You are on page 1of 4

29/3/2015 AutomationBasics:DistillationcolumnlooptuningISA

TRAINING & STANDARDS & CONFERENCES NEWS & TECHNICAL PROFESSIONAL


MEMBERSHIP CERTIFICATIONS PUBLICATIONS & EVENTS PRESS RELEASES RESOURCES TOPICS DEVELOPMENT STORE

HomeISAPublicationsInTechMagazine2014/NovDecAutomationBasics:Distillationcolumnlooptuning

Distillation column loop tuning AbouttheAuthor


HarleyJefferyhasworkedinthe
fieldofindustrialprocesscontrol
Trust the operator, but verify the basics formorethan40years,starting
ByHarleyJeffery withFisherControlsasavalve
applicationengineer.Hehas
beenacontrolsystemsengineer
Controllingdistillationcolumnsisatoughassignmentduetotheinteractingnatureoftheprocess
andengineeringmanagerwith
andtheupstream/downstreameffectsoflooptuning.Inthisexample,thecolumnbottomslevel
ControlSouthernforthepast34
controlhadstabilityproblems(cycling),causingmanualoperationandtheoperatorsconstant
years.Hisexperienceincludes
attention.
design,implementation,testing,
startup,andongoing
Thebottomslevelneededtobecontrolledtoasetpointthatminimizedthetimetheliquidsatinthe improvementofDCSs.Jeffery
receiverforqualitypurposes.However,ithadtomaintainafairlytightrangetopreventfloodingthe receivedaBSIEfromLouisiana
bottomtrayontherangeupperendandtomaintainsufficientpumpsuctionheadforthebottoms StateUniversity.
pumpoutforthelowerconstraint.

Previouslooptuninghadresultedinexcessivemovementstothebottomsflowthatadversely MoreAutomationBasics
loadedthedownstreamcolumn.Thus,theoperatorsrantheleveltoflowcascadecontrolloopsin PropermotorprotectionwithIEC
manualandconstantlymanipulatedtheoutflowtomaintainthebottomslevelwithinthedesired versusNEMA
range.Ofcourse,asothermattersrequiredoperatorattention,thebottomslevelwoulddrift.When Controlvalvesanupdate
theoperatorintervenedfortherequiredcorrection,theresultantupsetofthecolumnsaffected
TheArtofinstrumentselection
productionrateandquality.
AnhourwithDoctorFlowmeter
Discussionswithoperatingpersonnelrevealedthattheytypicallyletthelevelfloatwithinacomfort Hybridtemperaturecontrollers
range,becausetheprocesswouldfluctuatebutnottothepointofconcern.Ifitlookedlikeanupor offermoreversatility
downtrendwasdeveloping,thenasmallbumpwasmadetotheoutletflow,andtheywaitedfor ThermocouplesversusRTDs
correction.Ifthedeviationfromsetpointbecamegreater,thenlargerbumpstotheoutflowwere SeeallAutomationBasics
performed,andtheyagainwaitedforcorrection.Whenaskedtoplacethelevel/flowincascade Articles
operatingmode,thelevelloopwouldtypicallyovercorrectandbegintocycle.Theoutletflowloop
wasconsideredtobeworkingwelleventhoughitwasmanipulatedinmanual.
ReaderFeedback
Theoperatorshadlearnedfromexperiencewhatappearedtobeanonlineargaincontrolstrategy. Wewanttohearfromyou!Please
Trustingtheirjudgment,wethenproceededtoverifythebasicsbeforeimplementingthestrategy. sendusyourcommentsand
questionsaboutthistopicto
InTechmagazine@isa.org.
Basics Loop tuning and performance

Plant walk through
Thefirststepforloopperformancebenchmarkingiswalkingthroughtheprocesstoinvestigatethe
typeandinstallationoftheprocessmeasurementandfinalcontrolelements.Then,returningtothe
controlroom,werecordedthedistributedcontrolsystem(DCS)controlstrategyandtuning.The
loopinspectiondataformswereusedtodocumenttheasfoundcontrol(figure1).The
measurementswereselectedwiththelatesttechnologiesandappearedtobeinstalledwithgood
practices.Thecontrolvalveswereahighperformancedesignandofferednoconstraintstopush
loopresponseifneeded.Theasfoundtuningparameterswerequestionableandtherefore
confirmedourprocessofverifyingbaselevelloopperformance.

Figure1.Loopinspectiondataforms

TheDCSconfigurationapplicationwastracedtoverifyatypicalleveltoflowcascadestrategyper
figure2.TheDCSalsohasseveralbuiltincascadeapplicationenhancementsthatweactivated.

https://www.isa.org/intech/201412basics/ 1/4
29/3/2015 AutomationBasics:DistillationcolumnlooptuningISA

Figure2.Bottomsleveltoflowcascadestrategy

Designing and testing


Verifyingperformancebeginswithdesigningthelooptests,whichincludegatheringtimeseries
dataoftheprocessmeasurementstoanalyzetheoverallprocesstrends,statisticsforvariability,
andresponsetoupsets.Thenweperformindividualloopbumpteststogainprocessdynamicdata
andvalveperformance.WeelectedtousetheEnTechToolkittocollect,analyze,andhelptunethe
controlloops.However,thisplantusesFoundationFieldbus,whichcommunicatesdigitallytothe
DCS.Thetoolkitrequiresavoltagesignalfortheprocessinputs.Arecentlyaddedfeatureofthe
DCSusescharacterizableI/Omodules,sotheywereemployedtoreadtheprocessvariablesover
theDCScommunicationsstructureandreproducethevaluesonanalogoutput(AO)modules
(figure3).ThetoolkitwasthenconnectedtotheseAOsandsuccessfullyabletocollectpertinent
processdatafromtheDCS.

Figure3.CollectinganalogdatafromFoundationfieldbusdevices

Withthetoolkitgatheringprocessdata,wecouldcollecttimeseriesshorttermdataandovernight
runsforlongertermdataandsawthedisturbancetothelevelandvariousoperatorresponses.The
loopbumptestingconfirmedthattheslaveflowloopwascapableofaggressivetuningwithgood
responseto0.5percentbumps(figure4).

https://www.isa.org/intech/201412basics/ 2/4
29/3/2015 AutomationBasics:DistillationcolumnlooptuningISA

Figure4.Flowloopbumptest

Figure5.Nonlineargainalgorithm(NLG)

Loop tuning and verifying performance


Theflowloopwasretunedwiththenewparameterstominimizedelayinresponsetothemaster
levelcontrol.Withtheflowloopincascademode,theleveloutputbumpsverifiedan

integratingprocesswithalongdeadtime.Thelevelcontrollerwasretunedwithincreasedgainand
veryslowreset.Withtherevisedtuning,thelevelcontrollerwasthenputinautomaticwiththeflow
loopincascade.Wemonitoredtheperformance.Theresultsshowedcyclingwasminimized,and
thesetpointwasmaintainedwithinthedesiredlimits.

However,thestandardproportional,integral,derivative(PID)leveltoflowcascademanipulatedthe
flowtomaintainthesetpointwiththesameresponseforsmalldeviations,aswellaslargerones.
Thisdegreeofflowchangetothedownstreammovementcauseddisturbancestothehighly
interactivedistillationprocess.

Advanced control (trusting the operator)


Thequestionthenbecamehowtomaintainanacceptablebottomslevelwhileminimizing
downstreamdisturbancescausedbyoutletflowchanges.Returningtotheoperatorsmethod
manuallymakingsmallchangestotheoutletflowtomaintainthelevelwithinacomfortband
aroundthesetpointandonlymakinglargerflowchangesifthelevelwasapproachingaconstraint
hadthebenefitofminimizingflowdisturbancestothedownstreamcolumn.

Theoperatormethodlookedlikeagoodexampleofthenonlineargaincontrolalgorithm(figure5).
Thistechniqueusesagaparoundthesetpointwheresmallgainisineffect.Thisisequivalentto
theoperatornotchangingtheoutletflowiftheleveliswithinhisorhercomfortzone.However,as
thePVSPerrorincreases,thegainisincreasedtomaketheappropriatecorrectionagain,asthe
operatorseesthatthelevelis,infact,headedtoaconstraint,thenadditionaloutletflowisusedto
correctthelevel.

TheDCShasabuiltinNLGfunctionthatcanbeenabledonthePIDalgorithm:

NL_MINMODisthegainappliedwhentheabsolutevalueoftheerrorislessthanNL_GAP.To
getdeadbandbehavior,setNL_MINMODto0.
NL_GAPisthecontrolgap.WhentheabsolutevalueoftheerrorislessthanNL_GAP,KNL=
NL_MINMOD.
NL_TBANDisthetransitionbandoverwhichKNLislinearlyadjustedasafunctionoferror.
NL_HYSTisahysteresisvalue.UntiltheabsolutevalueoftheerrorexceedsNL_GAP+
NL_HYST,KNL=NL_MINMOD.OncetheabsolutevalueoftheerrorhasexceededNL_GAP+
NL_HYST,theabsolutevalueoferrormustreturntoavaluelessthanNL_GAPbeforeKNL

https://www.isa.org/intech/201412basics/ 3/4
29/3/2015 AutomationBasics:DistillationcolumnlooptuningISA
returnstoavalueofNL_MINMOD.IfNL_GAPis0,thenthevalueofNL_HYSThasnomeaning
(effectivelyassumedtobe0).

Wethenenabledthisfeatureandcollectedadditionaldata/observationsinordertotunethegap,
transitionband,minmod(gain),andhysteresis.Ingeneral,wetriedtomimictheoperatorsmethod
andaddasmootherimplementationoftheflowsetpointchanges.

Results
Placingthebottomslevelmastercontrollertoautomaticandtheslaveoutletflowcontrollerto
cascadewiththenonlineargainenabledandtuned,thelevelwasmaintainedwithacceptablelimits
aroundsetpointwhileminimizingtheflowtothedownstreamdistillationprocess.Recentfollowup
withtheplantsiteindicatedthattheloopsarestillinautomatic,allowingoperatorstospendtimein
moreproductiveefforts.

https://www.isa.org/intech/201412basics/ 4/4

You might also like