机械外文文献及翻译.docx

上传人:b****6 文档编号:15627810 上传时间:2023-07-06 格式:DOCX 页数:14 大小:248.48KB
下载 相关 举报
机械外文文献及翻译.docx_第1页
第1页 / 共14页
机械外文文献及翻译.docx_第2页
第2页 / 共14页
机械外文文献及翻译.docx_第3页
第3页 / 共14页
机械外文文献及翻译.docx_第4页
第4页 / 共14页
机械外文文献及翻译.docx_第5页
第5页 / 共14页
机械外文文献及翻译.docx_第6页
第6页 / 共14页
机械外文文献及翻译.docx_第7页
第7页 / 共14页
机械外文文献及翻译.docx_第8页
第8页 / 共14页
机械外文文献及翻译.docx_第9页
第9页 / 共14页
机械外文文献及翻译.docx_第10页
第10页 / 共14页
机械外文文献及翻译.docx_第11页
第11页 / 共14页
机械外文文献及翻译.docx_第12页
第12页 / 共14页
机械外文文献及翻译.docx_第13页
第13页 / 共14页
机械外文文献及翻译.docx_第14页
第14页 / 共14页
亲,该文档总共14页,全部预览完了,如果喜欢就下载吧!
下载资源
资源描述

机械外文文献及翻译.docx

《机械外文文献及翻译.docx》由会员分享,可在线阅读,更多相关《机械外文文献及翻译.docx(14页珍藏版)》请在冰点文库上搜索。

机械外文文献及翻译.docx

机械外文文献及翻译

与机械相关的外文及翻译

MultidisciplinaryDesignOptimization

ofModularIndustrialRobots by

UtilizingHighLevel CAD Templates

1、Introduction

Inthe designofcomplex andtightlyintegrated engineering products, it isessential tobeable to handleinteractions betweendifferentsubsystemsof multidisciplinarynature[1].Toachieveanoptimaldesign, aproductmust betreatedas acomplete systeminstead of developingsubsystemsindependently[2].MDO hasbeenestablishedasa convincingconcurrentdesignoptimizationtechnique indevelopmentofsuch complexproducts[3,4].

Furthermore,ithasbeenpointedoutthat,regardlessofdiscipline,basicallyall analysesrequireinformationthathastobeextractedfromageometrymodel[5].Hence,accordingto Bow-cutt[1],in ordertoenable integrateddesignanalysisandoptimizationitisofvitalimportancetobeabletointegrateanautomatedparametricgeometry generationsystemintothedesignframework.The automatedgeometrygenerationisakeyenabler for so-calledgeometry-in-the-loop[6]multidisciplinarydesignframeworks,wherethe CADgeometries canserveasframeworkintegratorsfor other engineering tools.

Toeliminatenoncreativework,methods forcreationandautomaticgenerationofHLCthavebeensuggestedbyTarkian[7].The principleofhighHLCtsis similartohighlevelprimitives(HLP)suggestedby LaRoccaand van Tooren [8],with theexceptionthatHLCtsarecreatedandutilizedin a CAD environment.Otherwise,thebasicsof bothHLPand HLCtcan,assuggestedbyLaRocca,be compared toparametric LEGOVRblockscontainingaset of designandanalysisparameters.Theseareproducedandstoredinlibraries,givingengineersor a computeragentthe possibility tofirsttopologicallyselect thetemplatesandthenmodifythemorphology,meaning the

shape,ofeachtemplate parametrically.

2、MultidisciplinaryDesign Framework

MDOisa“systematic approachto designspaceexploration”[17], theimplementationofwhichallowsthedesignertomaptheinterdisciplinaryrelations that existin asystem.Inthis work,the MDO framework consistsofa geometrymodel,a finiteelement(FE)model,a dynamicmodel andabasic costmodel.The geometrymodel providesthe analysistools withgeometricinput.The dynamicmodelrequiresmass properties suchasmass, centerofgravity,andinertia. TheFEmodelneeds the meshedgeometryoftherobotaswellastheforce andtorqueinteractionsbased onresultsof dynamicsimulations.

Highfidelity modelsrequire an extensiveevaluationtimewhichhas betakenintoaccount. Thisshortcomingis addressedbyapplying surrogatemodels forthe FE andthe CADmodels. Themodelsarebrieflypresentedbelow.

2.1HighLevel CADTemplate—GeometryModel

Traditionally,parametricCADismainly focused onmorphologicalmodificationsofthegeometry.However,thereisalimittomorphologicalparameterizationasfollows:

•Thegeometriescannotberadicallymodified.

•Increasedgeometriccomplexitygreatlyincreasesparameterizationcomplexity.

Thegeometrymodel oftherobotisgeneratedwith presavedHLCts,createdinCATIAV5.These are topologically instantiated withuniqueinternaldesignvariables. Topologicalparameterizationallowsdeletion, modification, andadditionofgeometricelementswhichleadsto a muchgreaterdesignspacecaptured.ThreetypesofHLCtsareusedto definetheindustrial robot topologically; DatumHLCtwhichincludeswireframe references requiredforplacementfor the ActuatorHLCTsandStructureHLCts,asseenFig.2.

Fig.2Anindustrialrobot(left)anda modularindustrial robot(right)

Thenamesofthereferencesthatmustbeprovidedforeach HLCtinstantiationare storedin theknowledgebase(seeAppen-dix A.4),whichissearchedthroughbytheinferenceengine. InAppendix A, pseudocodeexamplesdescribes howthereferencesareretrieved andhowtheyarestoredintheknowledgebase.

The processstartsbytheuser definingthenumberofdegreesoffreedom(DOF)oftherobot (seeFig. 3)and isrepeated untilthenumberofaxis(i)isequal totheuserdefined DOF.InordertoinstantiatethefirstStructure HLCt,twoDatum and twoactuatorinstances areneeded.ReferencesfromthetwoDatuminstanceshelporienting the structure inspace,whilethe geometries ofthe actuator instances, atboth endsofthelink,areusedtoconstructthe actuatorattachments,asseeninFigs.2 and3.Fortheremaininglinks,onlyonenewinstanceofbothdatum andactuator HLCts are required, sincethedatumandactuatorinstancesfromadjacent linksarealreadyavailable.AppendixA.2showsapseudocodeexampleofaninstantiationfunction.The first instantiateddatumHLCt is definedwithreferenceto theabsolutecoordinatesystem. TheremainingdatumHLCtinstances are placedinasequential order,wherethecoordinatesystem of previous instancesisusedasreference fordefiningthepositionin spaceaccordingtouser inputs(seealsoAppendixA.3). Furthermore,thetype ofeach actuatorand structureinstanceisuserdefined.

Fig.3 The highlevelCADtemplateinstantiationprocess

SinceitispossibletocreatenewHLCts inthe utilizedCADtool,the usersarenotforcedto merelychoosefrom the templates available. NewHLCtscanbecreated,placedinthedatabaseandparametricallyinserted intothemodels.

2.2DynamicModel

 Theobjective ofperformingdynamicsimulationofarobotistoevaluatesystemperformance,suchas predictingaccelerationand timeperformance,butit alsoyieldsloadsoneach actuatedaxis,neededforactuatorlifetimecalculationsandsubsequentstressanalysisbasedonFEcalculations.ThedynamicmodelintheoutlinedframeworkisdevelopedinModelicausingDymola, anditconstitutesaseven-axisrobotarmbasedon theModelicaStandard library[18].

 Thedynamic modelreceives inputfromthe geometry model,aswell asprovidingoutputtotheFEmodel, whichisfurtherdescribedinSec. 2.3.However,tobetterunderstandthecouplingsbetweenthemodels,theNewton–Eulerformulationwillbebrieflydiscussed. Inthisformulation,thelink velocitiesandacceleration areiterativelycomputed,forwardrecursively

When thekinematicpropertiesarecomputed,theforceandtorque interactions betweenthelinksarecomputed backward recursivelyfromthe lastto thefirstlink

2.3 FE SurrogateModel

  To computethestructural strengthoftherobot,FEmodels foreachrobotlink is createdutilizingCATIAV5,see Fig.4.ForeachHLCt, mesh andboundaryconditionsaremanuallypreprocessedinordertoallowforsubsequentautomation forFE-modelcreation.The timespentonpreprocessingeachFE-modelis thusextensive.Nonetheless,the obtained parametric FE-modelpavesway forautomatedevaluation ofawide span ofconcepts. Each robotlinkisevaluatedseparatelywiththeloadconditionsextractedfromthedynamicmodel.The force (fi-11andfi)andtorque(ţi-1andti) areapplied onthesurfaces wheretheactuators are attached.

2.4GeometricSurrogateModels.

  Surrogatemodelsarenumericallyefficientmodelstodeterminethe relationbetweeninputsandoutputsofa model[19].Theinput variablesforthe proposed applicationarethe morphologicalvariables thickness and linkheight as wellasatopologicalvariableactuatortype.Theoutputs ofthe surrogate modelsaremass m,InertiaI,andcenter ofgravityri,ci.

 Toidentifythemostsuitable type ofsurrogate modelfortheoutlinedproblem,arangeofsurrogate modelstypes are created andevaluatedusing 50 samples.Theprecisionof each surrogatemodeliscomparedwiththevaluesof theoriginalmodelwith 20new samples.Thecomparisonis madeusingtherelativeaverageabsoluteerror(RAAE)andrelativemaximum absoluteerror(RMAE)as specifiedbyShanetal.[20], aswellasthenormalizedrootmean squareerror(NRMSE), calculatedasseen in Eq.(3).All precisionmetricsare desiredtobe aslowaspossible,sincelowvalues mean thatthesurrogatemodelisaccurate

The resultingprecisionmetricscan beseen inAppendixB andthegeneral conclusionisthatanisotropic kriging[21],neuralnetworks [22],andradialbasisfunctions[23]arethe mostpromisingsurrogatemodels. Toinvestigatethe impactof increasing number ofsamples, additionalsurrogatemodelsofthosethree arefittedusing100 samples, andtheresultscompiledinAppendixB.TheresultingNRMSEsfor50and100samples foranistotropickriging,neural networks,andradialbasisfunctionscanbeseenin Fig.5.The figuresinsidetheparenthesesindicate the number ofsamples usedtofitthesurrogatemodels.

Fig.5 Graphofthe NRMSEsfordifferentsurrogate models,

fittedusing50 and 100samples

AccordingtoFig. 5,anisotropickriging outperforms the othersurrogatemodelsandthedoublingof thenumber of samples usedforfittingthe surrogatemodelincreasestheprecisiondramatically.

2.5FE SurrogateModels

ForgeneratingFEsurrogatemodels,theanisotropickriging wasalsoproventobe themostaccuratecomparedtothe methodsevaluatedin Sec.2.4.Here, one surrogatemodel iscreated foreachlink.Inputsarethickness, actuators,force (fi-11andfi)and torque (ţi-1and ti).Theoutputforeach surrogatemodelismaximumstress (MS).Ameanerrorof approximately9%isreachedwhenrunning1400 samplesfor eachlink.Thereasonfor thevast numberofsamples, comparedtogeometrysurrogatemodels,hastodowithamuchlargerdesignspace.

利用高水平CAD模板进行

模块化工业机器人的多学科设计优化

1介绍

指出,除了规则,基本上所有的分析都需要信息,而这些信息需要从一个几何模型中提取。

因此,根据Bowcutt[1]中,为了使综合设计分析和优化,最重要的是能够将在设计的复杂和紧密集成的工程产品的过程中,必须要有能力处理不同的子系统的多学科性质之间的相互作用。

达到一个最优的设计,一个产品必须被视为一个完整的系统,而不是正在开发子的独立系统。

此外,已经一个自动化的参数化几何生成系统融入到设计框架中。

自动化的几何生成对于所谓几何循环多学科设计框架是一个关键驱动因素,在这个框架中CAD几何图形可以作为框架连接者来连接其他工程工具。

消除没有创新的工作,Tarkian已经提出了创造和生成HLC

展开阅读全文
相关资源
猜你喜欢
相关搜索
资源标签

当前位置:首页 > 经管营销 > 经济市场

copyright@ 2008-2023 冰点文库 网站版权所有

经营许可证编号:鄂ICP备19020893号-2