanalytically,itwasconcludedthattheaccelerometercompensation

    wasacceptableforthisstudy.

    -40

    -30

    -20

    -10

    0

    10

    20

    0.1 1 10 100

    Frequency(Hz)

    Figure5.4-Open-loopmagnitudefrequencyresponseofthehydraulicsimulatorbefore

    applyingtheaccelerometercompensator(pointsaredatafromtheanalyzerandthesolid

    lineisthebestfitstraightlineapproximation;40dBperdecadeslope).

    )(

    )(

    )(

    dB

    jV

    j

    e

    v

  • 7/29/2019 Robot Hydraulic Control

    78/108

    63

    -40

    -30

    -20

    -10

    0

    10

    20

    0.1 1 10 100

    Frequency(Hz)

    Figure5.5-Open-loopmagnitudefrequencyresponseofthehydraulicsimulatorafter

    applyingaccelerometercompensator(pointsaredatafromtheanalyzerandthesolid

    lineistheexpectedstraightlineapproximationoftheplant,20dBperdecadeslope).

    5.4HydraulicSimulator

    Theexperimentalresultsofthehydraulicsimulatorsuchasopen-loopandclosed-loop

    frequencyresponses,open-loopcompensatedfrequencyresponse,open-looptransient

    responsesandtransientresponseerrorsarepresentedanddiscussedinthissection.

    5.4.1Open-LoopFrequencyResponses

    Theopen-loopbehaviorofthehydraulicsimulatorisshowninFigure5.5.Itisobserved

    thatthegainis-8dBandhasabandwidthofapproximately2to30 Hz.Anunknown

    dipinthemagnitudeoccursat25Hz;atthispointnoreasonableexplanationcanbe

    forwardedforthisbehavior.

    )(

    )(

    )(

    dB

    jV

    j

    e

    vc

  • 7/29/2019 Robot Hydraulic Control

    79/108

    64

    Thefrequencyresponseoftheopen-loophydraulicsimulatorforsmallerinput

    amplitudesof0.5v(asmallaccelerationof 2/5.11 srad )isshowninFigure5.6aandb.

    Itisevidentthattheperformancedeterioratesinthatthelo