Structural Dynamics Toolbox Primer: SDT Primer University of Bristol
Structural Dynamics Toolbox Primer: SDT Primer University of Bristol
Structural Dynamics Toolbox Primer: SDT Primer University of Bristol
UniversityofBristol
1/47
SDTPrimer
UniversityofBristol
FEModelling...............................................................................................................4
1.1 SDTdatastructure ................................................................................................ 5 1.2 Viewingmodelwithfeplot .................................................................................. 5 1.3 Meshingandmodelmanipulations.................................................................... 5 1.3.1 Explicitdefinition.............................................................................................. 5 1.3.2 FunctionalDefinition ....................................................................................... 7 1.4 Elements ............................................................................................................... 13 1.4.1 Rigidelements................................................................................................. 13 1.4.2 Masselements ................................................................................................. 14 1.4.3 ElementPropertyManipulation ................................................................... 14 1.4.4 Shelland3Delements .................................................................................... 15 1.5 Materialandelementproperties....................................................................... 15 1.5.1 MaterialProperties ......................................................................................... 15 1.5.2 SectionProperties ........................................................................................... 16 1.6 Loadsandboundaryconditions ....................................................................... 18 1.6.1 BoundaryConditions ..................................................................................... 18 1.7 Solving .................................................................................................................. 20 1.8 Postprocessing.................................................................................................... 23 1.8.1 feplot ................................................................................................................. 23 1.8.2 iiplot.................................................................................................................. 25 1.9 SDTmeshingexample........................................................................................ 25
Testing................................................................................................................27
2.1 TestPreparation .................................................................................................. 27 2.1.1 Sensorsandexcitation.................................................................................... 27 2.1.2 Impactexcitation............................................................................................. 28 2.1.3 Shakerexcitation ............................................................................................. 29 2.1.4 Sensorlocation ................................................................................................ 29 2.1.5 Wireframemodel ........................................................................................... 32 2.2 Testing................................................................................................................... 37 2.3 Modalextraction ................................................................................................. 37 2.3.1 XFarray ............................................................................................................ 37 2.3.2 iiplot.................................................................................................................. 38 2.3.3 SDOFpeakamplitudemethod..................................................................... 39 2.3.4 Modalmodel ................................................................................................... 41 2.4 Testdatavisualisation ........................................................................................ 41 2.4.1 Deformationarraydef...................................... Erreur!Signetnondfini. 2.4.2 Wireframeplot .................................................... Erreur!Signetnondfini. 2.5 SDTmodalidentificationexample ................................................................... 42
Correlation.........................................................................................................44
3.1 ModalAssuranceCriteria(MAC)..................................................................... 44 3.1.1 AutoMAC........................................................................................................ 45 3.1.2 StandardMAC ................................................................................................ 45 3.1.3 COMAC............................................................................................................ 45 3.1.4 eCOMAC.......................................................................................................... 45 3.2 ii_mac .................................................................................................................... 46
2/47
SDTPrimer
3.2.1 3.2.2 3.2.3
UniversityofBristol
3/47
SDTPrimer
UniversityofBristol
FEModelling
ThemeshgeneratedduringconstructionofanFEmodelisamathematical representationofthestructure.FEpackagesallowdefinitionofageometryin theformofnodes,lines,surfaceswhichareusedasaguideduringmeshing. Thereisadistinctdifferencebetweenmodelgeometryandthemodelmesh. Thegeometryofasquarebeamstructurecanbedescribedasfournodes (vertices)orfourlines.Bothdescriptionsareidenticalthesquareisdefined uniquely.
Figure1.1Squarebeamstructureandgeometrydefinition
Themeshisdefinedtoapproximatethebehaviourofthetruestructure.In thiscasebeamelementswouldbethemostsuitable,andwouldbedefined betweenthenodesoralongthelinesdefinedbythegeometry.Thereis, however,anelementofchoiceandthereforeengineeringjudgementinthis processandthesolutionisnotunique.Themaintradeoffduringmeshingis betweentheaccuracyrequiredandthecomputationalexpense. Oncemeshed,boundaryconditionsaredefinedtosimulatethemodel interactionswithoutsideinfluencessuchasaloadorarigidattachment andarestoredasacase.Thecasealongwiththegeometryandmeshmakeup themodel.Themodelcanthenbesolved,andthedesiredoutputextracted. GEOMETRY SOLUTION MESH OUTPUT MODEL CASE
Figure1.2Modelflowchart
4/47
SDTPrimer
UniversityofBristol
1.1 SDTdatastructure
AlldatainSDTarestoredinMatlabdatastructures.AFEmodelisaMatlab datastructurecontainingfields(seesdtweb('model')) Node matrixdescribingthemodelnodes Elt matrixoftheelementconnectivity pl matrixofmaterial il matrixofelementproperties Stack cellarraycontainingallotherinformation.
1.2 Viewingmodelwithfeplot
Thefeplotfunctiondealswithmodelviewing.cf=feplot(model) initializestheFEmodeldisplayedinthecurrentfigure.Thenthefiguremodel canbeaccessviacf.mdl,thestackviacf.Stackandthecasestackviacf.CStack. Thefecomfunctionthengivescommandstohandlewiththevisualization options.Propertiesofmodelmaybeviewedusingfecom(cf,'pro')which opensthefeplotpropertiesfigure(seesdtweb('dfeplot')).
1.3 Meshingandmodelmanipulations
Therearetwoapproachestothedefinitionofthenodalandelemental descriptionmatrices:explicitdefinitionandfunctionaldefinition.Whenusing theformerallnodesandelementsaredeclaredindividuallyandexplicitly. Functionaldefinitiontakesadvantageofcommandswhichallowthe extrusion,repetition,translationofrelativelysimplemodels(suchasthose definedexplicitly),enablingmorecomplexmodelstobeassembled.
1.3.1 Explicitdefinition
Whendefiningnodalandelementaldescriptionmatricesexplicitly,only nodesandelementmatricesarerequiredasallnodesandelementsare definedcollectively.Eachrowinnodeandelementmatricesrepresentsanew nodeorelement. Themeshingofasimpletrussstructurewillbeusedasanexamplefig1.3. 3 2 4 y Node X(m) y(m) z(m) 1 0 0 0 x 2 0 1 0 3 1 1 0 1 5/47
SDTPrimer
UniversityofBristol 2 1 0
Figure1.3trussstructuremesh
Thedatarequiredpernodeis:
model.Node -
NodeIDnodeidentificationnumber.
PIDcoordinatesystemnumberforposition.
DIDcoordinatesystemnumberfordisplacement.
GIDgroupidentificationnumber.
x,y,zglobalcoordinates.
AtthisstageonlyNodeIDandthecoordinatesarerequired.PID,DIDand GIDareallsettodefaultvaluesofzero.
model.Node=[1 2 3 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2 0 1 1 1 0 0 0 0];
6/47
SDTPrimer
UniversityofBristol
Agraphicalcheckcanbeperformedusingthefeplotcommandfig1.4.The desiredorientationcanbeselectedfromthefiguretoolbar.
cf0=feplot(model);
Figure1.4SDTplotoftrussmesh
Itisimportanttonoteatthisstagethatthetruenumberofvariablesthatcan beenteredforabeamelementfarexceedsthoseshownhere.Whenusingthe explicitdefinitionofelementsitisessentialthatthedimensionsofallrows andallcolumnsisequalthisincludesthefirstline.Inthecaseofabeamthe commandInf abs('beam1')accommodatessixcolumns.All otherrowsmustthereforehavesixcolumns.Thisiswhytheoptionscolumn, althoughnotbeingused,mustbedefined(aszerointhiscase)iffollowing rowsarelarger.Utilisationoftheadditionaloptionsisfareasierwhenusinga functionaldefinitionofthemodel.
1.3.2 FunctionalDefinition
Itisimpracticaltoexplicitlydefineallbutthesimplestofmodels.Anumber offunctionsareavailableintheSDTwhichallowmanipulationofgroupsof elementsintheassemblyofalargermorecomplexmodel.Thiscanbedonein
7/47
SDTPrimer
UniversityofBristol
apiecemealfashion,withsectionsofthecompletemodelbeingadded consecutivelytotheglobalnodeandelementdescriptionmatricesFEnode (sameformasmodel.Node)andFEelt (sameformasmodel.Elt). EachindividualadditiontothesematricesisdefinedinFEel0andtoalesser extentFEn0.FEn0islessimportantbecausebydefininganelementthe positionofthenodesisimplied,andFEnodeisthusupdatedbyFEel0.The femeshcommandisusedformostactions. feutilisanalternativeforfemeshcommands.Whilefemeshperforms operationsontheglobalvariablesFEnode,FEeltetc.feutildealswith modeldatastructures(modelwithfieldNode,Elt). 1.3.2.1 femeshbasiccommands TheSDTusescommandstringswhichdefinethespecificactionthatthe femesh(seesdtweb('femesh'))functionperforms.
femesh('command string'); femesh('command string1','command string2');
8/47
SDTPrimer
UniversityofBristol
Figure1.53Dtrussmesh
Aninitialdefinitionofthesquaregeometryisrequired.Thisisdoneusingan explicitdefinition.
FEnode=[1 2 3 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 1 0 0 0 0 0 ; ; ; ];
FEel0=[]; FEelt=[];
FEel0=[Inf 1 2 3 4
2 3 4 1
abs('beam1') 1 1 7 1 1 7 1 1 7 1 1 7
0 0 0 0
; ; ; ; ];
FEel0isusedasthemodelshallbebuiltupinpieces.Itisimportanttoreset theFEel0arraytozerobeforeanynewoperationstopreventrepetitionof existingelements.FEeltmustalsobeclearedofalldatafromprevious models. ThoseelementsinthearrayFEel0areautomaticallyselectedfor modification.Theywillbeaddedtotheglobalarrayslaterusingthe femesh('addsel')command. Thefirststepistoproduceonesideofthestructure.Twonodesdefinedatthe samecoordinatebydifferentelementscountsasasinglenode,andthose elementsarejoinedtogether.
femesh('repeatsel 10 1 0 0');
9/47
SDTPrimer
femesh('addsel');
UniversityofBristol
ThearraysFEeltandFEel0willnowcontaininformationforalltensquare sections.
femesh('transel 0 0 1'); femesh('addsel');
Figure1.6SDTplotof3Dtrussmesh
10/47
SDTPrimer
UniversityofBristol
Allnodeslyingontheline x=0 andwith y>=1wouldbeselected.These nodes,andanynewelementsdefinedbythesenodesbecometheselected nodesandelementsinFEn0andFEel0. 1.3.2.3.2 extrude Allowsextrusionoftheselectedelements.Requiredinputsarenumberof repetitionsoftheextrusionandthedisplacementvectorinwhichthe extrusionsoccur. Extrusionchangestheelementtype: mass beam beam quad quad hexa
femesh('extrude 5 1 0 0');
11/47
SDTPrimer
FEnode=[1 2 0 0 0 0 0 0 0 0 0 ; 0 1 0 ];
UniversityofBristol
Thissinglebeamelementcannowbeextrudedintoplateelements.
femesh('extrude 6 2 0 0');
Thesecondsurfacemustnowbemeshed.Thiscouldbedoneusingrotation butasecondexampleoftheadvancedcommandswillbeused.
FEel0=[]; femesh('objectbeamline', femesh('findnode x>=0 & y==0')); femesh('extrude 2 0 0 1'); femesh('addsel'); femesh('plotelt');
12/47
SDTPrimer
UniversityofBristol
Figure1.8SDTplotofrightangledstiffenermesh
TheSDTplotofthestiffenercanbeseeninfig1.8.Thedesiredorientationcan beselectedusingthefiguretoolbar.
1.4 Elements
Amodeldescriptionmatrixdescribesthemodelelements.model.Eltand FEeltare,forexample,modeldescriptionmatrices.Thedeclarationofa finiteelementmodelisdonethroughtheuseofelementgroupsstackedas rowsofamodeldescriptionmatrixandseparatedbyheaderrowswhosefirst elementisInfinMatlabandthefollowingaretheasciivaluesforthename oftheelement.Eachdescriptionofelementfollowingillustratesone descriptionmatrixcontainingtheflaglineandatypicallinedefiningone element.EltIdarepositiveintegersuniquelyidentifyingeachelement.The EltIdFix command(femeshorfeutilfunction)returnsamodelthat verifiestheunicityconstraint.
1.4.1 Rigidelements
Arigidelementisdefinedbetweentwonodes,thesecondofwhichis rigidlyconnectedtothefirst.
FEel0=[Inf abs('rigid') ; n1 n2 DofSel MatId ProId EltId];
n1node1ID.
n2node2ID.
DoFselselecteddegreesoffreedomthataretobeconnected.123are
translationaland456arerotationali.e.allDOFsareconnectedif DoFsel=123456.
13/47
SDTPrimer
UniversityofBristol
1.4.2 Masselements
Amasselementisdefinedatanode.
FEel0= [Inf abs('mass1') ; NodeID mxx myy mzz ixx iyy izz EltID]; mxx,myy,mzzmasscomponentsinthedirectionsgiven.
Ixx,Iyy,Izzinertiacomponentsaboutaxisgiven.
Masselementscanalsobedefinedusingthefemeshcommandobjectmass.
femesh('objectmass NodeID);
AmassiscreatedattheNodeIDgiven.Thiscanalsobecombinedwiththe findnodecommandinthesamewayasobjectbeamline.
1.4.3 ElementPropertyManipulation
Whendefiningelementsusingthefemeshcommandanumberoffields requiredmaynothavebeenentered.Theobjectmasscommandisagood example:onlytheNodeIDisrequested.Additionaldatacanbeaddedtothe FEel0matrixbeforeitisaddedtotheglobalmatrix. ThetablebelowholdsthefirstfewlinesofthefinalFEel0fromfemesh example1above. Inf 98 101 97 109 49 1 23 1 1 7 0 23 26 1 1 7 0 26 4 1 1 7 0 4 1 1 1 7 0 4 26 1 1 7 0 26 28 1 1 7 0
Table1.1FEel0example
SDTPrimer
UniversityofBristol
Thecomponentsoftheelementmatrixhavealreadybeendefined:
FEelt [Element type declaration NodeNumbers MatID ProID EltId OtherInfo]
1.4.4 Shelland3Delements
Wejustgivehere2examples forquad4element
FEel0= [Inf abs('quad4') 0; NodeID_1 NodeID_2 NodeID_3 NodeID_4 MatID ProID EltId];
fortetra4element
MatID ProID EltId]; FEel0= [Inf abs('tetra4') ; NodeID_1 NodeID_2 NodeID_3 NodeID_4
1.5 Materialandelementproperties
1.5.1 MaterialProperties
Thematerialmatrix,generallydefinedaspl,isthefieldmodel.plinthe modeldatastructure. Highlevelcallsusethem_elasticfunctionwhichcontainsadatabase. Typicalhighlevelcallsare
model.pl=m_elastic('dbval 1 Steel')
model.pl=m_elastic(model.pl,'dbval 2 Aluminum')
typetypeofmaterialbeingused,typicallyfe_mat('m_elastic',1,1).
15/47
SDTPrimer
EYoungsmodulus.
UniversityofBristol
nuPoissonsratio.
rhodensity.
1.5.2 SectionProperties
Alloftheelementsusedinafiniteelementmodelareapproximationstoareal structure.Theirgeometryhasbeensimplified(toalineinthecaseofabeam) andtheirbehaviourisapproximatedbydefinitionofsectionproperties.The propertiesrequiredtodefineanelementwillvarywithelementtype:abeam requiressecondmomentsofareawhereasaplaterequiresthicknessetc. Thepropertiesmatrix,generallydefinedasil,isthefieldmodel.ilinthe modeldatastructure.Eachrowoftheilmatrixrepresentsadifferentsection property.AllunitsareS.I. 1.5.2.1 Beam Highlevelcallsusethep_beamfunctionwhichcontainsadatabase.Typical highlevelcallsare
model.il=p_beam('dbval 1 rectangle .05 .01')
Thesecallsbuildmodel.ilwhereeachrowofthematrixrepresentsa differentbeamproperty.
16/47
SDTPrimer
UniversityofBristol
il=[ProID
type
I1 I2 A k1 k2]
ProIDelementpropertyIDasdefinedbyuser.
typetypeofproperty,typically,fe_mat('p_beam,1,1)
Jtorsionalstiffness.
I1secondmomentofareaaboutbendingplane1(forbeamonxaxisbp1in
planexy). I2secondmomentofareaaboutbendingplane2(forbeamonxaxisbp2in planexz). Aarea. Thesesixaretheonlyrequiredinputs.Theshearcorrectionfactorscanbe omittedorsettozeroifnotdesired. k1shearcorrectionfactorforbendingplane1. k2shearcorrectionfactorfortbendingplane2. 1.5.2.2 Plate
Highlevelcallsusethep_shellfunctionwhichcontainsadatabase.Typical highlevelcallsare
model.il=p_shell('dbval 1 mindlin .01')
17/47
SDTPrimer
UniversityofBristol
Thesethreearetheonlyrequiredinputsatthisstage.F,dand0cannotbe omittedasthethicknessisneeded.Defaultvaluesmustbegiven:[1 1 0].If kissettozerooromittedthedefaultwillbeused. ItisworthnotingthatastandardMatlabmatrixmusthavethesamenumber ofcolumnsineachrowandviceversa.Caremustbetakentoensurethatthis conditionisnotviolatedwhenstoringpropertiesofelementsofdifferent typesinthesamematrix.Additionalrowsorcolumnscanbefilledby enteringazero.Otherelementswillbedescribedfurtherinlatersections. Forexample:
il=[1 fe_mat('p_beam',1,1) 1e-6 2 fe_mat('p_shell',1,1) 1 1e-7 1 1e-7 0 1e-4 5e-3 0.8 ; 0 ];
1.6 Loadsandboundaryconditions
Alltheloadsandboundaryconditionsarehandledwiththefe_case functionandonlyoneglobalcaseispossibleatonetime.
1.6.1 BoundaryConditions
Beforeamodelcanbesolvedtheboundaryconditionsmustbeconsidered. 1.6.1.1 Degreesoffreedom Eachnodehassixdegreesoffreedom:threetranslationalandthreerotational. xtrans ytrans ztrans xrot yrot zrot 01 02 03 04 05 06
Table1.2Nodaldegreesoffreedom
18/47
UniversityofBristol
Bothofthesecommandsmustfollowtheformat:
Thefollowingsectionsdescribethemainentrytypes. 1.6.1.2.1 Fixdof Theentrytypeisfixdof(declaredDOFsfixed).Thenameisarbitrarily definedbytheuserandisusedasareferencetotheboundaryconditionbeing applied.ThedataholdsthelistofDOFsthatthecommandwillapplyto. Thereareanumberofoptionsonthetypeofdatathatcanbeentered. AglobaldeclarationofDOFscanbegiveni.e.[.01 .02 .06]' (columnvector)wouldactivatea2DsimulationwithonlyDOFsinx trans,ytransandzrotbeingactive. AspecificdeclarationofDOFscanbegivenasacolumnvector. Analgebraicexpressioncanbegivenbutitmustbeenclosedby invertedcommas.
model=fe_case(model, 'keepdof', '3-D', [.01 .02 .03 .04 .05 .06]' 'fixdof', 'leftedge', 'x==0');
DOFs.01to.06havebeendeclaredasactiveforallnodesthename3Dhas beengivenasthisrepresentsa3Danalysis.FinallyallDOFshavebeenfixed alongthelinex=0.Inthestructuresdefinedintheexamplesabovethis representstheleftedge,hencethenamegiven. Notethattheuseofakeepdofentry(declaredDOFsareactive)ispossible butnotencouraged. 1.6.1.2.2 StructuralLoading TheapplicationofaloadtotheFEmodelfollowsthesameformatasthe fixdofcommand: 19/47
SDTPrimer
UniversityofBristol
TheDOFdeclarationisacolumnvectoroftheDOFstowhichtheloadsapply. Theloadarraycontainsanumberofcolumnvectors,eachofwhichrepresents adifferentloadingcondition: 1.03 1 0 load = 0 1 DoF = 2.01 3.01 0 1 InthiscasethethreeDOFsbeingloadedare1z,2xand3x.Thefirstcolumnof thedefarrayreturnsaunitloadatnode1inthezdirection.Thesecond columnreturnsapositiveunitloadatnode2inthexdirectionandanegative atnode3inthexdirection.Thisisusedtodescribedistributedloads.
1.7 Solving
Thereareanumberofsolverfunctions,eachgivingadifferenttypeof solution.Thefe_simulfunctionallowstotreatstaticlinear,modalandtime solutions.Itisthehighlevelimplementationofthestandardsolversas fe_eigorfe_time. 1.7.1.1 fe_simul Thefe_simulfunctionisthegenericfunctiontocomputevarioustypesof response,aneasyaccesstospecializedfunctionstocomputestatic,modal(see fe_eig)andtransient(seefe_time)response.Onceyouhavedefineda FEMmodel,materialandelementsproperties,loadsandboundaryconditions callingfe_simul(seefe_mk)assemblesthemodel(ifnecessary)and computestheresponseusingthededicatedalgorithm. Thegenericcallis
[def,mo1] = fe_simul('command',model)
20/47
SDTPrimer
UniversityofBristol
Ofthefivepossibleoptionsonlythefirsttwoarerequired.Thefinalthree shouldbesettozero(thedefaultvalueswillbeassigned). solutionmethodintegerdefinessolutionmethod.Defaultmethodisafull solutionwhichcanonlybeusedforsimplemodels.Forlargermodelsuse Lanczossolver:solutionmethod=5. nmnumberofmodesrequired.Thedefaultvalueisafullanalysisi.e.one modeforeverydegreeoffreedom. 1.7.1.3 fe_mknl(andobsoletefe_mk) Thefe_mkfunctionreturnsthemassandstiffnessmatricesalongwiththe DOFdefinitionvector.
mo1 = fe_mknl(model);
allowsuseofdamping.
21/47
SDTPrimer Amoregeneralcallis
UniversityofBristol
[def,model] = fe_time(model)
Example:
22/47
SDTPrimer
UniversityofBristol
Example:
1.8 Postprocessing
1.8.1 feplot
ThefeplotfunctionallowsFEmodeldatabedisplayedgraphicallyand withanimation.Inordertomanipulatethefeplotatypicalcallis
cf=feplot(model,def);
23/47
SDTPrimer
UniversityofBristol
Thecurrentfigureisassignedamodelandasetofdeformationswhichapply tothatmodel.AnfeplotwindowwithaGUIisgenerated. ThedefstructurecontainsaDOFfield(columnvector),adeffield(matrix) whosenumberoflinesisthesameasDOFvectorandeachcolumnisasetof deformation,onemodaldeformationoronedeformationatonetimestep, anddataandlabfieldswhicharelabelsdefinition. Thetypeofplotwillvarydependingonthetypeofdataheldinthemodel anddeformationmatrices,andthereforethetypeofsolutionthatwasused. 1.8.1.1 NaturalFrequenciesandModeShapes Ananimatedplotofnaturalfrequenciesandmodeshapesfollowsfrom fe_eig:
def=fe_eig(model); cf=feplot(model,def)
24/47
SDTPrimer
UniversityofBristol
1.8.2 iiplot
iiplotallowsthefrequencyresponsetobeplotted.Thedataisstoredinto thefigureandcanbeaccessedthroughapointerci=iiplot;ci.Stack,see moredetailswithsdtweb diiplot#CurveStack.Therewillanumberof responses:oneateachlocationforeachloadpoint.Anumberofresponse propertiesarecalculatedincludedmagnitudeandphase.Theresponsetype canbetoggledusingtheiiplotGUI. Itcanbeusedinasimilarwayasfeplotwithregardstoassigningthecurrent figure.
1.9 SDTmeshingexample
Thefollowingexampleillustratesallfunctionsandcommandsthathavebeen utilisedtothispoint.Thestructurebeingmodelledistherightangled stiffeneralreadydescribed. Thecodeforthisexampleisalsoavailableelectronically.
% SDTtestex % INITIAL DECLARATION %--------------------FEnode=[]; % node and element matrices set to zero FEelt=[]; % prevents old data being added to the model FEel0=[]; FEn0=[]; FEnode=[1 0 0 0 0 0 0 2 0 0 0 0 .1 0]; % initial node declaration
% MESHING %--------FEel0=[]; femesh('objectbeamline',femesh('findnode x==0 & y<=1')); femesh('extrude 6 .2 0 0'); % beam element created and extruded to make first row % of plate elements femesh('repeatsel 2 0 .1 0'); % plates repeated to make one side of stiffener femesh('addsel'); % elements in FEel0 added to FEelt as group1 FEel0=[]; % FEel0 reset to zero femesh('objectbeamline', femesh('findnode x>=0 & y==0')); % beam elements created between nodes along line y=0 femesh('extrude 2 0 0 .1'); % beam elements extruded to plate elements femesh('addsel'); % elements in FEel0 added to FEelt as group2 % PROPERTIES %-----------pl=[1 1 7.2e9 il=[1 2 1
.3 1
2700 0
2.68e10]; 5e-3];
25/47
SDTPrimer
UniversityofBristol
femesh('set group 1 mat1 pro1'); femesh('set group 2 mat1 pro1'); % both groups defined as material 1 and section 1 % ASSEMBLE MODEL %---------------model=femesh('model'); model.il=il; model.pl=pl; % modelled assembled as structure matrix with required fields load=struct('DOF',[14.03],'def',[1]); % load condition defined at node 14 orientated along z axis model=fe_case(model, ... 'keepdof','3D',[.01 .02 .03 .04 .05 .06]',... % 1 'fixdof','left edge','x==0',... % 2 'dofload','endload',load,... % 3 'sensdof','tipsensor',[14.03]); % 4 % case 1 defined with: % 1 - all trans and rot DOFs active % 2 - left edge of the stiffener is fixed (x=0) % 3 - load applied to case % 4 - sensor defined at node 14 orientated in z axis % NATURAL FREQUENCIES AND MODE SHAPES %------------------------------------cf=feplot(model); model=cf.mdl; cf.Stack{'info','EigOpt'}=[5 10]; % deformation calculated for model % options: 5 - Lanczos solver % 10 - number of modes cf.def=fe_simul('mode',model); % TIME RESPONSE %--------------cf.Stack{'info','q0'}=[]; % initial conditions for time response set to zero % define options cf.Stack{'info','TimeOpt'}=... struct('Method','newmark','Opt',[.25 .5 0 .000002 10000 10],... ... 1 2 3 4 5 6 7 'NeedUVA',[1 1 0]); % 8 % options: % 1 - newmark solution selected % 2,3 - beta and alpha default values set % 4 - beginning of time simulation % 5 - time step % 6 - number of cycles % 7 - number of time steps for input force % 8 - what is computed (disp, vel, acc ...) def=fe_simul('time',model); % displacement and velocity fields calculated % model updated using case 1 boundary conditions cf.def=def; fecom('animtime'); % animate displacement.
26/47
SDTPrimer
UniversityofBristol
2 Testing
Thedynamicpropertiesofastructurearedeterminedduringtesting.FRFsare analysedtoproduceamodalmodelofthestructuretheeigenvalueand eigenvectormatrices. InordertoassessthemodeshapesbothpointandtransferFRFsarerequired. ApointFRFisaresponsetoexcitationatthesamelocation,atransferFRFisa responsetoexcitationatanalternativelocation.Therearetwoapproachesto obtainingtheFRFs: afixedexcitationpointandmeasuringresponseatanumberof locationse.g.ashakertest. afixed(orlownumberof)measurementpointandbyvaryingthe excitationlocatione.g.animpacttest. Thereisaminimumpermissibledatasetwhichwillallowdeterminationof themodalmodel.Alargerdatasetwillgiveahigheraccuracyanddefinition anFEmodel,forexample,calculatesresponsesateveryactivedegreeof freedomandsimulatesmodeshapesindetail.Howeverthisisclearly impracticalwiththesmallestofteststructuresandimpossibleindustrially whereDOFsnumberinthemillions.Thedeterminationofsensor/excitation placementisthereforeacrucialaspectoftesting. Therearealargenumberofpracticalissuesassociatedwithtesting,including thesimulationofboundaryconditionsandtherepeatabilityofexperiment. Theseissueswillnotbediscussedhere.TheuserisreferredtoModalTesting, Theory,PracticeandApplicationbyD.J.EwinsfromtheResearchStudies PressLtd.
2.1 TestPreparation
2.1.1 Sensorsandexcitation
Responseismeasuredatafixed(orlownumberof)measurementpoint(s) andbyvaryingtheimpactlocation. Thenumberofimpactlocationsrequiredwilldependonthecomplexityof thestructuralmodesbeinganalysed,andthedefinitionrequiredinthe measuredmodeshapes.Orientationofsensorsisalsocrucialasmostonly workinonedirectionanymotionorthogonaltothiswillnotbedetected.It isalsoimportanttoensurethattheaccelerometerisinsuchapositionastobe excitedbyallmodesofinterest.
27/47
SDTPrimer
UniversityofBristol
sensors
charge amp
CHANNEL 1 filters
analyser / PC
CHANNEL 2
Figure2.1Testsetupillustration
2.1.2 Impactexcitation
Excitationcanbeachievedbyanumberofmeans.Forsmallscalestructures impacttestingisrecommended.Ahammerwithforcetransducerisusedto excite,andanaccelerometerisusedtomeasuretheresponse. Singleormultiplesensorlocationsareuseddependingonthecomplexityof themodeshapesbeingmodelled.Formotioninonlyoneplaneitisoften possibletouseasingleaccelerometer.Theimpactlocationisvaried. Averagingisstronglyrecommendedwhencarryingoutanimpacttest. Theenergydeliveredbyahammerduringandimpactexcitationhasa frequencyspectrumasshowninfig2.2. 28/47
SDTPrimer
UniversityofBristol
amplitude
f c
frequency
Figure2.2Frequencyspectrumofhammerexcitation
2.1.3 Shakerexcitation
Excitationisachievedthroughuseofashaker,whichisdrivenbyasignal generator.Thereareanumberofsignaltypesthatareusedasdetailedbelow. Itisimportanttonotethatthesignalgeneratorandshakerisadynamic systeminitselfandsotheoutputoftheshakerwillnotmatchthesignal exactly.Theoutputoftheshakermustthereforebemonitoredusingaforce transducer. 2.1.3.1 Excitationsignals 2.1.3.1.1 Sine Excitationfollowstheprofileofasinewave.Itcanbeatafixedfrequency (sinewave)oroverafrequencyrange(sinesweep).Afteragivendurationthe signaloutputiszeroed,allowinganaturaldecayoftheshaker. 2.1.3.1.2 Chirp 2.1.3.1.3 Pseudorandomnoise
2.1.4 Sensorlocation
Thefollowingsectionreferstosensorlocation,butthedirectionsgivenapply equallytoimpactlocation. Agoodtestsetupwillenableallmodeshapesinthedesiredfrequencyrange tobedefined,withtheminimumofresponsemeasurements.AFEmodelcan
29/47
SDTPrimer
UniversityofBristol
Themseqcommandreturnsacolumnvector(sdof)containingtheoptimal sensorDOFs.Theninthecommandisauserinputandspecifiesthenumber ofsensorlocationsdesired.Thedeformationmatrixdefforthemodelis required.Theoptimisationofthesensorplacementwilltakeintoaccountall modesinthedeformationmatrix.AnoptionalinitialsetofDOFs(sdof0)can alsobedeclaredfromwhichtheoptimalwillbeselected. 2.1.4.2 fe_sensexample Acantileveredbeamwillbeusedasanexample.Themodeldimensionsare showninfig2.4. y 0.1m x 1m
Figure2.4Cantileveredbeammesh
30/47
UniversityofBristol
model=feutil('ObjectBeam 1 1',Node,linspace(0,1,10)); model.pl=[1 fe_mat('m_elastic','SI',1) 7.2e9 .3 2700 model.il=[1 fe_mat('p_beam','SI',1) 1e-9 1e-9 3e-9 cf0=feplot(model); model=cf0.mdl; model=fe_case(model,'keepdof','2D',[.01 .02 .06]',... 'fixdof','leftedge','x==0'); fecom('promodelviewon'); fecom('curtab Cases','outputs'); 0]; 1e-4];
Thefe_sensfunctionisusedtoplacetwosensors.
sdof=fe_sens('mseq 2 type',def);
31/47
SDTPrimer
UniversityofBristol
Inordertoplotthesensorlocationsthefeplotismanipulatedusingthefecom function.
fecom(cf0,';sub 2 1'); % feplot view subdivided - 2 rows, 1 column fecom(cf0,';sub 2 1 1;view3'); % for row 1: view 3 selected (isometric with z vertical) fecom(cf0,';sub 2 1 2;showsens;showarrow;view3;textDOF'); % for row 2: sensor location shown % arrows shown for sensor deformation % view 3 selected % DOF labels shown
Figure2.5SDTplotofcantileverbeamsensorplacement
2.1.5 Wireframemodel
Awireframemodeloftheteststructureisgeneratedtoenablecomparisonof thetestdatawiththeFEmodel.Thewireframemodelmapsoutthegeometry oftheteststructureintheformoflinesandnodeswhichencompassalltest
32/47
SDTPrimer
UniversityofBristol
Itisimportanttorememberthatthewireframeisnotameshlikethat generatedduringthemodellingstage.Therearenoelementsthenodesare connectedwithlineswhichhavenostructuralpropertiesandaretheretoaid visualisationofthedeformation. 2.1.5.1 Generatingthewireframe Itwillbeassumedatthispointthatsensorlocationscoincidewithnodesin theFEmodel. Thewireframeisdefinedbynodes,andtheconnectinglinesbetweenthem. Allsensornodesshouldbeincluded,alongwithanumberofadditional nodeswhichaidvisualization.Newnodesaredefinedateachlocation.A numberingconventionisrecommendedtoeasesegregationofthenewnodes fromtheold:ifthenumberoftheoriginalnodeis20,thenewwireframe nodebecomes1020.Obviouslythiscanonlybeusediftherearelessthan1000 nodesintheoriginalmodel. TheconnectionsbetweennodesaredefinedusingtheLandldrawarrays.The LarrayholdstheNodeIdnumbersforthosenodesbeingconnected,theline willbecontinuousbetweenallnodesgivenunlessseparatedbya0.Thedata inLisstoredintheldrawarraymultipleLscanbestoredwithoneineach rowofldraw.Thisenablessectionsofthewireframetobecreatedina piecemealfashion.
L=[1020 1023 1034 0 1012 1034 1039];
Representstwolinesbetweennodes:102010231034and101210341039.
ldraw(1,[1 82+[1:length(L)]])=[length(L) L];
33/47
SDTPrimer
UniversityofBristol
OncetheldrawarrayiscompleteitiscombinedwiththeFEmodelmesh usingthefemeshcommandaddtest.
femesh('addtest',node,ldraw);
ThisyieldsthefollowingsensorDOFs:
sdof=[35.02 21.03 18.03 32.02 19.03 33.02 16.03 30.02 35.03 21.02];
Itisassumedthatallofthesenodeslieinlocationsontheteststructurethat canbetested;nodes16,18,19,21,30,32,33,35mustbeincludedinthewire framemodel.ThenumberingsystemwillbeemployedwhichgivesNodeId 1016,1018etc. Ifonlythese8nodesareuseditisnotpossibletovisualizethemodeshape. Additionalnodesmustbeincludedinthewireframe. 1021 1019 1018 1016 1035
1030
1032 34/47
1033
SDTPrimer
UniversityofBristol
Figure2.7Nodetestlocations
Thefollowingwireframenodelocationsaresuitabletoproperlycapturethe modeshape. 1021 1019 1018 1016 1015 1013 1009 1007 1003 1035 1001 1033 1032 1030 1029
Figure2.8Wireframenodelocations
Thewireframenodearraymustfirstbedefined.Thisisdoneinexactlythe samewayastheFEnodedeclaration.
node=[1001 1003 1007 1009 1013 1015 1016 1018 1019 1021 1029 1030 1032 1033 1035 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.2 0.6 0.8 1.2 0 0.2 0.6 0.8 1.2 0 0.2 0.6 0.8 1.2 0 0 0 0 0 0.2 0.2 0.2 0.2 0.2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.2 0.2 0.2 0.2 0.2];
ldrawisdefinedusingL:
L=[101 103 107 109 1013]; ldraw(1,[1 82+[1:length(L)]])=[length(L) L]; L=[1015 1016 1018 1019 1021]; ldraw(2,[1 82+[1:length(L)]])=[length(L) L]; L=[1029 1030 1032 1033 1035]; ldraw(3,[1 82+[1:length(L)]])=[length(L) L];
Thethreestraightedgeshavebeendefinedfirst,eachinadifferentrowof ldraw.
L=[1015 101 1029 0 1016 103 1030 0 1018 107 1032 0 1019 109 1033 0 1021 1013 1035 0]; ldraw(4,[1 82+[1:length(L)]])=[length(L) L];
35/47
SDTPrimer
UniversityofBristol
Finallythe5Lshapedlinesareaddedtoldraw.Thewireframecannowbe addedtothemodelusingthefollowingcommand:
femesh('addtest',node,ldraw);
Figure2.9SDTwireframe
SDTPrimer
UniversityofBristol
nodeswouldcoincidewiththeFEnodes.Inarealtestthisisnotalways possible,sotheoffsetsbetweentestandFEnodesmustbeaccountedfor.
2.2 Testing
2.3 Modalextraction
ThemodalmodelisconstructedusingthepointandtransferFRFs.Thereare anumberofmethodsusedtoextractthisinformation,thesingledegreeof freedom(SDOF)peakamplitudemethodissupportedbySDT.
2.3.1 XFarray
FrequencyresponsedataisstoredintheXFarrayXF=iicom('curvexf').It hasastructurewhichispredefinedintheSDTandisreservedforfrequency analysis. XFissplitintosubarrays:XF(Test)andXF(IdFrf)containresponsedata. XF(IdMain)andXF(IdAlt)holdcurveshapedata.Thereareanumberof componentsineachwhichenablecontrolofmanyfrequencyresponse parameters.TheXFarraymustalsoidentifythepositionandorientationof bothsensorandexcitationforeachFRFgiven.Thisdataisstoredinthe degreeoffreedomsubarrayXF.dof. 2.3.1.1 Responsedata Thefrequencyresponseismeasuredatafinitenumberofpointswhichare commontoXF(Test)andXF(IdFrf).Thefrequencypointsarestoredin XF(n).w. XF(Test).xfistheprimaryresponseandwillholdtheactualmeasured response.Thelaterresponseswilloftenbetakenfromreducedmodels.These canbeplottedandcomparedwiththetestdataandassessedforaccuracy. Thiswillbediscussedlater. 2.3.1.2 Shapedata Thefrequencyresponseofarealstructuremustbebrokendownintoa mathematicalmodeltoenablecomparisonwiththeFEmodel.Theshapeof thefrequencyresponseplotisrepresentedbythepoles,whicharestoredin XF(IdMain).po.XF(IdAlt).poisanadditionalspaceallocatedforasecondset ofshapedata. 37/47
SDTPrimer
UniversityofBristol
EachrowofXF(IdMain).porepresentsapole.Column1isthefrequencyand column2isthemagnitudeatthepole. TheshapedatasectionoftheXFarrayalsoholdsthemodalmodel.Itisstored inXF(IdMain).res. XF('IDMAIN').res=IIres EachrowofIIresrepresentsamode(orpole).Eachcolumnrepresentsa differentresponsechanneli.e.theresponseatadifferentsensorlocation. 2.3.1.3 DOFdata ThesensorandexcitationDOFforeachFRFarerequiredincalculationofthe modeshapes.ThisdataismanuallydefinedbytheuserinXF(Test).dof.
XF('Test').dof=[response DOF, excite DOF, channel options]
TheformatoftheresponseandexcitationDOFfollowsthatinsection1.5.2.1. EachrowofXF(Test).dofholdsdataforadifferentFRF.Thechannelisan integerlabelwhichlinkstheDOFdatatothecorrespondingfrequency response(heldinXF(Test).xf).Thereareadditionaloptionsthatarenot requiredatthisstage. 2.3.1.4 XFoptions TheXFarrayholdsalldatarelatedtothefrequencyresponse.Thisincludes manyoptionseffectingplotting,datatypeetc.Theuserisreferredtothe MatlabSDThelpfilesseefunctionxfopt.
2.3.2 iiplot
Frequencyresponseplotsutilisetheiiplotfunctioninmuchthesamewayas FEmodelplotsusethefeplotfunction. TheXFarrayisintrinsicallylinkedtotheiiplotfunction.Callingtheiiplot functionwillautomaticallyplottheresponsedataheldintheXFarray.The datainthearraycanbemanipulatedusingtheiicomfunctionthisisthe equivalentofthefecomfunctionwhichisutilisedonfeplots.Theuseris advisedtousetheiicomcommand:
iicom('submagpha');
38/47
SDTPrimer
UniversityofBristol
2.3.3 SDOFpeakamplitudemethod
2.3.3.1 SDOFpeakamplitudetheory AnFRFforamultidegreeoffreedom(MDOF)systemwillcontainanumber ofpeaks.Eachpeakrepresentsaresonantfrequencyofaparticularmode.It canbeassumedthattheinfluenceofothermodesinthisregionisminimal. TheSDOFmethodutilisesthisassumptionbytreatingeachmode independently.ConsidertheFRFofmoder. |H()| || |H| 2 b r a
Figure2.6SDOFpeakamplitudemethod
kr
39/47
SDTPrimer
UniversityofBristol
EachmodeontheFRFisconsideredinturn. TheminimumdatarequirementisasinglepointFRF,andmultipletransfer FRFsforalllocationsrequiredinthetestmodeshape.Itisadvisabletouse moreFRFsthantheminimum(bothpointandtransfer)toimprovethemode shapeaccuracy. TheSDThasbuiltinfunctionstoaididentificationofthemodalmodel. 2.3.3.2 idcom Theidcomfunctionisusedinidentificationofthemodalmodel.Itutilises commandstrings:
idcom('command');
banduseagivenbandforfrequencyestimation.Bandcanbe:
lessthan1halfbandwidthasfractionoffrequencyestimate. greaterthat5numberoffrequencypointsinbandwidth.
frequencyinitialestimateofpolefrequency.
Example:
idcom('e
10
200');
40/47
SDTPrimer
idcom('ea 10 200');
UniversityofBristol
2.3.4 Modalmodel
Therearemanymethodswhichcanbeemployedtoextractthemodalmodel fromasetoftestdata.TheSDOFpeakamplitudemethodisthemostsimplest andcanattimesbeinaccurate.Therearealternativeapproachesand optimisationschemesthataresupportedinSDTwiththeaimofimproving themodalmodel.
2.4 Testdatavisualisation
Havingextractedthemodalmodelameanstovisualisethemodeshapesis required.Thedeformationsarelinkedinwiththewireframemodelto produceananimatedplot.ThemethodusedisastandardSDTfeplot. Itisassumedherethatawireframehasalreadybeendefinedandthatits componentsareselectedinFEel0.
femesh('plotel0'); fecom('showline'); cf=feplot;
41/47
SDTPrimer
UniversityofBristol
2.5 SDTmodalidentificationexample
Theexamplestructurebeingusedhereisatubularstructureshownbelow.
ThetestdatausedwasgeneratedusingSDTafulltestingexamplewillbe givenseparately.
% The data can be downloaded at % www.sdtools.com/contrib/kart_example.mat %-----------
% WIRE FRAME DEFINITION (present in WIREFRAME) built with %----------------------WIREFRAME.Node=[ ... 1 0 0 0 0.613 0 0 2 0 0 0 1.133 -0.198 0 3 0 0 0 1.548 -0.208 0 4 0 0 0 1.538 -0.487 0 5 0 0 0 1.133 -0.51 0 6 0 0 0 0.613 -0.742 0 7 0 0 0 0.32 0 0 8 0 0 0 0.772 -0.287 0 9 0 0 0 0.772 -0.521 0 10 0 0 0 0.33 -0.748333333333333 11 0 0 0 0.41 -0.218 0 12 0 0 0 0.38 -0.545 0 13 0 0 0 0.1 -0.220666666666667 14 0 0 0 0.1 -0.551666666666667 15 0 0 0 0 -0.217714285714286 16 0 0 0 0 -0.544285714285714 ]; % nodes for wire frame defined WIREFRAME.Elt=[];feplot(WIREFRAME) fecom('textnode');
0 0 0 0 0
% You can use the contextmenu (right click) cursor ... -> 3dline pick feplot('addline',[4 3 2 1 7 15 16 10 6 5 4]) % add a first line feplot('addline',[8 11 13 0 9 12 14 13 0 9 8 0 12 11]) % other line % POLE ESTIMATION %----------------[ci,XF]=idcom; %open interface and get pointers ci and XF load kart_example TESTDATA XF('Test').w=TESTDATA.w; % Frequencies XF('Test').xf=TESTDATA.xf; % Responses XF('Test').dof=TESTDATA.dof; % Dofs
42/47
SDTPrimer
iicom('submagpha'); % frequency response data plotted
UniversityofBristol
% poles must now be identified one by one idcom('e .01 44.6'); iicom('CurTabIdent') % pole estimated at freq of 44.6Hz with possible range of .01 percent % around that frequency. % estimate checked on freq plot, if correct it is added idcom('ea'); idcom('e .01 48.8');idcom('ea'); idcom('e .01 95.3');idcom('ea'); idcom('e .01 125.3');idcom('ea'); idcom('e .01 141');idcom('ea'); % five poles estimated in total and are stored in XF('IdMain').po. % modeshapes are based on narrowband estimate. idcom('est'); % build a broad-band identification % VISUALISATION %--------------cf=feplot; load kart_example WIREFRAME; cf.model=WIREFRAME; cf.def=XF('IdMain'); fecom('view3') % display identified shapes
43/47
SDTPrimer
UniversityofBristol
3 Correlation
Aquantifiablecorrelationbetweenexperimentalandanalyticalmodeshapes canbedeterminedbasedontheModalAssuranceCriteria(MAC).
3.1 ModalAssuranceCriteria(MAC)
Eachmodeshapeisavectorofnodaldisplacements.Thecorrelationbetween twovectorsaandbisgivenby:
{ b }T { b } MAC ( a , b ) = ({ b }T { b }).({ a }T { a })
2
Theproductofthisisascalarquantity.Ifaandbareidenticalthe numeratoranddenominatorareequal,givingaMACvalueof1.Ifthetwo vectorsareorthogonaltooneanotherthenumeratoris0andhencetheMAC valueiszero. EachvectorfromdatasetAiscomparedwitheachvectorfromdatasetB.The dataisdisplayedonaMACplot,seefig3.1. DatasetB Mode: 1 6 Mode: 1 MAC=1 DatasetA
6
Figure3.1MACplotexample
MAC=0
44/47
SDTPrimer
UniversityofBristol
3.1.1 AutoMAC
TheMACcanbeunreliablewhenavectorisunderpopulated.Thisisoften thecasewithexperimentaldatawherethenumberofDOFsincludedisfar lowerthanthetotal.Itisgoodpracticetocomparesuchdatawithitselfto gainanindicationofhowwellthevectorsaredefined.Thevaluesonthe leadingdiagonalwillequalunity.Highoffdiagonalterms(>0.2)indicate poordefinitionbetweenmodes.Ahighcorrelationbetweenmodeswhichare closeinfrequencyindicatesthepossibilityofcouplingbetweenmodes.
3.1.2 StandardMAC
Thisisthemostcommoncorrelationtestappliedinmodalanalysis.MAC valuesarecalculatedforallgridvaluesasindicatedinfig3.1.Automatic pairingofthemodesisoptional;basedonthehighestMACvalue.Theinitial pairingofmodesisbasedontheirfrequency,butdiscrepanciesbetween modelandtestarecommon.Itisalsopossiblethatatestmodewillhavea highcorrelationwithmorethanoneanalyticalmode,andasaresultwill appearmorethanonceontheMACplot.
3.1.3 COMAC
Thecoordinatemodalassurancecriteria(COMAC)isanextensionofthe MACandisusedtoidentifythoseDOFsresponsibleforloweringtheMAC. ForeachDOFthecorrelationisassessedbetweenanalyticalandexperimental modeshapesusingeverymode.Itrequiresthemodestobepairedusingthe standardMAC.
COMACf =
{
M m =1
amf
bmf
M
{
M m =1
T amf
amf } { bmf }
m =1 T bmf
WhereamfisthenodaldisplacementatDOFf,modemfromdataseta.As withthestandardMACavalueapproachingunityimpliesagoodcorrelation.
3.1.4 eCOMAC
TheEnhancedCOMAC(eCOMAC)isanadaptationoftheCOMAC.All modesarenormalisedtounity,thusmakingtheassessederrorinsensitiveto scaling.
45/47
SDTPrimer
UniversityofBristol
eCOMAC =
amf
bmf
amf =
bmf
amf am bmf = bm
2M
3.2 ii_mac
CorrelationisachievedintheSDTusingtheii_macfunction.
ii_mac(data set/s, 'command string');
3.2.1 AutoMAC
TheautoMACisaselfcorrelationtestandsoonlyonedatasetisrequired.
ii_mac(data A, 'MAC auto plot');
3.2.2 StandardMAC
ThestandardMACpairstheuncorrelatedvectors.
Ifthepaircommandinthestringisomittedadirectcorrelationwithout pairingisperformed.
3.2.3 COMAC
TheCOMACcommandautomaticallypairsthemodesusingtheMACpair command.TheequivalenteCOMACcommandisalsogiven.
ii_mac(data A, data B, 'comac plot'); ii_mac(data A, data B, 'comac e plot');
SDTPrimer
UniversityofBristol
47/47