船舶专业外文翻译船舶设计优化.docx

船舶专业外文翻译船舶设计优化.docx

《船舶专业外文翻译船舶设计优化.docx》由会员分享，可在线阅读，更多相关《船舶专业外文翻译船舶设计优化.docx（9页珍藏版）》请在冰豆网上搜索。

船舶专业外文翻译船舶设计优化

ShipDesignOptimization

ThiscontributionisdevotedtoexploitingtheanalogybetweenamodernmanufacturingplantandaheterogeneousparallelcomputertoconstructaHPCNdecisionsupporttoolforshipdesigners.TheapplicationisaHPCNonebecauseofthescaleofshipbuilding-alargecontainervesselisconstructedbyassemblingabout1.5millionatomiccomponentsinaproductionhierarchy.Theroleofthedecisionsupporttoolistorapidlyevaluatethemanufacturingconsequencesofdesignchanges.Theimplementationasadistributedmulti-agentapplicationrunningontopofPVMisdescribed

1AnalogiesbetweenManufacturingandHPCN

Thereareanumberofanalogiesbetweenthemanufactureofcomplexproductssuchasships,aircraftandcarsandtheexecutionofaparallelprogram.Themanufactureofashipiscarriedoutaccordingtoaproductionplanwhichensuresthatallthecomponentscometogetherattherighttimeattherightplace.Aparallelcomputerapplicationshouldensurethattheappropriatedataisavailableontheappropriateprocessorinatimelyfashion.

Itisnotsurprising,therefore,thatmanufacturingisplaguedbyindeterminacyexactlyasareparallelprogramsexecutingonmulti-processorhardware.Thishascausedanumberofresearchersinproductionengineeringtoseekinspirationinotherareaswheremanagingcomplexityandunpredictabilityisimportant.Anumberofnewparadigms,suchasHolonicManufacturingandFractalFactorieshaveemerged[1,2]whichcontainideasratherreminiscentofthosetobefoundinthefieldofMulti-AgentSystems[3,4].

Manufacturingtasksareanalogoustooperationscarriedoutondata,withinthecontextofplanning,schedulingandcontrol.Also,complexproductsareassembledatphysicallydistributedworkshopsorproductionfacilities,sothecomponentsmustbetransportedbetweenthem.Thisisanalogoustocommunicationofdatabetweenprocessorsinaparallelcomputer,whichthusalsomakescleartheanalogybetweenworkshopsandprocessors.

Theremainderofthispaperreportsanattempttoexploitthisanalogytobuildaparallelapplicationforoptimizingshipdesignwithregardtomanufacturingissues.

2ShipbuildingatOdenseSteelShipyard

OdenseSteelShipyardissituatedinthetownofMunkeboontheislandofFunen.Itisrecognizedasbeingoneofthemostmodernandhighlyautomatedintheworld.It

specializesinbuildingVLCC's（supertankers）andverylargecontainerships.Theyardwasthefirstintheworldtobuildadoublehulledsupertankerandiscurrentlybuildinganorderof15ofthelargestcontainershipseverbuiltfortheMaerskline.Thesecontainershipsareabout340metreslongandcancarryabout7000containersatatopspeedof28knotswithacrewof12.

OdenseSteelShipyardismorelikeashipfactorythanatraditionalshipyard.Theshipdesignisbrokendownintomanufacturingmoduleswhichareassembledandprocessedinanumberofworkshopsdevotedto,forexample,cutting,weldingandsurfacetreatment.Atanyonetime,upto3identicalshipsarebeingbuiltandanewshipislaunchedaboutevery100days.

Theyardsurvivesintheverycompetitiveworldofshipbuildingbyextensiveapplicationofinformationtechnologyandrobots,sotherearecurrentlyabout40robotsattheyardengagedinvariousproductionactivities.Theyardhasacommitmenttoresearchaswell,sothatthereareabout10industrialPh.D.studentsworkingthere,whoareenrolledatvariousengineeringschoolsinDenmark.

3Tomorrow'sManufacturingSystems

ThepenetrationofInformationTechnologyintoourliveswillalsohaveitseffectinmanufacturingindustry.Forexample,theInternetisexpectedtobecomethedominanttradingmediumforgoods.Thismeansthatthecustomercancomeintodirectdigitalcontactwiththemanufacturer.

Thedirectdigitalcontactwithcustomerswillenablethemtoparticipateinthedesignprocesssothattheygetaproductoverwhichtheyhavesomeinfluence.Theelementofunpredictabilityintroducedbytakingintoaccountcustomerdesiresincreasestheneedforflexibilityinthemanufacturingprocess,especiallyinthelightofthetendencytowardsglobalizationofproduction.Intelligentrobotsystems,suchasAMROSE,relyonthedigitalCADmodelastheprimarysourceofinformationabouttheworkpieceandtheworkcell[5,6].Thisinformationisusedtoconstructtaskperforming,collisionavoidingtrajectoriesfortherobots,whichbecauseofthehighprecisionoftheshipbuildingprocess,canbecorrectedforsmalldeviationsoftheactualworldfromthevirtualoneusingverysimplesensorsystems.Thetrajectoriesaregeneratedbynumericallysolvingtheconstrainedequationsofmotionforamodeloftherobotmovinginanartificialforcefielddesignedtoattractthetoolcentretothegoalandrepellitfromobstacles,suchastheworkpieceandpartsofitself.Finally,therearelimitstowhatonecangetarobottodo,sotheactualmanufacturingwillbeperformedasacollaborationbetweenhumanandmechatronicagents.

Mostindustrialproducts,suchasthewindmillhousingcomponentshowninFig.1,aredesignedelectronicallyinavarietyofCADsystems.

Fig.1.ShowingtheCADmodelforthehousingofawindmill.Themodel,madeusingBentleyMicrostation,includesboththework-pieceandtask-curvegeometries.

4Today'sManufacturingSystems

Theabovescenarioshouldbecomparedtotoday'srealitiesenforcedbytraditionalproductionengineeringphilosophybasedontheideasofmassproductionintroducedabout100yearsagobyHenryFord.Atypicalproductionlinehasthesamestructureasaserialcomputerprogram,sothatthewholeprocessisdrivenbyproductionrequirements.Thisrigidityisreflectedonthetypesoftop-downplanningandcontrolsystemsusedinmanufacturingindustry,whicharebadlysuitedtobothcomplexityandunpredictability.

Infact,themanufacturingenvironmenthasalwaysbeencharacterizedbyunpredictability.Today'smanufacturingsystemsarebasedonidealizedmodelswhereunpredictabilityisnottakenintoaccountbuthandledusingcomplexandexpensivelogisticsandbufferingsystems.

Manufacturersarealsobecomingawarethatoneoftheresultsofthetop-downserialapproachisanalienationofhumanworkers.Forexample,someofthecarmanufacturershaveexperimentedwithhavingteamsofhumanworkersresponsibleforaparticularcarratherthanperformingrepetitiveoperationsinaproductionline.Thismodelinfactbetterreflectstheconcurrencyofthemanufacturingprocessthantheassemblyline.

5ADecisionSupportToolforShipDesignOptimization

Largeshipsare,togetherwithaircraft,someofthemostcomplexthingseverbuilt.Acontainershipconsistsofabout1.5millionatomiccomponentswhichareassembledinahierarchyofincreasinglycomplexcomponents.ThusanysupporttoolforthemanufacturingprocesscanbeexpectedtobealargeHPCNapplication.

Shipsaredesignedwithbothfunctionalityandeaseofconstructioninmind,as

wellasissuessuchaseconomy,safety,insuranceissues,maintenanceandevendecommissioning.Onceafunctionaldesignisinplace,astepwisedecompositionoftheoveralldesignintoahierarchyofmanufacturingcomponentsisperformed.Themanufacturingprocessthenstartswiththeindividualbasicbuildingblockssuchassteelplatesandpipes.Thesebuildingblocksareputtogetherintoevermorecomplexstructuresandfinallyassembledinthedocktoformthefinishedship.

Thusaveryusefulthingtoknowassoonaspossibleafterdesigntimearethemanufacturingconsequencesofdesigndecisions.Thisincludesissuessuchaswhethertheintermediatestructurescanactuallybebuiltbytheavailableproductionfacilities,theimplicationsontheuseofmaterialandwhetherornottheproductioncanbeefficientlyscheduled[7].

Fig.2.showsschematicallyhowaredesigndecisionatapointintimeduringconstructionimpliesfuturecosts,onlysomeofwhichareknownatthetime.Thusadecisionsupporttoolisrequiredtogivebetterestimatesoftheimpliedcostsasearlyaspossibleintheprocess.

Simulation,bothofthefeasibilityofthemanufacturingtasksandtheefficiencywithwhichthesetaskscanbeperformedusingtheavailableequipment,isaverycompute-intenseapplicationofsimulationandoptimization.Inthenextsection,wedescribehowadecisionsupporttoolcanbedesignedandimplementedasaparallelapplicationbymodelingthemainactorsintheprocessasagents.

Fig.2.Economicconsequencesofdesigndecisions.Adesigndecisionimpliesafuturecommitmentofeconomicresourceswhichisonlypartiallyknownatdesigntime.

6Multi-AgentSystems

Thenotionofasoftwareagent,asortofautonomous,dynamicgeneralizationofanobject（inthesenseofObjectOrientation）isprobablyunfamiliartothetypicalHPCNreaderintheareaofscientificcomputation.Anagentpossessesitsownbeliefs,desiresandintentionsandisabletoreasonaboutandactonitsperceptionofotheragentsandtheenvironment.

Amulti-agentsystemisacollectionofagentswhichtrytocooperatetosolvesomeproblem,typicallyintheareasofcontrolandoptimization.Agoodexampleistheprocessoflearningtodriveacarintraffic.Eachdriverisanautonomousagentwhichobservesandreasonsabouttheintentionsofotherdrivers.Agentsareinfactaveryusefultoolformodelingawiderangeofdynamicalprocessesintherealworld,suchasthemotionofproteinmolecules[8]ormulti-linkrobots[9].Forotherapplications,see[4].

Oneoftheinterestingpropertiesofmulti-agentsystemsisthewayglobalbehaviorofthesystememergesfromtheindividualinteractionsoftheagents[10].Thenotionofeme