EPA/600/R-15/288 I April 2016
www.epa.gov/homeland-security-research
United States
Environmental Protection
Agency
Modeling and Experimental Testing
of Sediment Resuspension in Water
Distribution Storage Tanks
Office of Research and Development
National Homeland Security Research 0
-------
Executive Summary
Sediments in storage tanks have the potential to accumulate pathogens, metals, and other
hazardous materials. This report addresses the potential for sediments in storage tanks to be
transported back into water distribution systems. Computational fluid dynamics (CFD) models
were developed and three simulation studies were conducted to provide insight into sediment
resuspension processes in tanks. In addition, a pilot-scale experiment was conducted to validate
the model predictions. The results of this study highlight tank operating conditions which might
reduce resuspension and removal of sediments from tanks.
The simulation studies used a cylindrical, ground level, 11,000 m3 (3 million gallon) tank with a
single inlet/outlet as its model domain. Sediment was assumed to be distributed on the tank bottom
and was made up of 0.01, 0.1, or 1 mm diameter particles. CFD models were used to calculate
shear stresses on the tank bottom to predict if the particles would be resuspended from the tank
bottom and then entrained in the fluid flow, removed from the tank through the drain, or deposited
back on the tank bottom. Flow rates, the inlet/outlet location and diameter, filling or draining
cycles were varied in order to understand the effects of these key parameters on the resulting shear
stress and the potential for particle resuspension.
Key results of the simulation studies were as follows:
• Resuspension occurred under all operating conditions simulated; however less than 25%
of particles were resuspended and less than 1% were drained.
• Particle resuspension only occurred within a short distance of the inlet/outlet.
• Particle resuspension generally occurred immediately following the start of either the
filling or draining processes.
• Particle size, flow rate, and inlet/outlet location were found to be important factors for
particle resuspension:
o Smaller particles were more susceptible to resuspension.
o Higher flow rates yielded greater resuspension.
o Inlet/outlets located near the side wall yielded greater resuspension.
o Draining yielded greater resuspension of particles than filling.
• Particle resuspension was directly correlated to the amount of momentum flow, or jet
effect, through the inlet/outlet. Reducing the flow rate or increasing the diameter of the
inlet/outlet reduces the momentum flow and the potential for particle resuspension.
• A raised inlet pipe extending 15 cm (6 inches) above the tank bottom substantially
reduced the number of particles resuspended during filling, while a pipe height extending
30-61 cm (12-24 inches) reduced the number of particles drained from the tank.
-------
Small-scale experimental tests were performed to validate the resuspension model. Glass beads
and sand particles were placed in a 1.2 m (4 ft) diameter water-filled tank with a 2.0 cm (0.8 in)
diameter inlet/outlet to study resuspension and removal. Photos and videos were recorded before
and after each filling and draining event to determine where particles were resuspended from the
tank bottom. Tracer tests were also performed to characterize the flow patterns and velocity fields.
Finally, mitigation measures were investigated by raising the pipe inlet, which was normally flush
with the tank bottom, a short distance above the tank bottom.
Key results of the experimental tests were as follows:
• Measured and simulated velocities along tank bottom matched well in the region where
particles were resuspended.
• Resuspension of particles was only observed within ~1 cm from the inlet/outlet during
filling and draining cycles for the flow rates used in the study.
• Smaller particles and less dense particles yielded a greater resuspension.
• Model predictions of resuspension generally matched experimental data for glass beads,
and generally over predicted the amount of resuspension for silica sand. The non-
spherical shape of the sand may have reduced the amount of resuspension in the tests.
• A raised inlet height of 3-8% of the head of the water above the tank bottom was able to
completely mitigate particle movement near the inlet/outlet.
Based on the simulation and experimental studies conducted for this report, several strategies for
reducing resuspension and removal of sediments from tank bottoms were identified:
• constructing a raised inlet pipe 30 - 60 cm (12 - 24 inches) above the tank bottom,
• placing the inlet/outlet near the center of the tank rather than near the side wall,
• reducing the inlet/outlet flow rates,, and
• increasing the diameter of the inlet/outlet pipes.
-------
Acknowledgements
The National Homeland Security Research Center would like to acknowledge the following
organizations and individuals for their support in the development and/or technical review of this
report, or for sharing data used to generate this report:
Sandia National Laboratories
Clifford K. Ho, Joshua M. Christian, Eric Ching, Jason Slavin, and Jesus Ortega
EPA Office of Research and Development - National Homeland Security Research Center
Regan Murray, Jeff Szabo, Scott Minamyer, and James Goodrich
EPA Office of Research and Development - National Risk Management Research Laboratory
Lewis Rossman, Darren Lytle, and Michelle Simon
EPA Office of Water - Office of Groundwater and Drinking Water
Sean Conley, Julie Javier, and Heather Galada
Albuquerque Water Authority
German Andrade
Greater Cincinnati Water Works
Yeongho Lee
Bohannan Huston
Todd Burt
Questions concerning this document or its application should be addressed to:
Regan Murray
USEPA/NHSRC (NG 16)
26 W Martin Luther King Drive
Cincinnati OH 45268
(513)569-7031
Murray.Regan@epa.gov
-------
Disclaimer
The U.S. Environmental Protection Agency (EPA) through its Office of Research and
Development funded and collaborated in the research described here under an Inter-Agency
Agreement with the Department of Energy's Sandia National Laboratories. It has been subjected
to the Agency's review and has been approved for publication. Note that approval does not
signify that the contents necessarily reflect the views of the Agency. Mention of trade names,
products, or services does not convey official EPA approval, endorsement, or recommendation.
Sandia National Laboratories is a multi-program laboratory managed and operated by
Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the
U.S. Department of Energy's National Nuclear Security Administration under contract DE-
AC04-94AL85000.
NOTICE: This report was prepared as an account of work sponsored by an agency of the
United States Government. Neither the United States Government, nor any agency thereof, nor
any of their employees, nor any of their contractors, subcontractors, or their employees, make
any warranty, express or implied, or assume any legal liability or responsibility for the accuracy,
completeness, or usefulness of any information, apparatus, product, or process disclosed, or
represent that its use would not infringe privately owned rights. Reference herein to any specific
commercial product, process, or service by trade name, trademark, manufacturer, or otherwise,
does not necessarily constitute or imply its endorsement, recommendation, or favoring by the
United States Government, any agency thereof, or any of their contractors or subcontractors.
The views and opinions expressed herein do not necessarily state or reflect those of the United
States Government, any agency thereof, or any of their contractors.
Due to the complexity of some graphs, figures, equations, and tables, some information is not
amenable to screen readers. If you need assistance to access this information, please contact
Kathy Nickel (nickel.kathy@epa.gov).
-------
CONTENTS
Executive Summary 2
Acknowledgements 4
List of Figures 8
List of Tables 11
List of Acronyms and Abbreviations 12
List of Parameters and Variables 13
1. Introduction and Background 14
1.1. Type and Purpose of Water Storage Tanks 14
1.2. Typical Tank Operating Conditions 15
1.3. Water Quality Problems in Storage Tanks 16
1.4. Potential Human Health Impacts 16
1.5. Sediment Accumulation in Storage Tanks 17
1.6. Major Assumptions Used in this Report 18
2. Modeling 18
2.1. Hydraulic Model 18
2.2. Sediment Resuspension Models 20
2.2.1. Shields Model 20
2.2.2. Beheshti Model 22
2.2.3. Implementation of Sediment Suspension Models in CFD Model 23
2.3. Operational Study of Particle Movement during Filling and Draining Cycles 25
2.3.1. Geometric Configuration and Flow Parameters 25
2.3.2. Results of Operational Study 26
2.4. Parametric Study 36
2.4.1. Parametric Study Approach 36
2.4.2. Results of Parametric Study 38
2.5. Mitigation Measures 44
3. Experimental Testing 49
3.1. Physical Test Description 50
3.2. Test Results 57
3.2.1. Tracer Tests 57
3.2.2. Particle Resuspension Tests 59
3.3. Impact of Raised Inlet/Outlet on Particle Resuspension 66
4. Summary and Conclusions 68
4.1. Results of Operational Study 68
4.2. Results of Parametric Study 69
4.3. Results of Mitigation Measures 69
4.4. Results of Testing 69
-------
References 71
Appendix A: Hydraulic Model Evaluation 74
Appendix B: Particle Size Distribution 83
Appendix C: Surface Tension Model for Particles at the Air-Water Interface 88
Appendix D: Justification for Using a 2D Axisymmetric Model 91
Appendix E: Additional Results from Parametric Analysis 95
-------
List of Figures
Figure 1. Construction of an 11,000 m3 (3 million gallon) water distribution storage tank
in Albuquerque, NM 15
Figure 2. Schematic of the water distribution tank configuration used in the numerical
simulations 19
Figure 3. Graphical representation of the Shields model in which the dimensionless
critical shear stress, re * , is plotted as a function of the grain Reynolds
number, R * 21
Figure 4. Graphical representation of the Beheshti model 23
Figure 5. Mesh of 2D axisymmetric model used in operational study. The dashed outline
in the schematic on top shows the location of the computational domain 26
Figure 6. Flow pathlines colored by velocity magnitude (m/s) during filling and draining
cycles in high flow rate case (0.631 nvVs or 10,000 gpm) 27
Figure 7. Flow pathlines colored by velocity magnitude (m/s) during filling and draining
cycles in low flow rate case (0.215 m3/s or 3,400 gpm) 27
Figure 8. Particle traces during filling and draining cycles in high flow rate case 29
Figure 9. Particle traces during filling and draining cycles in low flow rate case 30
Figure 10. Pathlines and particle traces near the transmission line during draining 31
Figure 11. Distribution of particles deposited on tank bottom at end of fill/drain cycle 33
Figure 12. Initial positions of drained particles 33
Figure 13. Initial positions of particles entrained in flow at end of cycle 34
Figure 14. Final radial position vs. initial radial position of all particles entrained in flow
at end of cycle 34
Figure 15. Final radial position vs. initial radial position of all particles deposited on tank
bottom at end of cycle. The black line represents particles that remained
deposited throughout the entire cycle 35
Figure 16. Distribution of depositions overtime 35
Figure 17. Distribution of particle resuspensions overtime 36
Figure 18. Distribution of particle drains overtime 36
Figure 19. Half-symmetry tank domain used for the parametric study 37
Figure 20. Plan view of the tank bottom showing area around inlet/outlet line susceptible
to particle resuspension (shown in red) for the 24 inch line, center location,
high flow rate case during draining for different particle diameters of 1 mm
(left), 0.1 mm (center), and 0.01 mm (right) 39
Figure 21. Fraction of bottom wall area susceptible to particle resuspension as a function
of various parameters during filling 40
Figure 22. Fraction of bottom wall area susceptible to particle resuspension as a function
of various parameters during draining 40
Figure 23. Pareto charts of the standardized effects showing the relative importance of
each factor (and interactions) on particle resuspension during filling (left) and
draining (right) 41
Figure 24. Fraction of wall area susceptible to particle resuspension vs. momentum flux
for filling at the center line location for three particle diameters 42
Figure 25. Fraction of wall area susceptible to particle resuspension vs. momentum flux
for filling at the near wall line location for three particle diameters 43
-------
Figure 26. Fraction of wall area susceptible to particle resuspension vs. momentum flux
for draining at the center line location for three particle diameters 43
Figure 27. Fraction of wall area susceptible to particle resuspension vs. momentum flux
for draining at the near wall line location for three particle diameters 44
Figure 28. Raised inlet/outlet line (silt ring) installed by Bohannan Huston in an 11,000
m3 (3 million gallon) water tank installed in Albuquerque, NM 44
Figure 29. Filling at 200 seconds - pathlines colored by velocity magnitude 46
Figure 30. Draining at 200 seconds - pathlines colored by velocity magnitude 46
Figure 31. Filling at 200 seconds - variation of bottom wall shear stress with radial
distance from inlet line 47
Figure 32. Draining at 200 seconds - variation of bottom wall shear stress with radial
distance from inlet line 47
Figure 33. Percentage of resuspended particles as a function of raised inlet line height
after 200 seconds of filling 48
Figure 34. Percentage of particles drained as a function of raised inlet line height after
200 seconds of draining 48
Figure 35. Schematic (top) and photograph (bottom) of test apparatus 50
Figure 36. Top view of the bottom of the experimental tank 51
Figure 37. Cylinder used to restrict particles during particle emplacement 52
Figure 38. Valves used to control filling and draining 53
Figure 39. Drained particles collected in sieve 53
Figure 40. Dried particles were sorted by size using a sieve shaker (Dual H-4325) 54
Figure 41. Dried particles from sieve were poured into containers (left) and weighed
using a Scientech ZSA 210 scale (right) 54
Figure 42. Flow meter used to measure fill rate during filling 55
Figure 43. Tracer testing during draining. Note: Each ring is one centimeter apart 56
Figure 44. Example of tracer testing during filling. Numbers on the vertical plug are in
inches 56
Figure 45. Measured and simulated velocity along the tank bottom during draining as a
function of radial distance from the center of the drain 58
Figure 46. Example of photos of 0.853 - 1.68 mm silica sand particles on the tank
bottom near the inlet/drain 60
Figure 47. Example of photos of 1 mm glass bead particles on the tank bottom near the
inlet/drain 60
Figure 48. Average radial extent of resuspension from the edge of the drain for all tests.
Error bars represent one standard deviation about the mean 62
Figure 49. Mass fraction histogram and cumulative distribution (CDF) for #12 - #20
silica sand from manufacturer 63
Figure 50. Histogram and cumulative distribution function (CDF) for the mass fraction of
#12 - #20 silica sand collected after draining test 63
Figure 51. Draining scenarios: comparison of simulated (Solidworks™ software) and
experimental average radial extents of particle resuspension. Left: silica sand.
Right: glass beads. Error bars represent one standard deviation about the
mean 66
Figure 52. Draining scenarios: comparison of simulated (ANSYS Fluent®) and
experimental average radial extents of particle resuspension. Left: silica sand.
-------
Right: glass beads. Error bars represent one standard deviation about the
mean 66
Figure 53. Extended line installed above drain 67
Figure 54. Tracer concentrations from Roberts et al. experiment ( [6], Fig. 3.9) with 2.16
m/s inflow jet on left 78
Figure 55. Experimental (left) and simulated (right) contours of velocity magnitude in a
jet-stirred mixing tank (from Fig. 7 in [42]) 79
Figure 56. Schematic of jet penetration into quiescent fluid [40], the jet opening angle of
11.8° is universal regardless of inlet velocity and line diameter 80
Figure 57. 3D CFD jet velocity (m/s) in center line tank with inlet velocity of 0.33 m/s 80
Figure 58. Jet velocity (m/s) profile at quarter height of tank along radial distance of tank
for 0.33 m/s jet 81
Figure 59. Shear stress profiles along bottom of tank with center line 82
Figure 60. Sediment particles from the Colorado Tower, Columbus, Ohio. The sample
was sieved and weighed to develop a sediment weight distribution 83
Figure 61. Histogram of mass fraction against particle diameter 84
Figure 62. Complementary cumulative distribution function of particle diameter based on
number fraction 86
Figure 63. Bar graph displaying initial distribution of particles along bottom wall 87
Figure 64. Snapshot of model displaying particles suspended above the water surface 88
Figure 65. Schematic of surface tension forces on spherical particle at water/air interface 89
Figure 66. Contour map of air volume fraction illustrating that cells at the air-water
interface are not strictly water or strictly air 90
Figure 67. Comparison of shear stresses on tank bottom between a2D and 3D model 92
Figure 68. Velocity pathlines at 90 seconds for 3D Model 93
Figure 69. Velocity pathlines at 90 seconds for 2D Model 94
10
-------
List of Tables
Table 1. Range of storage facility inlet/outlet diameters and velocities (median values in
parenthesis) 16
Table 2. Tank and flow parameters used in operational study 25
Table 3. Summary of operational study results 28
Table 4. Experimental design matrix for cases simulated in the parametric study 38
Table 5. Raised inlet line cases run in the Fluent® software. In all cases, the flow
duration was 200 seconds and the inlet/outlet line velocity was 2.162 m/s,
yielding a flow rate of 0.631 nvVs (10,000 gpm) 45
Table 6. Flow parameters for simulation draining from 0.3 to 0.15 m (12 to 6 inches) 59
Table 7. Draining test results.* 59
Table 8. Filling test results.* 61
Table 9. Fill, drain, fill, drain results.* 61
Table 10. Draining scenarios: experimental radial extent of resuspension compared with
modeled radial extent of resuspension. Shear stresses for modeled results
obtained using the Solidworks™ software 64
Table 11. Draining scenarios: experimental radial extent of resuspension compared with
modeled radial extent of resuspension 64
Table 12. Filling scenarios: experimental radial extent of resuspension compared with
modeled radial extent of resuspension 65
Table 13. Extended line draining results 67
Table 14. Extended line filling results 68
Table 15. Comparison of tracer concentrations as a function of time during injection into
a water-filled tank using two different turbulence models 75
Table 16. Diameter range and mass fraction of each particle bin 83
Table 17. Particle diameter range, mass fraction, and number fraction of each bin 85
Table 18. Probability of occurrence of a particle with a diameter greater than or equal to
the specified diameter. The displayed diameters are the endpoints of each bin
(Table 16) 85
Table 19. Tank model and flow rate used in the 2D and 3D comparison. These values
were used in the actual operational study as well 91
Table 20. Summary of simulated regions susceptible to particle resuspension (shown in
red) based on bottom-wall shear stress for the different parametric cases 95
11
-------
List of Acronyms and Abbreviations
CCDF complementary cumulative distribution function
CDF cumulative distribution function
CFD computational fluid dynamics
EPA U.S. Environmental Protection Agency
gpm gallons per minute
NHSRC National Homeland Security Research Center
RANS Reynolds-averaged Navier Stokes
SST shear stress transport
UDF user defined function
US United States
VOF volume of fluid
12
-------
List of Parameters and Variables
Diameter of the inlet pipe (m)
d-max.i Maximum diameter of all particles in bin /' (m)
dmin,i Minimum diameter of all particles in bin /' (m)
drep,i Diameter of representative particle in bin / (m)
D Particle diameter (m)
D* Dimensionless grain diameter of a particle
^nominal Nominal diameter of particle (m)
g Gravitational constant (m/s2)
mf j Mass fraction of particle diameter bin /' (kg)
mj Mass of all particles in bin / (kg)
mmin,i Minimum mass of all particles in bin /' (kg)
mmax,i Maximum mass of all particles in bin /' (kg)
mrep,i Unit mass of representative particle in bin / (kg)
mtot Total mass of all particles (kg)
munit,i Unit mass of representative particle in bin /' (kg)
JVj Number of particles in bin/'
JVtot Total number of all particles
Nfi Number fraction of bin /'
Q Flow rate (m3/s)
/?* Grain Reynolds number
rdrain Radius of the drain line (m)
^extent Radial extent of particle emplacement (m)
Sf Corey Shape Factor
ws Settling velocity (m/s)
it* Shear velocity (m/s)
v Kinematic viscosity of water (m2/s)
p, pp Particle density (kg/m3)
ps Solid density (kg/m3)
Pbuik Particle bulk density (kg/m3)
pw Water density (kg/m3)
TC Dimensionless critical shear stress
T* Dimensionless shear stress of tank bottom
T Shear stress along tank bottom (Pa)
13
-------
1. Introduction and Background
Storage of treated drinking water serves several important functions. Storage facilities are
designed to help maintain system pressure, supply water for typical and emergency use, and
improve stability in treatment processes and pumping rates. However, storage of treated water can
diminish water quality and increase risks to human health. In particular, storage tanks have been
implicated in several waterborne disease outbreaks [20].
This report addresses sediments in storage tanks and the potential for the transport of these
sediments back into water distribution systems, possibly posing a human health risk. The objective
of this study is to better understand the tank operating conditions under which sediments might be
resuspended from tank bottoms and transported back into water distribution systems. To this end,
several computational fluid dynamics (CFD) models were developed and three simulation studies
conducted to provide insights into sediment resuspension processes in tanks. In addition, a pilot-
scale experiment was conducted to validate the model predictions. The results of this study
highlight tank operating conditions which might reduce resuspension and removal of sediments
from tanks.
CFD models can provide detailed simulations of flow fields and hydraulic behavior in large and
complex domains. CFD models are used to better understand important parameters and processes
in complex systems, and to optimize performance of these systems at specific sites. Previous
studies have used CFD models and experiments to investigate mixing characteristics in water tanks
[1-6]. CFD models have also been used to model sediment deposition, resuspension, and/or
transport in sewer systems, wastewater drainage channels, and rivers [7-16]. However, little
research has been performed to model and characterize sediment resuspension and transport in
water distribution storage tanks. This study begins to fill this gap.
This report is organized as follows. In the rest of this section, background information on water
storage tanks, typical tank operating conditions, common water quality problems, and tank
sediments is presented. In Section 2, the CFD modeling methodology and results are presented.
In Section 3, the experimental study and results are presented. Finally, conclusions are
summarized in Section 4.
1.1. Type and Purpose of Water Storage Tanks
There are many types of finished water storage facilities in use around the United States (US),
including elevated tanks, ground level tanks, standpipes, buried tanks, and hydropneumatic tanks.
The EPA 2006 Community Water System Survey of more than 1,300 US water systems estimated
that more than half of drinking water systems use elevated tanks (in which the tank bottom is above
ground level), about 38% of water systems use ground level tanks (in which the tank bottom is at
ground level), about 20% of systems use standpipes (which are ground tanks that are taller than
they are wide), 16% use buried tanks (in which some or all of the tank is below ground level), and
about 9% use hydropneumatic (or pressurized) tanks [17]. The average storage capacity of a water
system varies widely depending on the size of the population served, from an average of about
1,500 m3 (0.4 million gallons) for smaller sized water systems to an average of 560,000 m3 (149
million gallons) for systems serving more than 500,000 customers [17].
14
-------
Storage facilities are used to achieve many functions in water distribution systems beyond storing
water for customer use. Storage helps to provide adequate supplies for emergencies, such as fire-
fighting, power outages, main breaks, or other system failures. The elevation of water levels in
storage tanks helps to maintain pressures in the water distribution system and minimize
fluctuations caused by changes in customer demands or system operating conditions. Filling and
draining storage tanks on a regular basis throughout the day helps to equalize water flows in a
distribution system, allowing water treatment processes to proceed at a steady rate while still
meeting varying customer demands [18].
1.2. Typical Tank Operating Conditions
For this report, typical tank operating conditions for two US cities were determined through
personal communication with staff of Albuquerque Water Authority and Greater Cincinnati Water
Works. For these utilities, individual tank volumes vary from 1,900 m3 to 11,000 m3 (0.5 to 3
million gallons), while reservoir (i.e., very large ground level tank) volumes vary from 7,600 m3
to 300,000 m3 (2 to 80 million gallons). Tank diameters range from 6 to 37 m (20 to 120 feet) and
reservoir diameters from 30 to 55 m (100 to 180 feet). Figure 1 shows a new ground level tank
built in Albuquerque, New Mexico, in 2014.
Figure 1. Construction of an 11,000 m3 (3 million gallon) water distribution storage tank
in Albuquerque, NM.
For both utilities, tanks typically follow diurnal cycles of filling and draining, with draining
occurring most often during the day and filling occurring during the night when electricity needed
to pump water into tanks is less expensive. The length of fill cycles varies from 3 to 14 hours, and
drain cycles from 3 to 15 hours. Water levels typically are kept within 30-95% of tank capacity in
order to maintain adequate supplies of water for emergencies such as fires. Water levels in tanks
are always changing; in other words, the tank is always filling or draining.
Tank inlets bring water into tanks while tank outlets are used to drain water from tanks; often, a
single pipe serves as the inlet and outlet. The location of tank inlets and outlets vary significantly
from being located at the center of the tank bottom to the side of tank wall. Inflow and outflow
diameters and velocities are shown in Table 1.
15
-------
Table 1. Range of storage facility inlet/outlet diameters and velocities (median values in
parenthesis)
Tanks
Inlet diameter Outlet diameter Inlet velocity/
Flow Rate
20-76 cm
(61 cm)
8-30 in
36-61 cm
(41 cm)
14-24 inches
0.03-2.7 m/s
(0.24m/sor17m3/s)
0.1-9 ft/s
Outlet velocity/
Flow Rate
0.03-1.2 m/s
(0.27 m/s or 8.7 m3/s)
0.1-4 ft/s
Reservoirs 41-122 cm
(69 cm)
16-48 in
41-122 cm
(69 cm)
16-48 inches
0.03-3.0 m/s
(1.2m/sor100m3/s)
0.1-10 ft/s
0.15-2.4 m/s
(0.79 m/s or 72 m3/s)
0.5-8 ft/s
1.3. Water Quality Problems in Storage Tanks
By storing water for periods of time, the quality of the water in finished water storage tanks can
decrease. Over time, chlorine or other disinfectant residuals decrease, disinfection byproducts can
form, and taste and odor issues can develop. Other potential chemical problems in tanks include
an increase in pH, corrosion of tank materials, and precipitation and settling of iron and manganese
[18]. Microbiological problems can include or result from bacterial regrowth, nitrification,
sediment buildup, an increase in temperature, and introduction of worms, insects, bird droppings,
and other materials through vents, openings, and breaches in tank covers [18, 19]. Sediments can
cause water quality problems by increasing disinfectant demand, microbial growth, disinfection
by-product formation, and turbidity [20].
These water quality problems can be attributed to multiple factors. Long residence times in tanks
contributes to these chemical and microbiological problems. Poor mixing of water in tanks can
further exacerbate water quality problems [20-22]. Moreover, inadequate maintenance and
cleaning of tanks can lead to the introduction of contaminants or can facilitate the growth of
microbial contaminants. EPA and industry guidance recommend comprehensive inspection and
cleaning every 3 to 5 years. However, respondents to the 2006 Community Water System Survey
reported that many systems meet this goal, but some systems reported long periods without
comprehensive inspection and cleaning. In the survey, the average number of years reported
between cleaning tanks was 6.5 [17] during which contaminants can accumulate and potentially
pose human health risks.
Fortunately, steps can be taken to improve water quality in storage tanks. Water quality can be
improved by increasing mixing, reducing residence time, monitoring water quality on a routine
basis, inspecting, maintaining and cleaning tanks regularly, modifying inlet/outlet locations, and
optimizing operations by modeling retention times and mixing characteristics. More than half of
systems using at least one of these practices to maintain water quality in storage tanks [17].
1.4. Potential Human Health Impacts
Storage tanks are associated with a range of human health risks resulting from exposure to
waterborne pathogens, contamination due to breaches of tank integrity or contaminants generated
inside tanks from corrosion or disinfectant byproduct reactions. For example, a 1993 outbreak of
16
-------
Salmonella typhurium in Gideon Missouri resulted from bird droppings in a covered storage tank;
more than 600 residents became ill resulting in 7 deaths [23].
Similarly, a 2008 outbreak in Alamosa Colorado was attributed to animal waste contamination of
a ground level storage tank, resulting in 442 reported illnesses and one death [24]. Significant
amounts of sediment described as "black and tarry" or "sandy" were removed from Alamosa's
storage tanks during cleaning and disinfection processes. The last inspection of the implicated
tank had been conducted 11 years earlier; the report noted about an inch of sediment accumulated
at the bottom and recommended cleaning the tank every 3-5 years. However, the utility did not
appear to have a regular program for flushing, disinfecting or removing sediment from tanks.
Craun et al. [25] reported that 79 of 780 waterborne disease outbreaks recorded in the US between
1971 and 2006, about 10%, can be attributed to distribution system deficiencies. These
deficiencies include contamination of storage tanks, contamination of water mains, cross-
connections, and more. Of 33 waterborne disease outbreaks recorded in 2009-2010, there were
1040 related cases of illness [26]. While only a small number (12%) of the outbreaks were
attributed to distribution system deficiencies, most (68%) illnesses resulted from distribution
system deficiencies. In other words, the number of people who became ill was much higher for
distribution system problems, including storage tank problems, than for outbreaks resulting from
other types of water system deficiencies.
1.5. Sediment Accumulation in Storage Tanks
Sediments can be introduced into tanks through accidental contamination from the outside
environment due to improper tank maintenance, chemical precipitation, corrosion of tank
materials, or from the source water. Often, sediments accumulate on tank bottoms as a result of
low velocity zones. One study of three tanks that had not been cleaned for more than 7 years found
10-71 cm (4-28 inches) of sediments on the bottom of the tanks [22].
To help expand understanding of storage tank sediments, their potential for accumulating
contaminants and their risks to human health, a recent study collected sediments from 25 storage
tanks at 18 drinking water systems in 12 different US states [27]. Sediment samples were collected
from 3-5 locations within each tank. All of the tanks contained finished drinking water that
originated from ground or surface water sources. The sediments were characterized by total
organic carbon, organic matter, sand, silt, and clay content, grain size, pH, cation and anion
exchange capacity, bulk and particle density, and porosity. In addition, the chemical composition
of the sediments were analyzed; for example, in one sediment sample, iron, zinc, aluminum,
phosphorous, magnesium, silicon, and manganese were detected [28].
As part of this study, Lu et al. [29] studied 87 sediment samples from 18 US locations and analyzed
for a range of potential microbial pathogens. The study looked for both enteric pathogens present
in animal excreta as well as opportunistic pathogens that are known to cause health impacts from
drinking water exposure. Several opportunistic pathogens were detected: Mycobacterium species
were found in 89.9% of samples, Legionella species in 66.7%, Acanthamoeba species in 38.9%,
Pseudomonas aeruginosa in 22.2% and several other pathogens occurred in 5% or fewer samples.
In contrast, enteric pathogens like Escherichia coll O157, Salmonella, Giardia and
Cryptosporidium species were undetected, although Campylobacter was found in trace amounts.
17
-------
This study shows that opportunistic pathogens are present within storage tank sediments and raises
concerns about potential human health risks associated with tank sediments.
Another EPA study tested the adherence of several contaminants to tank sediments: nonradioactive
cesium, Cs-133 (a surrogate for radioactive Cs-137), lindane, an organochlorine pesticide,
Escherichia coli, and the attenuated strain Bacillus anthracis Sterne (a surrogate for pathogenic
Bacillus anthracis) [27]). All of the contaminants adhered to sediments to some degree: cesium
adherence ranged from 5-88%, lindane from 7-88%, E. coli from 42-100%, and B. anthracis from
31-100%. This study suggests that when sediments are present in tanks, both chemical and
biological contaminants are likely to adhere to sediments.
1.6. Major Assumptions Used in this Report
In the rest of this report, for both the CFD modeling studies and the experimental study, the starting
premise is that sediments are present at the bottom of the studied tanks. As such, this report does
not address the issues of how sediments accumulate on tank bottoms or how to prevent sediment
accumulation. From this assumption, this report focuses on understanding how and when
sediments are resuspended from the bottom of tanks, how they are transported within the tank, and
how sediments are removed from the tank back into a water distribution system.
For the CFD modeling studies, sediments are modeled as individual spherical particles of a certain
size and density. Chemical and biological reactions between particles are ignored, such as
adsorption or biofilm development.
A single tank configuration is studied for this report - a cylindrical ground level tank with a single
inlet/outlet pipe located at the bottom center of the tank and a storage volume of 11,400 m3 (3
million gallons). Many other tank types, shapes, configurations and sizes are used by drinking
water systems around the US, and while similar sediment transport processes and behavior may
be inferred from the current study, an accurate assessment would require further studies of these
alternate configurations.
2. Modeling
To understand sediment resuspension in tanks and the potential for transport of sediments back
into water distribution systems, three CFD modeling studies were performed: (1) an operational
study to determine the temporal and spatial transport and distribution of sediment particles during
both filling and draining cycles; (2) a parametric analysis to determine the impact of various
parameters and processes (inlet/outlet location and diameter, flow rate, particle size, and filling
vs. draining cycles) on the resulting shear stress and potential for particle resuspension along the
bottom wall; and (3) a study to determine the effectiveness of a raised inlet/outlet line (just above
the tank floor) in mitigating particle resuspension and removal.
2.1. Hydraulic Model
A hydraulic model for turbulent flow in a water distribution storage tank was developed using the
ANSYS® Fluent® CFD™ software (ANSYS, Inc., Canonsburg, PA). ANSYS® Fluent® is a
CFD code that can simulate single- and multi-phase flow phenomena in complex geometries.
18
-------
Fluent® has been validated in various applications relevant to water flow in tanks and used in
numerous simulations across all industries [7, 11, 13, 14, 30-32]. The Fluent® software has
advanced capabilities for simulating the flow paths of fluids within complex domains. The model
domain for this study was assumed to be an 11,000 m3 (3 million gallon) cylindrical, ground level
tank with a single inlet/outlet line at the bottom of the tank located either at the center or near the
wall (Figure 2).
For this study, the hydraulic model needed to capture typical dynamics observed in tanks: the
turbulent movement of water entering and exiting the tank, the complex flow patterns and eddies
created in the tank, and the changing levels of the water surface. In addition, in order to capture
the onset of particle movement, shear stresses along the tank bottom need to be predicted by the
model.
Within the Fluent® software, there are several turbulence models that can be used. Data from the
literature on tank flows and mixing tests were used to determine the most appropriate turbulence
model for this study (see Appendix A for additional details). Based on these verifications, the
Fluent® software's k-co Shear-Stress Transport (SST) turbulence model was determined to be the
best suited for this study. The Fluent® software solves the Reynolds-Averaged Navier-Stokes
(RANS) equations to determine for time-averaged turbulent flows. The SST model uses an
isotropic eddy viscosity value to solve for the Reynolds stress term in the RANS. The SST model
is capable of solving turbulence equations in the near-wall region as well as in the free-stream
region. The model was validated against experimental data found in the literature (see Appendix
A). The shear stress along the bottom wall, which is the critical parameter in determining particle
resuspension, was calculated by the hydraulic model.
The Fluent® software's two-phase Volume of Fluid (VOF) model was used to simulate the liquid
flow patterns in the tank and the movement of the top surface of the water during filling and
draining. The VOF model simulates two or more immiscible fluids (in this case, air and water) by
solving the fluid momentum equations and tracking the volume fraction of each fluid separately
in the computational domain. The VOF is often used to simulate liquid moving into a large volume
of air space (e.g., water flowing from a hose, motion of liquid after a dam break).
9.9 m (32 ft) height
4.9m (16 ft) initial
water level
38.4m (126 ft) tank
Figure 2. Schematic of the water distribution tank configuration used in the numerical
simulations. The inlet/outlet line (61 cm (24") or 91 cm (36") diameter) was located flush
with the bottom of the tank either in the center or near the side wall.
19
-------
2.2. Sediment Resuspension Models
With no existing mathematical model for sediment resuspension in water tanks, a literature review
was conducted to evaluate models developed for other applications. Beheshti and Ataie-Ashtiani
[33] summarize a number of models developed to predict sediment resuspension; they also derived
their own model - referred to hereafter as the Beheshti model. This section summarizes the criteria
and methods used in the Shields and Beheshti models, the two major approaches relevant to this
study, and describes how these methods were applied to simulate sediment resuspension in water
tanks.
2.2.1. Shields Model
For many decades, the Shields model has been used to predict the onset of motion of sediment
solids in rivers, open-water channels, and sewage systems [8-11,15, 16, 34-37]. The model reflects
the balance between the hydrodynamic forces of drag and lift that induce motion on a particle and
the force of submerged weight which resists motion on a particle. If the hydrodynamic forces are
greater than the forces of weight on a particle, then a critical threshold for movement has been
exceeded. In the Shields model, if the calculated dimensionless shear stress of the tank bottom, T*,
is greater than the dimensionless critical shear stress, then particle suspension will occur.
The Shields Model can be represented graphically by the Shields diagram, in which the
dimensionless critical shear stress, T*C, is plotted as a function of the grain Reynolds number, /?*
(Figure 3). Represented graphically, motion is induced if the dimensionless shear stress at a given
grain Reynolds number is above the Shields curve; conversely, particles will remain stationary if
the dimensionless shear stress is below the Shields curve. As shown in Figure 3, at low grain
Reynolds numbers, the dimensionless shear stress must be high in order to induce motion of the
particles. The Shields diagram is empirically based on a variety of experiments and materials (e.g.,
sand, minerals, glass beads, and steel shot) and should be generally applicable to flow in water
storage tanks. However, the model applies to non-cohesive granular sediment and particles, not
clays or muds that have electrostatic interactions. The equations do not account for armoring, in
which smaller particles may be trapped and protected under a layer of larger particles or biofilm.
Also, the model assumes uniformly shaped and uniformly sized particles, and does not account for
the random/transient nature of incipient motion.
20
-------
Shields Diagram
«
£
b.
ra
0.1
E
5
0.01
0.1
10
Grain Reynolds Number, R*
100
1000
Figure 3. Graphical representation of the Shields model in which the dimensionless
critical shear stress, T* , is plotted as a function of the grain Reynolds number, R*.
The Shields model defines the dimensionless shear stress, T*, the dimensionless grain Reynolds
number, /?*, and the shear velocity, it*, (in units of m/s) as follows:
T =
(1)
(2)
T
(Pw
(3)
where T is the shear stress on the bottom wall (Pa), ps and pw are the solid and liquid densities
(kg/m3), g is acceleration due to gravity (9.81 m/s2), D is the particle diameter (m), and v is the
kinematic viscosity of the liquid (m2/s). The shear stress at the tank bottom, T, is used to find the
grain Reynolds number, /?*, and the dimensionless shear stress of the tank bottom, T*. The
dimensionless critical shear stress, %*c, plotted in Figure 3 is a function of the grain Reynolds
number, /?*. Typically, in drinking water storage tanks, particle diameters would vary from 0.01
- 1 mm (see Appendix B), shear stresses from 0.1 to 10 Pa, and grain Reynolds numbers from 0.1
to 100.
The dimensionless critical shear stress, i*c, is represented by the piecewise function in Figure 3
according to the following equation for different values of the grain Reynolds number [38]:
21
-------
For/?, < 1.47, T* = .1166
1-.977482
(4)
For 1.47 < R, < 10, y = -.9078950 - 1.2326090 • x + .7298640 • x2
where y = log(jc) and x = \og(R*)
For 10 < R, < 400, T* = .0227 • fl;1568
-.0772426 -x3
For R, > 400,
T* = .06
2.2.2. Beheshti Model
Beheshti and Ataie-Ashtiani [33] performed more recent studies with new data sets and concluded
that the Shields method led to discrepancies because of the inability to address particle shape,
randomness of entrainment, and the difficult with defining criteria for incipient motion. In the
Beheshti model, the use of a dimensionless movability number as a function of dimensionless grain
diameter was found to provide better matches to the data. They state that the movability number,
which includes the settling velocity, provides an implicit inclusion of shape effects in the threshold
determination. The dimensionless grain diameter of a particle, /)*, is defined as
D.=D
PP
-Pw\
1/3
(5)
where pp is the particle density (kg/m3). The critical movability number of a particle is a function
of D* and represents the threshold of particle resuspension. It can be represented by the following
piecewise function:
9.6674Z)-157, D. < 10
0.4738D-226, D. > 10
(6)
Determining the actual movability number of a particle to compare with the critical movability
number requires first finding the particle's shear velocity, it*, and settling velocity, ws, by
T
(Pw
(7)
Av
BD
(8)
where
A = 53.5e-°-655/, B = 5.65e~2-5*5/, a = 0.7 + 0.9S/
The coefficients A and B are derived empirically and represent the effect of particle shape on
settling. Known as the Corey Shape Factor, 5^ is 0.7 in this study, which is typical of naturally
22
-------
occurring sediment particles. The actual movability number can then be determined by dividing
the shear velocity by the settling velocity, or u*/ws.
Figure 4 shows a plot of the critical movability number as a function of the dimensionless grain
diameter, D*. Similar to the Shields curve, if the actual movability number of a given particle is
greater than the critical movability number, then particle suspension will occur. In drinking water
storage tanks, typical dimensionless grain range from 0.25 - 25.
Beheshti Model
100
10
.
3
Z
o.i
1 10 100
Dimensionless Grain Diameter, D*
1000
Figure 4. Graphical representation of the Beheshti model (adapted from [33]) in which
the critical movability number is plotted as a function of the dimensionless grain
diameter.
2.2.3. Implementation of Sediment Suspension Models in CFD Model
For the CFD simulations conducted as part of this study, both the Shields and Beheshti sediment
transport models were implemented using C code consisting of user-defined functions (UDFs). A
UDF is defined by macros provided by the Fluent® software that allows the user to enhance normal
features and perform advanced tasks. For instance, UDFs can customize flow parameters, control
time step size during a transient simulation, alter material properties, obtain information about
specified grid mesh faces and cells, and modify boundary conditions, among other features.
In the operational simulation study (Section 2.3), UDFs were used to model particle movement
during a transient simulation of a representative fill/drain cycle. Particles were initially distributed
along the bottom wall of the tank (see Appendix B). Three different representative particle sizes
(0.01, 0.1, and 1 mm) were included in the simulations based on sediment samples collected from
several water distribution storage tanks (Appendix B). The size distribution of particles collected
from water tanks ranged from -0.01 - 1 mm. The density of the particles was assumed to be 2,650
23
-------
kg/m3 (silica sand). The UDF "injects" 6,000 particles consisting of equal numbers of 0.01, 0.1,
and 1 mm diameter particles on the tank bottom at the start of a simulation. Each size of particles
was distributed across the bottom face of the tank in the same locations, resulting in 2,000 different
initial positions and 3 different-sized particles at each position. The initial particle positions were
scaled towards the tank center since the particles further away from the inlet/outlet line were less
likely to experience resuspension and were thus of less interest (see Appendix B for more
information).
At each subsequent time step of the calculation, wall shear stresses were recorded at locations of
all particles touching the bottom wall. If the sediment suspension model predicted that particle
suspension would occur (e.g., the wall shear stress at the particle location was greater than the
critical shear stress), the particle was injected (allowed to move) into the fluid domain at that
location. Once injected, the motion of the particles was simulated in the Fluent® software using
a Lagrangian reference frame governed by a force balance on the particle. The force balance
includes a drag force that accounts for the velocity of the particle, velocity of the surrounding fluid,
particle size, and fluid properties [39]. If, on the other hand, a particle was in contact with the
bottom wall of the tank and the suspension model predicted a deposition (e.g., the wall shear stress
at the particle location was less than the critical shear stress), the particle would be recorded as
deposited in the UDF with an associated time stamp and particle identification number. Thus, the
number of particles deposited on the tank bottom, entrained in the flow and moving through the
tank, or drained from the outlet line were tracked and recorded at each time step.
To increase accuracy, the sediment suspension UDFs interpolated wall shear stresses between
centroids of the domain grid mesh faces comprising the bottom wall. In addition, critical particle
information for every particle that was either removed (drained) from the system or deposited
along the bottom wall, such as particle diameter, time of drain/deposition, time of preceding
resuspension, and particle position, were written to a text file for post-processing.
At the air-water interface, the Fluent® software does not by default include forces due to surface
tension. Therefore, a surface-tension model was created and implemented in a UDF to correctly
account for surface-tension forces on particles at the air-water interface, which prevented particles
from moving into the air phase from the liquid phase unrealistically. Appendix C describes the
surface tension model in detail.
In the parametric study (Section 2.4) and the small-scale validation study (Section 3), where
particles were not injected into the domain, UDFs were only used for post-processing after the
simulations were completed. The UDFs determined whether the wall shear stress at each bottom
wall mesh face exceeded the critical shear stress according to the Beheshti and Shields models, i.e.
particles would be resuspended at that face. The total area of particle resuspension was determined
by summing the individual areas of each face whose wall shear stress exceeded the critical shear
stress. In addition, the radial extent of resuspension was determined by dividing the bottom wall
into twenty individual regions of equal angular (azimuthal) span and averaging the radial extents
of resuspension over all regions. The averaging procedure accounted for small but noticeable radial
asymmetries in the bottom wall shear stresses.
24
-------
2.3. Operational Study of Particle Movement during Filling and Draining Cycles
The purpose of the operational study was to determine the temporal and spatial transport and
distribution of particles during typical tank operating conditions, which includes continuous filling
and draining cycles. Typical tank operating conditions were assumed to follow those described in
Section 1.2. The following sections describe the model development for the operational study and
the modeling results.
2.3.1. Geometric Configuration and Flow Parameters
In the operational study, two separate representative fill/drain cycles for a 38.4 m (126 ft) diameter,
9.75 m (32 ft) high Albuquerque Water Authority tank (11,300 m3 or 3 million gallons) were
simulated in the Fluent® software (Figure 2). Although tanks can have an inlet/outlet transmission
line located near the side walls, only the center-line tanks were modeled in the operational study,
allowing for a simpler and computationally efficient 2D axisymmetric domain. The axisymmetric
domain assumes that the flow is symmetric about the vertical centerline of the tank. Figure 5
shows the 2D domain and computational mesh used in the operational study simulations. The mesh
cells near the inlet/outlet and bottom wall are noticeably finer, which is necessary in order to
provide accurate flows and bottom wall shear stresses required to predict resuspension and
deposition of particles. The 2D axisymmetric mesh consists of 24,133 cells and 24,444 nodes.
Appendix D documents the validity of the 2D axisymmetric solution when compared to 3D
simulations. Grid convergence studies were also performed to ensure the adequacy of the grid
resolution. For this operational study, the 2D simulations of a single fill or drain cycle took several
weeks to complete. Full 3D simulations would take considerably longer (an order of magnitude
or more).
The flow parameters for the operational study are listed in Table 2. Two cases - a high flow rate
and a low flow rate - were simulated based on values provided by the two water utilities. Both
cases began with the fill stage, immediately followed by draining after 5,400 seconds (1.5 hours).
This time was sufficient to allow the particles to become suspended and recirculate through the
tank. The high flow case represents roughly the highest inlet/outlet velocity reported by the
Albuquerque Water Authority, while the low flow case represents the lower velocities.
Table 2. Tank and flow parameters used in operational study.
Both
Cases
Tank
Diameter
(ft / m)
126/38.4
Tank
Height
(ft / m)
32 / 9.75
Inlet/Oulet
Diameter
(in / m)
24 / 0.61
Fill Time
(s)
5400
Drain Time
(s)
5400
Initial Water
Level
50%
Case
High Flow
Low Flow
Flow Rate (gpm) / (m3/s)
10,000/0.631
3,400/0.215
Flow Velocity at Line (ft/s) / (m/s)
7.09/2.16
2.42 / 0.738
25
-------
Figure 5. Mesh of 2D axisymmetric model used in operational study. The dashed outline
in the schematic on top shows the location of the computational domain.
2.3.2. Results of Operational Study
Simulated flow pathlines - the trajectories that individual fluid particles follow - are displayed in
Figure 6 (high flow case) and Figure 7 (low flow case) at selected times (10, 5400, 5410, and
10,800 seconds). The pathlines are colored by the velocity of the fluid particles: red indicates
higher velocities and blue indicates lower velocities. During the filling stage, a circulation zone
forms early and grows in size until it occupies nearly the entire water-filled region of the tank.
Other, smaller, circulation zones can be seen as well. Once draining begins, much of the circulation
zone formed during filling is still present; however, flow near the inlet/outlet begins to split into
two directions: downward toward the outlet and upward in accordance with the circulation zone.
As the drain stage progresses, the circulation zone breaks down and the flow becomes highly
irregular far away from the inlet/outlet, characterized by rapid changes in direction and turns.
Chaotic flow was also observed in the 3D model during a drain-only run (see Appendix D).
26
-------
prim
f Draining-10,800
Figure 6. Flow pathlines colored by velocity magnitude (mis) during filling and draining
cycles in high flow rate case (0.631 m3/s or 10,000 gpm). Upside-down triangle denotes
water level. Flow pathlines above the water level denote velocity of the air phase.
Figure 7. Flow pathlines colored by velocity magnitude (mis) during filling and draining
cycles in low flow rate case (0.215 m3/s or 3,400 gpm). Upside-down triangle denotes
water level. Flow pathlines above the water level denote velocity of the air phase.
The pathlines in the filling stage of the high flow case (Figure 6) are slightly more stable and
orderly than in the low flow case (Figure 7). The primary circulation zone is also larger in the high
27
-------
flow case. However, both cases in the drain stage exhibit significantly high instabilities in the fluid
flow.
Figure 8 and Figure 9 display the particle traces of resuspended particles at selected times (every
2,000 seconds in addition to 10, 5400, and 10,800 seconds) for the high flow and low flow cases,
respectively. The red, green, and blue particles represent particle diameters of 1.0, 0.1, and 0.01
mm, respectively. The light blue mesh outline represents water, while the white space represents
air; any other colors in the mesh (near the water surface) represent cells in the process of switching
between the air phase and the water phase. When describing particle movement, particles are
categorized as resuspended (scoured from the tank bottom and entrained in the fluid flow), drained
(removed from the tank through the outlet pipe), or deposited back on the tank bottom. The
percentage of resuspended, drained, and deposited particles are summarized in Table 3.
Table 3. Summary of operational study results
Particle Size* (diameter)
Flow Rate**
% of particles resuspended
% of particles deposited
% of particles drained
% of particles drained whose
preceding resuspension occurred
during filling stage
% of particles entrained in flow at
end of filling (5400s)
% of particles entrained in flow at
end of draining (10800 s)
% of particles in upper stagnation
region
% of particles in lower recirculation
region
Maximum radial position at which
resuspension occurred (m)
0.01 mm
High
Flow
11.98%
7.97%
0.15%
0.15%
0.65%
3.87%
3.15%
0.23%
2.55
Low
Flow
8.70%
6.02%
0.32%
0.00%
0.38%
2.37%
1.68%
0.30%
2.25
0.1 mm
High
Flow
8.05%
5.72%
0.13%
0.00%
0.38%
2.20%
1.80%
0.20%
1.32
Low
Flow
4.27%
2.48%
0.25%
0.00%
0.07%
1.53%
1.15%
0.32%
0.79
1 mm
High
Flow
4.40%
2.62%
0.05%
0.00%
0.37%
1.73%
0.92%
0.47%
0.79
Low
Flow
1.45%
0.70%
0.28%
0.00%
0.00%
0.47%
0.47%
0.00%
0.49
All particle
diameters
High
Flow
24.43%
16.30%
0.33%
0.15%
1.40%
7.80%
5.87%
0.90%
2.55
Low
Flow
14.42%
9.20%
0.85%
0.00%
0.45%
4.37%
3.30%
0.62%
2.25
*6,000 particles were initially in the tank (2,000 each of 0.01, 0.1, and 1 mm diameter)
"High flow = 10,000 gpm (0.631 m3/s), Low flow = 3400 gpm (0.215 m3/s)
28
-------
Filling -10s
-
r~ •
Filling -4000s
";
•"*•*.
Draining - 6000 s
* " -jr* . •«*../
• •
* '•
i
Draining -10,000 s
.
•*•...*"
• ^
* ". " .'
* A * * **
Filling -2000s
•*«• "••
Filling -5400s
';
•.". . , ''T'
•
Draining - 8000 s
• t •
X" '••*• * '
J
Draining -10,800 s
• *» •
* • . ••
* • • • *
•. •/'• ' ,' . ' .• .
* » * * *
Figure 8. Particle traces during filling and draining cycles in high flow rate case (0.631
m3/s or 10,000 gpm). Particles colored by particle diameter (red = 1 mm, green = 0.1
mm, dark blue = 0.01 mm). Light blue mesh denotes the water domain.
29
-------
Filling-10s
Filling-2000s
Filling-4000s
Filling-5400s
Draining - 6000 s
Draining - 8000 s
Draining-10,000s
Draining-10,800s
Figure 9. Particle traces during filling and draining cycles in low flow rate case (0.215
m3/s or 3,400 gpm). Particles colored by particle diameter (red = 1 mm, green = 0.1 mm,
dark blue = 0.01 mm). Light blue mesh denotes the water domain.
In both the high- and low-flow cases, the particles are clumped together in the beginning of the
cycle and then steadily disperse, particularly during the draining cycle. The particles circulate
around the tank in the filling stage but begin to scatter less predictably in the drain stage, consistent
with the pathline patterns. Another common feature is the high proportion of 0.01-mm-diameter
particles entrained in the fluid flow. During the drain stage, both cases also show an increase in
the number of particles entrained in the flow compared to the fill stage.
30
-------
Due to its higher velocities and thus higher shear stresses, the high flow case exhibits a larger
number of resuspended particles and significantly more scatter in the particle traces. In addition, a
sizeable number of 1-mm-diameter particles become resuspended in the high flow case, compared
to only a few in the low flow case, which did not produce shear stresses along the bottom wall
sufficient to suspend the larger particles.
During the drain stage in both cases, a seemingly stationary particle trace can be seen near the
bottom of the tank connecting the inlet/outlet to the bottom wall. In fact, two tightly packed clusters
of particles are located in this region. One cluster is caught in a circulation zone within the
inlet/outlet; these particles are considered drained or removed from the tank. The other cluster of
particles remains stagnant on the bottom wall near the transmission line and is thus considered
deposited. These two recirculation patterns result from relatively sharp bends in the flow path near
the transmission line (Figure 10).
Figure 10. Pathlines and particle traces near the transmission line during draining. The
mesh is outlined in gray, the pathlines are red and orange, and the particles are red (1
mm), green (0.1 mm), and blue (0.01 mm). Two "stagnant" particle clusters are shown
here: one within the transmission line and one on the bottom wall. The pathlines can be
seen making relatively sharp bends in order to enter the transmission line, yielding
recirculation zones where the particles get trapped.
Figure 11 to Figure 18 illustrate additional results of the high- and low-flow cases, respectively.
Figure 11, Figure 16, and Figure 17 demonstrate that particles closer to the inlet/outlet are more
likely to experience resuspension. Figure 12 shows that the initial positions of particles that were
eventually drained during the simulation were close to the inlet/outlet (within one meter).
Similarly, Figure 13 shows the initial positions of particles that were entrained in the flow at the
end of the simulation; smaller particles close to the inlet are more likely to be entrained. Figure
12 shows that the smallest particles (0.01 mm) were drained closest to the tank center in the high
flow case, whereas the largest particles (1 mm) were drained closest to the tank center in the low
-------
flow case. In the high flow case, particles close to the tank center (including the largest particles)
were entrained in the flow during filling. While the larger particles were more prone to dropping
out of the entrainment at a location further from the inlet, the smaller particles tended to remain
entrained until the draining cycle, at which time they were more readily drained. In the low flow
case, the larger particles near the tank center were not entrained during the filling cycle, but during
the draining cycles, the shear stress near the inlet/outlet was greater, and the larger particles that
remained could be drained.
Figure 14 shows a high degree of scatter observed in the particles that were still entrained at the
conclusion of the cycle. The horizontal strip of symbols in both plots of Figure 14 represents the
particles stuck in the recirculation/stagnation zones referred to in Figure 10. This is a result of low
axial velocities (compared to the radial component) along the bottom wall away from the
transmission line (as seen in the velocity pathline figures), causing such particles to be "dragged"
along the bottom wall and toward the recirculation/stagnation zones. Furthermore, according to
Figure 15, among the particles that were resuspended at least once but ended up deposited on the
bottom wall at the end of the cycle, the majority were deposited close to their initial positions,
suggesting that most of these particles were deposited soon after being resuspended. This notion
is supported by Figure 16 and Figure 17, which show that the bulk of resuspensions and
depositions occurred during the same general times (shortly after the starts of filling at 0 s and
draining at 5,400 s). Figure 17 also indicates high shear stresses during draining due to the surge
of resuspensions immediately after the beginning of the drain stage.
Regarding the impact of different flow rates, the low-flow case resulted in a smaller number of
resuspensions than in the high flow case (Figure 17). Additionally, unlike in the high flow case
where depositions still occurred even after the drain stage was well underway (after about 6400
seconds in Figure 16), the low flow case exhibited no such depositions. This helps explain the
qualitative differences between the high- and low-flow cases in Figure 15. The former shows that
a modest fraction of the particles at rest on the bottom wall at the end of the cycle end up
considerably further away from the tank center relative to their initial positions; the latter displays
no such particles. Since a separate analysis revealed that no particles that were resuspended during
the low-flow fill stage were deposited during the drain stage, the above observations establish that
all particles that were deposited during the low-flow cycle were only briefly resuspended
beforehand. Despite the large number of resuspensions, the lower velocities and shorter periods of
entrainment prevented the particles from being transported large distances in the low-flow case.
Also, the total number of particles drained in the low-flow case was less than that of the high-flow
case. More particles were resuspended and redistributed away from the inlet/outlet in the high-
flow case, and fewer particles were available near the inlet/outlet when draining commenced.
Regarding the impact of different particle sizes, Figure 17 shows that the smallest particles (0.01
mm) were more prone to resuspensions than the larger particles (0.1 and 1 mm). However, the
number of particles eventually drained from the system was not largely dependent on particle size
in the low-flow case. Figure 18 shows that in the low-flow case, roughly the same number of
particles of each size were drained. This is caused by the resuspension of smaller particles near
the inlet line during filling. When draining commences, fewer smaller particles are near the drain
line. In contrast, fewer large particles (1 mm) are resuspended during filling in the low-flow case.
When draining commences, the larger shear stresses along the tank bottom enable resuspension
and draining of the larger particles that were previously immobile during filling, along with smaller
32
-------
particles located in regions outside the previously scoured region during filling. In the high-flow
case, fewer large particles are observed to drain (Figure 18). As shown in Figure 14, Figure 15,
and Figure 17, because of the larger momentum and shear stresses of the high-flow case, more
large particles are resuspended during filling and subsequently entrained or deposited to locations
further from the transmission line. Thus, fewer large particles are drained in the high-flow case.
In general, fewer particles were drained in the high-flow case due to more particles being entrained
and redistributed during filling.
Distribution of Particles Deposited on Tank Bottom at End of Cycle
Distribution of particles deposited on tank bot
300
300
4 6 8 10 12 14 16 18
Radial Distance from Tank Center (m)
4 6 8 10 12 14 16 18
Radial Distance from Tank Center (m)
Figure 11. Distribution of particles deposited on tank bottom at end of fill/drain cycle.
Initial Positions of Drained Particles
Initial Positions of Drained Particle:
t
£
•s
-------
Initial Positions of Particles Ent
led in Flow at End of Cycle
70
60 -
-§ 50 -
t
S. 4C
•s
S 30 -
1-1
10 -
0
High Flow Case
70
60 -
130 ^
E
i 2°-
10 -
0
Low Flow Case
0.4 0.6 0.8 1 1.2 1.4 1.6
Radial Distance from Tank Center (m)
Figure 13. Initial positions of particles entrained in flow at end of cycle
Final Radial Position vs. Initial Radial Position of all Particles Entrained in Flow at End of Cycle
Initial Radial Position of all Particles Entrained in Flow at End of Cycle
p ~
jj 18-
3 is -
•* -1/1
C 14 -
P
Q
H— 10 ~
OJ
u
ro
ts c
h b "
!™ 4 -
•o
ro
DC 2 -
"ni
C p.
High Flow Case
iP
*A
|b
A 1 mm
.1 mm
o .01 mm
A.
*2
A
O
0
c
jjj
c
-------
Final radial position vs. initial radial position of all particles deposited on tank bottom at end of cycle
Final Radial Position vs. Initial Radial Position of all Deposited on Tank Bottom at End of Cycle
Initial Radial Distance from Tank Center (m)
Initial Radial Distance from Tank Center (m)
Figure 15. Final radial position vs. initial radial position of all particles deposited on tank
bottom at end of cycle. The black line represents particles that remained deposited
throughout the entire cycle.
Distribution ofDepositio
8
u
t
S.
•s
V
E
z
OUU ~~
700 -
600 -
500 -
400 -
300 -
200 -
100 -
n
High Flow Case Blmm
• .1 mm
• .01 mm
-
2000
4000 5400 6000
Time (s)
8000 10000
VI
-------
Distribution of Resuspensions over Tim
Distribution of Resusper
VI
jmberof Pa rticl
z
900 -r
800 -
700 -
600 -
500 -
400 -
300 -
200 -
100 -
n
High Flow Case
• 1 mm
D .1 mm
• .01 mm
2000 4000 54006000 8000 10000
Time (s)
900
800 -
„ 700 -
_aj
£ 600 H
ro
°- 500 -
| 300 -
Z 200 -
100 -
0
Low Flow Case
0 2000 4000 5400 6000 8000 10000
Time (s)
Figure 17. Distribution of particle resuspensions overtime.
12
10 -
t
£
•5 6
2 -
High Flow Case
nber of Particles
l-> h
^ 01 oo o r-
3
2 -
n
Low Flow
Case
1
_
-j
1
-
nn :
• 1 mm
D .1 mm
• .01 mm
1 .
5400 6000 7000 8000 9000 10000 11000
Time (s)
5400
5400.5
5401 5401.5
Time (s)
5402
Figure 18. Distribution of particle drains overtime.
2.4. Parametric Study
The purpose of the parametric study was to determine the relative importance of various
parameters and processes, such as inlet/outlet location and diameter, flow rate, particle size, and
filling vs. draining cycles, on the resulting shear stress and potential for particle resuspension along
the bottom wall. When describing particle movement, particles are categorized as resuspended
(scoured from the tank bottom and entrained in the fluid flow), drained (removed from the tank
through the outlet pipe), or deposited back on the tank bottom.
2.4.1. Parametric Study Approach
The parametric study utilized a 3D half-symmetry model of a cylindrical water tank with a
diameter of 126 feet and height of 32 feet (see Figure 19). A 3D half-symmetry model (as opposed
to an axisymmetric model used in the operational study) was needed for the parametric study
because of the inclusion of a near-wall nozzle condition. The model had an initial water height of
16 feet (half full) for all studies. The two-phase volume-of-fluid model in the Fluent® software
36
-------
was used to simulate the two-phase water and air interactions at the water surface, and the k-co-
SST turbulence model was used to simulate turbulent flows in the tank. Approximately 670,000
model domain grid elements were used for this study. A grid-convergence study was performed
to ensure that that this was a sufficiently resolved mesh. Models were evaluated with different
element sizes, and each successive study refined elements around critical flow regions (regions
with distinct flow characteristics such as those near the nozzle location). Performance metrics
(particle velocity, nozzle velocity, flow patterns) were compared between studies to determine if
results changed between different mesh sizes. A sufficient grid was determined as the smallest
number of elements to produce an identical solution independent of the mesh size.
The transient modeling Courant number was kept as close to one as possible to ensure that an
accurate transient solution was being achieved. The transient model was run until the maximum
bottom-wall shear stresses were recorded (typically before 700 seconds of flow time), and the
recorded shear stresses were used in the sediment suspension models to determine where particle
resuspensions could occur in each scenario.
Figure 19. Half-symmetry tank domain used for the parametric study. The tank is 9.8 m
(32 ft) high with a diameter of 38 m (126 ft) and an initial water height of 4.9 m (16 ft).
An experimental design simulation matrix was created to define the parameter combinations
investigated in this study (see Table 4). Three two-level (low vs. high) factors were analyzed in
the CFD simulations for both filling and draining scenarios: inlet/outlet flow rate, diameter, and
location. The values chosen for each of the parameters were representative of the possible
minimum and maximum values during operation of similar sized water tank in Albuquerque, NM.
37
-------
Table 4. Experimental design matrix for cases simulated in the parametric study.
Design Study
Design Study
Flow Rate (gpm) /
(m3/s)
10,000/0.631
3,400/0.215
10,000/0.631
10,000/0.631
3,400/0.215
3,400/0.215
3,400/0.215
10,000/0.631
^^
Flow Rate (gpm) /
(m3/s)
10,000/0.631
10,000/0.631
500/0.0315
10,000/0.631
500/0.0315
500/0.0315
10,000/0.631
500/0.0315
Inlet Line Diameter
36/0.91
24/0.61
24/0.61
36/0.91
36/0.91
24/0.61
36/0.91
14/0.61
Draining
Inlet Line Diameter
Line Location
Near Side Wall
Center
Near Side Wall
Center
Center
Near Side Wall
Near Side Wall
Center
•
Line Location
Center
Near Side Wall
Near Side Wall
Center
Center
Near Side Wall
Near Side Wall
Center
2.4.2. Results of Parametric Study
The simulated shear stresses on the bottom of the water tank were analyzed to determine regions
that were susceptible to particle resuspension. Any shear stress (and particle size) resulting in a
movability number above the critical movability number in the Beheshti model was reported as a
susceptible region for particle resuspension. The results were reported as a contour plot in Figure
20 (red denotes regions susceptible to resuspension). Figure 20 shows the contours of the small
inlet/outlet diameter, high flow rate, and center location for three particle sizes. The results show
that smaller particles were more susceptible to resuspension for a given shear stress. Additional
contour plots similar to Figure 20 for all of the cases simulated in the parametric studies can be
found in Appendix E. The filling cases generally did not yield large shear stresses near the
transmission line, except for the cases where the line was near the wall during high flow rates. The
simulated shear stress near the transmission line was generally greater during draining, yielding a
greater region of potential resuspension.
38
-------
Figure 20. Plan view of the tank bottom showing area around inlet/outlet line susceptible
to particle resuspension (shown in red) for the 24 inch line, center location, high flow rate
case during draining for different particle diameters of 1 mm (left), 0.1 mm (center), and
0.01 mm (right).
Figure 21 and Figure 22 show the fraction of the tank bottom-wall area susceptible to resuspension
as a function of flow rate, inlet/outlet diameter, inlet/outlet location, and particle size. Results of
the tank filling scenario reveal that the high flow rate scenarios are most effective at inducing
particle resuspension, but only when the inlet/outlet is located near the tank wall. The case with
highest particle resuspension (>10%) for 0.01mm particles is the 24 inch line diameter, near wall
location, and high flow rate. The other particle sizes were only affected slightly during filling.
The draining scenarios were affected mostly by the high flow rate cases for all particle sizes
regardless of line location. The area susceptible to particle resuspension was also sensitive to the
particle size, with smaller particle sizes yielding larger areas of potential resuspension. These
charts include the entire area of the bottom of the tank for each particle size. Multiple particle
sizes can be resuspended in the same area of the tank. This chart does not explain the spatial
locations of resuspension - only the total amount of particles resuspended.
39
-------
t
re
Q.
o
re .2
12%
10%
8%
E
o 6%
0.01 mm particles
10.1 mm particles
11 mm particles
61 cm (24 in) Line
Diameter
Low Flow High Flow
91cm (36 in) Line
Diameter
Low Flow High Flow
61 cm (24 in) Line
Diameter
Low Flow High Flow
91 cm (36 in) Line
Diameter
Center Nozzle Location
Near Wall Nozzle Location
Figure 21. Fraction of bottom wall area susceptible to particle resuspension as a function
of various parameters during filling.
Low Flow High Flow
61 cm (24 in) Line
Diameter
Low Flow High Flow
91 cm (36 in) Line
Diameter
Low Flow High Flow
61 cm (24 in) Line
Diameter
Low Flow High Flow
91 cm (36 in) Line
Diameter
Center Nozzle Location
Near Wall Nozzle Location
Figure 22. Fraction of bottom wall area susceptible to particle resuspension as a function
of various parameters during draining.
40
-------
A statistical analysis was performed on the filling and draining cases to determine the relative
importance of each of the parameters on the fraction of area on the bottom of the tank that results
in particle resuspension. An analysis of variance (ANOVA) was performed to evaluate the impact
of input parameters (flow rate, inlet diameter, inlet location, and particle size) on the particle
resuspension. The ANOVA method evaluates the mean of the dependent variable (percentage of
area susceptible to particle resuspension) for all cases when a particular independent variable (e.g.,
flow rate, inlet diameter, inlet location, particle size) is set to its high or low value. If the means
are statistically different between these two settings based on the variance of the dependent
variable, the independent variable is considered statistically important. The interactions of
independent variables, say A and B, can also be evaluated by evaluating the mean of the dependent
variable at the high levels of both A and B.
Figure 23 shows a Pareto chart describing the relative importance of the input parameters on the
filling and draining cases.l Particle size, flow rate, and the interaction between particle size and
flow rate were found to be significant factors during both the filling and draining scenarios. The
inlet diameter, inlet location, and interactions among those parameters were not found to be
statistically significant in impacting the amount of particles resuspended.
Pareto Chart of the Standardized Effects for Filling
(response is % of Particles Resuspended, a = 0.05)
2.571
Pareto Chart of the Standardized Effects for Draining
(response is % of Particles Resuspended, a = 0.05)
2.57
D
A
AD
AC
CD
C
BC
B
BD
AB
7
A
D
AD
AC
BD
C
CD
B
BC
AB
?
Factor Name
A Flow Rate
B Inlet Diameter
C Inlet Location
D Particle Size
0.0 0.5 1.0 1.5 2.0 2.5
Standardized Effect
3.0 3.5
2 4 6 8 10 12
Standardized Effect
14
Figure 23. Pareto charts of the standardized effects showing the relative importance of
each factor (and interactions) on particle resuspension during filling (left) and draining
(right).
Figure 24 - Figure 27 show the fraction of bottom wall area susceptible to particle resuspension
as a function of momentum flow (jet effect) through the transmission line for different particles
1 The standardized effect of the independent parameters (A=Flow Rate, B=Inlet Diameter, C=Inlet Location,
D=Particle Size) is the absolute value of the t-statistic, which is equal to the coefficient (bl, b2, b3, b4) divided by
the standard error of the coefficient in the linear regression model: y = bO + bl*A + b2*B + b3*C + b4*D + b5*AB
+ b6*AC + b7*AD + b8*BC + b9*BD + blO*CD. The value of the threshold for the standardized effect to
determine significance corresponds to the cutoff in the t-distribution for an a of 0.05, where t is the (1 - a / 2)
quantile of a t-distribution with degrees of freedom equal to the degrees of freedom for the error term.
41
-------
sizes, line location, and filling vs. draining. The momentum flow (measured in Newtons, N) is
defined as the product of the mass flow rate and the momentum per unit mass:
Momentum Flow (N) = mv = pwv2A = -
where pw is the water density (1,000 kg/m3), Q is the flow rate of the water through the
transmission line (m3/s), and dmiet is the diameter of the transmission line (m). Results show that
particle resuspension is directly correlated with momentum flow through the transmission line.
Smaller particles are more susceptible to resuspension in all cases, and, during filling, the near-
wall inlet/outlet yielded greater particle resuspension than the center inlet/outlet. During draining,
both the near-wall and center inlet/outlet scenarios yielded similar particle resuspension trends,
with significantly greater particle resuspension occurring for the smallest (0.01 mm) particles.
These results indicate that reducing the momentum flow, or jet effect, through the orifice of the
inlet/outlet can reduce the potential for particle resuspension. The momentum flow can be lowered
by reducing the flow rate or increasing the inlet/outlet diameter. Particle resuspension may also
be reduced during filling by placing the inlet/outlet near the center of the tank rather than near the
side wall.
Fraction of bottom wall area
Fraction of bottom wall area susceptible to particle resuspension (%) at
c maximum particle resuspension time vs. momentum flow - Filling, Center
o
'in n 1 ">%,
-------
fG
-------
-------
For this study, a raised inlet configuration was added to the 2D axisymmetric tank model used in
the operational study to analyze its effects on particle resuspension. Extension height was varied,
while wall thickness was kept constant at 1 inch (2.54 cm) since it did not significantly affect
resuspension. The various cases are outlined in Table 5. As previously, in each case, 2,000 particles
of each of three different particle sizes (0.01, 0.1, and 1 mm) were included in the simulations. The
UDFs employed in the operational study were used in these simulations.
Table 5. Raised inlet line cases run in the Fluent® software. In all cases, the flow duration
was 200 seconds and the inlet/outlet line velocity was 2.162 mis, yielding a flow rate of
0.631 m3/s(10,000 gpm).
Filling/Draining
Filling
Filling
Filling
Filling
Draining
Draining
Draining
Draining
Extension Height Above
Bottom Wall
0.6 1m (24 in)
0.30m (12 in)
0.15m (6 in)
0
0.6 1m (24 in)
0.30m (12 in)
0.15m (6 in)
0.6 1m (24 in)
Filling and draining pathlines near a 0.30 m (12-inch) high raised inlet are displayed in Figure 29
and Figure 30, respectively. The raised inlet line forces the pathlines to make wide turns a moderate
distance away from the line, producing lower wall shear stresses adjacent to the line (Figure 31
and Figure 32). As a result, many resuspended particles were transported along the bottom wall
and were deposited immediately outside the raised inlet. Figure 33 illustrates that in the filling
cases, a line height of at least 0.15 m (6 inches) above the bottom wall prevented all of the 0.1 mm
and 1 mm particles from being suspended in the flow at the end of the run. A line height of 0.61
m (24 in) also prevented the 0.01 mm particles from being suspended. In the draining cases,
increasing the extension line height to 0.61 m (24 in) prevented the 0.1 and 1 mm particles from
being drained, but the smallest 0.01 mm particles continued to be susceptible to suspension and
subsequent drainage (Figure 34). In general, increasing the raised inlet line height above the base
of the tank appears to reduce the likelihood of suspending and draining particles, especially if they
are -0.1 mm or larger.
45
-------
2.209+00
2.099+00
1.988+00
1.878+00
1.769+00
1.659+00
1.548+00
1.438+00
1.329+00
1.219+00
1.108+00
9.908-01
8.808-01
7.709-01
6.609-01
5.508-01
4.409-01
3.309-01
2.209-01
HOe-01
O.OOe+00
Figure 29. Filling at 200 seconds - pathlines colored by velocity magnitude
3.148+00
2.98e+00
2.838+00
2.67e+00
2.51e+00
2.368+00
2.20e+00
2.04e+00
1.888+00
1.739+00
1.578+00
1.418+00
1.268+00
HOe+00
9.428-01
7.858-01
6.288-01
4.719-01
3.148-01
1.578-01
0.009+00
Figure 30. Draining at 200 seconds - pathlines colored by velocity magnitude
46
-------
l.E+01
l.E+00
To l.E-01
£ l.E-02
—•—No extension
—•—15 cm (6 in) extension
—*—30 cm (12 in) extension
—*— 61 cm (24 in) extension
Critical shear stress for resuspension
ro
-------
•o
Ol
•o
Ol
Q.
l/l
3
6%
5%
4%
Ol TO/
•g 3/°
t
01
U)
re
+j
c
Ol
y
Ol
Q.
2%
1%
0%
-0.01 mm
0.1 mm
-1 mm
10 20 30 40 50
Extension Line Height (cm)
60
70
Figure 33. Percentage of resuspended particles as a function of raised inlet line height
after 200 seconds of filling.
10 20 30 40 50
Extension Line Height (cm)
60
70
Figure 34. Percentage of particles drained as a function of raised inlet line height after
200 seconds of draining.
48
-------
3. Experimental Testing
Small-scale physical tests were performed to investigate particle resuspension during filling and
draining in a 1.2 m (4 ft) diameter water-filled plastic tank with a 0.02 m (0.8) diameter inlet
located in the center of the tank (see Figure 35). Two different particles (silica sand and glass
beads) with different densities were used in separate tests. During filling, different pumping flow
rates were investigated, but during draining, only gravity-induced draining was examined. Initially
for each test, the particles were distributed uniformly around the inlet line. Photos and videos were
recorded before and after each filling and draining event to determine where particles were
resuspended from the tank bottom. Tracer tests were also performed to characterize the flow
patterns and velocity fields. Finally, mitigation measures were investigated by raising the pipe
inlet, which was normally flush with the tank bottom, a short distance above the tank bottom. In
addition, a simulation study was conducted to compare CFD model predictions with the physical
test results.
49
-------
Water tank
Inlet/outlet line
—---> Valves to control and
divert flows during
filling or draining
Fill/drain line with
ball valve to divert
flow to sieves
Globe valve to
control fill rate
Bucket containing
sieve to collect
drained particles
u
Figure 35. Schematic (top) and photograph (bottom) of test apparatus.
3.1. Physical Test Description
Figure 35 shows the general configuration of the test apparatus. A globe valve was used to control
the flow rate during filling, and a flow meter used to measure the flow rate during both filling and
draining. Beneath the water tank are two ball valves, which were used to divert water either up
into the tank during filling, or out of the tank and to the drain during draining. The orange bucket
contained a sieve which would capture particles during draining; the bucket also had a hole in the
bottom, allowing water to drain out. The lights were used to illuminate the particles before and
after a test in order to take pictures of the initial and final distribution of particles, respectively. A
view of the bottom of the water tank is shown in Figure 36.
50
-------
Figure 36. Top view of the bottom of the experimental tank. The inlet/drain line can be
seen in the middle, surrounded by concentric circles. The circles were drawn 1 cm apart
from each other for spatial reference during the particle resuspension tests.
In order to test particle resuspension, three scenarios were identified: draining, filling, and a series
of filling and draining cycles (the draining scenarios only considered gravity drainage). For a
draining test, the tank was initially filled to a head of 0.3 m (12 inches), which was representative
of a full tank assuming a height/diameter aspect ratio of 1 :4. Next, a monolayer of particles was
evenly distributed around the drain. To obtain the monolayer of particles, a prescribed mass of
particles was initially measured to theoretically cover a given area around the drain based on the
particle mass per unit area. This calculation was conducted as follows:
To find mass per area:
~ Pbulk * "nominal
(9)
where
is the particle bulk density (2650 kg/m3 x 0.6 = 1590 kg/m3, assuming a porosity of
0.4) and Dnominai is the nominal diameter of the particle (m).
To obtain the desired area:
area = A= nRextent - nrdrain2
(10)
51
-------
where Rextent is the radial extent of the particle emplacement and rdrain is the radius of the drain
line. The desired radial extent was determined from preliminary tests. Finally, the necessary
mass of particles can be obtained as follows:
mass
mass = x A
area
(11)
The calculated mass of particles was then weighed and evenly distributed as a monolayer in the
water-filled tank around the drain using a cylindrical shell. Figure 37 shows the steel cylinder and
aluminum plug used to restrict the particles only to the desired area surrounding the drain. The
steel cylinder had a radius of about 10 centimeters, which covered a sufficiently large region
around the drain for the particle resuspension tests. After particle emplacement, the cylinder and
aluminum plug were removed (the particles remained stationary during this process).
DSCN1267.jpg
07.08.!
Figure 37. Cylinder used to restrict particles during particle emplacement.
Figure 38 shows the position of the valves required for each test. During a draining test, the lower
ball valve was closed and the upper ball valve was open, diverting the draining water towards the
sieve contained within the drain bucket.
52
-------
Open during draining,
closed during filling.
Open dunng filling,
closed during draining
Figure 38. Valves used to control filling and draining.
During draining, drained particles were collected in a set of sieves for particle-size and mass
analysis (Figure 39). Once the test had run to completion (when the water level reached six inches
(15 cm), or half the initial water height), the scoured region was measured radially from the drain,
at five to eight different locations. The drained particles were dried, distributed by diameter with
a sieve shaker, and weighed (Figure 40 and Figure 41).
Figure 39. Drained particles collected in sieve.
53
-------
38
Figure 40. Dried particles were sorted by size using a sieve shaker (Dual H-4325).
Figure 41. Dried particles from sieve were poured into containers (left) and weighed
using a Scientech ZSA 210 scale (right).
For the water-filling tests, the water tank was initially filled to a head of 0.15 m (6 inches), and a
monolayer of particles was distributed around the drain, as discussed before. The fill rate was
measured with the flow meter shown in Figure 42. The flow meter was a Blue-White Industries F-
1000 flow meter. Two flow rates were selected to use during fill tests, a low fill rate and a high fill
rate. The low fill rate was around 1.89 x 10"4 m3/s (about three gallons per minute) and the high
flow rate was 3.79 x 10"4 m3/s (about six gallons per minute).
54
-------
Figure 42. Flow meter used to measure fill rate during filling.
The tank was then filled to a head of nine inches. A final head of 0.225 m (9 inches) was sufficient
to observe the particle resuspension behavior during the filling test. Once a filling test had run to
completion, the scoured radial extent of the tank bottom was measured at five to eight radial
locations around the drain.
For a fill-drain-fill-drain test, the tank was initially filled to a head of 22.5 m (9 inches) and a
monolayer of particles was evenly distributed around the drain. The tank was then filled to 30 cm
(12 inches), drained to 22.5 cm, filled to 30 cm again, and finally drained to 22.5 cm. The range
from 22.5 to 30 cm (9 to 12 inches) was selected in order to keep the test time short, but still
provide maximum shear stresses and velocities during draining. Once the test had been run to
completion, the drained particles were dried, distributed by diameter with the sieve shaker, and
weighed. The radial extent of resuspension from the drain was also measured at five to eight radial
locations around the drain.
Tracer tests were also performed for both filling and draining scenarios prior to the particle
resuspension tests to evaluate the flow fields and fluid velocities near the tank bottom. Blue dye
(food coloring) was injected using a commercial baster at various positions along the bottom of
the tank (Figure 43). For draining tests, the dye was video recorded to obtain fluid velocities near
the bottom of the tank. Windows Movie Maker was used to analyze the velocity of the tracer as it
moved past the concentric rings spaced 1 cm apart. The time in Movie Maker is reported in
milliseconds, allowing for a precision within 0.03 seconds.
55
-------
DSCN1287.mov
Figure 43. Tracer testing during draining. Note: Each ring is one centimeter apart.
Tracer tests were also performed to evaluate the velocity near the inlet during filling. Results
showed that flow along the bottom wall within 1-2 millimeters from the edge of the drain
exhibited high velocities, but were low elsewhere along the bottom wall. Figure 44 shows an
example of a tracer test performed during filling, which revealed velocity distributions around the
inlet. In Figure 44, blue dye resting on the bottom moved very little as compared to the dye that
had entered the region directly above the drain. The fluid velocities outside the region within one
or two millimeters of the drain were very small, as compared to the fluid velocity directly next to
and above the drain. Resuspension tests conducted later confirmed this behavior.
Figure 44. Example of tracer testing during filling. Numbers on the vertical plug are in
inches.
56
-------
3.2. Test Results
3.2.1. Tracer Tests
Figure 45 shows the velocities measured along the tank bottom during draining with an initial
water head of 0.3 m (12 inches). The velocities were measured during the tracer tests described in
Section 3.1. Figure 45 shows that far away from the drain, fluid velocities were very low (<0.01
m/s). However, within about three centimeters from the center of the drain, fluid velocity increases
exponentially to greater than 0.1 m/s.
Solidworks™ Flow Simulation software and ANSYS® Fluent® CFD™ software were used to
simulate fluid velocities at varying radial distances from the drain. Results of simulated velocities
5 mm above the tank bottom using the Fluent® software and Solidworks™ software are also
shown in Figure 45. The Fluent® model used a two-phase volume-of-fluid model described in
Section 2.1. The Solidworks™ model assumed a single-phase system with a constant fluid
pressure boundary condition at the top of the domain (no air-water interface as in the Fluent®
model). The results show that for draining conditions, both models match the data well within
approximately 0.05 m from the drain. Thus, during draining, modeling flow and particle
resuspension near the drain with a single-phase constant-head model may be adequate at discrete
points in time. For filling, the two-phase model is necessary to accurately capture the changing
water level at the air-water interface and the resulting recirculating flow patterns.
57
-------
Fluid Velocity_Tracer, FLUENT, Solidworks.xlsx
0.1
(A
i
0)
A Test Data 9 July
• Test Data 8 Aug
FLUENT
^—Solidworks Flow
0.01
0.001
0.02
0.04
0.06 0.08
Distance from Drain (m)
0.1
0.12
0.14
Figure 45. Measured and simulated velocity along the tank bottom during draining as a
function of radial distance from the center of the drain. Error bars represent plus/minus
one standard deviation.
For comparison with the experimental results, a 0.3 to 0.15 m (12 to 6 inch) drainage simulation
was run with 3D reflection symmetry in the Fluent® software. To account for the continuous
decrease in the drain rate in the experimental tests, the velocity boundary condition at the outlet
was modified at every 0.0254 m (1 inch) change in the water level. The flow velocity in the
Fluent® simulation was linearly interpolated between the measured drain rate at the equivalent
water level and the measured drain rate after a 0.0254 m (1 inch) decrease in the equivalent water
level. The flow parameters are outlined in Table 6.
58
-------
Table 6. Flow parameters for simulation draining from 0.3 to 0.15 m (12 to 6 inches)
Water Height
(m - in)
0.3m (12 in)
0.28m (11 in)
0.25 m (10 in)
0.23 m (9 in)
0.2 m (8 in)
0.18m (7 in)
0.15m (6 in)
Measured Drain Velocity
(m/s)
0.804
0.791
0.781
0.771
0.756
0.745
0.737
Drain Velocity Input in Fluent® (m/s)
0.7977
0.7860
0.7757
0.7636
0.7510
0.7413
n/a
Flow Time (s)
0.000
108.046
217.699
328.811
441.690
556.464
672.729
3.2.2. Particle Resuspension Tests
During particle resuspension tests, all three scenarios described in Section 3.1 were conducted:
(1) draining, (2) filling, (3) fill-drain-fill-drain. For draining and fill-drain-fill-drain tests, the mass
drained and radial extents of resuspension were the two measured quantities. For filling, the radial
extent of resuspension was the only measured and recorded quantity. Table 7 - Table 9 summarize
the results of the particle resuspension testing. After each test, the mass of particles drained was
recorded using the methods described in Section 3.1. The radial extent of resuspension was
determined by measuring the radial distance of the scoured region around the drain. Several radial
measurements were taken and then averaged. In between tests, the tank was drained and any
remaining particles on the bottom of the tank were wiped toward the drain for removal before
filling the tank again with water.
The draining tests for the silica sand were consistent with the modeling studies that predicted
smaller particles were more likely to be resuspended as compared to larger particles of the same
density. As shown in Table 7, small silica sand particles (on the order of 0.05 to 0.1 mm diameter)
experienced more resuspension than the 0.853 to 1.68 mm silica sand as evidenced by a greater
mass fraction collected during draining and a larger scoured region around the drain.
Table 7. Draining test results.*
Particle Type
Silica Sand
Silica Sand
Glass Beads
Glass Beads
Particle
Density
(kg/m3)
2650
2650
2450
2450
Particle
Diameter (mm)
0.053-0.104
0.853-1.68
1
• 2 •
Number of
Tests Run
5
3
1
^^^^H
Particles
Collected
During
Draining (% by
mass)
4.17 ±2.46
2.79 ±0.08
4.20
• 2.72 •
Average Radial Extent
of Resuspension from
Drain Edge (cm)**
1.28 ±0.06
0.92 ±0.07
1.51 ±0.25
^H 1.74 ±0.26 ^H
* Uncertainty in measurements represented by plus/minus one standard deviation
** 5 -8 radial measurements per test
The radial extent of the resuspended/drained particles was greater for the smaller sized silica sand,
but statistically similar for the glass beads. Additionally, the draining tests show that particles with
a lower density were more likely to be resuspended. The 1 mm glass beads experienced a greater
59
-------
amount of resuspension when compared to the 0.853 - 1.68 mm silica sand. Additionally, the
scoured region was greater for the less dense glass beads as compared to the more dense silica
sand. Figure 46 and Figure 47 show examples of the radial extent of resuspension for silica sand
and glass beads during draining-only tests.
Figure 46. Example of photos of 0.853 - 1.68 mm silica sand particles on the tank bottom
near the inlet/drain before (left) and after the draining-only test (right). Concentric rings
are 1 cm apart.
Figure 47. Example of photos of 1 mm glass bead particles on the tank bottom near the
inlet/drain before (left) and after the draining-only test (right). Concentric rings are 1 cm
apart.
Filling tests revealed that particle resuspension generally occurred within a 0.2 - 0.5 cm radius
around the drain for all cases (Table 8). In this region, shear stresses were sufficiently high to
cause particle resuspension. Often this meant that particles were resuspended but quickly settled
outside the affected region. This region was consistent across the range of particle size and particle
density. Glass beads had a greater radial extent, but these values fall within one standard deviation
of silica sand values.
60
-------
Table 8. Filling test results/
Particle
Silica Sand
Silica Sand
Silica Sand
Silica Sand
Glass Beads
Glass Beads
Glass Beads
Glass Beads
Particle
Density
(kg/mA3)
2650
2650
2650
2650
2450
2450
2450
2450
Particle Diameter
(mm)
0.053-0.104
0.053-0.104
0.853-1.68
0.853-1.68
1
1
2
^H 2 ^|
T^-n T> .. TWT •_ f Average Radial Extent
Fill Rate Number of fn • f
/• A-I/N ^ , T, of Resuspension from
(mA3/s) Tests Run _ . *, , ...
v ' Dram Edge (cm)**
1.96E-04
4.01E-04
1.92E-04
3.65E-04
1.91E-04
3.81E-04
1.92E-04
3.84E-04
0.15 ±0.05
0.15 ±0.05
0.22 ±0.11
0.30
0.44 ±0.12
0.50
0.30
^^^^H 0.30 ^^H
* Uncertainty in measurements represented by plus/minus one standard deviation
** 5 -8 radial measurements per test
The fill-drain-fill-drain cycle tests yielded variable results when compared to the pure draining and
filling tests (Table 9). The larger silica particles experienced less resuspension than when the tank
was purely draining. One possible explanation is that the larger particles were resuspended by the
inflow during filling, but quickly fell outside the area that would be drained. Videos from the
filling test and cycle test support the initial hypothesis that the larger particles were kicked up and
quickly settled outside the area of action. Results of the fill-drain-fill-drain tests show that a larger
mass fraction of glass beads was drained when compared to the drain-only tests, but the radial
extents were statistically similar. It is possible that a small amount of particles were left over from
previous tests such that the mass collected from one test included particles from a previous test.
The radial extent of resuspension for all the tests are summarized in Figure 48.
Table 9. Fill, drain, fill, drain results.*
Particle
Silica Sand
Silica Sand
Silica Sand
Silica Sand
Glass
Beads
Glass
Beads
Glass
Beads
Glass
Beads
Particle
Density
(kg/mA3)
2650
2650
2650
2650
2450
2450
2450
2450
Particle
Diameter (mm)
0.053-0.104
0.053-0.104
0.853-1.68
0.853-1.68
1
1
2
2
Fill Rate
(mA3/s)
1.91E-04
3.79E-04
1.89E-04
3.66E-04
1.89E-04
3.85E-04
1.93E-4
3.83E-04
Number
of Tests
Run
2
2
1
1
1
1
2
1
Particles
Collected
during
Draining (%
by mass)
3.91±0.51
5. 53 ±0.46
1.57
1.54
5.11
4.87
7.32 ±0.19
5.76
Average Radial
Extent of
Resuspension
from Drain Edge
(cm)**
1.36 ±0.04
1.45 ±0.13
0.72 ±0.15
0.90 ±0.19
1.63 ±0.30
1.94 ±0.29
1.83 ±0.29
1.90 ±0.54
* Uncertainty in measurements represented by plus/minus one standard deviation
** 5 -8 radial measurements per test
61
-------
2.5
0>
Q.
i/i
0)
oc
"o
4.1
c
V
X
LU
|ro
E
u
'ro
Q
*O
0>
M
•o
LU
E
2
2
1.5
1
n.5
-------
0.022 0.049 0.064 0.090 0.178 0.337 0.711 1.500 2.000
Average Particle Diameter (mm)
Figure 49. Mass fraction histogram and cumulative distribution (CDF) for #12 - #20 silica
sand from manufacturer.
0.049 0.064 0.090 0.178 0.337 0.711
Average Particle Diameter (mm)
1.500
2.000
Figure 50. Histogram and cumulative distribution function (CDF) for the mass fraction of
#12 - #20 silica sand collected after draining test.
Simulations of the particle resuspension tests were performed using the Solidworks™ software
and the ANSYS® Fluent® CFD™ software. Both programs simulate shear stresses at various
radial locations from the drain, which can be used to predict particle resuspension based on the
Shields or Beheshti models. The simulated shear stresses were used to determine the radial extent
of particle resuspension, which was compared against experimental data. The results have been
compiled below, using both the Shields and Beheshti models for particle resuspension. Table 10
and Table 11 show the simulated radial extent of particle resuspension from draining scenarios
63
-------
using the Solidworks™ software and the Fluent® software, respectively. Figure 51 - Figure 52
provide bar charts that illustrate the data presented in Table 10 and Table 11. Table 12 shows the
simulated radial extent of particle resuspension during filling-only scenarios using the Fluent®
software.
Table 10. Draining scenarios: experimental radial extent of resuspension compared with
modeled radial extent of resuspension. Shear stresses for modeled results obtained
using the Solidworks™ software.
Particle
Particle Diameter Test Type
(mm)
Modeled Radial Extent of
Particle Resuspension from Edge
of Inlet/Drain (cm)
c-i • u T.» j • Beheshti
Shields Model ,„ , ,
Model
Experimental
Radial Extent of
Particle
Resuspension
from Edge of
Inlet/Drain (cm)
Silica Sand
Silica Sand
Glass Beads
Glass Beads
0.053-0.104
0.853-1.68
1
Draining
Draining
Draining
Draining
2.48-3.70
1.34-1.60
1.34-1.60
1.08-1.34
1.95-2.48
1.34-1.60
1.34-1.60
1.08-1.34
1.28 ±0.06
0.92 ±0.07
1.51 ±0.25
1.74 ±0.26
Table 11. Draining scenarios: experimental radial extent of resuspension compared with
modeled radial extent of resuspension. Shear stresses for modeled results obtained
using the ANSYS® Fluent® CFD™ software.
Particle
Silica Sand
Silica Sand
Silica Sand
Silica Sand
Silica Sand
Silica Sand
Glass Beads
Glass Beads
Glass Beads
Glass Beads
Glass Beads
Glass Beads
Particle
Diameter
(mm)
0.053-0.104
0.853-1.68
0.053-0.104
0.853-1.68
0.053-0.104
0.853-1.68
1
2
1
2
1
^^^^^H
Test Type
Draining
Draining
High Fill, Drain
High Fill, Drain
Low Fill, Drain
Low Fill, Drain
Draining
Draining
High Fill, Drain
High Fill, Drain
Low Fill, Drain
Low Fill, Drain
Modeled Radial Extent of
Particle Resuspension from Edge
of Inlet/Drain (cm)
c-i-- u n/r j i Beheshti
Shields Model - „ , ,
A/lnnpl
1TJ.UUC1
5-6 4-5
2-3 2-3
5-6 4-5
2-3 2-3
5-6 4-5
2-3 2-3
2-3 2-3
1-2 1-2
2-3 2-3
1-2 1-2
2-3 2-3
^^•-2 ^^^H-2
Experimental
Radial Extent of
Particle
Resuspension
from Edge of
Inlet/Drain (cm)
1.28 ±0.06
0.92 ±0.07
1.45 ±0.13
0.90 ±0.19
1.36 ±0.04
0.72 ±0.15
1.51 ±0.25
1.74 ±0.26
1.94 ±0.29
1.90 ±0.54
1.63 ±0.30
1.83 ±0.29
64
-------
Table 12. Filling scenarios: experimental radial extent of resuspension compared with
modeled radial extent of resuspension. Shear stresses for modeled results obtained
using the ANSYS® Fluent® CFD™ software.
Particle
Silica Sand
Silica Sand
Silica Sand
Silica Sand
Glass Beads
Glass Beads
Glass Beads
Glass Beads
Particle
Diameter
(mm)
0.053-0.104
0.853-1.68
0.053-0.104
0.853-1.68
1
2
1
^^^^^^B
Test Type
Low Flow Fill
Low Flow Fill
High Flow Fill
High Flow Fill
Low Flow Fill
Low Flow Fill
High Flow Fill
High Flow Fill
Modeled Radial Extent of
Particle Resuspension from Edge
of Inlet/Drain (cm)
c-i-- u n/r j i Beheshti
Shields Model ,. _ , ,
A/ionpl
1TJ.UUC1
0-
0-
0-
0-
0-
0-
0-
• 0-
0-1
0-1
0-1
0-1
0-1
0-1
0-1
^^^^^^^^^^B
Experimental
Radial Extent of
Particle
Resuspension
from Edge of
Inlet/Drain (cm)
0.15 ±0.05
0.22 ±0.11
0.15 ±0.05
0.30 ±0.01
0.44 ±0.12
0.30 ±0.01
0.50 ±0.01
0.30 ±0.01
Because the visible radial extent of resuspension is based on where the largest particles remained
after draining, nominal particle diameter used in the Shields and Beheshti prediction models was
set at the 95th percentile. As expected, the largest particles are less likely to experience any motion
due to the draining fluid; therefore particles in the 95th percentile would most likely be the observed
particles during radial extent measurements.
The predicted radial extents of particle resuspension during draining using Solidworks™ and
Fluent® software codes and the experimental data are shown in Figure 52- Figure 52. Both codes
tended to over predict the radial extent of particle resuspension for silica sand. The non-spherical
shape of the sand may have inhibited mobility and caused the radial extent of particle resuspension
in the experiments to be less. The Solidworks™ model generally under predicted the extent of
particle resuspension for glass beads during draining. During filling, both the simulations from
Fluent® and the experiments show that the radial extent of particle resuspension is significantly
less than during draining.
65
-------
Silica Sand Draining Test
Glass Bead Draining Test
0.053-0.104 0.853-1.68
Particle Size (mm)
• Shields Model
I Beheshti Model
Experimental Results
1 2
Particle Size (mm)
Figure 51. Draining scenarios: comparison of simulated (Solidworks™ software) and
experimental average radial extents of particle resuspension. Left: silica sand. Right:
glass beads. Error bars represent one standard deviation about the mean.
ANSYS FLUENT v. Experimental Results -
Silica Sand Draining Test
ANSYS FLUENT v. Experimental Results -
Glass Beads Draining Test
I Shields Model
I Beheshti Model
Experimental Results
0.053-0.104 0.853-1.68
Particle Diameter (mm)
1 2
Particle Diameter (mm)
Figure 52. Draining scenarios: comparison of simulated (ANSYS Fluent®) and
experimental average radial extents of particle resuspension. Left: silica sand. Right:
glass beads. Error bars represent one standard deviation about the mean.
3.3. Impact of Raised Inlet/Outlet on Particle Resuspension
Testing also included an evaluation of an inlet/outlet line raised 1 cm (0.4 in) above the tank bottom
to determine if the extension would reduce the amount of particle resuspension (see Figure 53).
66
-------
DSCN1409.jpg
Figure 53. Extended line installed above drain.
Table 13 and Table 14 show that the extended line mitigated the resuspension of particles during
both filling and draining. No region of scoured particles was observed near the inlet/drain. Only
the smallest of particles was resuspended and recovered during the silica sand drainage test, but
the mass of particles resuspended was significantly less than for a pure draining test. The mass of
silica sand resuspended with the extended line installed was about a tenth of the mass resuspended
without the extended drain.
Table 13. Extended line draining results.
Particle
Particle
Density
(kg/mA3)
Particle
Diameter
(mm)
Particles Collected
during Draining
(% by mass)
Average Radial Extent
of Resuspension from
Drain Edge
(cm)
Silica Sand
Silica Sand
Glass Beads
Glass Beads
2650
2650
2450
2450
0.053-0.104
0.853-1.68
2
1
0.55
0
0
0
0
0
0
0
67
-------
Table 14. Extended line filling results.
Particle
Silica Sand
Silica Sand
Silica Sand
Silica Sand
Glass Beads
Glass Beads
Glass Beads
Glass Beads
Particle
Density
(kg/mA3)
2650
2650
2650
2650
2450
2450
2450
2450
Particle
Diameter
(mm)
0.053-0.104
0.053-0.104
0.853-1.68
0.853-1.68
1
1
2
• 2 •
Fill Rate (mA3/s)
1.94E-04
3.82E-04
1.96E-04
3.82E-04
1.94E-04
3.79E-04
1.92E-04
3.79E-04
Average Radial Extent of
Resuspension from Drain Edge
(cm)
No visible extent
No visible extent
No visible extent
No visible extent
No visible extent
No visible extent
No visible extent
No visible extent
4. Summary and Conclusions
This report has presented studies of particle resuspension and movement in water distribution
storage tanks during filling and draining cycles. Two computational studies were performed:
(1) an operational study of a 11,000 m3 (3 million gallon) water tank was performed to determine
when and where different sized particles were resuspended from the tank bottom and transported
during filling and draining cycles at low and high flow rates, and (2) a parametric study to
determine the impact of parameters such as particle size, flow rate, inlet/outlet diameter and
location, and filling vs. draining on the shear stresses and potential for particle resuspension on the
tank bottom. Testing was also performed with a small-scale water tank to investigate particle
resuspension during filling and draining cycles with different particle sizes and densities (glass
beads and silica sand). The test results were compared to simulations of the physical tests
conducted in order to verify the modeling approach. Finally, the use of a raised inlet/outlet line
was investigated to mitigate particle resuspension. Results of these investigations are summarized
below.
4.1. Results of Operational Study
In the operational study, the resuspension and movement of different sized particles (0.01, 0.1, and
1 mm diameter) were simulated during subsequent filling and draining cycles in the tank. Two
different flow rates were used based on representative minimum and maximum flow rates provided
by the Albuquerque Water Authority. Key results of the numerical simulations were as follows:
• Particle resuspension from the tank bottom generally occurred immediately following the
start of either the filling or draining processes
• Smaller particles were more susceptible to resuspension and entrainment
• Once entrained in the fluid flow, particles were typically carried further away from the
inlet/outlet during the filling cycle, making them less susceptible to removal during
draining
• Greater shear stress during draining led to more particle resuspension than during filling
• Recirculation zones near the inlet/outlet were observed
68
-------
4.2. Results of Parametric Study
A parametric study was performed to determine the impact of particle size, flow rate, inlet/outlet
diameter and location, and filling vs. draining on the shear stress and potential for particle
resuspension on the tank bottom. Key findings were as follows:
• Particle size, flow rate, and inlet/outlet location were found to be important factors for
particle resuspension
o Smaller particles were more susceptible to resuspension, although the difference
was less during draining than filling
o Higher flow rates yielded greater resuspension
o Inlet/outlets located near the side wall (vs. at the center of the tank) yielded
greater resuspension
o Draining yielded greater resuspension of particles than filling
• Particle resuspension was directly correlated to the amount of momentum flow, or jet
effect, through the inlet/outlet. Reducing the flow rate or increasing the diameter of the
inlet/outlet reduces the momentum flow and the potential for particle resuspension.
• Placing the inlet/outlet near the center of the tank rather than near the side wall reduced
particle resuspension during filling.
4.3. Results of Mitigation Measures
In order to mitigate the potential for particle resuspension near the inlet/outlet line, a raised
inlet/outlet was investigated. The hypothesis was that this would reduce the shear stresses near
the inlet/outlet and reduce the potential for particle resuspension. The extension pipe configuration
(with varying extension heights) was added to the 2D axisymmetric tank model that had been
previously used in the operational study. Key results were as follows:
• During filling, a pipe height extending 6 inches (15 cm) above the center of the tank
bottom substantially reduced the number of particles resuspended
• During draining, a pipe height extending 12-24 inches (30-61 cm) above the center of
the tank bottom substantially reduced the number of particles drained from the tank
4.4. Results of Testing
Small-scale tests were performed to investigate particle resuspension during filling and draining
in a 4-ft (1.2 m) diameter water-filled tank with a 0.8 in (2 cm) diameter inlet/outlet located in the
center of the tank. Photos and videos were recorded before and after each filling and draining
event to determine where particles were resuspended from the tank bottom. Tracer tests were also
performed to characterize the flow patterns and velocity fields. Finally, mitigation measures were
investigated by raising the pipe inlet, which was normally flush with the tank bottom, 1 cm (0.4
in) above the tank bottom. Key results were as follows:
• Measured and simulated velocities along tank bottom matched well up to about 5 cm
from drain, including region where particles were resuspended
69
-------
o Velocities along the tank bottom were very small away from the inlet (<1 cm/s),
but increased exponentially within 2 - 3 cm from the inlet/outlet to above 10 cm/s
• Resuspension of particles was limited to within ~1 cm from the inlet/outlet during filling
and draining cycles for the flow rates used in the study.
• Smaller particles yielded a greater radial extent of resuspension from the inlet/outlet
during filling and draining cycles
• Less dense particles (glass beads) than the silica sand exhibited greater resuspension
• During the fill-drain-fill-drain cycle, fewer large particles were drained when compared
to the drain-only scenario. This is likely because during the fill cycle, particles close to
the inlet were resuspended and deposited further away. When the drain cycle
commenced, there were fewer particles near the inlet to be drained. However, a greater
fraction of the smaller particles were drained when compared to the drain-only scenario.
A possible reason may be that the small particles remained entrained during filling and
were subsequently drained, and the fill-drain-fill-drain sequence caused additional
perturbations and shear stress that enable a greater amount of smaller particles to be
drained.
• Model predictions of resuspension generally matched experimental data for glass beads,
and generally over predicted the amount of resuspension for silica sand. The non-
spherical shape of the sand may have reduced the amount of resuspension in the tests.
• Both modeling and experiments showed that a raised inlet reduced particle resuspension
and removal during filling & draining
o Minimum height to completely mitigate particle movement near the inlet/outlet
was found to be about 3 -8% of the head of water
• In the tests, an extension of 1 cm (0.39") mitigated particle movement
with a maximum head of water of 30 cm (12")
• In the models, an extension of-0.38 m (1.3 ft) mitigated particle
movement with a head of water of 4.9 m (16 ft)
70
-------
References
[1] Grayman, W.M., L.A. Rossman, C. Arnold, R.A. Deininger, C. Smith, J.F. Smith, and R.
Schnipke, 1999, Water Quality Modeling of Distribution System Storage Facilities,
American Water Works Association, Denver, CO.
[2] Mahmood, F., J.G. Pimblett, N.O. Grace, and W.M. Grayman, 2005, Evaluation of water
mixing characteristics in distribution system storage tanks, Journal American Water
Works Association, 97(3), p. 74-88.
[3] Tian, X.D. and P.J.W. Roberts, 2008, Mixing in Water Storage Tanks. II: With Buoyancy
Effects, Journal of'EnvironmentalEngineering-ASCE, 134(12), p. 986-995.
[4] Tian, X.D. and P.J.W. Roberts, 2008, Mixing in Water Storage Tanks. I: No Buoyancy
Effects, Journal ofEnvironmental Engineering-ASCE, 134(12), p. 974-985.
[5] Rossman, L.A. and W.M. Grayman, 1999, Scale-model studies of mixing in drinking
water storage tanks, Journal of Environmental Engineering-ASCE, 125(8), p. 755-761.
[6] Roberts, P.J.W., X. Tian, F. Sotiropoulos, and M. Duer, 2005, Physical Modeling of
Mixing in Water Storage Tanks, AWWA Research Foundation, Report #91112, Denver,
CO.
[7] Adamsson, A., V. Stovin, and L. Bergdahl, 2003, Bed shear stress boundary condition for
storage tank sedimentation, Journal of Environmental Engineering-ASCE, 129(7), p. 651-
658.
[8] Cheng, N.S. and Y.M. Chiew, 1999, Analysis of initiation of sediment suspension from
bed load, Journal of Hydraulic Engineering-ASCE, 125(8), p. 855-861.
[9] Choi, S.U. and S. Kwak, 2000, Probabilistic analysis of incipient motion of sediment
particles, in in Proceedings of the 8th International Symposium on Stochastic Hydraulics
ISSH2000, Beijing, China, July 25 - 28, 2000.
[10] Dey, S. and A. Papanicolaou, 2008, Sediment Threshold under Stream Flow: A State-of-
the-Art Review, KSCE Journal of Civil Engineering, 12(1), p. 45-60.
[11] He, C., J. Marsalek, and Q. Rochfort, 2004, Numerical modelling of enhancing
suspended solids removal in a CSO facility, Water Quality Research Journal of Canada,
39(4), p. 457-465.
[12] Stovin, V.R. and A.J. Saul, 1994, Sedimentation in Storage Tank Structures, Water
Science and Technology, 29(1-2), p. 363-372.
[13] Stovin, V.R. and A.J. Saul, 1998, A computational fluid dynamics (CFD) particle
tracking approach to efficiency prediction, Water Science and Technology, 37(1), p. 285-
293.
[14] Stovin, V.R. and A.J. Saul, 2000, Computational fluid dynamics and the design of
sewage storage chambers, Journal of the Chartered Institution of Water and
Environmental Management, 14(2), p. 103-110.
[15] Vanoni, V.A., 1975, Factors Determining Bed Forms of Alluvial Streams, Journal of the
Hydraulics Division-ASCE, 101(11), p. 1435-1440.
[16] Delleur, J.W., 2003, Hydraulics of sediment movement in urban drainage systems,
Journal of ~ Hydraulic Engineering-ASCE, 129(4), p. 251-252.
71
-------
[17] U.S. Environmental Protection Agency, 2009, Community Water System Survey.
Volume II: Detailed Tables and Survey Methodology, EPA/815/R-09/002, Office of
Water.
[18] Mays, L., 2001, Water Distribution Systems Handbook, Larry Mays (Editor), McGraw-
Hill, New York.
[19] Lamfers, J., 2011, Kansas Water System Contaminated due to Birds Accessing Storage
Tank, Kansas Lifeline., p. 38-41, November 2011.
[20] U.S. Environmental Protection Agency. Finished Water Storage Facilities. 2002 August
15, 2002 [cited 2015 March 3]; Available from:
http://www.epa.gov/ogwdw/disinfection/tcr/pdfs/whitepaper tcr storage.pdf
[21] Grayman, W.M., L.A. Rossman, R.A. Deininger, J.F. Smith, and R. Schnipke, 1999,
Water Quality Modeling of Distribution System Storage Facilities, AWWARF.
[22] Kirmeyer, G.J., L. Kirby, B.M. Murphy, N. P.P., M. K., and L. J.L., 1999, Maintaining
Water Quality in Finished Water Storage Facilities, AWWARF.
[23] Angulo, F.J., S. Tippen, DJ. Sharp, BJ. Payne, C. Collier, I.E. Hill, T.J. Barrett, R.M.
Clark, E.E. Geldreich, H.D. Donnell, and D.L. Swerdlow, 1997, A community
waterborne outbreak of salmonellosis and the effectiveness of aboil water order,
American Journal of Public Health., 87(4), p. 580-584.
[24] Safe Drinking Water Program (SDWP). Waterborne Salmonella Outbreak in Alamosa,
Colorado, March and April 2008: Outbreak Identification, Response, and Investigation.
Water Quality Control Division. Colorado Department of Public Health and
Environment, Denver, CO. November 2009 [cited 2015 March 9]; Available from:
http s: //www.colorado.gov/pacific/sites/default/files/WQ-DW-Publications-Alamosa-
Outbreak-Investigation-Report.pdf
[25] Craun, G.F., J.M. Brunkard, J.S. Yoder, and V.A. Roberts, 2010, Causes of Outbreaks
Associated with Drinking Water in the United States from 1971 to 2006, Clinical
Microbiology Reviews, 23(3), p. 507-528.
[26] Centers for Disease Control (CDC), 2013, Surveillance for Waterborne Disease
Outbreaks Associated with Drinking Water and Other Nonrecreational Water — United
States, 2009-2010, MMWR, 62(35), p. 714-720.
[27] U.S. Environmental Protection Agency. Adherence of Chemical, Biological, and
Radiological Contaminants to Sediments Found in Water Storage Tanks. EPA/600/S-
14/224 2014 [cited 2015 Mardch 9]; Available from:
http://cfpub.epa.gov/si/sijublic file download.cfm?p download id=521345.
[28] U.S. Environmental Protection Agency, 2015, Report on Physical and Chemical
Properties of Drinking Water Storage Tank Sediments, in preparation.
[29] Lu, J., I. Struewing, S. Yelton, and A. N., 2015, Molecular Survey of Occurrence and
Quantity of Legionalla spp., Mycobacterium spp., Pseudomonas aeruginosa and Amoeba
Hosts in Municipal Drinking Water Storage Tanks, Journal of Applied Microbiology (in
press).
[30] Wu, J.H. and S.F. You, 2011, Validation of FLUENT Type Simulation of Heat Pump
Water Tank Containing PCM, 2011 3rd World Congress in Applied Computing,
Computer Science, and Computer Engineering (Ace 2011), Vol 3, 3, p. 321-329.
[31] Patel, T., L. Gill, and M.G. Faram, 2011, Grit Removal from Wastewater Using
Secondary Currents in Open-Channel Flow around Bends, Journal of Environmental
Engineering-Asce, 137(11), p. 1026-1039.
72
-------
[32] Dautova, L.S., I. Milanovic, and K.J. Hammad, 2012, CFD Study of the Effect of Jet
Placement on Flow Patterns Inside a Jet Stirred Tank., in 2072 ASEE Northeast Section
Conference, University of Massachusetts Lowell, April 27 - 28, 2012.
[33] Beheshti, A.A. and B. Ataie-Ashtiani, 2008, Analysis of threshold and incipient
conditions for sediment movement, Coastal Engineering, 55(5), p. 423-430.
[34] Delleur, J.W., 2001, New results and research needs on sediment movement in urban
drainage, Journal of Water Resources Planning and Management-ASCE, 127(3), p. 186-
193.
[35] Hager, W.H. and G. Oliveto, 2001, Simplified shields condition for sediment transport
inception, Hydraulics of Rivers Water Works and Machinery, Vol II, Theme D,
Proceedings, p. 264-269.
[36] Mantz, P. A., 1977, Incipient Transport of Fine Grains and Flakes by Fluids - Extended
Shields Diagram, Journal of the Hydraulics Division-ASCE, 103(6), p. 601-615.
[37] Motamedi, A., H. Afzalimehr, and V.P. Singh, 2010, Evaluation of aNovel Approach to
Determine the Critical Shields Stress, Journal of Hydrologic Engineering, 15(11), p. 892-
900.
[38] Huber, W.C. and R.E. Dickinson, 1988, Storm Water Management Model, Version 4:
User's Manual, EPA/600/3-88/00la, Environmtnal Research Laboratory, U.S.
Environmental Protection Agency, Athens, GA.
[39] Morsi, S.A. and Alexande.Aj, 1972, Investigation of Particle Trajectories in 2-Phase
Flow Systems, Journal of Fluid Mechanics, 55(Sep26), p. 193-&.
[40] Cushman-Roisin, B., 2013, Environmental Fluid Mechanics, John Wiley & Sons, Inc.,
New York.
[41] McNaughton, K.J. and C.G. Sinclair, 1966, Submerged Jets in Short Cylindrical Flow
Vessels, Journal of Fluid Mechanics, 25(2), p. 367-375.
[42] Dautova, L.S., I. Milanovic, and K.J. Hammad, Year, CFD Study of the Effect of Jet
Placement on Flow Patterns Inside a Jet Stirred Tank, in 2072 ASEE Northeast Section
Conference, University of Massachusetts Lowell, April 27-28, 2012.
73
-------
Appendix A: Hydraulic Model Evaluation
Turbulence models are necessary to capture the chaotic flow patterns caused by the inlet jet when
pumping water into the tank. This appendix evaluates different turbulence models available for
multiphase flow in the Fluent® software. Comparisons to experiments with water jets were used
to evaluate the conditions representative of water being injected into a tank. Water jets have been
studied and classified based on their Reynolds number in several reports [40, 41]. A fully turbulent
jet is considered to have a Reynolds number of greater than 3,000 in the inlet. The Reynolds
numbers for the inflow in this study are all greater than 3,000.
Two options for modeling turbulence in the Fluent® software exist:
1. k-co SST: This model solves the Reynolds-Averaged Navier-Stokes (RANS) equations
using an isotropic eddy viscosity value to solve for the Reynolds stress values. This
model is capable of solving turbulence equations in the near-wall region as well as in the
free-stream region necessary in this problem.
2. Scale-Adaptive Simulation: This model solves the unsteady-RANS equations which
accounts for fluctuations in the turbulent flow as the equations progress. This model
utilizes the Reynolds stress model which individually solves all of the Reynolds stress
values for the transport equations without using the isotropic eddy viscosity value used in
the RANS equations. This model can be more accurate in solving for small turbulent
eddies and may be appropriate for this type of modeling.
The solutions from these two turbulence models were compared with experimental tank mixing
images given in Roberts et al. [4, 6] in which the geometric tank configuration (GC02) matched
those of the current study. Simulations were performed assuming two-dimensional axisymmetry
using the k-co SST and Scale-Adaptive Simulation models to determine which turbulence model
would be best suited for the computational studies conducted in this report. The modeling
conditions for this comparison were as follows:
• Mixture multiphase solution model (to evaluate mixing behavior in the Fluent® software)
• Three phase mixture: air, water, and tracer (water)
o Tracer was injected into the tank with a volume fraction of 1 with a velocity of
2.33 m/s (fully turbulent jet)
• 50,000 grid elements
Both the k-co SST and scale adaptive turbulence models qualitatively captured the tracer mixing
patterns within the tank (Table 15 and Figure 54). It was seen that the simulated flow patterns
followed the general movement of the tracer concentrations in Roberts et al. [6]. The tracer
concentrations were not matched completely, probably due to unknown diffusion parameters and
not being able to accurately capture the small turbulent eddies forming in the jet region. However,
those factors are not expected to significantly impact analysis of particle resuspension in terms of
the general advective flow patterns and simulated shear stress along the bottom of the tank. It is
not necessary to capture the tracer diffusive mixing and concentrations, allowing for the use of the
simpler VOF multiphase model. In addition, because the advective flow patterns were similar
74
-------
between the two turbulence models, small turbulent eddies (modeled in the scale-adaptive
simulation) are neglected in favor of using the simpler k-co SST model. While turbulent eddies are
expected near the inlet jet during filling of a tank, they are not expected to be significant during
draining of the water tanks. Further support of the k-co SST model comes from Dautova et al. [42].
They studied the impacts of a submerged jet in water on an impingement plate. They compare the
k-co SST model to a Reynolds Stress model as well and determined that the k-co SST model
provided more accurate answers when compared to experimental data.
Table 15. Comparison of tracer concentrations as a function of time during injection into
a water-filled tank using two different turbulence models.
Time (s)
k-w SST Turbulence Model
Scale Adaptive Turbulence Model
9.506-01
9.00e-01
8.506-01
8.006-01
7.50e-01
7.00S-01
6.506-01
6.006-01
5.506-01
5.006-01
4.506-01
4006-01
3.50«-01
3.006-01
2.506-01
2.006-01
1.506-01
1.006-01
6.006-02
0.006+00
9.506-01
9.009-01
8.506-01
8.006-01
7.506-01
7.006-01
6.506-01
6.006-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
1,006-01
5.006-02
O.OOe+00
75
-------
Time (s)
k-w SST Turbulence Model
Scale Adaptive Turbulence Model
27
•LOOe+OO
9.506-01
9.006-01
a.506-01
8.006-01
7.506-01
7.006-01
6.506-01
6.006-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
I.OOe-01
5.006-02
O.OOertO
1.006+00
9.506-01
9.006-01
8.506-01
8.006-01
7.506-01
7.006-01
6.506-01
6.006-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
1.006-01
5.006-02
O.OOe+00
9.006-01
8.506-01
8.006-01
7.506-01
7.006-01
6.506-01
6.006-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
1.006-01
5.006-02
O.OOe+00
76
-------
Time (s)
k-w SST Turbulence Model
Scale Adaptive Turbulence Model
57
I
LGOe+OO
9.506-01
9.008-01
8.506-01
8.006-01
7.506-01
7.006-01
6.506-01
6.00e-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
1.006-01
5.006-02
O.OQe+00
1
LOOe+00
9.50S-01
9.006-01
8.506-01
B.OOe-01
7.506-01
7.006-01
6.506-01
6.006-01
5.506-01
S.OOe-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
1.00S-01
S.OOe-02
O.OOe+00
71
1.006+00
9.506-01
9.00e-01
50e-01
a.DOe-01
7.506-01
7.006-01
6.50e-01
6.006-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3006-01
2.506-01
2.006-01
1.506-01
I.OOe-01
5.006-02
0.00e«o
1.006+00
9506-01
9.006-01
8.506-01
8.006-01
7.50e-01
7.006-01
6.50e-01
S.OOs-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
1.00e-01
5.006-02
O.OOe+00
77
-------
Time (s)
k-w SST Turbulence Model
Scale Adaptive Turbulence Model
144
1.00«+00
9.50*-01
9.006-01
8.506-01
8.006-01
7.506-01
7.00e-01
6.50S-01
G.OOe-01
5.506-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2506-01
2008-01
1.506-01
1.006-01
5.006-02
O.OOe+00
tOOe+00
9.506-01
9.006-01
8.506-01
8.006-01
7.506-01
7.006-01
6.50e-01
5.50e-01
5.006-01
4.506-01
4.006-01
3.506-01
3.006-01
2.506-01
2.006-01
1.506-01
I.OOe-01
5.006-02
O.OOe+00
Figure 54. Tracer concentrations from Roberts et al. experiment ([6], Fig. 3.9) with 2.16
mis inflow jet on left (red is concentration, blue is low concentration).
To further evaluate the turbulence models, a CFD model with a submerged jet in a cylindrical
container was compared to an existing CFD/experimental verification study. Replicating this
study with the same type of jet flow and turbulence model (k-co SST) gives confidence that the
CFD results can predict jet behavior in a submerged liquid within a cylindrical tank.
Results from Dautova et al. [42] were recreated using the k-co SST model. The geometry was
replicated for the center line location. The solution was transient with the k-co SST turbulence
model included. The jet flow enters the tank through a pipe, impinges on the bottom of the tank
and then the flow develops within the tank. The model has complex flow features due to the jet
impingement thus giving confidence that the CFD model can predict complex jet flow behavior.
The results were compared to those obtained in the Dautova report. Dautova was able reproduce
the behavior of experimental work in the CFD calculations (Figure 55).
78
-------
120
100
40 -
20 -
20 40
60
Figure 55. Experimental (left) and simulated (right) contours of velocity magnitude in a
jet-stirred mixing tank (from Fig. 7 in [42]).
The jet profile in the CFD tank studies can also be compared to general jet information provided
in [40]. It is reported that turbulent jets submerged in a quiescent fluid have the same jet opening
angle regardless of velocity and diameter of line. This universal jet opening angle is 11.8°
(Figure 56). The jet opening angle for the CFD studies is -12°. This value is found from
looking at the jet velocity profile in Figure 57 and Figure 58 and applying the appropriate
trigonometric equations, see equation below.
1 w
9 = tarT1 —
h
Where 9 is the jet opening angle, w (1.04 m) is the half width of the jet velocity profile, and h
(4.7 m) is the height where the jet profile is taken (note that this height is measured from the x
value [5*(diameter of line)/2] since the initial jet radius is not zero).
The CFD jet opening angle is close to the universal jet opening angle reported in [40] for
turbulent, submerged jets. This gives confidence that the CFD model is predicting the jet profile
overall shape without inclusion of small turbulent eddies.
79
-------
r,
plane of
orifice
5d
x=
L
entrainment
of ambient fluid
Figure 56. Schematic of jet penetration into quiescent fluid [40], the jet opening angle of
11.8° is universal regardless of inlet velocity and line diameter
Figure 57. 3D CFD jet velocity (mis) in center line tank with inlet velocity of 0.33 m/s
80
-------
0.4
0.35
0.3
<0.25
£• 0.2
"u
O
•5 0.15
0.1
0.05
•Center Nozzle Tank
-202
Radial distance across tank (m)
Figure 58. Jet velocity (mis) profile at quarter height of tank along radial distance of tank
for 0.33 m/s jet.
The Volume of Fluid (VOF) method and Mixture method were two main candidates for the
multiphase modeling required in this study. However, since tracer mixing is not important to
particle resuspension, the Mixture method is not required. The VOF method tracks the sharp
interface between phases. The interface in this study was the region where the air and water meet.
The jet can penetrate into the air region and the solver will track this water intrusion on the air
region. This model also solves for the flow within the water region of the tank. This multiphase
model was used for both tank filling and draining scenarios.
Initially, it was hypothesized that a two-phase (air, water) model would not be required after a
certain water head level. At some point the jet will stop penetrating the air region and dissipate
before reaching the air-water interface for inflow boundary conditions. However, after running
comparison simulations, it was concluded that the free surface plays an important role in the flow
patterns within the tank. During filling the jet penetrates the free surface for the tank sizes
considered. A single phase solution would not be plausible for the tank filling scenario. A two-
phase transient solution was used for all tank filling scenarios.
The shear stress profiles along the bottom of a center line tank for a steady single phase vs. transient
two-phase draining solution are shown in Figure 59. Results show that for regions near the drain,
the simulated shear stresses are similar between the two models. Away from the drain, the
simulated shear stresses deviate between the different models. During draining, as the free surface
starts to drop the water at the free surface develops both axial and radial velocity components
causing recirculation zones to form. A single phase draining simulation does not capture this
physical process. However, for the purposes of simulating particle resuspension during draining,
a steady single-phase model, which is much more computationally efficient than than the transient
two-phase model, may be appropriate. A steady single-phase Solidworks™ model was used to
simulate particle resuspension in the small-scale tests, and comparisons were shown to be good
(see Section 3.3).
81
-------
n
a.
l.E+00
l.E-i
Steady vs. Transient Drainage Comparison
•Steady
•Transient
20s
•Transient
100s
•Transient
200s
5 10 15
Radial Distance from Center (m)
20
Figure 59. Shear stress profiles along bottom of tank with center line : comparison of
steady-state single-phase drainage solution compared to transient two-phase drainage
solution.
82
-------
Appendix B: Particle Size Distribution
A representative particle size distribution was provided by the Colorado Tower in Columbus, Ohio
[28], where a sample of sediment particles was sieved and weighed (see Figure 60). The total
weight of all particles was 72.202 g. Table 16 and Figure 61 provide data collected on the mass
fractions and size distributions. The mass fractions of each particle diameter bin were converted
to number fractions based on the following procedure in order to determine an appropriate particle
size distribution to be used in the CFD models.
The total mass of each bin was calculated by multiplying the total mass of all particles by the mass
fraction of the bin.
™* = mtot x mfii
Figure 60. Sediment particles from the Colorado Tower, Columbus, Ohio. The sample
was sieved and weighed to develop a sediment weight distribution.
Table 16. Diameter range and mass fraction of each particle bin.
Bin Number
Particle
Diameter (mm)
Mass Fraction
1
>2
.5279
2
1-2
.1372
3
.25-1
.1657
4
.106-.25
.856
5
.75-. 106
.277
6
.53-75
.255
7
.45-.S3
.171
8
<.45
.132
83
-------
.106 .075 .053 .045 pan
Sieve Size (mm)
Figure 61. Histogram of mass fraction against particle diameter.
A representative diameter for each bin was then determined by a mass-weighted average of the
maximum and minimum diameters of the bin. The representative diameters of the smallest and
largest bins were calculated based on the ranges of the closest bins in size:
max,2-7
>4 rr
rep,2-7
"rep.l = "min.l + \^-max,2
"rep, 8 ~ dmin,7 ~ y-'-max.J
(12)
(13)
(14)
(15)
The number of particles in each bin was determined by dividing the mass of each bin by a
representative unit mass:
mrpr> i = p-n
rep,i r 3
Lrep,i
(16)
(17)
84
-------
The number of particles in each bin was divided by the total number of particles to yield the
number fraction of each bin:
not
(18)
(19)
The resulting number fractions are shown in Table 17.
Table 17. Particle diameter range, mass fraction, and number fraction of each bin.
Bin Number
Particle Diameter
(mm)
Mass Fraction
Number Fraction
1
>2
.5279
3.83e-5
2
1-2
.1372
3.99e-5
3
.25-1
.1657
3.36e-4
4
.106-
.25
.856
.0122
5
.075-
.106
.277
.0579
6
.053-
.075
.255
.1504
7
.045-
.053
.171
.2686
8
<.045
.132
.5106
Very few large particles made up the total number of particles in the sample; however, they
contributed greatly to the total mass simply because each large particle is orders of magnitude
heavier than the smaller particles.
The complementary cumulative distribution function (CCDF) displayed in Figure 62 describes the
probability of occurrence of a particle with a diameter greater than or equal to a specific value on
the x-axis. The x-values of each data point represent the minimum diameter of the corresponding
bin (except for the smallest bin); the y-values represent the cumulative number fraction of the
corresponding bin and of all larger bins. For instance, the rightmost data point indicates that
approximately fifty percent of the particles in a sample will have diameters greater than 45
microns. The data points plotted on this CCDF are shown in Table 18.
Table 18. Probability of occurrence of a particle with a diameter greater than or equal to
the specified diameter. The displayed diameters are the endpoints of each bin (Table 16).
Particle
Diameter (mm)
Probability of
Occurrence
2
3.83e-5
1
7.82e-5
0.25
4.14e-4
0.106
0.0126
0.075
0.0704
0.053
0.2208
0.045
0.4894
85
-------
l.E+00
l.E-01
£• l.E-02
15
n
.Q
2 l.E-03
l.E-04
l.E-05
0.01
0.1 1
Particle Diameter (mm)
10
Figure 62. Complementary cumulative distribution function of particle diameter based on
number fraction.
Implementing the Colorado Tower particle distribution in a CFD model would require a
considerable number of particles to simulate even a few large particles for the operational study.
Therefore, an injection of 6,000 particles consisting of equal numbers of 1-mm, 100-micron, and
10-micron diameter particles was used instead. Each division of particles was distributed across
the bottom face of the tank in the same locations, resulting in 2,000 different initial positions and
3 different-sized particles at each position. The initial particle positions were scaled towards the
tank center since the particles further away from the inlet/outlet line were less likely to experience
resuspension and were thus of less interest. The positions were determined by the following
equation:
R =
n"
1000
- + 0.3048
(20)
Where
a =
log[1000(19.2024 - .3048) - 1]
log 2000
R is the radial distance of the particle from the tank center (mm), and n is the particle index (1-
2000). For n = 1, R = 0.306 m, and for n = 2000, R = 19.2 m. Note that the bottom wall spans a
radial distance of 0.3048 meters to 19.2024 m. This scaled distribution allowed for a characteristic
representation of particle suspension and trajectories for a wide range of particle sizes. Figure 63
shows the initial distribution of particles along the tank bottom. The density of the particles was
assumed to be 2650 kg/m3 (silica sand).
86
-------
450
400
350
300
Initial Distribution of Particles along Tank Bottom at Beginning of Cycle
250
Q_
•6
200
150
100
50
6 8 10 12 14
Radial Distance from Tank Center (m)
Figure 63. Bar graph displaying initial distribution of particles along bottom wall.
87
-------
Appendix C: Surface Tension Model for Particles at the Air-Water Interface
Background
The operational study of particle movement (Section 2.3) is designed to determine the general
positions along the bottom of water storage tanks at which sediment particles are most likely to
be deposited or resuspended into the fluid flow. The trajectories of the resuspended particles are
then tracked until they are re-deposited. However, using the initial model, some of these particles
were observed to pass above the water surface into the air region (Figure 64). The Fluent®
software does not by default include forces due to surface tension at the air-water interface.
Therefore, a surface tension model was created and implemented by writing a user-defined
function (UDF).
Figure 64. Snapshot of model displaying particles suspended above the water surface.
Blue cells indicate water phase; white cells indicate air phase. Blue dots represent 10-
micron-diameter particles; green dots represent 100-micron-diameter particles.
Implementation
Figure 65 shows a schematic of the surface tension forces acting on a spherical particle at the
water/air interface. The UDF is defined by a macro that loops through all particles in the domain
at every particle time step. For each particle in the domain, the UDF instructs the Fluent®
software to find the cell in which the particle is located and determine the air volume fraction of
the cell (see Figure 66). If the air volume fraction is 0, indicating that the cell contains only
water, then no additional force is applied to the particle; if the air volume fraction is greater than
88
-------
0, then the particle undergoes a downward force due to surface tension (Fs,y). The magnitude of
the force is calculated as described below.
F = y(27ir) V V F = F cos(6)
s ' v ' s,y s x '
Figure 65. Schematic of surface tension forces on spherical particle at water/air
interface.
where
Y Surface tension of water at 20 °C (0.073 N/m)
Fs Force exerted by surface tension on the particle (N)
Fs,y Downward force exerted by surface tension on the particle (N)
r Radius of the particle (m)
P Perimeter of particle (m)
6 Contact angle between liquid and particle with respect to downward direction
The surface tension, y, is defined as a force per unit length. Thus, the maximum force exerted by
surface tension on the particle is equal to the surface tension times the perimeter of the particle, P:
Fs = y(27rr) (21)
The downward component of the surface tension force is obtained by multiplying the surface
tension force by the cosine of the contact angle shown in Figure 65:
Fs.y = Fs cos(9) = Y (27ir) cos(9) (22)
Assuming the contact angle equals zero (9 = 0), the maximum downward component of the surface
tension force is as follows:
Fs,y = Fs cos(9) = Y(2rcr) (23)
This model has been implemented in a user-defined function in the Fluent® software and has
successfully prevented particles from leaving the water phase during the operational simulations.
89
-------
Figure 66. Contour map of air volume fraction illustrating that cells at the air-water
interface are not strictly water or strictly air. The red cells indicate an intermediate air
volume fraction between 0 and 1.
90
-------
Appendix D: Justification for Using a 2D Axisymmetric Model
Initially, the operational study employed a 3D model with reflection symmetry, as in the
parametric study. However, preliminary testing revealed that using a 3D model would require
substantial computation time to provide an accurate representation of the particle trajectories for
sufficient flow time. Therefore, a 2D axisymmetric model was proposed instead due to the
significantly less computational memory required. The tank size, tank configuration, and inflow
rate used in this section are the same as for the operational study (Table 19). Two metrics were
used to compare the 2D and 3D models: wall shear stress on the tank bottom (see Figure 67) and
velocity pathlines (see Figure 68 and Figure 69).
The wall shear stress curves of both models show the greatest inconsistencies in two major regions:
around 5-10 meters away from the line and very close to the line. Five to ten meters away from
the line, the shear stresses predicted by both models failed to initiate movement of any of the
simulated particles, so the discrepancy between models can be neglected. Very close to the
inlet/outlet line (radial distance less than -2-3 meters), the discrepancy is considerably less.
The velocity pathlines are largely similar except in the region far away from the inlet/outlet and
outside the recirculation zone. However, preliminary testing revealed that particles do not travel
past the recirculation zone; therefore, the particle trajectories in both models should match well.
The 2D model was thus employed for the operational study, with a time step of .002 s to allow the
Fluent® software to finely capture fluid flow from mesh cell to mesh cell.
Table 19. Tank model and flow rate used in the 2D and 3D comparison. These values
were used in the actual operational study as well.
Parameter
Value
Tank Diameter
m(ft)
38.4m (126 ft)
Tank Height
m(ft)
9.8m (32 ft)
Line Diameter
m (in)
0.6 1m (24 in)
Inlet Flow Rate
(m/s)
0.631
Line Placement
Center
91
-------
l.OE+01
l.OE+00
Comparison of 2D and 3D Models
l.OE-06
l.OE-07
10 15
Radial Distance from Center (m)
-3D-20s
-3D-50s
3D-90s
-2D-20S
•2D-50S
2D-90S
2D-
180s
20
Figure 67. Comparison of shear stresses on tank bottom between a 2D and 3D model.
92
-------
1 OOe-01
9.506-02
9.00e-02
8.50e-02
S.OOe-02
7.50e-02
7.006-02
6.506-02
6.006-02
5.506-02
5.006-02
4.506-02
4.006-02
3.506-02
3.00e-02
2.506-02
2.006-02
1.506-02
LOOe-02
S.OOe-03
0.006+00
Water-Air
Interface
Pathlines Colored by Velocity Magnitude (mixture) (m/s) (Time=9.0000e+01)
Figure 68. Velocity pathlines at 90 seconds for 3D Model
93
-------
1.008-01
9.506-02
9.00e-02
8.50e-02
8.00e-02
7.506-02
7.00e-02
6.506-02
6.006-02
5.50e-02
5.006-02
4 50e-02
4.006-02
3.50e-02
3.006-02
2 506-02
2.00e-02
1.506-02
1 .OOe-02
5.006-03
O.OOe+00
Water-Air
Interface
Pathlines Colored by Velocity Magnitude (mixture) (m/s) (Time=9.0000e+01)
Figure 69. Velocity pathlines at 90 seconds for 2D Model
94
-------
Appendix E: Additional Results from Parametric Analysis
Table 20 presents a summary of the simulated regions along the bottom of the tank wall that were
susceptible to particle resuspension (shown in red) according to the Beheshti model assuming a
particle density of 2650 kg/m3. If the simulated shear stress along the bottom resulted in a
Beheshti movability number that was greater than the critical movability number, indicating a
potential for particle resuspension, the cell was colored red. The different cases shown in Table
20 correspond to the parametric studies discussed in Section 2.4.
Table 20. Summary of simulated regions susceptible to particle resuspension (shown in
red) based on bottom-wall shear stress for the different parametric cases (Table 4)
Case
Filling
1
High flow
Large
diameter
inlet
Near wall
Filling
2
Low flow
Small
diameter
inlet
Center
Area susceptible to
resuspension in red
(1 mm particle size)
None
Area susceptible to
resuspension in red
(0.1 mm particle size)
Area susceptible to
resuspension in red
(0.01 mm particle size)
95
-------
Area susceptible to
Case resuspension in red
(1 mm particle size)
Filling
3
High flow
Small
diameter
inlet
Near wall
Filling
4
High flow
Large
diameter
inlet
Center
Filling
5
Low flow
Large
diameter
inlet
Center
None
Area susceptible to
resuspension in red
(0.1 mm particle size)
Area susceptible to
resuspension in red
(0.01 mm particle size)
96
-------
Area susceptible to
Case resuspension in red
(1 mm particle size)
Filling
6
Low flow
Small
diameter
inlet
Near wall
Filling
7
Low flow
Large
diameter
inlet
Near wall
Filling
8
High flow
Small
diameter
inlet
Center
None
None
None
Area susceptible to
resuspension in red
(0.1 mm particle size)
None
None
Area susceptible to
resuspension in red
(0.01 mm particle size)
97
-------
Area susceptible to
Case resuspension in red
(1 mm particle size)
Draining
1
High flow
Large
diameter
inlet
Center
Draining
2
High flow
Large
diameter
inlet
Near wall
Draining
3
Low flow
Small
diameter
inlet
Near wall
None
Area susceptible to
resuspension in red
(0.1 mm particle size)
None
Area susceptible to
resuspension in red
(0.01 mm particle size)
98
-------
Area susceptible to
Case resuspension in red
(1 mm particle size)
Draining
4
High flow
Small
diameter
inlet
Center
Draining
5
Low flow
Small
diameter
inlet
Center
Draining
6
Low flow
Large
diameter
inlet
Near wall
None
None
Area susceptible to
resuspension in red
(0.1 mm particle size)
None
None
Area susceptible to
resuspension in red
(0.01 mm particle size)
99
-------
Case
Draining
7
High flow
Small
diameter
inlet
Near wall
Draining
8
Low flow
Large
diameter
inlet
Center
Area susceptible to
resuspension in red
(1 mm particle size)
None
Area susceptible to
resuspension in red
(0.1 mm particle size)
None
Area susceptible to
resuspension in red
(0.01 mm particle size)
100
-------
United States
Environmental Protection
Agency
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
------- |