• No results found

Design framework for multifunctional additive manufacturing: coupled optimization strategy for structures with embedded functional systems

N/A
N/A
Protected

Academic year: 2020

Share "Design framework for multifunctional additive manufacturing: coupled optimization strategy for structures with embedded functional systems"

Copied!
9
0
0

Loading.... (view fulltext now)

Full text

(1)

Design

framework

for

multifunctional

additive

manufacturing:

Coupled

optimization

strategy

for

structures

with

embedded

functional

systems

Ajit

Panesar

,

Ian

Ashcroft,

David

Brackett,

Ricky

Wildman,

Richard

Hague

FacultyofEngineering,UniversityPark,UniversityofNottingham,NG72RD,UK

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received4November2016

Receivedinrevisedform24February2017 Accepted21May2017

Availableonline23May2017

Keywords:

Optimization Additivemanufacturing Multi-functional Computationalmethods

a

b

s

t

r

a

c

t

Thedriverforthisresearchisthedevelopmentofmulti-materialadditivemanufacturingprocessesthat providethepotentialformulti-functionalpartstobemanufacturedinasingleoperation.Inorderto exploitthepotentialbenefitsofthisemergenttechnology,newdesign,analysisandoptimization meth-odsareneeded.Thispaperpresentsamethodthatenablesintheoptimizationofamultifunctionalpartby couplingboththesystemandstructuraldesignaspects.Thisisachievedbyincorporatingtheeffectsofa system,comprisedofanumberofconnectedfunctionalcomponents,onthestructuralresponseofapart withinastructuraltopologyoptimizationprocedure.Thepotentialoftheproposedmethodis demon-stratedbyperformingacoupledoptimizationonacantileverplatewithintegratedcomponentsand circuitry.Theresultsdemonstratethatthemethodiscapableofdesigninganoptimizedmultifunctional partinwhichboththestructuralandsystemrequirementsareconsidered.

©2017TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).

1. Introduction

Single-materialadditivemanufacturing(AM)processes,suchas

selectivelaser melting,enablethedesign ofgeometrically

com-plexparts.Multimaterialadditivemanufacturing(MMAM)further

expandsthis design freedomto includethe spatialvariation of

materialcompositionandenablemultifunctionalitythroughthe

volumeofapart.Multifunctionality,bydefinition,necessitatesthe

embeddingofactivesub-componentsinordertodeliveradditional

functionalcapability,suchaselectronic,electro-mechanical,

opti-cal,electromagnetic,chemical and thermal [1]. MacDonald and

Wicker[1],ina recentreviewarticle,identified multifunctional

additivemanufacturing(MFAM)asapivotaltechnologyin

advanc-ingthefuture ofAM.Effortshave beenmadebyresearchersto

develophybridsystemstoachieve MFAM,onesuchexample is

theworkbyLopesetal.[2]wherestereolithographyand direct

printtechnologiesarecombinedtorealizeadditivelymanufactured

electronicdevices.Such hybridapproaches oftenrequire

multi-plemachine/print-restartsandadditional manualor automated

accompanyingprocedures.Conversely,multi-headinkjetprinting,

apromisingMMAMprocessallowsforMFAMdesignstoberealized

withgreaterdegreeofmanufacturingfreedominasingleoperation

∗Correspondingauthor.

E-mailaddress:[email protected](A.Panesar).

byco-depositingstructuralandfunctionalinks.Recentworks[3–5]

havedemonstratedthepotentialofjettingforelectronics

appli-cationsand highlighted theconsiderable ongoing researchinto

materialsandprocessdevelopmentforMMAM.

StepstowardsexploitingthedesignfreedomsofAMhavealso

beenmade.OpenFab[6]definedaproceduretoefficientlygrade

mechanicalpropertiesthroughthevolumeofapartandPonche

etal.[7]proposedathreestepglobaldesignapproachwiththe

aimofbetterintegrationofthedesignrequirements(functional

specification)withtheAMprocess.However,therehasbeenlittle

workcarriedouttodateondevelopingthedesignphilosophies

torealize novelMFAM concepts.Theauthors considera closer

interplaybetweentheMMAMandtopologyoptimization(TO−

astructuraloptimizationtechniquethatiterativelyimprovesthe

materiallayoutwithinagivendesignspace,foragivensetofloads

andboundaryconditions[8])keytoprogressingtheMFAMdesign

paradigm.Onedirectbeneficiaryofthisistheareaof3Dprinted

electronics,asfabricationofruggedstructuresthatembed

non-traditionalelectronicsystemsinanarbitraryformbecomepossible

[2,9].Thisapproachhasthepotentialtopavethewayfor

light-weight,morecompact,betterintegratedandmoreoptimaldesigns.

TOwithembeddedcomponentshasbeeninvestigated

previ-ouslyforthe“integratedlayoutdesign”problem[10–13],where

theaimhasbeentofindboththeoptimalplacementand

orienta-tionofcomponentsandtheoptimalconfigurationofthematerial

simultaneously.Thiswasachievedbyiterativelychangingthe

geo-http://dx.doi.org/10.1016/j.addma.2017.05.009

(2)

Fig.1.Multifunctionaldesign−a)top-leveldiagramshowingcoupledoptimizationofstructureswithembeddedsystem,b)multi-materialjettedconceptprototypeshowing anoptimizedstructuralpartthatembedsaninternalsystem(comprisedofplacedcomponentsandtheassociatedrouting)intendedforthepurposeofstructuralhealth monitoring.

Fig.2.ExampleofaTOautomotivecomponent(transmissionmount)withembeddedsystems−demonstrating3Dplacementofcomponentsandassociatedroutinga) perspectiveview,b)topview.

metricaldesignvariablesdescribingthelocationandorientationof

componentswithinaTOregime.Inallthereportedinstances,the

implementationremainedlimitedtotheinclusionofrigid

compo-nentsthatsolelyintendedtoenhancethestructuralperformance.

Therefore,thisworksseekstomakeastepforwardtowardsthe

inclusionof functionalcomponents(orasystem) ina structure

enablingtheoptimizationofaMFAMdesign.Torealizethisaim,

i.e.optimizethedesignofamultifunctionalpart,couplingboththe systemdesign(whichisbasedonafunctionalperformancei.e.not

limitedtoonlycapturingstructuralormechanicalbehaviour)and

structuraldesign,asillustratedinFig.1,isneeded.

EarlierworkbytheauthorsdetailedaMFAMdesignframework

[9]torealizeafunctionalsystem,specificallyaimedat3DPrinted CircuitVolume(PCV)applicationsi.e.printedelectronicsintrue 3D−notlimitedto,forexample,printingonsurfacesasin[2]or thestacked2D(i.e.2.5D)PrintedCircuitBoard(PCB)paradigm[14]. Thisworkpresentedmethodsfortheintelligentplacementof func-tionalcomponentsatsuitablesites,andtheassociatedroutingfor

theconductivepathwayswithinapartmanufacturedusing

multi-materialjetting.Moreover,effortsweremadetointegratethese

aspectsofsystemdesignintoaTOproceduresuchthatthefinite

elementanalysis(FEA)conductedaspartofaTOaccountedfor

theupdatedmaterialproperties,reflectingsystemattributes[15].

Subsequently,thiswasextendedtobenefitfromabi-directional

couplingbetweentheTOandsystemdesignbutthe

implementa-tionremainedlimitedtoaspecificroutingmethodandsuffered

fromrobustnessissues[16].

ThecapabilityfordesigningMFAMconceptsforPCV

applica-tionisinitsinfancyandthereforethisworkfocusesondeveloping agenericcoupledoptimizationstrategyfortherealisationof

struc-tures with embeddedfunctional systems that are intended for

manufactureusingmulti-materialjetting.Thepapertakesthe

fol-lowing structure: firstly, a description of design for functional

systemsisprovided;secondly,thestructure-systemcoupling

strat-egy ispresented;thirdly, theheuristic definitionof thesystem

sensitivities is detailed (so that one can tackle a bi-directional

structure-system coupling);and lastly,theappropriatenessand

robustnessofthisstrategyisdemonstratedbyevaluatingand dis-cussingtheresultsforexampletestcases.

2. Methodology

Avoxelmodelingenvironmentischosenforseamless

transi-tionbetweensystemdesign,numericalanalysisandmanufacture

astheyallrelyondiscretizedvolumetric space[9].Specifically,

voxelsforsystemdesign,hexahedralelementsforFEAand2D

pix-elswithassociatedlayerthicknessintheraster-based(bmp)file

formatemployedinjetting.Adoptionofthevoxelmodeling

envi-ronmenteliminatestheneedformanualcomputer-aided-design

operations,includingconversiontothecommon

STereoLithogra-phyfileformatandassociatedslicing,whichiswellknowntobe

cumbersomeanderrorprone.

2.1. Functionalsystemdesign

The key enablers for making the functional system design

possibleare:(i)intelligentcomponentplacementand(ii)the

asso-ciatedconnectionsroutedbetweenthem.Forsimplicity,theseare

referredtoasplacementandrouting.Althoughadvancementsin

(3)

Fig.3.Flowchartshowingthecoupledoptimizationprocedure.

Fig.4.The2Dcantileverproblemwithfunctionalsitestobeconnectedtoformthe systemindicatedbycircles.

andmathematicalmethods,asreportedin[17–19],the

implemen-tationshavebeen,atbest,gearedatthe2.5Dparadigm.Manyofthe

PCBdesignstrategieshavebeenadaptedandcoupledwithglobal

evolutionaryoptimizationalgorithmstosolveoptimization

prob-lemsinotherfields.Examplesinclude,GeneticAlgorithms(GA)and

AntColonyOptimization(ACO),employedforpipe/cablerouting

problems[20–23]andoptimumplacementproblemsinstructural

healthmonitoringapplications[24,25].

Tosuccessfully realize structures withembedded functional

systems,oneneedstoexplorethetrue3Ddesignfreedomsoffered

byMFAMinanautomatedsenseandearlierworkbytheauthors[9]

discusseshowthiscanbeachieved.Thekeyelementsofthe

pro-posedmethodsandtheirfoundationalprinciplesaresummarised

below:

i.theintelligentplacementofcomponents− locationselection

basedonaperformanceand/orgeometrycriterion;component

orientationidentified usingskeletalinformation. Theskeletal

information (or the 1D medial axis) of a part’s topology is

obtainedusingathinningalgorithm,asdetailedin[26,27].

ii.thegenerationofconnectionstoformacircuit,i.e.routing−two

approachesproposedtoachievethisare:i)approximate

rout-ingwhichemploysDijkstra’salgorithm[28]tofindtheshortest

pathona1Dmedialaxis(i.e.skeletal)graph,ii)accurate

rout-ingwhichusesanimplementationoftheFastMarching(FM)

method[29].

Fig.2showsanexamplecapturingthefunctionalsystemdesign

capabilityin3D.Here,thehighlevelofaccuracyachievedusingthe

FMimplementationforidentifyingtheshortestroutesbetweena

setofcomponentsthatarenotlimitedtoplanarorientation

con-straints(asisthecasewith2.5DPCBparadigm)isdemonstrated. Inprinciple,itisbesttoperformroutingandplacement

concur-rentlybutastheyhavenesteddependencies,realizingasolution

toapracticalproblemmaybecomeunviable.Thereforethesetwo

stepsareoftentackledindependently.Withregardstothesystem

designconsideredhere,placementofa componentiskeptfixed

andthemethodofaccurateroutingisimplemented.Thesystem

optimizationproblemthereforebecomesaroutingoptimization

problemwiththeaimistoimprovethecircuitefficiencyby

low-eringenergylosses,whichareproportionaltotheresistance.This

isachievedbyidentifyingtheshortestpathsbetweencomponents

(asresistanceisproportionaltothepathlength)subjectto vari-ousdesignrulesandconstraints.Bydoingso,wealsominimisethe

utilizationoftheexpensivematerialusedforconductivetracks,

providingfurthereconomicbenefit.

2.2. Couplingstrategy

NotableTOmethods,suchasSIMP[8], BESO[30]and

level-set[31],havebeenprincipallydeveloped toidentifythestiffest

structure(throughcomplianceminimization)foragivenmass

con-straint,andthoughtheirimplementationshavebeendemonstrated

(4)

con-Fig.5.Effectsofsystemintegrationincludedintheprocessofstructuraltopologyoptimization−a)Sensitivitiesforstructure,b)sensitivitiesforinternalsystem,c)combined sensitivities,d)resultingcoupledsolution,ande)TOusingjuststructuralsensitivitiesforcomparison.

sideraclassiccantileverplateorbeamproblemandthereforethis workalsoutilizessuchastructuretoinvestigatetheproposed cou-plingstrategy.

BothBESOandlevel-setimplementationsrevealawell-defined

boundaryateveryiterationoftheoptimizationprocess,allowing

forembeddedsystemrealisation(asdiscussedin[9]).However,it isonlytheformerthatprovidesthestructuralelementalsensitivity

withinthevolumeoftheboundaryenablingtheoptimization

pro-cedureofFig.3 tocombinethiswiththecorrespondingsystem

sensitivities to govern the evolution of the combined

(struc-ture+system)solution.InthispapertheTOprocedure(indicated inFig.3)isbasedontherevisedBESOmethod[30,34,35]orwhat onemightcallthediscreteSIMPmethod[36].

Thecompliance(ameasureoftheinverseofstiffness)basedona

purelystructuraloptimizationproblemcanbestatedasminimise:

C= 1 2U

TKU (1)

subjectto:

VfN

i=1

Vixi=0 (2)

whereCisthetotalstrainenergy(commonlytermedcompliance),

KistheglobalstiffnessmatrixandUistheglobaldisplacement

vector,Viistheelementalvolumefraction,Vf isthestructural tar-getvolumefractionconstraint,andxi=xminor1,forvoidorsolid

regions,respectively.Forthiswork,xminwassetat1e-6.ThisBESO

formulationusesamaterialinterpolationscheme:

E(xi)=E1xpi (3)

whereE1istheYoung’smodulusforthesolidregion,andpisthe

penaltyexponent.Asdetailedin[22,23],andexplainedby

Bend-soeandSigmund[8],thesensitivityoftheobjectivefunctionwith respecttotheithelementis

C

xi

=−px

p−1 i

2 u

T ik

0

iui (4)

wherek0

i denotesthestiffnessmatrixanduirepresentsthe

dis-placementvectorfortheithelement.Therelativerankingofthe

elementalsensitivitiesforbothsolidandvoidregionelementscan

beexpressedas

S˛ i=

−1

p

C

xi

=

1 2u

T ik

0

iui when xi=1

xpmin−1

2 u

T ik

0

iui when xi=xmin

(5)

whereS˛

iisthestructuralsensitivitynumberforthei

thelement.It

canbeseenthatthesensitivitynumberforthesolidregionisequal totheelementalstrainenergy,andthesensitivitynumberforthe

voidregionisdependentonthevalueofp.Forthiswork,pwas

selectedtobe1,andthereforethestructuralsensitivitynumber

wasequaltothestrainenergyforbothregions.

2.3. Heuristicsensitivitydefinition

Under a generaloptimization framework,one would ideally

statetheobjectivefunctionintermsoftheproblemspecific vari-ablesinorderthatthesensitivitiesofdesignvariablescanbeeasily

obtained.However,incaseswheredoingthisbecomeschallenging

(5)

Itrlimit Optimizationiterationsafterwhichtheprocessisterminated 80

onecanattempttosolvethetwoproblemsasiftheywereoverlaid. Thisnecessitatesthedefinitionofappropriatesensitivitiesforeach classofproblemandtheirimplementationinacombinedobjective function.Asthisworkfocusesonexampleswheretheplacement locationofcomponentsis pre-determined/specified,thesystem associatedelementalsensitivitiescanbedeterminedexclusively fromtheroutingaspectofsystemdesignusing

R˛ i=

1 1+di

(6)

where,diisthedistance(Euclidianmeasure)betweenthe‘ith’ ele-ment(withinthediscretizeddesigndomain)andtheclosestpoint fromitontheroutedpaths.Thisassignsalowervalue(between0 and1)forelementsthatarefurtherawayfromaroutedpathanda valueof‘1tothoseelementswhichformaroute.

Thesystemelementalsensitivitiescanbecombinedwiththose ofthestructuralcounterpartusingaweightedsumapproachas shownin

C˛

i=ω1×S˛i+(1−ω1)×

R

˛i

(7)

where,S˛

irepresentsthenormalized(datarangebetween0and 1)structural elementalsensitivities(i.e.strainenergydensities) afterthresholdingtheoutliers(i.e.atpointofloadingandboundary conditions)andR˛

irepresentsthenormalizedsystemelemental sensitivities.ω1 ∈[0,1] isauserdefinedweighttocontrolthe rel-ativeinfluenceofthesystemdesignontheoverallcoupledsolution. Preliminaryworkbytheauthors[16]oncouplingsystemdesign

withTOconsideredthecantilever plateproblemof Fig.4.Four

componentlocations(shownascircles)wereidentifiedusingthe

resultanttopologyofastructure-onlyoptimization(seeFig.5e) sothattheeffectofthesysteminclusionontheresulting

struc-turewhenemployingacoupledoptimization,canbeinvestigated.

A pair-wiserouting between component locationsutilizing the

approximate routing method[9] (i.e.routes constrained tothe

medialaxisofthestructure)wasconsideredforsystemdesign.

ThecouplingemployedthealgorithmofFig.3andutilizedEqs.

(6)and(7)toassesssolutions.Fig.5presentstheeffectsofsuch

asystemintegrationwhenincludedintheprocessofTO.

How-ever,thisimplementationsufferedfromrobustnessissuesarising

fromtheunstableevolution(poorconvergence)oftheobjective

functionduringtheiterativeprocessoftheoptimization[16].This

wasessentiallyduetothesuddenchangesintheskeletal

topol-ogyresultingfromanydis-connectivityinstructuralmembers.It

isnoteworthythattheskeletaltopologygovernedthesystem

con-figurationintheaforementionedworkastheapproximateroutes

weremadetoadheretotheskeletaltopology. Inthis work,an

improvedcouplingstrategyisdevelopedthataddressesthis

prob-lemandusesamoreaccurateroutingmethod.

Thetwo key advancesmadein this papertowardsthe

cou-pledoptimizationalgorithmofFig.3are:i)animprovedheuristic definitionforsystemassociatedelementalsensitivitiesandii)an

improvedmethodofcombiningthesystemelementalsensitivities

Fig.6. Theconsideredextruded2D(2voxeldeep)cantileverproblem.

withthestructuralsensitivities.Fortheformer,theR˛

ivaluesare

‘bounded’toavoidunwantedinfluenceinareasofthetopologynot

pertinenttotherouting,bysettingR˛

itozerofordigreaterthan

thefilterradiusorRmin(seeTable1).

R˛ i=

1 1+di

(bounded) (8)

andforthelatter,anadaptiveparameterdefinedas

= S i R i (9)

whichismultipliedbyR˛

itoensureanappropriate

contribu-tionofthesystemassociatedelementalsensitivitiestowardsthe

combinedsensitivity,whichisnowdefinedas

C˛

i=ω1×S˛i+(1−ω1)×

×R˛

i

(10)

Although, Eqs. (7) and (10) share the same weighted sum

approachonthestructureandsystemsensitivitiesthatisnecessary fortheupdateofdesignvariables,itistheimprovedformulation intheheuristicsthattheauthorsevaluateanddiscussinthenext

sectionthatresultsinasignificantimprovementontheprevious

formulation.

3. Simulations,resultsanddiscussion

Thisworkraises tworesearch questionswhich allowfor an

assessmentoftheappropriatenessandrobustnessoftheproposed

couplingstrategy.

Researchquestion−I.Whatinfluencedoestheheuristic

sensi-tivitydefinition(specifically,theuseofR˛

i(bounded)and)have

onthecoupledsolution?

Researchquestion−II.Whetheritisadvantageoustoperform

acoupled(structure+system)optimizationornot?(Asa

compar-ativeuncoupledsolution,theauthorsrefertoperformingaTOto

obtainastructureandsubsequentlyaddingthesystemdesign)

Eachresearchquestionisbesttackledwithhelpofatestcase.

(6)

Fig.7. CoupledoptimizationforatargetVfof0.4witha)i(unbounded)and(notincluded),b)i(unbounded)and(included),c)i(bounded)and(notincluded), d)R˛

i(bounded)and(included).

Fig.8. CoupledoptimizationforatargetVfof0.25witha)i(bounded)and(notincluded),b)i(bounded)and(included).

showninFig.6(withtheleftedgefixedandavertically

down-wardforcebeingappliedtothebottomrightcorner).Theextrusion

extentofthedesigndomainwaskeptto2voxelsasthisallowed

formultipleroutestoexistbetweencomponentswithoutoverlaps.

Thisisbecauseroutesarenolongerlimitedtobeingplanarandcan existin3D.Twocomponentlocations(shownascircles)were iden-tifiedusingtheresultanttopologyofastructure-onlyoptimization soastobetterunderstandtheeffectofthesysteminclusiononthe

resultingstructurewhenemployingacoupledoptimizationwith

theimprovedheuristics.

Theparametersusedtosimulatethetestcasesarereportedin

Table1.Theseincludestandardvaluesforthematerialdensityof

solid(structure)andvoidregions,specifically,1andamuchlower valueof1e-6,respectively.Anintermediatevaluewasemployedfor

thematerialdensitycorrespondingtotherouting(system).This

choicewasmadesothatonlyaveryweakcontributioncouldbe

madebythefiniteelementsrepresentingtheroutesonthetotal

strainenergyofthestructure(i.e.measureofstructuralstiffness) asthesystemisnotintendedtobeloadbearingforthisstudy.Equal weightingwaschosenfortheobjectivefunctionformulationasthe focusofthisstudywasnotoninvestigatingtheinfluenceof weight-ingparameter(ω1).Theelementalstrainenergiesforthis study

(7)

Fig.9.Comparisoninresultsforuncoupledandcoupledoptimization−a)routingperformedona(converged)topologyoptimizedstructure,b)routingperformedasa coupled(structure+system)optimizationproblem.

Fig.10.Applicationtoamorecomplexproblem−a)coupledsolution,b)combinedsensitivity.

Nastran[37])onauniformmeshhavinganedgelengthof1unit

(Fig.6showsthestructureas128unitsby64unitsindimension).

ResultsforthesimulationofTestCaseI(addressingresearch

questionI)arepresentedinFig.7.Hereweanalyseandattempt

tounderstandtheeffectthatR˛

i(bounded)andhave

(indepen-dently)onthesolution.Fig.7ashowsacharacteristicstraight-line

routebetweenthetwo placementlocations,even whensimilar

weighting for structure and system contributions (i.e. ω1=0.5)

wereused.Thissystemdominatedsolutioncanbebestunderstood

byinspectingthecombinedsensitivities(C˛

i)ofthevoxels

repre-sentingtheroutewhichinthiscasehappenstoacquireanearly

constantandnotablyhighervaluefortheentireroute.Witha50%

weighting,thisshouldnotbethecaseasboththestructureand

systemshouldcontributemoreorlessequallytowardsthesolution.

InFig.7b, withtheinclusion oftheadaptive parameter, a

morebalancedcontributionofthesystemsensitivitiestowardsthe combinedsensitivitiesisachieved.Thisisevidentfromthe

vari-abilityin combinedsensitivityvalues forthevoxels comprising

theroute.However,eveninthiscaseacharacteristicstraight-line routeisobtainedforthefinalsolution,indicatingacleardominance ofthesystemsensitivitytowardsthecombinedsensitivity.Thisis duetoR˛

ibeingunbounded,resultinginanunnecessarilylarge

regionofinfluenceoverwhichthesystemsensitivitiesplayarole,

whichmakesthestructureneartheroutes unnecessarilybulky.

Thissystemsensitivitydominancemakesthepossibilityofelement

removalneartherouteverylow,causingthesystemsolutionto

havenonoticeablechangewiththeoptimizationiterations.

Conversely,a more intuitivesolutionwhere thesystem and

structureseemtoevolvecompetingwitheachothercanbeseen

inFig.7c-dwherethevaluesofR˛

iarebounded.However,only

marginaldifferencescanbeseenwhencomparingthevoxels

rep-resentingtheroutesinthecoupledsolutionsofFig.7candd.This canbeattributedtotherelativelyhighstructuralvolumecompared tothesystemvolume,makingtheeffectof(inFig.7d)less notice-able.Totestthishypothesisfurther,anothercoupledoptimization wasperformedbutthistimethetargetVf wassetas0.25instead

ofthepreviouslyusedvalueof0.4.

Table2

ComparisoninstructuralperformancebetweensolutionsofFig.8(withandwithout theinclusionof␭).

Fig.8a(␭−notincluded)Fig.8b(␭−included)difference

TotalStrainEnergy88.2 67.4 23.6%

Table3

Comparisoninperformancebetweenuncoupledandcoupledoptimization.

Uncoupledsolution Coupledsolution difference

Pathlength(pixels) 196 138 30%

TotalStrainEnergy 39.9 40.2 <1%

Max.displacement(pixels) 76.7 77.4 <1%

Withthelowervaluefortheratioofstructuretosystemvolume inthesolutionofFig.8(wherethecoupledoptimizationproblem

isperformedforatargetVf of0.25)ascompared tothatinthe

caseoftargetVf=0.4,amorenotableinfluenceofcanbeseen−

boththeresultingtopologyofthestructureandthepixels compris-ingtherouteareseentodevelopduringtheoptimizationwiththe

improvedheuristics.Itiscommonforsolutionstogettrappedin

localminimainanon-convexproblemspace,suchasthatseenin

discreteSIMPorBESO.Asaconsequence,theevolutionofthe solu-tionishistorydependent,thedifferentcontributionsfromrouting sensitivities(i.e.withorwithouttheincorporationof)playa dom-inantrole.Thisissupportedbythe23.6%differenceintotalstrain energyvalues(ameasureforcomplianceorinverseofglobal stiff-ness)observedforthesolutionsofFig.8aandb(valuesreportedin

Table2).Moreover,itisclearlyevidentfromthewell-distributed andsmoothlytransitioningvaluesforthecombinedsensitivitiesin

Fig.8bthat:i)havingR˛

ibounded,andii)theinclusionofthe

adap-tiveparameter,isnecessarytoobtainareliablesolutionasonly thencanoneensuretheappropriatecontributionofthesystem sen-sitivitiestowardsthecombinedsensitivitiesforamultifunctional problem.

ResultsforthesimulationofTestCaseIIaddressingresearch

questionII(i.e.whethertoperform(structure+system)coupled

(8)

pre-sentedquantitativelyinTable3.Thepixelscomprisingtheroute

areverydissimilarforcoupledanduncoupledsolutionsandthis

istobeexpectedasthesystemdesignisconstrainedtothefinal

topologyofthestructureintheuncoupledproblem.Byadoptinga

coupledproblemformulation,nearly30%reductioninpathlength

isachieved.Importantly,despitethevisualdifferencesinstructural topology,negligibledifference(<1%)inthevaluesfortotalstrain energyisobserved.Fromtheconsideredexample,itcanbeinferred

thatthesystemdesignbenefitsunderacoupledoptimization

for-mulationwhilsthavinglittleeffectonthestructuralperformance.

Togainabetterassessmentoftheeffects ofcouplingonthe

structuralperformance,amorecomplexroutingscenariowas

con-sidered where two component pairs are chosen such that the

placementlocationslieoutsidetheresultingstructurewhenone

considersthetopologyoptimizedgeometryofanuncoupled

prob-lem (specifically, topology of Fig. 9a). Fig. 10a shows how the

structuraltopologyhasevolvedtoaccommodatetheseroutes

join-ingthetwo componentpairsincase ofacoupledoptimization.

Fromapurelystructuralperformancestand-point,thetopologyof

Fig.10ahas15%highertotalstrainenergy(i.e.lessstiff)as com-paredtothatofFig.9a.However,itmustbenotedthatthelatter

hasundergoneasingle-objectiveoptimization(whereonlystrain

energyorinverseofstiffnesswasminimised)whilsttheformer

hasbeensubjectedtothecoupledoptimizationwhereaweighted

sumapproachisconsideredforoptimization.Suchacoupling,in

principle,shouldensurethebestmateriallayoutalong-side

opti-malsystemconfiguration(inthis work−shortestroutes).This

isbecausetheroutingbetweenthecomponentscan’tberealized

onthestructuraltopologyofFig.9aforthespecifiedplacement

locations.

Theproposedcouplingmethodhasshownpromisewhen

tack-lingMFAMproblemandas thecapabilityof thesystemdesign

employedfor this work hasalready beendemonstrated for 3D

applications[9],onecan considerthisstrategy withconfidence

(initsnativeformorbyextendingitfurtherbyapplyingto

differ-entclassesofproblemand/ordifferentoptimizationalgorithms)

tosolvesimilar(structure+system)coupledproblemsin3D.It

mustbenotedthatthismethodcatersforthemultipleobjectives

viaa singleweightedsumapproach andconsequentlywill

suf-ferfromthelimitationsofthisformulation.Whereweightedsum

approaches mayprovetobeinadequate, amore generic

multi-objectiveconsiderationsuchasthepareto-frontcriterioncanbe

considered.

4. Concludingremarks

Thispaperhaspresentedacoupledoptimizationformulation

whichallowsfortheoptimalmaterialandsystemlay-outtobe

identifiedasittacklesasystemdesignproblemoverlaidona struc-turaldesignproblem.Although,theimmediateapplicationforthis

developmentis enablingthe designof additively manufactured

(jetted)multi-materialpartswithembeddedfunctionalsystems,

forexampleastructuralpartwithelectricalcomponentryand con-ductivetracks,nevertheless,thestrategypresentedhereinshould beconsideredfortacklingamoregeneralclassofengineering prob-lems.Forinstance,civilengineeringstructures(buildings/bridges) thatincorporatesystems(pipes/cables).Thiscoupledoptimization

developmentmarksasignificantsteptowardsbeingabletoexploit

thedesignfreedomofferedbyMMAM.

Themaincontributionofthispaperistheimprovedheuristic

definitionthatallows fora moreappropriate couplingstrategy,

wherethesystemdesignisperformedconcurrentlywiththe

struc-turaloptimization.Thisisachievedbyaccommodatingtheeffects

ofsystemincorporationonthestructuralresponseofthepartat

everyiterationwithinamodifiedbidirectionalevolutionary struc-turaloptimization.

Thesimulationresultsfortheevaluatedextruded2Dcantilever

testcasesshowthesuitabilityoftheproposedcouplingmethod

where thesystem sensitivities,specificallyrouting sensitivities, arecombinedwiththestructuralsensitivitiesforamultifunctional designproblem.Theauthorsbelievethiscontributionwillprovide

thenecessary design-innovationand in-turn themanufacturing

incentivetorealizemultifunctionalAMorMFAMproducts.

Acknowledgment

ThisworkwassupportedbytheEngineeringandPhysical

Sci-encesResearchCouncil[grantnumberEP/I033335/2].

References

[1]E.MacDonald,R.Wicker,Multiprocess3Dprintingforincreasingcomponent functionality,Science353(6307)(2016)(p.aaf2093-aaf2093).

[2]A.J.Lopes,E.MacDonald,R.B.Wicker,Integratingstereolithographyanddirect printtechnologiesfor3Dstructuralelectronicsfabrication,RapidPrototyp.J. 18(2)(2012)129–143.

[3]J.Vaithilingam,E.Saleh,C.Tuck,R.Wildman,I.Ashcroft,R.Hague,P.Dickens, 3D-inkjetprintingofflexibleandstretchableelectronics,Proceedingsofthe 26thSolidFreeformFabricationSymposium(2015).

[4]E.Saleh,J.Vaithilingam,C.Tuck,R.Wildman,I.Ashcroft,R.Hague,P.Dickens, 3Dinkjetprintingofconductivestructuresusingin-situIRsintering, Proceedingsofthe26thSolidFreeformFabricationSymposium(2015). [5]S.-P.Chen,H.-L.Chiu,P.-H.Wang,Y.-C.Liao,Inkjetprintedconductivetracks

forprintedelectronics,ECSJ.SolidStateSci.Technol.4(4)(2015) P3026–P3033.

[6]K.Vidimce,S.-P.Wang,J.Ragan-Kelley,W.Matusik,OpenFab:a

programmablepipelineformulti-Materialfabrication,ACMTrans.Graphics 32(4)(2013)1–11.

[7]R.Ponche,J.Y.Hascoet,O.Kerbrat,P.Mognol,Anewglobalapproachtodesign foradditivemanufacturing,VirtualPhys.Prototyp.7(June(2))(2012)93–105. [8]M.P.Bendsoe,O.Sigmund,TopologyOptimization:Theory,Methodsand

Applications,Springer,2003.

[9]A.Panesar,D.Brackett,I.Ashcroft,R.Wildman,R.Hague,Designframework formultifunctionaladditivemanufacturing:placementandroutingof3D printedcircuitvolumes,J.Mech.Des.137(11)(2015).

[10]J.Zhu,W.Zhang,P.Beckers,Integratedlayoutdesignofmulti-component system,Int.J.Numer.MethodsEng.78(2009)631–651.

[11]J.H.Zhu,W.H.Zhang,Integratedlayoutdesignofsupportsandstructures, Comput.MethodsAppl.Mech.Eng.12(Jan)(2010)557–569.

[12]W.Zhang,L.Xia,J.Zhu,Q.Zhang,Somerecentadvancesintheintegrated layoutdesignofmulticomponentsystems,J.Mech.Des.133(10)(2011) 104503.

[13]T.Buhl,Simultaneoustopologyoptimizationofstructureandsupports,Struct. Multidiscip.Optim.23(5)(2002)336–346.

[14]E.Beyne,3Dsystemintegrationtechnologies,InternationalSymposiumon VLSITechnology,Systems,andApplications(2006)1–9.

[15]A.Panesar,D.Brackett,I.Ashcroft,R.Wildman,R.Hague,Designoptimization strategyformultifunctional3Dprinting,25thInterationalSolidFreeform FabricationSymposium(2014)592–605.

[16]A.Panesar,D.Brackett,I.Ashcroft,R.Wildman,R.Hague,Designoptimization formultifunctional3Dprintedstructureswithembeddedfunctionalsystems, 11thWorldCongressonStructuralandMultidisciplinaryOptimisation(2015). [17]N.Abboud,M.Grötschel,T.Koch,Mathematicalmethodsforphysicallayoutof printedcircuitboards:anoverview,ORSpectr.30(May(3))(2007)453–468. [18]T.F.Chan,J.Cong,J.R.Shinnerl,K.Sze,M.Xie,MultiscaleOptimizationinVLSI

PhysicalDesignAutomation,MultiscaleOptimizationMethodsand Applications(NonconvexOptimizationandItsApplications),vol.82, Springer-Verlag,2006,pp.1–67.

[19]S.Areibi,Z.Yang,EffectivememeticalgorithmsforVLSIdesign=Genetic algorithms+localsearch+multi-levelclustering,Evol.Comput.12(Jan(3)) (2004)327–353.

[20]J.-H.Park,R.L.Storch,Pipe-routingalgorithmdevelopment:casestudyofa shipengineroomdesign,ExpertSyst.Appl.23(Oct(3))(2002)299–309. [21]X.Ma,K.Iida,M.Xie,J.Nishino,T.Odaka,H.Ogura,Ageneticalgorithmforthe

optimizationofcablerouting,SystemsComput.Jpn.37(Jun(7))(2006)61–71. [22]G.I.F.Thantulage,AntColonyOptimizationBasedSimulationof3DAutomatic

Hose/pipeRouting.PhD,BrunelUniversity,2009.

[23]C.Velden,C.Bil,X.Yu,A.Smith,Anintelligentsystemforautomaticlayout routinginaerospacedesign,InnovationsSyst.SoftwareEng.3(May(2)) (2007)117–128.

(9)

References

Related documents

The aim of this study was to determine dermoscopy findings of alopecia areata that could be used as a clinical indicator of disease and also to evaluate the

We applied Pittsburgh Sleep Quality Index to screen the presence of sleep pathology and Epworth Sleepiness Scale to evaluate subjective daytime sleepiness; we used actigraphy and

Notch signaling regulates neuroepithelial stem cell maintenance and neuroblast formation in Drosophila optic lobe development. Stem cells: Units of development, units of

We propose a trajectory planning frame- work that utilises information of the traffic vehicles to identify safe driving zones on the road using potential field functions and a

Encounter ✦ healing, Martial Standard Action Melee touch Target: You or one adjacent ally. Effect: The target can spend a

The main contribution of this study is to decrease income inequality in risk sharing network such as informal insurance through an analytical method, this study

To explore the effectiveness of the early fusion ap- proach for fusing cross-modal features used in the retrieval model, we visualize the similarity scores (warmer colors =

The classification of loops is based on the way the loops flow on the hand (not the card), so that on a fingerprint card for the left hand, loops flowing toward the thumb