Q/SZJ191001-2017振動(dòng)噪聲領(lǐng)域有限元模型魯棒性分析規(guī)范N&VFiniteElementModelRobustAnalysis20XX20XX-12-31實(shí)施20XX-11-27發(fā)布XXXX汽車工程技術(shù)有限公司發(fā)布Q/SZXX-X-2017振動(dòng)噪聲領(lǐng)域有限元模型魯棒性分析規(guī)范范圍本標(biāo)準(zhǔn)規(guī)定了術(shù)語和定義、魯棒性等方面內(nèi)容。本標(biāo)準(zhǔn)適用于CAE部門NVH領(lǐng)域魯棒性分析內(nèi)容。規(guī)范性引用文件無。術(shù)語和定義本規(guī)范涉及到的術(shù)語及其定義如下：NVHVibrationNoiseandharshness；魯棒性設(shè)計(jì)性能對設(shè)計(jì)參數(shù)不敏感的稱為魯棒性設(shè)計(jì)。摘要如果性能對設(shè)計(jì)參數(shù)不敏感即稱為魯棒性設(shè)計(jì)。例如，怠速振動(dòng)的魯棒性設(shè)計(jì)，意味著怠速振動(dòng)性能對于設(shè)計(jì)變量不敏感，這些設(shè)計(jì)變量有懸置剛度、部件公差、材料屬性偏差和裝配誤差等，因此，魯棒性的評價(jià)，就是研究怠速振動(dòng)性能的敏感因素問題。Adesigniscalledrobustwhenitsintendedperformanceisinsensitivetovariations.Arobustdesignforidleshake,forexample,meansthattheidleshakeperformancefromthedesignisinsensitivetovariationsincludingmountratevariation(e.g.designtolerances,manufacturingvariations,etc.),partvariation,materialpropertyvariation,assemblyprocessvariation,etc.Arobustnessassessment,therefore,istostudythesensitivityoftheidleshakeperformance.流程圖有很多不同的方法來研究魯棒性設(shè)計(jì)。Thereareafewdifferentmethodsthatcanbeusedtoinvestigatetherobustnessofadesign.設(shè)計(jì)空間探究DesignSpaceExploration:基于過去的經(jīng)驗(yàn)，在給定的設(shè)計(jì)空間中理解NVH性能最優(yōu)效率的途徑方法就是使用拉丁超立方體抽樣的試驗(yàn)設(shè)計(jì)。這種方法不但可以提供最小、最大和平均性能參數(shù)水平，也能得到在給定設(shè)計(jì)空間中是否有更好的設(shè)計(jì)參數(shù)。在設(shè)計(jì)空間中有較多局部最優(yōu)解時(shí)候，基于優(yōu)化算法的梯度很難找到一個(gè)全局優(yōu)化解，在這種條件下，這種方法也是一種重要的全局優(yōu)化選擇。最終的結(jié)果存儲在Isight的數(shù)據(jù)庫文件中。Basedonpastexperience,usingtheDOELatinHyperCubetypeofapproachisaveryefficientwayofunderstandingtheNVHperformanceinagivendesignspace.Thisapproachnotonlycangiveanideaoftheminimal,maximum,andmeanperformance,butalsoprovideswhetherornotthereisanybetterdesigninthedesignspace.Thisapproachisalsoanimportantglobaloptimizationalternativewhengradientbasedoptimizationlgorithmshavedifficultyfindingaglobaloptimalduetohavingmanylocalminimainthespace.TheendresultsaretobecapturedinaniSIGHTdatabasefile(*.db).響應(yīng)面建模和應(yīng)用ResponseSurfaceModelsandApplication:響應(yīng)面建模ResponseSurfaceGeneration:設(shè)計(jì)空間探究完成后，數(shù)據(jù)庫能用于產(chǎn)生響應(yīng)面，可用的方法有很多，例如Kriging插值法、逐步回歸法和移動(dòng)最小二乘法法。響應(yīng)面還可以進(jìn)一步應(yīng)用于魯棒性優(yōu)化。包含了貢獻(xiàn)量、累計(jì)貢獻(xiàn)量和主要因素上的統(tǒng)計(jì)分析（一組變量分析結(jié)果）是曲面擬合的副產(chǎn)品。OncetherunsaredoneintheDesignSpaceExplorationstep,thedatabasecanthenbeusedtogenerateresponsesurfaces.TherearemanyoptionsavailableincludingKriging,StepwiseRegression,andMovingLeastSquares.Thesurfacecanbefurtherusedforoptimizationandrobustoptimization.Statisticalbreakdowns(asetofANOVAresults)intermsofpercentcontribution,cumulativepercentcontribution,andmaineffectsaretheby-productsofthesurfacefitting.變量分析AnalysisofVarianceANOVA:包含了貢獻(xiàn)量、累計(jì)貢獻(xiàn)量和主要因素的靈敏度被計(jì)算和提供給用戶。貢獻(xiàn)量顯示的參數(shù)是有最大影響。主要因子圖顯示了性能和各設(shè)計(jì)參數(shù)之間的關(guān)系，或者是線性的或者是非線性的。累計(jì)貢獻(xiàn)率可以用于查看到哪些部件會(huì)對性能有影響。這些標(biāo)準(zhǔn)可以在Kriging手冊中查到。有兩種方法來獲得變量分析結(jié)果：一種來自于存在的響應(yīng)面，另一種通過數(shù)據(jù)庫文件生成。作為一種標(biāo)準(zhǔn)輸出格式，性能的平均值和標(biāo)準(zhǔn)偏差也可以輸出。Thesensitivityiscalculatedandprovidedtotheuserintermsofpercentcontribution,cumulativepercentcontribution,andmaineffects.Thepercentcontributionidentifieswhichparametersaremostinfluential.Andthemaineffectplotsshowtherelationshipbetweentheperformanceandtheparametersincludingwhetheritislinearornonlinear,isproportionalordisproportional,etc.Thecumulativepercentcontributioncanbeusedtodeterminehowmanytermshavesignificantlycontributedtotheperformancechange.ThesearestandardoutputfromtheKrigingWizard.TherearetwowaystogettheANOVAresults:oneisfromanexistingresponsesurface;anotheristogenerateitfromadatabasefile.Alsoasastandardoutput,themeanandstandarddeviationoftheperformanceareprovided.篩選Screening:基于變量分析的累計(jì)貢獻(xiàn)量，影響最大的設(shè)計(jì)變量被提供出來。該變量后續(xù)會(huì)用于魯棒性優(yōu)化計(jì)算。BasedoncumulativepercentcontributionofANOVA,themostinfluentialdesignvariablescanbedetermined.Themostinfluentialdesignvariablesarethentobeusedforrobustoptimization.魯棒性優(yōu)化RobustOptimization:在前一步中產(chǎn)生的響應(yīng)面基礎(chǔ)上進(jìn)行魯棒性優(yōu)化計(jì)算。這一步中，魯棒性設(shè)計(jì)可以較快的完成。一旦完成分析，需要使用NASTRAN求解器進(jìn)行驗(yàn)證計(jì)算工作。目標(biāo)函數(shù)通常是方向盤和座椅的直接響應(yīng)。通常，響應(yīng)曲線通常和目標(biāo)曲線不同，前一種情況下，優(yōu)化值越小越好，后一種情況下，需要最小化最大的正向差異。建議使用多個(gè)變量進(jìn)行優(yōu)化設(shè)計(jì)（例如10個(gè)設(shè)計(jì)變量）。設(shè)計(jì)變量的上限和下限均是可行性的設(shè)計(jì)空間范圍。Robustoptimizationisdoneontheresponsesurfacesthataregeneratedfrompreviousstep.Thisway,robustdesignscanbeobtainedquickly.Oncetheyareobtained,avalidationneedstobedoneusingactualNastranruns.Theobjectivefunctionsare,veryoftenthedirectresponsesofsteeringwheelandseats.Manytimes,theobjectivefunctionsarethedifferencesbetweentheresponsesandthetargetcurve.Fortheformercase,theoptimalissmallerthebetter.Forthelaterone,itistominimizethelargestpositivedifferences(deltas).Itissuggestedtousemultipledesignalternatives(e.g.top10designalternatives)fromthedesignspaceexplorationsteptoassesstheirrobustness(e.g.mean,standarddeviation,andS/Nratios,etc.).Theconstraintsforthedesignparametersarethefeasibledesignspaceboundaryintermsoftheirlowerandupperbounds.工具描述整車分析中最常用的工具是MSC.Nastran軟件?？梢允褂貌考B(tài)綜合法求解來減小運(yùn)行時(shí)間。流程整合、流程自動(dòng)化和一整套的優(yōu)化過程可以使用優(yōu)化軟件Isight。ThebasictoolforvehiclesystemmodelisMSC/Nastran.TheuseofComponentModalSynthesis(CMS)oftheNastranmodelispreferredtodecreaseruntimestoamanageablelevel.Aprocessintegration,processautomation,andasuiteofoptimizationprogramssuchasiSIGHTisnecessary.建模和分析過程單位Units分析中所用單位制如下所示:Theunitsusedintheanalysisareasfollows:力(Force) N質(zhì)量(Mass) Mg長度(Length) mm時(shí)間(Time) S內(nèi)容ContentsNASTRAN中應(yīng)該包含代表噪聲變量的單元和卡片，例如CELASi卡片和PELISi卡片等。這些卡片不能包含在部件模態(tài)綜合法建模的超單元中。這些卡片需要使用文本格式記錄，以便優(yōu)化軟件Isight可以識別使用。TheNastrandatadeckshouldcontainelementsandcardsrepresentingdesignandnoisevariables,suchasCELASicards,PELASicards,etc.ThosecardsshouldNOTbeincludedinthe