P3 Applying Control Principles to Superheat

1、Applying Control Principles to Superheat Mark P. Taylor and John JJ Chen -LMRCUniversity of AucklandA Classical Feedback Control LoopWhatever the cause of temp variation,steam pressure change will compensatefor it and keep the temperature steady. A Classic Engineering Control Loop Supplier Customer

2、Process ControlVariables ProcessVariable InputOutput/OutcomeSignal tocontroller(deviation) ProductCharacterResponse intoProcess ControllerACD,Alumina concSensed byR, R changeRtarget,Rise,SlopeBeam Move For ACDFeed rate, forAlumina concAutomated Control HierarchyNormal Operation atOptimal Performance

3、Normal Operation SomewhereAbnormal Cell State-automatic response Abnormal Cell State-No auto response.Abnormal PotlineState/EmergencyLowestHighestPriorityHighestValue hereOrigins of Control in SmeltersIndirect Observations of Process .(R slope)Control design was SISO, despite multivariate nature.Ope

4、rational State not considered in Control.Tuning of Controllers is never-ending, as other parameters (sludge back-feed rate, I, bath vol, superheat) changeAbnormalities dealt with when serious damage.Trial and Error used - eg.arbitary setting of controller gains, control bands, dead bands.But Margins

5、 were US$1000/TStrategy Development to dateAlumina Control is paramount: - multiple point feeders, over/under/zero feed patterns Resistance change signal .Device control, damage limitation next.Heat Balance Control secondary being worked on now but limited by above strategies and sensing.Special Cau

6、se identification and diagnostics ?Optimum Cell Voltage/ACD for Process ? The Essence of Controlis understanding why Processesvary Controllers dont often incorporate this knowledgeThermal LimitFor Walls or anodesACD LimitCell VoltageLine AmperageFeasibleOperatingRegionWall Corrosion,Anode failureCE

7、loss, Spiking, Anode CD In an environment where longer term Variation in Energy Balance (due to Operating imperatives) is occurring,Superheat Control is essential.? Controllingthe Outcome Understandingthe Variation Observing Process StateMeasureSignalResponse Measurement System Capability Structural

8、 Variation By MaterialTransform.Cause Investigationand drill down .VisibilityofDataNatural,SpecialVariationOps, PCsT, T, on each potTE Anode sets,Tot. liq. depth, inputs, Lags.Cycles inT, T.OutliersSignals to Oper.Immediate actionfor abnormalityAl2O3 , CHole prep.Gage R+RMeasurement Planincludes:Pro

9、cedureAccountab.Gage R+RResponse Plan-includes:Find Cause+ CorrectEscalate +InvestigateKey Measure An effective Monitoring andControl Mechanism will be neededTrigger for Action (outside normal process variation)Embed understanding of the measurement system here And here Control limits calculated usi

10、ng the equationThe Control limit translates into “Control Ellipse” or “Confidence Ellipse”. The ellipse will include 68.2%, or 95%, of the population assuming the observations are from a bivariate normal distribution.So what is the natural variation ? (for one cell)No Ledge.No Cathode Joint Protecti

11、onPotential for Metal LeakageBath Freezes upStableLower CECathode RidgedHigh Heat LossCrust FallingLedge ProtectionCrust StableOptimum CEThere are process risks to be assessed as well Points outside the Elliptical Region are worthyof investigation and possibly action. Type I Error“False Alarm”Callin

12、g for action when in fact none would be requiredWhen no problem, but think there is a problemType II Error“Failed Alarm”Probability of failing to detect a shift in the process when one does occurDeclared no problem when there is really a problemINPUTPROCESSESPRODUCE AL,CONSUME C,Al2O3OUTPUTPROCESSES

13、CELL MAINTENANCE PROCESSESCONTROLPROCESSESWhich processes can impact on, or will be affected by control of Superheat ?Alumina, CEnergyCellsBathCoverAirAC PowerDraughtBeam raisingCell replacementCell Hardware maintHood replacementCrane maintenanceACD/ResistanceCell integrity monitoringBath levelMetal

14、 levelMetal purity Temp./T/AlF3Alumina conc.Cell DressingException CellsHot MetalEmissionsSkimmings, sweepings.Butt Bath, tapped bathAnode ButtsSpent Cell liningDirty cruciblesElectrolysis /reactionExcess C consumptionBack reactionFreeze/Melt of bath/ledgeUnperturbed process design is complex. So Co

15、ntrol Design needs to decouple/remove interactions, and to account for the response of the processes. Controllingthe Outcome Understandingthe Variation Observing Process StateMeasureSignalResponse Control, State, and Manipulated Variable Responses Signals for Corrective Action(op/maint) Closed Loop

16、Controller Model Audit and Learning on System + link to improvement Control Objective Understood Removecauses ofabnormal TT responseto Energy,or Comp.T, TL , Al2O3 pt.outside naturalvar.- immediate T, TL . processvelocity -reduce this Control wins+ failures plusIdent. causes A thermal shock of 30kW

17、increase in heat loss:on a bath mass of 5000 kg, s/w heat loss 40kWA thermal shock of 45kWon a bath mass of 5000 kg, s/w heat loss 40kWA thermal shock of 49kWon a bath mass of 5000 kg, s/w heat loss 40kWHow do existing engineering control loops work ?For any measured variable, an arbitrary definitio

18、n of its State is made.ArbitraryScaleFor MeasuredVariable(T, %AlF3)Variable 1 . Variable 2ExceptionstateExceptionstateTargetDeadBandUpper ControlLimitLower ControlLimitControlrangeControlrangeExceptions are too late - bad damage.Over-controlis common.Change inControl Variable(Usually R, AlF3 add etc

19、 )0Process Variable (maybe T, %AlF3)TargetControlTargetincreasedControlTargetreduced2. An arbitrary, piece-wise linear, proportional gain is defined.Dead BandProportional Controllers are the default. Gain is max responseDB is measurementerror. or something.So, Control of Superheat should be:Correcti

20、ve, not compensatoryEarly, not late.Damping variation, not over-controllingVisibly related to the processes (and causes) on the floorResponding to signals of abnormality, not the structure of the process designAble to identify control system actions which are impacting the variation (+ve,-ve)We stri

21、ve to put every Cell in a Multivariate Control WindowCell V/I feasible region, bounded by ACD, Heat Diss limits.Bath T/Composition feasible region, bounded by electrolyte phase equilibria, alumina dissolution.Materials Service limits for walls, electrodes, lining, cover.Electromagnetic and hydrodyna

22、mic stability of bath/metal interface.Electrode current density/distribution, resistance and over-voltage with parasitic and degrading chemical reactionsFreeze,MeltinfluencedBathDepthMetalDepthtimeCell Multidimensional SpaceAmpsACDSludgeV Instab.Al2O3diss.Al2O3consum.TTLCover Anode Consum.TargetZone

23、 forvariablesBut what is thewindow we areaiming for ?Derived Steady State Temperature/Superheat Operating Window90091092093094095096097098099010000510152015C Superheat5C SuperheatAre all cells in thisOperating Region ?SLUDGE SOLIDIFICATIONElectrolyte Temperature ( C )LOW CURRENTEFFICIENCYExcess AlF3

24、 concentration (wt%)Poor aluminaDiss.No ledgeHow long can cells live outside this window without process damage ? Cell B935940945950955960965970975980985791113151719xs AlF3 (%)Bath Temperature (C)20-Jan-03Cell A940945950955960965970975980985791113151719xs AlF3 (%)Bath Temperature (C)All the “Cell Bs

25、” are out here. Compensatory control hides the cause and magnifies its impact on the bottom line Why does Cell B operate with greater variation under the same process design ? Energy Composition when side- freeze PresentFreeze / melt Q sidewalls Temperature(bath) T Liquidus Superheat=-Chemical addit

26、ionsVoltage/R changeMetal tapping controlCover applicationFeeding disturbancesCONTROL ACTIONS are the likely CAUSE Controllingthe Outcome Understandingthe Variation Observing Process StateCriticality Risk (Type 2)anode failuresCECell walls Tampering Risk- (Type 1)Is the dominant cause known ?Is it p

27、art of the process design ?Will you increase variation further ?Can we act on variation inside +/- 3 ?YES is the Answer Control is a stream of Data, Analysis and DecisionsEach decision has a risk profile (probability of success, and consequences) Doing nothing is a decision option. with consequences

28、 The Control Design here is multivariate, and focused on removing causes of abnormalityNot proportional controlNot compensating for other variationInvolving the people who control the causesWhat type of Decision is appropriate?Compensatory making another change to compensate for the variation. (Ok i

29、f the cause and the compensating change are well understood, and the cause cannot be removed).But .“removing crane high tensile bolts which are failing and replace with heavier bolts ?” Corrective removing the source of this variation (immediate cause).Eg. Operator error on crane ? Root Cause Elimin

30、ation eg. Find the source of the overloading andremove this variation.Management System eg. What was the training system whichallowed operators to overload the crane during operation?What was the design criterion for loading and duty cycle of the crane ?Why is there a perceived need to apply this de

31、gree ororientation of loading ?CompensatoryCorrectiveRoot CauseManagement SystemsSpeedEffectivenessManagement SystemsRoot CauseImmediate CauseLoss / RecoveryTime SequenceRoot Cause AnalysisWhat type of Variation are we dealing with?Is the variation part of the way you normally run this process? (Is this the first time a bolt fell off the crane ?)A special cause can be addressed by a single action, eliminating a condition / error which will not recur. Special variation is not part of the way the process runs.A common cause is more probable, and often leads back to the management systems