ABE 527
Modeling of a West Lafayette chlorine disinfection
chamber at West Lafayette treatment plant
BY
Kanish Jindal
April 28, 2004
Introduction and theoretical background
West Lafayette waste water treatment plant is located at 500 south river road close to state
road IN 43. It was constructed in 1958, expanded in 1970 and had its final up-gradation in
1997. Currently it serves around 29,000 residents of West Lafayette and around 38,000
Purdue University students. The effluent is it’s treated as per EPA’s standard and then
discharged into Wabash river.
Arial view of West Lafayette Waste Water treatment plant*
(http://www.city.west-lafayette.in.us/departments/wastewater/)
Disinfection is the last process in waste water treatment process and involves the
inactivation of pathogenic microorganisms. This inactivation means biochemical alteration
of microorganism to prevent its replication or eliminate activity. Chlorination is the most
widely practiced disinfection process for water and waste water treatment. This is primarily
done in disinfection chambers where sufficient contact time is provided between treated
waste water and chlorine. The US EPA determines the effectiveness of these contactors for
disinfection by the CT method, where C is the concentration of the disinfectant at the outlet
of tank and T is the T10 value (The time required for 10% of the liquid to leave the tank, or
the time at which 90% of the liquid is retained in the tank and subject to at least disinfection
level of C). The contactor hydraulic efficiency can be measured by the ratio of T10 and
theoretical hydraulic residence time. The configuration of baffles, inlet and outlet conditions,
length and width of compartments influence this hydraulic efficiency. Computational fluid
dynamics analysis provides the better understanding fluid flow phenomena that affect
mixing, short circuiting, particle dispersion and other critical issues. It helps to eliminate
different configuration of the contactor which are inefficient and evaluate designs in terms
of mixing characteristics. It can help to reduce the cost of chlorine by increasing efficiency
of its disinfection reservoirs.
One of the greatest challenges in designing water treatment equipment is that its large size
makes it very expensive and time consuming to perform physical experiments.
Tracer test is conducted in full scale plants to calculate this residence time distribution
(RTD). The entire RTD can be used to predict the microbial inactivation in the disinfection
chamber. Another alternative of this full scale tracer test is to simulate it using CFD models
and obtain RTD which will thus same time and money. Two steps were involved in my
study. The first part was conducted at the West Lafayette Wastewater treatment plant
(WLWWTP) where tracer test was conducted using Rhodamine-wt to determine the
residence time distribution for the chamber. In the second part I simulated the tracer test
using Fluent and predicted the residence time distribution.
Objective:
The objective of this project is computational fluid dynamics(CFD) analysis of existing
chlorine contact chamber in WLWWTP using Fluent, predict its residence time distribution
and compare it with the experimental results.
Experimental Tracer Test:
During the experimental period the flow rate ranged from 7.46-14.63 MGD, and the average
flow rate during the experiment was 9.91 MGD (26.06 m3/min). Before starting experiment,
a solution of 10g of the fluorescent dye Rhodamine-wt was prepared. After that the
fluorescent dye Rhodamine-wt was added at the influent end of the chlorine contact
chamber as a pulse. The fluorometer was set up for the analysis of dye concentration.
Fluorometer readings were taken every minute over the course of tracer test, with t = 0
being defined as the time of the pulse addition. Fluorometer provided the quantification of
the concentration of dye in the contact chamber. The fluorometer data was downloaded in
real time to a laptop connected to it. The runtime of the test was recorded on a pocket
stopwatch. Samples were collected in flasks, and analyzed for the Rhodamine-wt that was
added to the system. Dye diffused and moved to the direction of flow. After 45 minute,
samples were taken for every minute at designated places. The test was conducted for 2~3
hydraulic detention times. During the experiment, flow rate as a function of time was
monitored and recorded at the times when fluorometer readings were taken. The tracer
plume was observed as it passed through the contact chamber and its behavior noted (the
observation of this fluorescent dye provided an opportunity for flow visualizationcharacteristics of the mixing behavior within the system until dye dispersed to the point
where it was no longer visible).
The picture of water treatment
Numerical Analysis:
FLUENT was used to simulate the flow through contactor. FLUENT uses finite volume
method to solve the differential equations governing fluid flow. It has in 7 in built models

In viscid

Laminar

k-epsilon

k-omega

Reynolds stress

Large eddy simulation
Following steps are followed numerical solution:

Divide of the domain into discrete control volumes using a computational grid.

Solve the governing integral equations for the conservation of mass and
momentum

Solve for energy, turbulence and chemical species if appropriate

Integration of the governing equations on the individual control volumes to
construct algebraic equations for the discrete dependent variables (``unknowns'')
such as velocities, pressure, temperature, and conserved scalars.

Linearization of the discretized equations and solution of the resultant linear
equation system to yield updated values of the dependent variables.
For my analysis I used RNG k-
Model. The RNG-based k-
turbulence model is derived
from the instantaneous Navier-Stokes equations, using a mathematical technique called
``renormalization group'' (RNG) methods. The analytical derivation results in a model with
constants different from those in the standard kfunctions in the transport equations for k and
model, and additional terms and
. Here k is the turbulent kinetic energy and
represents the turbulent energy.
The model solves separate transport equations for the turbulent velocity and length scales
through the dissipation of turbulent kinetic energy.
The first equation is turbulence kinetic energy equation and it expresses the transport of k in
terms of diffusive terms, generation terms and destruction of k due to dissipation (e) and
compressible effects. The turbulent viscosity uT is proportional to k2/e with a proportionality
constant that has the same value for all conditions. The term Ce1, which expresses the
production of energy, shows that the mean velocity gradients and buoyancy also contribute
to increase the dissipation. The term Ce2, expresses the destruction of dissipation. Under
large enough strains the normal stress u2 may become negative, being a non-real situation.
The realizable model refines the turbulence equations by ensuring that the normal stress u2 is
always positive. This is done through a different calculation of uT and adding a new term to
the dissipation equation.
Fluent allows you to choose either of two numerical methods, segregated and coupled. The
two numerical methods employ a similar discretization process (finite-volume), but the
approach used to linearize and solve the discretized equations is different. Segregated
Solution Method: Using this approach, the governing equations are solved sequentially (i.e.,
segregated from one another). Coupled Solution Method: The coupled solver solves the
governing equations of continuity, momentum, and (where appropriate) energy and species
transport simultaneously (i.e., coupled together). Governing equations for additional scalars
will be solved sequentially (i.e., segregated from one another and from the coupled set).
For this project I am using segregated solution method for solving the discretized equations.
Although the flow rate varied during the actual tracer test performed at the plant, for the
simplicity I am assuming the steady state flow conditions existed in the chamber. This
assumption is quiet reasonable as this variation in flow was less than 15%. Rhoda mine
(Inert organic dye) used in the actual experiment has been replaced by liquid water therefore
there is no chemical kinetics between the two phases.
Method:
I collected the actual plots or design drawings of the chlorine disinfection tank from
treatment plant. Tank being rectangular in shape, I used Cartesian co-ordinate system to
define various points on the tank. I used gambit as a pre-processor for modeling the tank.
The modeled was developed from top to bottom i.e. volumes were generated first by
subtracting, adding, splitting finite volumes. Volumes were added to obtain a final geometry
of the tank and then surfaces, edges and nodes were created on it. Uniform meshing was
applied to the whole domain. Hexagonal elements were used except for the corners. Due to
change in geometry wedge elements were used at the corners. Cooper meshing technique
was used which means the meshing (interval size) is defined at a surface and it swaps the
whole geometry with that mesh/grid structure. The file was exported as a mesh file for
processing. I selected segregated solver for solving the differential equations. Parameter for
model i.e. turbulence intensity and turbulence length were calculated. Appropriate boundary
conditions were applied at inlet, surface and outlet for solution and finally 1000 iterations
were used to arrive at a convergence.
Model description:
The following schematic shows the layout of contactor. Each chamber is 41.41 m in length.
The depth of water is 2.58 m and width of each chamber is 1.93 m. The effluent enters from
the inlet and Chlorine is injected at a distance from 14mts from the inlet boundary. A
hydraulic jump is provided for uniform mixing just below the injection.
For the same of simplicity I am assuming there’s turbulent zone for the first few 14 meters
in this tank. As a result of this there is uniform mixing of chlorine inside the tank. Plug flow
regime occurs beyond this hydraulic jump.
Outlet
Inlet
Point of injection
Schematic of Contactor
Outline of chlorine contact chamber
The above figure shows the outline of contactor as obtained in Gambit.
Meshing of the fluid volume:
Grid structure of the disinfection tank
Cooper meshing was employed for meshing the whole tank. An interval size of .4 was used
for meshing.
Hexagonal and wedge shape mesh elements
Inlet and outlet faces were meshed separately to ensure a good mesh density on the
boundaries. The meshed file is exported as neutral file for the analysis.
Boundary conditions:
The inlet boundary had uniform flux across the whole surface. The inlet velocity as
calculated from the flow coming in was used in the negative-x direction.
Inlet velocity = Flow coming in/Area of cross section = 0.0963m/sec
The inlet turbulence intensity I(turbulent kinetic energy) and turbulence length scale l were
calculated by:
k = kinetic energy = 0.0025 (vel) 2 m2/sec2
€ = length scale = 0.1643 (k) 3/2/ (0.1 * entrance width) m where k is the kinetic
energy as calculated in the first step.
Analysis of Tank: Processing
Fluent is very powerful in modeling fluid flow in disinfection tanks. The solution adaptive
grid capability is particularly useful for accurately predicting flow fields in regions with large
gradients. The neutral mesh file generated using gambit was imported in fluent.
Grid check is applied to see if there’s any error in geometry. The minimum volume obtained
came out to be positive. Water was used as a fluid material with a density of 998.2 kg/m3 and
viscosity of 0.001003 kg/(m.sec). For tracking the residence time again water was assumed as
an injector. Initially 1000 iterations were carried out. Convergence was achieved after 40
iterations. The following two plots show the snap shots of position of particles at two
different times.
Snap shot of particle trajectory at two different time intervals
Results
Residence time distribution
1.2
F(t) Residence time distribution
1
F(t) mins
0.8
0.6
0.4
0.2
0
0
20
40
60
80
100
120
Time (mins)
The residence time distribution of particle’s in contactor
The above graph represents the residence time distribution (F(t)). Residence time
distribution (F(t)) is the fraction of fluid elements in the effluent that reside in the
chamber less than a given time. It is a useful description of the flow characteristics.
F(t) is evaluated as the fraction of tracer mass that exits the reactor. Steady-state
hydraulics were assumed. It can be observed that, no tracer is detected in the
effluent immediately. But, as the time increases, the effluent concentration
gradually rises and the falls off after the tracer has begun to be exhausted from the
system.
From the graph, it can be observed that the ½ of the particles stay in the chamber
for a time period of around 55.5 minutes, The mean residence time distribution of
all the fluid elements of the tracer was around 55.5 mins.
The above RTD plot was obtained from the Fluent. Fluent gave a mean residence time of
27.13 min (1628 sec) for the contactor. The bends in the particle track’s can be
accounted for the bends and corners in the contactor. Six particles were traced in the
system to predict RTD. The mean residence time values are given as a numerical output
in fluent. The deviations in actual and experimental values are discussed later.
The following plot shows the velocity vector of water particles in the chlorine disinfection
chamber. It can be seen from the plot that the velocity of particles is relatively low at the
corners. This decrease in velocity at corners lead to dead zones in contact chambers and is a
matter of great concern. The dead zones can be eliminated by providing a transitional
geaomentry at the corners instead of sudden bends. The flow regime tends to deviate from
plug flow conditions due these sudden bends.
Velocity decreases at corners
Velocity in x direction
Sensitivity Analysis of parameters:
A linear sensitivity analysis was done to determine the sensitivity of turbulence length
and intensity on the RTD.
From these results, it shows that turbulence intensity and turbulent length scale are
not very sensitive to RTD but the variation in mean residence time is consistent with
the variation in turbulence intensity and turbulent length scale. Increase in turbulence
intensity resulted in increase in RTD, similarly increase in turbulent length scale resulted
in increased residence time. As the output predicted varied from the actual results the
model calibration and validations was not feasible.
Discussion
As from the velocity vector plot the velocity decreased at corners and ends which is in
consonance with the actual flow in the contactor. The results for RTD predicted from
Fluent showed some variation from the actual results. The input parameters turbulent
length scale and intensity did not significantly influenced the mean residence time of the
particles in the contactor. This variation in results may be accounted for various reasons.
The prediction on RTD depends upon geometry, boundary conditions, flow regime and
description of flow fluid. Any error in these parameters can lead to variation in RTD.
The width and depth of water were 1.93 m and 2.58 m respectively which is too narrow
resulting in significant effect of side walls and base of contactor. RNG k-€ model was
used which might not represent the true flow regime inside the contactor. The
assumptions for mixing condition of chlorine to treated influent might have caused some
deviations. Only one set of actual data was available for comparison. Any experimental
error (human and apparatus) while conducting this experiment can result in different
RTD. To conclude more research is required to adequately determine the flow
conditions inside the contactor. A better understanding of the effect of boundary
conditions, flow regime, model will help in predicting the residence time distribution.
References

Ducoste, Joel J and Peplinski Kevin Daniel. (2002) Modeling of Disinfection
Contactor of Hydraulics under Uncertainty. J of Environmental Engineering.

Metcalf and Eddy: Waste Water Engineering.

Fluent Help Manual