Session 3.2 - Michael Hyland - International Ultraviolet Association

Coquitlam UV Disinfection Project: Closing the (Vertical)
Loops on UVDGM Piping Geometry Guidelines
Mike Hyland, P.E.1 and Chris Schulz, P.E.2
Andrew de Boer, P. Eng, Duane Morris and Karen Tully, P. Eng.3
1
CDM Smith, 1218 3rd Avenue, Suite 1100, Seattle, WA 98101
CDM Smith, 555 17th Street, Suite 1100, Denver, CO 80202
3
Metro Vancouver, 4330 Kingsway, Burnaby, B.C. V5H 4G8
2
Thispapercomparesresultsofafull‐scalevelocityprofileanalysisforReactorLoopNos.1,4,and
8takenupstreamofeachultraviolet(UV)reactorbyADSEnvironmentalServices,summarizedin
aseparatereportdatedMarch28,2014tothepredictedvelocityprofile(s)fromdesign‐phase
effortsincludingphysicalmodeling.
Background
TheUltravioletDisinfectionGuidanceManual(UVDGM)1waspublishedbytheUnitedStates
EnvironmentalProtectionAgencyinNovember2006andprovidestechnicalinformationand
guidanceondesignandoperationofUVsystemsforreceivingdisinfectioncreditforcomplying
withfederaldrinkingwaterregulationsintheUnitedStates.TheCoquitlamUVDisinfection
FacilitywasdesignedtomeetdesignrequirementsoftheUVDGMsincecomparableguidanceis
notcurrentlyavailablefromHealthCanada.
AnimportanttopicintheUVDGMrelatestoinletandoutletpipingconfigurationsinstalledina
full‐scaleUVfacilityandhowtheyrelatetothepipingconfigurationusedduringvalidation
testingandassociatedimpactsonUVdosedelivery.Theoverallobjectiveistoensurethat"the
inletandoutletpipingtotheultravioletreactorintheultravioletfacilityresultsinanultraviolet
dosedeliverythatisequaltoorgreaterthantheultravioletdosedeliveredwhenthereactorwas
validated."TheUVDGMrecommendsthreeoptionsformeetingthisobjective:
1. MinimumfivepipediametersofstraightpipeupstreamofUVreactor:Thelengthof
straightpipeupstreamofeachUVreactorattheUVfacilityisthelengthofstraightpipe
usedinthevalidationtestingplusaminimumoffivepipediameters.
2. Identicalinletandoutletconditions:Inletandoutletconditionsusedduringvalidation
matchthoseusedatthewatertreatmentplantforatleasttenpipediametersupstream
andfivepipediametersdownstreamoftheUVreactor.
3. Velocityprofilemeasurement:Velocityofthewatermeasuredatevenlyspacedpoints
throughagivencross‐sectionoftheflowupstreamanddownstreamofthereactoris
1USEPA,UltravioletDisinfectionGuidanceManualfortheLongTerm2EnhancedSurfaceWaterTreatmentRule,Officeof
Water,USEPA,November2006.
within20percentofthetheoreticalvelocityforboththevalidationandwatertreatment
plantinstallations.
Whiletheseconditionswilllikelyresultinimprovedinlethydraulicsrelativetovalidationtest
conditions,theyrepresentconservative"rulesofthumb"criteriaandarenotsupportedby
empiricaldataorhydrauliccalculationstodemonstratetheyhavebeenoptimizedtoreflect
typicalUVsystempipinglayoutsorUVreactordesigns.Options1and2inparticularmayhave
significantfacilitylayoutandcapitalcostimplicationsfordesigninganewstand‐aloneUVfacility,
andmaynotbefeasibleforretrofitapplications,suchaswhenUVreactorsareinstalledinan
existingfilterpipegallery.AndthecriteriaforOption3,asitappliestooutletvelocityprofile
measurements,aredifficulttomeetforUVreactordesignsthatemployinternalbafflesfor
improvedUVdosedistributionwithinthereactorvessel.
FortheCoquitlamUVDisinfectionProject,analternativeapproachwasusedformeetingthe
UVDGMdosedeliveryobjectivecitedabove,whileallowingforflexibilityinthedesignofthe
verticallooppipingarrangementtomeetgeotechnicalandenvironmentalsiteconstraints.These
site‐specificdesignconstraintsdidnotpermitcompliancewithOptions1and2above,which
wouldhavesignificantlyincreasedtheconstructioncostsandrisksassociatedwiththeproject.
Thisapproachinvolvedthefollowingmodelingandfieldverificationstepscompletedduringthe
designandcommissioningphasesoftheproject:

Physicalmodeling.Aphysicalmodelatanapproximate5to1scalewasconstructedof
thecommoninlet/outletheaderandverticalpipelooparrangementtoevaluateand
optimizereactorinlethydraulicsfortheCoquitlamUVFacility.Thisworkwascompleted
duringthepreliminarydesignphasebyClemsonHydraulicsundersubcontracttoCDM
Smith.

CFDmodeling.Acomputationalfluiddynamics(CFD)modeloftheTrojanTorrentUV
reactorandfinalpipingarrangementfortheCoquitlamUVFacilitywasdevelopedto
predictvelocityprofilesupstreamofthereactoratdifferentdesignflowsandcompare
againstthephysicalmodelresults.ThisworkwascompletedbyTrojanTechnologiesand
resultsarepresentedinaValidationReportfortheTorrentUVreactor.

Fieldverification.VelocityprofilemeasurementsweretakenupstreamofselectedUV
reactorsduringvalidationandstart‐upofthefull‐scaleUVfacility.Thefour1.5‐inchvalved
outletslocatedupstreamofeachUVreactorinthelowerleveloftheUVReactorRoom
wereusedtotakethesemeasurements.ThisworkwascompletedbyADSEnvironmental
Systemsandispresentedintheirreport.
HydraulicAcceptanceCriteria
Acceptancecriteriaforinlethydraulicsshoulddemonstratethatvelocityprofilesimmediately
upstreamoftheUVreactorforthefull‐scaleUVsystemwillresultinaUVdosedeliverythatis
equaltoorgreaterthantheUVdosedeliverymeasuredduringvalidationtesting.Tomeetthis
requirement,pipelinevelocitymeasurementsforfull‐scaleandvalidationtestinstallations
shouldbetakenatevenlyspacedpointsthroughagivencross‐sectionoftheflowtogeneratea
datasetofinstantaneousvelocitymeasurements(typically20to40measurements).Thetwo
datasetscanthenbestatisticallyanalyzedtoverifythattheinlethydraulicsforthefull‐scaleUV
facilityarethesameorbetterthanforthevalidationtestinstallation,whichinturnshouldresult
inthesameorhigherUVdosedelivery.
Foracceptableinlethydraulics,thefollowingacceptancecriteriawereestablishedandself‐
imposedfortheCoquitlamUVDisinfectionProject:
1. Thenormalizedvelocityatanindividualsamplepointforthefull‐scaleinstallationshould
beequaltoorlessthan1.2(i.e.nomorethan20percenthigherthantheaverage
measuredvelocity).Thenormalizedvelocityiscalculatedasthemeasuredvelocityata
particularsamplepointdividedbytheaveragemeasuredvelocityfortheentiredataset,
andisunitless.
2. Thecoefficientofvariation(CoV)for"highvelocity"datapointsforthefull‐scale
installationshouldbeequaltoorlessthantheCoVfor"highvelocity"datapointsforthe
validationtestinstallation.Thehighvelocitydatapointsaredefinedasdatapointswith
normalizedvelocitiesequaltoorgreaterthan1.0.TheCoViscalculatedasthestandard
deviationofthehighvelocitydatapointsdividedbytheaveragemeasuredvelocityofthe
highvelocitydatapoints,expressedasapercentage.
3. The95thpercentilefor"highvelocity"datapointsforthefull‐scaleinstallationshouldbe
equaltoorlessthanthe95thpercentilefor"highvelocity"datapointsforthevalidation
installation.The95thpercentilewasselectedtoexcludeanyhighvelocitydatapoint
outliersattributabletoanalyticalvariabilityofthevelocitymeasurements.
ThefirstacceptancecriterionisrelatedtoOption3intheUVDGMguidancediscussedabove.Note
thatthesecondandthirdacceptancecriteriadonotinclude"lowvelocity"datapointsintheCoV
statisticalanalysis(i.e.,datapointswithnormalizedvelocitieslessthan1.0)sincelowvelocities
areindicativeoflongerUVexposuretimesandthereforedonotnegativelyimpactUVdose
delivery.
FieldTesting
ADSperformedvelocityprofiletestingofUVReactorLoopNos.1,4,and8fortheCoquitlamUV
Facility.Theseloopswereselectedtocomparevelocityprofilemeasurementsforpipeloops
locatedatthebeginningandendoftheinletpipeheader.Afullvelocityprofilewasperformedon
eachreactorloopusingfour1.5‐inchtapslocated23cmupstreamofthe1,200‐mminletflangeof
theTrojanTorrentUVreactor.ThetestingworkwascoordinatedwithMVoperationsstaffto
maintainaconstantflowthrougheachreactortestloopforapproximatelyonehourwhilethe
measurementswerebeingtaken.Thiswasadifficultobjectivetoachieveinpractice,sincethe
plantisanondemandfacilityandflowsvaryfromhourtohour.Consequently,flowstotest
reactorloopswereallowedtovarywithinapproximately10percentofthetargetflow.
VelocitymeasurementsweremadeusingaPitometerrodthatwasinsertedintothepipecarrying
flowunderapproximately40psiofpressure.Theendoftherodwasequippedwithtwoorifices,
onefacingupstreamandtheotherfacingdownstream.Thevelocityofwaterproduceda
differentialpressurebetweentheorificesthatwasmeasuredusingaRosemonttransmitterand
recordedonalaptopcomputer.Thewatervelocitywascalculatedusingastandardorifice
equation.Foreachtransect,localvelocitiesweremeasuredatthecenterandtenotherlocations.
TheADSreportgenerallyincludesthefollowinginformationforReactorLoopNos.1,4,and8
whichwereusedinthevelocityprofileanalysispresentedbelow.

PointvelocitytablesandtransectprofilegraphsforTaps1,2,3,and4forflowsof130and
170ML/day

Averagevelocityprofilesummarytables,averageprofilegraphsandpipecross‐section
graphsforflowsof130and170ML/day(total6tablesand4graphs)
Thenumberoflocalvelocitypointmeasurementsforeachflowtestrangedfrom33pointsfor
LoopNo.1(usingTaps2,3and4)to44pointsforLoopNos.4and8(usingTaps1,2,3,and4).
VelocityProfileAnalysis
Table1presentsthevelocityprofileanalysisforReactorLoopNos.1,4,and8fordesignflowsof
130and170ML/dayusingdatafromthepointvelocitytablesintheADSreport.Theseresults
arecomparedagainstthefollowingmodelingandvalidationtestingworkcompletedforthe
project.




PhysicalmodelforCoquitlamUVFacility(132ML/day)
PhysicalmodelbasedonUVDGMupstreamhydrauliccriteria(132ML/day)
CFDmodelforCoquitlamUVFacility(125ML/day)
OffsitevalidationtestingforTrojan'sTorrentUVreactor(134ML/day)
Table2presentsthesametypeofanalysis,exceptthatonly"highvelocity"datawereincludedin
thedatasetsforeachreactorloopinordertoverifycompliancewiththeacceptancecriteria
discussedabove.
Thestatisticalanalysispresentedinbothtableswasperformedusingthefollowingsteps:
1. Determinethenominalflowrateandsamplepointlocationsforeachdesigncondition
basedoninformationprovidedintheADSreportandotherreportspreparedforprevious
modelingandvalidationtestingwork.Thisinformationispresentedunderthe"General"
columninthespreadsheets.Notethatthenumberofradiiandsamplelocationsvaried
acrossthedifferentvelocityprofileinvestigations.Thenumberofvelocitysamplepoints
rangedfrom24to40points.
2. Entermeasuredvelocitydataatlocalizedsamplepointsforeachdesignconditionfrom
theADSreport(forreactorLoops1,4,and8)andotherinvestigations.Thisinformation
isprovidedforthematrixofsamplepoints(P1thruP8)andradii(R1thruR5)foreach
investigation.
3. Calculatetheaveragemeasuredvelocitybytakingtheaverageofthemeasuredvelocity
dataforeachinvestigation(seeGeneralcolumn).
4. Normalizethevelocitydatabydividingthemeasuredvelocityforeachsamplepointbythe
averagemeasuredvelocity.Ifthemeasuredvelocityataparticularsamplepointisthesame
astheaveragemeasuredvelocityfortheentiredataset,thenormalizedvaluewillbe1.0.
5. Performstatisticsonthenormalizeddatasetandrecordthemaximum,minimumand
averagevalues,andthestandarddeviationandCoV,foreachinvestigation.
6. Usetheinformationinbothtablestopreparefigurescomparingthevelocityprofiles
acrossthedifferentinvestigations,asdiscussedbelow.
Figures1through5presenthistogramsofthenormalizeddatasetsforreactorLoopNos.1,4,
and8fornominaldesignflowsof130and170ML/day(LoopNo.8wasonlytestedat
170ML/day).Eachfigurealsoincludesahistogramofthenormalizeddatasetforvalidation
testingoftheTrojanTorrentUVreactorforcomparisonpurposes.Thedataarepresentedin
orderfromhighesttolowestnormalizedvelocityvalues.Valuesabove1.0aredefinedas"high
velocity"valuesanddatabelow1.0are"lowvelocity"values.
Figure 1. Velocity Profile Histogram for Reactor Loop No. 1 (130 ML/day) Figure 2. Velocity Profile Histogram for Reactor Loop No. 1 (170 ML/day) Figure 3. Velocity Profile Histogram for Reactor Loop No. 4 (130 ML/day) Figure 4. Velocity Profile Histogram for Reactor Loop No. 4 (170 ML/day) Figure 5. Velocity Profile Histogram for Reactor Loop No. 4 (170 ML/day)
Thefollowingconclusionscanbedrawnfromtheabovehistograms:

Thehighestvelocitydatapoints(twobarsontheleftsideofeachgraph)forallfourreactor
loopswerelowerthanforthevalidationtestinstallation,whichmeanstheassociated
particletracksfortheformerwillseelongerUVexposuretimes.

Thetotalvelocitydatasetforthefourreactorloopswerelessthan120percentofthe
averagemeasuredvelocity,exceptfortwodatapointsforReactorLoopNo.8.By
comparison,thevalidationtestinstallationhadonedatapointthatdidnotmeetthis
criterion.

ThehighvelocitydatasetforReactorLoopNo.1at130ML/daywascomparabletothe
validationtestinstallation.However,theotherreactorloopsandflowconditionsshowed
highervelocityvalues.Theaveragenormalizedvelocityforthesedatapointsrangedfrom
1.07to1.13whereasthevalidationtestinstallationaveraged1.05.

Thelowvelocitydatasetforallfourreactorloopsshowedmorepointswithsubstantially
lowervelocitymeasurementsthanthevalidationtestinstallation,whichmeansthe
associatedparticletracksfortheformerwillseelongerUVexposuretimes.
Figure6comparesthecoefficientofvariationforthefourreactorloopsandallprevious
investigationsincludingphysicalmodeling,CFDmodelingandvalidationtesting.TheCoVswere
calculatedforallvelocitysamplepointsand"highvelocity"samplepointsforeachinvestigation.
Theresultsindicatethefollowing:

Velocitymeasurementsforthefourreactorloopsshowedhighervariability(CoVof8to
15percent)whenallvelocitysamplepointsareincluded.Thevariabilitywashigherfor
ReactorLoopNos.4and8,locatedattheendofthecommoninletheaderpipelinethanfor
ReactorLoopho.1,locatedatthebeginningoftheheaderpipeline.

Velocitymeasurementsforthefourreactorloopsshowedlowvariability(averageCoVof
5percent)whenlowvelocitydatapointsareexcludedfromtheanalysisacrossallfour
reactorloops.

Bycomparisonthevalidationtestinstallationshowedlowervariability(CoVof8.2percent)
whenalldatapointswereincludedintheanalysisandhighervariability(CoVof
6.5percent)whenlowvelocitydatawereexcluded.

Velocitymeasurementsforthephysicalmodelofthefull‐scaleinstallation(completedby
CDMSmithduringpreliminarydesign)comparedwellwiththefull‐scalemeasurements.
Figure 6. Comparison of Velocity Distribution Upstream of UV Reactor of Different Reactor Loops and Test Conditions ConclusionsandRecommendations
BasedontheforegoinganalysisofvelocityprofilesforthefourreactorloopsintheUV
DisinfectionFacilityandvalidationtestinstallation,wemayconcludethefollowingwithrespect
tocompliancewiththeacceptancecriteriaestablishedfortheproject:

AcceptanceCriteria1and2weremetforalltestedreactorloops,whichachievedlower
maximumvelocitydatapoints(belowthe1.2thresholdvalue)andhadalowercoefficient
ofvariationforhighvelocitysamplepointsthanthevalidationtestinstallation.

AcceptanceCriterion3wasmetforReactorLoopNos.1and4anddesignflowsof130and
170ML/day,achievingcomparable95thpercentilevelocityflowstothevalidationtest
installation.The95thpercentilevalueforReactorLoopNo.8at170ML/daywas7percent
higherthanvalidation,whichmaybeattributabletoreal‐timeflowchangesthroughthe
reactorloopduringthetest,asdiscussedearlier.
TwoadditionalpointsareworthnotingthatwillincreasetheeffectiveUVdosedeliveryduring
full‐scaleoperation:

Trojanhasincorporatedadiffuserplateimmediatelyupstreamofthelamparrayinthe
TorrentUVreactor,whichwilltendtoimproveflowdistributionacrossthereactorand
wasnotaccountedforinthevelocityprofileanalysisreportedherein(measurementstaken
upstreamoftheintegraldiffuserplantforthefull‐scaleinstallation).

Aballastpowerlevel(BPL)safetyfactorof1.06wasselectedduringcommissioning,which
reducedthetotalaccumulatedoff‐specvolumetolessthan0.02percentduringthetest
period.ThissafetyfactorwillincreasetheeffectiveUVdosedeliverytoensurethattheUV
disinfectiontargetof3‐logCryptosporidiuminactivationismetcontinuouslymetunder
variousplantflowandwaterqualityconditions.
Insummary,theanalysisofvelocityprofilemeasurementsforthefourreactorloops,asdetailed
intheADSReportandvalidationtesting,asdetailedinTrojan'svalidationtestreport,plusthe
decisiontouseaconservativeBPLsafetyfactorforroutineoperationoftheUVsystem,clearly
demonstratethattheinlethydraulicstoeachreactorloopwillprovideaUVdosedeliverythatis
equaltoorgreaterthantheUVdosedeliveredduringvalidation—akeytreatmentperformance
goalsetforthintheUVDGM.