A case-based reasoning approach for non-traditional machining ...
Transcript of A case-based reasoning approach for non-traditional machining ...
311
AdvancesinProductionEngineering&Management ISSN1854‐6250
Volume11|Number4|December2016|pp311–323 Journalhome:apem‐journal.org
http://dx.doi.org/10.14743/apem2016.4.229 Originalscientificpaper
A case‐based reasoning approach for non‐traditional machining processes selection
Boral, S.a, Chakraborty, S.a,* aDepartment of Production Engineering, Jadavpur University, Kolkata, West Bengal, India
A B S T R A C T A R T I C L E I N F O
Tosustaininthemoderneraofrapidmanufacturingdevelopment,itbecomesnecessarytogeneratecomplexshapesonmaterialswhicharehighlytempera‐ture and corrosion resistant, hard to machine, and have high strength‐to‐weightratio.Generationofcomplexshapesonthosematerialsusingconven‐tionalmachiningprocessesultimatelyaffectssurfacefinish,materialremovalrate, accuracy, cost, safety etc. Non‐traditional machining (NTM) processeshave the capability to machine those advanced engineering materials withsatisfactoryresults.But,selectionofthemostappropriateNTMprocessforaparticularmachiningapplicationisoftenacomplicatedtask.Case‐basedrea‐soning (CBR),adomainofartificial intelligence, isaparadigm for reasoningnewproblemsfromthepastexperience.InCBR,amemorymodelisassumedfor representing, indexing and organizing past similar cases, and a processmodelissupposedforretrievingandmodifyingthepastcasesandassimilat‐ing the new ones. This paper primarily focuses on the application of CBRapproachforNTMprocessselection.Basedondifferentprocesscharacteris‐tics and process parameter values, the past similar cases are retrieved andreusedtosolveacurrentNTMprocessselectionproblem.Forthis,asoftwareprototypeisdevelopedandthreerealtimeexamplesarecitedtoillustratetheapplicationpotentialityofCBRsystem.
©2016PEI,UniversityofMaribor.Allrightsreserved.
Keywords:Non‐traditionalmachiningprocessesProcessselectionArtificialintelligenceCase‐basedreasoning
*Correspondingauthor:[email protected](Chakraborty,S.)
Articlehistory:Received7June2016Revised25October2016Accepted12November2016
1. Introduction
With thedevelopmentofnewermaterialshaving improved thermal,mechanical and chemicalproperties,ithasnowbecomequitedifficulttomachinethosematerialsusingconventionalma‐chiningprocesses.Theseprocesses,generallybasedoncuttingandabrasionmechanism, incurhighermachining costwhile generating complex shape features on composites, ceramics andotheradvancedengineeringmaterials.Theachievedsurfacequalityanddimensionalaccuracyofthemachinedcomponentsarealsonotsatisfactory,andoftenfailtomeetthedesiredtarget.Inthesemachiningprocesses,unwantedmaterialfromtheparentworkpieceisgenerallyremovedemployingmechanicalenergy.Thisenergyissuppliedbymeansofacuttingtoolkeptincontactwiththeworkpiece,causingsheardeformationalongtheshearplane,leadingtochipformation.Newexoticworkmaterialsandcomplexgeometricalshapesonthosematerialshavebeenput‐tingmorepressureonthecapabilitiesoftheconventionalmachiningprocesses.Thisleadstothedevelopment and deployment of a new set ofmachining processes, popularly known as non‐traditionalmachining(NTM)processes.Intheseprocesses,unwantedmaterialisremovedfromtheparentworkpieceusingvariousformsofenergy,likechemical,thermal,mechanical,electri‐calorcombinationofthoseenergies.InanNTMprocess,thereisnodirectcontactbetweenthe
Boral, Chakraborty
312 Advances in Production Engineering & Management 11(4) 2016
cuttingtoolandtheworkpiece.Inabrasivejetmachiningprocess,excessmaterialisremovedbymeans ofmicroscopic chips and in electrochemicalmachining process by electrolytic dissolu‐tion. In laserbeammachiningprocess, there is evennoneedofanycutting tool. It isalsonotnecessarythatthecuttingtoolshouldbeharderthantheworkpiecematerialinanNTMprocess.Now‐a‐days, ithasbecomeeasier togeneratecomplexshapesonmaterials, likesteel, carbide,titaniumanditsalloys,ceramics,superalloys(Inconel718,hastelloy)etc.employingNTMpro‐cesses[1,2].Tilldate,therehavebeenapproximately20NTMprocessesdevelopedandappliedinmodernmanufacturingindustries.SelectionofthebestsuitableNTMprocessforaparticularworkmaterialandshapefeaturecombinationisgenerallymadebyadomainexpertonthebasisofvariousfactors,suchasworkpiecematerial,shapefeaturetobegenerated,materialremovalrate,surfacefinish,surfacedamage,cornerradii,tolerance,cost,safety,powerrequirementetc.Thus,anexpertinthisdomainmusthaveavastandin‐depthknowledgeaboutthecharacteris‐ticsandcapabilitiesofdifferentavailableNTMprocesses.But,inthepresentmanufacturingsce‐nario,most of the process engineers lack the requisite domain knowledge and availability ofexpertsisalsosometimesconstrained.
Usually,adomainexpertacquiresknowledgefromthepastexperienceaswellasfromotherreliablesources.Takingthisconceptasaplinth,whenanexpertattemptstoselectanNTMpro‐cess foragivenmachiningapplication,he/she justrecalls thesimilarpastsituationsandtheirsolutions.Thus,basedonthesimilarpastproblemsandtheirsolutions,newNTMprocessselec‐tioncasesaresolved.Thisentirecognitiveprocessofadomainexpert’sthinkinghasgivenbirthtoanewbranchofartificial intelligence(AI) technique,knownascase‐basedreasoning(CBR)approach.ThisCBRapproachisappliedhereforNTMprocessselection.Inthispaper,inordertochoosethemostsuitableNTMprocessforaspecificmachiningapplication,anexhaustivecase‐basecontainingthemachiningcharacteristicsofvariousavailableNTMprocessesandtheirper‐tinentprocessparametersisfirstcreated.Thesemachiningcharacteristicsandprocessparame‐terdataarelaterusedtoselectthefeasibleNTMprocessesaccordingtotheendrequirements.Theselectionprocedureisbasedonretrievalofthebestmatchedcasefromthecase‐baseusingthenearestneighbourhoodtechnique,whilecalculatingthesimilarityscorebetweentwocases.Thebestmatchedcase,whichisretrievedfromthecase‐baseaccordingtothevaluesofdifferentprocesscharacteristicsassetbytheprocessengineer/enduser,hasthesimilarityscoregreaterthantheothercases.ToautomateandsimplifytheapplicationofCBRapproachinNTMprocessselection, a softwareprototypehaving a graphical user interface (GUI) is designed anddevel‐opedinVisualBasic6.0.Thedevelopedsystemsimultaneouslyconsidersboththeuserrequire‐ments(productcharacteristics)andtechnicalrequirements(processcharacteristics)foragivenNTMprocessselectionproblem.
2. Literature review
Usingtwomulti‐attributedecisionmaking(MADM)tools, i.e.analytichierarchyprocess(AHP)and technique for order preference by similarity to ideal solution (TOPSIS), Yurdakul andCçogun [3] attempted to simplify theNTMprocess selectionprocedure for themanufacturingpersonnel.AlistoffeasibleNTMprocessessatisfyingtheusers’requirementswasfirstgenerat‐edandthoseprocesseswerethenrankedbasedontheirsuitabilitytomeetthedesiredmachin‐ing operation. An expert systemwas developed by Chakraborty andDey [4] for selecting thebestNTMprocessunder constrainedmaterial andmachining conditions. Itwould rely on thepriorityvaluesofdifferentcriteriaandsub‐criteriaforaspecificNTMprocessselectionproblem,andtheNTMprocesswiththehighestacceptabilityindexwasfinallyidentified.ChakrabortyandDey[5]developedaquality functiondeployment(QFD)‐basedexpertsystemforNTMprocessselection.Anoverallscore foreachof theNTMprocesseswasestimatedusingtheweightsex‐tractedfromthehouseofqualitymatrixforvariousprocesscharacteristics.TheoverallscoresofsomeoftheNTMprocessessimultaneouslysatisfyingcertaincriticalcriteriarequirementswereagaincomparedandtheNTMprocesshavingthemaximumscorewasfinallyselectedastheop‐timalchoice.Aweb‐basedknowledgebasesystemwasproposedbyEdisonChandraseelanetal.[6]foridentifyingthemostsuitableNTMprocesstomeetsomeinputparametricrequirements,
A case‐based reasoning approach for non‐traditional machining processes selection
Advances in Production Engineering & Management 11(4) 2016 313
liketypeoftheworkmaterial,shapeapplication,processeconomy,andotherprocesscapabili‐ties, likesurfacefinish,cornerradii,widthofcut,toleranceetc.SadhuandChakraborty[7]ap‐pliedaninputminimizedCharnes,CooperandRhodes(CCR)modelofdataenvelopmentanaly‐sisforNTMprocessselection.Employingweighted‐overallefficiencyrankingmethodofMADMtheory,theefficientNTMprocesseswereultimatelyrankedindescendingorderoftheirpriori‐ties.Temuçinetal.[8]designedafuzzydecisionsupportmodelforNTMprocessselectionwhileassessingthepotentialsofsomedistinctNTMprocesses.ChatterjeeandChakraborty[9]provedtheapplicationpotentialityofevaluationofmixeddata(EVAMIX)methodforsolvingNTMpro‐cess selection problems using three demonstrative examples. Roy et al. [10] integrated fuzzyAHPandQFDtechniquesforselectionofNTMprocessesbasedonsomepredefinedcustomers’perspectives. Temuçin et al. [11] solved theNTM process selection problem under fuzzy andcrisp environment, and proposed a decision supportmodel to guide the process engineers toexplorethepotentialsofsomedistinctNTMprocesses.Theapplicabilityoftheproposedmodelwasalsovalidated.KhandekarandChakraborty[12]applied fuzzyaxiomaticdesignprinciplesforselectionofNTMprocesses.Madićetal.[13]demonstratedtheapplicability,suitabilityandcomputational procedure of operational competitiveness ratings analysis (OCRA) method forsolvingNTMprocessselectionproblems.
Nowadays, CBR as a part of cognitive science, has been emerged out as an interesting re‐searchtopic.AmenandVomacka[14]employedCBRapproachasatoolforselectionofmaterialandheattreatmentprocessfromanexhaustivedatabasetosimplifythetaskofadesigner.Khe‐manietal.[15]appliedCBRapproachinfusedcastrefractorymanufacturingindustry.FangandWong[16]appliedahybridCBRapproachinagent‐basednegotiationforeffectivesupplychainmanagement.ArmaghanandRenaud[17]adoptedCBRapproachtoprovethecomplementarynatureofmulti‐criteriadecisionsandCBRapproach.Althoughthepastresearchersappliednu‐merousMADMmethodsanddevelopeddifferentdistinctdecisionaidsforselectionofNTMpro‐cessesforvaryingmachiningapplications,buttilldate,noattempthasbeenputforwardonse‐lectionofNTMprocessesusingCBRapproach.Thispaperthusproposesdevelopmentofadeci‐sionmakingmodelbasedonCBRapproachforselectingthebestsuitedNTMprocessforagivenmachiningapplication.It isobservedthatCBRis thecorrectandsimplestapproachinthisdo‐mainwhereavailabilityofexpertsissometimesconstrained.InCBRapproach,asetoffeasibleNTMprocessesisfirstretrievedfromthecase‐basesatisfyingtheworkmaterialandshapefea‐ture combination. Based on the user and technical requirements, it then identifies the bestmatchedNTMprocessfromthestoredsimilarcases.ThepastcasesarejustreusedhereforNTMprocessselectionforprovidingtheoptimalsolution.
3. CBR approach
Intelligence,beingapartof cognitivescience, canbedefinedas theprocess involving rationalandabstractthinking.Itisoftengoalorientedandpurposeful.Itconsistsofknowledgeandfeats,bothconsciousandunconscious,whichareacquiredthroughcontinuousstudyandexperience.TheAIisactuallytheintelligenceinmachines.Intelligentsystemisthebasementofknowledgeengineering.Itinvolvesseveraltasks,likeknowledgeacquisition,creationofaknowledgebase,knowledge representation and use of the acquired knowledge. The represented knowledge isbasicallyusedforreasoningorinference.InAI,knowledgeisrepresentedusingsymbolsalongwithheuristicsorrulesofthumb.Whileusingtheseheuristics,oneshouldnothavetorethinkwhenasimilarproblemisencountered.Theexpertsystemcanbedefinedasanintelligentcom‐puterprogramthatusesknowledgeandinferenceprocedurestosolveproblemsthatarediffi‐cultenoughrequiringsignificanthumanexpertisefortheirsolution.Basicallyinexpertsystem,knowledgeisrepresentedusing‘if‐then’rules.
TheCBRapproachisapartofAItechniquethatutilizesinformationstoredintheknowledgebase, when similar past problems are encountered again. It provides solution to the presentproblemthatisalmostsimilartothepast.InCBRapproach,aproblemisrepresentedasaninputinthepresentsituation.Itjustretrievesthemostsimilarcasetothenewonefromitscase‐base while calculating the similarity score over the defined parameters.Itfirstsearchesthecasehistory
Boral, Chakraborty
314 Advances in Production Engineering & Management 11(4) 2016
andchooses thatcasehavingtheclosestsimilarity to thecurrentproblem. InCBRsystem, thecase‐base iswell structured and documented. The case representationmay be flat, where allcases are represented at the same level, or it can be hierarchical, expressing relationship be‐tweencasesandsub‐cases.
TherearefourmajorstepsthatconstituteaCBRsystem,i.e.retrieve,reuse,reviseandretain.Thus,itisalsocalledas4‐RcycleorCBRcycle,asshowninFig.1.Whenaproblemoccursinthecurrentsituation,similarpastsituationsareretrievedfromthecase‐base.Reusingthepastcas‐es,apredictablesolutiontothecurrentproblemisthusprovided.Ifthereisaneedofanyrevi‐sion, theretrieveddataarerevisedandretainedasanewcase in thecase‐base for futureuse[18‐21].
Fig.1ACBRcycleor4‐Rcycle
Retrieving themostsimilarcasealongwith thesolution isbasedonsome logicalexpressions.The similaritybetween two cases is usuallymeasuredwith respect to eachparameter. It alsodependsonthetypeofparameter(beneficialornon‐beneficial)beingused.Thefollowingsarethemostcommonmethodsforcalculatingsimilaritybetweentwocases:
a Numeric:
Sim(a,b) |a–b|/Range (1)
whereRangeisthedifferencebetweentheupperandlowerboundariesofaset.
b Symbolic:
Sim(a,b)=1ifa=b
=0ifa≠b(2)
c Multi‐valued:
Sim ,∩∪
(3)
whereCardisthecardinality(size)ofaset.
d Taxonomy:
Sim ,common node ,min ,
(4)
wherehistheheight(numberoflevels)ofthespecifiedtaxonomytree.
TheproceduralstepsofaCBRapproacharepresentedasbelow:
a) Asolutionisfirstdefinedusingseveralparameters.Oneoftheparametersshouldbecho‐sencarefullysothatitwouldremainuniquethroughoutthedocumentationprocedure,e.g.casenumber.
b) Ahugesetofknownsolutions isput intothecase‐baseofCBRsystem.Anexistingdata‐basecanalsobeusedforthispurpose.
c) TheCBRsystemgenerallyreadsthedatabaseandorganizesacopyofitsown.
A case‐based reasoning approach for non‐traditional machining processes selection
Advances in Production Engineering & Management 11(4) 2016 315
d)Theusergenerallyformulatesaqueryaccordingtotheendrequirements.Alltheavailablevariablesarefirstdisplayed.Theuserhastheoptiontochooseallorfewvariablesbasedontheproblemstatement.Thequeryincludesthosevariablesassetbytheuser.Theuseralsohastheoptiontoallocatedifferentpriorityweightstotheconsideredvariables.
e) Asaresultof theuser‐definedquery,CBRsystemmaydisplayanumberofcasesor thebestmatchedcase.Itmayalsobepossiblethatnoneofthecaseswouldmatchthequeryexactly.
FavouringCBR techniqueas themostefficient tool forNTMprocess selection is a challengingtask, as several other approaches have already been available for the same purpose. It is ob‐servedfromtheavailableliteraturethatnoneoftheMADMmethods,likeAHP,EVAMIX,TOPSISetc. canprovidecompletesolutionwhen thedomain is ill‐structuredandmurky.TheworkingprincipleofCBRisbasedonsomeavailablespecificexperiencesinsteadofabstractedrules.Itisconsideredasausefultooliftheutilizationofpriorexperienceismorevitalthantoproduceathoroughlyoptimized solution according to the specifications.TheCBRapproachhasnoopti‐mizingpotentiality,butitcanbeusedforsearching,notforcalculations.Itsefficiencyisdeter‐minedbyfastretrievalofthemostsimilarcasesfromthecase‐base.TheprincipleofCBRalsostatesthatitcanfindthesimilaritiesbetweencasesbutnotreasons.So,itisunabletojudgehowimportanttheencountereddeparturesarethatcanbedeterminedonlybyanexperienceduser.
AcomparisonbetweentheexistingsearchtechniquesandtheadoptedCBRapproachiselu‐cidatedinTable1.
Table1ComparisonbetweendifferentsearchtechniquesandCBRapproach
Method FlexibilityOperationalapproach
Computationaltime
Programmingcomplexity
Decisionmaker’sinvolvement
Typeofdata
Geneticalgorithm
Medium(lackoflearn‐ingability)
Populationbasedprobabilisticsearchandoptimizationtechnique
High(basedonthedesiredaccuracyandterminationcriterion)
High HighNumerical
Artificialneuralnetwork
HighMimicstheworkingprincipleofbiologicalneurons
High Medium Medium Numerical
Simulatedannealing
Medium
Coolingprocessofmoltenmetalismodeledartificiallytoconstructanoptimizationalgorithm
Medium(basedonthedesiredaccura‐cyandterminationcriterion)
High HighNumerical
Expertsystem
Medium
Exactmatchingofinputandstoreddataproducingseveral‘if‐then’rulesforinference
Medium Medium HighBothnumericalandtextual
CBR High
Notionofsimilaritybetweenpresentandpriorstoredcases
Low Low MediumBothnumericalandtextual
4. CBR‐based approach for NTM processes selection
Although CBR approach has already been successfully applied in various fields ofmechanicalengineering, suchasmaterial selection,design selection,parts selection forautomobile indus‐triesetc.,noattempthasstillbeenmadeforitsapplicationinthedomainofNTMprocessselec‐tion.TheCBRapproachhasthepotential toprovidecomplete informationaboutacasewhereminimuminformationisavailabletotheuser.Ityieldsthebestresultswhentheuserprovidesdetailedqueryinformation.
Boral, Chakraborty
316 Advances in Production Engineering & Management 11(4) 2016
While selecting themost suitable NTM process for a particularmachining application, theprocess engineer has to consider severalmachining characteristics of the availableNTMpro‐cesses. InthedevelopedCBRapproach‐baseddecisionmakingmodel,nineNTMprocesses, i.e.abrasivejetmachining(AJM),abrasivewaterjetmachining(AWJM),electricdischargemachin‐ing(EDM), laserbeammachining(LBM),ultrasonicmachining(USM),electrochemicalmachin‐ing(ECM),electrochemicaldischargemachining(ECDM),plasmaarcmachining(PAM)andwireelectricdischargemachining(WEDM)aretakenintoconsideration.Astheprocesscharacteris‐tics,typeoftheworkpiecematerial,shapefeaturetobegenerated,materialremovalrate(MRR)(inmg/min),surfaceroughness(SR)(inµm),surfacedamage(SD)(inµm),tolerance(Tol)(inmm),overcut(OC)(inmm),cornerradii(CR)(inmm),taper(TP)(inmm/mm),cost(C)(inrela‐tive(R)priorityscale),power(P)(inkW)andsafety(S)(inRscale)areconsidered.Forcost,theRscaleissetas1‐lowest,2‐verylow,3‐low,4‐medium,5‐high,6‐veryhighand7‐highest.Ontheotherhand, forsafety, theRscale issetas1‐highlysafe,2‐safeand3‐attentionre‐quired. Asworkmaterials, a) aluminium, b) aluminium alloys, c) ceramics, d) composites, e)glass,f)steel,g)superalloysandh)titaniumareconsideredinthismodel.Theabove‐mentionedNTMprocessescangeneratea)hole(precision)(0.03mm≤D<0.13mm),b)hole(standard)(L/D≤20), c) hole (standard) (L/D>20), d) through cut (shallow) (t/w≤2), e) through cut(deep)(t/w>2), f) throughcavity(standard)(t/w>10),g) throughcavity(precision)(t/w≤10),h)pocket(shallow)(t≤1mm),i)pocket(deep)(t>1mm)andj)surfaceofrevolutionfea‐tureontheworkmaterial(whereListhelengthofthehole,Disthediameterofthehole,tisthethicknessandw is thewidthof themachined feature).Therelevantmachiningcharacteristicsdata for different NTM processes are accumulated from experimentations, machining datahandbooks and other reliable resources to create the corresponding case‐base. The collecteddataarethenorganizedinastructuredmannerinMSAccess.Thestep‐wiseoperationalproce‐duresof thedevelopedCBRsystem for selecting thebest suitedNTMprocess for aparticularmachiningapplicationaredepictedasfollows:
Step1:WhenthedevelopedCBRsystemstartstorun,thefirstscreen,asshowninFig.2,appearstotheenduserwherethetypeoftheworkmaterialtobemachinedandtypeoftheshapefea‐turetobegeneratedcanbechosenfromthegivenoptionsastheprimaryinputstothesystem.
Step2: After clicking ‘OK’ button, a list of feasibleNTMprocess(es) capable of generating thedesiredshapeonthespecifiedworkmaterialisdisplayed.Forthis,Eqn.(2)isutilizedforfilter‐ingandretrievingthedata.
Step3:Whentheuserpresses ‘Next’button,anotherscreen,asshowninFig.3, isdisplayedtoidentifythemostsuitableNTMprocessfromthelistoffeasibleprocesseswhilesatisfyingthesetmachiningrequirements.
Step4: In thisscreen, theenduserhas tochoose thedesiredprocesscharacteristicsbasedonwhichthefinalNTMprocessselectionismade.
Step5:When‘Enterrange’functionalkeyisclicked,therequirednumberofemptycellsareau‐tomaticallygeneratedwheretheinputrangesfortheselectedNTMprocesscharacteristicscanbeprovided.
Step6:Afterinputtingthedesiredrangesofvalues,pressingof‘BestNTMprocess’buttonidenti‐fiesthemostsuitableNTMprocessforthespecifiedmachiningapplicationwhilesatisfyingthesetcriteriavalues.ForretrievingthebestNTMprocessinthisstep,Eqn.(1)isemployed.
Step7:TheactualretrievedvaluesofallthetechnicalcharacteristicsforthebestmatchedNTMprocessarealsodisplayed.
Step8:When‘BestNTMprocess’buttonisclicked,thetechnicaldetails(tentativesettingsoftheassociatedprocessparameters)ofthebestmatchedNTMprocessarealsoavailable,asshowninFig.4.
A case‐based reasoning approach for non‐traditional machining processes selection
Advances in Production Engineering & Management 11(4) 2016 317
Althoughinstep5,thereisanoptionforenteringrangesofprocesscharacteristicvalues,butifthe developed CBR system does not find any datawithin those ranges from the case‐base, itwouldretrievethepossibledatanearesttothequeryset.ForaparticularNTMprocessselectionproblem,MRRisthesolebeneficialattributewhereitsvalueisalwaysrequiredtobemaximized.Ontheotherhand,SR,SD,Tol,TP,OC,CR,C,PandSarenon‐beneficialattributesrequiringtheirminimumvalues.Thebestmatchedcaseshouldhavethehighestsimilarityscore,whichiscalcu‐lated with respect to each of the process characteristics. After summing up these similarityscoresforthesetprocesscharacteristicsforeachcase,theNTMprocesshavingthehighestsimi‐larityscoreischosenasthemostsuitableoption.
5. Illustrative examples
5.1 Example 1: Standard hole on composite material
Inthisexample,standardholesaretobegeneratedonacompositematerial.Afterprovidingtheinputsofcompositeastheworkmaterialandhole(standard)astheshapefeatureoptionsintheprimary selectionwindowofFig. 2, a setof feasibleNTMprocesses consistingofAJM,AWJM,ECDM,ECM,EDM,LBMandUSMisdisplayed,when‘OK’buttonisclicked.Alltheprocessescangeneratestandardholesoncompositematerials.InthenextwindowofFig.3,MRR,SR,Tol,OC,CRandCareoptedasthemostimportantprocesscharacteristicsbasedonwhichthefinalNTMprocessselectionistobemade.Inthisexample,thedesiredinputrangesforthoseprocesschar‐acteristicsaresetasMRR100‐1000mg/min,SR2‐12µm,Tol0‐0.5mm,OC0‐0.05mm,CR0‐0.5mmandC1‐4(inRscale).Now,when‘BestNTMprocess’functionalbuttonisclicked,LBMpro‐cessisidentifiedasthebestmatchedcase,capableofmeetingthesetprocesscharacteristicval‐ues.Itisinterestingtoobservethatapartfromthesetprocesscharacteristics,valuesoftheoth‐erprocesscharacteristicsarealsoavailableforthebestmatchedNTMprocess.Inthisexample,theselectedLBMprocesscanachievevaluesofMRRas286.08mg/min,SRas2.63µm,SDas102µm,Tolas0.02mm,OCas0.001mm,CRas0.5mm,TPas0.05mm/mm,Cas1(inRscale),Pas0.23kWandSas3(inRscale).
InFig.4,theprocessengineercanalsohaveanideaaboutthesettingsofdifferentmachiningparametersofLBMprocess.Thesearethetentativeprocessparametricsettingsandforachiev‐ingthemaximummachiningperformance,fine‐tuningofthesesettingsisoftennecessary.ArealtimephotographofLBMprocessisalsoavailableinFig.4.
Fig.2PrimaryselectionwindowforExample1
Boral, Chakraborty
318 Advances in Production Engineering & Management 11(4) 2016
Fig.3BestNTMprocessforExample1
Fig.4DetailsofLBMprocess
5.2 Example 2: Standard through cavity on ceramics
Here,theprocessengineerwantstogenerateastandardthroughcavityonaceramicworkma‐terial. In theprimary selectionwindow, as shown inFig. 5, thedevelopedCBRapproach firstextractsfiveNTMprocesses,i.e.AJM,AWJM,EDM,USMandWEDMasthefeasibleoptionssatis‐fyingthesaidworkmaterialandshapefeaturecombinationrequirement.
InFig.6,MRR,SR,Tol,OC,CR,CandSarechosenasthemostimportantprocesscharacteris‐ticsbasedonwhichthefinalNTMprocessneedstobeselected.Basedontherangesofvaluesfortheseprocesscharacteristics,USMprocessisidentifiedasthebestmatchedcaseforthismachin‐ing application. For USM process, the attainable process characteristics are MRR as 131.96mg/min,SRas0.66µm,SDas25µm,Tolas0.014mm,OCas0.15mm,CRas0.08mm,TPas0.005mm/mm,Cas5(inRscale),Pas0.4kWandSas1(inRscale).
A case‐based reasoning approach for non‐traditional machining processes selection
Advances in Production Engineering & Management 11(4) 2016 319
Fig.5PrimaryselectionwindowforExample2
InFig.7,thetentativeparametricsettingsandthetechnicalspecificationsofUSMprocessalongwithitsactualphotographaredisplayedtoguidetheprocessengineertoachievethebestma‐chiningperformance.
Fig.6BestNTMprocessforExample2
Boral, Chakraborty
320 Advances in Production Engineering & Management 11(4) 2016
Fig.7DetailsofUSMprocess
5.3 Example 3: Shallow through cutting on steel
Inthisexample,ashallowthroughcuttingoperationneedstobeperformedonastandardsteelplate. For thisworkmaterial and shape feature combination, the CBR system first recognizesAJM, AWJM, ECM, EDM, LBM and PAM as the six feasible NTM processes, as shown in Fig. 8.Then,inFig.9,sevenprocesscharacteristics,i.e.MRR,SR,SD,Tol,OC,CRandCareidentifiedbytheprocessengineer forthefinalselectionof themostsuitedNTMprocess fortheconsideredmachiningapplication.Inthiswindow,therangesofvaluesofthesetprocesscharacteristicsarealsoprovided.
Fig.8PrimaryselectionwindowforExample3
A case‐based reasoning approach for non‐traditional machining processes selection
Advances in Production Engineering & Management 11(4) 2016 321
Fig.9BestNTMprocessforExample3
Fig.10DetailsofAJMprocess
ThedevelopedCBRsystemidentifiesAJMasthemostappropriateNTMprocessforgeneratingashallowthroughcutonsteelmaterial.InFig.9,valuesofvariousprocesscharacteristicsofAJMprocessareprovidedasMRR‐99.76mg/min,SR‐1.9µm,SD‐2.5µm,Tol‐0.03mm,OC‐0.001mm,CR‐0.1mm,TP‐0.005mm/mm,C‐4(inRscale),P‐0.22kWandS‐2(inRscale).InFig.10,thisCBRsystemalsoguidestheprocessengineerinsettingthemostdesiredvaluesofvari‐ousAJMprocessparametersforachievingtheoptimalmachiningperformance.But,dependingontheendrequirementsandavailabilityofthemachiningsetup,thoseAJMprocessparametersneedtobeaccuratelyadjusted.
In CBR approach, all the cases, along with the relevant parameters, are well-structured in the case-base. They are collected from real time experiments, machining handbooks and expert’s opinions. Hence, when a case is retrieved by CBR system, it is likely to be closely matched with the expert’s opinion. Moreover, CBR is a technique that can provide the closest solution to the input problem. In all these three examples, the final results provided by CBR system are well validated by the experts who have a vast capability of understanding and years of experience.
Boral, Chakraborty
In CBR approach, as the selection procedure is entirely based on similarity score calculation over several process characteristics for an efficient case retrieval, there is almost negligible probability that two cases have exactly the same similarity score. Moreover, in the developed CBR approach-based software prototype, the data type for the similarity score is considered as 'double'. Thus, two NTM processes may be apparently same from the logical as well as application point of view, but they are always slightly different based on the calculated similarity scores over several process characteristics by CBR system.
6. Conclusion In this paper, based on CBR approach, a decision making model is developed for selecting the most appropriate NTM process for a given work material and shape feature combination. The functional mechanism of CBR approach is based on retrieving and reusing the past similar cases from the case-base. In the case-base, numerous cases are stored from the real time experimental data which are later utilized to extract the case nearest to the given query set. It is observed that the CBR system can provide a reasonable solution to a given machining problem where there is a lack of expert knowledge. The developed model can be a pathway towards new research in the direction of NTM process selection. Its potentiality and solution accuracy are validated using three real time demonstrative examples. It can also guide a process engineer in setting various NTM process parameters for a specific machining operation, although fine-tuning of those par-ametric settings may sometimes be required. Its accuracy can further be increased while devel-oping a hybrid CBR system, incorporating more cases in the case-base or providing more options with respect to work material and shape feature choices. A validation of the results derived us-ing the CBR approach against the existing search mechanisms, like genetic algorithm, simulated annealing, artificial neural network etc. may be the future scope of this research paper.
References [1] El-Hofy, H. (2005). Advanced machining processes: Nontraditional and hybrid machining processes, The
McGraw-Hill Companies, New York, USA. [2] Pandey, P.C., Shan, H.S. (1981). Modern machining processes, Tata McGraw-Hill Publishing Com. Ltd., New Delhi. [3] Yurdakul, M., Cçogun, C. (2003). Development of a multi-attribute selection procedure for non- traditional ma-
chining processes, Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufac-ture, Vol. 217, No. 7, 993-1009.
[4] Chakraborty, S., Dey, S. (2006). Design of an analytic-hierarchy-process-based expert system for non-traditional machining process selection, The International Journal of Advanced Manufacturing Technology, Vol. 31, No. 5, 490-500, doi: 10.1007/s00170-005-0216-5.
[5] Chakraborty, S., Dey, S. (2007). QFD-based expert system for non-traditional machining processes selection, Expert Systems with Applications, Vol. 32, No. 4, 1208-1217, doi: 10.1016/j.eswa.2006.02.010.
[6] Edison Chandrasselan, R., Jehadeesan, R., Raajenthiren, M. (2008), Web-based knowledge base system for selec-tion of non-traditional machining processes, Malaysian Journal of Computer Science, Vol. 21, No. 1, 45-56.
[7] Sadhu, A., Chakraborty, S. (2011). Non-traditional machining processes selection using data envelopment analy-sis (DEA), Expert Systems with Applications, Vol. 38, No. 7, 8770-8781, doi: 10.1016/j.eswa.2011.01.088.
[8] Temuçin, T., Tozan, H., Vayvay, Ö., Harničárová, M., Valíček, J. (2014). A fuzzy based decision model for nontradi-tional machining process selection, The International Journal of Advanced Manufacturing Technology, Vol. 70, No. 9, 2275-2282, doi: 10.1007/s00170-013-5474-z.
[9] Chatterjee, P., Chakraborty, S. (2013). Nontraditional machining processes selection using evaluation of mixed data method, The International Journal of Advanced Manufacturing Technology, Vol. 68, No. 5, 1613-1626, doi: 10.1007/s00170-013-4958-1.
[10] Roy, M.S., Ray, A., Pradhan, B.B. (2014). Non-traditional machining process selection using integrated fuzzy AHP and QFD techniques: A customer perspective, Production & Manufacturing Research, Vol. 2, No. 1, 530-549.
[11] Temuçin, T., Tozan, H., Valíček, J., Harničárová, M. (2013). A fuzzy based decision support model for non-traditional machining process selection, Tehnički vjesnik – Technical Gazette, Vol. 20, No. 5, 787-793.
[12] Khandekar, A.V., Chakraborty, S. (2016). Application of fuzzy axiomatic design principles for selection of non-traditional machining processes, The International Journal of Advanced Manufacturing Technology, Vol. 83, No. 1, 529-543. doi: 10.1007/s00170-015-7608-y.
[13] Madić, M., Petković, D., Radovanović, M. (2015). Selection of non-conventional machining processes using the OCRA method, Serbian Journal of Management, Vol. 10, No. 1, 61-73, doi: 10.5937/sjm10-6802.
322 Advances in Production Engineering & Management 11(4) 2016
A case-based reasoning approach for non-traditional machining processes selection
[14] Amen, R., Vomacka, P. (2001). Case-based reasoning as a tool for materials selection, Materials & Design, Vol. 22, No. 5, 353-358, doi: 10.1016/S0261-3069(00)00105-9.
[15] Khemani, D., Selvamani, R.B., Dhar, A.R., Michael, S.M. (2002). InfoFrax : CBR in fused cast refractory manufac-ture, Advances in Case-Based Reasoning, 2416, 560-574, doi: 10.1007/3-540-46119-1_41.
[16] Fang, F., Wong, T.N. (2010). Applying hybrid case-based reasoning in agent-based negotiations for supply chain management, Expert Systems with Applications, Vol. 37, No. 12, 8322-8332, doi: 10.1016/j.eswa.2010.05.052.
[17] Armaghan, N., Renaud, J. (2012). An application of multi-criteria decision aids models for case-based reasoning, Information Sciences, Vol. 210, 55-66, doi: 10.1016/j.ins.2012.04.033.
[18] Slade, S. (1991). Case-based reasoning: A research paradigm, AI Magazine, Vol. 12, No. 1, 42-55. [19] Kolodner, J.L. (1992). An introduction to case-based reasoning, Artificial Intelligence Review, Vol. 6, No. 1, 3-34,
doi: 10.1007/BF00155578. [20] Aamodt, A., Plaza, E. (1994). Case-based reasoning: Foundational issues, methodological variations, and system
approaches, AI Communications, Vol. 7, No. 1, 39-59. [21] Watson, I. (1999). Case-based reasoning is a methodology not a technology, Knowledge-Based Systems, Vol. 12,
No. 5-6, 303-308, doi: 10.1016/S0950-7051(99)00020-9.
Advances in Production Engineering & Management 11(4) 2016 323