<%BANNER%>

Robust Control of Supercavitating Vehicles in the Presence of a Dynamic and Uncertain Cavity


PAGE 1

ROBUSTCONTROLOFSUPERCAVITATINGVEHICLESINTHEPRESENCEOF DYNAMICANDUNCERTAINCAVITY By ANUKULGOEL ADISSERTATIONPRESENTEDTOTHEGRADUATESCHOOL OFTHEUNIVERSITYOFFLORIDAINPARTIALFULFILLMENT OFTHEREQUIREMENTSFORTHEDEGREEOF DOCTOROFPHILOSOPHY UNIVERSITYOFFLORIDA 2005

PAGE 2

ACKNOWLEDGMENTS Iwouldliketoexpressmysinceregratitudetomycommitteec hairman,Dr.Andrew Kurdila,forhisinvaluableguidancethroughoutthecourse ofthisproject.Iwouldalsolike tothankhimforgivingmethisopportunitytoworkonsuchafa scinatingproject. Iwouldalsoliketothankmycommitteecochair,Dr.RichardC.L ind,forhisinvaluable guidanceandinspirationthroughouttheproject. IwouldliketothankDr.NormanFitz-Coy,Dr.BrianMannandDr. HaniphLatchman forservingonmycommittee.Iwouldalsoliketoshowmysince reappreciationtoDr. JohnDzielskiandJammulamadakaAnandKapardifortheirval uablecontributionstothis project.Iwouldalsoliketoexpressmygratitudetoallthem embers,pastandpresent,of theSupercavitationProject. IwouldalsoliketothanktheOfceofNavalResearchforthesu pportoftheresearch grantfortheproject. Onapersonalnote,Iwouldliketothankallmyfriendsandfam ilymemberswhose supporthelpedmetoaimtowardsmygoals. ii

PAGE 3

TABLEOFCONTENTS page ACKNOWLEDGMENTS ........................ii LISTOFTABLES ...........................vi LISTOFFIGURES ..........................ix ABSTRACT .............................x CHAPTER 1INTRODUCTION .........................1 1.1Cavitation ...................................1 1.2AspectsofSupercavitation ..........................3 1.3RelatedResearch ...............................4 1.4OverviewofThisDissertation ........................5 2NONLINEAREQUATIONSOFMOTION ................7 2.1CongurationoftheVehicle .........................7 2.1.1Body .................................7 2.1.2Cavitator ...............................8 2.1.3Fins ..................................9 2.1.4Maneuvering .............................9 2.2KinematicEquationsofMotion ........................10 2.2.1OrientationoftheTorpedo ......................10 2.2.2OrientationoftheCavitator .....................12 2.2.3OrientationofFins ..........................13 2.2.4AngleofAttackandSideslip .....................16 2.2.5KinematicEquations .........................20 2.3DynamicEquationsofMotion ........................22 2.3.1ForcesonCavitator ..........................24 2.3.2ForcesonFins ............................27 2.3.3GravitationalForces .........................30 2.3.4EquationsofMotion .........................31 2.3.4.1Forceequations ......................31 2.3.4.2Momentequations .....................32 2.3.4.3Orientationequations ...................32 2.3.4.4Navigationequations ...................32 iii

PAGE 4

3CAVITYANDPLANINGDYNAMICS ................34 3.1Munzer-ReichardtModel ...........................35 3.2LogvinovichCavityModel ..........................35 3.2.1LogvinovichTheoryofIndependentExpansion ...........35 3.2.2CavityCenterline ...........................36 3.3PlaningModel .................................38 3.4PlaningForceEquations ...........................39 3.5PlaningKinematics ..............................42 3.5.1CalculationofImmersion ......................42 3.5.2MethodofCalculationof a .....................42 4LINEARIZATION ........................47 4.1Linearization .................................47 4.1.1NeedforLinearization ........................47 4.1.2GenericFormofEquationsofMotion ................48 4.1.3SmallDisturbanceTheory ......................48 4.1.3.1Forceequations ......................50 4.1.3.2Momentequations .....................50 4.1.3.3Orientationequations ...................50 4.1.3.4Positionequations .....................50 4.1.4StabilityandControlDerivatives ...................50 4.2StateSpaceRepresentation ..........................55 5CONTROLDESIGNSETUP ....................59 5.1Open-LoopPerformancefortheFixedCavityModel ............60 5.2Closed-LoopProblem ............................63 5.3RobustnessoftheController .........................64 5.3.1Gainmargin .............................65 5.3.2Phasemargin .............................65 5.3.3Uncertaintyinparameters ......................65 5.3.4Controllerobjective ..........................66 5.3.5 analysis: ..............................66 6LQRCONTROL .........................67 6.1LQRTheory ..................................67 6.2LQRControlforFixedCavityModel: ....................71 6.2.1ControlSynthesis ...........................71 6.2.2NominalClosed-loopModel .....................73 6.2.2.1Model ...........................73 6.2.2.2Simulations ........................73 6.2.2.3Gainandphasemargins ..................76 6.2.3PerturbedClosed-loopModel ....................77 iv

PAGE 5

6.2.3.1Model ...........................78 6.2.3.2Simulations ........................78 6.2.3.3Gainandphasemargins ..................80 7 /H SYNTHESISCONTROL ...................82 7.1Uncertainty ..................................82 7.2SynthesisModel ................................83 7.3ControlObjectiveandConstraints ......................90 7.4 ControllerforFixedCavityModel .....................90 7.4.1Longitudinalcontroller ........................90 7.4.2Lateralcontroller ...........................93 8HOMINGCONTROL .......................98 8.1HomingControlusingProportionalNavigation ...............99 8.2ConstantMissileandTargetVelocity .....................100 8.3ConstantVelocityTargetandAcceleratingMissile ..............101 8.4AcceleratingTargetandMissile .......................102 8.5YawController ................................104 8.5.1YawcontrolusingtheLQRcontrollers ...............104 8.5.2Yawcontrolusingthe /H controller ................105 9CONCLUSION ..........................107 9.1Summary ...................................107 9.2FutureWork ..................................107 APPENDIX AREFERENCEFRAMESANDROTATIONMATRICES .........109 BNUMERICALTECHNIQUES ....................111 B.1InterpolationofForceData ..........................111 B.1.1Extrapolationscheme .........................111 B.1.2Cavitator ...............................112 B.1.3Fins ..................................113 B.2NumericalLinearization ...........................113 CTESTBEDDATA .........................115 C.1FinParametersfortheTestBedandFin-Data ................115 C.2CavitatorParametersfortheTestBedandCavitator-Data ..........116 REFERENCES ...........................119 BIOGRAPHICALSKETCH ......................122 v

PAGE 6

LISTOFTABLES Table page 5.1ControlParameters ..............................59 5.2ControlConstraints ..............................64 6.1GainandPhaseMarginwithLQRController ................77 6.2PercentageVariationinAMatrixdueto20%Variationin cl c ........78 6.3PercentageVariationinBMatrixdueto10%Variationin cl c ........79 6.4GainandPhaseMarginforPerturbedClosed-loopSystem:2 0%errorin cl fin 81 7.1GainandPhaseMarginsfortheLongitudinalController ..........93 7.2GainandPhaseMarginsfortheLateralController ..............96 B.1GridForExperimentalCavitatorData ....................112 B.2GridForExperimentalFinData .......................113 C.1TheFinVariablesRangeSet. .........................116 vi

PAGE 7

LISTOFFIGURES Figure page 1.1SupercavitatingVehicle ............................3 2.1SupercavitatingVehicle ............................7 2.2VehicleBodyisDividedInto4Sections. ...................8 2.3CavitatorandFins ...............................9 2.4Body-FixedandInertialFrames .......................11 2.5PrincipalPlanesofSymmetryfortheTorpedo ................11 2.6EulerAnglesofRotation ...........................11 2.7CavitatorReferenceFrame ..........................13 2.8RudderandFinReferenceFrames ......................14 2.9Rudder1FinReferenceFrames .......................15 2.10Rudder2FinReferenceFrames .......................16 2.11Elevator1FinReferenceFrames .......................17 2.12Elevator2FinReferenceFrames .......................17 2.13AngleofAttack( a )andSideslip( b ) .....................18 2.14Cavitator:(a)AngleofAttackandSideslipand(b)Hydro dynamicForces .26 2.15FinGeometry .................................27 3.1VehicleUndergoingTail-Slap. ........................39 3.2Cross-SectionofCavity ............................40 3.3PlaningSection ................................42 3.4TheAngle f ..................................44 3.5VelocityofTransomRelativetoFrame B ...................45 3.6Calculationof a withAssumptionthattheCavitySurfaceisPlanar .....46 5.1SimulinkModelforOpenLoopSimulation .................61 vii

PAGE 8

5.2Open-LoopResponseforTorpedo: w ; p ; q ..................61 5.3VariationofEigenvalueswithChangeinVelocity ..............63 5.4LoopGain ...................................66 6.1ControllerforTrackingwhenPlanthasanIntegrator. ............70 6.2ControllerforTrackingwhenPlanthasnoIntegrator. ............70 6.3EigenvaluesfortheClosed-LoopSystem ...................73 6.4PitchCommandTracking: q .........................74 6.5PitchCommandTracking: d c ; d c .......................74 6.6PitchCommandTracking: d e 1 ; d e 1 ......................74 6.7RollCommandTracking: p ..........................75 6.8RollCommandTracking: d c ; d c ........................75 6.9RollCommandTracking: d e 1 ; d e 1 .......................76 6.10BreakpointsforCalculatingtheLoop-GainforaTracking Controller ....76 6.11EigenvaluesforthePerturbedClosed-LoopSystem:20%E rrorin cl fin ...79 6.12Responsefor20%Variationin cl fin : p ; q ...................80 6.13Responsefor20%Variationin cl fin : d c ; d e 1 .................80 7.1CalculationofUncertainty ..........................83 7.2LinearFractionalRepresentation .......................88 7.3SynthesisModelfor Controller .......................89 7.4PitchAngleTrackingfortheNominalPlant .................92 7.5CavitatorandElevatorDeections ......................92 7.6FrequencyResponseofthePitchAngleController .............93 7.7ComparisonofResponsefortheNominalandPerturbedPlant s .......94 7.8RollAngleTrackingfortheNominalPlant ..................95 7.9CavitatorandElevatorDeections ......................96 7.10FrequencyResponseoftheRollAngleController ..............96 7.11ComparisonofResponsefortheNominalandPerturbedPlan ts .......97 viii

PAGE 9

8.1CollisionTriangle ...............................99 8.2VariousPossibilitiesofSequenceofCommandstoObtainY awRateControl 105 8.3SequenceofCommandstoObtainYawRateControl ............106 8.4YawAngleResponseoftheVehiclefortheCommandsShowninF igure8.3 106 A.1RotationofFrames ..............................109 B.1ShapeFunctionforOneDimensionalQuadraticScheme ..........112 C.1HydrodynamicForcesActingattheCenterofPressureofFin .......116 C.2Fin Cl and Cd TestBedData:SweepAngle=0.A)FinCl;B)FinCd ...117 C.3Fin Cl and Cd TestBedData:SweepAngle=15.A)FinCl;B)FinCd ..117 C.4Fin Cl and Cd TestBedData:SweepAngle=30.A)FinCl;B)FinCd ..117 C.5Fin Cl and Cd TestBedData:SweepAngle=45.A)FinCl;B)FinCd ..117 C.6Fin Cl and Cd TestBedData:SweepAngle=60.A)FinCl;B)FinCd ..118 C.7Fin Cl and Cd TestBedData:SweepAngle=70.A)FinCl;B)FinCd ..118 ix

PAGE 10

AbstractofDissertationPresentedtotheGraduateSchool oftheUniversityofFloridainPartialFulllmentofthe RequirementsfortheDegreeofDoctorofPhilosophy ROBUSTCONTROLOFSUPERCAVITATINGVEHICLESINTHEPRESENCEOF DYNAMICANDUNCERTAINCAVITY By AnukulGoel August2005 Chair:AndrewJ.KurdilaCochair:RichardC.LindMajorDepartment:MechanicalandAerospaceEngineering Underwatertravelisgreatlylimitedbythespeedthatcanbe attainedbythevehicles. Usually,themaximumspeedachievedbyunderwatervehicles isabout40m/s.Supercavitationcanbeviewedasaphenomenonthatwouldhelpustobre akthespeedbarrierin underwatervehicles.Theideaistomakethevehiclesurroun dedbywatervaporwhileit istravelingunderwater.Thus,thevehicleactuallytravel sinairandhasverysmallskin frictiondrag.Thisallowsittoattainhighspeedsunderwat er.Supercavitationisaphenomenonwhichisobservedinwater.Astherelativevelocity ofwaterwithrespectto thevehicleincreases,thepressuredecreasesandsubseque ntlyitevaporatestoformwater vapor.Supercavitationhasitsdrawbacks.Itisreallyhard tocontrolandmaneuverasupercavitatingvehicle.Therstpartofthisworkdealswith modelingofasupercavitating torpedo.Nonlinearequationsofmotionarederivedindetai l.Thelatterpartofthework dealswithndinginner-loopcontrollerstomaneuvertheto rpedo.Acontrollerisobtained byLQRsynthesisforpitchandrollratecontrol.Robustcontr ollersareobtainedby /H synthesisfortrackingpitchandrollanglecommands.Thero bustnessanalysisoftheLQR x

PAGE 11

controllersiscarriedoutbycalculatingthegainandphase margins.Simulationsofthe responseforaperturbedsystemalsohavebeenstudied.Thei nner-loopcontrollersare usedforguidanceandnavigationofthevehicle.Itisobserv edthatthepitchandrollangle controllerscanbeusedforyawratecontrolofthevehicle.T hisisappliedtothehoming guidanceproblem.Asimpliedcaseissolvedforhomingguid ance. xi

PAGE 12

CHAPTER1 INTRODUCTION Achievinghighspeedsisanimportantissueforunderwaterv ehicles.Eventhecommon fastestunderwatervehiclesarerestrictedtotravelatspe edsaround40ms 1 .Thereason forthisrestrictionisthedraginducedbyskinfriction.Whe nabodymovesinauid,a layeroftheuidclingstothesurfaceofthebodyandisdragg edwithit.Thisinteraction causeshighdragforcesonthebodyandiscommonlytermedski nfrictiondrag.Thedrag forceinwater,unlikeair,isdominatedbyskinfrictiondra gascomparedtoothersources suchaspressuredrag.Overtheyears,extensiveresearchha sbeendonetoboostthespeed ofunderwatervehicles.However,mostofthestudieswerema inlyfocusedonstreamlining thebodiesandimprovingtheirpropulsionsystems.Eventho ughthesehaveproventogive advancementsinspeed,therehasnotbeenaconsiderablered uctioninskinfrictiondrag. Inthelate1970's,theRussianNavywasabletoengineeratorp edo,theShkval(squall)[ 1 ], thatachievedaspeedover80ms 1 .Thisphenomenalimprovementinspeedwasmade possiblebysupercavitation.Theideawastoenvelopthetor pedoinagassothatithaslittle contactwithwater.TheShkvalwasabletoachieveatremendo usreductioninskinfriction dragandexhibithighspeed. 1.1Cavitation Asthespeedofanunderwatervehiclesincreases,i.e.,asth erelativevelocityofwater withrespecttothevehicleincreases,thepressuredecreas es.Thespeedcanbeincreased tothepointthepressuregoesbelowthevaporpressureofwat erandthenbubblesofwater vaporareformed.Thisprocessisknownascavitation.Cavita tionismostlyobservedat sharpcornersofabodywherethespeedcanreachhighmagnitu des.Aclassicexample forcavitationisatthetipofpropellers.Sincethepropell erisrotatingathighspeed,the 1

PAGE 13

2 relativevelocityatthetipsishighenoughsothatwateratt hetipsvaporizes.Cavitationhas beenextensivelyresearched,butremainsachallengeforun derwatervehicles. Cavitationcanbeharmfulinmanycases.Thecavitationregio nisusuallyturbulent. Whenthecavitationisnotsteady,thepressuredropismoment aryandveryquicklythe cavitationbubbleencountersaregionofhighpressurethat forcesthebubbletocollapse likeatinybomb.Thiscollapsecauseshighlevelsofnoisean dalsomaycauseconsiderable damagetothesurfaceofthebody.Thecavitybehaviorisgene rallydiscussedinterms ofadimensionlesscoefcientcalledthecavitationnumber s .Thisiscalculatedfrom Equation 1.1 s = p p c ( 1 = 2 ) r w V 2 (1.1) where r w isthedensityofuidand V isthespeedoftheuidrelativetothebody.The pressureinthecavity p c maybethatofvapor p V orvaporplusair.Thisnumberrelatesthe differenceinthepressurebetweenthecavityanditssurrou ndingstothedynamicpressure q =( 1 = 2 ) r w V 2 ofthefree-stream.Whencavitationoccurs,thecavitationn umberwill equalthenegativeofthelowestvalueofthecoefcientofpr essure.Thisvalueisreferred tothecriticalcavitationnumber.Thisisthevalueabovewh ichcavitationstartstooccurin theuid. Supercavitationisahigherstageofcavitationwhenthebod yisentirelysurrounded bythecavitationbubble.Thecavitationstartsasvaporbub blesneartheregionofhigh relativevelocity.Asthespeedisfurtherincreased,thebu bblesmergetoformalargearea ofvapor.Onfurtherincreaseinspeed,thewholeofthebodyi scoveredinvapor.This stageiscalledthesupercavitation.Atthispoint,thecond itionissimilartotravelinginair. Theskinfrictiondragistremendouslyreduced,andthushig hspeedcanbeattainedinthis phase.Figure 1.1 showsasupercavitatingtorpedotravelingunderwater.Itc anbeseenthat thewholeofitsbodyisenvelopedbyacavity.Thisisthekind ofvehiclethathasbeen studiedinthiswork.

PAGE 14

3 Cavity Fin Cavitator Figure1.1SupercavitatingVehicle[ 1 ] 1.2AspectsofSupercavitation Supercavitationisanextremeversionofcavitationinwhic hasinglebubbleisformed whichenvelopsthemovingobjectcompletely.Thecavitytak esontheshapenecessary toconservetheconstantpressureconditiononitsboundary .Theshapeisdeterminedby factorslikethebodycreatingit,thecavitypressureandth eforceofgravity.Withslenderaxisymmetricbodies,supercavitiestakeshapeofexten dedellipsoidsbeginningatthe forebodyandtrailingbehind,withthelengthdependentont hespeedofthebody[ 2 ].Only smallareasatthenoseandontheafter-bodyremainincontac twiththeliquid. Expressionsformaximumdiameterandlengthofthecavityas functionsofcavitator geometry,dragcoefcientandcavitationnumberareavaila ble.Thecavitationnumberand dragcoefcientthendeterminethecavitygeometry.Thesup ercavitymayconsistcompletelyofvapor(naturalcavity),orhavepartialnon-cond ensiblegascomposition(ventilatedcavity),whichisinjectedfromthenose.Thedeterm inationofcavitationnumber andhencecavitystructureinventilatedcavitationismore complicatedthanforanatural supercavity[ 3 ]. Ahighspeedsupercavitatingprojectile,whilemovinginth eforwarddirection,may rotateinsidethecavity.Thisrotationleadstoaseriesofi mpactsbetweentheprojectiletail andthecavitywall,calledthetail-slapphenomenon.Theim pactsaffectthetrajectoryas wellasthestabilityofmotionoftheprojectile.Despiteth eimpactswiththecavitywall, theprojectile,incertaincases,nearlyfollowsastraight linepath.May[ 4 ]hasadetailed analysisfromexperimentaldatafortrajectoriestakenbyu npoweredprojectilesuponwater entry.Alsothepatternsoftailslapforsuchprojectilesha vebeendiscussed.

PAGE 15

4 Thewettedforcesdeterminethevehiclespeedandvehiclest ability.Becausethewettedcontactareaissmall,unwantedchangesinthewettedcon tactareaarisingfromlocal cavitybreakdownmayresultinlargedestabilizingforcesa ndmomentsforasupercavitatingvehicle.Thewettedforcesmustbespeciedandunder stoodtodeterminevehicle maneuverability. 1.3RelatedResearch Researchintheeldofsupercavitationhasbeengoingonfrom theearly1900's.But earlierresearchwasaimedatreductionofcavitationsoast oreducenoiseandbodydamage. Inthe90'sthefocusshiftedtoexploitingtheeffectsofsup ercavitation. TheworkshowninMay[ 4 ]hasanextensivecollectionofexperimentaldataforvariationofforcesonvarioussupercavitatingshapes.Coefcie ntsofliftanddragareplotted withthevariationofcavitationnumberforshapeslikedisk s,cones,ogivesandwedges. TheworkdoneinthisresearchmakesuseofaCFDdatabaseprovi dedinFine[ 5 ].This databasehasvaluesforcoefcientsofliftanddragforconi calcavitators,whicharefunctionsofthehalfangleoftheconeandtheangleofattack.Thi sdatabasealsohascoefcients ofliftanddragforwedges,whicharefunctionsofwettedsur faceofthewedge,angleof attackandsweepbackangle(wediscussthedenitionofthes egeometricquantitiessuch astheangleofattackandhalfangleinthelaterchapters).T hisinformationisusefulto calculatetheforcesonnsofthetorpedo. Inlate90'salotofresearchwasdoneonthedynamicsofsuper cavitatingvehicles. WorkshowninKulkarniandPratap[ 6 ]andRandetal.[ 7 ]dealswithstudyingdynamics ofuncontrolledsupercavitatingprojectiles.Adynamicmo delforRAMICSandAHSUM hasbeendeveloped.Itisshownthattheseprojectilesrotat einsidethecavity.Thisrotation leadstoimpactsbetweenthetailoftheprojectileandtheca vitywall.Thefrequencyofthe impactincreaseswithtime.Theseprojectilesareveryshor trangeandhaveasmalltimeof ight,ontheorderofafewseconds.

PAGE 16

5 TheworkshowninDzielskiandKurdila[ 8 ]focusesontheformulationofacontrol problemforasupercavitatingtorpedo.Adynamicalmodelfo rancontrolledtorpedo haslikewisebeendeveloped.Themodelalsoincludesaformu lationforthecavity.Itis observedthattheweightofthebodyforcesittoskipinsidet hewallsofthecavityandthe vehicleisunstable.Acontrolsystemisdesignedforthetor pedoandresultsofclosed-loop simulationshavebeenpresented. 1.4OverviewofThisDissertation Thisworkaimsatformulatingacontroldesignforasupercav itatingtorpedo.Equations ofmotionforthetorpedoarederived.Linearcontrolmethod ologies,LQRand -synthesis, havebeenappliedtoobtainvariousrobustinner-loopcontr ollersforthetorpedo. Chapter 2 brieydescribesthecongurationofthesupercavitatingt orpedo.Thelocationanddimensionofthecontrolsurfaceshavebeendiscuss ed.Adetailedderivationofthe equationsofmotionforthetorpedohasbeencarriedoutinth elatersectionsoftheChapter 2 .Variousreferenceframeshavebeenusedtoobtainthekinem aticequationsofthetorpedo.DynamicequationsarederivedusingNewton'sLaws.Va riousforcesexperienced bydifferentregionsofthetorpedohavebeenexplained. Chapter 3 describesthetwocavitymodelsthathavebeenusedinsimula tions.Thecavitycontourequationsandthecavitykinematicsarepresent ed.Thehydrodynamicplaning forceactsonthehullofthevehicleandgenerallydestabili zesthemotionofthevehicle. Theplaningforceequationsandtheplaningkinematicsarep resentedintheinthesecond halfofthischapter. Chapter 4 describeslinearizationoftheequationsofmotionusingsm alldisturbance theory.Itisobservedthatthelinearization,evenforasim pletrim,straight-levelight,can beverycomplicated.Thus,numericalmethodsareusedforth ispurpose. Chapter 5 describesthecontrolsynthesisforthetorpedo.Open-loop dynamicsare shown.Theclosed-loopproblemandvariousconstraintsont hetorpedohavebeenstated.

PAGE 17

6 Chapter 6 formulatesaLinearQuadraticRegulator(LQR)controldesign forthetorpedo.Controllersforpitchandrollratecontrolofthetorpe doareobtained.Theresultsfor linearclosed-loopsystemandaperturbedlinersystemhave beenshown.Variousrobustnessanalysismethodsshowthatthecontrollerisrobustton umericalerrorsincoefcients ofliftanddrag. Chapter 7 describesrobustcontroldesignforthetorpedousing synthesis.Controllers forpitchandrollanglehavebeenobtained,thatarerobustt ovariousparametricandmultiplicativeuncertainties.Controllersforpitchandrollr atearealsoobtainedthatarerobust touncertaintiesinthecavitator. Oneofthemissionsofamissileistargethoming.Designofho mingcontrollerfora supercavitatingtorpedoisdiscussedintheChapter 8 Thesummaryofthecurrentworkispresentedintheconcludin gchapterofthedissertation,Chapter 9 .Adiscussionofthefutureworkisdiscussedinthenextsect ionofthe thesis. Appendices A and B describesomeofthenotationsandmethodsusedforderivati on andsimulationoftheequationsofmotion. AdescriptionoftheCFDdatabase[ 5 ]usedtoobtainnumericalvaluesforthecodehas beenpresentedintheAppendix C

PAGE 18

CHAPTER2 NONLINEAREQUATIONSOFMOTION Althoughsupercavitationcanbeaveryhelpfulphenomenon, itpresentssignicant challengesinmodelingandcontrolofsupercavitatingvehi cles.Asasignicantportionof thevehicleislocatedinthecavity,thecontrol,guidancea ndstabilityofthevehiclehave tobemanagedbyverysmallregionsinfrontandaftofthevehi cle.Alsogenerationofthe cavitycanbeasignicantproblem.Themajorproblemsassoc iatedwiththesupercavitatingvehiclesmaybesummarizedas generationandmaintenanceofcavity balancingtheweightofthevehicle controlandguidance stability Figure 2.1 isagureofasupercavitatingtorpedothatismodeledinthi swork.Themain partsofthetorpedoarethecavitatorinthefrontandthefou rnsintheaftportion.The cavitatorisusedtogenerateandmaintainthecavity.Theca vitatorandthefournstogether arealsousedforcontrolandstabilityofthevehicle. Figure2.1SupercavitatingVehicle[ 8 ] 2.1CongurationoftheVehicle 2.1.1Body Forsimulationpurposesthebodyofthetorpedohasbeendesi gnedwith4separate sections.Therstiscylindricalinshapewithaverysmallr adiuscomparedtothemain body.Thecavitatorpivotsatthefrontofthissectionandfo rmsthenoseofthevehicle. Thesecondisasectionofaconewithamuchlargerradiusthan thecavitator,thethirda 7

PAGE 19

8 Cavitator Fins I IIIIIIV Figure2.2VehicleBodyisDividedInto4Sections. cylindricalpartmakingupmostofthelengthofthevehicle. Thelastportionofthevehicle istheconicalsectionaccountingforthenozzle.Thesegeom etriccomponentsaredepicted inFigure 2.2 2.1.2Cavitator Thecavitatoristheelementthatgeneratesacavityaroundt hetorpedo.Severaltypes ofcavitators,includingcones,wedgesandplates,havebee ninvestigated[ 4 ].Thisproject willconsideraconicalcavitatorasshowninFigure 2.2 .Themainparameterthatdenes aconicalcavitatorisitshalf-angle.Thecoefcientsofli ftanddrag,asfunctionsofhalfangle,forthecavitatorhavebeengeneratedusingCFDandtab ulatedinFine[ 5 ]. Thecavitatorinthismodelhasonedegreeoffreedomdenedb yarotationangleabout anaxisperpendiculartoitsaxisofsymmetry.Theshapeandl ocationofthecavityare anonlinearfunctionofthiscavitatordeectionangleandh alfangleofthecone.This shapedeterminesthepositionwherethecavityintersectst hebodyofthevehicle.Thus,the amountofwettedareaofthevehicleisdependentontheseang les,whichinturndetermines theefciencyofsupercavitationachieved.Anefforthasal sobeenmadetodesigna2DOF cavitator. Asstatedearlier,alargeportionofthevehicleisinthecav ity.Theliftproducedbythe cavitatorcombinedwiththeliftproducedbythenshelpsin balancingtheweightofthe body.

PAGE 20

9 2.1.3Fins Althoughthecavitatoriscapableofprovidingenoughliftt osustainthebodyinwater, itdoesnotusuallyprovidesufcientforcesandmomentstos tabilizeandcontrolthevehicle.Forthispurposensarerequiredintheaftportionof thevehicle.Thenshelpin counteractingthemomentsproducedbythecavitatorandals oprovidingsomeamountof lifttosupporttheweightofthebody.Therearefournsplac edsymmetricallyalongthe girthofthevehicle,nearthetail.Thensalsoarethecontr olsurfaces,astheycanprovide differentialforces,thuscausingcontrolmoments.Twooft hensshowninFigure 2.3 are horizontal(placedparalleltotheaxisofrotationofcavit ator),calledelevatorsandareused toaffectthelongitudinaldynamicsforthevehicle.Theoth ertwonsarecalledtherudders andareusedtoaffectthelateraldynamicstothevehicle. Thenshavetwodegreesoffreedom,asweepbackangleandana ngleofrotation. Theseangleswillbeexplainedinlatersectionsofthischap ter. Figure2.3CavitatorandFins 2.1.4Maneuvering Themotionofthevehicleiscontrolledbyallvecontrolsur faces,thefournsandthe cavitator.Inastraightlinemotionthecavitatorandtheel evatorsbalancetheweightofthe vehicle.Apropulsionforceatthetailpushesthevehiclefo rward.Theruddersusuallyhave azerodeectioninsuchacase. Thevehicledependsonabank-to-turnstrategyformakingat urn,andcannotusetraditionalmissilestrategiessuchasskid-to-turn.Thisdepen dencearisesbecausethehullofthe vehicleisincapableofprovidingaliftforce.Thensareth emainliftgeneratingsurfaces. Adifferentialforcefromthenscanbeusedtogenerateafor cetowardsthecenterofthe turn.

PAGE 21

10 2.2KinematicEquationsofMotion Thedenitionofasuitablecoordinatesystemanddegreesof freedomisaprecursor toanyformulationofequationsofmotion.Thederivationof theequationsofmotionof multi-bodysystemsishighlysimpliedbydeningvariousr eferenceframesandrelations betweenthem.Appendix A describesbrieytheconceptofreferenceframesandrotati on matrices.Theseconceptswillbeusedextensivelyintheder ivationofequationsofmotion. Thederivationoftheequationsofmotionwillbedoneintwop arts.First,thekinematic equationswillbederived.Theseincludetheformulationof thepositionvectors,velocities andaccelerationsofvariouspointsonthetorpedo.Next,th edynamicequationswillbe derived.Finally,Newton'sLawsyieldthedynamicequation sofmotion. 2.2.1OrientationoftheTorpedo Sixdegreesoffreedom(DOF)arerequiredtodescribethepos itionandorientationof thetorpedowhenitisconsideredarigidbody.Ofthese,thre eDOFsgivethelocationofa pointxedonthetorpedo.Theother3DOFsgivetheorientati onofthetorpedoinaxed referenceframe.Thepositionofthetorpedoinareferencef ramecanalsobeobtainedby theintegrationofitsvelocityinthatreferenceframe. Thetorpedoisassumedtobemovinginanearth-xedreferenc eframe E ,centeredat anyconvenientlychosenpointanddescribedbythebasisvec tors ( ˆ e 1 ; ˆ e 2 ; ˆ e 3 ) .Theearth-xed referenceframeisshowninFigure 2.4 .Thevectorˆ e 3 pointsinthedownwarddirection, i.e.,thedirectionofthegravity.Thevectorsˆ e 1 andˆ e 2 areplacedinthehorizontalplane suchthatthebasisvectorsformaright-handedcoordinates ystem.Asshowninthegure, ˆ e 1 pointstotherightandˆ e 2 pointsoutsidetheplaneofthepaper.Abody-xedframe B is denedbythebasisvectors ( ˆ b 1 ; ˆ b 2 ; ˆ b 3 ) soastosimplifythederivationoftheequationsof motion.Theframe B iscenteredat O ,thecenterofgravityofthetorpedo,andmoveswith thetorpedo.Thereferenceframe B isshownintheFigure 2.4 .ItcanbeseeninFigure 2.5 thatthetorpedohastwoplanesofsymmetrynamely ˆ b 1 ˆ b 2 and ˆ b 1 ˆ b 3 .Theplane ˆ b 1 ˆ b 3 is calledthelongitudinalplaneandplane ˆ b 1 ˆ b 2 ,thelateralplane.

PAGE 22

11 ˆ e 1 ˆ e 3ˆ e2Oˆ b1ˆ b2ˆ b3 Figure2.4Body-FixedandInertialFrames ˆ b 1 ˆ b 3 ˆ b 2 B Figure2.5PrincipalPlanesofSymmetryfortheTorpedo ˆ x 2 ˆ e 3 ; ˆ x 3ˆ x1Y Y ˆ e 1ˆ e2ˆ x 3ˆ x1Q Q ˆ y 3 ˆ y 1ˆ x 2 ; ˆ y 2ˆ y 3 ˆ y 1 ; ˆ b 1ˆ y 2F F ˆ b 3 ˆ b 2 Figure2.6EulerAnglesofRotation

PAGE 23

12 Transformationfrom E to B canbeachievedby3rotations.Manysuchsetsofrotations arepossible.TherotationstepschosenherearetheEuleran glesofrotation,whicharethe yaw,pitchandroll.Astherearethreerotationstobeperfor med,twointermediatereference frameswithbasisvectors ( ˆ x 1 ; ˆ x 2 ; ˆ x 3 ) and ( ˆ y 1 ; ˆ y 2 ; ˆ y 3 ) willhavetobeintroducedtoperform thetransformation.Therotations,tobeperformedinorder ,arelistedbelow. 1.Rotatetheframe E aboutˆ e 3 throughayawangle, Y ,toobtaintheframe ( ˆ x 1 ; ˆ x 2 ; ˆ x 3 ) 2.Rotate ( ˆ x 1 ; ˆ x 2 ; ˆ x 3 ) aboutˆ x 2 throughapitchangle, Q ,toobtaintheframe ( ˆ y 1 ; ˆ y 2 ; ˆ y 3 ) 3.Rotate ( ˆ y 1 ; ˆ y 2 ; ˆ y 3 ) aboutˆ y 1 througharollangle, F ,toobtaintheframe B Figure 2.6 showstheaboverotationsinorder.Basedontheabovedeniti onofrotations, thetransformationmatrixfrom E to B canbewrittenasinequation 2.1 8>>>><>>>>: ˆ b 1 ˆ b 2 ˆ b 3 9>>>>=>>>>; = 266664 1000 C F S F 0 S F C F 377775 266664 C Q 0 S Q 010 S Q 0 C Q 377775 266664 C Y S Y 0 S Y C Y 0 001 377775 8>>>><>>>>: ˆ e 1 ˆ e 2 ˆ e 3 9>>>>=>>>>; = 266664 C Q C Y C Q S Y S Q C Y S F S Q C F S Y S F S Q S Y + C Y C F S F C Q C F S Q C Y + S F S Y C F S Q S Y C Y S F C F C Q 377775 8>>>><>>>>: ˆ e 1 ˆ e 2 ˆ e 3 9>>>>=>>>>; = E B 8>>>><>>>>: ˆ e 1 ˆ e 2 ˆ e 3 9>>>>=>>>>; (2.1) 2.2.2OrientationoftheCavitator Asdescribedearlier,thecavitatorhasonlyonedegreeoffr eedom.Itcanrotateinthe longitudinalplaneaboutanaxisparalleltothe ˆ b 2 axis.TheorientationofthecavitatorxedaxeswithrespecttothebodyxedaxesisshowninFigure 2.7 .Thedeectionofthe cavitatorisgivenbyanangle, d c ,whichispositiveforapositivecavitatorrotationabout the ˆ b 2 direction.Thegeometriccenterofrotationofcavitatoris denotedby P CP isthe centerofpressureforthecavitator,whichisatadistance D CP from P ,alongˆ c 1

PAGE 24

13 B A ˆ b 1 ˆ b 3ˆ c 1ˆ c 3C d c ˆ b 3 ˆ b 1ˆ c 1P CP D CP A d c Figure2.7CavitatorReferenceFrame FromFigure 2.7 ,therotationmatrixfrom B tocavitatorxedframe C canbewritten asinEquation 2.2 8>>>><>>>>: ˆ c 1 ˆ c 2 ˆ c 3 9>>>>=>>>>; = 266664 C d c 0 S d c 010 S d c 0 C d c 377775 8>>>><>>>>: ˆ b 1 ˆ b 2 ˆ b 3 9>>>>=>>>>; (2.2) 2.2.3OrientationofFins Figure 2.8 showstheorientationofthen-xedreferenceframes.Forc onvenience,all thenframesareindicatedbybasisvectors ( ˆ f 1 ; ˆ f 2 ; ˆ f 3 ) .Intexttheywillberepresentedas ( ˆ f 1 ; ˆ f 2 ; ˆ f 3 ) fin ,wheresubscript fin correspondstoaparticularn. Allthenshave2DOFs,namelythesweepbackangle( q fin )andthenrotation( d fin ). Thesecanbedenedas Sweepbackangle( q fin ):Thisparameteristheangleofrotationofanaboutits ˆ f 3 axis.Thedirectionofrotationforpositivesweepbackissu chthatthenismoved towardtherearportionofthetorpedo.Duetothisdenition ,thepositivesweepback isalongthenegative ˆ f 3 directionforallns.Sweepbackangledeterminestheamoun t ofnthatisenvelopedinthecavity.

PAGE 25

14 Rudder 1 Rudder 2 Elevator 1Elevator 2 FRONT VIEW TOP VIEW B ˆ f 1 ˆ f 1 ˆ b 3 ˆ b 1 ˆ b 2 ˆ f 2 ˆ f 3 ˆ f 3 ˆ f 2 B ˆ f 1 ˆ f 1 ˆ b 1 ˆ f 2 ˆ f 3 ˆ f 3 ˆ f 2 ˆ b 2 ˆ b 3 Figure2.8RudderandFinReferenceFrames FinRotation( d fin ):Thisparameteristheangleofrotationofthenaboutits ˆ f 2 axis, inpositivethe ˆ f 2 direction.Finrotationdeterminesthemagnitudeanddirec tionof theforcesactingonthens. Theorderofrotationintheabovecaseisimportanttoobtain thecorrectequations. Sweepbackhastobeperformedbeforenrotation.Aninterme diatereferenceframe G withbasisvectors ( ˆ g 1 ; ˆ g 2 ; ˆ g 3 ) isintroducedsoastosimplifythederivationofrotationma trixfrom B tothencoordinates.Sweepbackalignsthen-xedframesw iththeintermediateframe G andthenanrotationputsthenframeinactualorientation withthens. Asthesecondrotationisidenticalinallcases,thetransfo rmationmatrixfromframe G to nframe F fin canbewrittenasinEquation 2.3

PAGE 26

15 ˆ g 3 ˆ g 1 ˆ g 2 ; ˆ f 2 ˆ f 3 ˆ f 1 d R 1 q R 1 q R 1 q R 1 ˆ b 1 ˆ b 3 ˆ b 2 ˆ g 1 ˆ g 2 ˆ g 3 ; Figure2.9Rudder1FinReferenceFrames 8>>>><>>>>: ˆ f 1 ˆ f 2 ˆ f 3 9>>>>=>>>>; fin = 266664 C d fin 0 S d fin 010 S d fin 0 C d fin 377775 8>>>><>>>>: ˆ g 1 ˆ g 2 ˆ g 3 9>>>>=>>>>; fin (2.3) Therotationmatrixforsweepbackandthetransformationma tricesfrom B to F fin frame foreachofthenscanbederivedeasily,andaresummarizedb elow. Rudder1 Figure 2.9 showsthesweepbackandnrotationforrudder1.Thematrice s fortransformationfrom B to R1 canbederivedasinEquation 2.4 andEquation 2.5 8<: ˆ g 1 ˆ g 2 ˆ g 3 9=; R 1 = 24 C q R 1 0 S q R 1 S q R 1 0 C q R 1 0 10 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.4) 8<: ˆ f 1 ˆ f 2 ˆ f 3 9=; R 1 = 24 C d R 1 0 S d R 1 010 S d R 1 0 C d R 1 35 24 C q R 1 0 S q R 1 S q R 1 0 C q R 1 0 10 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.5) Rudder2 Figure 2.10 showsthesweepbackandnrotationforrudder2.The matricesfortransformationfrom B to R2 canbederivedasinEquation 2.6 and Equation 2.7 8<: ˆ g 1 ˆ g 2 ˆ g 3 9=; R 2 = 24 C q R 2 0 S q R 2 S q R 2 0 C q R 2 010 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.6) 8<: ˆ f 1 ˆ f 2 ˆ f 3 9=; R 2 = 24 C d R 2 0 S d R 2 010 S d R 2 0 C d R 2 35 24 C q R 2 0 S q R 2 S q R 2 0 C q R 2 0 10 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.7)

PAGE 27

16 ˆ b 1 ˆ b 3 ˆ g 2 ; ˆ b 2 ˆ g 1 q R 2 q R 2 ˆ g 1 ˆ g 2 ˆ f 1 ˆ f 2 d R 2 d R 2 ˆ g 3 q R 2 ˆ g 2 Figure2.10Rudder2FinReferenceFrames Elevator1 Figure 2.11 showsthesweepbackandnrotationforElevator1.The matricesfortransformationfrom B to E1 canbederivedasinEquation 2.8 and Equation 2.9 8<: ˆ g 1 ˆ g 2 ˆ g 3 9=; E 1 = 24 C q E 1 S q E 1 0 S q E 1 C q E 1 0 00 1 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.8) 8<: ˆ f 1 ˆ f 2 ˆ f 3 9=; E 1 = 24 C d E 1 0 S d E 1 010 S d E 1 0 C d E 1 35 24 C q E 1 S q E 1 0 S q E 1 C q E 1 0 00 1 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.9) Elevator2 Figure 2.12 showsthesweepbackandnrotationforElevator2.The matricesfortransformationfrom B to E2 canbederivedasinEquation 2.10 and Equation 2.11 8<: ˆ g 1 ˆ g 2 ˆ g 3 9=; E 2 = 24 C q E 2 S q E 2 0 S q E 2 C q E 2 0 001 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.10) 8<: ˆ f 1 ˆ f 2 ˆ f 3 9=; E 2 = 24 C d E 2 0 S d E 2 010 S d E 2 0 C d E 2 35 24 C q E 2 S q E 2 0 S q E 2 C q E 2 0 001 35 8<: ˆ b 1 ˆ b 2 ˆ b 3 9=; (2.11) Equations 2.2 to 2.11 willbeusedinlatersectionstotransformtheforcesonnsa nd cavitatortothebody-xedframe.2.2.4AngleofAttackandSideslip Thebody-xedreferenceframehasbeendened,buttheveloc ityofvariouspointson thebodyofthetorpedoisyettobedened.Thetorpedoiscons ideredasarigidbody.Ifthe velocityofacertainpointonarigidbodyisknown,theveloc ityatanyotherpointonthat

PAGE 28

17 ˆ g 3 ˆ g 1 ˆ g 2 ; ˆ f 2 ˆ f 3 ˆ f 1 d E 1 q R 1 ˆ b 1 ˆ g 1 ˆ g 2 ˆ g 3 ; ˆ b 3 ˆ b 2 q E 1 q E 1 Figure2.11Elevator1FinReferenceFrames ˆ b 1 ˆ g 1 ˆ g 1 ˆ g 2 ˆ f 1 ˆ f 2 d R 2 ˆ g 3 ˆ g 2 ˆ b 2 q E 2 d E 2 q E 2 ˆ g 3 ; ˆ b 3 q E 2 Figure2.12Elevator2FinReferenceFrames bodycanbefoundbyknowingtherotationmatrices.Thus, V = u ˆ b 1 + v ˆ b 2 + w ˆ b 3 willbe takenasthevelocityof CG ofthetorpedo.Thevelocityofotherpointsonthetorpedoca n bedenedsubsequently.Twoveryusefulparameters,angleo fattackandangleofsideslip canbedenedinconjunctionwiththeorientationofthebody axiswiththevelocityof CG Figure 2.13 showstheseparametersandtheirgeometricinterpretation Angleofattack, a ,canbedenedastheanglebetweentheprojectionofvelocit y V onto ˆ b 2 ˆ b 3 planeandthe ˆ b 1 axis.Angleofattackispositivewhenthenoseofthetorpedo pointsabovethevelocityvectorofthetorpedo.Asbefore,a nintermediateframe F given by ( ˆ f 1 ; ˆ f 2 ; ˆ f 3 ) canbedescribedbyrotationofthe B framebyanangle a .Thustherotation

PAGE 29

18 ˆ f 1 u w v ˆ b 1 ˆ b 2 ˆ b 3 a b V = u ˆ b 1 +( v )( ˆ b 2 )+ w ˆ b 3ˆ g1 Figure2.13AngleofAttack( a )andSideslip( b ) matrixfrom F body to B canbewritten. 8>>>><>>>>: ˆ b 1 ˆ b 2 ˆ b 3 9>>>>=>>>>; = 266664 C a 0 S a 010 S a 0 C a 377775 8>>>><>>>>: ˆ f 1 ˆ f 2 ˆ f 3 9>>>>=>>>>; body (2.12) Theangleofsideslip, b ,isdenedastheanglebetweentheactualvelocity V and theprojectionof V onto ˆ b 2 ˆ b 3 plane.Again,aframe G body canbedenedbyrotationof F body byanangleequalto b innegative ˆ f 3 direction,thusgivingarotationmatrixasin Equation 2.13 8>>>><>>>>: ˆ g 1 ˆ g 2 ˆ g 3 9>>>>=>>>>; body = 266664 C b S b 0 S b C b 0 001 377775 8>>>><>>>>: ˆ f 1 ˆ f 2 ˆ f 3 9>>>>=>>>>; body (2.13) Velocityof CG ofthetorpedointhe G body framecanbewrittenas V ˆ g 1 ,where V is magnitudeof V .Itwillbeseenthatdragandliftonthetorpedocanbeobtain edinthis frame.Thusatransformationfrom G body to B isimportant.ItisgivenbyEquation 2.14 8>>>><>>>>: ˆ b 1 ˆ b 2 ˆ b 3 9>>>>=>>>>; = 266664 C b C a C a S b S a S b C b 0 C b S a S a S b C a 377775 8>>>><>>>>: ˆ g 1 ˆ g 2 ˆ g 3 9>>>>=>>>>; body (2.14) UsingEquation 2.14 V canberewrittenasinEquation 2.15

PAGE 30

19 V = V ˆ g 1 = VC b C a | {z } u ˆ b 1 VS b | {z } v ˆ b 2 + VC b S a | {z } w ˆ b 3 (2.15) where V 2 = V 2 = u 2 + v 2 + w 2 .FromFigure 2.13 ,relationsbetweenthevelocitycomponentsandtheanglesofattackandsideslipcanbederived. tan a = w u (2.16) sin b = v V (2.17) Thoughthematrix G body B inEquation 2.14 hasbeendenedforthebody-xedreferenceframeandvelocityof CG ofthetorpedo,theequationisvalidforanyotherrigid partofthebodylikethensandcavitator.Thus,incaseofa n,thevelocity V would correspondtoapoint(likethetip,centerofpressureetc.) onthatn,and G fin B matrix wouldcorrespondtothen-xedreferenceframe.Inthiscas ethevelocityofcenterof pressureofthenwillbeusedtodenetheaboveparameters. Thus,obtaining a fin and b fin isatwostepprocess: 1.Obtainthevelocityofcenterofpressureofn. V C Pbody = V cg + E w B r cgCP (2.18) where V C Pbody isvelocityof CP ofnin B frame, V cg isthevelocityof CG ofthe torpedoin E frame, E w B isangularvelocityof B in E ,and r cgCP ispositionvector from CG to CP fin .Equation 2.18 canberewrittenasinEquation 2.19 8<: u fin v fin w fin 9=; B = 8<: u v w 9=; cg + ˆ b 1 ˆ b 2 ˆ b 3 pqr X fin Y fin Z fin (2.19) where r cgCP = X fin ˆ b 1 + Y fin ˆ b 2 + Z fin ˆ b 3 2.Transformthevelocity(in E )of CP ofnfromframe B toframeofcorresponding n.Thistransformationisobtainedbyusingrotationmatri cesderivedinEquations 2.3 to 2.11

PAGE 31

20 8<: u R 1 v R 1 w R 1 9=; R 1 = 24 C d R 1 0 S d R 1 010 S d R 1 0 C d R 1 35 24 C q R 1 0 S q R 1 S q R 1 0 C q R 1 0 10 35 8<: u R 1 v R 1 w R 1 9=; B (2.20) 8<: u R 2 v R 2 w R 2 9=; R 2 = 24 C d R 2 0 S d R 2 010 S d R 2 0 C d R 2 35 24 C q R 2 0 S q R 2 S q R 2 0 C q R 2 0 10 35 8<: u R 2 v R 2 w R 2 9=; B (2.21) 8<: u E 1 v E 1 w E 1 9=; E 1 = 24 C d E 1 0 S d E 1 010 S d E 1 0 C d E 1 35 24 C q E 1 S q E 1 0 S q E 1 C q E 1 0 00 1 35 8<: u E 1 v E 1 w E 1 9=; B (2.22) 8<: u E 2 v E 2 w E 2 9=; E 2 = 24 C d E 2 0 S d E 2 010 S d E 2 0 C d E 2 35 24 C q E 2 S q E 2 0 S q E 2 C q E 2 0 001 35 8<: u E 2 v E 2 w E 2 9=; B (2.23) ThelefthandtermsinEquations 2.20 to 2.23 givethevelocitycomponentsatthe CP forcorrespondingns,inthatnframe.ThesecanbeusedinE quations 2.16 and 2.17 to ndtheangleofattackandsideslipforaparticularn.2.2.5KinematicEquations Velocityofthe CG ofthetorpedohasbeendenedintheprevioussection.Thatv elocity denesthetranslationalkinematicsforthetorpedo.Analo goustothisquantity,angular velocityisrequiredtodenetherotationalkinematics.Th eangularvelocityofthehullhas components p q and r intheframe B E w B D = p ˆ b 1 + q ˆ b 2 + r ˆ b 3 (2.24) Asthetransformationfrom E to B hasalreadybeendenedintermsofrotationalmotions givebyEulerangles,theangularratescanalsobeobtainedb ydifferentiationofEuler angles.Thus,anotherformofEquation 2.24 canbewrittenasinEquation 2.25 E w B = Y ˆ e 3 + Q ˆ x 2 + F ˆ b 1 (2.25) TherotationmatricesfromEquation 2.1 canbesubstitutedintoEquation 2.25 toobtain Equation 2.26 E w B =( F S Q Y ) ˆ b 1 +( Y C Q S F + Q C F ) ˆ b 2 +( Y C Q C F Q S F ) ˆ b 3 (2.26)

PAGE 32

21 Equations 2.24 and 2.26 canbeequatedtoobtainEquation 2.27 8>>>><>>>>: p q r 9>>>>=>>>>; = 266664 S Q 01 C Q S F C F 0 C Q C F S F 0 377775 8>>>><>>>>: Y Q F 9>>>>=>>>>; (2.27) Equation 2.27 canberewrittenasinEquation 2.28 p = F S Q Y q = Y C Q S F + Q C F r = Y C Q C F Q S F (2.28) ToapplyNewton'sLaws,accelerationofthe CG isrequired.Thevaluesof p q r willbetheangularaccelerationsoftorpedoin B .Thetranslationalaccelerationcanbe calculatedbytimedifferentiationof V inNewtonianframe.Adifferentiationformulacan beusedtondthetimederivative,insomeframe,foravector denedinsomeotherrelated frame. d dt ( v ) I = d dt ( v ) B + I w B v (2.29) where,subscript I denotesNewtonian(inertial)frame,and B isthebody-xedframe. I w B isangularvelocityofthebody(orbody-xedframe)inthe I frame, v isthevelocityin I frame,ofthepointwhereaccelerationisdesired.Usingthe formulatheaccelerationof CM oftorpedoin E canbeobtained. E a CM = 8>>>><>>>>: u v w 9>>>>=>>>>; + ˆ b 1 ˆ b 2 ˆ b 3 pqr uvw (2.30) = 8>>>><>>>>: u + qw vr v + ur pw w + pv uq 9>>>>=>>>>; ˆ b (2.31)

PAGE 33

22 Similarly,therotationalaccelerationwillberequiredin theframe E .Thiscanbewritten analogoustoEquation 2.30 E a B = 8>>>><>>>>: p q r 9>>>>=>>>>; + ˆ b 1 ˆ b 2 ˆ b 3 pqrpqr = 8>>>><>>>>: p q r 9>>>>=>>>>; (2.32) Thepositionoftorpedocanbefoundbyintegratingtheveloc ity.Let ( x ; y ; z ) represent thecoordinatesof CG inframe E .Thus,thetimederivativeofthesecoordinatesin E should representthevelocitycomponentsofthetorpedoin E frame.Whenthesetimederivatives aretransformedtobody-xedframe,theywouldbeequivalen ttothevelocitycomponents inbody-xedframe. 8>>>><>>>>: x y z 9>>>>=>>>>; B = 8>>>><>>>>: u v w 9>>>>=>>>>; (2.33) Equation 2.1 canbesubstitutedinEquation 2.33 toobtainEquation 2.34 8>>>><>>>>: x y z 9>>>>=>>>>; E = 266664 C Q C Y C Q S Y S Q C Y S F S Q C F S Y S F S Q S Y + C Y C F S F C Q C F S Q C Y + S F S Y C F S Q S Y S F C Y C F C Q 377775 8>>>><>>>>: u v w 9>>>>=>>>>; (2.34) 2.3DynamicEquationsofMotion Nowthattheaccelerationsofvariouspartsofthetorpedoar eknown,Newton'sLaws canbeusedtoderivethedynamicequationsofmotion.Newton 'slawsstatethattherateof changeofmomentumisequaltothesumofexternalforceappli edonthebody.Equation 2.35 canbeobtainedfromNewton'slawsbyanassumptionthatthem assofthetorpedois constant.Thisassumptionisvalidforashortperiodoftime .Theequationsare

PAGE 34

23F = P = m a (2.35) where P isthelinearmomentum, m ismassofthebody, a istheaccelerationand F isallthe forcesofthebody.UsingEquation 2.31 inEquation 2.35 ,Newton'sLawsforthetorpedo canberewrittenasinEquation 2.36 m 8>>>><>>>>: u + qw vr v + ur pw w + pv uq 9>>>>=>>>>; ˆ b = F (2.36) Newton'slawscanbeextendedtorotation.Equation 2.37 aretheNewton'sLawsfor rotationalmotion.M = H = I CM a + E w B H (2.37) where H ( = I CM E w B )istheangularmomentum, I CM ismomentofinertiamatrixofthe body, a istheangularaccelerationand M isallthemomentsonthebody.Itshouldbe notedthatabovestatedNewton'slawsareonlyvalidwhenthe quantitiesarecalculatedin aninertialframeofreference.Thus,thequantitieshavebe encalculatedinframe E .Using Equation 2.32 ,theNewton'sLawforrotationcanbewrittenasinEquation 2.38 0BBBB@ I 1 00 0 I 2 0 00 I 3 1CCCCA 8>>>><>>>>: p q r 9>>>>=>>>>; + ˆ b 1 ˆ b 2 ˆ b 3 pqr I 1 pI 2 qI 3 r = M (2.38) Toderivetheequations,theforcesoneachofthepartswillb ecalculatedindividually, andthenexpressedinbody-xedreferenceframe,i.e.,summ ationwillbedoneinbody referenceframe.Therotationmatricesderivedinprevious sectionswillbeusedextensively forthispurpose.

PAGE 35

24 Varioustypesofforcesareexperiencedbyamovingtorpedoi nwater.Theseforcescan besummarizedasfollows: HydrodynamicForces :Thesearetheforcesexertedbytheuidonthetorpedo. Thustheyexistwhenevertheuidisincontactwithbody.For supercavitatingvehicle,mostofthebody(hull)isinsidethecavity.Onlyapor tionofthensandthe cavitatorareincontactwiththeuid.Thustheliftanddrag oncavitatorandnsare onlyhydrodynamicforces.Thecoefcientsofliftanddragf orthensandcavitator arefunctionsoftheirangleofattack,immersion,sweepbac kangle,angleofrotation etc.andaretabulatedinadatabase[ 5 ].Thisdatabasewillbeinterpolatedandextrapolatedtocalculatethenumericalvaluesfortheforces.When thebodyofthevehicle otherthanthensorthecavitatorcomesintocontactwithth ecavitywall,thebody experiencesforcescalledastheplaningforcesduringcont actwiththecavitywall. Detailsofthesekindofhydrodynamicforceswillbedealtwi thinChapter 3 F Hydrodynamic = F R 1 + F R 2 + F E 1 + F E 2 + F c + F planing (2.39) M Hydrodynamic = M R 1 + M R 2 + M E 1 + M E 2 + M c + M planing (2.40) GravitationalForces :Thisisthegravityforcesonthebody.Asthesummationof momentswillbetakenaboutthecenterofgravity,thegravit ationalforceswillnot contributetothesummationonmoments.Theywillappearonl yinsummationof forces. Propulsive :Thetorpedohasapropulsionsystem,whichissimilartotha tofrockets. Thelineofactionofthepropulsiveforceisassumedtobepas singthroughcenterof gravityandalong ˆ b 1 axis.Thusthisforcewillalsocontributetothesumofforce s, andnotmoments.Thepropulsiveforcesareprovidedbyburni ngthefuel,butfor simplicityitwillbeassumedthatthemassofthetorpedorem ainsconstant. Immersion: Thisforceresultsfromthevehiclebeingsubmergedunderwa ter.Since thecenterofbuoyancyisassumedtobecoincidentwiththece nterofgravitythis forcecontributesonlytotheexternalforceterm. Thetotalforcesandmomentsareexpressedintermsofthesec omponents. F = F Hydrodynamic + F Grav + F Prop + F Immersion (2.41) M = M Hydrodynamic (2.42) 2.3.1ForcesonCavitator Figure 2.14 showstheforcesactingonthecavitator.Coefcientoflift( cl c )anddrag ( cd c )forthecavitatorarefunctionsofangleofattack, a c ,andhalf-angle, g 1 2 ,ofthecavitator.Half-angle,foracone,istheanglemadebyaxisofthe conewithanylinepassing

PAGE 36

25 throughthevertexandlyinginthesurfaceofthethecone.Th isparameterdenesthemain geometryoftheconicalcavitator.Thevaluesof cl c and cd c aredeterminedusingCFDand tabulatedin[ 5 ].Thesevalueshavebeenextrapolatedtocalculateliftand drag. L c = 1 2 r V 2 c S c cl c ( a c ; g 1 2 ) (2.43) D c = 1 2 r V 2 c S c cd c ( a c ; g 1 2 ) (2.44) where S c isthecross-sectionalareaofthecavitatorbase.Directio nsoflift( L c )anddrag ( D c )areasshowninFigure 2.14 (b).Thesecanbetransformedtothebodyaxesusing 2.2 and 2.14 forthecavitator. 8>>>><>>>>: ˆ b 1 ˆ b 2 ˆ b 3 9>>>>=>>>>; = C B ( d c ) G cav C ( a c ; b c ) 8>>>><>>>>: ˆ g 1 ˆ g 2 ˆ g 3 9>>>>=>>>>; cav (2.45) Thustheforcesoncavitator,inbodyframe,canbewritten. F c = 8>>>><>>>>: F c ; x F c ; y F c ; z 9>>>>=>>>>; B = 8>>>><>>>>: D c ( a c ; g 1 2 ) 0 L c ( a c ; g 1 2 ) 9>>>>=>>>>; G cav = 266664 C d c 0 S d c 010 S d c 0 C d c 377775 266664 C b c C a c C a c S b c S a c S b c C b c 0 C b c S a c S a c S b c C a c 377775 8>>>><>>>>: D c ( a c ; g 1 2 ) 0 L c ( a c ; g 1 2 ) 9>>>>=>>>>; (2.46) where F c isa3x1matrixwitheachrowcorrespondingtoeachdirection in B basis.The forcesareassumedtobeactingat CP ofthecavitator.Oncetheforceshavebeencalculated, themomentaboutanypointcanbecalculated.

PAGE 37

26 M c = r CPcav F c (2.47) where r CPcav isthepositionvectorfromthatpointto CP ofcavitator.Itisassumedthatthe CP lieson ˆ b 1 whencavitatordeectionis0,anditscoordinateswithresp ecttobodyorigin O ,inthiscase,are ( X c ; 0 ; 0 ) .ThusfromFigure 2.7 ,thecoordinatesof CP canbewritten forthecasewhenthecavitatorisdeected. r CPcav = 8>>>><>>>>: X c + D CP C d c 0 D CP S d c 9>>>>=>>>>; body (2.48) Themomentsonthecavitatorinbody-xedcanbeobtainedbys ubstitutingEquations 2.46 and 2.48 inEquation 2.47 M c =[( X c + D CP C d c ) ˆ b 1 D CP S d c ˆ b 3 ] 266664 C d c 0 S d c 010 S d c 0 C d c 377775 266664 C b c C a c C a c S b c S a c S b c C b c 0 C b c S a c S a c S b c C a c 377775 8>>>><>>>>: D c ( a c ; g 1 2 ) 0 L c ( a c ; g 1 2 ) 9>>>>=>>>>; (2.49) b c a c ˆ c 1 ˆ c 3 ˆ g 1 ˆ c 2 (a) ˆ c 1 ˆ g 2 ˆ g 3 M c D c L c ˆ g 1 (b) CP Figure2.14Cavitator:(a)AngleofAttackandSideslipand(b )HydrodynamicForces

PAGE 38

27 wetted region D n ˆ f fin 3 L n ˆ f fin 1 ˆ f fin 2 V h a S L S 0 Figure2.15FinGeometry 2.3.2ForcesonFins FinforcesarealsoextrapolatedfromtheCFDdatabase[ 5 ],whichgivesthevaluesof coefcientsoflift( cl fin )anddrag( cd fin )forns.Thesevaluesarefunctionsofangle ofattack a fin ,nsweepback q fin ,nrotation d fin ,nimmersion I fin andngeometry. Figure 2.15 showstheseparametersgraphically,andtheycanbedeneda sfollows: FinGeometry:Thegeometryparametersfornsare L and S ,andwedgehalfangle ( h a ),asshowninFigure 2.15 .Theseparametersarexedaccordingtothevalues givenbythedatabase,soastocalculatetheforcesaccurate ly. FinImmersion:Asthenispartiallywettedbyuid,thewett edlengthcanberepresentedbyparameter S 0 alongn Y -axis.Theimmersion I fin canbedenedasthe ratioofthewettedlengthofthentoitstruelength. I fin =( S 0 = S ) fin (2.50) I fin determinesthetotalhydrodynamicforceonthen. FinRotation( d fin ):Asdenedearlier,thisisrotationaboutn ˆ f 2 axis.Thisdeterminesthedirectionofthehydrodynamicforce.Thusnrotat ionisusedasprimary controlforthetorpedo. FinSweepback( q fin ):Asdenedearlier,thisisrotationaboutn ˆ f 3 axis.Sweepback determinesthewettedsurfaceofthen,thusthehydrodynam icforceonthen. AngleofAttack:Angleofattackfortheniscalculatedasde scribedinFigure 2.15 andsection 2.2.4 Thedatabasegives cl fin and cd fin asafunctionof a fin q fin and I fin ,thusliftanddragon thenscanbecalculatedbythenormalizingfactor. L n = 1 2 r V 2 S 2 fin cl fin (2.51) D n = 1 2 r V 2 S 2 fin cd fin (2.52)

PAGE 39

28 Where S fin isthelengthofthenasshowninFigure 2.15 .Theseforceshavedirections asshowninFigure 2.15 .BeforesubstitutinginEquation 2.39 ,theseforceshavetobe transformedtobody-xedreferenceframe.Thisprocessinv olvesfollowingtworotations: 1.Rotatetheframe F fin (whichhas L n and D n alongitsbasisvectors)toalignitwith then-xedframeusingEquation 2.14 and 2.Rotatetheaboveobtainedn-xedframetoobtainthebodyxedframeusingEquations 2.3 to 2.11 Thustheforcesonthensinbody-xedframeaxiscanbeobtai ned. Rudder1 8<: F R 1 ; x F R 1 ; y F R 1 ; z 9=; B = 24 C q R 1 S q R 1 0 00 1 S q R 1 C q R 1 0 35 24 C d R 1 0 S d R 1 010 S d R 1 0 C d R 1 35 24 C b R 1 C a R 1 C a R 1 S b R 1 S a R 1 S b R 1 C b R 1 0 C b R 1 S a R 1 S a R 1 S b R 1 C a R 1 35 8<: D R1 0 L R1 9=; (2.53) Rudder2 8<: F R 2 ; x F R 2 ; y F R 2 ; z 9=; B = 24 C q R 2 S q R 2 0 00 1 S q R 2 C q R 2 0 35 24 C d R 2 0 S d R 2 010 S d R 2 0 C d R 2 35 24 C b R 2 C a R 2 C a R 2 S b R 2 S a R 2 S b R 2 C b R 2 0 C b R 2 S a R 2 S a R 2 S b R 2 C a R 2 35 8<: D R2 0 L R2 9=; (2.54) Elevator1 8<: F E 1 ; x F E 1 ; y F E 1 ; z 9=; B = 24 C q E 1 S q E 1 0 S q E 1 C q E 1 0 00 1 35 24 C d E 1 0 S d E 1 010 S d E 1 0 C d E 1 35 24 C b E 1 C a E 1 C a E 1 S b E 1 S a E 1 S b E 1 C b E 1 0 C b E 1 S a E 1 S a E 1 S b E 1 C a E 1 35 8<: D E1 0 L E1 9=; (2.55)

PAGE 40

29 Elevator2 8<: F E 2 ; x F E 2 ; y F E 2 ; z 9=; B = 24 C q E 2 S q E 2 0 S q E 2 C q E 2 0 001 35 24 C d E 2 0 S d E 2 010 S d E 2 0 C d E 2 35 24 C b E 2 C a E 2 C a E 2 S b E 2 S a E 2 S b E 2 C b E 2 0 C b E 2 S a E 2 S a E 2 S b E 2 C a E 2 35 8<: D E2 0 L E2 9=; (2.56) Equations 2.53 to 2.56 givethecomponentsofhydrodynamicforcesonnsinbody-x ed frame.Whatremainsistondthemomentoftheseforcesabout CG ofbody.Themoments canbeobtainedinanalogoustoEquation 2.47 M n = r nCG CP F n (2.57) Inthiscase,therootofnsisxedtobody,anditcanmovethu smovingthe CP ofn relativetoroot.Thepositionof CP fromrootisalsoknowwithrespecttoncoordinates. r nCG root = X fin root ˆ b 1 + Y fin root ˆ b 2 + Z fin root ˆ b 3 (2.58) r nroot CP = D x fin C P ˆ f 1 + D y fin C P ˆ f 2 (2.59) where ( ˆ f 1 ; ˆ f 2 ; ˆ f 3 ) isn-xedcoordinatesforcorrespondingn.Equations 2.58 and 2.59 canbeaddedbyusingmatricesgiveninEquations 2.3 to 2.11 .Thus,thepositionvector from CG to CP ofnscanbeobtained. Rudder1 8<: X R 1 Y R 1 Z R 1 9=; B = 8<: X root R 1 Y root R 1 Z root R 1 9=; B + 24 C q R 1 S q R 1 0 00 1 S q R 1 C q R 1 0 35 24 C d R 1 0 S d R 1 010 S d R 1 0 C d R 1 35 8<: D x R 1 C P D y R 1 C P 0 9=; R 1 (2.60) Rudder2 8<: X R 2 Y R 2 Z R 2 9=; B = 8<: X root R 2 Y root R 2 Z root R 2 9=; B + 24 C q R 2 S q R 2 0 00 1 S q R 2 C q R 2 0 35 24 C d R 2 0 S d R 2 010 S d R 2 0 C d R 2 35 8<: D x R 2 C P D y R 2 C P 0 9=; R 2 (2.61)

PAGE 41

30 Elevator1 8<: X E 1 Y E 1 Z E 1 9=; B = 8<: X root E 1 Y root E 1 Z root E 1 9=; B + 24 C q E 1 S q E 1 0 S q E 1 C q E 1 0 00 1 35 24 C d E 1 0 S d E 1 010 S d E 1 0 C d E 1 35 8<: D x E 1 C P D y E 1 C P 0 9=; E 1 (2.62) Elevator2 8<: X E 2 Y E 2 Z E 2 9=; B = 8<: X root E 2 Y root E 2 Z root E 2 9=; B + 24 C q E 2 S q E 2 0 S q E 2 C q E 2 0 001 35 24 C d E 2 0 S d E 2 010 S d E 2 0 C d E 2 35 8<: D x E 2 C P D y E 2 C P 0 9=; E 2 (2.63) Equations 2.60 to 2.63 givethepositionvectorfrom CG to CP ofthens.TheseequationsinconjunctionwithEquations 2.53 to 2.56 ,usedin 2.57 ,givestheexternalmoments onnsaboutthe CG M n = ˆ b 1 ˆ b 2 ˆ b 3 X fin Y fin Z fin F fin ; x F fin ; y F fin ; z (2.64) 2.3.3GravitationalForces Forsimplicity,mass( m )ofthetorpedoisassumedtobeconstantovertime.This isnotthecaseinrealitybuttheapproximationisreasonabl eforconsideringshorttime maneuvers.Accelerationduetogravity, g ,isalsoassumedtobeconstantastorpedomoves inspace.Thedirectionofgravityisgivenbyˆ e 3 axis.Thus,thegravitationalforcecanbe writtenasinEquation 2.65 F grav = mg ˆ e 3 (2.65)

PAGE 42

31 Equation 2.65 canbere-expressedinbodyframeofreferenceusingEquatio n 2.1 F grav = 266664 C Q C Y C Q S Y S Q C Y S F S Q C F S Y S F S Q S Y + C Q C F S F C Q C F S Q C Y + S F S Y C F S Q S Y S F C Y C F C Q 377775 8>>>><>>>>: 00 mg 9>>>>=>>>>; E = mg 8>>>><>>>>: S Q S F C Q C F C Q 9>>>>=>>>>; B (2.66) 2.3.4EquationsofMotion Nowthatamathematicalformulationofforcesonthetorpedo hasbeenachieved,these equationscanbesubstitutedintoEquations 2.36 to 2.42 toobtainthedynamicequationsof motion.Thustheforceequationsofmotioncanbesummarized asinEquation 2.67 2.3.4.1Forceequations m 8>>>><>>>>: u + qw vr v + ur pw w + pv uq 9>>>>=>>>>; B = F immersion + F planing + 8>>>><>>>>: F prop 00 9>>>>=>>>>; B + 8>>>><>>>>: F R 1 ; x F R 1 ; y F R 1 ; z 9>>>>=>>>>; B + 8>>>><>>>>: F R 2 ; x F R 2 ; y F R 2 ; z 9>>>>=>>>>; B + 8>>>><>>>>: F E 1 ; x F E 1 ; y F E 1 ; z 9>>>>=>>>>; B + 8>>>><>>>>: F E 2 ; x F E 2 ; y F E 2 ; z 9>>>>=>>>>; B + mg 8>>>><>>>>: S Q S F C Q C F C Q 9>>>>=>>>>; B + 266664 C d c 0 S d c 010 S d c 0 C d c 377775 266664 C b c C a c C a c S b c S a c S b c C b c 0 C b c S a c S a c S b c C a c 377775 8>>>><>>>>: D c ( a c ; g 1 2 ) 0 L c ( a c ; g 1 2 ) 9>>>>=>>>>; C (2.67) SomeissuesshouldbenotedforEquation 2.67 Theplaningforceshavenotyetbeencalculatedintheequati onsofmotion.The calculationoftheseforceswillbeexplainedinChapter 3 .Inthecaseofxedcavity modelforthetorpedo,theseforcesareneglectedbyassumpt ionthatthevehicleis centeredinthecavity.

PAGE 43

32 Thepropulsionforceisconstrainedtobealongnegative ˆ b 1 axis. 2.3.4.2Momentequations 266664 I 1 00 0 I 2 0 00 I 3 377775 8>>>><>>>>: p q r 9>>>>=>>>>; + ˆ b 1 ˆ b 2 ˆ b 3 pqr I 1 pI 2 qI 3 r = M planing + ˆ b 1 ˆ b 2 ˆ b 3 X R 1 Y R 1 Z R 1 F R 1 ; x F R 1 ; y F R 1 ; z + ˆ b 1 ˆ b 2 ˆ b 3 X R 2 Y R 2 Z R 2 F R 2 ; x F R 2 ; y F R 2 ; z + ˆ b 1 ˆ b 2 ˆ b 3 X E 1 Y E 1 Z E 1 F E 1 ; x F E 1 ; y F E 1 ; z + ˆ b 1 ˆ b 2 ˆ b 3 X E 2 Y E 2 Z E 2 F E 2 ; x F E 2 ; y F E 2 ; z + ˆ b 1 ˆ b 2 ˆ b 3 X c + D C P C d c 0 D C P S d c F c ; x F c ; y F c ; z (2.68) SomeissuesshouldbenotedforEquation 2.68 SomeofthetermsinEquation 2.68 areshownasadeterminant.Theyneedtobe expandedandwrittenascomponentsinbody-xedframesoast oequatethelefthandandright-handterms. Momentsduetogravitationdonotappearbecausethemoments aretakenabout CG Again,themomentduetoplaningforceshavenotyetbeenexpl ained.Also,the momentduetoplaningforcesistobeneglectedinthecaseof xedcavitymodel. 2.3.4.3Orientationequations 8>>>><>>>>: Y Q F 9>>>>=>>>>; = 266664 0 S F C Q C F C Q 0 C F S F 1 S F S Q C Q C F S Q C Q 377775 8>>>><>>>>: p q r 9>>>>=>>>>; (2.69) 2.3.4.4Navigationequations 8>>>><>>>>: x y z 9>>>>=>>>>; E = 266664 C Q C Y C Q S Y S Q C Y S F S Q C F S Y S F S Q S Y + C Y C F S F C Q C F S Q C Y + S F S Y C F S Q S Y S F C Y C F C Q 377775 8>>>><>>>>: u v w 9>>>>=>>>>; (2.70)

PAGE 44

33 Equations 2.67 to 2.70 areacompletesetofequationsofmotionsforthesupercavit ating torpedo.

PAGE 45

CHAPTER3 CAVITYANDPLANINGDYNAMICS Ifthebodyisimmersedinasteadyow,itisreasonabletoass umethattheshapeofthe cavityisconstantandaxiallysymmetric.Itmaynot,howeve r,besoascavitiesmaypulsate duetore-entrantjetslocatedattheaftendsofavehicle[ 4 ].Thecavityshapeisgenerally speciedbygivingthelength L andthemaximumdiameter d m ofthecavity,ortheneness ratio L = d m .Thediameterisspeciedasfunctionofthearclengthalong thecenterlineof thecavity.Variousmodels,analytical,empiricalandnume rical,havebeendevelopedover theyearstodeterminetheabovementionedparametersforac avity.Waid[ 9 ]derived formulaeforthecavityshapesofdisksfrommeasurementsma deonthecavities.May[ 4 ] listsvariouscavitymodelsbasedonexperimentaldataforg eometriessuchasdisks,cones, spheres,etc.Manymodelsexistforsupercavitybehaviorba sedonslender-bodytheory, boundary-elementmethods,andReynolds-averagedNavier-S tokessolvers.However,for time-domainsimulationofvehicledynamics,itisdesirabl etousesimplemodels.Thisis particularlythecasewhenwestudytheguidance,controlan dstabilityoftheoverallbody. Twocavitymodelshavebeenusedinthesimulations.Theyare theMunzer-reichardt cavitymodelandtheLogvinovichcavitymodel.Bothcavitymo delsassumethecrosssectionofthecavityasbeingcircularinshape.Theradiuso fthecavityisspeciedas afunctionofthearclengthalongthecavitycenterline.The modelingofthecavity,for simulationpurposes,comprisesofthesecavitymodelsandt hedescriptionofthecavity centerlinebasedonmemoryeffects. 34

PAGE 46

35 3.1Munzer-ReichardtModel Reichardtshowedthatthenenessratioofthecavity L = d max isafunctiononlyofthe cavitationnumber s .Inotherwords,theratioisindependentofthecavitatorsh apeand isonlyfunctionof s .TheformulaeforthecavitydimensionsaccordingtotheMun zerReichardt[ 10 ]modelare R = R n 1 L L max 2 # 1 2 : 4 L max = 2 R n p C d C 1 : 24 s 1 : 123 0 : 6 R n = R C p Cd C 1 : 35 s 0 : 93 (3.1) IntheEquations 3.1 L max isthelengthofthecavity, L isarclengthalongthecavitycenterline, C d C isthecavitatordragcoefcient, R C istheradiusofthecavitatorand R thecavity radiusatthatarclengthonthecavitycenterline. 3.2LogvinovichCavityModel 3.2.1LogvinovichTheoryofIndependentExpansion Logvinovichhasmadethefollowingfundamentalobservatio n:“Eachcross-sectionof thecavityexpandsrelativetothepathofthebody-centeral mostindependentlyofthesubsequentorprecedingmotionofthebody...”[ 11 ]. Logvinovich'scavitycontourequationsareanotherexampl eofthekindwhichconsider thecavitytobeintheshapeofanellipsoid.Thefollowingfo rmulasaretakenfromLogvinovich[ 11 ]forthecavityshape R = R k vuut 1 1 R 21 R 2k 1 t t k 2 k R = R 2k R t k 1 R 21 R 2k 1 k 1 t t k 1 t t k 2 ( 1 k ) k (3.2) IntheEquations 3.2 R k denotesthemaximumcavityradius, R and R t k denotecavity radiusataparticularpointintimeonthecavitycenterline t timetakenforcavityatthat pointtoevolvefrominception, t k isthetimetakenbycavitatortotravelonebodylength

PAGE 47

36 ofthevehicle, k isacorrectionfactorand R isthetimerateofchangeofthecavityradius. Explicitassumptionismadeofthecavitatorbeingofdisksh apedinderivingtheabove formulae.However,inourcasethecavityduetoaconeshaped cavitatorisapproximated withtheoneduetoadiskshapedcavitator.Theformulaeinth eaboveformmakeit possibletocalculatethecavityprolealsowhen t = t k > 1,butusuallywhen t = t k > 1 : 5the boundariesofthecavityareundeterminable.Theystartbre akingupandfoambeginsto form.At t = 0thecontourexpressedbyEquation 3.2 ismatchedwiththatoftheleading partoftheempiricalcavity R = R n 1 + 3 x R n 1 3 (3.3) where R n isthecavitatordiameter.Themaximumcavityradius R k canbecalculated usingtheformula R k = R n 0 : 82 1 + s s (3.4) Byselecting x 1 = 2 R n asourmatchingpointweobtain R 1 = 1 : 92 R n .Acavity”correction factor”, k ,isalsorequiredinthecalculations.Fortheselection x 1 = 2 R n and k = 0 : 85, agoodcorrelationcanbeobtainedbetweentheformulaeinEq uation4.2andtheexperimentaldata.Undertheseconditionsthecavityhalflengthc anbeapproximatedusingthe formulainEquation 3.5 [ 11 ]. L k = R n 1 : 92 s 3 (3.5) Thecavityhalflengthwillbeusedtoapproximatethevalueo f t = t k as L = L k .Thevalidity ofthisapproximationcanbeseenfromEquation 3.10 .Fromtheaboveformulaewecan calculatethebasicdimensionsofthecavity.However,duri ngsimulationthecoordinates ofcavitycenterlineneedtobelocated.Thecenterlineisre quiredinordertodeterminethe nalshapeandpositionofthecavitywithrespecttothebody .Effectslike buoyancy and downwash needtobeconsideredfordeterminingthenalshapeofcavit ycenterline. 3.2.2CavityCenterline Thearclengthoftheactualpathisassumedtobeapproximate lyequaltothedistance fromthecavitatortotheplaningpointprojectedontothebo dyaxis.Thetimelaginconsideringthememoryeffectsshapingthecavitycanbeobtain edbytheexpression

PAGE 48

37 t = L u (3.6) Where,Listhedistancealongthecavitycenterline.Theabov evalueisobtained througha2 nd orderapproximation u avg = 1 2 ( u ( t )+ u ( t t )) = 1 2 u ( t )+ u ( t ) u ( t ) t = u ( t ) 1 2 u t (3.7) Theaveragevelocityoverthetimeperiod [ t ; t t ] multipliedbythetimeperiodwill givethedistancecovered, L ,intime t .Thedistance L canbewrittenas L = u ( t ) 1 2 u t t ) 1 2 u t 2 + u t L = 0(3.8) Now,solvethequadraticequationfor t t = u q u 2 4 1 2 u ( L ) u = u p u 2 2 uL u (3.9) Takingthelimitof u approachingzeroandusingtheL'Hospitalruleforcalculat ing limitsforindeterminateformswehave lim u 0 t = L u (3.10) Fromonepointofview,thevariationinthecavityshapeatap ointinspaceisduetothe memoryeffectsassociatedwiththepassageofcavitatorthr oughthatpointataprevious time.Thustocalculatethecavityshape,thecavitatorposi tionateachtime-stepisstored. Thecavitycenterlineisformedbythepositionofthecavita tortipatcurrenttimeandthe previous(n-1)time-steps,where'n'isthenumberofpoints onthefulllengthofthecavity. Wecanwrite 8>>>><>>>>: x i y i z i 9>>>>=>>>>; = 8>>>><>>>>: x c ( t ( i 1 ) D t ) y c ( t ( i 1 ) D t ) z c ( t ( i 1 ) D t ) 9>>>>=>>>>; (3.11)

PAGE 49

38 where ( x i ; y i ; z i ) arethecoordinatesofthe i th pointonthecavitycenterline, ( x c ; y c ; z c ) are thecoordinatesofthecavitatortipand D t isthetimedifferentialbetweenpoints i and i 1. Thebuoyancyforceactsonthecavityandthecenterlineisdi stortedovertime.Onewayof representingthebuoyancyforceisbyaconstantaccelerati onactingagainstgravity.Now takingintoaccounttheeffectofbuoyancyonthecavitycent erline,wehave 8>>>><>>>>: x i y i z i 9>>>>=>>>>; = 8>>>><>>>>: x c ( t ( i 1 ) D t ) y c ( t ( i 1 ) D t ) z c ( t ( i 1 ) D t ) 1 2 b (( i 1 ) D t ) 2 9>>>>=>>>>; i = 1to n (3.12) where b istheconstantupwardaccelerationduetobuoyancyontheca vity.Incontrast Logvinovich[ 11 ]hasadifferentformofequationsforthebuoyancyanddownw ash.They aregivenbythefollowingexpressions h buoyancy ( X )= g p V 2ZX 0 O k R 2 ( x ) dx (3.13) h downwash ( X )= F pr V 2ZX 0 dx R 2 ( x ) (3.14) Intheseequations, X isthepositiononthebodycenterlineinframe B R isradiusof cavityasfunctionofpositionalongbodycenterline, h buoyancy and h downwash representthe verticaldisplacementsduetobuoyancyanddownwashrespec tivelyatapointonthebody centerline, O k isthevolumeofthecavityenclosedbythecross-sectionatt hatpointon thebodycenterlineand V isthevelocityofthecenter-of-massofthebody.Inthepres ent workthecavitymodelislimitedtotheeffectssuchasbuoyan cy,downwashandmemory effects.Othercomplexphenomenaassociatedwiththeshape ofthecavityareignoredin thecurrentinvestigation. 3.3PlaningModel Thestabilityofthesupercavitatingvehicledoesnotdepen dsimplyonthehydrodynamiccoefcientsasinthecaseofafullywettedmissile,bu tratheronthemomentsasso-

PAGE 50

39 ciatedwiththenoseandtheaft.Thecontactoftheaftofthev ehiclewiththecavitywall givesrisetotheplaningforce.Thisforce,dependingonthe strengthofthetail-moment itgenerates,maysendthebodyintotail-slap.Whensufcien trestoringnosemoments existthebodymayceaseoscillatingbutstillhaveapartoft hehullprotrudingoutsidethe cavity.Inthiscasepartoftheliftrequiredbythevehiclet omaintainit'sightisprovided bytheplaningforce.Thepathsanunpoweredprojectilemayt akeforvarioussuchplaning congurationshasbeengiveninMay[ 4 ].Theliftneededtosupporttheaft-sideofthe bodymaybegeneratedbyeitherthens,theplaningforceora combinationoftheboth. InthepresentworktheHassan'splaningforcemodelhasbeen used.Figure 3.1 showsthe vehicleundergoingtail-slap.Theplaningforce( F P )andmoment( M P )areshownactingat transomoftheplaningsection. Figure3.1VehicleUndergoingTail-Slap. 3.4PlaningForceEquations Wagner'splaningtheory[ 12 ]assumesplaningandcontinuousimmersiontobeidentical.Thetheoryismainlydevelopedfor2-dimensionalbodie sundergoingplaning.Using potentialow,theimmersionforceonaatwedgeiscalculat edassumingthattheforceis duetouidtransportintoaspraysheet.Theforceperunit-l engthofthecylinderiscomputedandthenintegratedovertheentirewettedarea.Thefo llowingaretheequationsthat characterizethissituationasgivenbyWagner[ 12 ].

PAGE 51

40 R r h Cavity wall BodyWall Figure3.2Cross-SectionofCavity Assume h =( R r ) tan 2 q 2 ) q = 2tan 1 p h D h = h r D = R r (3.15) Intheseequations h isthedepthofthepenetrationofthebodyoutsidecavityatt ransom., R istheradiusofthecavityattransomand r isthevehiclebodyradiusattransom.Figure 3.2 denesthesevariables.Theforceisgivenby P = m ¨ h + h dm dt 1 + h 1 + 2 h (3.16) where dm dt = rp R 2 1 + cos 2 q 2 sin 2 q 2 = rp R 2 ( 1 D D + h 2 ) (3.17) Letting dm dt = m h h t (3.18) Wecanwrite P = m ¨ h + h 2 m h 1 + h 1 + 2 h (3.19)

PAGE 52

41 Theterm m ¨ h isusuallyneglected.Ifweassume h ( x )= h 0 x tan ( a ) l = h 0 tan ( a ) (3.20) m s =Zl 0 m h j h = h 0 r tan ( a ) dx = rp R 2 tan ( a ) 1 D D + h 0 2 (3.21) Itcanbeconcludedthattheplaningforceisgivenby F p = h 2 m s r + h 0 r + 2 h 0 (3.22) Hassan[ 13 ],basedontheWagner'stheory,hasgiventhefollowingseto fplaningforce, momentandskin-frictionequations.Theterm F p denotestheplaningforce, M p theplaning momentand F f theplaningskin-frictionforce. F p = 1 2 r w V 2 p R 2 sin a cos a 1 R r h 0 + R r 2 # r + h 0 r + 2 h 0 M p = 1 2 r w V 2 p R 2 cos a cos a h 20 h 0 + R r r + h 0 r + 2 h 0 (3.23) Dummyvariables u c u s and S w areintroducedintotheskin-frictionequation.These variablesaredenedbythefollowingequations u c = q h 0 R r u s = 2 r p ( R r ) h 0 S w = 4 r ( R r ) tan a ( 1 + u c 2 ) ATan ( u c ) u c + r 3 2 ( R r ) tan a u s 2 0 : 5 ASin ( u s )+ 0 : 5 u s p ( 1 u s 2 ) (3.24) Thenalformoftheskin-frictionequationcanbewrittenas F f = 0 : 5 r V 2 cos a cos a S w Cd (3.25) NotethatHassan'sequationfor F p canbederivedfromtheWagner'sequationfor F p in Equation(6.8)byusingthevalue h = Vsin a

PAGE 53

42 3.5PlaningKinematics 3.5.1CalculationofImmersion Themainassumptioninvolvedinthecalculationoftheimmer sionisthatthecrosssectionofthecavityiscircular.Strictlyspeakingthecav itycross-sectionwouldundergo deformationduetogravityandothereffects.Itmaybecomee llipticalinshape[ 14 ].Using thisassumptionwecan,bysimplegeometriccalculations, ndtheimmersiondepth, h 0 ,at transomasfollows h 0 = r R + R t (3.26) InEquation 3.26 R t isthedistancebetweenthecentersofthecircularcross-se ctionsofthe cavitywallandthebody.Becausethecross-sectionsarecirc ular,alinepassingthrough centerofcavityandbodyalsopassesthroughthepointofdee pestimmersion. 3.5.2MethodofCalculationof a Thissectiondescribesthecalculationoftheangle a thatappearsintheplaningtheory. Itisimportantnottoconfusethis a withthenumerousplacesthat a representstheangleof attack.Figure 3.3 showstheangle a astheangleofinclinationbetweencavitycenterline andbodycenterline. 3 1 a Rq V0 Body CavityWall h0 Figure3.3PlaningSection Denethefollowingterms, V t :changeincavitatorvelocityovertheinterval [ t t ; t ] expressedindynamics frame B .Subscript't'denotestransomorplaningsection. V :Velocityofcavitatoratcurrenttime.

PAGE 54

43 V c :Velocityofcavitatoratthetimethecavityattheplanings ectionformed. R c :Locationofcavitycenterlineinthedynamicreferencefra me. V c = V t ( t )+ V ( t ) ˆ V c = V c k V c k ˆ V c isthetangenttothecavitycenterlineattheplaningpoint. Itisalsothenormaltothe cross-section.Notethatcalculatingthesetypesofdeect ionsimpliesthatcross-sections areelliptical;thiswillbeignoredforthetimebeing. Theangle a thatappearsintheplaningtheoryistheanglebetweenthebo dycenterline andthecavitycenterline.Thetheoryassumesthatcavityce nterlineandcavitysurfaceare parallel.Thisisnotthecaseifthecavityradiusischangin g.Theangle a iscomputed byndingthetransformationmatrixthatrotatestheveloci tyvector ˆ V c sothatitistheunit normal ˆ i = f 1 ; 0 ; 0 g T intheframe B .Thistransformationmatrixrotatesavectorinthe dynamicsframe B tothecavitynormalframe.Thedeningequationis: T 1 ˆ V c = 8>>>><>>>>: 100 9>>>>=>>>>; Therotationanglesfor T 1 canbefoundusing 8>>>><>>>>: ˆ V c x ˆ V c y ˆ V c z 9>>>>=>>>>; = T 1 1 8>>>><>>>>: 100 9>>>>=>>>>; = 8>>>><>>>>: cos ( q ) cos ( y ) cos ( q ) sin ( y ) sin ( q ) 9>>>>=>>>>; T 1 = T 3 ; 2 ; 1 | {z } denes rotation sequence ( y ; q ; f )

PAGE 55

44 Weknowthattworeferenceframesinspacecanberelatedusin gtheEulerangles.The aboveequationusestheEulerangles y q and f ,inthatorder,torelatetheframes B andthe cavitynormalframe.Theseanglesshouldnotbeconfusedwit htheEuleranglesusedin orientationequationsforthevehicleinotherpartsofthis thesis. Itfollowsthat: 1 : ) q = sin 1 V c z 2 : ) y = sin 1 ˆ V c y cos q If R c denotesthevectorthatdenesthelocationofthecavitycen terlinerelativetotheframe B ,then R ? = T 1 ( R c ) isthelocationoftheframe B relativetothecavity-normalframe. y R^ z Body CavityWall f Figure3.4TheAngle f tan ( f )= R l y R l z (3.27) Thetransformationmatrix T 2 = T 3 ; 2 ; 1 ( y ; q ; f ) willtransformavectorfromtheframe B toaframewith ˆ i perpendiculartothecavitycross-sectionwith ˆ k pointedtowardthebody centerline.Forcircularcross-sections,thispointalsop assesthroughthepointofdeepest immersion.

PAGE 56

45 Introduceanotherrotationtoaccountfortherotationofth ecavitysurfaceduetoexpansionofthecavity sin ( q x )= R x k V c k Letting T 3 = T 3 ; 2 ; 1 ( 0 ; q x ; 0 )= 266664 cos ( q x ) 0 sin ( q x ) 010 sin ( q x ) 0 cos ( q x ) 377775 (3.28) and T = T 3 T 2 we'vegotourtransformationfromdynamicscoordinatestot heframein whichtheplaningtheoryisdened.Theangle a appearingintheFigure 3.3 isobtainedfrom cos ( a )= < T ˆ i ; ˆ i > ) cos ( a )= T 11 (3.29) Thisistheanglebetweenthevehiclecenterlineandthecavi tysurfaceattheplaningpoint Figure3.5VelocityofTransomRelativetoFrame B Weneedtoincludetheeffectofthevelocityofthetransomon theangle a V n = w R p (3.30) Intheaboveequation, V n isthevelocityoftransomrelativetotheframe B R p isthe positionvectorofthetransominthe B givenby L p ; 0 ; 0 T ,where L p isthepositionof

PAGE 57

46 transomonthe 1 axis.Inordertoincludetheeffectofthevelocityoftranso m, V n ,onthe angle a V n isprojectedontothevector ˆ R ? .Thentheangle q R ,asshowninFigure 3.3 ,is computedbythefollowingequation sin q = < V n ; ˆ R ? > V (3.31) Intheaboveequation V denotesthevelocityofthevehicle.So,theanglealphaused in theplaningforceequationcanbenallywrittendownas a = a + q (3.32) Figure3.6Calculationofangleofplaneinclination a withassumptionthatthecavitysurfaceisplanar. Asimplewaytocalculateareasonablyaccuratevalueof a duringcomputationsisby theEquation(6.19). tan ( a )= L h 0 (3.33) Here,weareassumingthecavitywalltobeplanar.Figure 3.6 showstheangleofplane inclination, a ,withtheplanarassumption.

PAGE 58

CHAPTER4 LINEARIZATION 4.1Linearization 4.1.1NeedforLinearization Theequationsofmotionforthetorpedoareidenticaltoairp laneequationsofmotion buttheforcestermsontheright-handsideoftheseequation saredifferent.Thus,thelinearizationcanbecarriedoutsimilarly,asshowninNelson[ 15 ].Theequationsofmotion, asinthecaseofasupercavitatingtorpedo,arerepresented byasetofrst-orderdifferential equations. x = f ( x ; u ) (4.1) using f : n n asanonlinearfunctionofatime-varyingvector x 2 n and u 2 n Forcontroldesign,thesystemdynamicsareobservedatsome trimconditionsbygiving perturbationstostatesofthesystematthattrim.Thedynam icsassociatedwiththese perturbationsareobtainedbylinearization. Anadvantagebylinearizationisthatmostofthecontrolmet hodologyisbasedonlinear equationsofmotion.Acontrollerisdesignedinitiallyfor thelinearsystemandthentested fortheactualnonlinearsystem.Yet,therearefewdisadvan tagesforthisprocess Linearizedequationscanpredictthesystemperformanceon lyinasmallrangeofoperations.Thelinearizedequationschangeastheoperating pointofsystemchanges, thusmakingitdifcultforsimulatingtruebehaviorofsyst em. Informationrelatingtononlinearitieslikehysteresis,b acklash,coulombfriction,discontinuitiesetc.maybelostbylinearizingthesystem. Acontrollerthatisgoodforlinearizedsystemmighthaveve rypoorperformancefor thenonlinearequations.Aniterativeprocessmaybeneeded tondacontrollerthat isgoodfornonlinearequations. 47

PAGE 59

48 4.1.2GenericFormofEquationsofMotion TheequationsofmotioninEquations 2.67 and 2.70 canbewritteninsimpliedform usingsumsoftotalforcesandmomentsactingonthebody. m ( u + qw vr + gS Q )= X m ( v + ru pw gC Q S F )= Y m ( w + pv qu gC Q C F )= Z (4.2) I x p + qr ( I z I y )= L (4.3) I y q + rp ( I x I z )= M (4.4) I z r + pq ( I y I x )= N (4.5) 8>>>><>>>>: Y Q F 9>>>>=>>>>; = 266664 0 S F C Q C F C Q 0 C F S F 1 S F S Q C Q C F S Q C Q 377775 8>>>><>>>>: p q r 9>>>>=>>>>; (4.6) 8>>>><>>>>: x y z 9>>>>=>>>>; E = 266664 C Q C Y C Q S Y S Q C Y S F S Q C F S Y S F S Q S Y + C Y C F S F C Q C F S Q C Y + S F S Y C F S Q S Y S F C Y C F C Q 377775 8>>>><>>>>: u v w 9>>>>=>>>>; (4.7) Theseequationsofmotionsarecoupledbythestatevector, x ,andaredependentonthe controlvector, u x = f u ; v ; w ; p ; q ; r ; Y ; Q ; F ; x ; y ; z g u = f q R 1 ; q R 2 ; q E 1 ; q E 2 ; d R 1 ; d R 2 ; d E 1 ; d E 2 ; d c ; F prop g (4.8) 4.1.3SmallDisturbanceTheory Thesmalldisturbancetheorywillbeusedforlinearization ofequationsofmotion. Accordingtothetheorythelinearizationwillbecarriedab outanoperatingpoint(reference ightcondition),i.e.,theequationsthusderivedwillbev alidforthetorpedooperatingat andnearthatvalueofvector x .Theoperatingpointischosentocorrespondtothetrim conditioninEquation 4.9

PAGE 60

49 x 0 = f u 0 ; v 0 ; w 0 ; p 0 ; q 0 ; r 0 ; Y 0 ; Q 0 ; F 0 ; x 0 ; y 0 ; z 0 g = f u 0 ; 0 ; 0 ; 0 ; 0 ; 0 ; 0 ; Q 0 ; 0 ; 0 ; 0 ; 0 g (4.9) Thiscorrespondstostraightandlevelightwithconstantv elocity.Asthetorpedomaybe travelingatotherightconditions,thelinearizationatt hoseconditionswouldbecarriedout numerically,whichwillbeexplainedinlatersections.Ava lueof u 0 isfoundnumerically, thatsatisestheequationsofmotionforagivenvalueof x 0 .Thenadisturbanceof D x isgiventotheequationsofmotionfromthereferencecondit ionthuschangingtheight conditionsto x 0 + D x .Severalassumptionsaremadetocarryoutthelinearizatio n: Thedisturbancesfromreferenceightconditionaresmall. Thusthetermsinvolvinghigherorderofdisturbances D willbeneglected.Furthermorerstorderterms involving D willbeapproximatedasinEquation 4.10 Sin ( D )=( D ) Cos ( D )= 1 (4.10) Thepropulsiveforcesandmassareassumedtobeconstant. Planingandimmersionforcesareneglectedforthisanalysi s.However,thefollowing analysisisjustanexplanationofthemethod.Eventually,t helinearizationiscarried outnumerically.Thus,inthenumericalmethodtheplaningf orcesareincluded, unlessspecied. Thelinearizationprocedureispresolvedfortheforceequa tionin ˆ b 1 direction.This equationrelatestheforce, X ,tothestates. m ( u + qw ru + gS Q )= X (4.11) UsingtheightconditionfromEquation 4.9 inEquation 4.11 givesthevalueofforceat thereferencetrimcondition. mgS Q 0 = X 0 (4.12) Theperturbationequation,i.e.,theequationwithightco nditiondisturbedby D x canbe obtainedbysubstitutingtheperturbedightconditionint oEquation 4.11 m [ d dt ( u 0 + D u )+( q 0 + D q )( w 0 + D w ) ( r 0 + D r )( u 0 + D u ) + gS ( Q 0 + DQ )]= X 0 + D X (4.13)

PAGE 61

50 Equations 4.12 and 4.13 canbecombinedtogivethelinearizedformofEquation 4.11 for straightandlevelightcondition. m ( D u + g DQ C Q 0 )= D X (4.14) Proceedinginasimilarwayallotherequationsofmotioncan belinearized.ThelinearizedequationsforstraightlevelightareshowninEqua tion 4.15 toEquation 4.18 4.1.3.1Forceequations m ( D u + g DQ C Q 0 )= D X m ( D v + u 0 D r g DF C Q 0 )= D Y m ( D u u 0 D q + g DQ S Q 0 )= D Z (4.15) 4.1.3.2Momentequations I x D p = D L I y D q = D M I z D r = D N (4.16) 4.1.3.3Orientationequations D Y = D r C Q 0 D Q = D q D F = D p + T Q 0 D r (4.17) 4.1.3.4Positionequations D x = S Q 0 u 0 DQ + C Q 0 D u + S Q 0 D w D y = D v D z = C Q 0 u 0 DQ S Q 0 D u + C Q 0 D w (4.18) 4.1.4StabilityandControlDerivatives Thevariationsintotalforceandmomentareoftendifcultt ocompute.Thesevariations inforcescanbecalculatedbychainruleforderivatives.As statedinEquation 4.8 ,theseare functionsofstatevariables x andcontrolvariables u .Thusforexample D X canbewritten bychainruleasinEquation 4.19

PAGE 62

51 D X = X u D u + X v D v + X w D w + X p D p + X q D q + X r D r + X Y DY + X Q DQ + X F DF + X prop D F Prop + X q R 1 q R 1 + X q R 2 q R 2 + X q E 1 q E 1 + X q E 2 q E 2 + X d R 1 d R 1 + X d R 2 d R 2 + X d E 1 d E 1 + X d E 2 d E 2 + X d c d c (4.19) wherethesubscripted X representsitspartialderivativewithrespecttoitssubsc ript. X u = X u x = x0 (4.20) Eachofthesesubscriptedvariablesthathaveasubscriptof statevariableareknownas stabilityderivativesandtheoneswithacontrolvariablea ssubscriptareknownasacontrol derivative.Therecanbeasmanystabilityderivativesasth ereareforcesandstateand controlvariables.Manyofthesearenegligible,depending onthereferenceightcondition. Thesedependenciesareknownusuallybyexperimentationor numericalcalculations.For example,theforce, X ,isobservedtodependmainlyonasubsetofthestateandcont rol variable.Thusonlythestabilityderivativesthatcorresp ondtothesevariableshavetobe retainedinEquation 4.19 ,whenstraightandlevelightisconsidered. X = funct ( u ; w ; d E 1 ; d E 2 ; q E 1 ; q E 2 ; d c ; F prop ) (4.21) Thenextproblemiscalculatingnumericalvaluesofthesede rivatives.Inmostcasesit iseasytocalculatethesenumericallyorusingasymbolicma nipulationsoftware.Forsome referencepoints,itispossibletodothecalculationmanua lly.Thecalculationof X u willbe donemanuallyforstraightandlevelight. X u = u ( F R 1 ; x + F R 2 ; x + F E 1 ; x + F E 2 ; x + F c ; x + F prop ; x ) (4.22) ExpressionsforeachofthetermsinEquation 4.22 havebeenderivedinChapter 2 .For example,theexpressionfortheforceoncavitatorisshowni nEquation 4.23 F c ; x = C d c 0 S d c 266664 C b c C a c C a c S b c S a c S b c C b c 0 C b c S a c S a c S b c C a c 377775 8>>>><>>>>: D c ( a c ; g 1 2 ) 0 L c ( a c ; g 1 2 ) 9>>>>=>>>>; C (4.23)

PAGE 63

52 InEquation 4.23 a c b c ,andthus L c and D c aretheonlyfunctionsof u .Thusthepartial derivativeswithrespectto u canbeobtained. u F c ; x = C d c 0 S d c 266664 S b c C a c b c u C b c S a c a c u S a c S b c a c u + C a c C b c b c u C a c a c u C b c b c u S b c b c u 0 S b c S a c b c u + C b c C a c a c u S a c C b c b c u + C a c S b c a c u S a c a c u 377775 8>>>><>>>>: D c ( a c ; g 1 2 ) 0 L c ( a c ; g 1 2 ) 9>>>>=>>>>; C + 266664 C d c 0 S d c 010 S d c 0 C d c 377775 266664 C b c C a c C a c S b c S a c S b c C b c 0 C b c S a c S a c S b c C a c 377775 8>>>><>>>>: u D c ( a c ; g 1 2 ) 0 u L c ( a c ; g 1 2 ) 9>>>>=>>>>; C (4.24) Itcanbeseenthat a c u b c u L c u and D c u aretermsrequiredtobecalculated.Thesecanbe calculatedfromequations 2.16 and 2.17 ,whicharerestatedinEquations 4.25 toEquation 4.27 tan ( a c )= w c u c (4.25) tan ( b c )= v c V c (4.26) V 2 c = u 2c + v 2c + w 2c (4.27) ThevelocitycomponentsinEquation 4.27 canbefoundusingEquation 2.2

PAGE 64

53 8>>>><>>>>: u c v c w c 9>>>>=>>>>; C = 266664 C d c 0 S d c 010 S d c 0 C d c 377775 8>>>><>>>>: u c v c w c 9>>>>=>>>>; B (4.28) 8>>>><>>>>: u c v c w c 9>>>>=>>>>; B = 8>>>><>>>>: u v w 9>>>>=>>>>; B + ˆ b 1 ˆ b 2 ˆ b 3 pqr x c y c z c (4.29) Nowthevelocitycomponentscanbeobtainedforthereferenc eightconditionthatis statedinEquation 4.9 8>>>><>>>>: u c v c w c 9>>>>=>>>>; C = 8>>>><>>>>: C d c u 0 0 S d c u 0 9>>>>=>>>>; (4.30) Thevariationof a c canbeobtainedbydifferentiatingEquation 4.25 atreferenceight condition. sec 2 ( a c ) d a c = u c dw c w c du c u 2c ) d a c = u c dw c w c du c u 2c + w 2c (4.31) d a c = C d c dw c u 0 S d c du c u 0 at ( x 0 ; u 0 ) (4.32) Similarlythevariationof b c canbeobtainedbydifferentiatingEquation 4.26 atreference ightcondition. d b c = ( V c dv c v c dV c ) V c p V 2 c + v 2c = dv c u 0 at ( x 0 ; u 0 ) (4.33) Now,usingEquations 4.28 and 4.29 ,variationofvelocitycomponentsofcavitatorwith respectto u canbeobtained. u c u = C d c v c u = 0 w c u = S d c (4.34)

PAGE 65

54 Finally,combiningEquations 4.32 to 4.34 ,thevariationsof a c and b c withrespectto u can beobtainedatreferenceightcondition. a c u = C d c u 0 w c u S d c u 0 u c u = C d c S d c u 0 S d c C d c u 0 0 (4.35) b c u = 1 u 0 v c u 0 (4.36) Clearly,twoofthederivativesthatarerequiredtocalculat estabilityderivativeshavebeen obtained.Itwaspreviouslystatedthatliftanddragarecal culatedusingthecoefcientof liftanddrag. L c = 1 2 r V 2 c S c cl c D c = 1 2 r V 2 c S c cd c (4.37) Theseforcescanbedifferentiatedbychainrulethederivat ivewouldbelikeinEquation 4.38 L c u = 1 2 r S c 2 V c cl c V c u + V 2 c cl c a c a c u (4.38) TheEquation 4.38 requirestwoderivatives.Oneofthederivativesiscalcula tedinEquation 4.35 .TheotherderivativecanbecalculatedusingEquation 4.27 V c u = u q u 2c + v 2c + w 2c = u c u c u + v c v c u + w c w c u p u 2c + v 2c + w 2c = 1 (4.39)

PAGE 66

55 Thusthederivativeoftheliftanddragforcescanbeobtaine d. L c u = r S c V c cl c (4.40) D c u = r S c V c cd c (4.41) Thusallthederivativesrequiredtocalculateright-hands ideofEquation 4.24 havebeen calculated.Allthetermsonright-handsideoftheEquation 4.20 canbecalculatedin asimilarmanner.Itistedioustocalculatethederivatives analyticallyinsuchaway.The complexityincreasesforotherightconditions.Forpract icalpurposesthesederivativesare calculatedusingsymbolicmanipulationsoftwareslike Mathematica orbyusingnumerical methods.Thenumericalmethodsusedtocalculatethederiva tiveshavebeendescribedin Appendix B .The Mathematica codeforlinearizationisdescribedinChapterAppendix C 4.2StateSpaceRepresentation Equations 4.15 to 4.18 areacompletesetoflinearizedequationsofmotionforthe torpedo.Theycanberepresentedinamoreconvenientformkn ownastheStateSpace Form.Thestatespaceequationsareasetofrst-orderdiffe rentialequations. x = Ax + Bu y = Cx + Du x 2 n ; u 2 p ; y 2 m A 2 n n ; B 2 n p C 2 m n ; D 2 m p (4.42) Equation 4.42 isageneralizedformofstatespacerepresentationforanys ystem.Eachof thetermsintheequationshasaparticularimportanceforde scribingthedynamicsofthe system. StateVariablex: Thestatevariablesforasystemareasetofvariables,whenk nown attime t 0 andalongwithinput u ,aresufcienttodeterminethestateofthesystem atanytime t > t 0 .Allthestatesofthesystemneednotbemeasurable.

PAGE 67

56 InputVariableu: Thisisthecontrolsurfacedeections. OutputVariabley: Theoutputvariablesarethemeasuredparameters.Thesemay ormaynotbesameasthestatevariables.Theoutputvariable sareusuallyconsidered tobemeasurablebutsometimestheyareestimated. Thematrices A ; B ; C and D mayeitherbeconstantortime-varyingfunctions. Inthecaseofthesupercavitatingtorpedo,thestatevector isofsize12(n)andthe controlvectorisofsize10(p). x = D u D w D q DQD v D p D r DFDYD x D y D z u = d c d E 1 d E 2 d R 1 d R 2 q E 1 q E 2 q R 1 q R 2 D F prop (4.43) Someofthesecontrolsmaynotbeneededforsomemaneuvers.F romthelinearizedequationsitcanbeobservedthatthestatevariablesarenotcoup ledbythestates f Y ; x ; y ; z g Thesefourstatescanberemovedfromtheanalysisfornow.Th esystembecomesa8state system.Thesestatescanbefurtherdividedintolongitudin alandlateral-directionaldynamics.Thevariables D u ; D w ; D q ; DQ correspondtolongitudinaldynamics,whichalsomeans thedynamicsin ˆ b 1 ˆ b 3 plane.Thevariables D v ; D p ; D r ; DF correspondtolateraldynamics, whichisthedynamicsin ˆ b 1 ˆ b 2 plane.Sometimesthelateralandlongitudinalequationsca n bedecoupled.Thusifthetorpedoismakingapullclimb/desc enttoacertaindepth,usually itsdynamicscanberepresentedbylongitudinalstatevaria bles.Theplantmatrix A canbe dividedintofourparts. A = 264 A long A coup 1 A coup 2 A latd 375 (4.44) When A isdividedasinequation 4.44 ,whereeachelementisa4 4matrix, A long wouldcorrespondtolongitudinaldynamicsand A latd wouldcorrespondtolateraldynamics. A coup 1 and A coup 2 arecouplingmatrices.Itisrequiredthatthecouplingmatr ices becomenegligiblefortheequationstobedecoupled.Ifthes epartsarenotnegligible, theequationscannotbedecoupled,anda8statemodelwillbe requiredtobeconsidered.

PAGE 68

57 Fromlinearizedequations,thefourpartsofthe A matrixforthetorpedocanbewrittenas inEquation 4.45 toEquation 4.48 A long = 266666664 X u m q 0 + X w m w 0 + X q m gC Q 0 + X Q m q 0 + Z u m Z w m u 0 + Z q m gC F 0 S Q 0 + Z Q m M u I y M w I y M q I y M Q I y 00 C F 0 0 377777775 (4.45) A latd = 266666664 Y v m Y p m u 0 + Y r m gC Q 0 C F 0 + Y F m L v I x L p I x L r ( I z I y ) q 0 I x L F I x N v I z N p ( I y I x ) q 0 I z N r I z N F I z 01 C F 0 T Q 0 q 0 C F 0 S Q 0 r 0 S F 0 S q 0 C Q 0 377777775 (4.46) A coup 1 = 266666664 r 0 + X v m X p m v 0 + X r m X F m p 0 + Z v m v 0 + Z p m Z r m gS F 0 C Q 0 + Z F m M v I y M p ( I x I z ) r 0 I y M r ( I x I z ) p 0 I y M F I y 00 S F 0 S F 0 q 0 C F 0 r 0 377777775 (4.47) A coup 2 = 266666664 r 0 + Y u m p 0 + Y w m Y q m gS Q 0 S F 0 + Y Q m L u I x L w I x L q ( I z I y ) r 0 I x L Q I x N u I z N w I z N q ( I y I x ) p 0 I z N Q I z 00 S F 0 T Q 0 S F 0 q 0 + C F 0 r 0 377777775 (4.48) Similarly B isa8 10matrix,whoseelementsarejustthecontrolderivativesa ccordingto theirlocationsinthematrix.Therst4rowscorrespondtol ongitudinaldynamicsandthe last4correspondtolateraldynamics. B long = 266664 X d c m X F prop ; x m ... . ... 0 0 377775 4 10 (4.49)

PAGE 69

58 B latd = 266664 Y d c m Y F prop ; x m ... . ... 0 0 377775 4 10 (4.50) Nowthecompletestatespacerepresentationforthetorpedo canbewrittenasinEquation 4.51 whichgivestwosetsofequations.Therstsetisthelongitu dinalequationsandthe secondsetisthelateral-directionalequations. x long = D u D w D q DQ T x latd = D v D p D r DF T u = d c d E 1 d E 2 d R 1 d R 2 q E 1 q E 2 q R 1 q R 2 D F prop 8><>: x long x latd 9>=>; 8 1 = 264 A long A coup 1 A coup 2 A latd 375 8 8 8><>: x long x latd 9>=>; 8 1 + 264 B long B latd 375 8 10 u 10 1 (4.51)

PAGE 70

CHAPTER5 CONTROLDESIGNSETUP Thischapterdealswiththecontroldesignforthetorpedode scribedinpreviouschapters.Variousparametersassociatedwiththecontrolarere statedinTable 5.1 Table5.1ControlParameters Longitudinal Lateral State u w q Q v p r F Y Control d c d e 1 d e 2 d r 1 d r 2 Itshouldbenotedthat Y hasbeenincludedinthestatesthoughitwasobservedin thestatematricesthatallothervariablesareindependent of Y .Itwillbeseenlaterthat theinclusionof Y inthefeedbackstateshelpsinimprovementofperformance. Also, forlongitudinalcontrol,twoelevatorsandthecavitatora rerequired.Similarlyforlateral direction,theruddersshouldbeenoughforcontrol.Adevia tionfromtheserequirements willbeobservedinsomeofthecontrollers,mainlytoimprov eperformanceandprovide stability. Therearevariouscontrolmethods,likelinearquadraticre gulator(LQR)synthesis, synthesisetc.,whichcanbeusedtodesignacontroller.Eac hofthesemethodshasadvantagesanddisadvantages.LQRmethodgivesaconstantgainco ntrollerwhichisbasedon minimizationofaquadraticperformanceindexandconsider stheproblemofrobustness onlyintermsofgainandphasemargins. -synthesisdealswithrobustnesswithrespectto awidevarietyofuncertaintiestominimizeaninnity-norm matrixbuttheresultingcontrollercanbeofhighorder.Regardlessofcomplexityandrob ustness,eachdesignmethod presentsdifculties.Thischapterexplainsvariousprobl emsassociatedwiththecontrol synthesisandthesystemmodelusedforsynthesisofthecont roller. 59

PAGE 71

60 5.1Open-LoopPerformancefortheFixedCavityModel Initiallytheequationsofmotionforthetorpedohavebeenl inearizedforstraightand levelightataforwardvelocityof75 ms 1 x = f 75 ; 0 ; 0 ; 0 ; 0 ; 0 ; 0 ; 0 ; 0 ; 0 ; 0 ; 0 g (5.1) Itisfoundthatthecavitatorandtwoelevatorsaresufcien ttomaintaintheabovetrim. Itisalsoassumedthatthevalueofpropulsiveforcerequire dtomaintainthistrimisxed attherequiredvalue. u = f d c ; d e 1 ; d e 2 ; d r 1 ; d r 2 ; F prop g = f 0 : 0067 ; 0 : 0106 ; 0 : 0106 ; 0 ; 0 ; 4 : 0023 e + 03 g (5.2) wherethedeectionsaregiveninradians,and F prop isinNewtons.Astheseparameters areobtainednumerically,itmaynotbeaperfecttrim.Thesy stemmayhavesomenonzeroaccelerations,andconsequentlymaytendtodeviatefr omstraightandlevelight.To checkthis,theopen-loopdynamicsaresimulatedatthiscon ditionwithoutanyfeedback. Figure 5.1 showstheSimulinkmodelusedforopen-loopsimulation.The controlsurface deectionsarexedattheirtrimvaluesforthissimulation .Theclosed-loopsystemis obtainedusingtheequationsofmotionthatwerederivedinCh apter 2 .Thestatederivatives arethenintegratedtoobtainthestateatthenextinstant. Figure 5.2 showstheopen-loopresponseforthetorpedoatthistrimcon dition.Itcan beseenthattheopen-loopsystemisunstable.Thesimulatio niscarriedoutatthetrimthat isshowninEquation 5.1 ,i.e.,allthevaluesshowninthesegureshavetobezero.Th e systemisseentohaveoscillationaboutnon-zerostates. Thestateandcontrolmatricesobtainedforthistrimcondit ionareshowninEquations 5.3 to 5.6 .Thereare5controlvariables, f d c ; d e 1 ; d e 2 ; d r 1 ; d r 2 g .Thematricescorresponding tothelateraldynamicsareofdimension5becausethestate Y isincludedinthelateral dynamics.Thusthelateralstatesarenow f v ; p ; r ; F ; Y g

PAGE 72

61 State Feedback In1 state For Plotting simt Time MATLABFunction NL Equation Torpedo 1 s Integrator Out Control at Trim Clock Figure5.1SimulinkModelforOpenLoopSimulation 0 20 40 60 80 100 0 5 10 15 20 time(s)w (ms -1 ) 0 20 40 60 80 100 -0.04 -0.03 -0.02 -0.01 0 0.01 time(s)p (rad s -1 ) 0 20 40 60 80 100 0 0.5 1 1.5 2 2.5 3 3.5 4 time(s)q (rad s-1) Figure5.2Open-LoopResponseforTorpedo: w ; p ; q

PAGE 73

62 A long = 266666664 4 : 52041 : 54171 : 3110 9 : 8100 0 : 2616 15 : 764878 : 58880 0 : 00001 : 2077 3 : 56140 001 : 00000 377777775 (5.3) B long = 266666664 32 : 301069 : 0608 69 : 060800 406 : 0942 303 : 3736303 : 373600 158 : 4675 45 : 153145 : 153100 00000 377777775 (5.4) A latd = 266666666664 12 : 0422 0 : 0002 71 : 60119 : 81000 0 : 1813 54 : 22810 : 300400 1 : 1437 0 : 0025 1 : 252800 01 : 0000000 001 : 000000 377777777775 (5.5) B latd = 266666666664 000 366 : 60511366 : 60511 0 14297 : 086 14297 : 086 17276 : 994 17276 : 994 0 1 : 4129523 1 : 412952354 : 561629 54 : 561629 0000000000 377777777775 (5.6) Thelongitudinaleigenvaluesare f 21 : 1414 ; 4 : 5137 ; 1 : 8262 ; 0 : 0178 g andthelateraleigenvaluesare f 0 ; 54 : 2289 ; 0 : 0002 ; 6 : 6472 + 7 : 2683 ; 6 : 6472 7 : 2683 g .The eigenvaluesclearlyshowthatthesystemisunstable.Itcan alsobeseenthatthelongitudinaldynamicshavenooscillatorymodes.Figure 5.3 showsthevariationofeigenvalues forthetorpedofordifferentvelocities.Statevaluesare xedexceptforforwardvelocity, whichisvariedfrom60ms 1 to90ms 1 .Theguresshowthatthevariationissmalland mostoftheeigenvaluesstayinnegativehalfofcomplexplan e.

PAGE 74

63 -30 -20 -10 0 10 -1 -0.5 0 0.5 1 Longitudinal Egienvalues for u=60:5:90 Real ( l long )Imag ( l long ) (a)Longitudinal -80 -60 -40 -20 0 20 -10 -5 0 5 10 Lateral Egienvalues for u=60:5:90 Real ( l latd )Imag ( l latd ) (b)Lateral Figure5.3VariationofEigenvalueswithChangeinVelocity 5.2Closed-LoopProblem Asstatedearlierthecontrolproblemcanbesubdividedinto variousproblems.Each canbesolvedtogetanalcontroller.Theultimategoalofth econtrollerdesignisto achieveadesiredtrajectoryorreachaparticularpointwit hminimizationofsomeperformancecriteria.Theachievementofthisgoalrequiresaddre ssingmaneuvering,trimming, guidanceandnavigation.Thisthesiswillconsiderthebasi cproblemofmaneuvering.So theproblemistobeabletotrackacertainpitchandrollcomm andwhilemaintainingcertainperformancecriteria.Theperformancecriteriathatt hecontrollerisrequiredtomeet are: Trackastepcommandinpitchorrollrateofsizeupto30 deg = s Maintainanovershootlessthan15%. Havearisetimeoflessthan0.6sec. Haveasteadystateerroroflessthan5%. Besidesmeetingtheabovementionedperformancecriteria,t hecontrollerisalsorequired tostabilizetheclosed-loopsystem. Variousboundsareplacedonthecontrolsurfacedeections andrates.Thesebounds arelistedinTable 5.2 .Theboundsareincludedinthenonlinearsimulationsandit is requiredthatthereisnosaturationofdeectionortherate satleastforthecommands havingtherate30 deg = s

PAGE 75

64 Table5.2ControlConstraints CavitatorDeection 15 d c + 15 CavitatorRate 25 = s d c + 25 = s ns 60 d f + 60 FinRate 25 = s d f + 25 = s Thrust 0 F prop 30000 N Anactuatormodelisincludedinthesystemsoastoconstrain theratesofthecontrol surfacemotion.Thisactuatorisrealizedasalowpasslter A c = 80 s + 80 .Additionofthis ltersynthesizesacontrollerthathasslowercontroldee ctions. Let q comm ( t ) beafunctionoftime,deningthedesiredpitchrateprole. Theaimof thecontrolleristondacontrollaw u ( t ) thatyieldsanachievedpitchrate, q achi ( t ) ,to minimizetheoptimizationproblemstatedinEquation 5.7 nd u ( t ) thatminimizes z ( t )= j q achi ( t ) q comm ( t ) j subjectto u min u u max u min u u max x = Ax + Bu (5.7) where, u min and u max arelowerandupperboundsoncontroldeections.Thequanti ties u min and u max arelowerandupperboundsoncontroldeectionrates. Theproblembecomesadisturbancerejectionproblem,whent hecommandedvariable is0foralltime.Thisisanoptimizationproblem,wherethes tatespaceequationisan equalityconstraintandthecontrolsurfaceboundsareineq ualityconstraints. 5.3RobustnessoftheController Acontrolsystemthatisdesignedtoaccommodatetheuncerta intiesinamathematical modeliscalledarobustcontrolsystem.Usuallytherespons eofamodeldoesnotaccurately matchtheresponseofthetruesystem.Arobustcontrolsyste mshouldgivethedesired

PAGE 76

65 performancenotonlyduringthesimulationsofthemodel,bu tforthetruesystemalso. Variousparameterscanbeintroducedinthemodeltosimulat euncertainties.Random noisecanbeaddedtooutputsignaltosimulatemeasuremente rrors,thegainsinsignals canbechangedtomodeluncertaintyinresponseetc.Gainand phasemarginsaregenerally usedtopredicttherobustnessofacontrolsystem.Physical ly,thesearejustthefactors bywhichthefeedbackgaincanbeincreasedandyethaveastab lerealsystem.Aformal denitionofthesecanbegivenbyusingafrequencyanalysis forafeedbacksystem. 5.3.1Gainmargin Figure 5.4 showsatypicalfeedbacksysteminvolvingaplant,P,andaco ntroller,K. Thegainforthesystemindottedregionisknownastheloopga in.Theloopgainis importantbecauseitdeterminesstability.Errorsinthepr edictedloopgaincouldcause errorsinpredictedstability.Thegainmarginisthelarges tfactorbywhichthisgaincanbe increasedandstillhaveastablesystem.Physically,itmea nsiftheresponseofthetorpedo foragivenelevatorinputishigherbyafactorofthegainmar gin,thetorpedoisstillstable. Thegainmarginisusuallyexpressedindecibel(db)units,a ndcanbeeasilyobtainedfrom theBodeplotsforthesystem.Thegainmarginisameasurement ofthemagnitudeonthe Bodeplot,atthepointwherethephaseis180 o 5.3.2Phasemargin Gainisavalidrobustnesscriteriawhenthesystemhasreale igenvalues.Butusually theeigenvalueshaveimaginarycomponentsandthusthephas eisalsoaconcern.Phase marginisthemeasureofthemaximumpossiblephaselagbefor ethesystembecomes unstable.FromtheBodeplot,phasemarginisthephasewhenth emagnitudeofthegainis zero.5.3.3Uncertaintyinparameters Anotherfactorthatcandeterminetherobustnessofacontro llerisitsresponsetoerrors inknownparameters.Asstatedearlier,thecoefcientsofl iftanddragarecalculatedfroma CFDdatabase.Theaccuracyofthemodeldependsonaccuracyof thisdata.Robustnessof

PAGE 77

66 + KP Figure5.4LoopGain acontrollercanbeassessedbyintroducingerrorsinthedat aandcheckinghowitperforms. Thefollowingvariationshavebeenintroducedinthesystem tocheckforperformanceof thesystemwithintrinsicuncertainties: 20%errorin C l ofCavitator. 20%errorin C d ofCavitator. 20%errorin C l ofalltheFins. 20%errorin C d ofalltheFins. 5.3.4Controllerobjective Intermsofrobustness,thecontrollerisrequiredtomeetva riousperformanceobjectives.Theseobjectivecanbesummarizedas: Theclosed-loopsystemshouldhaveagainmarginofatleast6 dB. Theclosed-loopsystemshouldhaveaphasemarginsofatleas t45 deg 5.3.5 analysis: FortheH / analysis[ 16 ]itisdesiredthatthepeak valuefortheclosed-loopsystem becloseto1.Thisensuresgoodrobustnessatleastequaltot hevaluesofuncertaintiesin thesynthesismodel.

PAGE 78

CHAPTER6 LQRCONTROL 6.1LQRTheory Thetrackingproblem,liketheonegiveninEquation 5.7 ,canbesolvedbyusinga combinationoffeedbackandfeedforwardcontrol[ 17 ].TheLinearQuadraticRegulator (LQR)problemistondanoptimalfeedbackmatrix K suchthatthestate-feedbacklaw u = Kx minimizesthelinearquadraticcostfunctionshowninEquat ion 6.1 J ( u )= Z0 ( x T Q x + u T R u + 2 x T N u ) dt (6.1) ThebasicideaofLQRcontrolistobringthestateofthesyste mclosetozero.Alinear systemcanberepresentedinthestatespaceformasinEquati ons 6.2 and 6.3 .Thematrices A and B arethestateandcontrolmatrices.Thevariable x representsthestatevector, y is theoutputvectorand u istheinputvector. x = A x + B u (6.2) y = x (6.3) TheLQRcontrollerisrealizedbyaconstantgainmatrix K ,suchthatthefeedback u = Kx makes x gotozero.Byamodicationtothislaw,theLQRmethodcanalso be usedfortracking.Thestatevector x isofsize n x = f x 1 ; x 2 ; ; x n g T (6.4) Letthetrackingproblembeforthestate x 1 totrackastepcommand r 1 .Theideaisto make ( x 1 r 1 ) gotozerousingaLQRcontroller.Thenewcontrollawcanbech osenasin 67

PAGE 79

68 Equation 6.5 u = K 0BBBBBBB@ x 1 r 1 x 2 ... x n 1CCCCCCCA (6.5) Equation 6.2 canberewrittenbysubstitutingthenewcontrollaw. x = A x + B u = A x B K 0BBBBBBB@ x 1 r 1 x 2 ... x n 1CCCCCCCA (6.6) Forsimplicity,assumethatthereisonlyonecontrol, u (thisisdifferentfromvelocity u ). Thecontroller K isofsize n 1anditcanbeexpandedinitselements. K =[ k 1 ; k 2 ; ; k n ] (6.7) Equation 6.6 canberewrittenbysubstitutingthe K initsexpandedform. x = A x B [ k 1 ; k 2 ; ; k n ] 0BBBBBBB@ x 1 r 1 x 2 ... x n 1CCCCCCCA = A x B Kx + Bk 1 r 1 =( A B K ) x + Bk 1 r 1 (6.8) Itshouldbenotedthatthecommand r 1 isastepcommand.Thesteady-statedynamicsof thesystemcanbeobtainedfromEquation 6.8 x ( )=( A B K ) x ( )+ Bk 1 r 1 (6.9)

PAGE 80

69 TheerrordynamicscanbeobtainedbysubtractionEquation 6.9 fromEquation 6.8 x ( t ) x ( )=( A B K ) x ( t ) x ( ) (6.10) e =( A B K ) e (6.11) where e =( x ( t ) x ( )) .So,thetrackingproblemiscastasaregulatorproblem.The new statevectoristhesteady-stateerror e ,whichismadezerousingtheregulator.Figure 6.1 showstheblockdiagramforthisclosed-loopsystem.Itisre quiredthattheclosed-loop systemhasanintegratorsomewheresoastomakethesteady-s tateerrorgoto0[ 17 ].That is, e hastogotozeroratherthan e soastoachievegoodtracking.Figure 6.2 showsthenew congurationofasystemthathasnointegratorandthusanin tegratorhastobeenincluded duringdesign.Thus,theintegraloftheactualerrorhastob emadetogotozerosoasto achieveagoodtracking. e =Z( r 1 x 1 ) (6.12) Thestatespaceequationforthesystemwiththismodicatio ncanbewritten. x = A x + Bu e = r 1 x 1 = r 1 C x (6.13) where x 1 = C x .Itcanbeseenthattheerrorequationissimilartostateequ ation.Thus e canbeconsideredasanotherstate,.i.e,thesystemnowhas n + 1stateswithstatevector x = f x 1 ; x 2 ; ; x n ; e g T .Soanewformulationofthestatespaceequationcanbewritt en, x = 264 A 0 C 0 375 x + 264 B 0 375 u + 264 01 375 r 1 (6.14) ) x = A x + Bu + Ir (6.15) whichissimilartoEquation 6.8 .Theerrordynamicsofthissystemrepresenttheformof statespaceequations,forwhichaLQRcontrollercanbederi ved.TheLQRcontroller K willbeaconstantmatrixofsize n + 1asthesystemnowisofsize n + 1.

PAGE 81

70 K =[ k 1 ; k 2 ; ; k n k k n + 1 ] (6.16) Then,thenewcontrollawcanbewrittenasinEquation 6.17 u = K x = [ k 1 ; k 2 ; ; k n k k n + 1 ] 264 x e 375 = [ k 1 ; k 2 ; ; k n ] x +[ k n + 1 ] e = Kx + k I e (6.17) whichisrepresentedinFigure 6.2 + k 1 K r 1 x = Ax + B h y = xx Figure6.1ControllerforTrackingwhenPlanthasanIntegrat or. + + k I K x = Ax + B u y = xRr 1 x C Figure6.2ControllerforTrackingwhenPlanthasnoIntegrat or.

PAGE 82

71 6.2LQRControlforFixedCavityModel: 6.2.1ControlSynthesis Thetorpedosystemdoesnothaveanintegratorinthesystem. Atrackingcontrollercan beobtainedfromLQRmethodbytheprocessdescribedinSecti on 6.1 .Thecontrolleris obtainedforatrimstateofstraightandlevelightat75ms 1 .Thelinearizeddynamicsare rstseparatedintothelongitudinalandlateraldynamicsa sgiveninTable 5.1 .Thecontrols usedinlongitudinaldirectionarethecavitatorand2eleva tors.Thecontrolsinlateral directionarethe2rudders.Itisobservedthatforstraight andlevelightthelongitudinal andlateraldynamicsarepracticallydecoupled. Oncethestateandcontrolmatriceshavebeenobtained,them ainvariablesthattheLQR controllerdependsonaretheweightingmatrices Q R and N .Inthiscasethecrosscoupling matrix N ischosentobe0.Thematrices Q and R penalizethecostfunctionforhigherstate andcontrolvaluesrespectively.Ahighervaluein Q matrixwouldcauseabettertracking.Alarger R wouldconstrainthecontrolsurfacedeection.Anoptimumc ombination ofthematricesisobtainediteratively,soastogetgoodtra ckingwithachievablecontrol deections. Thematricesforlongitudinalpitchratetrackingaregiven inEquation 6.18 Q long = diag ([ 0 ; 0 ; 0 ; 0 ; 10 ]) R long = diag ([ 5 ; 4 ]) (6.18) Therstfournumbersinthe Q long correspondtoweightingsonthefourlongitudinalstates. Theyarechosentobe0.Wedonotwanttorestrictthestatesfr omchanging.Thisis especiallyimportantforweightingson q and Q .Aweightingonthesevariableswould restrictthemfromchanging.Thelastnumber,10,isweighti ngontheerror.Thisischosen tobehightopenalizethetrackingerror.Ahighervalueofwe ightingwouldgiveabetter tracking,butitisseenthatitwouldrequireveryhighcontr olrates.Therstnumberin theweightingmatrix R long ,5,correspondstocavitatorweightingandthenumber4isfo r

PAGE 83

72 elevatorweighting.Elevatorweightingischosentobesmal lersoastoencouragethe controllertousemoreelevatorthanthecavitator.Thisgiv esamorestableperformance. Thecontrolmatricesobtainedforthelongitudinaldynamic saregiveninEquations 6.19 and 6.20 k I = 264 1 : 1182 0 : 9681 375 (6.19) K = 264 0 : 00000 : 00400 : 1016 0 : 00001 : 4195 1 : 1184 0 : 0000 0 : 0042 0 : 0995 0 : 0000 1 : 39811 : 1308 375 (6.20) Similarprocessisinvolvedinthedesignofthelateralcont roller.Initiallythelateral controllerisdesignedwithonlyfourstatefeedback,and Y isneglectedinthefeedback.In thiscaseitisfoundthatthetorpedohashighsidewashandde viatesconsiderablyfromthe originalpath,evenwhenthepitchangleis0.Toavoidthis, Y isincludedinthefeedback states.Itisalsoobservedthatapenaltyontheyawmotionca usesthecontrollertocommandaveryfastcontrolsurfacemotion.Also,acontinuousc orrectionofcontrolsurface deectionisrequiredtopreventtheyawmotionentirely.Th usanoptimumcombinationof theweightingmatricesisobtainedthatwouldpreventavery highyawmotionbutwould stillhaveslowcontrolsurfacemotion. Q latd = diag ([ 0 ; 0 ; 0 ; 0 ; 0 ;: 1 ]) R latd = diag ([ 1000 ; 1000 ]) (6.21) Therst5numberscorrespondto5statesandthelastnumberi sweightingfortheerror. The R latd isofdimension2asonlytheruddersareincludedinthesynth esis.Theweighting ontheruddersishighasitisobservedthattherollrateisve rysensitivetotherudder deection.Thecontrolmatricesobtainedforthelateraldy namicsaregivenintheEquations 6.22 and 6.23

PAGE 84

73 k I = 264 0 : 0071 0 : 0071 375 (6.22) K = 10 3 264 0 : 0005 0 : 1253 0 : 01320 : 0019 0 : 0000 0 : 0005 0 : 1254 0 : 01320 : 0026 0 : 0000 375 (6.23) Thefeedbackmatrix K forlateraldynamicsisofsize2 5,whichisshowninEquation 6.23 6.2.2NominalClosed-loopModel6.2.2.1Model Figure 6.3 showstheeigenvaluesfortheclosed-looplongitudinaland lateralsystems. Itcanbeseenthatbothsystemsarestableasalltheeigenval uesareinthelefthalfof thecomplexplane.Also,eachofthedynamicshasoneeigenva lueattheorigin,whichis introducedduetheintegratorinthesystem. -1000 -800 -600 -400 -200 0 200 -15 -10 -5 0 5 10 15 Real ( l long )Imag ( l long ) (a)Longitudinal -1000 -800 -600 -400 -200 0 200 -8 -6 -4 -2 0 2 4 6 8 Real ( l latd )Imag ( l latd ) (b)Lateral Figure6.3EigenvaluesfortheClosed-LoopSystem 6.2.2.2Simulations Theresponseofthevehicletoalongitudinalcommandissimu latedandshowninFigures 6.4 to 6.6 .Theseguresshowtheresponseforapitchratedoubletof15 deg = s .The risetimeforthepitchratecommandof15 deg = s is0.18sandthereisanovershootof 11.53%.Thesteady-stateerroris.8%.Thecontrollerisabl etocommandpitchratesas highas30 deg = s .Itisobservedthatthevehiclemotionisconnedtolongitu dinalplane

PAGE 85

74 0 5 10 15 -20 -10 0 10 20 time(s)q (deg s-1) Figure6.4PitchCommandTracking: q 0 5 10 15 -1 -0.5 0 0.5 1 time(s)dc (deg) 0 5 10 15 -30 -20 -10 0 10 20 time(s)rate dc (deg s-1) Figure6.5PitchCommandTracking: d c ; d c only.Thisshowsthatthecontrollerallowspurelongitudin almotiontobeuncoupledfrom thelateralmotion. 0 5 10 15 0 0.5 1 1.5 2 time(s)d e1 (deg) 0 5 10 15 -15 -10 -5 0 5 10 15 20 time(s)rate de1 (deg s-1) Figure6.6PitchCommandTracking: d e 1 ; d e 1

PAGE 86

75 Theresponseofthevehicletoalateral,rollrate,commandi sshowninFigures 6.7 to 6.9 .Arollratecommandof15 deg = s isachievedin.52swithanovershootof0%and asteady-stateerrorof0.09%.Thecontrollerisabletocomm andarollratemotionofas highas50 deg = s beforeasaturationofcontrolsurfacerateisreached.Itis observedthat thereissomelongitudinalmotioninthiscase.Thislongitu dinalmotionhasbeenreduced byinclusionof Y inthefeedbackstatestothecontroller.Itcanbeseenthatt herudder deectionforarollratecommandissmall.Thisisexpecteda sthetermscorrespondingto rollratefromrudderareanorderof3timeslargerthanthete rmscorrespondingtopitch ratefromelevators.Itisassumedthatthecontrolsurfaced eectionisachievable. 0 5 10 15 -20 -10 0 10 20 time(s)p (deg s-1) Figure6.7RollCommandTracking: p 0 5 10 15 0 0.2 0.4 0.6 0.8 1 time(s)dc (deg) 0 5 10 15 -0.5 0 0.5 1 1.5 2 2.5 3 time(s)rate dc (deg s-1) Figure6.8RollCommandTracking: d c ; d c

PAGE 87

76 0 5 10 15 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 time(s)d e1 (deg) 0 5 10 15 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 time(s)rate de1 (deg s-1) Figure6.9RollCommandTracking: d e 1 ; d e 1 +1 3 2+ K x r a k 1Rr 1y = x x = Ax + B u Figure6.10BreakpointsforCalculatingtheLoop-GainforaTr ackingController 6.2.2.3Gainandphasemargins TheLQRtrackingsystemshowninFigure 6.2 isobviouslymorecomplexthanthe systemshowninFigure 5.4 .Thus,theloopgaincanbedenedinmanywaysinthiscase. Theblockdiagramcanbebrokenatdifferentpointssoastosi mplifyittotheformshown inFigure 5.4 .Figure 6.2 isredrawninFigure 6.10 whichshowsthepossiblebreakpoints forthissystem.Forunderstanding,theoutputofplantPisd ividedintotwoparts,oneisthe achievedvalueofthecommandedvariable( r a )andtheotherisremainingstatesoftheplant P( x ).Thebreakpointsarenumbered1to3.Thesystemcanbebroke nateachofthese pointstogivealoopgain.Thesegainswillbenamedouter-lo op,inner-loopandall-loop gainsrespectively. GainandPhasemarginsforeachoftheabovepossiblebreakpo intshavebeencalculatedforboththelongitudinalandlateralcontrollers.Ta ble 6.1 liststhegainandphase

PAGE 88

77 Table6.1GainandPhaseMarginwithLQRController Longitudinal GainMargin(db) PhaseMargin(deg) 1 21.056(at47.498rad/s) 64.846(at9.0625rad/s) 2 327.87(at0rad/s) 77.118(at25.925rad/s) 3 57.606(at20.845rad/s) Lateral GainMargin(db) PhaseMargin(deg) 1 22.964(at0rad/s) 2 250.51(at0rad/s) 3 50.36(at0rad/s) marginsforthetorpedowithLQRcontrollerthatwasobtaine dinprevioussections.All marginsarequitehighandmeetthedesiredconditionsof6dB forgainand45 deg forphase margin. Also,thelateralcontrollerisunabletostabilizetheunst ablespiralmode.Thusthe closed-loopsystemisinherentlyunstableduetothispolea ndwouldconsequentlyhave negativegainmargin.Numeroussimulationsshowthattheaf fectofspiralmodeisnegligible, i.e.,thetimetodoublefortheinstabilityisconsiderably largerthanthemaneuveringtime ofthetorpedo.So,theclosed-loopsystemmodelisreducedb yremovingthespiralmode fromthemodel.ThegainandphasemarginsinTable 6.1 areforthisreduced-ordersystem andreecttherobustnessofthedominantdynamics.6.2.3PerturbedClosed-loopModel Aperturbedsystemmodelisformedbyaddinganerrortotheva luesofcoefcients ofliftanddragforthensandcavitator.Newvaluesoftrimd eectionareobtainedfor theperturbedmodelandthusanewsetof A and B matricesisobtained.Tables 6.2.3 and 6.3 showthepercentagevariationoftheelementsof A and B matricesfora20%change incoefcientsofliftcavitator.Thiscomparisonisdonefo rcaseswithchangesinother coefcientsalso.Inallcases,fewelementsinthestateand controlmatriceschange.In mostcases,thechangeinelementsof A and B matricesisalinearfunctionofthechange

PAGE 89

78 Table6.2PercentageVariationinAMatrixdueto20%Variati onin cl c u w q Q v p r F Y u 0.46 0.62 w 5.52 6.86 1.58 q 1.05e5 34.8 13.58 Q v p r F Y inacoefcient.Forexample,inTable 6.2.3 thereare8termsthatshowavariationdue toa20%variationincoefcientofliftofthecavitator.The termA(3,1)showsalarge variationbutitsnumericalvalueisnegligible.ThetermA( 3,2)showsa34%variation butthistermisalsosmallcomparedtootherterms.Remaining termsinthematrixshow verysmallvariation.Sometermsinthe B matrixshowa20%variation.Thussometerms incontrollabilitymatrixchangeconsiderably.Thiswould meanthatforanerrorinthese coefcients,theresponsewouldshowsomedifferenceincon trolsurfacedeection.Asit isobservedthattheclosed-loopsystemhasgoodgainandpha semargins,thiseffecton B matrixshouldnotbeofmuchconcern.6.2.3.1Model Figure 6.11 showstheeigenvaluesfortheperturbedclosed-looplongit udinalandlateral systems.Anerrorof-20%isincludedinthevalueofcoefcie ntofliftforthens.Itcan beseenthatthelongitudinaldynamicsshowsomeperturbati oninthedampingwhilethe lateralsystemrelativelyunchanged.6.2.3.2Simulations Theperformanceofthecontrollersisstudiedusingthesimu lationwithaperturbed systemmodel.Anerrorisassumedinthevaluesofvariouscoe fcientsandacorrection

PAGE 90

79 Table6.3PercentageVariationinBMatrixdueto10%Variati onin cl c d c d e 1 d e 2 d r 1 d r 2 u w 20 q 20 Q v p r F Y -1000 -800 -600 -400 -200 0 -15 -10 -5 0 5 10 15 RealImag ( l long ) (a)Longitudinal -1000 -800 -600 -400 -200 0 200 -8 -6 -4 -2 0 2 4 6 8 RealImag ( l latd ) (b)Lateral Figure6.11EigenvaluesforthePerturbedClosed-LoopSyste m:20%Errorin cl fin factorisadded.Responseoftheclosed-loopnonlinearsyste misnotmuchaffectedbythe variationsincoefcientsofliftanddrag.Itisobservedth atthecontrollercommandsthe systemtoanewtrimstatewhichisalsoastraightandleveli ght,withchangeinspeed andcontroldeections.Afterthat,thesystemfollowsapit chorrollcommandaswell asbefore.Figures 6.12 to 6.13 showtheresponseforonesuchcase.Inthiscasearoll doubletiscommandedtothesystem,andthereisanerrorof 20%inthevalueof cl fin Itcanbeclearlyseenthatthevehiclehasgonetoanothertri mstateandthenitfollowsthe commandequallywell.Thereisalmostnochangeinthetrajec toryofthevehicle.The

PAGE 91

80 0 10 20 30 -6 -4 -2 0 2 4 6 p (deg s -1 )time(s) true+20% error-20% error 0 10 20 30 -8 -6 -4 -2 0 2 4 6 q (deg s-1)time(s) true+20% error-20% error Figure6.12Responsefor20%Variationin cl fin : p ; q 0 10 20 30 -0.5 0 0.5 1 d c (deg)time(s) true+20% error-20% error 0 10 20 30 0.4 0.6 0.8 1 1.2 de1 (deg)time(s) true+20% error-20% error Figure6.13Responsefor20%Variationin cl fin : d c ; d e 1 controlsurfacedeectionsaresimilarwithaconstantoffs et.Suchresponsehasalsobeen checkedforothercases.Theaffectoferrorissimilarinall cases. 6.2.3.3Gainandphasemargins Table 6.4 liststhegainandphasemarginsfortheperturbedclosed-lo opsystem.The perturbedsystemalsohasgoodgainandphasemargins.Compar ingthevalueswithTable 6.1 ,itcanbeseenthattherearesmallchangesinthevaluesexce ptforthelateralall-loop. Thelastvalueisincreasedto showinganimprovementfortheperturbedsystem. Fromtheanalysisoftheperturbedclosed-loopsystemitcan besaidthatthelinear modelisrobusttovariousuncertaintiesinthesystem.

PAGE 92

81 Table6.4GainandPhaseMarginforPerturbedClosed-loopSys tem:20%errorin cl fin Longitudinal GainMargin(db) PhaseMargin(deg) 1 21.193(at49.599rad/s) 68.981(at8.7966rad/s) 2 320.33(at0rad/s) 77.605(at25.925rad/s) 3 60.305(at22.552rad/s) Lateral GainMargin(db) PhaseMargin(deg) 1 24.391(at0rad/s) 2 278.23(at0rad/s) 3 (at0rad/s)

PAGE 93

CHAPTER7 /H SYNTHESISCONTROL 7.1Uncertainty Themainparameters(besidesthecontrol)thatdictatethef orcesonthethetorpedo arethecoefcientoflift, cl ,anddrag, cd ,ofthecavitatorandns.Thevaluesofthese parametersareobtainedusingaCFDcodein[ 5 ].Thesevalueshavealsobeenobtained theoreticallyandexperimentallyin[ 4 ].AcomparisonoftheCFDandpredictedresults isshowninthegures 7.1 .Itcanbeseenthatthereisadifferenceofatleast1orderof magnitude,inthevalueof cd ,betweenthetwodatasets.Also,asupercavitatingowisa2 phaseowwithpartialwaterandpartialvapor.Thusthehydr odynamicsforsupercavitating vehiclesareuniqueandneedmoreinvestigation.Itisclear thatthereisaneedforinclusion ofuncertaintyforthevaluesof cl and cd ,whicharemainparametersthatdeterminethe hydrodynamics. Acontrollerdesignedbasedon orH theoryisarobustcontroller,whichdealswith errorsanduncertaintiesinthesystem,implicitly.Basicid eaof synthesisistoreduce thegainfromerrorordisturbancetotheerrorintracking.I nthedesignof controller fortorpedo,anuncertaintyisassumedinthecoefcientsof lift, cl c ,andthecoefcientof drag, cd c ,forthecavitatoronly.Thefollowingformulationforthes ynthesismodelhas beenderivedindetailforthelongitudinalplant.Theformu lationforlateralplantisvery similar,differencebeingonlythestatesandcontrols.Als o,thefollowingformulationis forapitchangletrackingcontroller.Theformulationfort rackinganyotherstate,saypitch rate q ,wouldbesimilar.Thefeedbacktothecontrollerwill,inth atcase,bethetracking errorin q 82

PAGE 94

83 Figure7.1CalculationofUncertainty 7.2SynthesisModel Thestatespaceformulationforlongitudinaldynamicscanb egivenas x = Ax + B u y = Cx (7.1) where x = f D u ; D w ; D q ; DQ g u = f d c ; d e g y = x C = I 4 A =[ a ij ] 4 4 B =[ b ij ] 4 2 (7.2) where, I 4 representsanidentitymatrixofsize4 4.Forsimplicity,the B matrixischosen tobe4 2,i.e.,othercontrolsareassumedtobexedattheirtrimva lues.Thustheonly controlsaredeectionofcavitatorandthedeectionofthe elevators.Theelevatorsare

PAGE 95

84 assumedtohaveasymmetricdeectionforalongitudinaltra ckingcommand.Thusonly oneelevatorcanbeincludedinthesynthesismodel.Notetha tthesymbol u isusedto representthecontrolvector. u willalsobebeusedtorepresenttheforwardvelocityofthe torpedo. Asdescribedearlier,parametricuncertaintyisassumedin thecoefcientsoflift( cl c cl e ) anddrag( cd c cd e )ofthecavitatorandtheelevators.Let W 1 and W 2 representpercentage errorsin cl e and cd e respectively.Let W 3 and W 4 representpercentageerrorsin cl c and cd c respectively.Thusavalueof W 1 = 0 : 1wouldmeanaupto10%errorinvalueof cl e andso on.Now,thetruevalueofthesecoefcientscanbewrittenas cl e = cl e ( 1 + d cle W 1 ) cd e = cd e ( 1 + d cde W 2 ) cl c = cl c ( 1 + d clc W 3 ) cd c = cd c ( 1 + d cdc W 4 ) (7.3) where, cl e cd e cl c and cd c arethevaluesofthesecoefcientsfromthedatabase,i.e., their nominalvalues. d cl and d cd arevariables,whosenumericalvaluedeterminestheactual errorinthesecoefcients.Thus 1 d cle ; d cde 1 1 d clc ; d cdc 1 (7.4) Elementsof A and B matriceswillbefunctionsoffunctionofcoefcientsoflif tanddragof cavitator,andsomeothertermswhichareoflesserimportan ceforthisanalysis.Intheory, theparametricformcanbewrittenasexplicitfunctionof cl and cd i.e. A = A 0 + cl e A 1 + cd e A 2 + cl c A 3 + cd c A 4 B = B 0 + cl e B 1 + cd e B 2 + cl c B 3 + cd c B 4 (7.5) Thisformulationof A and B matricescanbetoughtoobtainforgeneralightcondition, butitcanbefoundnumericallyMATLAB,asdescribedintheApp endix B .Asymbolic

PAGE 96

85 derivationofthesematricesisdoneusingMATHEMATICA,buti tisfoundthattheterms areverylongandtheresultsrunsthroughpages.Thedescrip tionofvariouscodeshasbeen giveninAppendix C .Thenumericalvaluesfromboththecodesmatch,thusleadin gtoa vericationofthenumericalmethod.Now,usingequation 7.3 in 7.5 ,wehave A = A 0 + cl e ( 1 + d cle W 1 ) A 1 + cd e ( 1 + d cde W 2 ) A 2 + cl c ( 1 + d clc W 3 ) A 3 + cd c ( 1 + d cdc W 4 ) A 4 B = B 0 + cl e ( 1 + d cle W 1 ) B 1 + cd e ( 1 + d cde W 2 ) B 2 + cl c ( 1 + d clc W 3 ) B 3 + cd c ( 1 + d cdc W 4 ) B 4 (7.6) or A = A 0 + d cle A 1 + d cde A 2 + d clc A 3 + d cdc A 4 B = B 0 + d cle B 1 + d cde B 2 + d clc B 3 + d cdc B 4 (7.7) where A 0 = A 0 + cl e A 1 + cd e A 2 + cl c A 3 + cd c A 4 A 1 = cl e W 1 A 1 A 2 = cd e W 2 A 2 A 3 = cl c W 3 A 3 A 4 = cd c W 4 A 4 (7.8) and B 0 = B 0 + cl e B 1 + cd e B 2 + cl c B 3 + cd c B 4 B 1 = cl e W 1 B 1 B 2 = cd e W 2 B 2 B 3 = cl c W 3 B 3 B 4 = cd c W 4 B 4 (7.9) Withthistransformationthestatespaceformshowninequat ion 7.1 canbewrittenas x =( A 0 + d cle A 1 + d cde A 2 + d clc A 3 + d cdc A 4 ) x +( B 0 + d cle B 1 + d cde B 2 + d clc B 3 + d cdc B 4 ) u (7.10)

PAGE 97

86 Inaboveequation,eachofthetermscorrespondingtouncert ainty,arelikeinputstothe system.Forexample, d cl A 1 x islikeaninput,andisofsize4 1.Thistermcanbereplace byaninputvector w 1 ofsamesize.Alsothetermmultiplyingtheuncertaintycanb ewritten asanoutputofthesystem.Thusletsmakeafurthertransform ationasfollows z 1 = 266666664 z 11 z 12 z 13 z 14 377777775 = A 1 x w 1 = 266666664 w 11 w 12 w 13 w 14 377777775 = 266666664 d cle 000 0 d cle 00 00 d cle 0 000 d cle 377777775 z 1 (7.11) similarly z 2 =[ z 2 i ] T = B 1 uw 2 =[ w 2 i ] T = Diag [ d cle ] 4 4 z 2 z 3 =[ z 3 i ] T = A 2 xw 3 =[ w 3 i ] T = Diag [ d cde ] 4 4 z 3 z 4 =[ z 4 i ] T = B 2 uw 4 =[ w 4 i ] T = Diag [ d cde ] 4 4 z 4 z 5 =[ z 5 i ] T = A 3 xw 5 =[ w 5 i ] T = Diag [ d clc ] 4 4 z 5 z 6 =[ z 6 i ] T = B 3 uw 6 =[ w 6 i ] T = Diag [ d clc ] 4 4 z 6 z 7 =[ z 7 i ] T = A 4 xw 7 =[ w 7 i ] T = Diag [ d cdc ] 4 4 z 7 z 8 =[ z 8 i ] T = B 4 uw 8 =[ w 8 i ] T = Diag [ d cdc ] 4 4 z 8 (7.12) for i =1to4,and Diag [] representsadiagonalmatrixwiththediagonalelementsgiv enby thosein [] andremainingelementsaszeros.Usingabovetransformatio ninequation 7.10 wegetanewstatespaceformas x = P 11 x + P 12 u y = P 21 x + P 22 u (7.13)

PAGE 98

87 where u = w 1 w 2 w 3 w 4 w 5 w 6 w 7 w 8 u T34 1 y = z 1 z 2 z 3 z 4 z 5 z 6 z 7 z 8 y T36 1 (7.14) and P 11 = A 0 4 4 (7.15) p 12 = I 4 I 4 I 4 I 4 I 4 I 4 I 4 I 4 B 4 34 (7.16) p 21 = 26666666666666666666666664 A 1 0 4 4 A 2 0 4 4 A 3 0 4 4 A 4 0 4 4 I 4 37777777777777777777777775 (7.17) p 22 = 26666666666666666666666664 0 4 32 0 4 2 0 4 32 B 1 0 4 32 0 4 2 0 4 32 B 2 0 4 32 0 4 2 0 4 32 B 3 0 4 32 0 4 2 0 4 32 B 4 0 4 32 0 4 2 37777777777777777777777775 (7.18)

PAGE 99

88 P z 1 ; z 2 z 3 ; z 4 z 5 ; z 6 z 7 ; z 8 w 5 ; w 6 w 3 ; w 4 w 1 ; w 2 w 7 ; w 8 2664 I 8 d cl e 000 0 I 8 d cd e 00 00 I 8 d cl c 0 000 I 8 d cd c 3775 x u Figure7.2LinearFractionalRepresentation ThusThenewsystemis34inputand36outputsystem.Thisrepr esentationforthe systemisknowasLinearFractalRepresentation,andcanbesh ownbyablockdiagramas showningure 7.2 Withthisformulations,a synthesiscontrollercanbefoundasdescribedin[ 16 ].Figure 7.3 showsthesynthesismodelforthesystemshowninequation 7.13 .Toaccountfor feedbacksignaluncertainty,amultiplicativeuncertaint yisaddedinthefeedbacksignalfor Q .The W m representsthemultiplicativeuncertainin Q ,whosevaluesis W m D m .Thus, Q actual = Q nominal ( 1 + W m D m ) (7.19) Variousltershavebeenaddedtosystemtoputconstraintso nthecontrolsurfacesand improveperformance.Theseltersarelikethe Q and R matricesforandLQRcontrol synthesis.Thesechoiceoftheseltersdonebothbytrialan derrorandalsosomeanalysis oftheirfrequencyresponse. W p :Thisischosensoastopenalizetheperformanceerror.This isusuallyalowpass lter.Thusthelterhasahighvalueatlowfrequenciesandt hevaluedropsdownat higherfrequencies.Thevalueisobtainedbytrialanderror W Kc and W Ke :Thesearethelteroncontrolcommands.Theseareusuallyh ighpass lter.Thecontrolsurfacemotionallowedistheinverseoft heselter.Thusahigher controlmotionwillbeallowedalowfrequenciesandlowcont rolmotionathigh frequencies.

PAGE 100

89 + P e P AutoPilotInput Q 0 e k f d c ; d e 1 ; d e 2 g w q Q Noise 5 1 W K W N K u Q p W P Ac f d c ; d e 1 ; d e 2 g D m W m d m z 1 ; z 2 z 3 ; z 4 z 5 ; z 6 z 7 ; z 8 w 5 ; w 6 w 3 ; w 4 w 1 ; w 2 w 7 ; w 8 2664 I 8 d cl e 000 0 I 8 d cd e 00 00 I 8 d cl c 0 000 I 8 d cd c 3775 Figure7.3SynthesisModelfor Controller W N :Thissignalistoaddsensornoiseinthesystem.Inthissynt hesis,thisischosen asaconstantvalue.Abetterapproximationofthiscanbedon ebyknowingsome sensorcharacteristics. Ac :Thislterrepresentsanactuatormodel.Thisrepresentst heactuatordynamics inthefrequencydomain. Theinputs d andoutputs e forthesynthesismodelare d = wnoisepilot e = ze p e k (7.20)

PAGE 101

90 Theaimofa orH controlleristotryandreducethegainfrominputstooutput s.The problemforH controllercanbestatedas Find K suchthatmin K jj e jj 2 jj d jj 2 (7.21) Thusavalueof1for jj d jj 2 wouldmeantominimize jj e jj 2 .Thisishowa H or problem issetup. Anadvancedapproachtosameproblemisgivenby theory.Thisisknownasthe DK-iterationmethod[ 16 ].Abovedescribedsynthesisisforalongitudinalpitchang lecontroller.Thesynthesisforothercontrollersdesignedisve rysimilar,withsomedifferences inuncertaintyandsignals.Thedifferenceswillbemention edincorrespondingsections. 7.3ControlObjectiveandConstraints Therearevariousobjectivesthatacontrollershouldmeet. Twoproblemsconsidered forthexedcavitycasearepitchandrollangletracking.Pi tchrateandrollratetracking controllersareobtainedforthecaseofdynamiccavity.The controllerisrequiredthemeet followinggoals: Anovershootoflessthan10%onnominalandperturbednonlin earmodels. Asteadystateerroroflessthan10%fornominalandperturbe dnonlinearmodels. Thecavitatordeectionissaturatedat 15 andthedeectionofnsissaturatedat 25 Theratesofcontrolsurfacesaresaturatedat 25 Itisalsodesiredtohaveacontrollerwithlownumberofstat es. 7.4 ControllerforFixedCavityModel 7.4.1Longitudinalcontroller Thelongitudinalcontrollerisdesignedtotrackapitchang lecommand.Figure 7.3 exactlyrepresentsthesynthesismodelforthecontrolsynt hesisofthecontrollerforpitch angletracking.Thelongitudinalstatesare f u ; w ; q ; Q g andthecontrolsurfacesarethe

PAGE 102

91 cavitatorand2symmetricelevators.Followingarethevalu esoftheweightingfunctions forthepitchratetrackingcontroller. W p =Thisischosentohaveahighvaluealowfrequencies.Thusth econtrolleris penalizedforanerroratlowfrequency,butnopenaltyathig hfrequencies. W p = 1 : 8 s + 0 : 1 (7.22) Thusapenaltyof18ontheperformanceerrorat0frequencyan daweightingof0 foraperformanceerrorats= W Kc :Thecontrollerarechosensoastominimizehighfrequencym otionofthe controlsurfaces.Thevalueofthiscavitatorweightingisc hosentobe W Kc = 20 s s + 100 (7.23) Thishasavalueof0ats=0and20ats= .Thusacavitatormotionof radisallowed atalowfrequencybutonly0.05(=1/20)radisallowedatahig hfrequency.Thevalue atlowfrequencyisnotaconcernasitisfoundthatthecontro llercommandssmall cavitatorusually. W Ke :Sameasthepreviousargument,theelevatorweightingisch osentobe, W Ke = 100 ( s + 0 : 1 ) s + 100 (7.24) Thishasavalueof0.1ats=0and100ats= .Thusacavitatormotionof10rad (=1/0.1)isallowedatalowfrequencybutonly0.01(=1/100) radisallowedatahigh frequency. W N :Thisisa5inputand5outputlter.Thevalueischosentobe0 .01. Forsynthesis,theuncertaintiesarechosentobe, W 1 = W 2 = 0 : 6 ) 60%parametricuncertaintyinthecoefcientsofliftanddr agof theelevator. W 3 = W 4 = 0 : 5 ) 50%parametricuncertaintyinthecoefcientsofliftanddr agof thecavitator. Wm = 0 : 1 ) 10%multiplicativeuncertaintyinthepitchangle Q Withtheaboveparameters,acontrollerisobtainedusingMA TLAB[ 16 ].Thecontroller has9statesandtheclosedloopsystemhaspeak valueof1.15.Figure 7.4 showsthetime responsefora10degpitchanglecommandwiththiscontrolle r.Itcanbeseenthatthe systemhasasmallovershootof5%fromthesteady-statevalu e.Therisetimeis1.8s andthesteadystateerroris6%.Figure 7.5 showsthecavitatorandelevatordeection

PAGE 103

92 0 5 10 15 -10 -5 0 5 10 time(s) Q (deg) commandlinearNL Figure7.4PitchAngleTrackingfortheNominalPlant 0 5 10 15 -5 0 5 time(s) d c (deg) linearNL 0 5 10 15 -20 -10 0 10 20 time(s) rate d c (deg s -1 ) linearNL 0 5 10 15 -5 0 5 time(s) d e1 (deg) linearNL 0 5 10 15 -20 -10 0 10 20 time(s) rate d e1 (deg s -1 ) linearNL Figure7.5CavitatorandElevatorDeections forabovedoubletcommand.Thecommandsasseenareinbounds forbothlinearand nonlinearmodels. Figure 7.6 showsthefrequencyresponseofthecontroller.Thecontrol lerhas5inputs and2outputs,thusthereare10linesshowninthegure.Asex pected,thecontrollerhasa highgainatlowfrequencyandthegaindropsdownathighfreq uencies. Table 7.1 showsthegainandphasemarginsforthe4feedbackstates.Th esemargins validatethefactthatthecontrollerobtainedisfairlyrob ustwithrespecttothechosen uncertainties.

PAGE 104

93 10 -2 10 0 10 2 10 -5 10 0 10 5 MagnitudeFrequency (rad/s) 10 -3 10 -2 10 -1 10 0 10 1 10 2 10 3 -400 -200 0 200 Frequency (rad/s)Phase Figure7.6FrequencyResponseofthePitchAngleController Table7.1GainandPhaseMarginsfortheLongitudinalControl ler signalGainMargindB(atrad/s)PhaseMargindeg(atrad/s) u 34.801(4.735) w 13.222(31.11)87.26(2.7536) q 68.372(32.0848) Q 27.144(25.064)75.483(3.3884) Figure 7.7 showsthecomparisonoftheresponseforthenominalplantan damodelof perturbedplant.Theperturbationintheplantisa20%incre seinthevaluesoftheliftand dragforboththecavitatorandelevator.7.4.2Lateralcontroller Thelateralcontrollerisdesignedtotrackarollanglecomm and.Theblockdiagram forthelateralcontrollerisverysimilartothelongitudin alcontrollerinFigure 7.3 .The differenceisthattheautopilotcommandisnowarollanglec ommand,thefeedbackstates are f v ; p ; r ; F g andtherollangletrackingerror.Thecontrolsurfacesfort hiscasearethe2 elevatorsandthe1DOFcavitatorisnotcapableofaffecting therollmotion.Theelevators areassumedtomoveasymmetricallyforarollcommand,thusa sbefore,onlyoneelevator

PAGE 105

94 0 5 10 15 -10 -5 0 5 10 Time (s)Q (deg) CommandNominalPerturbed 0 5 10 15 -4 -2 0 2 4 Time (s)d c (deg) NominalPerturbed 0 5 10 15 -20 -10 0 10 Time (s)Rate d c (deg/s) NominalPerturbed 0 5 10 15 -4 -2 0 2 4 Time (s)d e1 (deg) NominalPerturbed 0 5 10 15 -20 -10 0 10 Time (s)Rate d e1 (deg/s) NominalPerturbed Figure7.7ComparisonofResponsefortheNominalandPerturbe dPlants canbeincludedinthecontrolsynthesis.Thustherearefoll owinguncertaintiesinthe system, W 1 = W 2 = 0 : 2 ) 20%parametricuncertainyinthecoefcientsofliftanddra gof theelevators. W m = 0 : 1 ) 10%multiplicativeuncertaintyinthefeedbackofrollangl e F Thustheuncertaintyblockisofsize17,ascomparedtoasize 33forthelongitudinal controller.Variousdesignweightingarechosentobe, W p :Apenaltyof20ontheperformanceerrorat0frequencyandaw eightingof0 foraperformanceerrorats= W p = 2 s + 0 : 1 (7.25) W Ke :Thishasavalueof0.025ats=0and25ats= .Thusacavitatormotionof 40radisallowedatalowfrequencybutonly0.04(=1/20)radi sallowedatahigh frequency.

PAGE 106

95 0 5 10 15 -10 -5 0 5 10 Time(s) F (deg) commandlinearNL Figure7.8RollAngleTrackingfortheNominalPlant W Ke = 25 ( s + 0 : 1 ) s + 100 (7.26) W N :Thisisa5inputand5outputlter.Thevalueischosentobe0 .01. Withtheaboveparameters,acontrollerisobtainedusingMA TLAB[ 16 ].Thecontroller has8statesandtheclosedloopsystemhaspeak valueof0.72.Figure 7.8 showsthetime responsefora10degrollanglecommandwiththiscontroller .Itcanbeseenthatthesystem has0%fromthesteady-statevalue.Therisetimeis1.2sandt hesteadystateerroris3%. Figure 7.9 showsthecavitatorandelevatordeectionforabovedouble tcommand.The commandsasseenareinboundsforbothlinearandnonlinearm odels. Figure 7.10 showsthefrequencyresponseofthecontroller.Thecontrol lerhas5inputs and1output,thusthereare5linesshowninthegure.Asexpe cted,thecontrollerhasa highgainatlowfrequencyandthegaindropsdownathighfreq uencies. Table 7.2 showsthegainandphasemarginsforthe4feedbackstates.Th esemargins validatethefactthatthecontrollerobtainedisfairlyrob ustwithrespecttothechosen uncertainties. Figure 7.11 showsthecomparisonoftheresponseforthenominalplantan damodelof perturbedplant.Theperturbationintheplantisa20%incre seinthevaluesoftheliftand dragfortheelevators.

PAGE 107

96 0 5 10 15 0.2 0.3 0.4 0.5 0.6 Time (s) d c (deg) linearNL 0 5 10 15 -5 0 5 Time(s) rate d c (deg s -1 ) linearNL Time(s) 0 5 10 15 0.6 0.8 Time(s) d e1 (deg) linearNL 0 5 10 15 -5 0 5 Time(s) rate d e1 (deg s -1 ) linearNL Figure7.9CavitatorandElevatorDeections 10 -2 10 0 10 2 10 -5 10 0 10 5 Log MagnitudeFrequency (radians/sec) 10 -2 10 0 10 2 -300 -200 -100 0 100 Phase (deg)Frequency (radians/sec) Figure7.10FrequencyResponseoftheRollAngleController Table7.2GainandPhaseMarginsfortheLateralController signalGainMargindB(atrad/s)PhaseMargindeg(atrad/s) v 83.526(138.97) p 93.144(117.36) r 71.186(3.6056) F 24.417(208.13)93.3(17.45)

PAGE 108

97 0 5 10 15 -10 -5 0 5 10 Time (s)F (deg) CommandNominalPerturbed 0 5 10 15 0 0.5 1 1.5 Time (s)d c (deg) NominalPerturbed 0 5 10 15 -4 -2 0 2 4 Time (s)Rate d c (deg/s) NominalPerturbed 0 5 10 15 0.5 1 1.5 Time (s)d e1 (deg) NominalPerturbed 0 5 10 15 -2 -1 0 1 2 Time (s)Rate d e1 (deg/s) NominalPerturbed Figure7.11ComparisonofResponsefortheNominalandPerturb edPlants

PAGE 109

CHAPTER8 HOMINGCONTROL Homingcontrolofmissilesisawidelystudiedproblemtheli terature.Itisachallenging problemandnumerouspartialsolutionshavebeendocumente dintheliterature.Themain challengingaspectsofthehomingcontrolare: Insufcientmeasurementofthetargetheading,velocityet c. Non-constantmissileandtargetvelocity. Highspeedandfastmaneuveringtargets[ 33 ]. Missilemaneuvering. Measurementdelay. Estimationoftime-to-go(TTG)[ 19 ].TTGisthetimetakefromcurrenttimetothe collision. Predictionofmiss-distance[ 28 ]. Variousapproacheshavebeensuggestedtosolvethemissile homingproblem.Proportional Navigation(PN)isoneofthemostwidelyusedguidancemetho dologiesbecauseofits simplestructureandeaseofimplementation.PNisbasedont hefactthat:collisionwiththe targetcanbeguaranteediftheturnrateoflineofsight(LOS )remainszero.Inotherwords, Amethodofhomingnavigationinwhichthemissileturnratei sdirectlyproportionaltothe turnrateinspaceoftheLOS.VariousPNsolutionshavebeenp resentedin[ 21 22 31 32 ]. AdvancedPNschemeshavebeenpresentedin[ 31 25 ].Nonlinearcontrolmethodologies likefuzzycontrol[ 20 30 ],slidingmodecontrol[ 23 24 27 ]andmoderntechniqueslike neuralnetworksandH control[ 29 26 ]havebeenproposedintheory.Thesemodern methodologieshaveimplementationissuesduetocomplicat edstructureofthecontroller. Ageneralizedguidancelaw,whichcanbereducedtomanyofth eotherlaws,ispresentedin 98

PAGE 110

99 [ 34 ].Analyticalexpressionsforcapturearea,missilecomman dedaccelerationandhoming timedurationarederivedinclosedformin[ 34 ]. 8.1HomingControlusingProportionalNavigation ProportionalNavigation(PN)isoneoftheeasilyimplement ablemethodologiesfor missilehomingcontroldesign.Asdescribedearlier,PNisb asedonregulatingtherateof turnoftheLOS.PNismainlyeffectivewhenthedistancebetw eenthemissileislarge.As themissileclosesin,theerrorinmeasurementhasalargera ffect.AlthoughPNhasanadvantageofeaseofimplementation,itsuffersfrompoorpred ictionoftimetogoestimation [ 19 ]andanon-zeromissdistanceintheterminalphase.Thus,ot herlinearmethodologies willbeimplementedintheterminalphase. ThePNproblemcanbesubdividedintovarioussimplerproble ms: Constantmissileandtargetvelocity. Targetmoveswithconstantvelocitywhilethemissilemoves withconstantacceleration. Constantaccelerationforbothmissileandtarget Eachofthesimplerproblemswillbesolvedindependently.Be foretheproblemscanbe attacked,itisworthwhiletounderstandaboutthetargetan dmissilepathgeometry,collision triangle.ThecollisiontriangleisshownintheFigure 8.1 .Considerthemissileismoving Reference Line Target Missile Collision 1 3 M M S M S T T R V M V T Figure8.1CollisionTriangle withavelocity V M andthetargetismovingwithavelocity V T ,alongthedirectionsasshown

PAGE 111

100 intheFigure 8.1 .Theobjectareonacollisioncourse,ifattheirtrajectori escoincideat somepoint,atsomenaltime t f .Fromthecollisiontrianglewehave: S M sin f T = S T sin f M = R sin ( f T f M ) (8.1) where S M S T and R arethedistancesasshowninthegure. f M and f T aretheangles betweentheLOSandthemissileandtargetvelocityvectors. Equation 8.1 willbeused frequentlyinupcomingsectionsforderivationofcontroll aws. 8.2ConstantMissileandTargetVelocity Thesimplestcaseforhomingmissileguidanceiswhenthemis sileandtargetaremovingwithaconstantvelocity. S M = V M t f ; S T = V T t f (8.2) where t f isthetime-to-go,thetimetocollision.UsingthisinEquat ion 8.1 ,wegetthe relationbetweenthemissileandtargetvelocities. V T sin f T V M sin f M = 0(8.3) ThisisthevelocityrelativetotheLOS. R s = V T sin f T V M sin f M (8.4) where, s istheLOSorientationwithsomeinitialreferenceline.Thu s,ifamissileison collisioncoursewithatarget,theLOSrateofturniszero. s = 0 ; R 6 = 0(8.5) Incasetheaboveisnotsatised,thesimpleststrategyisto compensatetheLOSturnrate, byanoppositeyawofthemissile.Thisisthebasicguidancel awforPN. g M = N s (8.6) where, N istheguidancegain.Intermsofthelateralacceleration, a M = V M g M (normalto V M ),PNbecomes a M = NV M s = N 0 V c s (8.7) where V c = R = V T cos f T + V M cos f M istheclosingspeed,and N 0 = NV M = V c

PAGE 112

101 8.3ConstantVelocityTargetandAcceleratingMissile Ingeneral,themissileacceleratesduringtheboost.Theta rgetcanstillbeassumedto bemovingwithaconstantvelocity.Assumingthatthemissil ehasaconstantacceleration a M t f isthetime-to-go: V Mf = V M + a M t f S M ( t f )= V M t f + 1 2 a M t 2 f = V Mf t f 1 2 a M t 2 f V M = S M ( t f ) t f = V M + 1 2 a M t f = V Mf 1 2 a M t f (8.8) where, V Mf isthenalvelocityofthemissile, S M ( t f ) shownin 8.1 isnowdependenton theaccelerationandtime-to-go, V M istheaveragespeedofthemissile. Againusing 8.1 t f = R sin f M V T sin ( f T f M ) R sin f T S M ( t f ) sin ( f T f M )= 0 (8.9) TheequationoftheLOScanbeobtainedbytakingthecomponen tsofthemissileandtarget velocityparallelandperpendiculartotheLOS. R = V T cos f T V M cos f M R s = V T sin f T V M f M (8.10) TheLOSequationscanbeusedintheEquation 8.9 toeliminatethemissileparameters V T and f T t f = R sin f M R s cos f M R sin f M = R R s c tg f M R R s +( V M V M ) sin f M = 0 (8.11) Astheseequationsusethecollisiontriangle,theyjointly deneafunctionthatvanishes whenthemissileisonthecollisioncoursewiththetarget.T heaveragevelocity V M canbe eliminatedusingtheEquations 8.8 R s 1 2 a M t f sin f M = 0(8.12)

PAGE 113

102 FromEquation 8.11 (i), R t f = R s c tg f M R = 0.Thustheequationofcollisioncourseis writtenasinEquation 8.13 ( R s c tg f M R ) s 1 2 a M sin f M = 0(8.13) Incasethesearenotsatised,theguidancetocollisionlaw canbeformulatedtodrivethe equationto0. a M = N 0 ( R s c tg f M + V c ) s 1 2 a M sin f M (8.14) Foramissilewithconstantvelocity,theguidancelawreduc estoPN. 8.4AcceleratingTargetandMissile Nowconsideramorecomplicatedsituation,whereinboththe missileandthetarget haveconstantaxialacceleration(ordeceleration).Theeq uationsforthemissileandtarget canbewrittenintermsoftheaccelerations. V Mf = V M + a M t f ; S M = V M t f + a M t 2 f = V Mf t f a M t 2 f V Tf = V T + a T t f ; S T = V T t f + a T t 2 f = V Tf t f a T t 2 f (8.15) TheEquations 8.15 canbesubstitutedintheEquations 8.1 and 8.10 toobtaintheequations ofLOSintermsofvelocity. R sin f M t f ( V Tf 1 2 a T t f ) sin ( f T f M )= 0 R sin f T t f ( V Mf 1 2 a M t f ) sin ( f T f M )= 0 (8.16) From(i)wecanobtainthetime-to-go. t f = R sin f M ( V Tf 1 2 a T t f ) sin ( f T f M ) = V Tf a T t f V Tf 1 2 a T t f t f (8.17) where t f isthetime-to-goforthecasewhen a T = 0. t f = R sin f M R s cos f M R sin f M = R V c + R s c tg f M (8.18) Equation 8.19 ,secondorderequationintermsoftime-to-go,issolvedtoo btainthetimeto-go. 1 2 a T t 2 f ( a T t f + V Tf ) t f + V Tf t f = 0(8.19)

PAGE 114

103 Thenaltargetspeedcanbereplacedbytheinstantaneousta rgetspeedusingtheEquation 8.15 1 2 a T t 2 f + V T t f V T t f = 0(8.20) Again,usingtheEquations 8.1 and 8.10 ,weobtaintheequationalongthecollisioncourse. R s + V M V T S M S T sin f M = 0(8.21) Equation 8.15 canbesubstitutedtoobtainthecollisioncourseintermsof themissileand targetvelocities. R s + 1 2 a M V Tf a T V Mf V Tf 1 2 a T t f t f sin f M = 0(8.22) Equation 8.17 and 8.18 canbefurthersimpliedasshowninequation 8.23 R t f = R s cos f M R sin f M sin f M V Tf 1 2 a T t f V Tf a T t f (8.23) Equation 8.23 canbesubstitutedinEquation 8.21 toobtainEquation 8.24 ( V c + R s c tg f M ) s 1 2 ( a M V Tf a T V Mf )( V Tf a T t f ) ( V Tf 1 2 a T t f ) 2 sin f M = 0(8.24) Equations 8.15 canbeusedtoreplacethenalvelocitiesintermsofinstant aneousvelocities. ( V c + R s c tg f M ) s 1 2 a M V M a T V T t f t f 2 sin f M = 0 t f t f = 1 + p 1 + 2 z z ; z = a T V T t f (8.25) BasedonthecollisioncourseEquation 8.25 ,theguidancetocollisionlawcanbestatedas inEquation 8.26 a M = N 0 ( V c + R s c tg f M ) s 1 2 a M V M a T V T t f t f 2 sin f M # t f t f = 1 + p 1 + 2 z z ; z = a T V T t f (8.26)

PAGE 115

104 8.5YawController Ithasbeenobservedthatitisverydifculttodesignayawco ntrollerforthemodel ofthetorpedothatisstudiedinthisthesis.Themainreason forthisisthatthetorpedo isequippedwitha1DOFtorpedo.The1DOFcavitatormodelisv eryeffectiveinthe longitudinaldirectionbutdoesnotcontrolthelateraldyn amics.Moreover,theuseofthe4 nsdoesnotgiveagoodyawratecontroller.Thus,itisimper ativethatindirectmethods havetobeusedtoachieverequiredyawrate. Themostobviousapproachforyawratecontroller,istogive acertainrolltothemissile followedbyacertainpitch.Thesecommandsarefollowedbys imilarnegativecommand, soasthemakethevehiclestraightandlevel.Thissequenceo fcommands,causesaresidual yawinthevehicle.8.5.1YawcontrolusingtheLQRcontrollers ThecontrollersobtainedinChapter 6 areusedtoobtainthedesiredyawratecontroller. Duetoproblems,withthiscontroller,otherLQRcontroller sforyawandpitchanglehave alsobeenobtained. Varioussequenceofcommandshavebeentriedtoobtainayawr atecontrol.Theseare showninFigure 8.2 .Itisobservedthattheyawcontrollerhasvariousproblems ,whichever combinationistried.Thevariousproblemsthatwereobtain edwhileusingthesecontrollers are: Inmanycasestherewasnoresidualyawangleobtainedatthee ndofthecommand sequence.Thecontrollerbroughtthevehicletotheinitial trimstateattheendof command. Insomeofthecasesaresidualyawangleisobtained.Inthese casestheyawangleis eitherincreasingordecreasingrapidlyattheendofthecom mand.Inotherwords,the vehiclenolongerhasastraight-levelighttrimattheendo fthecommandsequence. Insomeofthecaseswhenthereisaresidualyawangle,theyaw anglegoestozero exponentiallyandthenthevehicleistrimmedattheorigina ltrimstate. Insomeofthecases,thevehiclehasaconstantresidualyawa ngleattheendofthe commandsequence.Thevehiclealsohasaconstantpitchangl einthesecases.

PAGE 116

105 Insomeothercasesthevehiclegoesunstableasthecontrolr atelimitsarereached. Allthepossiblecombinationsofthepitch-rollcommandsha vebeentestedfortheLQR controller,andallofthosehaveatleastoftheaboveproble ms. 0 5 10 15 0 2 4 6 8 10 12 14 16 18 20 time (s)command roll anglepitch angle 0 5 10 15 -20 -15 -10 -5 0 5 10 15 20 time (s)command roll ratepitch rate Figure8.2VariousPossibilitiesofSequenceofCommandstoO btainYawRateControl 8.5.2Yawcontrolusingthe /H controller The controllerobtainedintheChapter 7 giveamuchbetterperformancefortheYaw control.Again,thesequenceofcommandsisapositiverolla nglecommand,followedbya positivepitchanglecommand.ThisisshownintheFigure 8.3 .Thesequenceofcommands givesaresidualyawangleattheendofthecommandsequence. Moreover,thereareno

PAGE 117

106 residualvaluesforanyotherstates,i.e.,thevehiclehase ffectivelychangeditsbearing.The orientationanglesforthevehicleareshownintheFigure 8.4 0 2 4 6 8 10 0 2 4 6 8 10 12 14 16 18 20 time (s)command roll anglepitch angle Figure8.3SequenceofCommandstoObtainYawRateControl 0 2 4 6 8 10 -5 0 5 10 15 20 time (s)response yqf Figure8.4YawAngleResponseoftheVehiclefortheCommandsSh owninFigure 8.3

PAGE 118

CHAPTER9 CONCLUSION 9.1Summary Adynamicalmodelforasupercavitatingvehiclehasbeenobt ained.Thevehicleis foundtobeopen-loopunstable,andacontrollerforstabili zingthepitchandrollratemotion hasbeenobtained.TheLQRcontrollershowsgoodtrackingpe rformanceforthevehicle andallthecontrolobjectivesaremet.Thecontrollerisals ofoundtoberobusttoerrorsin cavitypredictionandvelocitychanges.Thisrobustnessis furtherdemonstratedbythefact thattheclosed-loopsystemhashighgainandphasemargins. Robustcontrollerforpitch androllangleareobtainedby /H synthesis.Thecontrollergivegoodperformanceand showgoodrobustness. Methodologiesforhomingnavigationhavebeenformulated. Inthesimplestcase,constantvelocitymissileandtarget,theguidancetocollisio nisproportionalnavigation.The guidancetocollisionlawsformorecomplicatedcaseshaveb eenformulated.Yawrate controllerisdesignedforhomingguidance.TheLQRcontrol lershaveseveralcoupling problemwhenusedforyawratecontrol.Therobustpitchandr ollanglecontrollersare usedasinner-loopforyawratecontrol.Agoodresidualyawi sobtainedinthiscase. 9.2FutureWork TheLQRcontrollersobtainedaretypicallyknownasthe`inn er-loop'controllers.An outer-loopcontrollerisalsoneededforguidanceandnavig ation.Theideaisthattheouterloopcontrollercanbemodeledfortrackingthetrajectoryi nspace,basedontheclosed-loop dynamicsoftheinner-loopmodel.Thisissimilartousingth erobustcontrollerfortheyaw control. 107

PAGE 119

108 HomingControl:Homingcontrolmethodologiesforsimplecas eshavebeenformulated.Thesemethodologiesrelyontheaccuracyofsensorsa ndtime-to-goprediction. Thesestrategiesareefcientwhenthetargetisclosetothe missile.Methodologiesforcase whenthemissileandtargetareseparatedbylargedistances havetobepredicted. Sensorcharacteristics:Thecurrentmodelassumesallthes tatesareavailableforthe feedback.Adetailedstudyofthesensorsisrequired.Theav ailablesensorsandtheir characteristics,likedelay,rangeetc.,affecttheactual performanceofthecontroller.The vehicleisequippedwithMEMSsensorsthatmeasurethedista nceofthecavitywallfrom thevehiclewalls.Thesesensorscanbeusedtopredicttheca vityshape,immersionetc. besidesusingtheequationsforthecavity. CavityControl:Thevehicleissurroundedbyadynamiccavity. Thevehicledynamics andstabilityarehighlydependentonproductionandsusten anceofthecavity.Thusa detailedstudyofthecavityisrequired.Intheory,thecavi tycanalsobeusedtocontrolthe vehicle.Thisisachallengingproblemandrequiresfurther study.

PAGE 120

APPENDIXA REFERENCEFRAMESANDROTATIONMATRICES x 2 y 2 x 3 y 3 x 1 x 2 q FigureA.1RotationofFrames Figure A.1 showstwoframesX f x 1 x 2 x 3 g andY f y 1 y 2 y 3 g .YisrotatedfromXbyan angle q aboutx-axis.ThusthebasisvectorsofframeYcanbewritten intermsofbasis vectorsofXframe. y 2 = x 2 cos ( q )+ x 3 sin ( q ) y 3 = x 2 sin ( q )+ x 3 cos ( q ) (A.1) Thisrelationcanalsobeexpressedintermsofmatrices. 8>>>><>>>>: y 1 y 2 y 3 9>>>>=>>>>; = 266664 1000 cos ( q ) sin ( q ) 0 sin ( q ) cos ( q ) 377775 8>>>><>>>>: x 1 x 2 x 3 9>>>>=>>>>; (A.2) Thiswasacaseofsimplerotation.Thematrixaboveinsquare bracketsisknownas therotationmatrixfromXtoYandisrepresentedasX Y.Therotationmatrixcanbe generalizedforacasewhenthetworeferenceframesarearbi trarilyoriented. 109

PAGE 121

110 X Y = 266664 ( y 1 ; x 1 )( y 1 ; x 2 )( y 1 ; x 3 ) ( y 2 ; x 1 )( y 2 ; x 2 )( y 2 ; x 3 ) ( y 3 ; x 1 )( y 3 ; x 2 )( y 3 ; x 3 ) 377775 (A.3) where(,)meansthedotproductofthetwovectors.Thus, Y = X Y X (A.4)

PAGE 122

APPENDIXB NUMERICALTECHNIQUES B.1InterpolationofForceData Thissectiondescribesthenumericaltechniqueusedtoobta inthevaluesofcoefcients ofliftanddragforcavitatorandns.B.1.1Extrapolationscheme Forabetterresult,aquadraticinterpolation/extrapolat ionschemeisused.Thus3points wouldberequiredtoobtainaninterpolatedorextrapolated datavalue.Figure B.1 shows theshapefunctionsusedforonedimensionalinterpolation .Say,points f x i 1 ; x i ; x i + 1 g are usedtondthevalueofafunction f atpoint x .Thevalueof f ( x ) wouldbegivenbya parameter a andthethreeshapefunction N 1, N 2and N 3. N 1 = 1 + ( 2 x i 1 x i x i + 1 ) ( x i x i 1 ) a + ( x i + 1 x i 1 ) ( x i x i 1 ) a 2 N 2 = ( x i + 1 x i 1 ) 2 ( x i 1 x i )( x i x i + 1 ) a + ( x i + 1 x i 1 ) 2 ( x i 1 x i )( x i x i + 1 ) a 2 N 3 = ( x i 1 x i ) ( x i x i + 1 ) a + ( x i 1 x i + 1 ) ( x i x i + 1 ) a 2 (B.1) where,thevalueofashapefunctioncanbeobtainedbynding thevalueof a a = x x i 1 x i + 1 x i 1 (B.2) a 2 [ 0 ; 1 ] forx 2 [ x i 1 ; x i + 1 ] a < 0 forx < x i 1 a > 1 forx > x i + 1 Thus a 2 [ 0 ; 1 ] forinterpolationanditisgreaterthan1orlessthan0forex trapolation. f ( x )= N 1 f ( x i 1 )+ N 2 f ( x i )+ N 3 f ( x i + 1 ) (B.3) 111

PAGE 123

112 0 0.2 0.4 0.6 0.8 1 -0.5 0 0.5 1 AlphaShape FunctionN1 N2 N3 X i-1 X X i i+1 FigureB.1ShapeFunctionforOneDimensionalQuadraticSche me TableB.1GridForExperimentalCavitatorData HalfAngle( h a deg) f 15,30,45,60,75,90 g AngleofAttack( a c deg) f 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15 g Thismethodcanbeextendedfor2Dand3Dasincaseforcavitat orandnsrespectively. B.1.2Cavitator Thecoefcientsoflift( cl c )anddrag( cd c )forthecavitatorarefunctionsofhalfangle ( h a )ofcavitatorconeandangleofattackforcavitator( a c ).TheCFDdata[ 5 ]isavailableforcombinationofpointsgiveninTable B.1 .Equation B.3 canbeextendedfor2D cavitator. f ( a c ; h a )= 3i = 1 3j = 1 N ( 1 ; i ) N ( 2 ; j ) f ( a c ( i ) ; h a ( j )) (B.4) a c ( i ) Valueof a c at i th node h a ( j ) Valueof h a at j th node N ( 1 ; i ) i th Shapefunctionfor a c N ( 2 ; j ) j th Shapefunctionfor h a

PAGE 124

113 TableB.2GridForExperimentalFinData Immersion( S f ) f 0.1,0.3,0.5,0.7,0.9 g Sweepback( q f deg) f 0,15,30,45,60,70 g AngleofAttack( a f deg) f 0,1,2,3,4,5,6,7,8,9,10,12,15 g DatanotAvailablefor 1. S 0& S < 0 : 1& q 0 2. S 0 : 1& S < 0 : 3& q 30 3. S 0 : 3& S < 0 : 5& q 45 4. S 0 : 5& S < 0 : 7& q 60 5. S 0 : 7& S < 0 : 9& q 70 6. S 0 : 9& S 1& q 0 7. a 0 8. a > 15 B.1.3Fins Thecoefcientsoflift( cl fin )anddrag( cd fin )forthensarefunctionsofangleof attack( a f )forn,immersion( S f )andsweepbackangle( q f ).TheCFDdataisavailable forcombinationofpointsgiveninTable B.2 .Equation B.3 canbeextendedfor3Dn. f ( S f ; q f ; a f )= 3i = 1 3j = 1 3k = 1 N ( 1 ; i ) N ( 2 ; j ) N ( 3 ; k ) f ( S f ( i ) ; q f ( j ) ; a f ( k )) (B.5) S ( i ) Valueof S f at i th node q f ( j ) Valueof q f at j th node a f ( k ) Valueof a f at k th node N ( 1 ; i ) i th Shapefunctionfor S f N ( 2 ; j ) j th Shapefunctionfor q f N ( 3 ; j ) k th Shapefunctionfor a f B.2NumericalLinearization Numericallinearizationcanbedonebythe`linmod'command intheMatlabSimulink toolbox.Thiscanalsobedonebynotingthat,thetermsinthe A and B matricesarethe derivativesofstaterateswithrespecttostatesandcontro ls.Forexample,suppose x 0 and u 0 representthestateandcontrolvaluesattrim.Itshouldben otedthat x 0 isa9 1(excluding

PAGE 125

114 thepositions f x ; y ; z g )vectorand u 0 is5 1(cavitatorandfourns). x 0 = u 0 w 0 q 0 Q 0 v 0 p 0 r 0 F 0 Y 0 T u 0 = d c 0 d e 1 0 d e 2 0 d r 1 0 d r 2 0 T (B.6) Theequationsofmotionareoftheformasinequation B.7 x = f ( x ; u ) (B.7) wherethefunction f isavectorfunctionhaving9outputsandthereare9states.T hecode fornonlinearequationofmotion,takes x and u asinputsandgivethevalueof x asoutput. Let e deneaverysmallchange.Nowtheelement A ( i ; j ) canbecalculatedasinequation B.8 A ( i ; j )= f ( x 0 + e ( j ) ; u 0 ) i f ( x 0 ; u 0 ) i e 1 i ; j 9(B.8) where, e ( j ) meansamatrixofsize x 0 withallzerosexcept j th element,whichisequalto e ,and f i representsthe i th elementofvector f .Anelement B ( i ; j ) alsocanbeobtainedin asimilarway. B ( i ; j )= f ( x 0 ; u 0 + e ( j )) i f ( x 0 ; u 0 ) i e 1 i 9 ; 1 j 5(B.9) where, e ( j ) meansamatrixofsize u 0 withallzerosexcept j th element,whichisequalto e

PAGE 126

APPENDIXC TESTBEDDATA C.1FinParametersfortheTestBedandFin-Data Test-beddataforthehydrodynamicliftanddragcoefcient softhenhasbeenobtained fromEngineeringTechnologyCenter(ETC),OfceofNavalResea rch(ONR).Theelevator ngeometryparameters,S/L,S/dandb,wereselectedaccord ingtotheInterimTestBed DesignReviewofMay[ 4 ].Thebodycavityisassumedtobeaxi-symmetric.Thepoint aboutwhichthensrotateisassumedtobelocatedatanaxial distancedownstreamofthe cavitatorequalto85%ofthelengthofthetorpedo.Inthena nalysis,thebodycavityis treatedasaconstantdiametercylinderinthevicinityofth ens.Theeffectsofgravityare notincludedinanyofthenforcecalculations. Eachsetofforces(andliftanddragcoefcients)iscalcula tedforagivensetofthe followingparameters: s ,S/L,S/d, b a q ,andSo/S.Forthepresentdata-set,S/L,S/d, b and s areallheldxed.Theimmersion,So/S,isdenedforzeroswe epbackangle, q =0.Theliftanddragforcesaredenedatthelocationofthec enterofpressureofthe nandarenormalizedby:1 = 2 r V 2 S 2 and1 = 2 r V 2 S 3 ,respectively,whereVisthesteady forwardspeedofthetorpedo.TheTable C.1 givesthevaluesofthenvariablesforwhich experimentaldataisavailable.ThegraphsinFigures C.2 C.7 ,givetheplotsofthetestbed datafordifferentvaluesofthesweepbackangle. Thedatafromthisdatabaseisusedtondthevaluesofcoefc ientsofliftanddragfor thens.TheextrapolationschemeforthisisshowninAppend ix B 115

PAGE 127

116 FigureC.1HydrodynamicForcesActingattheCenterofPressur eofFin. TableC.1TheFinVariablesRangeSet. s 0.029 a 0 o 10 o ; 12 o ; 15 o q 0 o ; 15 o ; 30 o ; 45 o ; 60 o ; 70 o S 0 / S 0 : 1 ; 0 : 3 ; 0 : 5 ; 0 : 7 ; 0 : 9 C.2CavitatorParametersfortheTestBedandCavitator-Data TheCavitatorhastheshapeofacone.Themainparametersonwh ichthecoefcients dependaretheconehalf-angle, g 1 2 ,andangleofattack, a C .Theliftanddragforcesact atthecenterofpressureoftheconicalcavitator.Sincethe cavitatorissmallcomparedto theoverallsizeofthetorpedothecenterofpressureisassu medtobestaticforvarying wettedareasofthecone.Thecavitatorhasonedegreeoffree dom,thepitchangle d C Experimentaldataforthecavitatorconehasalsobeenobtai nedfromETC,ONR.The graphinFigure ?? givestheliftanddragcoefcienttestbeddatafortheconic alcavitator. Thedatafromthisdatabaseisusedtondthevaluesofcoefc ientsofliftanddragfor thecavitator.Theextrapolationschemeforthisisshownin Appendix B

PAGE 128

117 L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r0.00r Clr Submersion (So/S)r -0.020r 0.000r 0.020r 0.040r 0.060r 0.080r 0.100r 0.120r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r Attack Angle 9 r Attack Angle 10 r Attack Angle 6 r Attack Angle 7 r(A) L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r0.00r Cdr Submersion (So/S)r -0.005r 0.000r 0.005r 0.010r 0.015r 0.020r 0.025r 0.030r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r(B) FigureC.2Fin Cl and Cd TestBedData:SweepAngle=0.A)FinCl;B)FinCd L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r15.00r Fzr Submersion (So/S)r -0.010r 0.000r 0.010r 0.020r 0.030r 0.040r 0.050r 0.060r 0.070r 0.080r 0.090r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r15.00r Clr Submersion (So/S)r -0.020r 0.000r 0.020r 0.040r 0.060r 0.080r 0.100r 0.120r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r Attack Angle 9 r Attack Angle 10 r Attack Angle 6 r Attack Angle 7 r Interpolated Valuesr(A) L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r15.00r Cdr Submersion (So/S)r -0.010r 0.000r 0.010r 0.020r 0.030r 0.040r 0.050r 0.060r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r Interpolated Valuesr(B) FigureC.3Fin Cl and Cd TestBedData:SweepAngle=15.A)FinCl;B)FinCd L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r30.00r Clr Submersion (So/S)r -0.020r 0.000r 0.020r 0.040r 0.060r 0.080r 0.100r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r Attack Angle 9 r Attack Angle 10 r Attack Angle 6 r Attack Angle 7 r(A) L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r30.00r Cdr Submersion (So/S)r -0.005r 0.000r 0.005r 0.010r 0.015r 0.020r 0.025r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r(B) FigureC.4Fin Cl and Cd TestBedData:SweepAngle=30.A)FinCl;B)FinCd L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r45.00r Clr Submersion (So/S)r -0.010r 0.000r 0.010r 0.020r 0.030r 0.040r 0.050r 0.060r 0.070r 0.080r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r Attack Angle 9 r Attack Angle 10 r Attack Angle 6 r Attack Angle 7 r(A) L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r45.00r Cdr Submersion (So/S)r -0.002r 0.000r 0.002r 0.004r 0.006r 0.008r 0.010r 0.012r 0.014r 0.016r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r(B) FigureC.5Fin Cl and Cd TestBedData:SweepAngle=45.A)FinCl;B)FinCd

PAGE 129

118 L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r60.00r Clr Submersion (So/S)r -0.010r 0.000r 0.010r 0.020r 0.030r 0.040r 0.050r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r Attack Angle 9 r Attack Angle 10 r Attack Angle 6 r Attack Angle 7 r(A) L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r60.00r Cdr Submersion (So/S)r -0.001r 0.000r 0.001r 0.002r 0.003r 0.004r 0.005r 0.006r 0.007r 0.008r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r(B) FigureC.6Fin Cl and Cd TestBedData:SweepAngle=60.A)FinCl;B)FinCd L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r70.00r Clr Submersion (So/S)r -0.025r -0.020r -0.015r -0.010r -0.005r 0.000r 0.005r 0.010r 0.015r 0.020r 0.025r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r Attack Angle 9 r Attack Angle 10 r Attack Angle 6 r Attack Angle 7 r(A) L/SrWedgeHAngrLp/LrCavNorSwpBckAngr 0.25r10.0r0.35r0.029r70.00r Cdr Submersion (So/S)r -0.001r -0.001r 0.000r 0.001r 0.001r 0.002r 0.002r 0.003r 0.003r 0.0r0.2r0.4r0.6r0.8r1.0r Attack Angle 0r Attack Angle 1r Attack Angle 2r Attack Angle 3r Attack Angle 4r Attack Angle 5r Attack Angle 8 r Attack Angle 12r Attack Anlge 15r(B) FigureC.7Fin Cl and Cd TestBedData:SweepAngle=70.A)FinCl;B)FinCd

PAGE 130

REFERENCES [1]S.Ashley, WarpdriveUnderwater ,2002, http://www.diodon349.com/KurskMemorial/Warpdrive underwater.htm ,accessed:March2002. [2]I.N.Kirschner,D.C.Kring,A.W.Stokes,N.E.Fine,N.E.U hlman,Jr., “Controlstrategiesforsupercavitatingvehicles,”AnteonCorporation/ETC,North Kingstown,2001. [3]R.Kuklinski,C.HenochandJ.Castano,“Experimentalstudy ofventilatedcavities ofdynamictestmodel,”NavalUnderseaWarfareCenter,Cav200 1:SessionB3.004, 2001. [4]A.May, WaterEntryandCavity-RunningBehaviorofMissiles ,Arlington,VA, SEAHACTR75-2,NavalSeaSystemsCommand,1975. [5]N.Fine, SixDegree-of-FreedomFinForcesfortheONRSupercavitatin gTestBed Vehicle ,AnteonCorporation,2000, http://www.anteon.com ,accessed:September 2000. [6]S.S.KulkarniandR.Pratap,“Studiesondynamicsofasupe rcavitatingprojectile,” AppliedMathematicalModeling ,vol.24,pp.113–129,2000. [7]R.Rand,R.Pratap,D.Ramani,J.CipollaandI.Kirschner,“Imp actdynamics ofasupercavitatingunderwaterprojectile,”in ProceedingsoftheThirdInternationalSymposiumonPerformanceEnhancementforMarineApp lications ,Newport,RI,pp.215–223,1997, http://tam.cornell.edu/randpdf/Rand pub.html ,accessed:March2002. [8]J.DzielskiandA.J.Kurdila,“Abenchmarkcontrolprobl emforsupercavitating vehiclesandaninitialinvestigationofsolutions,” JournalofVibrationandControl vol.9,Jul.2003,pp791-804. [9]R.L.Waid,“Cavityshapesforcirculardisksatanglesofat tack” CITHyd. Rpt. E-73.4,1957. [10]H.Reichardt,“Lawsofcavitationbubblesataxiallysym metricbodiesinaow [sic],”MinistryofAircraftProductionVolkenrode,MAP-V G,ReportsandTranslations766,OfceOfNavalResearch,Arlington,VA. [11]G.V.Logvinovich, HydrodynamicsofFlowswithFreeBoundaries ,Kyiv, Trudy TsAGI, 1980. 119

PAGE 131

120 [12]H.Wagner, UberStossundGleitvarangangeanderOberachevonFlussig kaiten Berlin, ZAMM ,12(4),193215. [13]S.E.Hassan, AnalysisofHydrodynamicPlaningForcesAssociatedwithCavi ty RidingVehicles ,NUWC-NOTTechnicalMemorandum90085,Jul.1999. [14]D.R.Stinebring,M.L.Billet,J.W.LindauandR.F.Kunz, DevelopedCavitationCavityDynamics ,AppliedResearchLaboratory,StateCollege,PA16804,USA. [15]R.C.Nelson, FlightStabilityandAutomaticControl ,Boston,MA,McGrawHill, 1997. [16]G.Balas,R.Chiang,A.PakcardandM.Safonov, MuControlToolboxforUsewith Matlab ,Natick,MA,TheMathworks,Inc.,2001. [17]K.Ogata, ModernControlEngineering ,UpperSaddleRiver,NJ,PrenticeHall, 2002. [18]S.Wolfram, TheMathematicaBook ,4ed.,NewYork,NY,ThePressSyndicateof theUniversityofCambridge,1999. [19]M.Tahk,C.RyooandC.Hangju,“Recursivetime-to-goestima tionforhoming guidancemissiles,” IEEETransactiononAerospaceandElectronicSystems ,vol. 38no.1,Jan.2002,pp.13–24. [20]C.S.Shieh,“Nonlinearrulebasedcontrollerformissil eterminalguidance,” IEEE ProceedingsonControlTheoryandApplications ,vol.150no.1,Jan.2003,pp 45–48. [21]Z.Rio,“Designofclosedloopoptimalguidancelawusing neuralnetworks,” ChineseJournalofAeronautics ,vol.15no.2,May2002,pp98-102. [22]Y.Ochi,K.ItohandK.Kanai,“Missileguidancelawdesi gnvia -synthesis,” TransactionsoftheJapanSocietyforAeronauticalandSpac eSciences ,vol.45 no.148,Aug.2002,pp102-107. [23]S.Jian-Mei,“Modelreferencevariablestructureauto pilotdesignforhomingmissile,” JournalofBeijingInstituteofTechnology ,vol.10no.4,Dec.2001,pp364369. [24]D.ZhouandC.Mu,“Motiontrackingsliding-modeguidanc eofahomingmissile withbearings-onlymeasurements,” TransactionsoftheJapanSocietyforAeronauticalandSpaceSciences ,vol.43no.141,Nov.2000,pp130-136. [25]P.Gurl,M.JodorkovskyandM.Guelman“Neoclassicalg uidanceforhoming missiles,” JournalofGuidance,ControlandDynamics ,vol.24no.3,pp452-459. [26]P.A.IglesiasandT.J.Urban,“Loopshapingdesignform issileautopilots:controllercongurationsandweightinglterselection,” JournalofGuidance,Control andDynamics ,vol.23no.3,May-June2000,pp516-525.

PAGE 132

121 [27]D.Zhou,C.MuandW.Xu,“Adaptiveslidingmodeguidanceo fahomingmissile,” JournalofGuidance,ControlandDynamics ,vol.22no.4,Jul-Aug1999,pp589594. [28]J.ShinarandT.Shima,“Nonorthodoxguidancelawdevel opmentforintercepting maneuveringtargets,” JournalofGuidance,ControlandDynamics ,vol.25no.4, Jul-Aug2002,pp658-666. [29]C.YangandH.Chen,“NonlinearH robustguidancelawforhomingmissiles,” JournalofGuidance,ControlandDynamics ,vol.21no.6,Nov-Dec1998, pp882-890. [30]C.Lin,Y.Chen,“Designoffuzzylogicguidancelawagains thigh-speedtarget,” JournalofGuidance,Control,andDynamics ,vol.23no.12000.p17–25, [31]T.L.SongandT.Y.Um,“Practicalguidanceforhomingmi ssileswithbearingsonlymeasurements,” IEEETransactionsonAerospaceandElectronicSystems ,vol. 32no.1,Jan.1996,pp434–443 [32]D.Rew,M.TahkandH.Cho,“Short-timestabilityofpropor tionalnavigationguidanceloop,” IEEETransactionsonAerospaceandElectronicSystems ,vol.32no.3, Jul.1996,pp1107–1115. [33]G.T.LeeandJ.G.Lee,“Improvedcommandtoline-of-sig htforhomingguidance,” IEEETransactionsonAerospaceandElectronicSystems ,vol.31no.1,Jan.1995, pp506–510. [34]Ciann-Dong,F.HsiaoandF.Yeh,“Generalizedguidancel awforhomingmissiles,” IEEETransactionsonAerospaceandElectronicSystems ,vol.2,1989,pp 197–212.

PAGE 133

BIOGRAPHICALSKETCH AnukulGoelwasborninLucknow,India,onMarch3 rd ,1978,andraisedinHyderabad, India.AnukulattendedtheIndianInstituteofTechnology, locatedinMumbai,India,where hereceivedaBachelorofTechnologydegreeinaerospaceengi neeringin2000. AnukuljoinedtheDepartmentofAerospaceEngineeringatth eUniversityofFlorida inAugust2000andgotaM.S.inaerospaceengineeringinDece mber2002.Since2002, AnukulhasattendedtheCollegeofEngineeringattheUnivers ityofFlorida,Gainesville,to pursuehisPh.D.inaerospaceengineering.Duringthistime heworkedasateachingassistantandaresearchassistantintheMechanicalandAerospac eEngineeringDepartmenton apart-timebasis.Hisresearchinterestsincludecontrols anddynamicsandoptimization. 122


Permanent Link: http://ufdc.ufl.edu/UFE0010119/00001

Material Information

Title: Robust Control of Supercavitating Vehicles in the Presence of a Dynamic and Uncertain Cavity
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0010119:00001

Permanent Link: http://ufdc.ufl.edu/UFE0010119/00001

Material Information

Title: Robust Control of Supercavitating Vehicles in the Presence of a Dynamic and Uncertain Cavity
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0010119:00001


This item has the following downloads:


Full Text









ROBUST CONTROL OF SUPi i': / AITATING VEHICLES IN THE~ PRESii .:-E OF
DYINAMICl AND UNCERTAIN CAVITY














By

ANUKUL GOEL


A DIIi'.i iT'; iON PRESENTED TO TH-IE GRADUATE SCHOOL
OF THIE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUI~il- i NTS FOR THE~ DEGREE OF
DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA~














ACKNO W1V i ii: -1'- il ii `TS


I would like to : : my sincere gratitude to my committee chairman, Dr. A~ndrew

Kurdila, for hi s invaluable guidance throughout the course of this proj ect. I would also liket

to thank him for giving me this opportunity to work on such a i : :::.:i T::. proj ect.

I would also like to thank my committee cochair, Dr. Richard C. Lind, for his invaluable:

guidance and inspiration throughout the i :- : i

I would like to thank Dr. Normnan Fitz-Coy, Dr. Brian Mann and Dr. Haniph Latchman

for : on m~y committee. I would also like to showv my sincere .:~.:..: -..i: to D~r.

John IDzielsk~i and Jamnmulamnadakia Anand Kapardi for their valuable contributions to this

i 1 w .iIvould also like to .... m~y gratitude to all the members, past and present, of

the Supercavitation r:l

I would also like to thank the Office of Naval Research for the support of the research

grant for the :.:;

On a personal note, I would like to thank all my friends and family members whose:

support helped me to aim towards my goals.















TABLE OF CONTENTS


page


ACKNOWLEDGMENTS ......i

LIST OFTABLES . vi

LIST OF FIGURES . ix

ABSTRACT ......x

CHAPTER

1 INTRODUCTION . 1

1.1 Cavitation . . 1
1.2 Aspects of Supercavitation . 3
1.3 Related Research . . 4
1.4 Overview of This Dissertation . 5


2 NONLINEAR EQUATIONS OF MOTION . 7

2.1 Configuration of the Vehicle . 7
2.1.1 Body . 7
2. 1.2 Cavitator . 8
2.1.3 Fins . 9
2. 1.4 Maneuvering . 9
2.2 Kinematic Equations of Motion . . 10
2.2.1 Orientation of the Torpedo . . 10
2.2.2 Orientation of the Cavitator . . 12
2.2.3 Orientation of Fins . . 13
2.2.4 Angle of Attack and Sideslip . . 16
2.2.5 Kinematic Equations . . 20
2.3 Dynamic Equations of Motion . . 22
2.3.1 Forces on Cavitator. . . 24
2.3.2 Forces on Fins . . 27
2.3.3 Gravitational Forces . . 30
2.3.4 Equations of Motion . . 31
2.3.4.1 Force equations . . 31
2.3.4.2 Moment equations . . 32
2.3.4.3 Orientation equations . . 32
2.3.4.4 Navigation equations . . 32










3 CAVITY AND PLANING DYNAMICS


3.1 Munzer-Reichardt Model . . 35
3.2 Logvinovich Cavity Model . . . 35
3.2.1 Logvinovich Theory of Independent Expansion . . 35
3.2.2 Cavity Centerline . . 36
3.3 Planing Model. . . 38
3.4 Planing Force Equations . . 39
3.5 Planing Kinematics . . 42
3.5.1 Calculation of Immersion . . 42
3.5.2 Method of Calculation of C O.... . . 42

4 LINEARIZATION . 47

4.1 Linearization . . 47
4.1.1 Need for Linearization . . 47
4. 1.2 Generic Form of Equations of Motion . . 48
4.1.3 Small Dishractuerbac Theory .. .. . . 48
4.1.3.1 Force equations . . 50
4.1.3.2 Moment equations . . 50
4.1.3.3 Orientation equations . . 50
4.1.3.4 Position equations . . 50
4. 1.4 Stability and Control Derivatives . . 50
4.2 State Space Representation . . 55

5 CONTROL DESIGN SETUP . 59

5.1 Open-Loop Performance for the Fixed Cavity Model . . 60
5.2 Closed-Loop Problem . . 63
5.3 Robustness of the Controller . . 64
5.3.1 Gain margin . . 65
5.3.2 Phase margin . . 65
5.3.3 Uncertainty in parameters . . 65
5.3.4 Controller objective. . . 66
5.3.5 pu analysis: 66

6 LQR CONTROL . 67

6.1 LQRTheory.. .. .. ................... .. 67
6.2 LQR Control for Fixed Cavity Model: 71
6.2.1 Control Synthesis. . . 71
6.2.2 Nominal Closed-loop Model . . 73
6.2.2.1 Model . . 73
6.2.2.2 Simulations . . 73
6.2.2.3 Gain and phase margins . . 76
6.2.3 Perturbed Closed-loop Model . . 77











6.2.3.1
6.2.3.2
6.2.3.3


Model
Simulations
Gain and phase margins


7 p/lHo SYNTHESIS CONTROL

7.1 Uncertainty
7.2 Synthesis Model. .
7.3 Control Objective and Constraints
7.4 p Controller for Fixed Cavity Model
7.4.1 Longitudinal controller .
7.4.2 Lateral controller .


8 HOMING CONTROL

8.1 Homing Control using Proportional Navigation .
8.2 Constant Missile and Target Velocity . .
8.3 Constant Velocity Target and Accelerating Missile .
8.4 Accelerating Target and Missile ....
8.5 Yaw Controller .... ..... ..
8.5.1 Yaw control using the LQR controllers ..
8.5.2 Yaw control using the p/~Ho controller .

9 CONCLUSION ...

9.1 Summary ....
9.2 Future Work . .


. . 99
100
. . 101
102
104
. 104
. 105


107

107
107


APPENDIX


A REFERENCE FRAMES AND ROTATION MATRICES ..


B NUMERICAL TECHNIQUES ...

B.1 Interpolation of Force Data . .
B.1.1 Extrapolation scheme . .
B.1.2 Cavitator ....
B.1.3 Fins .. .. .
B.2 Numerical Linearization ....


C TESTBED DATA ...

C.1 Fin Parameters for the Test Bed and Fin-Data .....
C.2 Cavitator Parameters for the Test Bed and Cavitator-Data


REFERENCES ...


BIOGRAPHICAL SKETCH ...


.109


111

... .111
... .111
.... .. .12
.........13
.... .. .13


115

.... .. .15
. .. .. .. .16


119


S 122














LIST OF TABLES


Table page

5.1 Control Parameters . . 59

5.2 Control Constraints . . 64

6.1 Gain and Phase Margin with LQR Controller . . 77

6.2 Percentage Variation in A Matrix due to 20% Variation in clc . . 78

6.3 Percentage Variation in B Matrix due to 10% Variation in clc . . 79

6.4 Gain and Phase Margin for Perturbed Closed-loop System: 20% error in clfi, 81

7.1 Gain and Phase Margins for the Longitudinal Controller . . 93

7.2 Gain and Phase Margins for the Lateral Controller . . 96

B.1 Grid For Experimental Cavitator Data . . 112

B.2 Grid For Experimental Fin Data . . 113

C.1 The Fin Variables Range Set. . . 116














LIST OF FIGURES


Figure


page


Supercavitating Vehicle.

Supercavitating Vehicle.

Vehicle Body is Divided Into 4 Sections.

Cavitator and Fins .

Body-Fixed and Inertial Frames

Principal Planes of Symmetry for the Torpedo.

Euler Angles of Rotation

Cavitator Reference Frame

Rudder and Fin Reference Frames

Rudder 1 Fin Reference Frames

Rudder 2 Fin Reference Frames

Elevator 1 Fin Reference Frames.

Elevator 2 Fin Reference Frames.

Angle of Attack (u) and Sideslip (P)

Cavitator: (a) Angle of Attack and Sideslip and I

Fin Geometry

Vehicle Undergoing Tail-Slap.

Cross-Section of Cavity.

Planing Section

The Angle .

Velocity of Transom Relative to Frame B .

Calculation of a with Assumption that the Cavit


1.1

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

2.10

2.11

2.12

2.13

2.14

2.15

3.1

3.2

3.3

3.4

3.5

3.6


(b) Hydrodynamic Forces


y Surface is Planar .


5.1 Simulink Model for Open Loop Simulation









5.2

5.3

5.4

6.1

6.2

6.3

6.4

6.5

6.6

6.7

6.8

6.9

6.10

6.11

6.12

6.13

7.1

7.2

7.3

7.4

7.5

7.6

7.7

7.8

7.9

7.10

7.11


Open-Loop Response for Torpedo: w, p, q

Variation of Eigenvalues with Change in Velocity

Loop Gain.

Controller for Tracking when Plant has an Integrator.

Controller for Tracking when Plant has no Integrator.

Eigenvalues for the Closed-Loop System .

Pitch Command Tracking : q.

Pitch Command Tracking : Sc. 6c

Pitch Command Tracking : Se, Bel -

Roll Command Tracking: p .

Roll Command Tracking: Sc. c -

Roll Command Tracking: Se, Bel .

Breakpoints for Calculating the Loop-Gain for a Tracking Controller.

Eigenvalues for the Perturbed Closed-Loop System: 20% Error in clfi,

Response for 20% Variation in clfi,: p, q .

Response for 20% Variation in clfi,: 6c,6el

Calculation of Uncertainty

Linear Fractional Representation.

Synthesis Model for pu Controller.

Pitch Angle Tracking for the Nominal Plant

Cavitator and Elevator Deflections

Frequency Response of the Pitch Angle Controller

Comparison of Response for the Nominal and Perturbed Plants.

Roll Angle Tracking for the Nominal Plant .

Cavitator and Elevator Deflections

Frequency Response of the Roll Angle Controller.

Comparison of Response for the Nominal and Perturbed Plants.









8.1 Collision Triangle . . 99

8.2 Various Possibilities of Sequence of Commands to Obtain Yaw Rate Control 105

8.3 Sequence of Commands to Obtain Yaw Rate Control . . 106

8.4 Yaw Angle Response of the Vehicle for the Commands Shown in Figure 8.3 106

A.1 Rotation of Frames . . 109

B.1 Shape Function for One Dimensional Quadratic Scheme . . 112

C.1 Hydrodynamic Forces Acting at the Center of Pressure of Fin. . 116

C.2 Fin Cl and Cd Test Bed Data: Sweep Angle = 0. A) Fin Cl; B) Fin Cd .. 117

C.3 Fin Cl and Cd Test Bed Data: Sweep Angle = 15. A) Fin Cl; B) Fin Cd 117

C.4 Fin Cl and Cd Test Bed Data: Sweep Angle = 30. A) Fin Cl; B) Fin Cd 117

C.5 Fin Cl and Cd Test Bed Data: Sweep Angle = 45. A) Fin Cl; B) Fin Cd 117

C.6 Fin Cl and Cd Test Bed Data: Sweep Angle = 60. A) Fin Cl; B) Fin Cd 118

C.7 Fin Cl and Cd Test Bed Data: Sweep Angle = 70. A) Fin Cl; B) Fin Cd 118















Abstract of D~issertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of ii ::1 i: .: i i:y

ROBUST CONTROL OF SUPli iAV ITATING VEHICLES IN THE~ PRESii .:-E OF
DYINAMICl AND UNCERTAIN CAVITY

B~y

Anukul Goel

August I


Chair: Andrew J. Kurdila
Cochair: Richard C. Lind
..I~~~ ..-I=1.0...i Mechanical and Aerospace Engineering


Ulnderwvater travel is greatly limited by the speed that can be attained by the vehicles.

L : :II :i, the maximnum speed achieved by underwater vehicles is about 40 mn/s. Supercav-

itation can be viewed as a phenomenon that would help us to break the speed barrier in

underwater vehicles. The idea is to make the vehicle surrounded by water vapor while it

is traveling underwater. Thus, the vehicle actually travels in air and has very small skin

friction drag. This allowi~s it to attain high speeds underwater. ":: --: : l.:IT :: is a phe--

nomenon which is observed in water. As the relative velocity of water with ** i to

the vehicle increases, the pressure decreases and .::1 : ::- ::11 it ;-- -:I to form water

vapor. '::i cavitationn has its drawbacks. It is really hard to control and maneuver a su-

percavitating vehicle. The first part of this work deals with modeling of a :i :7.1:

torpedo. Nonlinear; .:::l : of motion are derived in i. i : i The latter part of the work

deals with finding inner-loop controllers to maneuver the torpedo. A controller is obtained

by ILQR synthesis for pitch and roll rate control. Robust controllers are obtained by i:. ;!

synthesis for tracking pitch and roll angle commands. The robustness analysis of the LQR









controllers is carried out by calculating the: gain and phase margins. Simulations of the:

response for a perturbed system also have been r::-IThe inner-loop controllers are

used for guidance and navigation of the vehicle. It is observed that the pitch and roll angle

controllers can be used for yawv rate control of the vehicle. This is applied to the horning

guidance problem. A simplified case is solved for horning guidance.















CHAPTER 1
INTRODUCTION


Achieving high speeds is an important issue for underwater vehicles. Even the common

fastest underwater vehicles are restricted to travel at speeds around 40 ms-1. The reason

for this restriction is the drag induced by skin friction. When a body moves in a fluid, a

layer of the fluid clings to the surface of the body and is dragged with it. This interaction

causes high drag forces on the body and is commonly termed skin friction drag. The drag

force in water, unlike air, is dominated by skin friction drag as compared to other sources

such as pressure drag. Over the years, extensive research has been done to boost the speed

of underwater vehicles. However, most of the studies were mainly focused on streamlining

the bodies and improving their propulsion systems. Even though these have proven to give

advancements in speed, there has not been a considerable reduction in skin friction drag.

In the late 1970's, the Russian Navy was able to engineer a torpedo, the Shkval (squall) [1],

that achieved a speed over 80 ms-1. This phenomenal improvement in speed was made

possible by supercavitation. The idea was to envelop the torpedo in a gas so that it has little

contact with water. The Shkval was able to achieve a tremendous reduction in skin friction

drag and exhibit high speed.

1.1 Cavitation

As the speed of an underwater vehicles increases, i.e., as the relative velocity of water

with respect to the vehicle increases, the pressure decreases. The speed can be increased

to the point the pressure goes below the vapor pressure of water and then bubbles of water

vapor are formed. This process is known as cavitation. Cavitation is mostly observed at

sharp corners of a body where the speed can reach high magnitudes. A classic example

for cavitation is at the tip of propellers. Since the propeller is rotating at high speed, the









relative velocity at the tips is high enough so that water at the tips vaporizes. Cavitation has

been extensively researched, but remains a challenge for underwater vehicles.

Cavitation can be harmful in many cases. The cavitation region is usually turbulent.

When the cavitation is not steady, the pressure drop is momentary and very quickly the

cavitation bubble encounters a region of high pressure that forces the bubble to collapse

like a tiny bomb. This collapse causes high levels of noise and also may cause considerable

damage to the surface of the body. The cavity behavior is generally discussed in terms

of a dimensionless coefficient called the cavitation number o. This is calculated from

Equation 1.1.
P~ Pc
(1/2)pwYo V3.

where p, is the density of fluid and V. is the speed of the fluid relative to the body. The

pressure in the cavity pc may be that of vapor pv or vapor plus air. This number relates the

difference in the pressure between the cavity and its surroundings to the dynamic pressure

(q, = (1/2)pwV3~) of the free-stream. When cavitation occurs, the cavitation number will

equal the negative of the lowest value of the coefficient of pressure. This value is referred

to the critical cavitation number. This is the value above which cavitation starts to occur in

the fluid.

Supercavitation is a higher stage of cavitation when the body is entirely surrounded

by the cavitation bubble. The cavitation starts as vapor bubbles near the region of high

relative velocity. As the speed is further increased, the bubbles merge to form a large area

of vapor. On further increase in speed, the whole of the body is covered in vapor. This

stage is called the supercavitation. At this point, the condition is similar to traveling in air.

The skin friction drag is tremendously reduced, and thus high speed can be attained in this

phase. Figure 1.1 shows a supercavitating torpedo traveling underwater. It can be seen that

the whole of its body is enveloped by a cavity. This is the kind of vehicle that has been

studied in this work.
















Figure 1.1 Supercavitating Vehicle [1]


1.2 Aspects of Supercavitation

Supercavitation is an extreme version of cavitation in which a single bubble is formed

which envelops the moving object completely. The cavity takes on the shape necessary

to conserve the constant pressure condition on its boundary. The shape is determined by

factors like the body creating it, the cavity pressure and the force of gravity. With slen-

der axisymmetric bodies, supercavities take shape of extended ellipsoids beginning at the

forebody and trailing behind, with the length dependent on the speed of the body [2]. Only

small areas at the nose and on the after-body remain in contact with the liquid.

Expressions for maximum diameter and length of the cavity as functions of cavitator

geometry, drag coefficient and cavitation number are available. The cavitation number and

drag coefficient then determine the cavity geometry. The supercavity may consist com-

pletely of vapor (natural cavity), or have partial non-condensible gas composition (ven-

tilated cavity), which is injected from the nose. The determination of cavitation number

and hence cavity structure in ventilated cavitation is more complicated than for a natural

supercavity [3].

A high speed supercavitating projectile, while moving in the forward direction, may

rotate inside the cavity. This rotation leads to a series of impacts between the proj ectile tail

and the cavity wall, called the tail-slap phenomenon. The impacts affect the trajectory as

well as the stability of motion of the projectile. Despite the impacts with the cavity wall,

the projectile, in certain cases, nearly follows a straight line path. May [4] has a detailed

analysis from experimental data for traj ectories taken by unpowered proj ectiles upon water

entry. Also the patterns of tailslap for such proj ectiles have been discussed.









The wetted forces determine the vehicle speed and vehicle stability. Because the wet-

ted contact area is small, unwanted changes in the wetted contact area arising from local

cavity breakdown may result in large destabilizing forces and moments for a supercavi-

tating vehicle. The wetted forces must be specified and understood to determine vehicle

maneuverability.

1.3 Related Research

Research in the field of supercavitation has been going on from the early 1900's. But

earlier research was aimed at reduction of cavitation so as to reduce noise and body damage.

In the 90's the focus shifted to exploiting the effects of supercavitation.

The work shown in May [4] has an extensive collection of experimental data for vari-

ation of forces on various supercavitating shapes. Coefficients of lift and drag are plotted

with the variation of cavitation number for shapes like disks, cones, ogives and wedges.

The work done in this research makes use of a CFD database provided in Fine [5]. This

database has values for coefficients of lift and drag for conical cavitators, which are func-

tions of the half angle of the cone and the angle of attack. This database also has coefficients

of lift and drag for wedges, which are functions of wetted surface of the wedge, angle of

attack and sweepback angle (we discuss the definition of these geometric quantities such

as the angle of attack and half angle in the later chapters). This information is useful to

calculate the forces on fins of the torpedo.

In late 90's a lot of research was done on the dynamics of supercavitating vehicles.

Work shown in Kulkarni and Pratap [6] and Rand et al. [7] deals with studying dynamics

of uncontrolled supercavitating projectiles. A dynamic model for RAMICS and AHSUM

has been developed. It is shown that these proj ectiles rotate inside the cavity. This rotation

leads to impacts between the tail of the proj ectile and the cavity wall. The frequency of the

impact increases with time. These proj ectiles are very short range and have a small time of

flight, on the order of a few seconds.









The work shown in Dzielski and Kurdila [8] focuses on the formulation of a control

problem for a supercavitating torpedo. A dynamical model for a fin controlled torpedo

has likewise been developed. The model also includes a formulation for the cavity. It is

observed that the weight of the body forces it to skip inside the walls of the cavity and the

vehicle is unstable. A control system is designed for the torpedo and results of closed-loop

simulations have been presented.

1.4 Overview of This Dissertation

This work aims at formulating a control design for a supercavitating torpedo. Equations

of motion for the torpedo are derived. Linear control methodologies, LQR and pu-synthesis,

have been applied to obtain various robust inner-loop controllers for the torpedo.

Chapter 2 briefly describes the configuration of the supercavitating torpedo. The loca-

tion and dimension of the control surfaces have been discussed. A detailed derivation of the

equations of motion for the torpedo has been carried out in the later sections of the Chapter

2. Various reference frames have been used to obtain the kinematic equations of the tor-

pedo. Dynamic equations are derived using Newton's Laws. Various forces experienced

by different regions of the torpedo have been explained.

Chapter 3 describes the two cavity models that have been used in simulations. The cav-

ity contour equations and the cavity kinematics are presented. The hydrodynamic planing

force acts on the hull of the vehicle and generally destabilizes the motion of the vehicle.

The planing force equations and the planing kinematics are presented in the in the second

half of this chapter.

Chapter 4 describes linearization of the equations of motion using small disturbance

theory. It is observed that the linearization, even for a simple trim, straight-level flight, can

be very complicated. Thus, numerical methods are used for this purpose.

Chapter 5 describes the control synthesis for the torpedo. Open-loop dynamics are

shown. The closed-loop problem and various constraints on the torpedo have been stated.









Chapter 6 formulates a Linear Quadratic Regulator (LQR) control design for the tor-

pedo. Controllers for pitch and roll rate control of the torpedo are obtained. The results for

linear closed-loop system and a perturbed liner system have been shown. Various robust-

ness analysis methods show that the controller is robust to numerical errors in coefficients

of lift and drag.

Chapter 7 describes robust control design for the torpedo using pu synthesis. Controllers

for pitch and roll angle have been obtained, that are robust to various parametric and mul-

tiplicative uncertainties. Controllers for pitch and roll rate are also obtained that are robust

to uncertainties in the cavitator.

One of the missions of a missile is target homing. Design of homing controller for a

supercavitating torpedo is discussed in the Chapter 8.

The summary of the current work is presented in the concluding chapter of the disser-

tation, Chapter 9. A discussion of the future work is discussed in the next section of the

thesis.

Appendices A and B describe some of the notations and methods used for derivation

and simulation of the equations of motion.

A description of the CFD database [5] used to obtain numerical values for the code has

been presented in the Appendix C.















CHAPTER 2
NONLINEAR EQUATIONS OF MOTION


Although supercavitation can be a very helpful phenomenon, it presents significant

challenges in modeling and control of supercavitating vehicles. As a significant portion of

the vehicle is located in the cavity, the control, guidance and stability of the vehicle have

to be managed by very small regions in front and aft of the vehicle. Also generation of the

cavity can be a significant problem. The major problems associated with the supercavitat-

ing vehicles may be summarized as

* generation and maintenance of cavity
* balancing the weight of the vehicle
* control and guidance
* stability

Figure 2.1 is a figure of a supercavitating torpedo that is modeled in this work. The main

parts of the torpedo are the cavitator in the front and the four fins in the aft portion. The

cavitator is used to generate and maintain the cavity. The cavitator and the four fins together

are also used for control and stability of the vehicle.





Figure 2.1 Supercavitating Vehicle [8]

2.1 Configuration of the Vehicle
2.1.1 Body

For simulation purposes the body of the torpedo has been designed with 4 separate

sections. The first is cylindrical in shape with a very small radius compared to the main

body. The cavitator pivots at the front of this section and forms the nose of the vehicle.

The second is a section of a cone with a much larger radius than the cavitator, the third a





















Figure 2.2 Vehicle Body is Divided Into 4 Sections.


cylindrical part making up most of the length of the vehicle. The last portion of the vehicle

is the conical section accounting for the nozzle. These geometric components are depicted

in Figure 2.2.

2.1.2 Cavitator

The cavitator is the element that generates a cavity around the torpedo. Several types

of cavitators, including cones, wedges and plates, have been investigated [4]. This project

will consider a conical cavitator as shown in Figure 2.2. The main parameter that defines

a conical cavitator is its half-angle. The coefficients of lift and drag, as functions of half-

angle, for the cavitator have been generated using CFD and tabulated in Fine [5].

The cavitator in this model has one degree of freedom defined by a rotation angle about

an axis perpendicular to its axis of symmetry. The shape and location of the cavity are

a nonlinear function of this cavitator deflection angle and half angle of the cone. This

shape determines the position where the cavity intersects the body of the vehicle. Thus, the

amount of wetted area of the vehicle is dependent on these angles, which in turn determines

the efficiency of supercavitation achieved. An effort has also been made to design a 2DOF

cavitator.

As stated earlier, a large portion of the vehicle is in the cavity. The lift produced by the

cavitator combined with the lift produced by the fins helps in balancing the weight of the

body.









2.1.3 Fins

Although the cavitator is capable of providing enough lift to sustain the body in water,

it does not usually provide sufficient forces and moments to stabilize and control the ve-

hicle. For this purpose fins are required in the aft portion of the vehicle. The fins help in

counteracting the moments produced by the cavitator and also providing some amount of

lift to support the weight of the body. There are four fins placed symmetrically along the

girth of the vehicle, near the tail. The fins also are the control surfaces, as they can provide

differential forces, thus causing control moments. Two of the fins shown in Figure 2.3 are

horizontal (placed parallel to the axis of rotation of cavitator), called elevators and are used

to affect the longitudinal dynamics for the vehicle. The other two fins are called the rudders

and are used to affect the lateral dynamics to the vehicle.

The fins have two degrees of freedom, a sweepback angle and an angle of rotation.

These angles will be explained in later sections of this chapter.







Figure 2.3 Cavitator and Fins
2.1.4 Maneuvering

The motion of the vehicle is controlled by all five control surfaces, the four fins and the

cavitator. In a straight line motion the cavitator and the elevators balance the weight of the

vehicle. A propulsion force at the tail pushes the vehicle forward. The rudders usually have

a zero deflection in such a case.

The vehicle depends on a bank-to-turn strategy for making a turn, and cannot use tradi-

tional missile strategies such as skid-to-turn. This dependence arises because the hull of the

vehicle is incapable of providing a lift force. The fins are the main lift generating surfaces.

A differential force from the fins can be used to generate a force towards the center of the

turn.









2.2 Kinematic Equations of Motion

The definition of a suitable coordinate system and degrees of freedom is a precursor

to any formulation of equations of motion. The derivation of the equations of motion of

multi-body systems is highly simplified by defining various reference frames and relations

between them. Appendix A describes briefly the concept of reference frames and rotation

matrices. These concepts will be used extensively in the derivation of equations of motion.

The derivation of the equations of motion will be done in two parts. First, the kinematic

equations will be derived. These include the formulation of the position vectors, velocities

and accelerations of various points on the torpedo. Next, the dynamic equations will be

derived. Finally, Newton's Laws yield the dynamic equations of motion.

2.2.1 Orientation of the Torpedo

Six degrees of freedom (DOF) are required to describe the position and orientation of

the torpedo when it is considered a rigid body. Of these, three DOFs give the location of a

point fixed on the torpedo. The other 3 DOFs give the orientation of the torpedo in a fixed

reference frame. The position of the torpedo in a reference frame can also be obtained by

the integration of its velocity in that reference frame.

The torpedo is assumed to be moving in an earth-fixed reference frame E, centered at

any conveniently chosen point and described by the basis vectors (edi, e^2, e^3). The earth-fixed

reference frame is shown in Figure 2.4. The vector ^3 points in the downward direction,

i.e., the direction of the gravity. The vectors e01 and e^2 are placed in the horizontal plane

such that the basis vectors form a right-handed coordinate system. As shown in the figure,

el points to the right and e^2 points outside the plane of the paper. A body-fixed frame B is

defined by the basis vectors (jby, j2, 3 ~) so as to simplify the derivation of the equations of

motion. The frame B is centered at O, the center of gravity of the torpedo, and moves with

the torpedo. The reference frame B is shown in the Figure 2.4. It can be seen in Figure

2.5 that the torpedo has two planes of symmetry namely fil2 and fil3. The plane $183 iS

called the longitudinal plane and plane $182, the lateral plane.






































Figure 2.5 Principal Planes of Symmetry for the Torpedo


Figure 2.6 Euler Angles of Rotation


i: ex

e3s
Figure 2.4 Body-Fixed and Inertial Frames









Transformation from E to B can be achieved by 3 rotations. Many such sets of rotations

are possible. The rotation steps chosen here are the Euler angles of rotation, which are the

yaw, pitch and roll. As there are three rotations to be performed, two intermediate reference
frames with basis vectors (1, 2, 3) and (fi1, 2, 3) will have to be introduced to perform

the transformation. The rotations, to be performed in order, are listed below.

1. Rotate the frame E about ^3 through a yaw angle, Y, to obtain the frame (1, 2,23).
2. Rotate (1, 2, 3) about 2 through a pitch angle, 8, to obtain the frame (fi, f2, 3)
3. Rotate (ftl,f2, 3) about ft through a roll angle, 4, to obtain the frame B.

Figure 2.6 shows the above rotations in order. Based on the above definition of rotations,
the transformation matrix from E to B can be written as in equation 2.1i.


CO 0 -SO

0 10 -

SO 0 CO

CGSY

SAS8SY +t CYCO

CAS8SY CY SO


S

0


YSf

~Y CY

0

-SG

SOC8

CCOC8


1 0 0

0 A O

0 -SO CA

COCY

CYSSOS CASY

CASOCY +t SOSY



E _B e82

e^3


(2.1)


2.2.2 Orientation of the Cavitator

As described earlier, the cavitator has only one degree of freedom. It can rotate in the

longitudinal plane about an axis parallel to the b2 axis. The orientation of the cavitator-
fixed axes with respect to the body fixed axes is shown in Figure 2.7. The deflection of the

cavitator is given by an angle, Sc, which is positive for a positive cavitator rotation about

the b2 direction. The geometric center of rotation of cavitator is denoted by P. CP is the

center of pressure for the cavitator, which is at a distance ACP from P, along St.


1 e3



dI
e^3




















b3





Figure 2.7 Cavitator Reference Frame

From Figure 2.7, the rotation matrix from B to cavitator fixed frame C can be written

as in Equation 2.2.
di C6c 0 -S6c bi(22
do = 0 1 0 b22

cl8c3 S6c 0 C6c b

2.2.3 Orientation of Fins

Figure 2.8 shows the orientation of the fin-fixed reference frames. For convenience, all
the fin frames are indicated by basis vectors (fi,}2}). In text they will be represented as

(fi,f2,f3) ps, where subscript fin corresponds to a particular fin.
All the fins have 2 DOFs, namely the sweepback angle (8fi,) and the fin rotation (6ip,).
These can be defined as

*Sweepback angle (6 ps,): This parameter is the angle of rotation of a fin about its f3
axis. The direction of rotation for positive sweepback is such that the fin is moved
toward the rear portion of the torpedo. Due to this definition, the positive sweepback
is along the negative f3 direction for all fins. Sweepback angle determines the amount
of fin that is enveloped in the cavity.










Ruddei 1)









Ruddel 2)



FRONT VIEW








Elevator 2 )





TOP VIEW




Figure 2.8 Rudder and Fin Reference Frames


Fin Rotation (6r;,): This parameter is the angle of rotation of the fin about its f2 axis,
in positive the f2 direction. Fin rotation determines the magnitude and direction of
the forces acting on the fins.

The order of rotation in the above case is important to obtain the correct equations.

Sweepback has to be performed before fin rotation. An intermediate reference frame G

with basis vectors (gl,Ag2,3) is introduced so as to simplify the derivation of rotation ma-

trix from B to the fin coordinates. Sweepback aligns the fin-fixed frames with the interme-

diate frame G and then a fin rotation puts the fin frame in actual orientation with the fins.

As the second rotation is identical in all cases, the transformation matrix from frame G to

fin frame F piz can be written as in Equation 2.3.








gzfi
f

It.


t)R1


~3 A


Figure 2.9 Rudder 1 Fin Reference Frames


(2.3)


The rotation matrix for sweepback and the transformation matrices from B to Ffin frame
for each of the fins can be derived easily, and are summarized below.

*Rudder 1 Figure 2.9 shows the sweepback and fin rotation for rudder 1. The matrices
for transformation from B to R1 can be derived as in Equation 2.4 and Equation 2.5.


~1
2

-ceR1
-S6eR1
0


-C6R1
-S6R1
0


S6eR1
-ceR1
0


(2.4)


(2.5)


CiR1
0
SSR1


S6R1
-ceR1
0


*Rudder 2 Figure 2.10 shows the sweepback and fin rotation for rudder 2. The
matrices for transformation from B to R2 can be derived as in Equation 2.6 and
Equation 2.7.


(2.6)


(2.7)


~1

2 eR
3se
- R2


-SeR2
ceR2
0


ftCS~fin


0 -S~fin g1

1 0 g2
0 CS fin g3 fi


#2
I 1 3 R1

}2
I fl 73 R1


1
2
$3


-SSR1
0
C6R 1 _


g2 =-S6R2 0 C6R2 -R20SR2
g3 R2 1 0

ft C8R2 0 -SSR2
)3 R2 SSR2 0 C8R2 _


1
2
$3























Figure 2. 10 Rudder 2 Fin Reference Frames


Elevator 1 Figure 2.11 shows the sweepback and fin rotation for Elevator 1. The
matrices for transformation from B to El can be derived as in Equation 2.8 and
Equation 2.9.


g2 =-S6eEl C6El 0 E (2.8)
g3 El 0 0 -1 3
CEl 0 -SSE1 CE S6l0 b
S-S6eEl C8El 0 (2.9)
3~~~ El -ce S8E 0sE CSE 0 0 -1

Elevator 2 Figure 2.12 shows the sweepback and fin rotation for Elevator 2. The
matrices for transformation from B to E2 can be derived as in Equation 2.10 and
Equation 2. 11.


g2 =-S6E2 -C6E2 0 8, (2.10)
g3 E2 0 0 1 8
I ft CSE2 0 -SSE2 CE SE2 0 b
}2 0 1 0 -SOE2 -C8E2 O Lo(2.11)
73 E2 S8E2 0 CSE2 1 8

Equations 2.2 to 2.11 will be used in later sections to transform the forces on fins and
cavitator to the body-fixed frame.

2.2.4 Angle of Attack and Sideslip

The body-fixed reference frame has been defined, but the velocity of various points on

the body of the torpedo is yet to be defined. The torpedo is considered as a rigid body. If the

velocity of a certain point on a rigid body is known, the velocity at any other point on that














43,Z gR1z1









I~

II





eE2 ~E2
22






Figure 2. 12 Elevator 2 Fin Reference Frames


body can be found by knowing the rotation matrices. Thus, V = ubl t +vb2 + wb3 will be

taken as the velocity of CG of the torpedo. The velocity of other points on the torpedo can

be defined subsequently. Two very useful parameters, angle of attack and angle of sideslip

can be defined in conjunction with the orientation of the body axis with the velocity of CG.

Figure 2. 13 shows these parameters and their geometric interpretation.
Angle of attack, a, can be defined as the angle between the projection of velocity V

onto $2 3 plane and the $1 axis. Angle of attack is positive when the nose of the torpedo

points above the velocity vector of the torpedo. As before, an intermediate frame F given

by (fi, f2, 3) can be described by rotation of the B frame by an angle a. Thus the rotation




















Figure 2. 13 Angle of Attack (u) and Sideslip (P)

matrix from Fbody to B can be written.

by Ca 0 -Sa


b3 Sa 0 Ca f3


The angle of sideslip, P, is defined as the angle between the actual velocity V and

the projection of VT onto babJ3 plane. Again, a frame Gbody can be defined by rotation of

Fbody by an angle equal to P in negative f3 direction, thus giving a rotation matrix as in

Equation 2. 13.
1l Cp -Sp 0 f


2 = s cp o (2.13)


Velocity of CG of the torpedo in the Gbody frame can be written as VAl, where V is

magnitude of V. It will be seen that drag and lift on the torpedo can be obtained in this
frame. Thus a transformation from Gbody to B is important. It is given by Equation 2.14.

by CpCa CuSp -Sa g1


by CpSa SuSp Ca g3
body (.4

Using Equation 2.14 V can be rewritten as in Equation 2.15.









V = Vit

= VC~ubiV~pb+VCBub3(2.15)


where V2 = V2 = 2 -t 2 -2. From Figure 2. 13, relations between the velocity compo-

nents and the angles of attack and sideslip can be derived.

tan a = (2.16)
-v
sin 0 = (2.17)

Though the matrix Gbody-B in Equation 2.14 has been defined for the body-fixed ref-

erence frame and velocity of CG of the torpedo, the equation is valid for any other rigid

part of the body like the fins and cavitator. Thus, in case of a fin, the velocity V would

correspond to a point (like the tip, center of pressure etc.) on that fin, and GfinB matrix

would correspond to the fin-fixed reference frame. In this case the velocity of center of

pressure of the fin will be used to define the above parameters. Thus, obtaining afin and

Pfin is a two step process:

1. Obtain the velocity of center of pressure of fin.

VCl'body = Veg +tE 0B X TEcgCP (2.18)

where VCPbodv is velocity of CP of fin in B frame, Veg is the velocity of CG of the
torpedo in E frame, E B is angular velocity of B in E, and regCP is position vector
from CG to CPfin. Equation 2. 18 can be rewritten as in Equation 2. 19.

fufin u by b2 b3
vfin v + p r (2.19)
Wfin B c Xfin Yfin Zfin

where regCP = Xfinby + Yfninb2 + Zfinb)3

2. Transform the velocity (in E) of CP of fin from frame B to frame of corresponding
fin. This transformation is obtained by using rotation matrices derived in Equations
2.3 to 2. 11.









MR1 ),[R1 0 -SSR CR1 0 S6~R1 VR1
vR1 0 R1 0 -S6R CR1 0 -C6R1 UR1 (.0
wR1 SSR1 0 C8R1 0 -1 0 wR1
SuR2 C6R2 0 -SSR2 -COR2 0 -S6eR2 UR2

PR20 0 -SR2 CR2 PR (2.21)
al :R2-SSR2 0 C8R2 0 1 0 O

vEl ),[0 1 0 -S6El C8El 0 vEl2.2
wEl S8El 0 CSE1 0 0 -1 wEl

( E2E2 2h 00 -SSE2-CE2 S6~E2 0 uE2
vE2 0 12 0 -S6E2 -C6E lE2 0 vE2 (.3

wEE2 2-S8E2 0 CSE2 __ 0 0 1 wE2 B

The left hand terms in Equations 2.20 to 2.23 give the velocity components at the CP

for corresponding fins, in that fin frame. These can be used in Equations 2.16 and 2.17 to

find the angle of attack and sideslip for a particular fin.

2.2.5 Kinematic Equations

Velocity of the CG of the torpedo has been defined in the previous section. That velocity

defines the translational kinematics for the torpedo. Analogous to this quantity, angular

velocity is required to define the rotational kinematics. The angular velocity of the hull has

components p, q and r in the frame B.

E B API 1t qi2 -t r3 (2.24)


As the transformation from E to B has already been defined in terms of rotational motions

give by Euler angles, the angular rates can also be obtained by differentiation of Euler

angles. Thus, another form of Equation 2.24 can be written as in Equation 2.25.

E B = e3 -t X2 1t~ (2.25)

The rotation matrices from Equation 2.1 can be substituted into Equation 2.25 to obtain

Equation 2.26.


Eo B= (@ SOYG )bi+( C8SQ+8t CO)&2+( CCOC~-8SA)&3


(2.26)








Equations 2.24 and 2.26 can be equated to obtain Equation 2.27.


p -SO 0 1


Pr C8CO -S~O0


Equation 2.27 can be rewritten as in Equation 2.28.

p = 0- SOY

q = YC8SQ +t OCO (2.28)
r = YC8CO 8SO

To apply Newton's Laws, acceleration of the CG is required. The values of pi, q, r
will be the angular accelerations of torpedo in B. The translational acceleration can be

calculated by time differentiation of V in Newtonian frame. A differentiation formula can

be used to find the time derivative, in some frame, for a vector defined in some other related
frame.
d. d..
(v)l -(v) +' oB xv (2.29)
dt \,I dt \/B

where, sub script I denotes Newtonian (inertial) frame, and B is the body-fixed frame. lo3

is angular velocity of the body (or body-fixed frame) in the I frame, v is the velocity in

I frame, of the point where acceleration is desired. Using the formula the acceleration of

CM of torpedo in E can be obtained.

u, 8, 82
E~ ~ CM = (2.30)





= -+ ur pw (.1

wi +t pv uq








Similarly, the rotational acceleration will be required in the frame E. This can be written
analogous to Equation 2.30.





(2.32)

Pr



The position of torpedo can be found by integrating the velocity. Let (x,y,z) represent
the coordinates of CG in frame E. Thus, the time derivative of these coordinates in E should

represent the velocity components of the torpedo in E frame. When these time derivatives
are transformed to body-fixed frame, they would be equivalent to the velocity components
in body-fixed frame.



yx =( wv (2.33)


Equation 2.1 can be substituted in Equation 2.33 to obtain Equation 2.34.

x COCY C89P -SO

y = CYSSOS8- CSY SAS8SY +CYCO SCOC8 v (.4
zx CASOCY+SOSY CASOSY SOCY COC8 w

2.3 Dynamic Equations of Motion

Now that the accelerations of various parts of the torpedo are known, Newton's Laws
can be used to derive the dynamic equations of motion. Newton's laws state that the rate of

change of momentum is equal to the sum of external force applied on the body. Equation
2.35 can be obtained from Newton's laws by an assumption that the mass of the torpedo is
constant. This assumption is valid for a short period of time. The equations are








FP=P
(2.35)
= ma

where P is the linear momentum, m is mass of the body, a is the acceleration and F is all the

forces of the body. Using Equation 2.3 1 in Equation 2.35, Newton's Laws for the torpedo

can be rewritten as in Equation 2.36.



m vt ur pw =F(.6

w +t pv uq
;I u-t q v =F 2.3b

Newton's laws can be extended to rotation. Equation 2.37 are the Newton's Laws for
rotational motion.

M c= II
(2.37)
= Icnnu+tE 0B x H

where H (= IcMnE B) is the angular momentum, IcMn is moment of inertia matrix of the

body, a is the angular acceleration and M is all the moments on the body. It should be
noted that above stated Newton's laws are only valid when the quantities are calculated in
an inertial frame of reference. Thus, the quantities have been calculated in frame E. Using

Equation 2.32, the Newton's Law for rotation can be written as in Equation 2.38.


li0 0 pi by b2 b3

0 I2 0 q + p r = M (2.38)

0 I3 7 P 293

To derive the equations, the forces on each of the parts will be calculated individually,
and then expressed in body-fixed reference frame, i.e., summation will be done in body

reference frame. The rotation matrices derived in previous sections will be used extensively
for this purpose.









Various types of forces are experienced by a moving torpedo in water. These forces can

be summarized as follows:

* Hydrodynamic Forces: These are the forces exerted by the fluid on the torpedo.
Thus they exist whenever the fluid is in contact with body. For supercavitating ve-
hicle, most of the body (hull) is inside the cavity. Only a portion of the fins and the
cavitator are in contact with the fluid. Thus the lift and drag on cavitator and fins are
only hydrodynamic forces. The coefficients of lift and drag for the fins and cavitator
are functions of their angle of attack, immersion, sweepback angle, angle of rotation
etc. and are tabulated in a database [5]. This database will be interpolated and extrap-
olated to calculate the numerical values for the forces. When the body of the vehicle
other than the fins or the cavitator comes into contact with the cavity wall, the body
experiences forces called as the planing forces during contact with the cavity wall.
Details of these kind of hydrodynamic forces will be dealt with in Chapter 3.

FHydrodynantic = FR1 -tFR2 -tFEl FE2 ec- Fplaning (2.39)
MHydrodynantic = MR1 -tMR2 -tME1 -tME2 -tMc Mplaning (2.40)

Gravitational Forces: This is the gravity forces on the body. As the summation of
moments will be taken about the center of gravity, the gravitational forces will not
contribute to the summation on moments. They will appear only in summation of
forces.

Propulsive: The torpedo has a propulsion system, which is similar to that of rockets.
The line of action of the propulsive force is assumed to be passing through center of
gravity and along byl axis. Thus this force will also contribute to the sum of forces,
and not moments. The propulsive forces are provided by burning the fuel, but for
simplicity it will be assumed that the mass of the torpedo remains constant.

Immersion: This force results from the vehicle being submerged underwater. Since
the center of buoyancy is assumed to be coincident with the center of gravity this
force contributes only to the external force term.

The total forces and moments are expressed in terms of these components.

F = FHydrodynantic -t Gray FProp -tFlninersion (2.41)

M = MHydrodynantic (2.42)

2.3.1 Forces on Cavitator

Figure 2.14 shows the forces acting on the cavitator. Coefficient of lift (clc) and drag

(cdc) for the cavitator are functions of angle of attack, occ, and half-angle, yl, of the cavi-

tator. Half-angle, for a cone, is the angle made by axis of the cone with any line passing









through the vertex and lying in the surface of the the cone. This parameter defines the main

geometry of the conical cavitator. The values of cle and cde are determined using CFD and

tabulated in [5]. These values have been extrapolated to calculate lift and drag.


Le =2pif Secl,(ae,7 y) (2.43)

De =~2plSecd (ae, y ) (2.44)

where Sc is the cross-sectional area of the cavitator base. Directions of lift (Le) and drag

(Dc) are as shown in Figure 2.14(b). These can be transformed to the body axes using 2.2
and 2. 14 for the cavitator.





b1, = C_B(sc) x Gcav_C(ue,,i) x gl82, (2.45



Thus the forces on cavitator, in body frame, can be written.



Faxv






= (2.46)

Le (Rc,7 4)
SGear

C6c 0 S6c cpc~ac CcaSpc -SW Dc(c

= 1 0 -Spc Ocp 0 0t

-S6c 0 C6c OcpSue SucSic Cac -Le(Rc,7 4)

where Fe is a 3xl matrix with each row corresponding to each direction in B basis. The

forces are assumed to be acting at CP of the cavitator. Once the forces have been calculated,

the moment about any point can be calculated.








Me = rCPcar X Fec


(2.47)


where rCPcar is the position vector from that point to CP of cavitator. It is assumed that the
CP lies on by when cavitator deflection is 0, and its coordinates with respect to body origin

O, in this case, are (Xc, 0, 0). Thus from Figure 2.7, the coordinates of CP can be written
for the case when the cavitator is deflected.


rCPcar =


bodv


x c +PCGc
0


(2.48)


The moments on the cavitator in body-fixed can be obtained by substituting Equations 2.46
and 2.48 in Equation 2.47.


Me = [(Xc t kPC~c~j by -kPSrCcb3] x


(2.49)


Ccpuc ,
-Sic
C~cSuc


CcaSic
Oc
SucSic


Le/ De c

(a) (b)

Figure 2. 14 Cavitator: (a) Angle of Attack and Sideslip and (b) Hydrodynamic Forces


C6c 0 S6c
0 10

-S6c 0 C6c

-Suc -Dc (Orc. y)
0 0(

Cac -Le (acG 4y~)









L Lan




S a ha







Figure 2. 15 Fin Geometry

2.3.2 Forces on Fins

Fin forces are also extrapolated from the CFD database [5], which gives the values of

coefficients of lift (clf;,) and drag (cdf;,) for fins. These values are functions of angle

of attack afin, fin sweepback 8fi;,, fin rotation 6f;,, fin immersion Ii,, and fin geometry.

Figure 2. 15 shows these parameters graphically, and they can be defined as follows:

* Fin Geometry: The geometry parameters for fins are L and S, and wedge half angle
(ha), as shown in Figure 2.15. These parameters are fixed according to the values
given by the database, so as to calculate the forces accurately.

Fin Immersion: As the fin is partially wetted by fluid, the wetted length can be rep-
resented by parameter So along fin Y-axis. The immersion If,, can be defined as the
ratio of the wetted length of the fin to its true length.
If,, = (So/S) p;, (2.50)
Ip;, determines the total hydrodynamic force on the fin.

Fin Rotation (87;;,): As defined earlier, this is rotation about fin f2 axis. This deter-
mines the direction of the hydrodynamic force. Thus fin rotation is used as primary
control for the torpedo.

Fin Sweepback (8f;;;): As defined earlier, this is rotation about fin f3 axis. Sweepback
determines the wetted surface of the fin, thus the hydrodynamic force on the fin.

Angle of Attack: Angle of attack for the fin is calculated as described in Figure 2. 15
and section 2.2.4.

The database gives clfi, and cdfi, as a function of my;,,, 8p,, and Ip,,, thus lift and drag on

the fins can be calculated by the normalizing factor.


Lan = ~p2 pVS inc (2.51)
1
Din = 2pV S injcdpin (2.52)









Where Sfin is the length of the fin as shown in Figure 2. 15 These forces have directions

as shown in Figure 2.15. Before substituting in Equation 2.39, these forces have to be

transformed to body-fixed reference frame. This process involves following two rotations:

1. Rotate the frame Ffin (which has Lan and Dfin along its basis vectors) to align it with
the fin-fixed frame using Equation 2.14 and

2. Rotate the above obtained fin-fixed frame to obtain the body-fixed frame using Equa-
tions 2.3 to 2. 11.

Thus the forces on the fins in body-fixed frame axis can be obtained.

*Rudder 1


-ceR1 -S6R1 0
0 0 1
S6R1 -ceR1 0 _
cRIcaR1 caRISPR1
-SpR1 c R1
cRISaR1 SaRISPR1



-ceR2 -S6R2 0
0 0 1
-S6R2 C6R2 0 _
CPR2CtR2 CaR2SPR2
-SBR2 c R2
CPR2SaR2 SaR2SPR2


C8R1
0
_-SSR1
-SaR1
0
CaR1 _



CsR2
0
-SSR2
-SaR2
0
CaR2 _


(2.53)









(2.54)


CSE1
0
-SSE1
-SaE1
0
CaE1


-COEl
-S6El c
0
E10tcal
-SBE1
CE1SaE1


S
8
0


(2.55)


FR1~






* Rudder 2

FR2,x







* Elevator 1

(FE1,x


0 SSR1
1 0
0 C8R11

-DR1-L1





0 SSR2
1 0
0 C8R2
-DR2
0
-LR2


6El 0
El 0


c El PE
SaEE1SE1


0 S8El
1 0
0 CSE1

-DE








*Elevator 2

FE2,x -C6E2 -S6E2 0 CSE2 0 S8E2
FE2L, S6eE2 8E2
FE2,z B 0 0 1 -SSE2 0 CSE2
( ) (2.56)
-SB2CE2 CESE2 0SE 0DE
CE2SaE2 SaE2SPE2 CaE2_ -LE2

Equations 2.53 to 2.56 give the components of hydrodynamic forces on fins in body-fixed
frame. What remains is to find the moment of these forces about CG of body. The moments
can be obtained in analogous to Equation 2.47.

Mfin =r nG-CP x Ffin (2.57)

In this case, the root of fins is fixed to body, and it can move thus moving the CP of fin
relative to root. The position of CP from root is also know with respect to fin coordinates.

r nG-~root t, 1' ; 2 + Z$1 t3 (2.58)

r ot-P ; 7 2(2.59)

where (ft,}2,}) is fin-fixed coordinates for corresponding fin. Equations 2.58 and 2.59

can be added by using matrices given in Equations 2.3 to 2.11. Thus, the position vector
from CG to CP of fins can be obtained.

Rudder 1

XR1 ~ Yiot 8R1 -S6R1 0
YR 1 oo'3tR10 0
ZR1 B Z ,ot B -S6R1 -ceR1 0
(2.60)
C8R1 0 SSR1 ~:
0 1 0 Ayo
-SSR1 0 C8R1 0 R

Rudder 2
XR2 X coPt -C6R2 -S6R2 0
YR2 = 7 ootR -
ZR2 B Z~o~t B --S6R2 ceR2 0
C8R2 0 SSR2 2XC ~ (.1
0 1 0 Ay(
-SSR2 0 C8R20











_ceEl
-S6El

CSE1 0
0 1
-SSE1 0


-S6El
cE1



CSE1 _


(2.62)


YE2 = root +t S6E EX2X' CE2 -S6E2 0
ZE2 BZrroot B-0 0 1
(2.63)
CSE2 0 S8E2 ~ E~
0 10O
-SSE2 0 CSE20

Equations 2.60 to 2.63 give the position vector from CG to CP of the fins. These equa-

tions in conjunction with Equations 2.53 to 2.56, used in 2.57, gives the external moments

on fins about the CG.


Man, =


by

Xfin

Ffin ,x


Ffin~v


Zfin

Ffin ,z


(2.64)


2.3.3 Gravitational Forces

For simplicity, mass (m) of the torpedo is assumed to be constant over time. This

is not the case in reality but the approximation is reasonable for considering short time

maneuvers. Acceleration due to gravity, g, is also assumed to be constant as torpedo moves

in space. The direction of gravity is given by e^3 axis. Thus, the gravitational force can be

written as in Equation 2.65.


Fgray = mg^3


(2.65)


* Elevator 1


SXEl ot El = 7 oot + -X~O
ZE1 Zroot B





* Elevator 2


0

-1


0 El








Equation 2.65 can be re-expressed in body frame of reference using Equation 2. 1.

COCY CGSY -SO

F gray = CY ~SOS8- CSY SASOSY +CCOC SCOC 0g
CASOCY +SOSY CAS8SY SCY CIOC8 m
E(2.66)

= mg -SOC8
SCOC


2.3.4 Equations of Motion
Now that a mathematical formulation of forces on the torpedo has been achieved, these

equations can be substituted into Equations 2.36 to 2.42 to obtain the dynamic equations of
motion. Thus the force equations of motion can be summarized as in Equation 2.67.
2.3.4.1 Force equations

ui +tqw vr -prop FR1,x
m vt ur pw = imrinFlnn+ 0 + FR1,y

w + tpv -uq 0FR1,z

FR2,x FE1,x FE2 ,x -SO
FR2,y + FEyI +( FE~Sy +mg SOC8O + (2.67)
FR 2 ,z FE 1, z FE 2 z CICO

C6c 0 S6c Ccpuca CucSic -Suc -Dc(ac,T y)
0 1 0 -Spc Ccp 0 0t

-S6c 0 C6c C~cSuc SucSic Cac -Le(ac,71)
2C
Some issues should be noted for Equation 2.67.

*The planing forces have not yet been calculated in the equations of motion. The
calculation of these forces will be explained in Chapter 3. In the case of fixed cavity
model for the torpedo, these forces are neglected by assumption that the vehicle is
centered in the cavity.





















































COSY -SG

CASf SAS8SY +CYCO SCOC8

SOSY CAS8SY SOCY COC8


(2.70)


*The propulsion force is constrained to be along negative by axis.

2.3.4.2 Moment equations

li0 0 p1 by~ bg3 by bg b3

0 IO qr +p qr = Mplaning+t XR1 YR1 ZR1

0 0 Is r I p Igq I3r FR1,x FR1,y FR1,z

by bg b3 by bg b3

XR2 YR2 ZR2 + XEl YEl ZE1 + (2.68)

FR2, FRy F2,z FE1,x FElyI FE1,z

by bg b3 by bg bi3

XE2 YE2 ZE2 +Xc +t AcCPC, 0 -ACPSGc

FE2x F2, FEzFc,x Fcy Fc,z

Some issues should be noted for Equation 2.68.

Some of the terms in Equation 2.68 are shown as a determinant. They need to be
expanded and written as components in body-fixed frame so as to equate the left-
hand and right-hand terms.

Moments due to gravitation do not appear because the moments are taken about CG.

Again, the moment due to planing forces have not yet been explained. Also, the
moment due to planing forces is to be neglected in the case of fixed cavity model.

2.3.4.3 Orientation equations


O O pocc
e =~S 0 ~ C -SO q~ (2.69)


2.3.4.4 Navigation equations

x COCY

y =I CYSSOS8-
i CASOCY +t


u

v









Equations 2.67 to 2.70 are a complete set of equations of motions for the supercavitating

torpedo.















CHAPTER 3
CAVITY AND PLANING DYNAMIC S


If the body is immersed in a steady flow, it is reasonable to assume that the shape of the

cavity is constant and axially symmetric. It may not, however, be so as cavities may pulsate

due to re-entrant jets located at the aft ends of a vehicle [4]. The cavity shape is generally

specified by giving the length L and the maximum diameter d,, of the cavity, or the fineness

ratio L/d,,. The diameter is specified as function of the arc length along the centerline of

the cavity. Various models, analytical, empirical and numerical, have been developed over

the years to determine the above mentioned parameters for a cavity. Waid [9] derived

formulae for the cavity shapes of disks from measurements made on the cavities. May [4]

lists various cavity models based on experimental data for geometries such as disks, cones,

spheres, etc. Many models exist for supercavity behavior based on slender-body theory,

boundary-element methods, and Reynolds-averaged Navier-Stokes solvers. However, for

time-domain simulation of vehicle dynamics, it is desirable to use simple models. This is

particularly the case when we study the guidance, control and stability of the overall body.

Two cavity models have been used in the simulations. They are the Munzer-reichardt

cavity model and the Logvinovich cavity model. Both cavity models assume the cross-

section of the cavity as being circular in shape. The radius of the cavity is specified as

a function of the arc length along the cavity centerline. The modeling of the cavity, for

simulation purposes, comprises of these cavity models and the description of the cavity

centerline based on memory effects.









3.1 Munzer-Reichardt Model

Reichardt showed that the fineness ratio of the cavity L/dinax is a function only of the

cavitation number o. In other words, the ratio is independent of the cavitator shape and

is only function of o. The formulae for the cavity dimensions according to the Munzer-

Reichardt [10] model are


R = R,L, 1


Lazx 2 ,, [124 -0. (3.1)

R,, = RcClcdel.350-.9

In the Equations 3.1 Lza is the length of the cavity, L is arc length along the cavity center-

line, Cd, is the cavitator drag coefficient, RC is the radius of the cavitator and R the cavity

radius at that arc length on the cavity centerline.

3.2 Logvinovich Cavity Model

3.2.1 Logvinovich Theory of Independent Expansion

Logvinovich has made the following fundamental observation: "Each cross-section of

the cavity expands relative to the path of the body-center almost independently of the sub-

sequent or preceding motion of the body..." [1 l].

Logvinovich's cavity contour equations are another example of the kind which consider

the cavity to be in the shape of an ellipsoid. The following formulas are taken from Logvi-

novich [11] for the cavity shape


R = Rk RR 1 2k1 1--tk


RR2\ 1 t\ t
R = k 1 1--I 1-- (3 .2)
R, Rg i tkj tk

In the Equations 3.2 Rk denotes the maximum cavity radius, R and Rt, denote cavity

radius at a particular point in time on the cavity centerline, t time taken for cavity at that

point to evolve from inception, tk is the time taken by cavitator to travel one body length









of the vehicle, K is a correction factor and R is the time rate of change of the cavity radius.

Explicit assumption is made of the cavitator being of disk shaped in deriving the above

formulae. However, in our case the cavity due to a cone shaped cavitator is approximated

with the one due to a disk shaped cavitator. The formulae in the above form make it

possible to calculate the cavity profile also when t/tk > 1, but usually when t/tk > 1.5 the

boundaries of the cavity are undeterminable. They start breaking up and foam begins to

form. At t = 0 the contour expressed by Equation 3.2 is matched with that of the leading

part of the empirical cavity 3x
R- = n (3.3)


where Rn is the cavitator diameter. The maximum cavity radius Rk can be calculated

using the formula 1=, +08 i-o(3


By selecting xl = 2Rn as our matching point we obtain R1 = 1.92Rn. A cavity "correction

factor", K, is also required in the calculations. For the selection xl = 2Rn and K = 0.85,

a good correlation can be obtained between the formulae in Equation 4.2 and the experi-

mental data. Under these conditions the cavity half length can be approximated using the

formula in Equation 3.5 [l l]. L I (1.923.5


The cavity half length will be used to approximate the value of t/tk as L/Lk. The validity

of this approximation can be seen from Equation 3.10. From the above formulae we can

calculate the basic dimensions of the cavity. However, during simulation the coordinates

of cavity centerline need to be located. The centerline is required in order to determine the

final shape and position of the cavity with respect to the body. Effects like buoyancy and

downwa~sh need to be considered for determining the final shape of cavity centerline.

3.2.2 Cavity Centerline

The arc length of the actual path is assumed to be approximately equal to the distance

from the cavitator to the planing point projected onto the body axis. The time lag in con-

sidering the memory effects shaping the cavity can be obtained by the expression









z = L (3.6)


Where, L is the distance along the cavity centerline. The above value is obtained

through a 2tad order approximation



= u~) +u~t)- ut~r(3.7)


The average velocity over the time period [tl t z] multiplied by the time period will

give the distance covered, L, in time t. The distance L can be written as

L= u -t- &TT--&T+u-=0(38

Now, solve the qluadratic equation for t,

-ui u2 4(- ) (L) ui V'u2 2u
I = (3.9r)

Taking the limit of ii approaching zero and using the L'Hospital rule for calculating

limits for indeterminate forms we have

lim I=4 (3.10)


iFromn one y... ..t of view, the variation in the cavity shape at a point in space is due to the

memory effects associated with the i : :- -- of cavitator through that point at a i :; T us

time. Thus to calculate the cavity i the cavitator position at each time-step is stored.

The cavity centerline is formed by the i : .I II :: of the cavitator tip at current time and the

previous (n-1) time i i where 'n' is the number ofl 1...1= on the full length of the cavity.
We can write


yi = ye ( i- 1At)(3.11)

yizi z 1 i ) /









where (xidyi,zi) are the coordinates of the ith point on the cavity centerline, (xc,,7,zc) are

the coordinates of the cavitator tip and At is the time differential between points i and i 1.

The buoyancy force acts on the cavity and the centerline is distorted over time. One way of

representing the buoyancy force is by a constant acceleration acting against gravity. Now

taking into account the effect of buoyancy on the cavity centerline, we have


xi xc (t (i 1)At)


zi zc (t -(i ) At) b((i ) At)

where b is the constant upward acceleration due to buoyancy on the cavity. In contrast

Logvinovich [1 1] has a different form of equations for the buoyancy and downwash. They

are given by the following expressions

hbuoyancy (X) = 7 v2J R-bXdx (3.13)


F ,xdx
hdownwash(X) = -, (3.14)
npf -o R2(x)

In these equations, X is the position on the body centerline in frame B, R is radius of

cavity as function of position along body centerline, hbuovancy and hdownwash represent the

vertical displacements due to buoyancy and downwash respectively at a point on the body

centerline, Ok is the volume of the cavity enclosed by the cross-section at that point on

the body centerline and V is the velocity of the center-of-mass of the body. In the present

work the cavity model is limited to the effects such as buoyancy, downwash and memory

effects. Other complex phenomena associated with the shape of the cavity are ignored in

the current investigation.

3.3 Planing Model

The stability of the supercavitating vehicle does not depend simply on the hydrody-

namic coefficients as in the case of a fully wetted missile, but rather on the moments asso-









ciated with the nose and the aft. The contact of the aft of the vehicle with the cavity wall

gives rise to the planing force. This force, depending on the strength of the tail-moment

it generates, may send the body into tail-slap. When sufficient restoring nose moments

exist the body may cease oscillating but still have a part of the hull protruding outside the

cavity. In this case part of the lift required by the vehicle to maintain it's flight is provided

by the planing force. The paths an unpowered proj ectile may take for various such planing

configurations has been given in May [4]. The lift needed to support the aft-side of the

body may be generated by either the fins, the planing force or a combination of the both.

In the present work the Hassan's planing force model has been used. Figure 3.1 shows the

vehicle undergoing tail-slap. The planing force (Fp) and moment (Mp)are shown acting at

transom of the planing section.



Plaming








Figure 3.1 Vehicle Undergoing Tail-Slap.


3.4 Planing Force Equations

Wagner's planing theory [12] assumes planing and continuous immersion to be identi-

cal. The theory is mainly developed for 2-dimensional bodies undergoing planing. Using

potential flow, the immersion force on a flat wedge is calculated assuming that the force is

due to fluid transport into a spray sheet. The force per unit-length of the cylinder is com-

puted and then integrated over the entire wetted area. The following are the equations that

characterize this situation as given by Wagner [12].










-- a it w l

Body Wall






Figure 3.2 Cross-Section of Cavity

Assume


h = (R- rta2/ -> 6 = 2tan


h = hs A =R- r (3.15)

In these equations h is the depth of the penetration of the body outside cavity at transom. ,

R is the radius of the cavity at transom and r is the vehicle body radius at transom. Figure

3.2 defines these variables. The force is given by

P = m*h;_ hl ~~t t2 (3.16)


where

dm = pnR2 1 + COS2 Sin2e



= pniR2 (3.17)


Letting

dm* am* ah
-(3.18)
dt ah at

We can write


Sh 1 +t 2h j~( 9








The term m*h~ is usually neglected. If we assume

h(x) = ho -xtan(u)
I = (3.20)
tan(u)

lam* pnsR22
mf = h hho-rtan(,~dx = tan(u) 1 -A h (3.21)

It can be concluded that the planing force is given by

,=Y- ( r +t ho 2n/(.2


Hassan [13]i, based on the Wagner's theory, has given the following set of planing force,

moment and skin-friction equations. The term Fp denotes the planing force, Myp the planing
moment and Ff the planing skin-friction force.

1 R-r r-tho\
FP 2 ho+-tR r) jr+2ho

1 hr +t ho
Mlr = pw V2" (nR2) cosucostIh tRr t h (3.23)

Dummy variables uc, us and S, are introduced into the skin-friction equation. These
variables are defined by the following equations




s r (3.24)
Sw = 4r ((1+uc2)ATan(uc) M)

+ 2~(R-r)tn Os .5) ASin(u,) + 0.5us (1 us

The final form of the skin-friction equation can be written as

Ff. = 0.5pV'"cosolcosuSwCdt (3.25)

Note that Hassan's equation for Fp can be derived from the Wagner's equation for Fp in
Equation (6.8) by using the value h = Vsina.









3.5 Planing Kinematics

3.5.1 Calculation of Immersion

The main assumption involved in the calculation of the immersion is that the cross-

section of the cavity is circular. Strictly speaking the cavity cross-section would undergo

deformation due to gravity and other effects. It may become elliptical in shape [14]. Using

this assumption we can, by simple geometric calculations, find the immersion depth, ho, at

transom as follows
ho = r R+-Rt (3.26)

In Equation 3.26, Rt is the distance between the centers of the circular cross-sections of the

cavity wall and the body. Because the cross-sections are circular, a line passing through

center of cavity and body also passes through the point of deepest immersion.

3.5.2 Method of Calculation of a

This section describes the calculation of the angle a that appears in the planing theory.

It is important not to confuse this a with the numerous places that a represents the angle of

attack. Figure 3.3 shows the angle a as the angle of inclination between cavity centerline

and body centerline.



3 Body







Figure 3.3 Planing Section


Define the following terms,


y, : change in cavitator velocity over the interval [t z, T] expressed in dynamics

frame B. Subscript 't' denotes transom or planing section.


V : Velocity of cavitator at current time.










Vc:Velocity of cavitator at the time: the: cavity at the i .1 ::1:: section formed.


Rc : Location of cavity centerline in the dynamic reference frame.









TF is the tangent to the cavity centerline at the planing point. It is also the normal to the

cross-section. Note that calculating i1:; r i-- of deflections l:-:T:;-1 that cross-sections

are elliptical; this will be ignored for the time being.

The angle a that :i i :: in the planing theory is the angle between the body centerline

and the cavity centerline. The theory assumes that cavity centerline and cavity surface are:

parallel. This is not the case if the cavity radius is changing. The angle a is : :::i: .

by finding the transformation matrix that rotates the: velocity vector Vc so that it is the unit

normal i" = {1,010}"T in the fr-ame B3. This transformation matrix rotates a vector in the

dynamics frame BZ to the cavity normal frame. The: defining .;:: .0. :: is:


o1







The rotation angles for T~ can be found using



fx 1 cos(6) ... 9)










defines
10tation
sequence








Introduce another rotation to account for the rotation of the cavity surface due to ex-

pansion of the cavity
Rx,
sin(ex) =
| | cl

Letting


cos(6x) 0 -sin(6x)
T3 = T3,2,l (0, 8x, 0) = 0 1 0 (3.28)

sin(6x) 0 cos(6x)

and T = T3 T2 we've got our transformation from dynamics coordinates to the frame in

which the planing theory is defined.

The angle a appearing in the Figure 3.3 is obtained from

costu) = < Ti, l > -> cos(u) = Tr y (3.29)

This is the angle between the vehicle centerline and the cavity surface at the planing point


Body





v. Wal

Figure 3.5 Velocity of Transom Relative to Frame B


We need to include the effect of the velocity of the transom on the angle a.

V,,= o x R, (3.30)

In the above equation, V,, is the velocity of transom relative to the frame B. Rp is the

position vector of the transom in the B given by { -Lp,, 0, O} where LI, is the position of









transom on the 1 axis. In order to include the effect of the velocity of transom, F),, on the

angle a, F), is projected onto the vector R ~. Then the angle eR, as shown in Figure 3.3, is

computed by the following equation

sin6 = V(3.31)

In the above equation V denotes the velocity of the vehicle. So, the angle alpha used in

the planing force equation can be finally written down as

a = u+6t (3.32)




Cavity wall



lfho





Body Section

Figure 3.6 Calculation of angle of plane inclination a with assumption that the cavity sur-
face is planar.

A simple way to calculate a reasonably accurate value of a during computations is by

the Equation (6.19). L
tan(u) = (3.33)

Here, we are assuming the cavity wall to be planar. Figure 3.6 shows the angle of plane

inclination, a, with the planar assumption.















CHAPTER 4
LINEARIZATION

4.1 Linearization

4.1.1 Need for Linearization

The equations of motion for the torpedo are identical to airplane equations of motion

but the forces terms on the right-hand side of these equations are different. Thus, the lin-

earization can be carried out similarly, as shown in Nelson [15]. The equations of motion,

as in the case of a supercavitating torpedo, are represented by a set of first-order differential

equations.

ri = f (x, u) (4.1)

using f : 2" 2" as a nonlinear function of a time-varying vector x E "Z and u E 2".

For control design, the system dynamics are observed at some trim conditions by giving

perturbations to states of the system at that trim. The dynamics associated with these

perturbations are obtained by linearization.

An advantage by linearization is that most of the control methodology is based on linear

equations of motion. A controller is designed initially for the linear system and then tested

for the actual nonlinear system. Yet, there are few disadvantages for this process

* Linearized equations can predict the system performance only in a small range of op-
erations. The linearized equations change as the operating point of system changes,
thus making it difficult for simulating true behavior of system.

Information relating to nonlinearities like hysteresis, backlash, coulomb friction, dis-
continuities etc. may be lost by linearizing the system.

A controller that is good for linearized system might have very poor performance for
the nonlinear equations. An iterative process may be needed to find a controller that
is good for nonlinear equations.








4.1.2 Generic Form of Equations of Motion

The equations of motion in Equations 2.67 and 2.70 can be written in simplified form

using sums of total forces and moments acting on the body.

m (u +tqw vr +t gS8) = X

m (v-tru pw -gC8SA) = Y (4.2)

m(wi!-tpy -qu -gC8CO) = Z

Ixt pt + r (I, ly) = L (4.3)

ly@FFx- )= M (4.4)

Iz qI I)= N (4.5)



8O = )-I 46
A 1 SO COr


x COCY CGSY -SO

y = CYSSOS8- CSY SAS8SY +CYCO S~IOCv(47
z ~CASOCY +SOSY CASOSY SCY CCOCw
E( wv 4 7
These equations of motions are coupled by the state vector, x, and are dependent on the

control vector, u.

x = {u, v, w, p, q, r, Y, 8, 4, x, y, z }
(4.8)
u = {61, 6R2, 6E1, 6E2, 8R1, 6R2, 8E 1, E2, 6c, Fprop

4.1.3 Small Disturbance Theory

The small disturbance theory will be used for linearization of equations of motion.

According to the theory the linearization will be carried about an operating point (reference

flight condition), i.e., the equations thus derived will be valid for the torpedo operating at
and near that value of vector x. The operating point is chosen to correspond to the trim

condition in Equation 4.9.









xo ={uo, vo, WO,Po, go, ro, Yo, Oo, Go, xo,yo, zo }
(4.9)
= {uo, 0, 0, 0, 0, 0, 0, Go, 0, 0, 0, O }

This corresponds to straight and level flight with constant velocity. As the torpedo may be

traveling at other flight conditions, the linearization at those conditions would be carried out

numerically, which will be explained in later sections. A value of uo is found numerically,

that satisfies the equations of motion for a given value of xo. Then a disturbance of Ax

is given to the equations of motion from the reference condition thus changing the flight

conditions to xo +t Ax. Several assumptions are made to carry out the linearization:

* The disturbances from reference flight condition are small. Thus the terms involv-
ing higher order of disturbances A will be neglected. Furthermore first order terms
involving A will be approximated as in Equation 4.10.

Sin(A) = (A)
(4.10)
Cos(A) = 1

The propulsive forces and mass are assumed to be constant.

Planing and immersion forces are neglected for this analysis. However, the following
analysis is just an explanation of the method. Eventually, the linearization is carried
out numerically. Thus, in the numerical method the planing forces are included,
unless specified.

The linearization procedure is resolved for the force equation in by direction. This

equation relates the force, X, to the states.

m (u tqw -ru -tgS8) =X (4.11)

Using the flight condition from Equation 4.9 in Equation 4.11 gives the value of force at
the reference trim condition.

mgS~o = Xo (4.12)

The perturbation equation, i.e., the equation with flight condition disturbed by Ax can be

obtained by substituting the perturbed flight condition into Equation 4. 11.

m [ (uo +t AM) + (40 +t Af) t o +t A) (r0 +t A) (uo +t AM)
dt (4.13)
+tgS(Oo +t AG)] = Xo +t AK









Equations 4. 12 and 4. 13 can be combined to give the linearized form of Equation 4. 11 for

straight and level flight condition.


m (Au+- gAGC~o) = AK (4.14)

Proceeding in a similar way all other equations of motion can be linearized. The lin-

earized equations for straight level flight are shown in Equation 4. 15 to Equation 4. 18.

4.1.3.1 Force equations

m (Au+- gAGC~o) = AK

m (Av +t uoir g^0000) = AY (4.15)

m (Au -uo~q tgAGS~o) = AZ

4.1.3.2 Moment equations

Ix Ap = AL

ly~q = API (4.16)

Izr = AN

4.1.3.3 Orientation equations


C~o

AG = Aq (4. 17)

AQ = Ap+- T~o~r

4.1.3.4 Position equations

Ax= -SO,,an,,G+-COI ~n u+S~o~w

Ay = Av (4.18)

Az = -CO,,an,,G- SOI ,Au+tC~o~w

4.1.4 Stability and Control Derivatives

The variations in total force and moment are often difficult to compute.These variations

in forces can be calculated by chain rule for derivatives. As stated in Equation 4.8, these are

functions of state variables x and control variables u. Thus for example AX can be written

by chain rule as in Equation 4.19.










K =X,Au+_tXvOv+tX,~ -w +X,Ap+XAq+Xr,

(4.19)
+ XeR1 6R1 +t X8R2 6R2 +t X8E1 OEl 1+ X8Ez 6E2

+ XaR1 R1+XaR 6R2+-tXaE16E 1+-XaE 6E2+-tXac i

where the sub scripted X represents its partial derivative with respect to its sub script.

aX
x, = (4.20)
au X=XO

Each of these subscripted variables that have a subscript of state variable are known as

stability derivatives and the ones with a control variable as subscript are known as a control

derivative. There can be as many stability derivatives as there are forces and state and

control variables. Many of these are negligible, depending on the reference flight condition.

These dependencies are known usually by experimentation or numerical calculations. For

example, the force, X, is observed to depend mainly on a subset of the state and control
variable. Thus only the stability derivatives that correspond to these variables have to be

retained in Equation 4. 19, when straight and level flight is considered.

X = funct(u, w, SEl s E2, 6Els E2, 6c, Fprop ) (4.21)

The next problem is calculating numerical values of these derivatives. In most cases it

is easy to calculate these numerically or using a symbolic manipulation software. For some

reference points, it is possible to do the calculation manually. The calculation ofX, will be

done manually for straight and level flight.


B (FR1,x +t FR2,x +t FE1l,x +t FE2,x +t Fe,x +t Fprop ,x) (4.22)

Expressions for each of the terms in Equation 4.22 have been derived in Chapter 2. For

example, the expression for the force on cavitator is shown in Equation 4.23.




Fc~x= Ci 0 Sc -Sc Ci 0 (4.23)

















































The velocity components in Equation 4.27 can be found using Equation 2.2.


In Equation 4.23, ac, Pc, and thus Le and De are the only functions of u. Thus the partial

derivatives with respect to ze can be obtained.


acx=


-cCac 0~ r Sc, a







0~s, + 0~a ~

-Le(ac,Y ) -SC


- Dc(ac,: -S
0T N I


0 S6c Ocpuca CcaSpc -Suc
1 0-P -Sic Oc

0 C6c OcpSuc SucSic Cca


(4.24)


It can be seen that a s, ,L and D are terms required to be calculated. These can be

calculated from equations 2.16 and 2.17, which are restated in Equations 4.25 to Equation

4.27.


tan(ac) = '"
21c


(4.25)

(4.26)

(4.27)


tan(Pc) =
Ce

V;2 =t + -t + w; W









C6c 0

0 1


u6 o




w
BI

be obtained


-S6c

0

C6c


(4.28)


by $2 ~3

p q r (4.29)

xc ye zc

for the reference flight condition that is


L IB

Now the velocity components can

stated in Equation 4.9.


2 ucdwe we due
sec2 (ac)due=

ucdwe we due
due =
u? +t w,
CGcdwe SGcdue
due =
uo uo


(4.31)



(4.32)


at (xo, uo)


Similarly the variation of Pc can be obtained by differentiating Equation 4.26 at reference

flight condition.


(Vcdvc vcdYc)
di =
c Vc Y V
dvc
= at (xo, uo)


(4.33)


Now, using Equations 4.28 and 4.29, variation of velocity components of cavitator with

respect to u can be obtained.


auc
= C6c


= 0


= S~c


(4.34)


uc CGcuo





The variation of ac can be obtained by differentiating Equation 4.25 at reference flight

condition.









Finally, combining Equations 4.32 to 4.34, the variations of ac and Pc with respect to u can

be obtained at reference flight condition.


ue C6Sc Sw S6,C Bu
C~cSc S~C~c(4.35)
uo uo


apc 1 avc
Bu uo 8"(4.36)
_O

Clearly, two of the derivatives that are required to calculate stability derivatives have been

obtained. It was previously stated that lift and drag are calculated using the coefficient of

lift and drag.


Le = ~plc2Sccle
2 (4.37)
De = ~p~c2Scdc


These forces can be differentiated by chain rule the derivative would be like in Equation

4.38.

-ta -rpSc 2V~ccle + Vc (4.38)

The Equation 4.38 requires two derivatives. One of the derivatives is calculated in Equation

4.35. The other derivative can be calculated using Equation 4.27.




;uc +v +weu lr? (4.39))
u? + v + w









Thus the derivative of the lift and drag forces can be obtained.


aLe
= pScYccle (4.40)
aD,
= pScYccdc (4.41)
au

Thus all the derivatives required to calculate right-hand side of Equation 4.24 have been

calculated. All the terms on right-hand side of the Equation 4.20 can be calculated in

a similar manner. It is tedious to calculate the derivatives analytically in such a way. The

complexity increases for other flight conditions. For practical purposes these derivatives are

calculated using symbolic manipulation software like Mathematica or by using numerical

methods. The numerical methods used to calculate the derivatives have been described in

Appendix B. The Mathematica code for linearization is described in Chapter Appendix C.


4.2 State Space Representation

Equations 4.15 to 4.18 are a complete set of linearized equations of motion for the

torpedo. They can be represented in a more convenient form known as the State Space

Form. The state space equations are a set of first-order differential equations.

i = Ax+-tBu

y = Cx+-tDu

x E Zn,u E ~Z,ye Zm (4.42)

Ae Znx Zn,Be Znx Zp

C E Zmx Zn,De Zmx X F

Equation 4.42 is a generalized form of state space representation for any system. Each of

the terms in the equations has a particular importance for describing the dynamics of the

system.

*State Variable x: The state variables for a system are a set of variables, when known
at time to and along with input u, are sufficient to determine the state of the system
at any time t > to. All the states of the system need not be measurable.









* Input Variable u: This is the control surface deflections.
* Output Variable y: The output variables are the measured parameters. These may
or may not be same as the state variables. The output variables are usually considered
to be measurable but sometimes they are estimated.
The matrices A, B, C and D may either be constant or time-varying functions.

In the case of the supercavitating torpedo, the state vector is of size 12 (n) and the

control vector is of size 10 (p).


x =[n nA Aw Aq AG Av, Ap Ar Y Ax Ay Az
(4.43)
u = ic8El 6E2 6R1 6R2 8E1Bl BE R1 BR2 bfypro

Some of these controls may not be needed for some maneuvers. From the linearized equa-

tions it can be observed that the state variables are not coupled by the states {Y,x,y,z}.

These four states can be removed from the analysis for now. The system becomes a 8 state

system. These states can be further divided into longitudinal and lateral-directional dynam-
ics. The variables Au, Aw, Aq, AG correspond to longitudinal dynamics, which also means

the dynamics in $1bi3 plane. The variables Av, Ap, Ar A correspond to lateral dynamics,

which is the dynamics in $1b2 plane. Sometimes the lateral and longitudinal equations can

be decoupled. Thus if the torpedo is making a pull climb/descent to a certain depth, usually

its dynamics can be represented by longitudinal state variables. The plant matrix A can be

divided into four parts.


A = AopAlong Acoupl~lt (4.44)



When A is divided as in equation 4.44, where each element is a 4 x 4 matrix, Along

would correspond to longitudinal dynamics and Alatd would correspond to lateral dynam-

ics. Acoupl and Acoup2 are coupling matrices. It is required that the coupling matrices

become negligible for the equations to be decoupled. If these parts are not negligible,

the equations cannot be decoupled, and a 8 state model will be required to be considered.































































(4.49)


Xsc
m

Blong .

0 --


From linearized equations, the four parts of the A matrix for the torpedo can be written as

in Equation 4.45 to Equation 4.48.

x, xw +x g~+x
m 90 mt mw- m gO tX
o+z, z, uo +z -gC~oS~o + ze~
Alon m m m m (4.45)


0 0 C~o 0


-uo + -gC~oC~o +
Lr-(l4 )qo Le
Ix Ix



C~oT.~o 'ocoose0,,1 ,,1~H-

vo +x

~P zv -gS~oC~o $
m~ m
Z)ro Mr(I-4;oM


-S~o --S ogo -- C


m
Lp
Ix
Np lyx)90

1


Alatd








Acoupl


(4.46)








(4.47)








(4.48)


ro +x







Ix

N,
0r -


xz










L,
Ix
Nw
0


-gS~oSpo +~
Le

No


S~ogo +t C~oro


z



~oro


S4(o YTOo


Similarly B is a 8 x 10 matrix, whose elements are just the control derivatives according to

their locations in the matrix. The first 4 rows correspond to longitudinal dynamics and the

last 4 correspond to lateral dynamics.


XFprop,x




0x1









Tsc YFprop,x

Blatd = (4.50)

-4x10

Now the complete state space representation for the torpedo can be written as in Equation

4.51 which gives two sets of equations. The first set is the longitudinal equations and the

second set is the lateral-directional equations.


Xlong = Ihu Aw Aq AG]

xlatd = Av Ap Ar A'

H = 16c El: 6E2 6R1 6R2 B8El 8E12 BR1 BR2 EFprop (4.51)


Xilong Along Acoupl1 Xlong Blong 1x

Ailatd Acu2Alt latd Blatd
8xl 8x8 8xl 81















CHAPTER 5
CONTROL DESIGN SETUP


This chapter deals with the control design for the torpedo described in previous chap-

ters. Various parameters associated with the control are restated in Table 5.1.

Table 5.1 Control Parameters


Longitudinal Lateral
State u, w, q, 8 v, p, r, Control Sc, 6el, 6e2 6rl, 6r2


It should be noted that Y has been included in the states though it was observed in

the state matrices that all other variables are independent of Y. It will be seen later that

the inclusion of Y in the feedback states helps in improvement of performance. Also,

for longitudinal control, two elevators and the cavitator are required. Similarly for lateral

direction, the rudders should be enough for control. A deviation from these requirements

will be observed in some of the controllers, mainly to improve performance and provide

stability.

There are various control methods, like linear quadratic regulator (LQR) synthesis, pu-

synthesis etc., which can be used to design a controller. Each of these methods has advan-

tages and disadvantages. LQR method gives a constant gain controller which is based on

minimization of a quadratic performance index and considers the problem of robustness

only in terms of gain and phase margins. pu-synthesis deals with robustness with respect to

a wide variety of uncertainties to minimize an infinity-norm matrix but the resulting con-

troller can be of high order. Regardless of complexity and robustness, each design method

presents difficulties. This chapter explains various problems associated with the control

synthesis and the system model used for synthesis of the controller.









5.1 Open-Loop Performance for the Fixed Cavity Model

Initially the equations of motion for the torpedo have been linearized for straight and

level flight at a forward velocity of 75 msl


x= {75, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, O} (5.1)

It is found that the cavitator and two elevators are sufficient to maintain the above trim.

It is also assumed that the value of propulsive force required to maintain this trim is fixed

at the required value.

H= ( cel,~e2,6rl, 6r2, Fprop
(5.2)
= {0.0067, 0.0106, -0.0106, 0, 0,4.0023e+-t03}

where the deflections are given in radians, and Fprop is in Newtons. As these parameters

are obtained numerically, it may not be a perfect trim. The system may have some non-

zero accelerations, and consequently may tend to deviate from straight and level flight. To

check this, the open-loop dynamics are simulated at this condition without any feedback.

Figure 5.1 shows the Simulink model used for open-loop simulation. The control surface

deflections are fixed at their trim values for this simulation. The closed-loop system is

obtained using the equations of motion that were derived in Chapter 2. The state derivatives

are then integrated to obtain the state at the next instant.

Figure 5.2 shows the open-loop response for the torpedo at this trim condition. It can

be seen that the open-loop system is unstable. The simulation is carried out at the trim that

is shown in Equation 5.1, i.e., all the values shown in these figures have to be zero. The

system is seen to have oscillation about non-zero states.

The state and control matrices obtained for this trim condition are shown in Equations

5.3 to 5.6. There are 5 control variables, {Sc, 6e, 8e,2, ir, in }. The matrices corresponding

to the lateral dynamics are of dimension 5 because the state Y is included in the lateral

dynamics. Thus the lateral states are now {v, p, r, Y}.























1 ~~~NL Equation~opd Integrator
at For

simt

Clock
Time


Figure 5.1 ::::::i::. 1 Model for Open Loop Simulation


state
r Plotting


Control ~
Trim


20


15



5 '


0 01








-0 03

-000 20 40 6 8 00
times)


oO 20 40 0 t9
4
3 5



P25

1 5


Fi::- 5.2, O -- : rLoop i- i..: for Torpedo: w?,p, q


VVVVVVVVWVVWvywywywwyvi










-4.5204

-0.2616

0.0000





-32.3010

-406.0942

158.4675

0


-12.0422

-0.1813

1.1437
-

0


1.5417

-15.7648

1.2077

0


1.3110

78.5888

-3.5614

1.0000


-9.8100

0

0

0


Along









Blong









Alatd =


(5.3)









(5.4)









(5.5)


69.0608 -69.0608

-303.3736 303.3736

-45.1531 45.1531

0 0


-0.0002

54.2281

-0.0025

1.0000

0


-71.6011

0.3004

-1.2528

0

1.0000


9.8100

0

0

0

0


0

-14297.086

-1.4129523

0

0


0

14297.086

1.4129523

0

0


-366.60511

-17276.994

54.561629

0

0


366.60511

-17276.994

-54.561629

0

0


Blatd


(5.6)


The longitudinal eigenvalues are {-21.1414, -4.5137, 1.8262, -0.0178} and the lat-

eral eigenvalues are {0, -54.2289, 0.0002, -6.6472 +t 7.2683, -6.6472 7.2683 }. The

eigenvalues clearly show that the system is unstable. It can also be seen that the longi-

tudinal dynamics have no oscillatory modes. Figure 5.3 shows the variation of eigenvalues

for the torpedo for different velocities. State values are fixed except for forward velocity,

which is varied from 60 ms-1 to 90 ms-1. The figures show that the variation is small and

most of the eigenvalues stay in negative half of complex plane.










1 Longitudinal Egienvalues for u=60:5:90 10 Lateral Egienvalues for u=60:5:90


0.55



.E E
-0.5 -


-1 10
-0 -20 -10 ~~og 0 10 -0 -60 -~a (La -20 0 20

(a) Longitudinal (b) Lateral

Figure 5.3 Variation of Eigenvalues with Change in Velocity



5.2 Closed-Loop Problem

As stated earlier the control problem can be subdivided into various problems. Each

can be solved to get a final controller. The ultimate goal of the controller design is to

achieve a desired trajectory or reach a particular point with minimization of some perfor-

mance criteria. The achievement of this goal requires addressing maneuvering, trimming,

guidance and navigation. This thesis will consider the basic problem of maneuvering. So

the problem is to be able to track a certain pitch and roll command while maintaining cer-

tain performance criteria. The performance criteria that the controller is required to meet

are:

* Track a step command in pitch or roll rate of size up to 30 deg/s.
* Maintain an overshoot less than 15%.
* Have a rise time of less than 0.6 sec.
* Have a steady state error of less than 5%.

Besides meeting the above mentioned performance criteria, the controller is also required

to stabilize the closed-loop system.

Various bounds are placed on the control surface deflections and rates. These bounds

are listed in Table 5.2. The bounds are included in the nonlinear simulations and it is

required that there is no saturation of deflection or the rates at least for the commands

having the rate 30 deg/s.












Cavitator Deflection -15" I Sc I +t15"
Cavitator Rate -25" s I Sc I +t25" s
fins -60" I 6f I +60"
Fin Rate -25" s I 8f I +25" s
Thrust 0 < F,, < 30000N


Table 5.2 Control Constraints


An actuator model is included in the system so as to constrain the rates of the control

surface motion. This actuator is realized as a low pass filter, Ac = ~. Addition of this

filter synthesizes a controller that has slower control deflections.

Let qcomm(t) be a function of time, defining the desired pitch rate profile. The aim of

the controller is to find a control law u(t) that yields an achieved pitch rate, qachi(t), to

minimize the optimization problem stated in Equation 5.7.

find u(t)

that minimizes 5(t) = |qachi(t) qcomm (t) |

subj ect to ui


i = Ax+-tBu

where, umin and umax are lower and upper bounds on control deflections. The quantities

uin, and umax are lower and upper bounds on control deflection rates.

The problem becomes a disturbance rej section problem, when the commanded variable

is 0 for all time. This is an optimization problem,where the state space equation is an

equality constraint and the control surface bounds are inequality constraints.

5.3 Robustness of the Controller

A control system that is designed to accommodate the uncertainties in a mathematical

model is called a robust control system. Usually the response of a model does not accurately

match the response of the true system. A robust control system should give the desired









performance not only during the simulations of the model, but for the true system also.

Various parameters can be introduced in the model to simulate uncertainties. Random

noise can be added to output signal to simulate measurement errors, the gains in signals

can be changed to model uncertainty in response etc. Gain and phase margins are generally

used to predict the robustness of a control system. Physically, these are just the factors

by which the feedback gain can be increased and yet have a stable real system. A formal

definition of these can be given by using a frequency analysis for a feedback system.

5.3.1 Gain margin

Figure 5.4 shows a typical feedback system involving a plant, P, and a controller, K.

The gain for the system in dotted region is known as the loop gain. The loop gain is

important because it determines stability. Errors in the predicted loop gain could cause

errors in predicted stability. The gain margin is the largest factor by which this gain can be

increased and still have a stable system. Physically, it means if the response of the torpedo

for a given elevator input is higher by a factor of the gain margin, the torpedo is still stable.

The gain margin is usually expressed in decibel (db) units, and can be easily obtained from

the Bode plots for the system. The gain margin is a measurement of the magnitude on the

Bode plot, at the point where the phase is 1800.

5.3.2 Phase margin

Gain is a valid robustness criteria when the system has real eigenvalues. But usually

the eigenvalues have imaginary components and thus the phase is also a concern. Phase

margin is the measure of the maximum possible phase lag before the system becomes

unstable. From the Bode plot, phase margin is the phase when the magnitude of the gain is

zero.

5.3.3 Uncertainty in parameters

Another factor that can determine the robustness of a controller is its response to errors

in known parameters. As stated earlier, the coefficients of lift and drag are calculated from a

CFD database. The accuracy of the model depends on accuracy of this data. Robustness of




















Figure 5.4 Loop Gain

a controller can be assessed by introducing errors in the data and checking how it performs.

The following variations have been introduced in the system to check for performance of

the system with intrinsic uncertainties:

* +20% error in Cl of Cavitator.
* +20% error in Cd of Cavitator.
* +20% error in Cl of all the Fins.
* +20% error in Cd of all the Fins.

5.3.4 Controller objective

In terms of robustness, the controller is required to meet various performance objec-

tives. These objective can be summarized as:

* The closed-loop system should have a gain margin of at least 6 dB.
* The closed-loop system should have a phase margins of at least 45 deg.

5.3.5 pu analysis:

For the Hoo/p analysis [16] it is desired that the peak pu value for the closed-loop system

be close to 1. This ensures good robustness atleast equal to the values of uncertainties in

the synthesis model.















CHAPTER 6
LQR CONTROL

6.1 LQR Theory

The tracking problem, like the one given in Equation 5.7, can be solved by using a

combination of feedback and feedforward control [17]. The Linear Quadratic Regulator

(LQR) problem is to find an optimal feedback matrix K such that the state-feedback law

u = -Kx minimizes the linear quadratic cost function shown in Equation 6.1.


J,(u)= (x'Te Qx +uTRu + 2x Nu)dt (6.1)


The basic idea of LQR control is to bring the state of the system close to zero. A linear

system can be represented in the state space form as in Equations 6.2 and 6.3. The matrices

A and B are the state and control matrices. The variable x represents the state vector, y is

the output vector and u is the input vector.

x = Ax +t Bu (6.2)

y =x (6.3)

The LQR controller is realized by a constant gain matrix K, such that the feedback

u = -Kx makes x go to zero. By a modification to this law, the LQR method can also be

used for tracking. The state vector x is of size n.

x={x1, x2, ---, x,,} (6.4)

Let the tracking problem be for the state xl to track a step command rl. The idea is to

make (xl rl) go to zero using a LQR controller. The new control law can be chosen as in









Equation 6.5.



u = -K (6.5)




Equation 6.2 can be rewritten by substituting the new control law.

i = Ax+-tBu

X1 r1

x2 (6.6)
= Ax BK





For simplicity, assume that there is only one control, u (this is different from velocity u).

The controller K is of size n x 1 and it can be expanded in its elements.


K= kl,k2,---,k,, (6.7)

Equation 6.6 can be rewritten by substituting the K in its expanded form.




i = Ax B kl, k2, -, k;,

(6.8)


= Ax BKx+-tBklrl

= (A BK)x+-tBkl rl

It should be noted that the command rl is a step command. The steady-state dynamics of

the system can be obtained from Equation 6.8.


i(- )= (A -BK)x() )+ Bklrl


(6.9)








The error dynamics can be obtained by subtraction Equation 6.9 from Equation 6.8.

i(t) -i( ) = (A K) (x(t) -x( )) (6. 10)

e = (A BK)e (6.11)

where e = (x(t) x( o)). So, the tracking problem is cast as a regulator problem. The new

state vector is the steady-state error e, which is made zero using the regulator. Figure 6.1

shows the block diagram for this closed-loop system. It is required that the closed-loop

system has an integrator somewhere so as to make the steady-state error go to 0 [17]. That

is, e has to go to zero rather than e so as to achieve good tracking. Figure 6.2 shows the new

configuration of a system that has no integrator and thus an integrator has to been included

during design. Thus, the integral of the actual error has to be made to go to zero so as to

achieve a good tracking.
e = (ri- xt) (6.12)


The state space equation for the system with this modification can be written.

i = Ax+-tBu

e"= ri xi (6.13)

= r Cx

where xl = Cx. It can be seen that the error equation is similar to state equation. Thus e"

can be considered as another state, .i.e, the system now has n+- 1 states with state vector

i = {xl,x2, ,Xny e) T. So a new formulation of the state space equation can be written,

AO B
i= i+: u+ ri (6.14)


-> = Ai+tBu+Ir~ (6.15)

which is similar to Equation 6.8. The error dynamics of this system represent the form of

state space equations, for which a LQR controller can be derived. The LQR controller K,

will be a constant matrix of size n+- 1 as the system now is of size n+- 1.






70

K= [kl,k2, ---, kn||kn 1] (6.16)

Then, the new control law can be written as in Equation 6.17.
u = -Kxj


=-[k, k2 k Ilk~l] 11(6.17)

= -[kl, k2,- -, kn]x+t [-kn 1]e"
= -Kx -tkre

which is represented in Figure 6.2.

rl kltti ~y=x




I K


Figure 6.1 Controller for Tracking when Plant has an Integrator.


rl + x + Bu X




I K





Figure 6.2 Controller for Tracking when Plant has no Integrator.









6.2 LQR Control for Fixed Cavity Model:

6.2.1 Control Synthesis

The torpedo system does not have an integrator in the system. A tracking controller can

be obtained from LQR method by the process described in Section 6.1. The controller is

obtained for a trim state of straight and level flight at 75ms 1. The linearized dynamics are

first separated into the longitudinal and lateral dynamics as given in Table 5.1. The controls

used in longitudinal direction are the cavitator and 2 elevators. The controls in lateral

direction are the 2 rudders. It is observed that for straight and level flight the longitudinal

and lateral dynamics are practically decoupled.

Once the state and control matrices have been obtained, the main variables that the LQR

controller depends on are the weighting matrices Q, R and N. In this case the cross coupling

matrix N is chosen to be 0. The matrices Q and R penalize the cost function for higher state

and control values respectively. A higher value in Q matrix would cause a better track-

ing. A larger R would constrain the control surface deflection. An optimum combination

of the matrices is obtained iteratively, so as to get good tracking with achievable control

deflections.

The matrices for longitudinal pitch rate tracking are given in Equation 6.18.

Long = diag( [0, 0, 0, 0, 10])
(6.18)
Rlong = diag([ 5, 4])

The first four numbers in the Qlong correspond to weightings on the four longitudinal states.

They are chosen to be 0. We do not want to restrict the states from changing. This is

especially important for weightings on q and 8. A weighting on these variables would

restrict them from changing. The last number, 10, is weighting on the error. This is chosen

to be high to penalize the tracking error. A higher value of weighting would give a better

tracking, but it is seen that it would require very high control rates. The first number in

the weighting matrix Rlong, 5, corresponds to cavitator weighting and the number 4 is for









elevator weighting. Elevator weighting is chosen to be smaller so as to encourage the

controller to use more elevator than the cavitator. This gives a more stable performance.

The control matrices obtained for the longitudinal dynamics are given in Equations 6. 19

and 6.20.

1.1182

S-0.9681 (.9

K=I-0.0000 0.0040 0.1016 -0.0000 1.4195 -1.1184
0.0000 -0.0042 -0.0995 -0.0000 -1.3981 1.1308 (.0


Similar process is involved in the design of the lateral controller. Initially the lateral

controller is designed with only four state feedback, and Y is neglected in the feedback. In

this case it is found that the torpedo has high sidewash and deviates considerably from the

original path, even when the pitch angle is 0. To avoid this, Y is included in the feedback

states. It is also observed that a penalty on the yaw motion causes the controller to com-

mand a very fast control surface motion. Also, a continuous correction of control surface

deflection is required to prevent the yaw motion entirely. Thus an optimum combination of

the weighting matrices is obtained that would prevent a very high yaw motion but would

still have slow control surface motion.

Qlatd = diag([0, 0, 0,0, 0,.1])
(6.21)
Rlatd = diag([1000, 1000])

The first 5 numbers correspond to 5 states and the last number is weighting for the error.

The Rlatd is of dimension 2 as only the rudders are included in the synthesis. The weighting

on the rudders is high as it is observed that the roll rate is very sensitive to the rudder

deflection. The control matrices obtained for the lateral dynamics are given in the Equations

6.22 and 6.23.










-0.0071 (.2

-0.0071


K = 0-3 0.0005 -0.1253 -0.0132 0.0019 -0.0000 (6.23)
0.0005 -0.1254 -0.0132 0.0026 -0.0000

The feedback matrix K for lateral dynamics is of size 2 x 5, which is shown in Equation

6.23.

6.2.2 Nominal Closed-loop Model

6.2.2.1 Model

Figure 6.3 shows the eigenvalues for the closed-loop longitudinal and lateral systems.

It can be seen that both systems are stable as all the eigenvalues are in the left half of

the complex plane. Also, each of the dynamics has one eigenvalue at the origin, which is

introduced due the integrator in the system.





-10
-~~oo -84
5solP~~O 2 0 ~~0a70




(a) Longitudinal (b) Lateral
Figure 6.3 Eigenvalues for the Closed-Loop System


6.2.2.2 Simulations

The response of the vehicle to a longitudinal command is simulated and shown in Fig-

ures 6.4 to 6.6. These figures show the response for a pitch rate doublet of 15 deg/s. The

rise time for the pitch rate command of 15 deg/s is 0.18s and there is an overshoot of

11.53%. The steady-state error is .8%. The controller is able to command pitch rates as

high as 30 deg/s. It is observed that the vehicle motion is confined to longitudinal plane







741


20


10



0
-10


-20 5 10 15
times)

Figure 6.4 P~itch Command T~rackiing : q

1 20

0.5 ~ l1 10





-0.5 ~ 1-
-20


-1 5 times) 10 15 -30 5 times) 10 15


Figure 6.5 Pitch C~ommand Tracking : C Sc


only. This shows that the controller allows pure longitudinal motion to be uncoupled from

the lateral m~otion.




1.15



1.5~ 10 -





00 5 times) 101 105 times) 105


Figure 6.6 Pitch Command Tracking : Sen Bel









The response of the vehicle to a lateral, roll rate, command is shown in Figures 6.7 to

6.9. A roll rate command of 15 deg/s is achieved in .52s with an overshoot of 0% and

a steady-state error of 0.09%. The controller is able to command a roll rate motion of as

high as 50 deg/s before a saturation of control surface rate is reached. It is observed that

there is some longitudinal motion in this case. This longitudinal motion has been reduced

by inclusion of Y in the feedback states to the controller. It can be seen that the rudder

deflection for a roll rate command is small. This is expected as the terms corresponding to

roll rate from rudder are an order of 3 times larger than the terms corresponding to pitch

rate from elevators. It is assumed that the control surface deflection is achievable.











0 5 10 15
times)


Figure 6.7 Roll Command Tracking: p


O 5 10 15 -05L5 10 15
times) times)


Figure 6.8 Roll Command Tracking: Sc, Sc

















0.4 -2
0.3 -2.5

0. 5 times) 101 05 times) 105


Figure 6.9 Roll Command Tracking: Se, Bel1

rl + y=x '3






K






Figure 6.10 Breakpoints for Calculating the Loop-Gain for a Tracking Controller


6.2.2.3 Gain and phase margins

The LQR tracking system shown in Figure 6.2 is obviously more complex than the

system shown in Figure 5.4. Thus, the loop gain can be defined in many ways in this case.

The block diagram can be broken at different points so as to simplify it to the form shown

in Figure 5.4. Figure 6.2 is redrawn in Figure 6.10 which shows the possible breakpoints

for this system. For understanding, the output of plant P is divided into two parts, one is the

achieved value of the commanded variable (ra) and the other is remaining states of the plant

P (x). The break points are numbered 1 to 3. The system can be broken at each of these

points to give a loop gain. These gains will be named outer-loop, inner-loop and all-loop

gains respectively.
Gain and Phase margins for each of the above possible break points have been calcu-

lated for both the longitudinal and lateral controllers. Table 6.1 lists the gain and phase









Table 6.1 Gain and Phase Margin with LQR Controller


Longitudinal
Gain Margin(db) Phase Margin (deg)
1 21.056(at 47.498 rad/s) 64.846(at 9.0625 rad/s)
2 327.87(at 0 rad/s) 77. 118(at 25.925 rad/s)
3 o 57.606(at 20.845 rad/s)
I ~Lateral
Gain Margin(db) Phase Margin (deg)
1 22.964(at 0 rad/s) o
2 250.51 (at 0 rad/s) o
3 50.36 (at 0 rad/s) o


margins for the torpedo with LQR controller that was obtained in previous sections. All

margins are quite high and meet the desired conditions of 6dB for gain and 45 deg for phase

margmn.

Also, the lateral controller is unable to stabilize the unstable spiral mode. Thus the

closed-loop system is inherently unstable due to this pole and would consequently have

negative gain margin. Numerous simulations show that the affect of spiral mode is negligible,

i.e., the time to double for the instability is considerably larger than the maneuvering time

of the torpedo. So, the closed-loop system model is reduced by removing the spiral mode

from the model. The gain and phase margins in Table 6.1 are for this reduced-order system

and refl ect the robustness of the dominant dynamics.

6.2.3 Perturbed Closed-loop Model

A perturbed system model is formed by adding an error to the values of coefficients

of lift and drag for the fins and cavitator. New values of trim deflection are obtained for

the perturbed model and thus a new set of A and B matrices is obtained. Tables 6.2.3 and

6.3 show the percentage variation of the elements of A and B matrices for a 20% change

in coefficients of lift cavitator. This comparison is done for cases with changes in other

coefficients also. In all cases, few elements in the state and control matrices change. In

most cases, the change in elements of A and B matrices is a linear function of the change













u w q Ovpr Y
u 0.46 0.62
wl 5.52 6.86 1.58
q 1.05e5 34.8 13.58
v



p


Table 6.2 Percentage Variation in A Matrix due to 20% Variation in clc


in a coefficient. For example, in Table 6.2.3 there are 8 terms that show a variation due

to a 20% variation in coefficient of lift of the cavitator. The term A(3,1) shows a large

variation but its numerical value is negligible. The term A(3,2) shows a 34% variation

but this term is also small compared to other terms. Remaining terms in the matrix show

very small variation. Some terms in the B matrix show a 20% variation. Thus some terms

in controllability matrix change considerably. This would mean that for an error in these

coefficients, the response would show some difference in control surface deflection. As it

is observed that the closed-loop system has good gain and phase margins, this effect on B

matrix should not be of much concern.

6.2.3.1 Model

Figure 6. 11 shows the eigenvalues for the perturbed closed-loop longitudinal and lateral

systems. An error of -20% is included in the value of coefficient of lift for the fins. It can

be seen that the longitudinal dynamics show some perturbation in the damping while the

lateral system relatively unchanged.

6.2.3.2 Simulations

The performance of the controllers is studied using the simulation with a perturbed

system model. An error is assumed in the values of various coefficients and a correction















Sc 6el 6e2 r 6r2


w 20
q 20

v


r


Table 6.3 Percentage Variation in B Matrix due to 10% Variation in clc


15 6
10





-5
-4
-10
-6

-11 00 -80 -60-400-20 10 -0 60-00-0 0
Real Real

(a) Longitudinal (b) Lateral

Figure 6. 11 Eigenvalues for the Perturbed Closed-Loop System: 20% Error in clfin




factor is added. Response of the closed-loop nonlinear system is not much affected by the

variations in coefficients of lift and drag. It is observed that the controller commands the

system to a new trim state which is also a straight and level flight, with change in speed

and control deflections. After that, the system follows a pitch or roll command as well

as before. Figures 6.12 to 6.13 show the response for one such case. In this case a roll

doublet is commanded to the system, and there is an error of +20% in the value of clfin.

It can be clearly seen that the vehicle has gone to another trim state and then it follows the

command equally well. There is almost no change in the trajectory of the vehicle. The










C C

4---- +20% error 4~ ---- +20% error
-20% error I -20% error




0. -2 a-
"-2 ~ 1 I -4
-44
-6

CO10time(s) 203 0time(s)203


Figure 6. 12 Response for 20% Variation in clpn, P, 4

1 1.2
---- +20% error ---- +20% error
S -20% error 1 -20% error
0.5

9 90.8


o 0 0.6



-0.50 1 20 30 0.40L 10 20 30
times) times)


Figure 6.13 Response for 20% Variation in clpn: 6c, Bel


control surface deflections are similar with a constant offset. Such response has also been

checked for other cases. The affect of error is similar in all cases.

6.2.3.3 Gain and phase margins

Table 6.4 lists the gain and phase margins for the perturbed closed-loop system. The

perturbed system also has good gain and phase margins. Comparing the values with Table

6.1, it can be seen that there are small changes in the values except for the lateral all-loop.

The last value is increased to showing an improvement for the perturbed system.

From the analysis of the perturbed closed-loop system it can be said that the linear

model is robust to various uncertainties in the system.






81


Table 6.4 Gain and Phase Mtargin for Perturbed Closed-loop System: : error in clfin















CHAPTER 7
p/lHo SYNTHESIS CONTROL

7.1 Uncertainty

The main parameters (besides the control) that dictate the forces on the the torpedo

are the coefficient of lift, cl, and drag, cd, of the cavitator and fins. The values of these

parameters are obtained using a CFD code in [5]. These values have also been obtained

theoretically and experimentally in [4]. A comparison of the CFD and predicted results

is shown in the figures 7. 1. It can be seen that there is a difference of at least 1 order of

magnitude, in the value of cd, between the two datasets. Also, a supercavitating flow is a 2

phase flow with partial water and partial vapor. Thus the hydrodynamics for supercavitating

vehicles are unique and need more investigation. It is clear that there is a need for inclusion

of uncertainty for the values of cl and cd, which are main parameters that determine the

hydrodynamics.

A controller designed based on pu or Hoo theory is a robust controller, which deals with

errors and uncertainties in the system, implicitly. Basic idea of pu synthesis is to reduce

the gain from error or disturbance to the error in tracking. In the design of pu controller

for torpedo, an uncertainty is assumed in the coefficients of lift, clc, and the coefficient of

drag, cde, for the cavitator only. The following formulation for the synthesis model has

been derived in detail for the longitudinal plant. The formulation for lateral plant is very

similar, difference being only the states and controls. Also, the following formulation is

for a pitch angle tracking controller. The formulation for tracking any other state, say pitch

rate q, would be similar. The feedback to the controller will, in that case, be the tracking

error mn q.











Plot of Cd values from Database and Maytest Data
:- May Test Date
SDataba~se-Immersion=0J 1
-- Daab~ase-lImmrson0 3
--Database-Imerasron=0~ 5
Database-Immerston=0 7
DatabaSe-Immer~con=0 9


ne


i1 4-


:: L
c~z-





; Y
-


Figure 7.1 Calculation of Uncertainty


7.2 Synthesis Model

The state space formulation for longitudinal dynamics can be given as


x = Ax +tBu


y =Cx
where


x = {Au, Aw,A~q, AG}


u = {0,8e}


(7.1)


(7.2)


where, I4 represents an identity matrix of size 4 x 4. For simplicity, the B matrix is chosen

to be 4 x 2, i.e., other controls are assumed to be fixed at their trim values. Thus the only

controls are deflection of cavitator and the deflection of the elevators. The elevators are


A = [aij]4x4 B = [b,7]4x2









assumed to have a symmetric deflection for a longitudinal tracking command. Thus only

one elevator can be included in the synthesis model. Note that the symbol u is used to

represent the control vector. u will also be be used to represent the forward velocity of the

torpedo.

As described earlier, parametric uncertainty is assumed in the coefficients of lift(clc, cle)

and drag(cde, cde) of the cavitator and the elevators. Let W1 and W2 represent percentage

errors in cle and cde respectively. Let W3 and W4 represent percentage errors in clc and cdc

respectively. Thus a value of W1 = 0. 1 would mean a up to 10% error in value of cle and so

on. Now, the true value of these coefficients can be written as

cle = cle (l +t 6cieW1)

cde = cde(1+8t cdeW2)
(7.3)
cle = cle (1 +t ScicW3)

cdc = cdc(1+8t cdcW4)

where, cle, cde, cle and cdc are the values of these coefficients from the database, i.e., their

nominal values. 6ct and Scd are variables, whose numerical value determines the actual

error in these coefficients. Thus


(7.4)


Elements ofA and B matrices will be functions of function of coefficients of lift and drag of

cavitator, and some other terms which are of lesser importance for this analysis. In theory,

the parametric form can be written as explicit function of cl and cd i.e.

A = Ao tcleA t +cdeA2 +tclcA3 +tcdcA4
(7.5)
B = Bo +t cleB1 +t cdeB2 +t clcB3 +t cdcB4

This formulation of A and B matrices can be tough to obtain for general flight condition,

but it can be found numerically MATLAB, as described in the Appendix B. A symbolic









derivation of these matrices is done using MATHEMATICA, but it is found that the terms

are very long and the results runs through pages. The description of various codes has been

given in Appendix C. The numerical values from both the codes match, thus leading to a

verification of the numerical method. Now, using equation 7.3 in 7.5, we have

A = Ao+- cle (1+ ScieW1 )A t +cde ( 1+8cde W2)A2 tC, +l( 1+8t ciW3)A3+ -cdc (1+ -tcdc W4)A4

B = Bo +t cle ( 1 +t Sci W1)B1 +t cde ( 1 +t cde W2)B2 +t clc ( 1 +t 8ci W3)B3 +t cdc ( 1 +t cdc W4)B4

(7.6)



A = Ao +t6IciA t +8cdeA2 +8 cicA3 +8 cdcA4
(7.7)
B = Bo +t6IciB1 +8 cdeB2 +8 cicB3 +8 cdcB4

where

Ao = Ao +tcleA t +cdeA2 +tclcA3 +tcdcA4

Al = cleW1AI

A2 = cde W2A2 (7.8)

A3 = clcW3A3

A4 = cdcW4A4

and

Bo = Bo +tcleB1 +tcdeB2 +tclcB3 +tcdcB4

B1 = cleW1BI

B2 = cde W2B2 (7.9)

B3 = clcW3B3

B4 = cdcW4B4

With this transformation the state space form shown in equation 7.1 can be written as

x = (Ao +t6IciA t +8cdeA2 +8 cicA3 +8 cdcA4)X
(7.10)
+ (Bo +t6cieB1 +8 cdeB2 +8 cicB3 +8 cdcB4) M









In above equation, each of the terms corresponding to uncertainty, are like inputs to the

system. For example, 6clAix is like an input, and is of size 4 x 1. This term can be replace

by an input vector wl of same size. Also the term multiplying the uncertainty can be written

as an output of the system. Thus lets make a further transformation as follows


11l

212

213

214

wit



w13

W14


Aix


(7.11)


Scie 0 0 0

0 Scie 0 0

0 0 Scie 0

0 0 0 Scie


similarly


z2 = [Z2i]

z3 = [Z3i]

4 4Zi T

zg = [Zsi]

Zg6 6Zi T

z7 = [Z7i]"

z8 8zi T


[w2i]"

[wati]"

[wqi] T

[wsi]"

[wsi T

[7i] T

[W~i T


SBiut

A2x

B2Ue

= A3x

= B3u

A41

B421


= Diag[8ie]4x4 2

'Diag[6cde 4x4 3

: Diag[6cde 4x4 4

= Diag[8cic]4x4 5

= Diag[8cic]4x4 6

: Diag~[cT._, 1]Jx4 7

: Diag~[cT._, I]Jx4 8


(7.12)


for i=1 to 4, and Diag[] represents a diagonal matrix with the diagonal elements given by

those in [] and remaining elements as zeros. Using above transformation in equation 7.10,

we get a new state space form as


x = P~11x+Piu

P = P21x+-tP22u


(7.13)









where:

U = w" w12 11'3 11'4 5/' M6 M7 *'

7 4x (7.14)
I = :zl ~I z3 z4 5j 6 ;7 8g .716

and
P?11[ = .x4 (7.15)



pig = 4 4 4 4 4 4 4 4 ;13 (7.16)


At

041x4

AZ

04x4

=A3 (7. 17)

04x4

A4






04Ix32 04x2

041x32 R1


04x32 x


/ = 04x32 04x2,(. 8

04x32 L3

04 x32 .- x2

04x32

04x32 '. x2













I___ ___ ___ ____ ___I I_





x u
Figur 7.2Linea Frationl Repesenatio
Thus-- The--- ne sytmi 4iptad3 outu rsyse.Thsrpesnain oh

systm i kno asLiner Factl Reprsenttion an a esonbabokdarma
shown in figure 7.2.___


With- thsfomlain,Z aW snhsscnrlecabefudsdsribe n[6.Fg



ure 7.3 shows the sythsi model fr the system showsn i nation 71.T con o







feedback signal uncertainty, a multiplicative uncertainty is added in the feedback signal for

8. The W, represents the multiplicative uncertain in 8, whose values is W,A,,. Thus,


Oactual = S~oninal(1 +t WmA,,, (7.19)


Various filters have been added to system to put constraints on the control surfaces and

improve performance. These filters are like the Q and R matrices for and LQR control

synthesis. These choice of these filters done both by trial and error and also some analysis

of their frequency response.

* W,: This is chosen so as to penalize the performance error. This is usually a low pass
filter. Thus the filter has a high value at low frequencies and the value drops down at
higher frequencies. The value is obtained by trial and error.

WKc and WKe: These are the filter on control commands. These are usually high pass
filter. The control surface motion allowed is the inverse of these filter. Thus a higher
control motion will be allowed a low frequencies and low control motion at high
frequencies.












Isc 0 0
0 Is Sc4, O 0
0 0 Isc
0 0 0 I Scdc


Auto Pilot
Input Oo


Figure 7.3 Synthesis Model for pu Controller


\' :y: This signal is to add sensor noise in the system. In this synthesis, this is chosen
as a constant value. A better approximation of this can be done: by knowt~ing some:
sen sor character sticGs.

Ac: This filter .. an actuator model. This represents the actuator dynamics
in the frequency domain.

The ::i.::i d and .: Hi.=:: e for the synthesis model are


noise, ..-7-


(7.20)


ek
WK


d= w


e = z eg ek