2/22/2015
1
Caro
lloT
em
pla
teW
ate
rWave
.pptx
Application of CFD toWastewater Process Engineering
Randal W. Samstag, P.E., B.C.E.EPrincipal Technologist
Ed Wicklein, P.E.Senior Technologist
Carollo Engineers
Caro
lloT
em
pla
teW
ate
rWave
.pptx
2
“If we know what is happening withinthe vessel, then we are able to predictthe behavior of the vessel as a reactor.Though fine in principle, the attendantcomplexities make it impractical to usethis approach.” – Octave Levenspiel(1972)
Computational fluid dynamics (CFD) changesthis picture. Using CFD, we can computethree-dimensional velocity fields and followinteractions of reactants and productsthrough a tank. We can use this informationto optimize tank geometry.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
3
CFD can calculate the velocity fields.
Hydraulic Model
– Continuity (massconservation)
– Momentumtransport
– k-epsilon turbulencemodel
– Control volumesolution scheme
0i
i
X
U
t
i
jj
it
ij
ij
i FXX
U
X
P
X
UU
t
U
21
2kCt
kMbk
jk
t
j
i
i
SYPPX
k
XkU
Xk
t
)()(
S
kCPCP
kC
XXU
Xtbk
jk
t
j
i
i
2
231 )()()(
Caro
lloT
em
pla
teW
ate
rWave
.pptx
4
3D Solids Transport ModelIndividual Control Volume
2/22/2015
2
Caro
lloT
em
pla
teW
ate
rWave
.pptx
5
3D transport models can be implementedby user defined functions (UDF) in Fluentor other commercial software.
Solids transport UDF
– Solids Transport
– Vesilind settling
– Density couple
– Viscosity impact
)*exp(* CkVV os
))w-(*C-(1/w 1
))(_*132.0(^ CUDSIC
z
CV
X
C
XX
CU
t
Cs
is
t
ii
i
)(
Caro
lloT
em
pla
teW
ate
rWave
.pptx
6
3D transport models can be implementedby user defined functions (UDF) in Fluentor other commercial software.
Biokinetic Models
– ASM Models
– Advanced oxidationModels
– Disinfection models
Sobremisana, Ducoste, de los Reyes III(2011)
Caro
lloT
em
pla
teW
ate
rWave
.pptx
7
CFD is well established for analysis ofhydraulic components.
Often flows are split between paralleltreatment components as well
Caro
lloT
em
pla
teW
ate
rWave
.pptx
8
Pump Stations – Optimize IntakeHydraulics
Adverse Hydraulics:
Vortices
Pre-rotation
Turbulence
Velocity Distribution
Lead to:
Decreased capacity
Cavitation
Excessive wear
2/22/2015
3
Caro
lloT
em
pla
teW
ate
rWave
.pptx
9
Screening / Headworks
CFD can optimize design
Screen channel flowbalance
Screen flow distribution
Infer grit deposition fromvelocity profiles
Caro
lloT
em
pla
teW
ate
rWave
.pptx
10
Grit Management
Vortex grit systemefficiency is a functionof approach and exitvelocity
Aerated grit tanks requireproper sizing andbaffling to preventshort circuiting
Grit deposition can beinferred from velocityprofiles and neutraldensity particletracking
Caro
lloT
em
pla
teW
ate
rWave
.pptx
11
Flow Splitting
Flow Splitting is critical tooptimize the capacity ofparallel treatmentcomponents
CFD analysis can be rigidlid or free surface
Caro
lloT
em
pla
teW
ate
rWave
.pptx
12
Head losses through complex systems with non-uniform approach conditions can be investigated.
2/22/2015
4
Caro
lloT
em
pla
teW
ate
rWave
.pptx
13
But CFD can also be used for analysis oftransport processes.
Often flows are split between paralleltreatment components as well
Caro
lloT
em
pla
teW
ate
rWave
.pptx
14
Primary Clarification
CFD can be used toinvestigate geometricinfluences on solidsremoval.
CFD can be used toinvestigate sludgeconsolidation problems.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
15
Aeration Tanks
Multiphase modeling canbe used to investigatewater-air flows.
Dissolved air transfermodels can beincorporated.
Solids transport andbiokinetic models can beincorporated.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
16
Secondary Sedimentation
Clarifiers require effectiveinlet energy dissipation.
Baffles can aid insedimentation.
Density currents dominateflow field, therefore acustom transport model isrequired.
CFD analysis of activatedsludge sedimentation isvery well established.
2/22/2015
5
Caro
lloT
em
pla
teW
ate
rWave
.pptx
17
Disinfection - UV
Hydraulics are critical:
Flow split betweentrains
Flow distribution
Head losses
Dose models can beincorporated whendeveloping newdesigns.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
18
Mixing Systems
Mixing Used When:
Combining FluidStreams
Chemical Additions
Minimizing Stagnation
Typical Mixing Systems
Natural Diffusion
Passive Baffles
Aeration
Mechanical
Pumped Jets
Caro
lloT
em
pla
teW
ate
rWave
.pptx
19
DigestersGood mixing is critical to performance
Improved kinetics more gas productionReduces foamingEfficient mixing saves power
Caro
lloT
em
pla
teW
ate
rWave
.pptx
20
Case Studies Using Transport Modeling
Comparison of UDF transport model to out-of-the-box commercial multiphase model
Activated sludge lamella clarifiers
Use of UDF model to evaluate sedimentation inlets
Use of UDF models to evaluate activated sludgemixing
2/22/2015
6
Caro
lloT
em
pla
teW
ate
rWave
.pptx
21
Model Comparison Case Study:Transport Modeling versus Mixture Model*
Transport modeling treats concentrations as apassive scalar quantity that react and aretransported through the fluid grid.
z
CV
z
C
zx
C
xz
CV
x
CV
t
Csszsx
zx
* Wicklein and Samstag (2009)
Caro
lloT
em
pla
teW
ate
rWave
.pptx
In transport modeling the solids settlingvelocity is treated empirically.
kCs eVV 0
(Vesilind Equation)
Caro
lloT
em
pla
teW
ate
rWave
.pptx
23
In the Fluent mixture model settling isdependent on particle size.
q
sss
d
18
2
g
fv
sdrag
msss
(from Stokes Law)
Caro
lloT
em
pla
teW
ate
rWave
.pptx
24
Comparison of Field Data to MixtureModel Results for a Rectangular Tank
Field Data
Mixture model
2/22/2015
7
Caro
lloT
em
pla
teW
ate
rWave
.pptx
25
Comparison of Field and TransportModel Results for Rectangular Tank
Field Data
TransportModel
Caro
lloT
em
pla
teW
ate
rWave
.pptx
26
Comparison of Field to Mixture Model forRadial Flow (Circular) Tank.
Field Data
MixtureModel
Caro
lloT
em
pla
teW
ate
rWave
.pptx
27
Comparison of Field to UDF TransportModel (Radial Flow Tank)
Field Data
UDF Model
Caro
lloT
em
pla
teW
ate
rWave
.pptx
28
Transport versus MixtureModelingConclusions
Either transport modeling or mixture modeling canreasonably predict solids transport, if properlycalibrated.
We have a lot of data to calibrate transport modelsusing empirical settling velocity.
We have very little data to know the appropriateparticle size to use in a mixture model for activatedsludge.
2/22/2015
8
Caro
lloT
em
pla
teW
ate
rWave
.pptx
29
Sedimentation Case Study:
Activated Sludge Lamellas*
*Samstag, Wicklein, Lee (2012)
Caro
lloT
em
pla
teW
ate
rWave
.pptx
30
The Boycott Effect has been used as thebasis for the PNK theory.
Boycott (1920) observed adifference in apparentsettling rate of blood inslanted tubes.
PNK theory: Settling isenhanced by the ratio ofthe projected area ofinclined plates or tubes.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
31
Does this apply to flow-throughactivated sludge lamella clarifiers?
Coarse Grid Custom Model 2D and 3D Commercial Models
Caro
lloT
em
pla
teW
ate
rWave
.pptx
32
Results: No difference between tankswith and without lamella plates!
With lamella plates Without lamella plates
2/22/2015
9
Caro
lloT
em
pla
teW
ate
rWave
.pptx
33
Sedimentation Case Study:Clarifier Inlet ComparisonExisting Configuration
Caro
lloT
em
pla
teW
ate
rWave
.pptx
34
Clarifier Inlet ComparisonExisting Configuration
Caro
lloT
em
pla
teW
ate
rWave
.pptx
35
Alternative Inlet Configurations
Caro
lloT
em
pla
teW
ate
rWave
.pptx
36
Alternative Velocity Vector Plans
2/22/2015
10
Caro
lloT
em
pla
teW
ate
rWave
.pptx
37
Alternative Velocity Plans
Caro
lloT
em
pla
teW
ate
rWave
.pptx
38
Alternative Velocity ProfilesC
aro
lloT
em
pla
teW
ate
rWave
.pptx
39
Comparison Solids Profiles
Caro
lloT
em
pla
teW
ate
rWave
.pptx
40
Mixing Case Study:Jet mixing and aeration in a sequencing batchreactor (SBR)*
415,350 mixed tetrahedralcells
2,108,308 nodes
Inlet flow into jet nozzles
Outlet flow to pump suction
Air added as second phase
Solids transport, settling,and density impactmodeled by UDF
*Samstag, Wicklein, et al. (2012)
Mesh projected onto modelsurfaces.
2/22/2015
11
Caro
lloT
em
pla
teW
ate
rWave
.pptx
41
Velocity profiles for pumped mixing andaeration
Simulated Pumped Mixing Profile Simulated Aeration Profile
Caro
lloT
em
pla
teW
ate
rWave
.pptx
42
Comparison of pumped mix velocityprofiles for increasing jet velocities
Existing(2.5
m/secJet)
3.0m/sec
Jet
3.5m/sec
Jet
4.0m/sec
Jet
Caro
lloT
em
pla
teW
ate
rWave
.pptx
43
Comparison of solids profiles forincreasing jet velocities
Existing(2.5
m/secJet)
3.0m/sec
Jet
3.5m/sec
Jet
4.0m/sec
Jet
Caro
lloT
em
pla
teW
ate
rWave
.pptx
44
Min / Max Deviations from Average
Layer
Average TSS Concentration (mg/L)
2.5 m/sec 3.0 m/sec 3.5 m/sec 4.0 m/sec
Top 1,208 1,404 2,102 2,155
2 2,385 2,331 2,280 2,285
3 2,519 2,374 2,322 2,308
4 2,538 2,422 2,448 2,387
5 2,554 2,518 2,526 2,443
6 2,604 2,620 2,511 2,456
Bottom 3,008 2,806 2,559 2,500
Average 2,402 2,353 2,392 2,362Max DeviationFrom Average
(%) 50% 40% 12% 9%
2/22/2015
12
Caro
lloT
em
pla
teW
ate
rWave
.pptx
45
Comparison of Power Levels at DifferentJet Velocities
Jet Velocity Mix Criterion
PowerLevel
(hp/MG)
PowerLevel
(W/m3)
2.5 m/sec jetvelocity
50% MaxDeviation 39 7.7
3.0 m/sec jetvelocity
40% MaxDeviation 66 13.0
3.5 m/sec jetvelocity
12% MaxDeviation 105 20.7
4.0 m/sec jetvelocity
< 10% MaxDeviation 156 30.8
To meet a 10 percent deviation criterion would require fourtimes more power than currently installed.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
46
Comparison of Power Levels to OtherMixing Devices
Mixer Reference Mix Criterion
PowerLevel
(hp/MG)
PowerLevel
(W/m3)
4.0 m/sec jet This study< 10% MaxDeviation
156 30.8
Large Propeller inRacetrack
Carollo FieldVisit
Little MLSSseparation
5 ~1
Surface MixingImpeller
CarolloWitnessed Test
0.6 m/sec (2fps) bottomvelocity
39 7.6
Hydrofoil mixerOtun et al.(2009)
< 30% MaxDeviation
39 7.6
Hyperboloidmixer
Otun et al.(2009) < 11% Max
Deviation20 4.0
Caro
lloT
em
pla
teW
ate
rWave
.pptx
47
Frequent CFD practice for mixing is toassume neutral density.
The influence of solids on thevelocity pattern is ignored.
A velocity profile is thencalculated assuming clearwater.
It is then assumed that a givenminimum velocity (2.5 ft/min)will be sufficient to providemixing.
But it is SOLIDS that we are tryingto mix. They aren’t typicallymodeled.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
48
Comparison of density-coupled and neutraldensity simulationsDensity-coupled
Solids transport modelcalculates the local solidsconcentration based onflow regime.
The influence of the localsolids concentration onthe local density is theniteratively calculated.
This approach was verifiedby the field solids profiletest data.
2/22/2015
13
Caro
lloT
em
pla
teW
ate
rWave
.pptx
49
Comparison of density-coupled and neutraldensity simulationsNeutral Density
Solids transport modelcalculates the local solidsconcentration based onflow regime.
Influence of the local solidsconcentration on the localdensity was turned off.
This approach over-predicted measured solidsmixing.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
50
Comparison of density-coupled andneutral density simulations
Density-coupled Neutral density
Neutral density simulation dramatically over-predicts thedegree of mixing.
Caro
lloT
em
pla
teW
ate
rWave
.pptx
51
CFD and Process EngineeringConclusions
CFD is well established and important for analysisof hydraulic components.
There is growing appreciation that CFD can be apowerful tool for analysis of the impact ofgeometry and hydrodynamics on processperformance.
Results from studies of lamella settlers, radial flowclarifier inlets, and activated sludge mixing showthat CFD can establish important conclusions forprocess engineering.