2/23/2015

1

W116: Combining the Power of Whole-PlantSimulators and CFD for the Dynamic Modeling ofWWTPs

Presentation: Available Clarifier Models - Setting up aCFD model - Calibration Procedures - Data needed

Alonso Griborio, Hazen and SawyerRandal Samstag, Carollo Engineers

Outline Background

1-D models

2-D models

3-D models

Effective use of all levels of models

Model calibration and validation

Data needed

Calibration procedures

Output validation test

Case study, 1-D vs. 2-D vs. 3-D

Modeling of Clarifiers -Background

Conservation Equations

1. Continuity (Conservation of Liquid)

• All levels

2. Conservation of Momentum (F = ma)

– 2D and 3D

3. Conservation of Mass for all solids anddissolved substances

– 1D, 2D and 3D

4. Conservation of Energy (including Heat)

– 2D and 3D

2/23/2015

2

Background in ClarifierModeling

Secondary settling tank functions

• Clarification

• Thickening

• Storage

• Flocculation

Processes on SecondarySettling Tanks

• Hydrodynamics

• Settling

• Turbulence

• Sludge Rheology

• Flocculation

• Heat Exchangeand Temperature

Secondary Settling TankModels

• Surface overflow rate (early approaches)

• Box Model: 1-D Idealized Solids Flux

• 1-D Models: Drift Flux Model

• 2-D and 3-D Hydrodynamic Models

Early Approaches

• Hazen (1904)

– Discrete settling in a plug flow in the horizontalplane

• Camp-Dobbins model (1944,1946)

– Modification of Hazen model to include verticalmixing

• State Point Analysis

– Box type model that assumes the limiting flux canbe estimated based on a Vesilind type settingequation.

2/23/2015

3

Hazen

H

L

VsU

R

Removal = R/H=VsL/(UH)=Vs/SOR

State Point Analysis

SOR x (MLSS)*(1+a)

UR x XR

SORxESS

LimitingFluxHorizon

SFL

0

10

20

30

40

50

60

0

10

00

20

00

30

00

40

00

50

00

60

00

70

00

80

00

90

00

100

00

110

00

120

00

130

00

140

00

150

00

So

lid

sF

lux

(lb

/d/s

f)

Solids Concentration (mg/L)

Overflow - 4 clarif iers

Underflow - 4 clarifiers, RAS Flow = 50%

Settling Flux - SVI = 120

Settling Flux - SVI = 270

Overflow - 5 clarif iers

Underflow - 5 clarifiers, RAS Flow = 50%

1-D models• Vertical flow in

layers

– Vitasovic et al(VesilindEquation)

– Kinnear (2 phaseflow)

SOR

Vs

SOR+UFR

Xj

j+1jj-1

UFR

2-D Models• Larsen developed the first CFD type of

secondary clarifier model (1977)• LaRock; McCorquodale et al; Rodi et al

presented 2D primary and secondary clarifiermodels from 1980-2000

• Griborio and McCorquodale (2004) developed ageneral public domain SST model (2Dc) thatcouples solids and hydrodynamics, five types ofsettling, flocculation, non-Newtonian flow, andcompression rate.

2/23/2015

4

2-D

City of Windsor – physical-chemical CSO treatment.

3-D Models

• Use of commercial CFD package such asFLOW3D and FLUENT

– Richardson (2000) used FLOW3D with uncoupledsolids and hydrodynamics

– CCNY have developed a comprehensive 3Dmodel (FLUENT) that couples solids andhydrodynamics, five types of settling, flocculation,non-Newtonian flow, and compression rate. Theirmodel has been developed with an extensive fieldtesting program.

3D

CCNY – Wards Island NY

Remarks

• All of the four levels of modelling are stillin use;

• When used within their limitations thesemodel are still providing usefulinformation for the designers and/oroperators of clarifiers.

2/23/2015

5

Roles of ModelsLevel Strength Application Weakness

Hazen Simple Discrete settling (Grit) Ignoreshydrodynamics

LimitState

Simple Zone settling(SST preliminary designand operational)

Ignoreshydrodynamics

1D Computationspeed

All types of settlingincluding 2 phase flows

Ignoreshydrodynamics

2D Computationspeed comparedto 3D; runs onlaptop.

All clarifiers wherethere is a dominantflow direction; dynamicsimulations.

Ignores lateral non-uniformity in solidsand momentum.

3D Completeness ofgoverningequations; highspatialresolution.

All clarifiers. Steadystate simulations wherea dominant flowdirection cannot beassumed.

Long executiontimes; high level ofexpertise required.

The 2Dc Clarifier Model -Overview

• A quasi three-dimensional clarifier model

• Developed at the University of New Orleans (UNO)

• Based upon more than 30 years of experience onCFD modeling of clarifiers and published research

• State-of-the-Art tool that has beensuccessfully applied to projects in:

– Canada

– USA

– Japan

– Korea

– Australia Concentration (g/L): 0.01 0.02 0.05 0.11 0.23 0.52 1.14 2.50 5.50

8.5 ft14.8 ft 14 ft

Slope = 8.33%

Previous 2D hydrodynamic modelslimitations

• Settling Velocities

• Flocculation

• Compression Rate

• Temperature Simulationand Heat Exchange

Model Calibration

2/23/2015

6

Calibration or Validation?Definitions

•Calibration: Initial trials toadjust model parametersto reproduce fieldconditions (either longterm data or field testingdata).

•Validation: Tests to confirmthat a model is representingfield conditions. Forexample by independentstress tests with differentflow or settling conditions oroperating data

Input and Output Parameters forModel Calibration and Validation

•Input parameters:

• Settling velocity testparameters

• Flow measurements

• MLSS tests

• Flocculation parameters

• Fractionation

• Simulation parameters

• Others like temperature, dryfloc density, atmosphericparameters, etc.

•Output parameters:

• ESS

• Solids profile

• Velocity profile

• Dye behavior

• RAS SS

• Blanket depth

• Solids fractiondistribution

Different Levels of Calibration

•Four Levels of Calibration:

•Level 1

•Level 2

•Level 3

•Level 4

•Level 1 to Level 4decreasinguncertainty

Different Levels of CalibrationLevel 1

•Calibrate the model to historical plant data

•Minimum data needed: flows (influent andRAS), MLSS, effluent TSS and SVIs

•Settling coefficients defined based on the SVI

•Calibration normally consists in adjustingbasic settling parameters to match historicaleffluent TSS

2/23/2015

7

Different Levels of CalibrationLevel 2

•Similar to CRTC protocol (Dye testing is optional)

•Full-scale testing

•Basic settling parameters, e.g., Vo, K, SVI, are measured

•Data collected include: flows (influent and RAS), MLSS,effluent TSS, RAS TSS, sludge blanket depth, FSS andDSS

•Calibration normally consists in adjusting Takacsequation K2 or the wastewater fractionation and discretesettling velocities

Different Levels of CalibrationLevel 3

•Full-scale testing: average flow and stress testing

•Zone settling and compression rate parameters aremeasured

•Flocculation kinetics is determined

•Particle size distribution and discrete settling are estimated

•All data collected on Level 2 plus solids profiling. Timeseries of all parameters are gathered

•Calibration normally consists in adjusting particle sizedistribution and/or turbulence model parameters

Different Levels of CalibrationLevel 4

•All parameters measured during Level 3 Calibration

•Accurate measurement of particle size distribution (e.g.,image analysis, laser reflection, laser diffraction, acousticspectroscopy, etc.)

•Dye test

•Measurement of Velocity (e.g., drogue test, acousticdoppler)

•Measurement of the floc’s dry specific gravity

•Calibration normally consists on adjusting discrete settlingvelocities or turbulence model parameters

Input Parameter Tests

• Settling velocity testing

• Flow measurement

• MLSS measurement

• Density measurement: lock exchange

• Dispersed solids / flocculation tests

• Particle size distributions

2/23/2015

8

Sludge Settling Velocity Tests

•Goal:

• Establish settlingvelocity at the time offield tests

•Sensitive to:

• Column shape (Dick1975)

• Mixing intensity

• Temperature

Settling velocity modelsused in clarifier modeling

)()( min2min1 XXkXXk eeVoVs Vs = Vo e-KX

0

1

2

3

4

5

6

7

8

9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Concentration (g/L)

Settlin

gVelo

city

(m/h

)

Takacs Vesilind Field

Settling Velocity Data Fits

CKos eVV

Takacs versus FlocculationSubmodel

•Takacs Equations:•Calibration of K2•K2 is geometry specific•Cannot predict effect on ESS ofchanges in geometry•Does not explicitly includediscrete settling•Less physically based•Simpler formulation and shortercomputation time

•Flocculation Model:•Measured KA & KB.

•Measured particle fractionation.

•Can predict effect on ESS ofchanges in geometry.

•Includes discrete settling.

•More physically based.

•More complex formulation andlonger computation time.

}{ )()( min2min1 CCKCCKos eeVV

2/23/2015

9

Thickening/Compression• Two phase phenomenon

• Similar to soil consolidation

• Weight of overburden results in a piezometricgradient that causes liquid phase to migrate.

• As the solids consolidate, the water beingdisplaced impedes the solids movement.Kinnear used the Karman-Kozeny equation todescribe this; the problem is the number ofcalibration parameters needed to use thisapproach.

Sludge Compression Model(Kinnear 2002)

220

3

15

)1(

)1())(1(

S

zPog

Vs

m

gfl

for ≥g

l and f are the liquid and floc densities, is the porosity, Eg is the porosity at the gel

pointSo = 6/dp is the specific surface area,dp is the particle diameter, is the fluid dynamic viscosity,Po is an empirical coefficient, and g are the solids and gel solid fraction respectively.

Settling Velocity Models Based on a

Two-Phase Approach

XneX

XnkVs 2

41 )1(

Cho et al. (1993)

X

ekVs

nX

Cho et al. (1993) – Simplified Model

These models predict zone and compression settling but faildrastically in the discrete zone

Settling Velocity Research (McCorquodaleet al. 2004)

0.100

1.000

10.000

0 2 4 6 8 10 12 14 16 18 20

X (g/L)

ZS

V(m

/h)

Hindered Compression Measured

2/23/2015

10

Six Months of Settling Velocities

0.1

1

10

0 1000 2000 3000 4000 5000

MLSS mg/L

Vs

m/h

Zone settling and compressionrate

Zone Settling:

Vs = 10.54e-0.40X

R2

= 0.99

Compression Rate:

Vs = 3.20e-0.184X

R2

= 0.98

0.1

1

10

0 5 10 15

Sample Concentration (g/L)

Settlin

gV

elo

city

(m/h

)

Zone Settling

Compression Rate

Expon. (Zone Settling)

Expon. (Compression

Rate)

Vesilind-TakacsRAS SS

MLSS

2Dc settling velocity sub-model –Discrete settling

• Three types of flocs:– Big Flocs

(Vs ≥ 6 m/h)

– Medium Flocs(1.5 m/h ≤ Vs < 6 m/h)

– Small Flocs Procedure based on(Vs < 1.5m/h) a Flux Analysis

• Threshold for hindered settling 1000 to 1400 mg/L

• Threshold for discrete settling 500 to 650 mg/L

• Non-settleable particles FlocculatedSuspended Solids (FSS)

Direct Measurement

2Dc Clarifier model – Settlingregions based on SS concentration

Total Suspended SolidsConcentration (X) Settling Region Settling Model

X ≤ FSS* Non-settleable Vs= 0

FSS* < X ≤ Discrete Threshold Discrete settlingIndividual floc

settling velocity

Discrete Threshold <X ≤ Hindered Threshold

Flocculentsettling

Transition zone

Hindered Threshold <X ≤Compression Threshold

Hindered settlingExponential

model

X > Compression ThresholdCompression

settlingExponential

model

2/23/2015

11

The flocculation sub-modelincludes

• Aggregation of primary-dispersed particlesdue to shear induced flocculation includingflock breakup model

• Aggregation of primary-dispersed particlesdue to differential settling flocculation

• Implicitly models the filtration of particles inthe sludge blanket

Shear induced flocculation(Parker et al., 1970)

Where: X is the MLSS concentration (g/L),G the root-mean-square velocity gradient (s-1),KA a floc aggregation coefficient (L/g),KB a floc breakup rate coefficient (s

m-1),m the floc breakup rate exponent (dimensionless), andn is the primary particle number concentration (g/L).

GnXKGXKdt

dnA

mB

Calibration consists in determining KA and KB

Differential settling flocculation

122

1

2

2

1

21

211 212

3SSds VV

d

C

d

dCCk

dt

dC

Subscript 1 is for unflocculated–primary particlesand subscript 2 is for flocculated–flocs particlesC is concentrationd is cross sectional diameter is densityKds is a kinetic constant between 1 & 2Vs is the settling velocities

UNO 2Dc Clarifier Model:Calibration and Validation of

the Model

2/23/2015

12

Model calibration

• Adapting the model to reproduce fieldor experimental data:

– Identify the geometry and operational parametersto be evaluated

– Measure the settling properties of the sludge:• Discrete

• Zone

• Compression settling

– Measure the flocculation kinetic constants

– Adjust fractionation and/or compressibility

– Adjust the diffusion coefficients

Model calibration – Identify and inputthe clarifier geometry and operational

parameters

Marrero WWTP

SOR = 1.0 m/h ~ 590 gpd/ft2

MLSS = 2800 mg/L, RAS SS = 8400 mg/L

RAS= 50%

Concentration (g/L): 0.01 0.02 0.05 0.11 0.23 0.52 1.14 2.50 5.50

14 ft

Radius = 58 ft

8.5 ft14.8 ft

Discrete settling

The settling velocities of largeand medium flocs are found bydirect measurement (visualinspection) in a column batchtest using a light source, ascale and a stopwatch

Zone and compressionsettling

Xkos eVV

1 Hindered Threshold (Xh) < X ≤ Compression Threshold

Xkcs

ceVV X > Compression Threshold

2/23/2015

13

Determination of Settling Properties

0

5

10

15

20

25

30

0 20 40 60 80 100

Time (min)

Solid

s-l

iquid

inte

rface

heig

ht

(cm

)

Vs

Vs = 30e-0.45 X

R2 = 0.98

0

2

4

6

8

0 1 2 3 4 5 6

Concentration (g/L)

ZS

V(m

/h)

Field Data Expon. (Field Data)

Xk

os eVV

Measured Settling Properties

Discrete Zone

• f1 = 0.742, Vs1= 10.8 m/h

• f2 = 0.255, Vs2=3.0 m/h

• f3 = 0.003, Vs3=0.7 m/h

Hindered and Compression Zone

• Vo = 10.54 m/h, K1= 0.40 L/g

• Vc = 3.20 m/h, Kc= 0.18 L/g

• FSS* = 4.3 mg/L

Marrero WWTP SST

• MLSS = 2800 mg/L

• SOR = 1 m/h

Calibration of the flocculationsub-model

• Collect a MLSS sample (About 15.0 Liters)

• Use a six-paddle stirrer and fill each jar with 2.0 L of mixedliquor (avoiding unnecessary delays)

• Assign a flocculation time to each jar, e.g., 0, 2.5, 5,10, 20, 30 minutes

• Mix the samples at a G ofapproximately 40 s-1

• Allow the sample to settlefor 30 minutes

• Take a supernatantsample from each jar

• Measure the TSS

Calibration of the flocculationsub-model

tGXK

A

Bo

A

Bt

AeK

GKn

K

GKn

Xtkg

O eaCaC )(

Wahlberg et al. (1994) La Motta et al. (2003)

KA = 7.4 x 10-5 L/g SS

KB = 8.00 x 10-9 s

X = 2800 mg/LG = 40 s-1

0

10

20

30

40

50

60

70

0 2 4 6 8 10 12 14 16 18 20

Flocculation Time (min)

Supern

ata

nt

SS

(mg/L

)

Equation 2.39 Data

C = 4.3 + (60.7)·e-0.49728·t

C = a+(CO-a)·e-kg·t·X

2/23/2015

14

Particle Size Distributions•McCorquodale et al. developed a semi-empirical simple method to estimatefractionation and discrete settling for 3 particle sizes

•City College New York Modified the method

•Accurate methods for measuring PSD include:

• Image analysis

• Laser reflection

• Laser diffraction

• Acoustic spectroscopy

•In Situ vs. Ex Situ Techniques

Particle Size Distribution

Output Validation Tests

• Solids profile testing

• Velocity profile testing

• Dye transport testing

– RTD

– Continuous dye snapshot

• Sludge blanket monitoring

Solids Profile Measurement

•Sampling Method

• Larsen:• Kemmerer

• Crosby:• Solids Distribution

Test - Sample pumps

• Esler:• Optical device

2/23/2015

15

Solids Profile VisualizationSolids Profile Comparison to Simulation

Field Test (Crosby SD test) Simulation (2DC)

Velocity Profile Measurement

•Larsen built his ownultrasonic velocity probe

•Commercial probes: ADV

•Drogues

•Concerns:• Low velocities

• Probe sensitivity

• Difficult to hold still!

Velocity Profile Visualizations

2/23/2015

16

Dye Tests: Residence TimeDistribution Tests

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0.00 0.50 1.00 1.50 2.00 2.50

C/C

o

N = 2.3West SideSouth SideEast SideNorth SideAverage

Dye Tests: Flow Pattern DistributionTest (Crosby and Bender 1984)

Sludge Blanket Monitoring

• Dynamic monitoring of sludge blanket

• Sludge judge

• Difficulties: What is the thresholdconcentration of the “sludge blanket?

Conclusions: Output validation tests

• Solids profiles

• Relatively easy to measure

• Directly comparable to model results

• Velocity profiles

• More difficult to measure directly

• Dye tests

• Useful for flow distribution issues

• Continuous test not commonly used

• Dynamic blanket monitoring

• Useful for rough monitoring of test conditions

• Not as quantitative as solids profiles

2/23/2015

17

Case Study Number 1:Comparison of Models

• One-dimensional (1D) Model (StatePoint)

• Two-dimensional (2D) Model (UNOCFD)

• Three-dimensional (3D) Model (ZhouCFD)

1D Model (Clariflux)

• Developed by Carollo Engineers

• Solves solids flux equations based onmeasured settling velocity coefficients (orSVI)

• Calculates state point for steady stateoperation

– SOR Line

– MLSS Line

– RAS line

State Point ModelClariflux Model Results

Test Condition (33% RAS)

State Point

0

5

10

15

20

25

30

35

40

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000

Suspended Solids Conc., mg/l

So

lids

Flu

x,l

b/s

f.da

y

ThickeningFailure

2/23/2015

18

Clariflux Model ResultsIncreased RAS

State Point

0

5

10

15

20

25

30

35

40

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000

Suspended Solids Conc., mg/l

So

lids

Flu

x,lb

/sf.d

ay

ThickeningOK

2D Model – UNO Model

• Developed by Griborio, McCorquodale, andassociates at the University of New Orleans

• Two-dimensional model based on– Vorticity / stream function model

– Turbulent hydraulics

– Radial flow coordinates (axi-symmetric)

– Discrete Settling

– Hindered Settling

– Compression

– Flocculation

2D Model ResultsModel Runs

• All tests run at 3,600 mg/L

• Settling coefficients as measured in thefield

• Dynamic model runs

• 350 minutes simulation time (5.8 hours)

2D Model ResultsTest Calibration Results

Field UNO Model

2/23/2015

19

2D Model ResultsSummary of Model Runs

3D Model (Zhou CFD)

• Developed by Siping Zhou and J. A.McCorquodale

• Three-dimensional solution based on

– Control volume model

– K-epsilon turbulence model

– Generalized coordinates

– Hindered settling

– Density-coupled solids transport

– No flocculation or compression modeling

Existing Inlet

• Four openings invertical feed pipe

• No energydissipating feed-well

Prototype of Improved Inlet

2/23/2015

20

3D Model Results (SVI 110)

Optimized

Existing

3D Model ResultsInlet Improvements (SVI 190)

3D Model Results (SVI 190)

Optimized

Existing

3 D Model ResultsSummary of Model Runs

Conclusions from Modeling

• 1D model– Confirmed clarification capacity at 3.5 mgd

– Clarifiers can be RAS limited at 33%

• 2 D models– Reasonable verification of field tests

– Inboard effluent launders slightly better than end laundersand Stamford baffle

• 3D Model– Increased RAS rate without optimized inlet caused

clarification failure

– Optimized inlet can result in significant improvement withhigher RAS ratio

2/23/2015

21

Case Study 2: Comparison ofTangential Versus Puzzled Inlet

Fluent UDF Model

• Commercial CFD with k-epsilonturbulence model

• User-defined functions (UDF) addsettling velocity, flocculation, solidstransport, and density coupling

• Two or three-dimensional

• Capable of very refined grids

Fluent UDF ModelOverall Solids Profile

Tangential Inlet Puzzled Inlet

Fluent UDFModel Center-well Velocity Profiles

Tangential Inlet Puzzled Inlet

2/23/2015

22

Fluent UDF ModelInlet Velocities

Tangential Inlet Puzzled Inlet