连铸坯生产过程控制设计外文

ORIGINALARTICLE

Automaticsystemforintelligentsupportofcontinuouscastbilletproductioncontrolprocesses

O.S.Logunova&I.I.Matsko&I.A.Posohov&S.I.Luk’ynov

Received:24December2013/Accepted:11June2014/Publishedonline:1July2014#Springer-VerlagLondon2014

AbstractThedevelopmentofanautomaticsystemfortheintelligentsupportofcontinuouscastbilletproductioncontrolprocessesisdescribed.Themotivationforthedevelopmentofthesystemisthatitshouldimprovetheefficiencyofproductionfacilitiesandminimizethepos-sibilityofproducinginferiorandunacceptablequalityproducts.Atheoreticalanalysisoftheinformationrelatingtothequalitycontroloftheprocessesandthefinishedproductsispresented,enablingtheidentificationofthesourcesofinformation,methodsofinformationacquisi-tion,andtechniquesforprocessingittoensureimprovedproductquality.Thedevelopmentofmathematicalsup-portisdescribedforaprogramanalyzerthatautomaticallyandreliablyidentifiesthedefectsandqualityofthecon-tinuouslycastbillets.Theapplicationofgraphicinforma-tionacquisitionandprocessingtechniquesconcerningthequalityofthemetalproductsisalsopresented.Thede-velopmentofmathematicalandsoftwaresupportisde-scribedforthesetpointadjustmentmoduleoperatingintheautomaticsystemfortheintelligentsupportofthemultistagecontinuouscastbilletproductioncontrolfacil-ity.Thismodulemakesuseofadaptivefuzzytreeswithdynamicstructurestoprovidescientificallygroundedanalysisoffactorscausingbilletdefects.Theintroductionofthedevelopedsystems,includingpracticalissues,intotheoperationofaproductionfacilityisexplained.Thestudyidentifiesthegenerallackofautomaticsystemsthatencompassandcontrolthewholeproductionchainonthebasisofproductquality.Typicalsavingsresultingfromqualityimprovementsinacontinuouscastbilletproduc-tionfacilitycanapproachamillionrubles.

O.S.Logunova:I.I.Matsko(*):I.A.Posohov:S.I.Luk’ynovNosovMagnitogoskStateTechnicalUniversity,Av.Lenina38,Magnitogorsk,ChelyabinskRegion455000,Russiae-mail:matskoigor@gmail.com

KeywordsCastbilletproduction.Qualitycontrol.Fuzzysetdecision-making.Automaticproductionsystems.Billetdefectidentification.Imageanalysis

1Thescopeoftheproblem

InRussia,themodernindustrypresentsnewrequirementsinmultistagemanufacturingcontrolsystemsdeterminedbytheintroductionofnewprioritiesoutlinedbythecountry’sstatepolicy.Onesuchpriorityisthedevelopmentofinformationandtelecommunicationtechnologiesrepresentinganessentialpartofautomaticcontrolsystemsforlargeindustrialenter-prises.Theapplicationofnewmodulesinautomaticcontrolsystemsformultistagemanufacturingprocessesimprovestheefficiencyofproductionfacilitiesandminimizesthepossibil-ityofinferiorqualityproducts.

Fromtheviewpointofcontroltheory,multistagetechnol-ogyofcontinuouscastbilletproductionisacomplexsubject.Thistypeoftechnologyrequiresacontrolsystemwhichprovidesreal-timemonitoringofproductqualityaswellasintelligentsupportofdecision-makingintheproductioncon-trolprocesses.

Developmentandimplementationofnewmodules,supplementingthecontrolsystemsalreadyinoperation,makeitnecessarytousegraphicinformationobtainedintheprocessofqualityassessmentoffinishedandsemi-finishedproducts.Specialistsinautomaticprocesscontrolsystemshavegainedconsiderableexperienceinthefieldoftheoryandthepracticalapplicationofgraphicinformationanddecision-making.Theissuesofimageacquisition,processing,andsegmentationhavebeendiscussedinpapersbybothforeignandRussianscientists.ThepapersofGonzalezandWoods[1],ShapiroandStockman[2],andothersdeterminedthedevelopmentofthemathematicaltheoryinthefieldofgraphicinformationimprovementandsegmentation.Thefieldof

1408decision-makingonthebasisoftreestructureswasdevelopedbyQuinlan[3],Janikow[4],Hastieetal.[5],andBerestnevaandMuratova[6].Inthefieldofpracticalapplicationoffuzzysetsandfuzzylogictheory,thepapersofJantzen[7],Timothy[8],Espositoetal.[9],andDulicheva[10]arerelevant.

However,inspiteofnumerousstudiesandpapersinthefieldofautomaticprocesscontrolsystemsofcontinuouscastbilletproduction,therearestillsomeurgentproblemstobesolved:&Thelackofautomaticsystemstoprovidecontrolofthecompleteprocessingchainonthebasisofinformationaboutproductquality.

&

Thelackofgraphicinformationacquisitionandprocess-ingtechniquesconcerningthequalityofmetalproductsusinglow-contrastimagescontainingirregular(shaped)elements.

&

Thelackofapplicationprogrampackagesforintelligentsupportofdecision-makinginmultistagemanufacturingautomaticcontrolsystemsdevelopedusingadaptivefuzzydecisiontreeswithdynamicstructuresthatconsiderprod-uctquality.

2Objectivesoftheresearch

Takingintoaccounttheseproblems,themainobjectiveofthisresearchwastoreducetheincidenceoflow-qualityproductsbyusinganautomaticsystemofintel-ligentsupportformultistagemanufacturingcontrolde-velopedonthebasisofadaptivefuzzydecisiontreeswithdynamicstructures.

Toachievethisobjective,theauthorshadto:&

Carryouttheoreticalanalysisoninformationontheauto-maticcontrolsystemforcontinuouscastbilletproductiontodeterminethesourcesandmethodsofacquiringandprocessinggraphicinformation;

&

Developmathematicalsupportforthecontinuouscastbilletqualityanalyzerincludingaformaldescriptionoftheareaboundariesofthetemplateimage,algorithmstoimproveandsegmentit,andafuzzytreestructureofbilletdefectclassificationusingthequalityandquantityofirregular(shaped)objects;

&

Developthestructureoftheautomaticsystemforin-telligentsupportofcontinuouscastbilletproductioncontrol.Inaddition,todevelopthemathematicalsup-portandsoftwareforthedecision-makingmoduleusedinprocesssetpointadjustmentintheautomaticsystemforintelligentsupportofmultistagecontinuouscastbilletproductiononthebasisofadaptivefuzzytreeswithdynamicstructures;and

IntJAdvManufTechnol(2014)74:1407–1418

&

Developorganizationalandtechnicalsolutionsfortheoperationandmaintenanceoftheautomaticsystemin-cludingproductionengineerworkstationdesign,selectionofahardwareplatformforprogramanalyzeroperation,andestimatedeconomicefficiencyoftheoperationofthedevelopedmodules.

Theinvestigationisconcernedwiththeautomaticsys-temofqualitycontrolformultistagecontinuouscastbilletproduction.Thefocuswasmainlyoninformation,math-ematicalandsoftwaresupportincludingthemathematicaldescriptionofgraphicinformation,andadaptivefuzzydecisiontreeswithdynamicstructuresdevelopedfortheintelligentsupportofdecision-makingprocessesforon-linecorrectionofcontrolactions.

3TheoreticalanalysisofinformationontheautomaticcontrolsystemforcontinuouscastbilletproductionAsaresultofthetheoreticalanalysis,itwasfoundthatitsarchitecturewasmadeupofautomaticcontrolsystemsforthevariousprocessstagesincludingsteelmanufactureinelectricarcfurnaces,steeltreatmentintheladlefurnace,andsteelcastinginthebilletcontinuouscastingmachines.Theprocesscontrolsubsystemspecifiestheoperatingscheduleandprod-uctqualitycontrolparameters.Thesubsystemcontainsallthetechnologicalregulatoryinformationfordeterminingandforwardingprocessrequirementstomachineautomaticpro-cesscontrolsystemsforeachsub-stageoftheproductionprogram[11].Theinformationistransferredtothequalitycontrolsystem.Toestablishthemaincomponentsofalltheprocessesofthetechnologicalchainofcontinuouscastbilletproduction,theauthorscarriedoutacomplexanalysisofthecontrolsystemforeachprocessstage.Theanalysiswascar-riedoutonthebasisofexperimentalobservationsandastudyofthetechnicalinstructions,standards,andmetrologicaldata.Themainsourceofinformationformakingdecisionsonthecauseofinferiorqualitybilletsistheheatlogcompletedforeachproductionstage[12].

Theheatlogcontainstheresultsofthevisualtestofthetemplate.Thetemplateselectionschemeforsquareandrect-angularbilletsisgiveninFig.1.Thetechnologicalinstructionstatesthatthetemplatemustbeselectedafterthefifthbilletofthefirstbatch.Ifnecessary,anextratemplateselectionofotherbilletsfromanyotherbatchcanbecarriedout.Whenfulfillingsomecriticalorders,templatesaretakenfromeverysecondbilletofthebatch.

Theacquiredtemplateissenttotheplateandsheetlabora-torywhereitisprocessedandevaluatedaccordingtoOST14-4-73andOST14-1-235-91[13,14].However,theinforma-tionaboutthequalityofthebilletestimatedinthelabisquitesubjective.Analysisoftheresultsmadeitpossibletoidentify

IntJAdvManufTechnol(2014)74:1407–1418Fig.1Commonschemeoftemplateselectionforacontinuouscastbillet:aasquarebillet;barectangularbillet

thefollowingdrawbacksinthecontrolsystemforcontinuouscastbilletproduction:&Thelackofanautomaticsystemdevelopedforcontinuouscastbilletqualityestimation,whichresultedingeneratingdataoflowreliability,integrity.andrelevance.

&Thelackofcommunicationwiththequalitycontrolde-partmentsbecausetheinformationaboutbilletqualitywasnotreliableenough.

&

Thelackofmathematicalsupportfortheimprovementandsegmentationofcontinuouscastbillettemplateim-agesandtheirsulfurprints,defectlocationintheseim-ages,andclassificationofthediscovereddefects.

&Thelackofprogramanalyzersdevelopedforautomaticprocessingofgraphicinformationfordecision-makinginthetechnologicalprocess.

&

Thelackofanintegratedscientificdecision-makingsys-temstructurefortheinterdependentprocessesofsteelmelting,secondarysteelmaking(processinginaladlefurnace),andcontinuouscastingofsteel,whichresultsincontrolmismatchingateachstage.

Thedisadvantagesabovemakeitdifficulttoorganizeefficientcontrolofcontinuouscastbilletproductionusinginformationaboutthefinalproductquality.Itresultsintheincidenceofinferiorqualitybilletatabout0.03%ofthetotalbloomoutputbecauseofthereductionintheefficiencyofthemetallurgicalfacilities.

4Mathematicalsupportforacontinuouscastbilletquality

analyzer

Toanalyzethegraphicinformation,researchersoftheOJSC“MagnitogorskIronandSteelWorks”laboratoriescollected

1409

71imagesofsulfurprintsand384templatephotos.TheacquiredinformationisclassifiedinFig.2.

TheinformationisrepresentedbytheimagesgiveninFig.3.Thedefectsarerepresentedbyirregular(shaped)objectsoflowbrightness.Intotal,thedefectsshowninFig.4,namelydotnon-uniformity,axialporosity,perpendic-ularbilletedgesandblowholes,wereconsidered.

Todevelopamathematicaldescriptionofthetemplateimage,weintroducedthefollowingassumptionsthatdidnotchangeeitherthesemanticorphysicalessenceoftheobjectunderstudy:&Thecolorimagehasadotmatrixdrawing(morethanthreehundreddotsperinch).

&

Theimageisaparallelogram,andanaffinecoordinatesystemisintroducedforthisimagewithaunitsegmentofonepixel.

&

Theimageispositionedaccordingtotherulethattheupperedgeoftheimageisalignedwiththebilletsideplanelocatedsidewiseontheminorradiusofthebilletcontinuouscastingmachine.

&

Thecolorimageismatchedwiththelocationschemefortheinternaldefectsofthebillet;anobliquesystemofcoordinateswithΔhincrementsisintroducedintotheimagetodecomposethephotointoregionswithuniformbrightness.

Themathematicalmodelofacolorimageconsistsof:

1.Ananalyticaldescriptionoftheimageareaboundaries

(Fig.3c):

&

Thedotnon-uniformityregionABFEgivenby:8

>> sinðα−βÞ!cosðβÞ

⋅ctgðβÞ;>>ð1Þ

:x⋅tgðαÞ≤y≤x⋅tgðαÞþ0;84⋅bsinðβ−αÞcosðαÞ

;:&TheaxialdefectsregionMNOPgivenby:

8>>

sinðα−βÞ! cosðβÞ⋅ctgðβÞ≤x≤y−0;6⋅asinðα−βÞ!cosðβÞ

⋅ctgðβÞ;>>:x⋅tgðαÞþ0;4⋅bsinðβ−αÞcosðαÞ≤y≤x⋅tgðαÞþ0;6⋅bsinðβ−αÞcosðαÞ

;:ð2Þ

&

TheblowholesregionABCD–(ABFE∪MNOP)whereaisthebilletwidthinmm,bisthebilletheightinmm,andα–∠XOX′,β–∠XOY′.

1410

Fig.2Classificationofsourceinformationforanalysis

IntJAdvManufTechnol(2014)74:1407–1418

2.Theanalyticaldescriptionoftheimagecolormatrixis

givenby:

fðx;yÞ¼

cðx;yÞ

;cmax

ð3Þ

wheref′(x0,y0)isthefunctionoftherelativecolorafternoisesuppressionatthepoint(x0,y0),andHisanarrayofweightcoefficients.

6.Equationforimagebinarization:

&

Preliminaryimageinverting:

&00

1−fðx;yÞwhenQ−fðx;yÞ>h;

Iðx;yÞ¼:

1−Qinallothercases;

ð8Þ

3.Uniformitycriteriaoftheirregular(shaped)region:

jfðx;yÞ−fðx;yþ1Þjð4Þ

whereGistheconstraintconstantdeterminedempirically.4.Expressionsforthepropertycalculation:

&

Imagerecognitioncriterion:fðx;yÞð5Þ

&

Imagebinarization:

&

1whenIðx;yÞ>1−Q;

:Bðx;yÞ¼

0inallothercases:

ð9Þ

&Imagecontrastcriterion:

x∈½0;a󰀂;y∈½0;b󰀂

max

ffðx;yÞ−QgwhereQisthevalueoftheStrehlratio,andGкрisthecriticaldifferencebetweenthevaluesoftherelativecolorofthepointandtheStrehlratio,whichmakesitpossibletoselectauniformregion.

5.Equationfornoise-suppressingmaskontheimage:

x0þ1X

y0

X

fðx0;y0Þ¼

0

fðx;yÞ⋅Hðx;yÞ;ð7Þ

x¼x0−1y¼y0−1

Amethodofimageacquisitionwasdevelopedtoimprovethereliabilityofthedata.TheschemeforthedevelopedmethodisgiveninFig.5.Thedevelopedmathematicalde-scriptionforimagescomprisingthedatabasemadeitpossibletoturnfromunformalizeddescriptionsofimageobjectstotheirquantitativeprototypes(Figs.6and7).

Formalizedquantificationoftheestimationresultsmadeitpossibletodevelopanautomaticclassificationsystemoftheinternaldefectsincontinuouscastbillet.AclassificationalgorithmwasdevelopedonthebasisoftheC4.5algorithmandthealgorithmforimprovingthedecision-makingprocessusingafuzzydecisiontree.Tomakeaforestoffuzzydecisiontreesthattakeintoaccounttheclassificationtableresultsusingattributedqualitycharacters,theresearchgroupintroducedlinguisticvariablestodescribedefectparameters.Thesein-cludedwidth,length,relativewidth,relativelength,defectarea,maximumarea,relativearea,length/widthratio,width/lengthratio,quantity,andbelongingtoregions.

IntJAdvManufTechnol(2014)74:1407–14181411

󰀁Fig.

3Graphicproductqualityinformationfedintotheautomaticprocesscontrolsystem:asulfurprint,btemplatephoto,сtheschemeoftemplateimagedecompositionintopartswheref(x,y)isthevalueoftherelativecoloratthepoint(x,y),c(x,y)isthecolordigitalcodeatthepoint(x,y),andсmaxisthemaximumcolorcodeintheimage

Figure8showsamembershipfunctionofalinguisticvariable“Defectsize”obtainedfromexpertevaluationbylaboratorytechnicianswhohavebeenworkinginthisfieldformorethan10years.

IfthedefectsizeinFig.8isreferredtoassmall,itmeansthatallprocessesworkedsteadily;ifitisreferredtoasaver-age,itisnecessarytointroducesomeminorcorrectionsintotheprocessvariables;andifitisreferredtoaslarge,itisnecessarytotakeurgentmeasuresandinterfereintheprocessasthebilletisofinferiorqualityandcannotbefurtherproc-essedorshippedtothecustomer.

Thetreestructuredevelopedtoestimatetheaxialporos-ityofacontinuouscastbilletisgiveninFig.9.Thecompleteforestoffuzzydecisiontreescontainsfrom800to2,000leaves.BranchesinFig.9weremarkedforadefectofaveragesize,apartofwhichislocatedinregion3.If10%ofthedefectislocatedinregion3,thetemplategradeofmembershiptothetargetclass(qualitytemplate)is0.334,correspondingto2.5pointsusingtheOSTclassification.

However,obtainingaqualityevaluationonthebasisofonlythedecisiontreeisnotsufficientformakingadecisionon

Fig.4Elementofasulfurprintcontainingcharacteristicdefectcatego-riesandshapesonascaleof1in8:adotnon-uniformity,bcracksperpendiculartothewideside,clongitudinalcracks

1412IntJAdvManufTechnol(2014)74:1407–1418

Fig.5Theschemeforthedataacquisitionmethodonthequalityofcontinuouscastbilletfortheautomaticprocesscontrolsystem

thevaluesofthevariablesoftheproductionprocess,whichiswhyitwasdecidedtodevelopmathematicalandsoftwaresupportforadecision-makingmodule.Themodulewillberesponsibleformakingdecisionsonsetpointadjustments.

5ThestructureoftheautomaticsystemforintelligentsupportofcontinuouscastbilletproductioncontrolTakingintoaccountthestructuralfeaturesoftheautomaticcontrolsystemforcontinuouscastbilletproductionaswellas

IntJAdvManufTechnol(2014)74:1407–14181413

Fig.6Unformalizeddescriptionschemeforirregular-shapedobjectsonthetemplateimage

stablebatchproduction,thestructureofanautomaticsystemfortheintelligentsupportofcontinuouscastbilletproductioncontrolattheelectricsteelmakingshopofOJSC“MMK”wasdeveloped(Fig.10).

InthegloballoopsoftheEAFandtheladlefurnace(1,2),thecontrolledobjectisliquidmetalandinthebilletcontinu-ouscastingmachine(3),thecontrolledobjectisabillet.Asampleintheformofatemplateand(or)itssulfurprintistransferredtothemoduleforautomaticevaluationofcontin-uouscastbillettemplates.Havingprocessedtheimage,themodulesendstheexpertanalysisoftemplatesurfacemacrodefectstothedecision-makingblockforsetpointad-justmentofthevariousstagesofthecontinuouscastbilletproductionchain[15].

Thedecision-makingblockgeneratesinstructionsoncor-rectionstobemadeinthecontrolsystemwhicharetrans-ferred,time-lagged,totheblocksofmathematicalmodelsofgloballoopsprocesses.Aservicesimulatingtestiscarriedoutonthebasisofthenumerousoperatingprocesstechnologymodels[16–19].Theresultsoftheservicesimulatingtestsontheadjusteddataandforecastedbilletqualityaretrans-ferredtoeachdecision-makingblock.Thedecisiononsetpointadjustmentismadebyanexpertinthescientificallybasedresultsofmodelingandforecasting.Thedecisionistransferredtoeachsubsystemofthegloballoop,withtimelagsresultingfromthetimeintervalsnecessaryforperformingoperationsonfurtherstagesoftheliquidmet-al,billet,orsampleprocessing.

Takingintoaccounttheminimumandmaximumvaluesofthelagtimeaswellastheprocessingtimefortheheatineachmachine,weestablishedthatitwouldtakethesignalforthesetpointadjustmentfrom5to7heatcyclestoreachtheEAF,from3to7heatcyclestoreachtheladlefurnace,andfrom2to3heatcastingsforthesignaltoreachthebilletcontinuouscastingmachine.

Fig.7Unformalizedandformalizeddescriptionschemeforirregular-shapedobjectsonthetemplateimage

1414IntJAdvManufTechnol(2014)74:1407–1418

Fig.8Graphofmembershipfunctionforclassificationofbilletdefect

Providedthatthebilletbatchproductionconsistsofthemelting,processing,andcastingof25–50heatcyclesperbatch,theproposedtechnologicaladvancementmakesitpossibletoreducetheincidenceofinferiorqualitybillets.ThefollowingsymbolsareintroducedinFig.10:block1isthesteelmeltingcontrolloopintheelectricarcfurnaces,block2isthemeltprocessingcontrolloopintheladlefurnaces,block3isthecontinuouscastingcontrolloopinthebilletcontinuouscastingmachines,Zjiareinitialsetpointsfortheithmachine,Zijaretheadjustedinitialsetpointsfortheithmachine,Ziniarethecalculatedvaluesofthelocalcontrolloopmodesfortheithmachine;Voutisthebilletyield,I3isthesampleintheformofatemplateand(or)itssulfurprint;{OE}aretheexpertevaluationsofmacrodefectsonthetemplatesurfacebyOSTclassifications,Kiistheinstructiontoadjustsetpointsfortheithmachine,Oiistheforecastmacrodefectsevaluationfortheithmachine,{Zpi}areparametervaluesofthelocalcontrolloopsfortheithmachineobtainedasaresultofprocesssimulation,andΔZjiaretheadjustedsetpointsfortheithmachine,wherejisthesetpointnumberandiisthemachinesymbol.Inthegloballoops(1,2,3),theprocessmodelblockacceptsinstructionstoadjustthesetpoints.

K

Anadaptivefuzzytreewithadynamicstructurewasde-velopedforthedecision-makingblock(Fig.11).Thefollow-ingfactorsweredeterminedaslinguisticvariablesforthedecision-makingtreesonpossiblecausesofdefects:heat,weight,metaltemperatureaftertappingfromtheEAF,metaltemperatureaftersecondarytreatment,metaltemperatureinthetundish,metaloxidation/deoxidationstate,steelcastingspeed,andsteelchemistry.

TakingintoaccountthattheleavesofthedevelopedtreeareasdepictedinFig.11,thedecisionmakesuseofonlythetwobranchesmarkedbythedashedline.Providedthatthewaterconsumptioninthesecondarycoolingzoneis444l/minandthecastingspeedis3.5m/min,thedecision0.14belongstotheupperbranch(A)and0.86belongstothelowerbranch(Fig.12).Forthisdecision,onecancalculatethegradeofmembershipinthequalitybilletclass:

ÀÁð77:81⋅1þ5:05⋅0Þ⋅0:14þð13:2⋅1þ5:71⋅0Þ⋅0:86

¼0:357:δDj¼

ð77:81þ5:05Þ⋅0:14þð13:2þ45:71Þ⋅0:86

ð10Þ

Thisgradeofmembershipcorrespondsto2.5pointsusingOST[14,15].Thebilletisconsideredtobeofgoodqualityifthisvalueislessthan2points.Therefore,itisnecessarytocorrectthedecisionabove.AccordingtotheC4.5algorithm,thefollowingstepsshouldbetaken:1.Controlparametersadjustment:

δðАÞ¼

77:81⋅1þ5:05⋅0

¼0:94;

77:81þ5:05

13:2⋅1þ45:71⋅0

¼0:22;

13:2þ45:71

ð11Þ

δðBÞ¼

ð12Þ

Fig.9Pathofmotioninthedecision-makingprocessonthebasisofafuzzytreestructuretoevaluatetheaxialporosityofthecontinuouscastbillet

IntJAdvManufTechnol(2014)74:1407–14181415

Fig.10Structureoftheautomaticsystemforintelligentsupportofcontinuouscastbilletproductioncontrol

1416

Fig.11Partofthemotionpathinthedecision-makingprocessonthebasisofthefuzzytree

structureforcontinuouscastbilletqualityforecast

IntJAdvManufTechnol(2014)74:1407–1418

wheretheleafofbranchBhastheleastinfluenceontheobjectivefunction.

2.Selectionof“Waterconsumptioninthesecondarycooling

zone”crosspointforthebranchofthediscoveredleafbecauseithasthelowestvalueofthegradeofmembership.

3.Determinationofthenon-zerogradeofmembershipfor

allsubunitsof“Waterconsumptioninthesecondarycoolingzone”crosspoint.

4.Determinationofthemaximumsubunitmembership

gradeforbranchAsubunit.Waterconsumptioninthesecondarycoolingzoneof320l/minshouldbeselectedasthisdecisionhasagradeofmembershipinbranchAequalto1.

5.Evaluationoftheselecteddecision:

ÀÁð77:81⋅1þ5:05⋅0Þ⋅1δDj¼¼0:94:

ð77:81þ5:05Þ⋅1

ð13Þ

Thisdecisionmeetsalltherequirementsnecessaryforqualitybilletproduction.

6.Usingbinaryanalysis,itispossibletodeterminethatif

waterconsumptionis438l/min,thedecisionhasagradeofmembershipinqualitybilletclassequalto0.5corre-spondingto2points.

Thus,reducingwaterconsumptionfrom444to438l/minimprovesthebilletqualityfrom2.5to2points.Thedecisionmakesuseofonlytwodashedlinebranches.Decision0.14belongstotheupperand0.86tothelowerbranch.Forthisdecision,theproducedbillethasagradeofmembershipforaqualitybilletof0.357correspondingto2.5pointsusingOST.

Thebilletisconsideredtobeofgoodqualityifitscoreslessthan2points.Therefore,itisnecessarytocorrectthedecision.Accordingtothedevelopedalgorithm,controlparametersareadjustedinthefollowingway:Coolantconsumptionisreducedfrom444to438l/min;thus,billetqualityevaluationisreducedto2pointsmakingitpossi-bletoqualifythebilletasaqualityone.

Fig.12Membershipfunctiongraphforclassificationof:asteelcastingspeed;bwaterconsumptioninthesecondarycoolingzone

IntJAdvManufTechnol(2014)74:1407–1418Asthedevelopedtreeisadaptive,itcanberelearntincasethetechnologicalprocesschangesanditcanreacttotheexamplesthathavenotbeendefinedinthetree.Thisfeaturemakesitpossibletochangetheimportanceofaparameterafteranumberoftreeadaptationstointroduceshorterdecisionbranches,thusreducingthenumberofparameterstobeadjusted.

6Organizationalandtechnicalinstructionsforimplementationofresearchresults

Toimplementthecontrolactionadjustmentfromtheprogramanalyzermoduleanddecision-makingblockoneachmachine,itisnecessaryto:&Integratetheprogramanalyzeranddecision-makingblockmodulesintothecorporateinformationsystemoftheenterprise;

&

Installthedevelopedsystemforproductionprocess-esmonitoringintoworkstationsoftheEAFoperator,ladlefurnaceoperator,billetcontinuouscastingma-chineoperator,theHeadoftheEAFshop,theHeadofdispatchcontrol,headsoflaboratories,andplant-operatingengineersofthecentralcontrollaboratory;&Updatetheplant-operatingengineerworkstationbyinstallingasuperaccurategesturecontroller;and

&

Trainpersonneltomaintainandoperatetheinstalledmod-ulesoftheprogramanalyzeranddecision-makingblock.

Thehardwarecomplexityoftheplant-operatingengineerworkstationrequiresaPCwithamodernCPUthathasaprocessorspeedofatleast1.6GHzandrandomaccessmemoryofatleast1Gb,asuperaccurategesturecontrollercapableofsurfaceimageacquisitionwitharesolutionofatleast300dpi(dotsperinch),andtheabilitytodevelopadotcloudwithanaccuracyofupto0.01mm.Thedecision-makingblockneedsaserverwithaprocessorspeedofatleast1.6GHzandfourprocessorcoresandrandomaccessmemoryofatleast4Gb.Thedevelopedautomaticsystemfortheintelligentsupportofcontinuouscastbilletproductioncontrolmadeitpossibletoextendthestructureofthemonitoringsystem(Fig.13).Itwasproposedtointroduceadditionaldialogboxestodisplaythehistoryofinternaldefectsdevelopmentanddecisionsforcontingencyprevention.

ThefollowingsymbolswereintroducedinFig.13:1istheinformationaboutthevaluesoftechnicalparametersforeachproductionunitobtainedfromtheproductionsitelevel,2istheinformationabouttheprocessstateobtainedusingahu-manoperatorinthemonitoringsystem,3isthemeltproducedintheEAF,4isthemeltproducedintheladlefurnace,5isacontinuouscastbillet,6isasampleimageforqualityevalu-ation,7istheinformationaboutbilletquality,and8isthe

1417

Fig.13Developmentschemeforthebilletproductionprocessmonitor-ingsystem

informationaboutcurrentbilletqualityandthepossibilityofsetpointvalueschanging.

Theintroductionoftheresearchfindingswillresultinestimatedsavingsof981,000rublesperannumfromthereductionininferiorqualitybilletsandclaimsfromcustomers.

7Conclusions

Thefollowingcanbeconcludedonthebasisoftheresearchwork:

1.Theoreticalanalysismadeitpossibletoestablishthe

drawbacksofthecurrentsystemandjustifythemeanschosenforitsimprovement.Consequently,methodstolocate,acquire,andprocessinformationfordevelopingtheautomaticsystemforintelligentsupportofcontinuouscastbilletproductioncontrolmakinguseofinformationonthequalityofthefinishedproductwereproposed.2.Mathematicalsupportwasdevelopedfortheprogram

analyzerofthecontinuouscastbilletqualityfortheauto-maticprocesscontrolsystem.Thedevelopedprogrammodulemakesitpossibletoautomaticallyrecognizeandclassifybilletdefects,reducetheinfluenceofhumanfactorsonbilletqualityevaluation,andimprovethereli-abilityofinformationusedintheautomaticprocesscon-trolsystemwhenitmakesdecisionsonproductionpro-cessadjustment.

3.Mathematicalandsoftwaresupportwasdevelopedforthe

setpointadjustmentmoduleoperatingintheautomaticsystemforintelligentsupportofmultistagecontinuouscastbilletproductioncontrolonthebasisofanadaptivefuzzytreewithadynamicstructure.Whenthedevelopedprogrammodulewasinstalledintheoperatingautomaticprocesscontrolsystem,itextendedthesystemof

1418automaticmonitoringandcontrolandprovidedscientifi-callygroundedanalysisoffactorscausingbilletdefects.4.Organizationalandtechnicalinstructionsweredeveloped

fortheoperationoftheautomaticsystemofintelligentsupportofmultistagecontinuouscastbilletproductioncontrolmodule.Thisincludedthearrangementoftheindustrialengineer’sworkstation,thechoiceofthehard-wareplatformfortheoperationoftheprogramanalyzer,andthecalculationoftheestimatedsavingsasaresultoftheoperationofthedevelopedmodules.Introductionofthedevelopedsolutionsintotheoperatingautomaticpro-cesscontrolsystemwillsaveupto981,000rublesperannumbecauseofthereductionininferiorqualitybillets.TheprogrammodulesweretestedatRTCAusferr,KonsOMSKS,andinthelaboratoryforphysical,mechanical,andmetallographictestingofplateandrodsteelatOJSC“MagnitogorskIronandSteelWorks”.

AcknowledgmentsThisresearchwascarriedoutwithintheframeworkoftheFederaltargetprogram“Scientificandpedagogicalstaffofinno-vativeRussia”fortheperiodof2009–2012undercontractП2402signedon18November2009.ItwassupportedbytheChelyabinskRegionGovernment(in2011and2012)andtheRectorofFSBEIHPE“NosovMagnitogorskStateTechnicalUniversity.”

References

1.GonzalezR,WoodsR(2005)Digitalimageprocessing.Tehnosphera,Moscow

2.ShapiroL,StockmanG(2006)Computer.BI-NOM.KnowledgeLaboratory,Moscow

3.QuinlanJR(1993)C4.5:programsformachinelearning.MorganKaufmannPublishersInc,SanMateo

4.JanikowCZ(1998)Fuzzydecisiontrees:issuesandmethods.IEEETransSystManCybern1:1–14

5.HastieT,TibshiraniR,FriedmanJ(2009)Theelementsofstatisticallearning:datamining,inferenceandprediction.SpringerScience+BusinessMedia,NewYork

6.BerestnevaOG,MuratovaEA(2004)Developmentoflogicalmodelsusingdecisiontrees.ProceedingsofTomskpolytechnic

IntJAdvManufTechnol(2014)74:1407–1418

university.PublishinghouseofTomskpolytechnicuniversity,Tomsk2:154–160

7.JantzenJ(2013)Foundationsoffuzzycontrol:apracticalapproach.Wiley,Chichester

8.TimothyJR(2010)Fuzzylogicwithengineeringapplications.Wiley,Chichester

9.EspositoFA,MalebraD,SemeraroG(1997)Comparativeanalysisofmethodsforpruningdecision.IEEETransPatternAnalMachIntell5:476–496

10.DulichevaYuY(2002)Strategyofdecisiontreereduction(survey).TavricheskyBullComputSciMath1:10–17

11.

LogunovaOS,MatskoII,PosochovIA(2013)Integratedsystemstructureofintelligentmanagementsupportofmultistagemetallur-gicalprocesses.VestnikNosovMagnitogorskStateTechUniv5:50–55

12.

LogunovaO.S.,KaptsanF.V.,SuspitsinV.G.,PavlovV.V.,NurovH.H.(2006)Automaticworkstationorganizationinthesystemofcontinuouscastbilletmacrostructurequalityevalua-tioninthearc-furnacemeltingshopoftheOJSC«ММК».Steel№11:80–82.

13.OST14-4-73(1973)Steel.Methodofcontinuouscastbilletmacro-structurecontrol.MinistryofferrousmetallurgyoftheUSSR,Moscow

14.

OST14-1-235-91(1992)Steel.Methodofcontinuouscastbilletmacrostructurecontrolforrolledsectionandhollowbilletproduction.InsteadofОСТ14-4-73inthechapterdevotedtocontinuouscastsquarebillet.Ministryofmetallur-gyoftheUSSR,Moscow

15.

MatskoII,BelyavskyAB,LogunovaOS(2012)Proceedingsofuniversities.Ferrousmetallurgyorganizationoffeedbackincontrolsystemofcontinuouscastbilletproduction.Proceedingsofuniversi-ties.FerrousMetallurgy11:55–60

16.

LogunovaOS,MatskoII,SafonovDS(2012)Thermalstatesimula-tionoftheinfinitebodytakingintoaccountdynamicboundaryconditionsofthethirdkind.BullSouth-UralStateUnivMathModelProgramSer27:74–85

17.

KolokoltsevVM,SinitskyEV,VolkovSY(2011)Analyticalandtechnical/engineeringcriteriaofcastingabrasivewearresistanceevaluationasafactoroftheirquality.VestnikNosovMagnitogorskStateTechUniv2:16–18

18.

LogunovaOS,PavlovVV,MatskoII,PavlovIV(2012)Empiricalmodelofresidualelementcontentinsteelwhenthreecomponentburdenisusedintheprocessofsteelproductioninelectricarcfurnace.JMinWorldExpress1(2012):21–26

19.

YachikovIM,ZaretskayaEM(2011)Analysisofmagneticfieldbehaviorinthevicinityofelectricarcfurnaceelectrodesbymeansofmathematicalmodeling.Proceedingsofuniversities.FerrousMetall1:18–21