机械外文文献及翻译.docx
- 文档编号:24784680
- 上传时间:2023-06-01
- 格式:DOCX
- 页数:14
- 大小:243.95KB
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)》请在冰豆网上搜索。
机械外文文献及翻译
与机械相关的外文及翻译
MultidisciplinaryDesign Optimization
ofModular IndustrialRobotsby
UtilizingHighLevelCADTemplates
1、Introduction
Inthedesignofcomplexandtightlyintegratedengineeringproducts, itisessentialtobe abletohandleinteractionsbetweendifferent subsystemsofmultidisciplinarynature [1].To achieveanoptimaldesign,a productmustbe treatedasacompletesysteminstead ofdeveloping subsystemsindependently [2].MDOhasbeenestablished asa convincingconcurrent design optimization techniqueindevelopmentof suchcomplexproducts [3,4].
Furthermore,ithas beenpointedout that,regardless ofdiscipline, basicallyall analysesrequireinformationthathastobeextractedfroma geometrymodel[5].Hence, accordingtoBow-cutt[1],inordertoenableintegrateddesign analysis andoptimizationitisofvitalimportance tobeable to integratean automatedparametricgeometrygenerationsystemintothedesign framework.Theautomated geometrygenerationis a key enabler for so—calledgeometry-in—the-loop[6]multidisciplinary designframeworks,where theCADgeometriescan serveasframeworkintegratorsfor other engineeringtools.
Toeliminatenoncreativework,methodsforcreationandautomaticgenerationofHLCthavebeen suggestedbyTarkian[7].The principleofhigh HLCtsissimilartohighlevelprimitives(HLP) suggestedbyLaRocca andvanTooren [8],with theexception that HLCtsarecreatedandutilizedinaCAD environment。
Otherwise,thebasicsofbothHLP andHLCtcan,assuggestedbyLaRocca, becomparedto parametricLEGOVRblockscontaininga setofdesignandanalysisparameters.Theseareproduced andstoredinlibraries, givingengineersor a computer agent thepossibilitytofirsttopologicallyselectthetemplatesand thenmodifythemorphology,meaningthe
shape,of each templateparametrically.
2、MultidisciplinaryDesignFramework
MDOisa“systematicapproachto designspaceexploration”[17], theimplementationofwhichallows thedesignertomap theinterdisciplinaryrelationsthatexistinasystem.Inthiswork,theMDO framework consistsofageometrymodel,afinite element(FE)model,adynamicmodelandabasiccostmodel.Thegeometry modelprovidestheanalysis toolswithgeometricinput。
Thedynamic modelrequires massproperties such as mass,centerofgravity,and inertia。
TheFEmodelneeds the meshedgeometry ofthe robot aswellas theforceandtorque interactions basedonresultsofdynamicsimulations。
Highfidelitymodels requireanextensiveevaluation timewhichhasbetaken intoaccount。
This shortcoming isaddressedbyapplying surrogate modelsforthe FEandtheCAD models.Themodelsarebriefly presentedbelow.
2。
1 HighLevel CADTemplate-GeometryModel
Traditionally,parametricCADismainlyfocused onmorphologicalmodifications of thegeometry.However,thereisalimitto morphologicalparameterizationasfollows:
•The geometriescannotberadicallymodified.
•Increasedgeometric complexity greatly increasesparameterizationcomplexity。
The geometrymodel ofthe robotis generatedwith presaved HLCts,createdinCATIAV5. These aretopologicallyinstantiatedwithuniqueinternal designvariables。
Topological parameterizationallowsdeletion,modification, and additionofgeometricelementswhichleads to a muchgreaterdesign spacecaptured.Threetypes ofHLCtsareused todefinetheindustrial robottopologically;DatumHLCtwhich includeswireframe references requiredforplacementfor theActuatorHLCTsandStructure HLCts,asseenFig.2.
Fig.2Anindustrial robot(left) and amodular industrialrobot(right)
The namesofthereferencesthatmust beprovided foreachHLCtinstantiationarestoredin theknowledge base(see Appen—dix A.4),whichis searchedthroughbytheinference engine. InAppendixA,pseudocodeexamplesdescribeshowthereferences areretrievedand how theyarestoredin theknowledgebase.
Theprocess startsby the user definingthenumberof degreesoffreedom(DOF)oftherobot (seeFig.3)andis repeated untilthe numberofaxis (i)is equal totheuserdefinedDOF。
In orderto instantiate thefirstStructureHLCt, two Datumandtwoactuator instancesareneeded. ReferencesfromthetwoDatuminstanceshelp orientingthe structureinspace,whilethegeometriesoftheactuator instances, atbothends ofthelink,areused to construct theactuator attachments,as seeninFigs. 2and3.Fortheremaininglinks,only onenewinstanceofbothdatumandactuatorHLCtsarerequired,sincethedatum andactuator instancesfrom adjacent linksare alreadyavailable.AppendixA.2shows apseudocode exampleofan instantiationfunction.The firstinstantiateddatumHLCtisdefinedwithreferenceto the absolutecoordinatesystem.The remainingdatumHLCt instances areplaced inasequentialorder,wherethecoordinate systemofpreviousinstances isused asreferencefordefining thepositioninspaceaccordingtouserinputs(seealsoAppendixA。
3).Furthermore,thetypeofeachactuator andstructureinstanceisuserdefined。
Fig.3 Thehighlevel CADtemplateinstantiationprocess
Since itispossibletocreate newHLCtsintheutilizedCAD tool,theusersarenot forcedtomerelychoosefromthetemplatesavailable.NewHLCts canbecreated,placedinthedatabaseandparametricallyinsertedinto themodels。
2。
2DynamicModel
The objectiveof performingdynamicsimulationofa robotis toevaluatesystemperformance,suchaspredictingaccelerationandtime performance, butitalsoyieldsloadson eachactuated axis,neededforactuatorlifetimecalculations andsubsequentstressanalysisbasedonFEcalculations。
Thedynamicmodelintheoutlinedframework isdeveloped in Modelicausing Dymola,andit constitutesaseven-axisrobotarm based ontheModelicaStandardlibrary[18].
The dynamicmodelreceivesinputfromthegeometry model,as wellas providing outputto theFEmodel,whichis furtherdescribedin Sec。
2.3.However,to betterunderstandthecouplings betweenthemodels, theNewton–Euler formulationwillbebriefly discussed。
Inthisformulation, thelink velocitiesandaccelerationareiteratively computed,forwardrecursively
When the kinematicproperties arecomputed,theforceand torqueinteractions betweenthelinksare computedbackwardrecursivelyfromthelasttothefirst link
2。
3FE SurrogateModel
Tocomputethestructural strengthoftherobot,FE models foreach robotlinkis createdutilizingCATIAV5,seeFig. 4。
ForeachHLCt,meshandboundary conditionsaremanuallypreprocessedinordertoallowforsubsequentautomationforFE—modelcreation. The timespenton preprocessing each FE-modelisthusextensive.Nonetheless,theobtainedparametricFE-model paveswayforautomatedevaluationofawidespanofconcepts.Eachrobotlinkis evaluatedseparatelywiththe load conditionsextracted fromthedynamicmodel.The force (fi—11and fi)andtorque(ţi—1andti)areappliedonthesurfaceswheretheactuatorsareattached。
2。
4GeometricSurrogateModels.
Surrogatemodelsare numericallyefficientmodelstodetermine therelationbetweeninputsand outputsof a model[19]。
The inputvariables fortheproposedapplicationarethemorphological variablesthickness andlinkheightas well asatopologicalvariableactuator type。
Theoutputsofthesurrogatemodelsaremassm, InertiaI,and centerof gravity ri,ci。
Toidentifythemostsuitabletypeofsurrogatemodel forthe outlinedproblem,arangeofsurrogatemodels types arecreatedandevaluatedusing50samples. Theprecisionofeachsurrogatemodelis comparedwiththevaluesoftheoriginalmodelwith20newsamples.The comparison ismade usingthe relative averageabsolute error (RAAE)andrelativemaximumabsoluteerror(RMAE)asspecifiedbyShan etal.[20],aswellasthenormalizedroot meansquareerror(NRMSE), calculated asseeninEq。
(3).Allprecisionmetrics aredesiredtobeaslowas possible,sincelowvaluesmeanthatthesurrogatemodelisaccurate
TheresultingprecisionmetricscanbeseeninAppendixBandthe generalconclusionisthatanisotropic kriging[21],neuralnetworks[22],and radialbasis functions[23]arethe most promisingsurrogatemodels。
To investigate theimpactofincreasingnumberofsamples, additionalsurrogate models ofthosethreearefittedusing100samples,andtheresultscompiledinAppendix B。
TheresultingNRMSEsfor50and100samplesforanistotropickriging,neuralnetworks, andradial basisfunctions canbeseeninFig。
5。
Thefiguresinsidetheparenthesesindicatethenumber ofsamplesusedtofitthesurrogatemodels。
Fig。
5 GraphoftheNRMSEs fordifferentsurrogatemodels,
fittedusing50and100 samples
According toFig.5,anisotropickriging outperformstheothersurrogatemodels andthedoublingof thenumberof samplesusedforfitting thesurrogatemodelincreases theprecisiondramatically。
2。
5 FE SurrogateModels
ForgeneratingFEsurrogatemodels,theanisotropickrigingwasalso provento bethemost accuratecomparedtothemethodsevaluatedinSec.2。
4。
Here,one surrogatemodeliscreated for eachlink.Inputsare thickness,actuators, force(fi-11andfi)andtorque (ţi-1andti)。
Theoutputforeachsurrogatemodelis maximumstress(MS).A mean errorof approximately9%is reached whenrunning1400samplesforeach link。
Thereasonforthe vastnumber ofsamples,comparedtogeometrysurrogatemodels,has todo withamuchlarger designspace.
ﻬ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HLCt的
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 机械 外文 文献 翻译