Table Of ContentPROCESSCONTROLDESIGNANDANALYSISFORWASTEWATER
DISINFECTION,STREAMNEUTRALIZATION,ANDRUN-TO-RUNSTRATEGIES
By
CHRISTOPHEREDWARDMEREDITH
ADISSERTATIONPRESENTEDTOTHEGRADUATESCHOOL
OFTHEUNIVERSITYOFFLORIDAINPARTIALFULFILLMENT
OFTHEREQUIREMENTSFORTHEDEGREEOF
DOCTOROFPHILOSOPHY
UNIVERSITYOFFLORIDA
2003
Copyright2003
by
ChristopherEdwardMeredith
ToBeckyandRuby
ACKNOWLEDGMENTS
Iwouldliketoexpressmysincereappreciationtomyadvisor,OscarCrisalle,
withoutwhosesupportthisdissertationwouldnothavebeenpossible. Iamalsograteful
fortheopportunityhehasgivenmetoobtainaminorinelectricalengineering.
IwishtothankProfessorsHaniphLatchman,SpyrosSvoronos,andJasonWeaver
forservingonmysupervisorycommittee. Dr.Svoronoswasparticularlyhelpfulinthe
developmentofthewastewaterchlorinationproject,aswereChuckFenderandProfs.
PaulChadikandRobertThieke.
IthankH.MikeMahon,VaradharajanBasker,GiovaniNunes,JonEngelstad,
SerkanKincal,andChuckBaab,whocamebeforeandBrianRemarkwhowillfollow.
Theyallhavenotonlycontributedgreatlytomyresearchbuthavebeengoodfriends.
Finally,Iwishtothankmymother,Sheila,andmybrother,Doug,fortheir
unendingsupport.
IV
11
TABLEOFCONTENTS
page
ACKNOWLEDGMENTS
iv
ABSTRACT
viii
CHAPTER
INTRODUCTION
1 1
2 WASTEWATERCHLORINATIONUSINGANOVELCONTROL 6
2.1 Introduction 6
2.2 ProcessModelofaTypicalWastewaterDisinfectionPlant 7
2.2.1 KineticModelofChlorineReactioninaWastewaterTreatmentPlant 10
2.2.2 DosageAreaModel 13
2.2.3 ContactAreaModel 17
2.2.4 SimplifiedModelofaPlugFlowReactor 18
2.2.5 LinearizedState-SpaceModelofaPlugFlowReactor 2
2.3 PlantDataFittingandParameterEstimation 22
2.4 ControlDesignandAssessmentofPerformance 27
2.4.1 DescriptionandTuningoftheSlavePIController 3
2.4.2 DescriptionandTuningoftheMasterPIController 37
2.4.3 DescriptionandTuningoftheMasterPIControllerandSmithPredictor....44
2.4.4 DescriptionandTuningoftheMasterPIControllerandAdaptiveSmith
Predictor — 48
2.5 NovelProcessModification DynamicWeir 52
2.5.1 DynamicWeirControllerDesign 55
2.5.2 Closed-LoopResponseoftheEntireModelforStepChangesinFlowrate
UsingaDynamicWeir 60
2.5.3 Closed-LoopResponseoftheEntireModelforVaryingFlowrateUsinga
DynamicWeir 64
2.5.4 RobustnessAnalysisfortheModel-BasedResidenceTimeController 66
2.5.5 PropagationofErrorAnalysisforthePIFeedback-ControlledSystem 69
2.5.6 AdditionalBenefitsoftheDynamicWeirDesign 72
2.6 Conclusions 75
3 STABILITYOFPI-CONTROLLEDRUN-TO-RUNSTATICSYSTEMSWITH
SIGNIFICANTDELAY 79
v
1
3.1 Introduction 79
3.2 DerivationofaPIRun-to-RunController 80
3.3 PlantStabilityunderFeedbackControl 82
3.3.1 NominalStabilityRegion 83
3.3.2 RobustnessAnalysis 87
3.3.3 TuningCorrelations 91
3.4 PlantStabilityunderPIControlwithaSmithPredictor 94
3.4.1 Closed-loopdynamicsunderPIcontrolwithaSmithPredictor 95
3.4.2 StabilityRegionasaFunctionofModelUncertainty 96
3.5 Conclusions 99
4 CONTROLOFPHUSINGANOVELOPTICALSENSOR 101
4.1 Introduction 101
4.2 ModelingandSimulationofapHControlSystem 104
4.2.1 ProcessModelofpHControlinaCSTR 106
4.2.2 TraditionalElectrodes 11
4.2.3 NovelOpticalSensors 116
4.3 Closed-LoopSimulations 122
4.3.1 Noise-FreeClosed-LoopPerformance 127
4.3.2 Closed-LoopPerformancewithRealisticNoise 135
4.4 DesignofanExperimentalSystem 141
4.4.1 Apparatus 142
4.4.2 MeasurementNoiseAnalysis 145
4.5 Conclusions 150
5 CONCLUSIONSANDFUTUREWORK 152
APPENDIX
A DISTRIBUTEDPARAMETEREQUATIONDERIVATION 157
B PERFORMANCEASSESSMENTOFTHESTANDARDFEEDBACK
CONTROLCONFIGURATION 160
C REVIEWOFTHEOPERATIONOFADISCRETECONTROLLER
APPLIEDTOACONTINUOUSSYSTEM 164
DCONDITIONSFORWHICHTHETRANSFERFUNCTION(3-8)HAS
UNSTABLEPOLE-ZEROCANCELLATIONS 165
E VALUESOFaANDpFORTHE2-PARAMETERCURVEFITSOFTHE
FUNCTIONSGENERATEDBYTHEJURYTEST 167
F CONDITIONSFORWHICHTHETRANSFERFUNCTION(3-38)HAS
UNSTABLEPOLE-ZEROCANCELLATIONS 168
GREVIEWOFWEAKACID-STRONGBASECHEMISTRY 170
vi
HAPPLICATIONOFDESCARTES'RULEOFSIGNS 171
I DEVELOPMENTOFTHEEXHAUSTIVENUMERICALSEARCHUSEDTO
OPTIMIZECONTROLPERFORMANCE 174
J MOVINGAVERAGEANDEXPONENTIALFILTERS 177
LISTOFREFERENCES 180
BIOGRAPHICALSKETCH 185
AbstractofDissertationPresentedtotheGraduateSchool
oftheUniversityofFloridainPartialFulfillmentofthe
RequirementsfortheDegreeofDoctorofPhilosophy
PROCESSCONTROLDESIGNANDANALYSISFORWASTEWATER
DISINFECTION,STREAMNEUTRALIZATION,ANDRUN-TO-RUNSTRATEGIES
By
ChristopherEdwardMeredith
December2003
Chairman: OscarD.Crisalle
MajorDepartment: ChemicalEngineering
Thewastewaterchlorinationprocessinfixed-sizeopen-channelreactorsisa
challengingcontrolproblembecauseofthepresenceofavariabledeadtimecausedby
changesintheinletflowrate. Theapproachproposedhereinvolvestheutilizationofa
movingweirasanactuatorinanovelresidencetimecontrolscheme,designedtoensure
thatfluidpassingthroughthereactorexperiencesaconstantresidencetimeindependent
offlowrate. Thedynamicweircontrolloopisconceivedtosupplementexistingcontrol
structuresthatarechargedwithmanipulatingthedosageofchlorine. Simulationsstudies
showthatbothfeedforwardandfeedbackmanifestationsoftheresidencetimecontroller
arehighlyeffectiveatimprovingchlorinedosagecontrolandarehighlyrobustwith
respecttoweir-flowmodelingandmeasurementerrors. Infact,thesecontrollershavethe
potentialtoimprovethecontrolperformancedramaticallywhilesavingasmuchas45%
ofthechlorineusedfortreatment.
viii
Anautomaticcontrolsystemdescribedasrun-to-runisoneinwhichadiscrete
controllerisappliedtoaninherentlydiscreteprocess,andwheretheperformance
measurementsaremadeonlyafterallprocessingiscompleted. Acompletetheoretical
analysisisdevelopedforthecaseofastaticplantunderproportional-integralcontrol,
resultinginquantitativedescriptionsofthenominalandrobuststabilityregions,and
recommendedtuningheuristics. TheanalysisforthecaseofaSmithPredictordesign
showsthatverysmallmodelingerrorsintheestimatedmeasurementdelaymaycause
closed-loopinstability. Therefore,theuseofaSmithPredictorforrun-to-runsystems
sufferingfromuncertaintyinthedelayisnotrecommended.
Aweak-acid/strong-baseneutralizationcontrolsystemutilizingtwoalternative
pHsensorsismodeledandanalyzed. ThefirstsensorisatraditionalpHelectrodewhose
outputsignalisproportionaltopH,andthesecondsensorisafiber-opticoptrodethat
producesasignalthatisproportionaltohydrogenionconcentration. Systematic
simulationstudiesshowthatnoinherentperformanceimprovementisobservedwhen
usingtheoptrodeinsteadoftheelectrodesensorinthefeedbackloop.
IX
CHAPTER
1
INTRODUCTION
Therearethreetypesofprocesscontrolstudiesaddressedinthisdissertation,
namely,adesignproject,atheoreticalproject,andanexperimentalproject.Thedesign
projectexamineswastewaterchlorinationcontrol,adistributed-parametersystemthatis
typicallydifficulttocontrolwithstandardstrategiesduetothepresenceofhigh
variationsinflowrate. Itisshownthatanoveldesignmodificationbasedonan
automaticallyadjustedmovingweircanleadtodramaticallyimprovedperformance.In
thetheoreticalstudy,single-inputsingle-outputrun-to-rununitoperationsare
investigatedforthecaseofastaticplant. Therun-to-runcontrollerformiscontrasted
withastandarddiscretecontrollerform,exposingsubtledifferencesinbehavioral
characteristicsunderclosed-loopcontrol. Finally,theexperimentalprojectseeksto
improvetheperformanceofpHcontrolsystemsthatusetraditionalpHelectrodes.
Knowntobehighlynon-linear,theelectrodemayhavelimitedperformancepotentialasa
result. Anovellinearopticalsensorisintroducedasanalternative.
Chapter2presentsadesignstudyinvolvingastandardwastewaterchlorination
basin. Itisshownthattraditionalcontrolschemeshavedifficultymaintainingtheoutlet
concentrationataspecifiedvalue. Thedifficultyiscausedbytheeffectthatthevarying
flowrateofwastewaterhasonthedeadtime(residencetime)ofthesystem. Becausethe
volumeofthereactorisconstant,residencetimeisinverselyproportionaltoflowrate,
andthelattercanvarybyasmuchasanorderofmagnitude. Specifically,thevariable
residencetimecausespoorclosed-loopperformance. Thesystemismodifiedby
1