Reducing Booster-Pump-Induced Contaminant Intrusion in Indian Water Systems with a

advertisement
Reducing Booster-Pump-Induced Contaminant
Intrusion in Indian Water Systems with a
Self-Actuated, Back-Pressure Regulating Valve
by
David Donald James Taylor
BASc., University of Toronto (2011)
Submitted to the Department of Mechanical Engineering
in partial fulfillment of the requirements for the degree of
Master of Science in Mechanical Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2014
c Massachusetts Institute of Technology 2014. All rights reserved.
β—‹
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Mechanical Engineering
May 9, 2014
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alexander H. Slocum
Pappalardo Professor of Mechanical Engineering
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
David E. Hardt
Chairman, Department Committee on Graduate Theses
2
Reducing Booster-Pump-Induced Contaminant Intrusion in
Indian Water Systems with a Self-Actuated, Back-Pressure
Regulating Valve
by
David Donald James Taylor
Submitted to the Department of Mechanical Engineering
on May 9, 2014, in partial fulfillment of the
requirements for the degree of
Master of Science in Mechanical Engineering
Abstract
Intermittently-operated water systems struggle to equitably and effectively distribute
clean water to customers.
One common customer response to intermittency is to
supplement the water system’s pressure by using a household, or residential, booster
pump. When such booster pumps are directly connected to the water utility’s supply
pipe, without an underground isolation tank (sump), they often induce negative pressure in the supply pipe which increases the flow rate. Unfortunately, where leakage
rates are high, this negative pressure also increases the risk of contaminant intrusion.
This thesis presents the iterative design and field testing of a patent-pending,
full-bore, back-pressure regulating valve. The valve’s simple mechanism relies on a
stabilized collapsing tube, or ‘Starling Resistor,’ which when installed at a customer’s
connection, controls the flow rate and prevents booster pumps from creating negative
pressure in the supply pipe. In collaboration with the Delhi Jal Board and several
private partners, the valve’s performance was verified in two rounds of field trials in
neighborhoods of New Delhi, India including Pitampura, Azad Market, Vivek Vihar,
Malvia Nagar, and Vasant Vihar.
Using a crossover study, the valve was found to
reduce the total contamination risk across all 19 tested houses during supply times by
a median of 80%. The valve prevented 96% of pressure below -1 meter and an average
of 53 minutes per day, per connection of total negative pressure.
In an estimated
worst-case scenario for contaminant intrusion, the presence of the valve reduced the
contamination risk by two orders of magnitude at six customer connections — enough
to correspond to significant reductions in health risks.
Thesis Supervisor: Alexander H. Slocum
Title: Pappalardo Professor of Mechanical Engineering
3
4
Acknowledgments
[Noah’s] Ark should float in Grade 1, and by the time students leave Grade
12..., someone has to have sunk the ark for them!
(Denis Lamoureux)
I am eternally grateful to both my parents for the blessing of both floating and
sinking Noah’s Ark, simultaneously and perpetually. You have role modeled for me
that intellectual rigor is the beginning of faith, not its end.
You have given me a
sticky and wrestling faith; thank you.
To my sister for promising to fly down to Boston and slap me in the face if I
became pretentious, to my brother for sharing your love and humour with me always,
and to both of you, my best friends: front seats should always be designed for three.
To my former colleagues at HydraTek & Associates, thank you for teaching me so
much of what I know about water and water systems; you allowed me to approach
this problem with the confidence of familiarity.
To Professor Gutowski, thank you for allowing me to defer my admission to MIT,
my thesis would not have been what it is without the experiences I was able to have
before coming here.
To Professor Slocum, I am grateful for your wise suggestions, great design advice,
but most of all for your fierce care and compassion for your students.
To all my fellow graduate students in PERG, I often tell people my thesis and
design are like the story of Stone Soup: you were frequent contributers of ideas and
suggestions to this soup; it could not have happened without you.
For all of my
lunch-table friendships, I am especially grateful to Anthony Wong for persistently
and warmly insisting that “everyone needs to eat; come have lunch.”
I very grateful to the MIT Tata Center and to the Tata Trust for the funding,
freedom, and fellowship that enabled this research. To Dr. Nevan Hanumara, thanks
for caring deeply about my project, always striving to connect me to the right person,
and carefully editing many documents for me.
To Mark Belanger and the Edgerton Student Shop, you enabled ideas to take form
5
and be tested; thank you Mark, for your humour, warmth, and great help.
Thank you to the four Sloan Business students who helped develop a commercialization plan and facilitated many great connections in Delhi: Michael Alley, Ross
Frasier, Deepika Goyal, and Horacio Leal. Additional thanks go to Sally Miller and
Wesley Cox who helped assemble the alpha prototypes.
To my supporters in India, I owe a special set of thank-yous, for as a foreigner I
would neither have identified this project, nor been able to test my product without
your help and gracious hospitality.
This topic and project would not have seen the light of day without the help,
advice, and guidance of Mr. Manish Kumar, of Tata Consulting Engineers.
I am
grateful to you and the whole TCE Delhi office for sharing their your, perspective
and expertise with me.
I am indebted to Mr. Babbar, from the Delhi Jal Board (DJB), for your patience,
wisdom, and support for this project from its inception.
To the DJB’s Pitampura
staff, you were instrumental in gathering the January field data.
Specifically I am
grateful to Mr. Thakur, Mr. Sandeep Sharma, and Mr. Yogendra Singh for not only
connecting me to Mr. PK Jain, but for also allowing two valves to be installed on
your office connections. I am additionally grateful to Mr. PK Jain and especially to
Mr. Tej Pal who facilitated nine household connections inside the DJB colony.
I owe an enormous thanks to each and every household that gave their informed
consent to the water utility, without your permission, this thesis would not have been
possible. Specifically, I am grateful to Mr. Ankush, for not only allowing me to test
the valve on your house, but for introducing me to all your neighbors and making all
five Azad Market connections possible. Further, I am grateful to Mr. Manoj, your
willingness to innovate and try new products is contagious — your East Delhi house
continues to have a working alpha prototype connected to it; thank you.
To the supporters of the March field trial in South Delhi, I am grateful for the
partnership and field assistance, specifically Mr. Mappa and especially to Mr. Kévin
Beulé and Mr. Ashish Shenoy for your patient, enduring, and innovative help. Additionally I am grateful to Mr. El Hassane and to Mr. Saurabh, for your insights on
6
the project and assistance gathering data.
7
THIS PAGE INTENTIONALLY LEFT BLANK
8
Contents
1
Introduction
23
2
Background
25
2.1
Contaminant intrusion
2.3
2.4
3
4
25
2.1.1
Leakage rates indicate the pathway size for contaminant intrusion 26
2.1.2
Contaminants may be present because of the unplanned locations of water pipes . . . . . . . . . . . . . . . . . . . . . . . .
28
Pipe pressure can deter contamination
. . . . . . . . . . . . .
28
Intermittent water systems . . . . . . . . . . . . . . . . . . . . . . . .
30
2.2.1
. . . . . . . . . . . . . .
31
Residential booster pumps . . . . . . . . . . . . . . . . . . . . . . . .
33
2.3.1
Booster pump configurations . . . . . . . . . . . . . . . . . . .
34
Upgrading intermittent water systems and its effect on booster pumps
39
2.4.1
Incremental improvements in intermittent supply
. . . . . . .
39
2.4.2
A transition to continuous supply . . . . . . . . . . . . . . . .
40
2.1.3
2.2
. . . . . . . . . . . . . . . . . . . . . . . . . .
Effects of intermittent water supplies
Prior art
43
3.1
Household contamination prevention devices . . . . . . . . . . . . . .
43
3.2
Self-actuated, full-bore pressure regulating valves
. . . . . . . . . . .
46
3.3
Collapsible tubes or ‘Starling Resistors’ . . . . . . . . . . . . . . . . .
48
The iterative design of a self-actuating, full-bore, back-pressure regulating valve
51
9
4.1
4.2
Preliminary concepts and their evaluation
. . . . . . . . . . . . . . .
52
4.1.1
Keeping upstream pressure positive . . . . . . . . . . . . . . .
52
4.1.2
Minimizing hydraulic losses
54
4.1.3
Concept generation and selection
. . . . . . . . . . . . . . . .
55
Early iterations: searching for stable throttling . . . . . . . . . . . . .
55
4.2.1
Full factorial experiment to determine insert geometry and tube
choice
4.3
4.4
5
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
Alpha prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
4.3.1
Durability . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
4.3.2
Withstanding system pressure . . . . . . . . . . . . . . . . . .
69
4.3.3
Tamper resistant
70
4.3.4
Safe for drinking water contact
. . . . . . . . . . . . . . . . .
71
4.3.5
Laboratory fatigue test . . . . . . . . . . . . . . . . . . . . . .
71
4.3.6
Field test
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
74
4.3.7
Improvements needed . . . . . . . . . . . . . . . . . . . . . . .
74
Beta prototype
. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
75
4.4.1
Increasing the valve’s toughness . . . . . . . . . . . . . . . . .
77
4.4.2
Preventing set-point drift . . . . . . . . . . . . . . . . . . . . .
79
4.4.3
Detailed design of the beta prototype . . . . . . . . . . . . . .
82
4.4.4
Laboratory fatigue test . . . . . . . . . . . . . . . . . . . . . .
85
4.4.5
Field test
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
4.4.6
Improvements needed . . . . . . . . . . . . . . . . . . . . . . .
87
Experimental and numerical methods
89
5.1
89
Experimental methods
. . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.1
Laboratory setup and experimental method
. . . . . . . . . .
89
5.1.2
Fatigue test setup . . . . . . . . . . . . . . . . . . . . . . . . .
92
5.1.3
Field testing setup and method
. . . . . . . . . . . . . . . . .
92
5.2
Hydraulic notation and units . . . . . . . . . . . . . . . . . . . . . . .
98
5.3
Quantifying the intrusion risk
98
. . . . . . . . . . . . . . . . . . . . . .
10
5.4
5.3.1
Water leakage and its pressure dependency . . . . . . . . . . .
100
5.3.2
Quantifying the potential for contaminant intrusion . . . . . .
101
5.3.3
Defining the contaminant intrusion ratio
. . . . . . . . . . . .
106
5.3.4
Contaminant concentration at the consumer’s tap . . . . . . .
107
5.3.5
Connecting contaminant intrusion to health risks
. . . . . . .
109
Other numerical methods . . . . . . . . . . . . . . . . . . . . . . . . .
110
5.4.1
Monte Carlo sensitivity study
110
5.4.2
Determining upstream system pressure during booster pump
events
5.4.3
6
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
111
Determining the supply hours of intermittently supplied houses 113
Results
115
6.1
Effect of the valve on a single pump cycle . . . . . . . . . . . . . . . .
117
6.2
Case study 1: applying the metrics to a house in Old Delhi (J02)
. .
119
. . . . . . . . . . . . . . . . . .
119
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
119
6.3
6.2.1
Duration of negative pressure
6.2.2
CIR
6.2.3
Health risk
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
121
Sources of uncertainty in the data . . . . . . . . . . . . . . . . . . . .
123
6.3.1
Variability in baseline characteristics of customer connections
123
6.3.2
Case study 2: the effects of variability on a house in Lado Sarai
(M05)
6.4
6.5
6.6
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123
Total aggregate impact of the valve . . . . . . . . . . . . . . . . . . .
127
6.4.1
Duration of negative pressure
. . . . . . . . . . . . . . . . . .
127
6.4.2
CIR
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
129
Aggregate impact of the valve on continuously-supplied houses . . . .
129
6.5.1
Reduction in the duration of negative pressure . . . . . . . . .
132
6.5.2
Reduction in CIR . . . . . . . . . . . . . . . . . . . . . . . . .
135
6.5.3
Reduction in health risk
138
. . . . . . . . . . . . . . . . . . . . .
Aggregate impact of the valve on intermittently supplied houses
6.6.1
Intermittent connections require data filtering
11
. . .
138
. . . . . . . . .
139
6.7
6.8
7
6.6.2
Duration of negative pressure during supply hours . . . . . . .
141
6.6.3
CIR during supply hours . . . . . . . . . . . . . . . . . . . . .
144
6.6.4
Health risk during supply hours . . . . . . . . . . . . . . . . .
148
Valve’s aggregate impact whenever water was being supplied . . . . .
150
6.7.1
Duration of negative pressure
. . . . . . . . . . . . . . . . . .
150
6.7.2
CIR
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
150
6.7.3
Health risk
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
The necessity of a pilot study and recommendations for its scoping
.
Conclusions and future work
153
154
157
Bibliography
160
A Measuring the precision and accuracy of the pressure loggers
167
A.1
Measurement method . . . . . . . . . . . . . . . . . . . . . . . . . . .
167
B
Other ideas for future work
171
C
Impact of the valve on each house
173
12
List of Figures
2-1
Sewage and water pipes cross contaminating . . . . . . . . . . . . . .
29
2-2
Degrading intermittency of supply in four large Indian cities
. . . . .
31
2-3
Range of supply hours in select cities in India
. . . . . . . . . . . . .
32
2-4
A typical household pipe configuration and why booster pumps are
required
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
2-5
RBP configuration 1: official booster pump connection
. . . . . . . .
35
2-6
RBP configuration 2: booster pump before and after the sump . . . .
36
2-7
RBP configuration 3: booster pump is directly connected without a
sump . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
3-1
Air gap as a method for stopping negative pressure
. . . . . . . . . .
44
3-2
Standard vacuum breaker
. . . . . . . . . . . . . . . . . . . . . . . .
45
3-3
Barometric loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
3-4
Self-actuated back-pressure valve with diaphragm
. . . . . . . . . . .
47
3-5
Self-actuated full-bore valve
. . . . . . . . . . . . . . . . . . . . . . .
48
4-1
Overly complex variable set point valve . . . . . . . . . . . . . . . . .
54
4-2
Bend and constriction losses in some types of commercial valves
. . .
54
4-3
Structured brainstorming sketches, part 1 . . . . . . . . . . . . . . . .
56
4-4
Structured brainstorming sketches, part 2 . . . . . . . . . . . . . . . .
57
4-5
Selected concept: a deflating tube as a back pressure valve
. . . . . .
58
4-6
Detail of open and closed state of collapsible tube . . . . . . . . . . .
58
4-7
Observed instability in unmodified tube . . . . . . . . . . . . . . . . .
58
4-8
Three prototypes attempting to remove instability . . . . . . . . . . .
60
13
4-9
Optimal insert shape . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
4-10 Optimal insert cross-section geometry . . . . . . . . . . . . . . . . . .
61
4-11 Linear Opening insert geometry . . . . . . . . . . . . . . . . . . . . .
61
4-12 Linear Side Profile insert geometry
61
. . . . . . . . . . . . . . . . . . .
4-13 Total seconds of flutter for each design parameter
. . . . . . . . . . .
62
4-14 Minimum upstream pressure allowed for each design parameter . . . .
62
4-15 Bill of materials and exploded view of alpha prototype
. . . . . . . .
65
. . . . . . . . . . . . . . . . . . . . . . .
66
4-17 Alpha prototype insert part drawing; units of inches . . . . . . . . . .
67
4-18 Alpha prototype’s strain transition layers . . . . . . . . . . . . . . . .
68
4-19 Detail of alpha prototype’s two possible external support mechanisms
70
4-16 3D CAD of alpha prototype
4-20 Process steps in the outsourced casting of the urethane inserts for the
alpha prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
72
4-21 Identifying where fatigued prototypes leaked . . . . . . . . . . . . . .
73
4-22 Detail of failure location and suggested cause in fatigue test
. . . . .
73
. . . . . . . . . . . . . . . . . . .
74
4-23 Alpha prototypes after field testing
4-24 Alpha prototypes’ field test required a flexible hose to reduce the shear
load on the valve; from bottom to top, pressure logger, alpha prototype,
check valve to air, flexible hose, booster pump
. . . . . . . . . . . . .
4-25 55% of alpha field units failed in shear at the threads
. . . . . . . . .
75
76
4-26 Alpha prototype’s set-point drift because of a slow leak into the reference chamber
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
76
4-27 The peel force was dramatically reduced by turning the membrane
inside-out at each end of the middle layer . . . . . . . . . . . . . . . .
80
4-28 Section views of the beta prototype . . . . . . . . . . . . . . . . . . .
81
4-29 Beta prototype’s exploded view
. . . . . . . . . . . . . . . . . . . . .
82
. . . . . . . . . . . . . . . . . . . .
83
. . . . . . . . . . . . . . . . . . . . . . .
83
4-32 Beta prototype’s middle layer part drawing . . . . . . . . . . . . . . .
84
4-33 Field modification of the beta prototype
87
4-30 Beta prototype’s assembly stages
4-31 Beta prototype’s outer pipe
14
. . . . . . . . . . . . . . . .
5-1
Laboratory test setup . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
5-2
Laboratory experimental setup photo . . . . . . . . . . . . . . . . . .
91
5-3
Layout of the fatigue test; ‘R’ are the two regulators; pressures were
typically set to 10m and 17m
5-4
. . . . . . . . . . . . . . . . . . . . . .
93
Built fatigue testing apparatus; blue tubes carry 17m pressure for the
inside of the valves; pink tubes carry 10m pressure for the outside of
the valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
93
5-5
Pressure loggers from Hydreka . . . . . . . . . . . . . . . . . . . . . .
94
5-6
Field test schematic and predicted pressure performance
. . . . . . .
95
5-7
Three prototypes attempting to remove instability . . . . . . . . . . .
95
5-8
Field equipment arrangement for beta prototype; currently taking control measurements — valve is bypassed . . . . . . . . . . . . . . . . .
96
Spacial modeling of contaminants near leaky pipes . . . . . . . . . . .
100
5-10 Pressure drop along the service connection . . . . . . . . . . . . . . .
104
5-11 Two examples of upstream pressure estimation . . . . . . . . . . . . .
112
5-12 Determining supply times
. . . . . . . . . . . . . . . . . . . . . . . .
114
6-1
Effect of the valve on different types of house connections . . . . . . .
118
6-2
Negative pressure reduction example from J02 in Old Delhi . . . . . .
120
6-3
Contamination risk of J02
. . . . . . . . . . . . . . . . . . . . . . . .
122
6-4
How the valve caused a longer duration of negative pressure at M05 .
125
6-5
Evidence of a mean shift in the sensor at M05
. . . . . . . . . . . . .
126
6-6
Severe negative pressure reduction . . . . . . . . . . . . . . . . . . . .
128
6-7
Inconclusive impact of duration of negative pressure . . . . . . . . . .
130
6-8
Inconclusive impact on CIR
. . . . . . . . . . . . . . . . . . . . . . .
131
6-9
Time spent at negative pressure on continuous connections . . . . . .
132
5-9
6-10 Detail of pressure history at J07; highlights the anomaly during the
control measurement
. . . . . . . . . . . . . . . . . . . . . . . . . . .
134
6-11 Time spent at negative pressure on continuous connections; adjusted
for J07 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
134
6-12 Accounting for the pressure anomaly in M01 . . . . . . . . . . . . . .
135
6-13 Evidence of a drift in the valve’s set point at M01 . . . . . . . . . . .
136
6-14 Valve’s ability to change CIR values for continuously pressurized houses 137
6-15 Variability in the supply pressure to intermittent houses was larger
than the effect of the valve . . . . . . . . . . . . . . . . . . . . . . . .
141
6-16 Inconclusive impact on negative pressure duration on intermittentlysupplied houses
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6-17 Inconclusive impact on total CIR on intermittently-supplied houses
.
142
143
6-18 Negative pressure reduction during supply hours on intermittent houses 145
6-19 Unusual performance of M04 . . . . . . . . . . . . . . . . . . . . . . .
146
6-20 CIR reduction during supply hours on intermittent houses; red dots
are expected values, red error bars are 99% range and blue error bars
are the 95% range
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
6-21 Reduction of negative pressure during supply times
6-22 Reduction of CIR during supply times
147
. . . . . . . . . .
151
. . . . . . . . . . . . . . . . .
152
A-1
Accuracy and precision or pressure loggers
. . . . . . . . . . . . . . .
168
C-1
Pressure summary: Azad Market: J01
. . . . . . . . . . . . . . . . .
174
C-2
Pressure summary: Azad Market: J02
. . . . . . . . . . . . . . . . .
175
C-3
Pressure summary: Azad Market: J03
. . . . . . . . . . . . . . . . .
176
C-4
Pressure summary: Azad Market: J04
. . . . . . . . . . . . . . . . .
177
C-5
Pressure summary: Azad Market: J05
. . . . . . . . . . . . . . . . .
178
C-6
Pressure summary: Pitampura: J06 . . . . . . . . . . . . . . . . . . .
179
C-7
Pressure summary: Pitampura: J07 . . . . . . . . . . . . . . . . . . .
180
C-8
Pressure summary: Vivek Vihar: J08 . . . . . . . . . . . . . . . . . .
181
C-9
Pressure summary: Pitampura: J09 . . . . . . . . . . . . . . . . . . .
182
C-10 Pressure summary: Pitampura: J10 . . . . . . . . . . . . . . . . . . .
183
C-11 Pressure summary: Pitampura: J11 . . . . . . . . . . . . . . . . . . .
184
C-12 Pressure summary: Pitampura: J12 . . . . . . . . . . . . . . . . . . .
185
C-13 Pressure summary: Vasant Vihar: M01 . . . . . . . . . . . . . . . . .
186
16
C-14 Pressure summary: Vasant Vihar: M02 . . . . . . . . . . . . . . . . .
187
C-15 Pressure summary: Shivalik: M03 . . . . . . . . . . . . . . . . . . . .
188
C-16 Pressure summary: Shivalik: M04 . . . . . . . . . . . . . . . . . . . .
189
C-17 Pressure summary: Lado Sarai: M05
. . . . . . . . . . . . . . . . . .
190
C-18 Pressure summary: Lado Sarai: M06
. . . . . . . . . . . . . . . . . .
191
C-19 Pressure summary: Lado Sarai: M07
. . . . . . . . . . . . . . . . . .
192
17
THIS PAGE INTENTIONALLY LEFT BLANK
18
List of Tables
2.1
Water balance definitions . . . . . . . . . . . . . . . . . . . . . . . . .
26
2.2
CPHEEO static pressure requirements
. . . . . . . . . . . . . . . . .
38
4.1
Initial functional requirements and design parameters . . . . . . . . .
52
4.2
Logic of a back-pressure regulating valve with a set point of X
. . . .
52
4.3
Design variables considered in search of stable operation
. . . . . . .
59
4.4
Insert geometry test matrix; repeated for silicone and latex soft tube
61
4.5
Alpha prototype’s functional requirements and design parameters
. .
64
4.6
Alpha fatigue test matrix: tube material and inner fillet radius . . . .
72
4.7
Beta prototype’s functional requirements and design parameters . . .
78
4.8
Beta fatigue test matrix: clearance for membrane and the insert’s inner
fillet radius
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
5.1
Laboratory measurement equipment . . . . . . . . . . . . . . . . . . .
91
5.2
House connection characteristics . . . . . . . . . . . . . . . . . . . . .
97
5.3
The assumptions behind Equation 5.14 . . . . . . . . . . . . . . . . .
108
5.4
The effects of Equation 5.14’s assumptions . . . . . . . . . . . . . . .
108
5.5
WHO’s example thresholds for
E. coli
contamination risk . . . . . . .
110
5.6
Assumptions required to connect CIR values to health risks . . . . . .
110
5.7
Allowable IC and CIR levels . . . . . . . . . . . . . . . . . . . . . . .
111
5.8
Measured supply times . . . . . . . . . . . . . . . . . . . . . . . . . .
113
6.1
Summary of negative pressure reduction on continuously supplied houses136
6.2
Summary of contamination reduction on continuously supplied houses
19
137
6.3
Reduction in health risk at continuously supplied houses
. . . . . . .
138
6.4
Booster pump induced pressure characteristics on intermittent houses
140
6.5
Duration of negative pressure at intermittently-supplied houses during
supply hours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.6
Summary of contamination reduction on intermittently supplied houses,
during supply hours . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.7
148
148
Reduction in health risk at intermittently supplied houses, during supply hours
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
149
6.8
Summary of negative pressure reduction during supply hours . . . . .
151
6.9
Summary of contamination reduction during supply hours
. . . . . .
153
6.10 Reduction in health risk during supply hours . . . . . . . . . . . . . .
153
6.11 Required supply pressure to limit the valve’s flow reduction . . . . . .
154
A.1
169
Total variation in the logging equipment
20
. . . . . . . . . . . . . . . .
Nomenclature
𝑄𝐼
3
Average Daily Intrusion Volume [π‘š /π‘‘π‘Žπ‘¦ ]
𝑄𝐿
3
Daily averaged leakage rate for a given section of pipe [π‘š ]
π‘Ž
The distance along the supply pipe at which pressure becomes negative [π‘š]
𝐴eq
2
Equivalent area of a single orifice to account for all leaks [π‘š ]
πΆπ»π‘Š
Hazen-Williams constant
𝐷
Pipe diameter [m]
𝑓
Friction factor [unitless]
𝑔
2
Gravitational acceleration [π‘š/𝑠 ]
𝐻loss
Frictional pressure head loss [m]
π»π‘œ
The contaminant pressure, positive if it forces them inside the pipe [m]
𝐻𝑝
Pressure in the pipe [m]
𝐻𝑆
Supply pressure at the distribution main [π‘š]
𝐻𝐡𝑃
Pressure at the suction inlet of the booster pump [π‘š]
𝐻𝑅𝑆
Duration of water supply [hours/day]
𝐿𝐢
2.5
Empirical leakage coefficient [π‘š /𝑠]
𝐿𝑃
The length of pipe [m]
21
𝐿𝐿
Leakage level [%]
𝑄
3
Volumetric flow rate [π‘š /𝑠]
𝑄𝐼
3
The volumetric flow rate of contaminant intrusion into the pipe [π‘š /𝑠]
𝑄𝐿
Volumetric flow rate of leaking water
𝑇
Duration of Observation [s]
𝑑
Elapsed Time [s]
𝑇′
The duration of either the control or treatment measurement at a given house
[days]
𝑇𝑇
The total duration of the control and treatment trials at a given house [days]
𝑉𝐼′ (𝑑)
The relative total volume of intruded contaminants over a given time interval
[π‘š
3
/𝐿𝐢 ]
𝑉𝐿′ (𝑑)
The relative total volume of leakage over a given time interval [π‘š
π‘₯
Length along pipe from supply towards consumer’s house [m]
CFU
A measure of contamination; colony forming units [/100mL]
CIR
Contaminant Intrusion Ratio [unitless]
IC
Intrusion concentration
V
Average fluid velocity [m/s]
22
3
/𝐿𝐢 ]
Chapter 1
Introduction
This thesis began with a desire to make a contribution to improving the quality of
water delivered to the 279 million people in India who have access to piped water on
their premises [1]. Most water distribution systems in India operate intermittently
[2, 3], which makes them more likely to recontaminate the distributed water than
continuously-operated water systems [4, 5].
As part of the MIT Tata Center for Technology and Design, the author traveled
to India on five separate occasions and met with water-system designers, operators,
and users in Mumbai, Bangalore, and New Delhi. A pivotal moment in the project’s
scoping occurred when one system engineer admitted that he only knew what the
average pressure in his system was when the electricity was not working. While the
city’s electricity was working, households would connect small pumps directly to the
water network, rapidly withdrawing water and reducing system pressure. However,
when the electricity was not working, these pumps could not be used and the system
pressure increased to a known level. To increase withdrawal rates, these residential
booster pumps create a negative pressure in the pipe leading to the house; unfortunately, this negative pressure increases the risk of contamination.
Rapid withdrawal rates lead to dramatic reductions in water pressure at the extreme ends of the water system and lead to unequal distribution of water [6].
To
protect households from contamination and to protect the water system from low
and unequal pressure, it is typically mandatory in India to install an isolation tank,
23
called a sump, between the city’s water supply pipe and any residential booster pump.
A more detailed consideration of the connection between contamination, pressure, and
booster pumps can be found in Chapter 2.
Some cities are better at enforcing the prohibition against using a booster pump
without a sump than others. Specifically, because of population density and political
precedent, the regulations regarding sumps have not been enforced in New Delhi.
Therefore, an alternative way of protecting customers from contamination, and protecting New Delhi’s water system from low and unequal pressure is required. Existing
options for such protection were reviewed for their applicability to Delhi’s water system and were found to be either easily tampered with or too costly; a brief summary
of such prior art is included in Chapter 3.
Working in collaboration with New Delhi’s water utility, the Delhi Jal Board
(DJB), a patent-pending, self-actuating, full-bore, back-pressure regulating valve was
developed and tested. The valve’s simple mechanism relied on a collapsing tube that
would close when negative pressure was present inside the pipe. The major innovation
in this valve’s design was a specially-shaped insert which prevented oscillations, which
are typically found in such collapsing tubes. This stabilized-collapsing-tube mechanism was encapsulated in a rigid valve shell.
The valve’s design and the design’s
evolution are documented in Chapter 4.
The experimental and numerical methods for testing and interpreting the valve’s
performance, presented in Chapter 5, were used to measure the valve’s performance
in laboratory flow tests, fatigue tests, and field demonstrations.
Field tests of two
design iterations were conducted in New Delhi in January and March of 2014; the
successful results of these tests are presented and discussed in Chapter 6.
While the original application for the valve designed herein was New Delhi’s water
utility, it holds potential for any water system in which customers use booster pumps
without isolation tanks, or any local where it might be desirable to replace isolation
tanks with an equivalent valve that is smaller and cheaper. Further conclusions and
opportunities for future work are presented in Chapter 7.
24
Chapter 2
Background
While the bulk of this thesis focuses on the effects of residential booster pumps and
a water valve to reduce the risk of contamination, this section explores the issue
of water contamination from first principles.
First, it considers the mechanism of
contamination in water distribution systems; second, it specifically explores the consequences of intermittent water supply; third, it addresses how booster pumps are
typically installed and the effects of these installation method; and finally, this chapter concludes with a description of the two current and competing opinions on how
to best address the problems associated with intermittent water distribution, noting
that in until water is distributed continuously and at uniformly high pressure, the
problem of residential booster pumps must be addressed.
2.1 Contaminant intrusion
The intrusion of contaminants into water distribution systems has caused significant
public health concerns and several historical crises [7]. Intrusion requires three things
[8]:
1. a pathway for contaminants to enter the water system,
2. contaminants to be present in the vicinity of the pathway, and
3. a pressure gradient such that the contaminants flow into the water system.
25
Table 2.1: Water balance definitions, unaccounted for water comprises of real losses
and apparent losses; adapted from [9]
This section will sequentially evaluate whether each of these requirements is met in
intermittent water systems in general, in India, and then specifically in New Delhi.
2.1.1 Leakage rates indicate the pathway size for contaminant
intrusion
The first of three requirements for contaminant intrusion to occur is that a pathway
for contamination must exist. Were a physical pathway to exist between the inside of
a water pipe and the outside, one would expect that most of the time, such a pathway
would have clean water leaking out of it. The proxy of leakage rates is therefore used
to assess the size and location of potential pathways for contamination.
While all water systems leak, leakage levels in water systems in India are typically extremely high and therefore many pathways for contamination are present.
Because the exact level of physical leakage in a water system is difficult to measure,
many proxy indicators are used; their overlapping definitions are captured in Table
2.1. Unaccounted for water (UFW) is a frequently used metric because it is easily
calculated, but it obscures the magnitude of physical leakage by combining it with
inaccurate water meters and unauthorized connections. Another common metric is
non-revenue water (NRW), which further obscures physical leakage by including metered but unpaying customers.
The WHO and UNICEF have estimated an average level of UFW in urban water
systems across Asia, Africa, and South and Central America as 40% [10]; within this
UFW, they estimate that physical losses comprise the majority.
26
In India in 2007,
self-reported UFW levels of 20 water utilities averaged 32% [3].
In New Delhi in
2011, non-revenue water was reported by Delhi’s water utility, the Dehli Jal Board
(DJB), as 52% [11].
Delhi’s physical losses were estimated to be 20-25% [11].
For
comparison, New York and Toronto average UFW levels of 15% [12].
A simple percentage comparison however, hides how large the problem is. Cities
in India on average supply water for only four hours per day, often at low pressure [3].
To equitably compare leakage rates across different cities, the leakage rates must be
scaled according how long each system was pressurized and at what pressure. Using a
simple leak dependency where flow is proportional to the square root of pressure [13],
we find Equation 2.1 (for a more thorough treatment of leakage equations see Section
5.3.1, for more information about pressure notation and units see Section 5.2). Where
𝐿𝐿
is the leakage level in percent,
and
𝐻
𝐻𝑅𝑆
is the number of hours per day of supply,
is the average pressure during supply times.
𝐿𝐿adjusted
(𝐻𝑅𝑆)adjusted
= 𝐿𝐿old
(𝐻𝑅𝑆)old
√οΈ‚
𝐻adjusted
𝐻old
(2.1)
Converting the average leakage levels in India to their equivalent 24-hour amounts
would make them six times higher. Taking New Delhi’s leakage rate as an example,
if one optimistically assumes an average supply of 6 hours per day at 7m of pressure,
then to keep those sames pipes pressurized to 17m for 24 hours per day, would create
a leakage level of 125%, i.e. it would require 125% of the current total water usage
to go towards leakage alone. Adding customer demand to this calculation, it would
therefore require twice the current amount of water to supply all customer needs and
to keep the New Delhi’s current pipe system continuously pressurized.
It must therefore be concluded that in existing water systems in India, significant
pathways for contaminant intrusion exist.
While upgrading these systems would
reduce many of the pathways for contamination, some leakage will always exist and
so the pathway for contaminants can never be completely removed. Keeping water
safe from contaminant intrusion must therefore rely on avoiding one or both of the
27
other two requirements for contaminant intrusion.
2.1.2 Contaminants may be present because of the unplanned
locations of water pipes
Eliminating the presence of contaminants is likely infeasible. Officially, Indian standards insist on three meters of horizontal separation between a water pipe and sewers
or other contamination sources. When water pipes need to crossover a contamination source, the water pipe must be at least 0.5m above the contamination source
[14, p.388]. Even where safety distances are maintained, during and after a rainfall,
soils can become saturated with rainwater. In one detailed field study of water contamination in an Indian city, water contamination was found to increase significantly
immediately after major rainfalls [4].
To make matters worse, in most cities in India, the drainage, sewer, and water
pipe networks have been developed independently, haphazardly, and very shallowly,
preventing adherence to the required separation.
One field engineer even reported
that plumbers will often route the water pipe through (not over or under) an open
channel sewer to make the job go faster. A typical example of this cross-contamination
potential is shown in Figure 2-1. It therefore seems inevitable that contaminants are
present in the vicinity of water pipes; if intrusion is to be avoided, another prevention
strategy must be taken.
2.1.3 Pipe pressure can deter contamination
Where a pathway exists and contaminants are present, pipe pressure is the last line
of defense against contaminant intrusion.
Intrusion can only occur when the con-
taminants outside the pipe have a higher pressure than the water inside the pipe.
This occurs either when the pressure in the pipe goes below atmospheric (negative
pressure), or when the contaminants around the pipe become positively pressurized.
Low water system pressure led to water contamination in eight case studies compiled by Lee and Schwab [7]. Contaminant pressure is usually less than the depth of
28
Figure 2-1: A leaking sewage pipe drips onto an array of household water service
connection pipes, ensuring contaminants are in present around the pipe
the pipe underground and can occur when the pipe has been laid below the ground
water level (unlikely in urban centers in India), when soils become temporarily saturated with rainwater, or when sewers plug and the sewage becomes pressurized. One
Indian engineer reported that the typical first response to customer contamination
complaints is to check to see whether a sewer near the customer’s house has become
plugged. To overcome the risk of contamination, distribution systems should maintain
a minimum safety pressure greater than the expected maximum external contaminant
pressure.
Worse than low pressure, is no pressure. While pipes are unpressurized, they have
no defense against the intrusion of gravity-driven, percolating, contaminants. Intermittent water systems, are by definition, frequently unpressurized. On average, water
systems in India are unpressurized for 20 hours per day [3]. When these systems are
turned on, the initial water flowing through them must flush out the contaminants
which have been filling the system for the past 10 to 20 hours. These contaminants
typically exit the system through consumers’ taps. Kumpel and Nelson found that
29
contamination indicator organisms were present in concentrations that were an order of magnitude higher during the initial flushing period than during steady-state
operations [4]. Compounding this quality concern, many consumers use an overhead
storage tank which fills during supply hours. Such houses are often plumbed to directly fill the tank whenever there is water pressure, mixing therefore the lower quality
flushing water with the higher quality steady-state water. While not within the scope
of this thesis, creating a new system for automatically storing the initial lower quality
water separately from the cleaner water represents a possible improvement to the
status quo.
The most serious source of contamination however, comes from negatively-pressurized
pipes. Negative gauge pressure by definition means that the pressure inside the pipe
is less than atmospheric pressure; it therefore acts to actively pull contaminants into
the pipe.
Section 5.3 develops the required theory to quantify the relative risks of
unpressurized and negatively-pressurized systems.
Given the difficulty in preventing the first two requirements for contaminant intrusion, positive pressure must be maintained. The impossibility of maintaining positive
pressure in intermittently supplied water is one of the main reasons many in this sector
advocate for a transition back to continuously operated (24/7) water distribution.
2.2 Intermittent water systems
Despite the contamination danger, in Africa, Latin America, and the Caribbean, onethird of distribution systems operate intermittently [12]; in Asia, half do [12]; and,
most major water systems in India do [3]. These distribution systems were neither
intended to be nor designed to be operated intermittently [6], but are forced to do so
because the combination of leakage rates and consumer demand exceeds the available
water supply.
Water suppliers forced to adopt the strategy of intermittent water supply, might
initially only need to shut the system down for a few hours per day.
Consumers
would withdraw a similar volume of water as during continuous supply (only at a
30
Figure 2-2: Historically, maintaining an intermittent system that provides water for
the majority of the day has not been sustainable, reproduced from [15]
faster rate), but the leakage levels would be reduced, according to Equation 2.1. This
reduction in leakage level would make up the supply deficit. The longterm viability
of this strategy however, is dubious; historically the amount of time a utility in India
has been able to supply water has degraded over time, as set out in Figure 2-2,
reproduced from [15].
While the mechanism for this degradation is debatable, the
historical viability of turning off the system for only a few hours per day in India is
not. Therefore, most cities in India supply water for approximately four hours per
day or less, as summarized in Figure 2-3.
2.2.1 Effects of intermittent water supplies
In addition to the primary concern of higher contamination risks, when compared to
continuous water supply, intermittent water supply does the following:
1. distributes water less equitably,
2. causes the water network to degrade more quickly, and
3. passes on more adaptation costs to the consumer.
High frictional losses during supply times create severe geographic inequalities in
the pressure and duration of water supply [6, 16], which disadvantage consumers at
31
Figure 2-3: Most cities in India supply water for approximately four hours per day
or less, reproduced from [3, p.21]
the extremities of the water system whose pressure would already have been lower
than customers closer to supply reservoirs. This phenomenon arises because the total
consumption of water after a transition to intermittent water supply does not change
dramatically.
Demand is therefore concentrated in shorter periods of time, and so
flow rates increase. As pressure loss is approximately proportional to the flow rate
squared [13], even a reduction to 12 hours of supply per day which could lead to a
doubling of flow rates, would translate into a quadrupling of pressure losses.
In the filling and draining of water systems, the sudden change from open-channel
flow to pressurized flow can create large pressure transients and cause accelerated pipe
wear and bursts [17, 18].
While in continuously-operated systems these dangerous
events occur infrequently, they can happen as often as twice per day in intermittently
operated systems.
One water system engineer in India reasoned that since most
consumer taps are left open in intermittent systems, they act as distributed air release
valves that dampen this effect. Transient expert Bryan Karney however, was skeptical
32
that this would provide significant damping [19].
This assertion will be verified in
future work as a possible opportunity for improving the quality and durability of the
existing water networks. Nevertheless, it remains true that the intermittent operation
of a water system accelerates its degradation.
Previous studies have focused in detail on the adaptations that consumers are
forced to make in response to lower pressure and shorter supply times as a method
of quantifying the indirect costs of intermittent supply to consumers [20, 21, 22, 23].
These studies often aimed at determining how much households would be willing to
pay for the provision of higher quality of water. Such lists of consumer adaptations
often include:
1. water filtration or boiling to address the lower quality;
2. time spent collecting water, or waiting for the system to turn on;
3. purchasing alternative (and possibly lower quality) water to supplement supply;
4. a storage tank (often on the roof ) to make water available during non-supply
hours; and sometimes
5. a residential booster pump to supplement the system pressure and/or increase
the flow rate.
This final adaptation where consumers use residential booster pumps is of special
relevance to this thesis and is the focus of the next section.
2.3 Residential booster pumps
A residential booster pump (RBP) is a centrifugal pump in the home of a water
consumer.
They are typically used when consumers wish to supplement either the
pressure or the flow rate provided by the water utility. They range in size from 0.5
to 1.5 horsepower (hp), with the most common being 0.5 and 1hp. Some methods of
connecting the RBP can lead to an increased risk of contamination. This section first
33
Storage Tank
No flow
Supply Pressure
Pressure
Low flow
Moderate flow
Water
Meter
M
Ground Level
Distribution
Main
2 Story
Home
Service Connection Pipe
Figure 2-4: A household connection without a RBP overlaid with the pressure profile
for house connections at different stories; upper stories and the roof tank do not
receive acceptable flow rates and require a booster pump
details how pumps are typically configured and then elaborates on the risks associated
with these configurations.
2.3.1 Booster pump configurations
Figure 2-4 depicts how a house would typically be connected to a high pressure water
supply without a RBP. The pipe’s pressure profile is overlaid on the diagram, where
the vertical axis is the pressure head and the horizontal axis is the location along the
pipe. The slope of this pressure profile indicates the direction and magnitude of flow;
water flows in the direction of decreasing pressure in proportion to the square root of
the slope. When the supply pressure is less than the height of the building, residents
of upper stories receive little or no water.
To overcome this problem, consumers
typically use RBPs. RBPs can be connected in one of three ways:
1. Figure 2-5 shows the officially-sanctioned configuration. After passing through
the water meter, the water discharges into an underground reservoir called a
sump. The booster pump draws water from the sump and pushes it into the
overhead tank. Note that the booster pump can only draw water from the sump
and therefore the RBP neither affects the flow rate into the sump, nor creates
a negative pressure in the service connection.
34
Tank
Pressure
Supply Pressure
Moderate flow
Water
Meter
Figure 2-5:
RBP configuration 1:
P
M
Ground Level
Distribution
Main
2 Story
Home
Service Connection Pipe
Sump
the officially sanctioned method of connecting
a booster pump; the pump only draws water from the sump and cannot create a
negative pressure in the service connection
2. Figure 2-6 shows a modification of the “official” configuration.
Where sumps
exist but consumers are not content with the rate at which the sump fills, a
second booster pump is often installed upstream of the sump. The sole purpose
of this upstream pump is to increase the flow rate into the sump.
Since the
upstream booster pump is connected directly to the service connection pipe, it
increases flow by creating a negative pressure at its inlet. This negative pressure
spreads upstream in the service connection, represented by the shaded orange
region in Figure 2-6.
3. Figure 2-7 shows another common configuration.
Where houses do not have
sumps, the pump connects the service connection pipe directly to the overhead
tank.
This configuration can also create negative pressure upstream of the
booster pump.
Configurations 2 and 3 are problematic because each can induce negative pressure,
which creates four connected problems:
1. it increases significantly the risk of contamination as discussed in Section 2.1.3;
2. it increases flow rate which, when many people have booster pumps, further
lowers the system pressure and increases distribution inequality;
35
P
Supply Pressure
Pressure
M
Tank
= Booster Pump
= Water Meter
2 Story
Home
Very large flow
M
Ground Level
Distribution
Main
Service Connection Pipe
P
P
P<0
Sump
Figure 2-6: RBP configuration 2: a booster pump is located before the sump and fills
it faster; the pump after the sump fills the overhead tank; the flow rate is increased
by creating a negative pressure which also increases contamination risk
P
Supply Pressure
Pressure
M
Tank
= Booster Pump
= Water Meter
2 Story
Home
Very large flow
M
Ground Level
Distribution
Main
Service Connection Pipe
P
P<0
Figure 2-7: RBP configuration 3: booster pump is directly connected without a sump;
this configuration can also cause negative pressure
36
3. it creates
pump wars, in which neighbors with bigger pumps receive more water;
and
4. it prevents system operators from being able to use water pressure as a means
of controlling customer flow rates.
Wherever intermittent systems exist, booster pumps are likely to be found [21,
22, 24, 25].
Specifically in India, their use is frequent [20, 21, 22]; however local
governments vary in their insistence on the use of the “official” configuration style.
This thesis’s geographic focus on New Delhi arose because of Delhi’s unique position
regarding booster pumps.
During approximately 40 house inspections in New Delhi by the author, configurations 2 and 3 were frequently observed, with a ratio of approximately 1:2. Configuration 1 was observed in New Delhi only once. Despite their widespread use, because
of the severe consequences of configurations 2 and 3, the Delhi Jal Board Act has
specifically prohibited the connection of residential booster pumps to the service connection pipe [26, §19(3)]. The DJB is however mandated by the Indian Central Public
Health and Environmental Engineering Organization (CPHEEO) to supply pressure
at the point where the service connection meets the distribution main (the location of
a check valve called a
ferrule ) in accordance with the Table 2.2.
Because of municipal
building codes in Delhi and the high population density, water must be supplied at
17m of pressure in order to ensure that apartments on the third story (Indian second
story) can receive water.
Since the DJB cannot meet this requirement, enforcing
the prohibition against booster pumps has been described as politically infeasible;
enforcing the prohibition would cut off consumers from water. Could these connections at least be forced to construct sumps? In densely populated areas, retroactively
constructing sumps would be prohibitively expensive.
A section on booster pumps would not be complete without three quotes from
upper managers at the Delhi Jal Board which summarize the practitioners’ perspective
on the problem:
“It [the problem of booster pumps] f_cks my hydraulics”
37
Table 2.2: CPHEEO static pressure requirements, from [14, p.361 §10.3.3]
Height of Building
Required Static Pressure
Single story
7m
Two story
12m
Three story
17m
“Booster pumps are like a disease [once one person has one, they spread]”
“Booster pumps are our biggest problem”
New Delhi’s water system, therefore, required a method for making booster pump
configurations 2 and 3 safer and more equitable without the space or cost of an
underground sump.
More specifically, this could only be accomplished if booster
pumps were prevented from inducing negative pressure in the service connection pipe
upstream of them. Therefore, the key objective of the design presented in
this thesis was to limit the ability of booster pumps to create negative
pressure in the service connection pipe, thereby decreasing contamination
and increasing distribution equality.
Should a solution to this problem be found, it would have applications beyond just
New Delhi; were it to become an approved alternative to configuration 1, households
across India would have a smaller and cheaper way of safely connecting their booster
pumps.
A solution to this problem would, therefore, benefit households and the
water utility in New Delhi, in other intermittently-supplied cities in India, and in any
intermittently-supplied city that uses booster pumps.
During the development of this project, however, a number of practitioners questioned the long-term relevance of the problem of booster pumps given the current
policy push for intermittent water systems to convert to continuous water systems.
The final section in this chapter seeks to review the current strategies for improving
intermittency and to determine the likelihood that each strategy would mitigate the
problem of booster pumps.
38
2.4 Upgrading intermittent water systems and its effect on booster pumps
To address the concerns of water quality and distribution equity associated with
intermittently supplied water, two kinds of solutions have been proposed: the first
proposes to create incremental improvements in intermittent systems, and the second
proposes methods to eliminate intermittency altogether. The author’s conversations
with many practitioners in the Indian water sector have suggested that the latter
approach has captured much of the policy focus. Where incremental improvements are
being made to intermittent water systems and where the transition to high pressure
and continuous supply has not yet been completed, the problem of booster pumps is
likely to persist; in such circumstances, solutions to the booster pump problem will
continue to be relevant.
2.4.1 Incremental improvements in intermittent supply
Advocating for the incremental option, a leading expert in intermittent water systems,
Professor Vairavamoorthy, has argued that intermittent systems are inevitable and
therefore finding better ways of operating them must be a priority [27, p.191].
To
mitigate equity concerns he recommended:
“Equitable distribution of the limited quantity of water is the keystone
of the entire design process...and is a non-negotiable design objective. This
objective is achieved mainly through effective pressure management.”
To mitigate quality concerns, Vairavamoorthy has advocated for more detailed
modeling of contaminants and intrusion [28, 29].
This would allow for prioritized
incremental improvements in water safety by, for example, replacing high-risk pipes
or lining open sewers that cross many water connections.
Residential booster pumps installed in configurations 2 and 3 threaten Vairavamoorthy’s recommended improvements to quality and equity because during supply
times, booster pumps actively increase the contamination risk and increase flow rates,
39
which ultimately makes effective pressure management more difficult. The incremental improvement approach, therefore, would not negate the need to safely connect
booster pumps; instead it stands to benefit from safely-connected booster pumps.
2.4.2 A transition to continuous supply
From the currently dominant perspective, many, including the World Bank’s Water
and Sanitation Program (WSP), have advocated for a complete transition back to
continuously supplied systems [30], thereby addressing the quality and equity concerns of intermittent water supply.
The WSP has demonstrated this approach in
three projects in pockets of cities in Karnataka. These pockets successfully made the
transition to continuous water supply and found that water quality went up [4], while
the total water requirement went down [30].
During the execution of these three demonstration projects in Karnataka, it was
found to be more economical to replace the whole water systems than to follow
through on the originally planned replacement of 60% of the networks.
While the
results from this demonstration look positive, it remains unclear as to how feasible
such a project would be without external support and oversight from the WSP.
Building on the success of these demonstration projects, several cities in India,
including New Delhi and Nagpur, have undertaken major public-private partnerships (PPPs) to attempt to transition sections of their distribution networks from
intermittently- to continuously-operated. Most of these PPPs do not plan to create
an entirely new distribution network, but rather to replace significant portions of the
old one and thereby incrementally transition to 24/7 water provision.
An incremental transition to continuously-supplied water would require extensive
leak repair efforts.
Taking New Delhi as an example, if only 7m of pressure and
continuous supply were desired, to maintain the current leakage rate of 20%, using
Equation 2.1, a 75% reduction in the amount of leaks in the current system would
be required. Making the transition to 24/7 and 17m of pressure, would require an
84% reduction in the amount of leaks in the system. To address 84% of leaks, which
have unknown locations within the system, would likely be more difficult than a total
40
system replacement, which explains why this was done in the WSP project.
When a water system supplies high-pressure water continuously, the problem of
booster pumps will be negated; residentially-sized booster pumps cannot create flow
rates high enough to create a negative pressure while the supply pressure is 17m;
therefore, they will be rendered safe, independent of their connection configuration.
However, before and during the transition to continuous high-pressure supply, booster
pumps will continue to threaten water quality and distribution equity. Designing and
testing a device to limit these consequences that is smaller and cheaper than an
underground sump is the focus of this thesis.
41
THIS PAGE INTENTIONALLY LEFT BLANK
42
Chapter 3
Prior art
This thesis sought to improve the water quality and distribution equity of urban water
distribution networks in India by eliminating the ability of booster pumps to create
negative pressure in the service connection pipe.
To achieve this aim, the design
process detailed in Chapter 4 identified the specific need for a simple and inexpensive, self-actuating, full-bore, back-pressure regulating valve.
To contextualize the
designs of Chapter 4, this chapter summarizes the existing household contamination
prevention devices; it then reviews existing self-actuated, full-bore, and back-pressure
regulators; and, finally it concludes with a brief discussion of a collapsing tube, which
was the critical component of the final valve design.
3.1 Household contamination prevention devices
In continuously-supplied water systems, one possible source of contamination comes
from cross-connections. A cross-connection is any connection in the water distribution
system, typically in a customer’s home, where the supply of potable water comes in
contact with a non-potable liquid, as when, for example, a garden hose dilutes a bucket
of liquid fertilizer. If the potable water pressure were to temporarily become negative,
contaminants could be drawn back (back-flow) into the potable water supply.
To
address this public health risk, the US EPA has published the Cross-Connection
Control Manual [31]. While this contamination mechanism is different than the one
43
Figure 3-1: An air gap offers excellent protection from negative pressure, but introduces a potential source of contamination and pressure loss; reproduced from [32]
caused by booster pumps, both are caused by negative pressure, and therefore, some
of the safety equipment designed to prevent contamination from cross-connections
was considered as possibly applicable to the problem of booster pumps. Specifically,
three devices recommended by the US EPA were reviewed as part of this section: an
air gap, a vacuum breaker, and a barometric loop.
The US EPA commended the use of an air gap, as in Figure 3-1, as “extremely
effective” for the prevention of back-flow [32]. This is the same mechanism employed
in the mandatory isolation tank in Indian water systems, referred to there as a ‘sump.’
This solution, however, poses three possible problems: first, the space for such a tank
may not always be present; second, the tank creates the possibility for contamination
to enter the water supply; and finally, all of the water pressure entering the tank is
lost. In India, where an air gap is used, it is typically in a large underground reservoir
that can be difficult to clean and costly to build.
Another safety mechanism recommended by the US EPA was a vacuum breaker,
one such device is shown in Figure 3-2. When vacuum (negative) pressure is present
upstream from the device, the vent is pulled open, air enters the system, and the
vacuum is broken, i.e. negative pressure was removed. While this device is very effective at preventing back-flow, the problem of booster pumps has different functional
requirements. Customers with booster pumps may occasionally wish to create a negative pressure so as to increase their flow rate; the safety device must prevent this
and therefore must be difficult to tamper with. The vacuum breaker in Figure 3-2
44
Figure 3-2: A vacuum breaker can prevent negative pressure, but it is easy to tamper
with and introduces air into the system; reproduced from [32]
could be easily defeated by blocking its vents. Another contextual difference between
back-flow and booster pumps is that some houses use booster pumps that are not only
above-ground level, but above the water pressure level. For example, an apartment
might use a booster pump located 12m above the ground to draw water from the
supply pipe which had 10m of supply pressure. To operate, the booster pump would
create a negative pressure for several meters below it in the above-ground supply
pipe. This negative pressure would not cause contaminant intrusion since it would
only occur above the ground where the presence of contaminants is extremely unlikely. Therefore the ideal safety device would only prevent negative pressure in the
supply pipe below ground, and would allow elevated pumps to continue to operate.
The US EPA’s style of vacuum breaker would introduce air into the system, which
would cause an elevated booster pump to lose its prime and the envisioned apartment
would receive no water; this option was therefore not pursued.
The final potentially relevant device recommendation by the US EPA was a barometric loop, shown in Figure 3-3. If water with a negative pressure enters the bottom
of the barometric loop, by traveling upwards its pressure is further reduced and it
will eventually vaporize.
Once vaporized, the two sides of the loop are no longer
hydraulically coupled and the negative pressure on one side of the loop cannot exert
45
Figure 3-3: A barometric loop prevents negative pressure by taking advantage of the
vapor pressure of water; unfortunately it requires long stretches of pipe; reproduced
from [32]
a pull force on the other side of the loop. The height of the loop is specified so that
if positively-pressurized water enters the loop, it will not vaporize at the top, but
negatively-pressurized water will. This device is robust and extremely simple. Unfortunately, if implemented in an intermittent system, each time the system was turned
off, the loop would drain and would only work again when the system provided more
than 10m of pressure. Many cities in India do not consistently provide 10m of pressure and therefore this idea too was abandoned. From this brief survey of back-flow
prevention devices, houses in India are in need of something as reliable as a vacuum
breaker, but that does not entrain air into the system.
3.2 Self-actuated, full-bore pressure regulating valves
This thesis sought to design a valve that was self-actuating, full-bore, and pressure
regulating.
Combinations of two of these three characteristics are reviewed in this
section, but no combination of all three was found.
Were one to be found, it is
likely that it would be extremely complex, unlike the design of the valve described in
Chapter 4.
As any valve closes, it forces fluid through a smaller and smaller area, increasing
46
Figure 3-4:
Self-actuated back-pressure valve with a large diaphragm; reproduced
from [34]
the local jet velocity and therefore increasing the frictional losses through the valve.
A regulating valve is designed to operate in many positions along its closing stroke
[33].
A pressure regulating valve is designed to maintain a minimum or maximum
pressure either upstream or downstream from the valve.
To make efficient use of
a regulating valve’s stroke, Skousen’s Valve Handbook recommends that regulating
valves be undersized, making even the initial parts of the stroke have a noticeable
effect on the frictional losses through the valve [33].
Since many regulating valves
follow Skousen’s advice about being undersized, full-bore (i.e. valves that do not
reduce the area that the fluid has to pass through) regulating valves are much less
common. Still, the most common type of regulating valve is a gate valve [33], which
if externally actuated, can be full-bore. Further, an externally-actuated, regulating
ball valve could also be full-bore. So while full-bore regulators are rare, they do exist.
Self-actuated regulators also exist. Self-actuated gate valves typically have a limited stroke and use a large diaphragm to overcome the frictional forces associated with
changing the valve’s position. An example of such a valve is pictured in Figure 3-4,
reproduced from [34]; it uses a set of sliding plates with matching orifices to regulate
the flow.
Self-actuating, full-bore valves are commonly used as steam safety devices, where
the full-bore allows the system to release pressure more quickly. One such valve is
shown in Figure 3-5.
They are frequently actuated with pilot lines, and therefore
cannot throttle the flow.
47
Figure 3-5: Self-actuated full-bore valve is pilot operated, designed for emergency
relief and not for regulation; reproduced from [35]
The combination of all three has not yet been found: self-actuating, full-bore, and
regulating. Should such a combination exist, it is likely that it would use a very large
diaphragm or complicated pilot lines. To address the problem laid out in Chapter 2,
a simple, full-bore, self-actuated, back-pressure regulator is required. The designed
valve used a stabilized collapsing tube to accomplish this.
3.3 Collapsible tubes or ‘Starling Resistors’
The design of a self-actuating, full-bore, back-pressure regulating valve was realized
with a stabilized collapsing tube.
The use of a collapsing tube as a regulator is
frequently referred to as a “Starling Resistor” [36] after E.H. Starling, who used one
in his experimental apparatus, where the tube only allowed flow through it when the
upstream pressure was greater than or equal to the pressure on its outside [37]. Such
a tube acts as a self-actuated, full-bore, back-pressure regulator. If subjected to an
external pressure of atmospheric pressure, the collapsing tube would prevent negative
48
pressure from passing upstream, and therefore meets the design requirements.
Unfortunately, these resistors are known to have an unstable regime in which
they oscillate rapidly [38].
No information however was found about methods or
mechanisms, theoretical or practical, for removing this instability. While this thesis
does not provide the theoretical framework to explain why the design outlined in
Chapter 4 stabilized the collapsing tube, a specially-shaped insert placed inside the
collapsing tube was found to eliminate much of the instability, preventing it from
inducing severe and frequent transients into the system.
49
THIS PAGE INTENTIONALLY LEFT BLANK
50
Chapter 4
The iterative design of a
self-actuating, full-bore,
back-pressure regulating valve
The design of a patent-pending, self-actuating, full-bore, back-pressure regulating
valve used an iterative design process.
It was developed to increase water quality
and distribution equity by eliminating the negative pressures caused by residential
booster pumps in India.
This chapter outlines the four distinct design cycles that
occurred:
1. an initial rapid prototyping and concept generation stage developed the key
strategy: a collapsing soft tube;
2. early experiments found the optimal method for stabilizing the collapse of the
soft tube: a rigid insert;
3. the alpha prototype took this strategy and packaged it into a product that could
be tested in houses in India; and finally,
4. the beta prototype attempted to improve upon two issues encountered during
the field trial of the alpha prototype.
51
Table 4.1: Initial functional requirements and design parameters
F1
F2
Functional Requirement
Design Parameter
Keep pressure upstream
Back-pressure regulating valve, with a set point
from the house positive
of atmospheric pressure, i.e. X=0 in Table 4.2
Minimize hydraulic
Use a full-bore valve design
losses through the device
F3
Durable
Simple, self-actuated, low part count
F4
Inexpensive
Simple, self-actuated, low part count
Table 4.2: Logic of a back-pressure regulating valve with a set point of X
Pressure > X
Pressure < X
Open
Close
4.1 Preliminary concepts and their evaluation
The four core functional requirements that drove this design process and the design
parameters selected to meet these requirements are detailed in Table 4.1.
4.1.1 Keeping upstream pressure positive
To ensure the pressure upstream from the house remained positive, two possible
strategies were identified to reduce the efficacy of the booster pump: add a check
valve to air that would allow air into the system (entrain air) when it was negatively
pressurized, causing two-phase flow; and implement a back-pressure regulating valve
that would throttle the flow, inducing cavitation. While the check valve was promising because it would avoid forcing the booster pump to cavitate, it was ultimately
abandoned after its inclusion in some of the January 2014 field trial because:
1. The air traveled from the check valve to the booster pump at the velocity of
the water and therefore created a significant delay in the control loop, which
varied between houses. On houses with a far distance between the valve and
the pump, the system would cyclically oscillate (hunt) — instead of finding an
52
equilibrium state.
The pump would entrain a pipe full of air before the air
reached the pump; when the air did reach the pump, the flow would then stop
until the pump had cleared the air; finally, the pump would start again, induce
a negative pressure, and the cycle would repeat.
2. The air pathway could double as a contaminant pathway;
3. The air caused consumers’ water to sputter and splash out of the spout; and
4. Some RBPs were elevated and ran on a vacuum above ground level, and air
caused these RBPs to lose their prime.
Instead of entraining air, the final design throttled the pressure upstream from
the pump to reduce its efficacy.
The use of pressure as a control variable had less
hunting problems because the pressure signal traveled at the speed of sound in the
fluid, not at the fluid’s velocity. To use pressure as a control variable, a back-pressure
regulating valve was designed. A back-pressure regulating valve has the same logic
as a pressure relief valve, whose logic is summarized in Table 4.2. However, unlike a
pressure relief valve, whose default state is closed, this back-pressure regulating valve
would operate in the open position most of the time. If the valve were placed at the
consumer’s house, perhaps next to the water meter, then the set point pressure ‘X’
in Table 4.2 would simply be atmospheric pressure.
If instead, the valve were designed to be placed underground and upstream from
the consumer’s house, its set point logic would need to be a function of the flow rate
to adjust for the variable pressure drop between the valve and the household. While
complicated, a regulating valve with a flow-dependent set point has been implemented
for large pipes, as in Figure 4-1, reproduced from [39, pp. 266-267]. The underground
placement of the valve was ultimately abandoned, however, after feedback from a
Tata Center for Technology and Design design-review in August 2013 in Mumbai,
which highlighted the relative difficulty of inspecting and maintaining underground
equipment.
53
Figure 4-1: Overly complex variable set point valve, reproduced from [39]
(b) Fluid is forced through a restriction in
(a) Fluid takes two 90 ∘ bends in globe valve, this poppet valve, even while open, reproreproduced from [33]
duced from [40]
Figure 4-2: Bend and constriction losses in some types of commercial valves
4.1.2 Minimizing hydraulic losses
Two major sources of friction occur as fluid passes through a valve:
bending and
constriction, as in Figure 4-2.
For turbulent flow, constriction causes losses proportional to the constricted diameter to the fifth power [13].
Where the cross section is not circular, a modified
diameter, called the hydraulic diameter,
the cross sectional area and
𝑃
𝐷𝐻 ,
defined in Equation 4.1, is used.
𝐴
is
is the wetted perimeter of the flow path [40, p. 92].
𝐷𝐻 =
4𝐴
𝑃
(4.1)
The minimum friction therefore comes from a circular cross section with no restriction
in diameter, known as a full-bore or full-port design [33].
54
4.1.3 Concept generation and selection
Pursuant of a passively-actuated, low-friction, back-pressure regulating valve with a
set point of atmospheric pressure, a structured brainstorm was conducted to generate
concepts. Select rough sketches are shown in Figures 4-3 and 4-4. Imposed on these
sketches is the reason they were not selected. Key criteria for selecting between the
strategies included low hydraulic losses, simple control logic, and low part count.
Some promising concepts had large unrestricted flow areas, but large wetted
perimeters, and so these designs were eliminated. Ultimately, the full-bore concept of
the deflating tube, shown in Figure 4-5, was selected. When the fluid was negatively
pressurized, the pressure outside the soft tube caused it to collapse and restrict the
flow. When the fluid was positively pressurized, the tube stayed inflated and the fluid
saw a full-bore valve, thus fulfilling both functional requirements with only one part.
4.2 Early iterations: searching for stable throttling
The first embodiment of the deflating tube concept was simply a very flexible tube
attached to barbed fittings and inserted upstream from a booster pump. The tube
stayed fully open when positively pressurized, and collapsed closed when negatively
pressurized; these two states are depicted in Figure 4-6. A flow test confirmed that
the tube prevented the negative pressure downstream from it from traveling to its
upstream side, as measured by the pressure recordings shown in Figure 4-7. While
the pressure performance of this tube met the functional requirements F1, F2, and
F4 from Table 4.1, it had two issues which were thought to reduce the fatigue life —
thereby possibly compromising the durability requirement of F3:
1. a large stress concentration at the downstream boundary condition between the
flexible tube and the rigid support, marked by the small radius of curvature at
the pinch points in the tube in Figure 4-6a; and
2. an unstable regime in which the tube would flutter open and closed, inducing
large pressure fluctuations into the system, as can be seen in Figure 4-7.
55
90 degree bend
Stable poppet’s are complex
Fold and Unfold
Pressure and momentum
forces hard to balance.
180 degree bend
Complex control logic
Wetted perimeter large
Inflate and Deflate
180 degree bend
Momentum influence?
Possible
Stretch or Pinch
How does edge seal?
Friction proportional
to length less efficient
than diameter
High part count
Figure 4-3: Structured brainstorming sketches, part 1
56
Pull or Collapse
Wetted perimeter is
large
Drag deflection
Wetted perimeter is large
Pinch points
Twist
Residual tension in tube
is vulnerable to creep
Snap
Possible
Wetted perimeter is large
Wetted perimeter is large
Figure 4-4: Structured brainstorming sketches, part 2
57
Figure 4-5: Selected concept: a deflating tube as a back pressure valve
(a) Negatively pressurized closed tube
(b) Positively pressurized open tube
Figure 4-6: Detail of fully closed and fully open collapsible tube. Flow is left to right
1
0
−1
Pressure (m)
−2
Upstream of Valve
Between Valve and Pump
−3
−4
−5
−6
−7
−8
−9
70
71
72
73
74
75
Time (s)
Figure 4-7:
Negative pressure was prevented, but rapid pressure oscillations were
observed in the throttled upstream pressure with the unmodified flexible tube
58
Table 4.3: Design variables considered in search of stable operation
Variable
Effect
Pre-buckling the tube
Different collapse pattern, but equally unstable
Reinforcing the tube
Instability removed
Smoothing the rigid to flexible transition
Equally unstable, but less stress concentrations
The supplemental functional requirement that the valve should stably throttle
the flow sought to address this second concern. To meet this additional functional
requirement, three design parameters were experimented with. The success of each
one is summarized in Table 4.3. Three of the prototypes used to gather these results
are pictured in Figure 4-8. The reinforcing tube with a gradual transition into the
flexible region, shown in Figure 4-8b, proved to both eliminate the instability and to
minimize the stress concentrations caused by the flexible tube getting sucked into the
downstream end.
4.2.1 Full factorial experiment to determine insert geometry
and tube choice
Using a full-factorial experiment, the optimal combination of insert size and soft tube
material was found to be a silicone rubber tube where the inside diameter of the
soft tube matched the cross-sectional perimeter of the insert exactly; the final insert
geometry is shown in Figure 4-9. The performance did not depend significantly on
the shape of the opening.
The parameter variation is summarized in Table 4.4.
‘Insert Size’ refers to the
cross-sectional perimeter at the center of the insert (shown in green in Figure 410). It was measured in the percentage that the flexible tube would have to stretch
circumferentially to occlude the pipe, calculated assuming no slippage between the
insert and the soft tube — later found to be an unrealistic assumption.
The over-sized by +40% inserts performed so poorly that they were excluded
from the presented results. The undersized (-20%) inserts were prone to fluttering,
59
(a) Pre-buckled design
(b) Stiffening support for the inside of the tube
(c) Gradual transition from rigid to unsupported
Figure 4-8: Three prototypes attempting to remove instability
A
R0.05
R0.05
A
16
0.53
0.43
SECTION A-A
1.45
(a) CAD render
0.84
0.75
0°
(b) Part drawing; units of inches
UNLESS OTHERWISE SPECIFIED:
DIMENSIONS ARE IN INCHES
TOLERANCES:
FRACTIONAL
ANGULAR: MACH
BEND
TWO PLACE DECIMAL
THREE PLACE DECIMAL
Figure 4-9: Optimal insert shape
INTERPRET GEOMETRIC
TOLERANCING PER:
PROPRIETARY AND CONFIDENTIAL
THE INFORMATION CONTAINED IN THIS
DRAWING IS THE SOLE PROPERTY OF
<INSERT COMPANY NAME HERE>. ANY
REPRODUCTION IN PART OR AS A WHOLE
WITHOUT THE WRITTEN PERMISSION OF
<INSERT COMPANY NAME HERE> IS
PROHIBITED.
MATERIAL
SolidWorks Student Edition.
For Academic Use Only.
NEXT ASSY
60
5
USED ON
APPLICATION
4
NAME
DATE
DRAWN
TITLE:
CHECKED
ENG APPR.
MFG APPR.
Q.A.
COMMENTS:
SIZE DWG. NO.
REV
Old Insert Correct
cross section 0.05ra
A
FINISH
3
SHEET 1 OF 1
SCALE: 1:2 WEIGHT:
DO NOT SCALE DRAWING
2
1
Figure 4-10: Optimal insert was one where the cross sectional perimeter matches the
inner circumference of the soft tube (0.75inπœ‹
= 2.36in)
Table 4.4: Insert geometry test matrix; repeated for silicone and latex soft tube
Insert Size
+40%
+20%
0%
Linear Opening, as in Figure 4-11
Linear Side Profile, as in Figure 4-12
(a) Top view
(b) Side view
Figure 4-11: Linear Opening insert geometry
(a) Top view
(b) Side view
Figure 4-12: Linear Side Profile insert geometry
61
-20%
Min
-4
-5
Linear Open Linear Slope Linear Open Linear Slope Linear Open Linear Slope
Total Seconds of Flutter
-20
0
20
18
16
14
12
10
8
6
4
2
0
20
Latex Rubber
Silicone Rubber
Linear Open Linear Slope Linear Open Linear Slope Linear Open Linear Slope
-20
0
20
Figure 4-13: Total seconds of flutter for each design parameter
Minimum Upstream
Pressure (psi)
0
-1
-2
Latex Rubber
Silicone Rubber
-3
-4
-5
Linear Open Linear Slope Linear Open Linear Slope Linear Open Linear Slope
-20
0
20
20
18
16
14 Figure 4-13. The +20% inserts allowed brief, but severe negative pressure
shown in
Latex Rubber
12
to pass upstream, as shown in Figure 4-14. Finally, silicone rubber
was selected
Silicone
Rubber as
10
the membrane
material because it performed better in every category tested. This
8
6
data therefore suggests that a silicone rubber tube should be used with an insert
4
of matching
size (0% stretch required). The CAD drawing of this optimal insert is
2
0
shown in Figure 4-9.
Linear Open Linear Slope Linear Open Linear Slope Linear Open Linear Slope
Total Seconds of Flutter
Figure 4-14: Minimum upstream pressure allowed for each design parameter
The experimental data did not conclusively suggest one opening style or the other.
-20
0
20
It was hypothesized that the linear slope opening (defined in Figure 4-12) would be
more resistant to flutter as the membrane stretched over time and therefore it was
used instead of the linear opening (defined in Figure 4-11). Should there be a desire
62
to change the opening geometry in the future, there is no reason given here not to.
Also of note, the latex rubber proved much more resistant to tearing than the silicone;
so should there be a future issue with tearing, the material choice described above
should be revisited.
Both the alpha and the beta prototypes used the insert shape detailed in Figure 49 and silicone for the soft tube. It should be noted, however, that because of a version
tracking issue in the CAD files, the January field trial was conducted with a +20%
insert. This mistake was corrected for the March trial.
4.3 Alpha prototype
The alpha prototype sought to convert the stabilized tube design of Section 4.2.1 into
a product that could be tested in an Indian household. Table 4.5 details the functional
requirements created to describe this design objective. F4c-F7 (see Table 4.5) were not
requirements in the earliest prototypes, and therefore the design required adjustment.
The conclusion of this design process was the product summarized with the bill
of materials and exploded assembly view in Figure 4-15 and the 3D CAD model in
Figure 4-16. The design featured an insert designed to minimize fatigue in the soft
tube, strain transition sections between the soft tube and rigid supports, a supporting
wire mesh to carry the positive pressure load, and a sealed outer shell to isolate the
device from overly curious customers.
The only moving part was the silicone rubber tube of hardness A-35, shown pink
in Figure 4-16.
When water was positively pressurized, the tube stayed inflated,
and did not obstruct the flow, but when water was negatively pressurized, the tube
collapsed through the cutaway in the insert (green in Figure 4-16), and throttled the
flow.
4.3.1 Durability
To minimize the likelihood of fatigue failure, the early prototypes used a gradual
transition from the rigid pipe to the partially supported collapsing tube, as in the
63
Table 4.5: Alpha prototype’s functional requirements and design parameters
F1
Functional Requirement
Design Parameter
Sustain positive
Back-pressure regulating valve, with of
back-pressure
atmospheric, i.e. if P
if P
F2
Minimize frictional losses
>
<
0, then close more;
0, then open.
Use a full-bore valve design
through the valve when
open
F3
F4a
Inexpensive
Simple and self-actuated, low part count
Durable: minimize
Simple and self-actuated, low part count
complexity
F4b
Durable: throttle flow
Use the stabilizing insert strategy
stably
F4c
F5
Durable: minimize strain
Use a gradually opening insert and have
mismatches and stress
strain transition sections at tube
concentrations
attachment points
Withstand cyclic loading of
Soft tube balloons under high pressure,
17m positive pressure and
provide support on the outside of the tube
full vacuum
F6
Be difficult to tamper with
No access to tube from outside. No
atmospheric vents that could be plugged
F7
Safe for drinking water
As a first prototype use NSF or FDA
contact
approvable materials
64
ITEM NO.
PART
DESCRIPTION
This part has NPT threads on
both ends and a cutaway of
linear slope type
Used as strain transitioning
material, this short tube is 1/16"
thick and has an ID=0.75"
QTY.
1
Insert with full radius
2
A-50 Si-Rubber
Outer
3
A-35 Si-Rubber Hose
This is the valve's actuator. It
collapses into the pipe under
negative pressures. t=1/16";
ID=0.75"
1
4
5
A-50 Si-Rubber Hose
Brass Ferrule
t=1/16"; OD=0.75"
Crimps the assembly together
2
2
6
PVC Reducer
Reduces from 2" to 0.5"
unthreaded pipe. McMaster
6826K192
2
7
2" PVC Pipe
Nominally 2" schedule 40
1
Constructed of rolled mesh, acts
as a positive pressure stress relief
for the soft rubber hose (3)
1
8
Mesh tube
7
1
3
2
4
1
2
5
6
8
Figure 4-15: Bill of materials and exploded view of alpha prototype
SolidWorks Student Edition.
For Academic Use Only.
65
(a) Doubly sectioned view
(b) Section view
(c) Outer shell
Figure 4-16: 3D CAD of alpha prototype
66
SECTION C-C
C
1/2" NPT male
C
R0.05
A
R0.05
A
B
0.13
SECTION A-A
5.47
6.53
7.28
°
DETAIL D
0.25 SCALE 5 : 1
86
.6
1.98
2.72
3.78
DETAIL B
SCALE 2 : 1
9.25
UNLESS OTHERWISE SPECIFIED:
NAME
DATE
Figure 4-17: Alpha prototype
insert
drawing; units of inches
DRAWN
DIMENSIONS
ARE IN INCHESpart
TOLERANCES:
FRACTIONAL
ANGULAR: MACH
BEND
TWO PLACE DECIMAL
THREE PLACE DECIMAL
INTERPRET GEOMETRIC
CHECKED
TITLE:
ENG APPR.
MFG APPR.
Q.A.
PROPRIETARY
AND CONFIDENTIAL of Section View C-C
TOLERANCING
middle
segment
inPER:the insert’s
part drawing in Figure 4-17.
COMMENTS:
THE INFORMATION CONTAINED IN THIS
DRAWING IS THE SOLE PROPERTY OF
<INSERT COMPANY NAME HERE>. ANY
REPRODUCTION IN PART OR AS A WHOLE
WITHOUT THE WRITTEN PERMISSION OF
<INSERT COMPANY NAME HERE> IS
PROHIBITED.
SIZE DWG. NO.
REV
SolidWorks Student Edition.
Another
possible
location was hypothesized to be at the sealing joint
For Academic
Usefatigue
Only.
USED ON
NEXT ASSY failure
MATERIAL
AInsert for thesis
FINISH
APPLICATION
DO NOT SCALE DRAWING
SCALE: 1:2 WEIGHT:
SHEET 1 OF 1
between 5the rigid support 4and the soft tube,
mismatch. The1 soft
2
3 because of a strain
tube was made of silicone with Shore A-35 hardness. To ease the transition, a stiffer
shore A-50 tube, also of silicone, was used at the interface between the rigid insert
and the soft tube, and between the soft tube and the rigid outer crimp. These layers
and the assembly stages are detailed in Figure 4-18.
To form a durable seal between the insert and the soft tube, a crimp was used to
provide a compressive force. For creep resistance, a barbed profile was used on the
inside face of the insert. Upon later analysis, however, it was revealed that while the
mechanical integrity of a barbed joint under cyclic loading is resistant to creep, the
seal — which relies on the compressive force — is not. The barb design is shown in
Detail B of the insert’s part drawing in Figure 4-17.
To control stress concentrations in the soft tube, a minimum bend radius was
incorporated into the insert design, so that even in the fully-closed position, the tube
would not need to bend sharply. A fillet radius of 0.05" was used to approximate a
67
Figure 4-18: Alpha prototype’s strain transition layers; far-left the inner strain transition tube is attached, mid-left the flexible tube is stretched over the inner strain
transition material, mid-right outer strain transition sleeves are installed, far-right
the whole assembly is rolled in wire mesh and crimped
68
fully-rounded edge, as shown in Detail D of Figure 4-17. This choice however would
make injection molding the part as a single piece more difficult. Without the inner
fillet radius, it could be made using only a center pull. If instead the insert were made
as two pieces, then the inner fillet would be insignificant. The necessity of the inner
fillet was inconclusively investigated in the fatigue test of Section 4.3.5.
4.3.2 Withstanding system pressure
In urban areas in India, drinking water should typically be supplied at 17 meters of
pressure, see Table 2.2; in intermittently-supplied areas, this pressure cycles with the
system’s intermittency. The valve was therefore required to withstand 17m of cyclic
pressure. To meet this requirement additional support for the soft tube was required.
This additional support, however, could not compromise the tube’s ease of collapse.
Three concepts were investigated to achieve this goal:
1. Create a custom soft tube with a braid inside of it; the braiding material would
be easily deformable, but strong under tension. This concept was not pursued
because prototyping it would have been extremely expensive and a manufacturer
could not be found for a braided tube as soft as A-35.
2. Put the soft tube inside a wire mesh that would fit under the crimped end
connections. For prototyping ease, a rolled mesh sheet was used instead of a
custom mesh tube.
3. Put the soft tube inside a rigid tube with breathing holes. In this concept, the
crimp could be replaced with the rigid tube itself.
Figure 4-19 depicts Options 2 and 3. To get the external tube to act as a crimp, a sixjaw chuck was used to crush its ends. Unfortunately, this led to an uneven distribution
of compressive forces as might be expected from the shape of the deformed tube seen
on the right of Figure 4-19.
Under a burst test, Option 2 failed at 48m of water
pressure, while Option 3 did not fail at 61m, the maximum pressure of available shop
69
Figure 4-19: Detail of alpha prototype’s two possible external support mechanisms
air. Taking burst pressure as a proxy for robustness of the design, the mesh option
was pursued.
4.3.3 Tamper resistant
To make the design tamper resistant, two risks needed to be eliminated. First, the
actuating membrane could be punctured or cut if exposed to the user. Second, the
valve’s control logic involved throttling to atmospheric pressure; were this pressure to
be calculated using a reference port, for example, plugging the reference port could
limit the valve’s functionality.
Using a sealed reference-pressure chamber mitigated both of these risks.
If the
valve had no openings to the outside, puncturing the membrane or plugging the valve
would be more difficult. An infinitely large reservoir of gas at atmospheric pressure
would allow the system to function as if there were an external port.
reservoirs follow the ideal gas law:
𝛿𝑃
𝑃
=
Finite-sized
𝛿𝑉
. The actuated valve was estimated to
𝑉
have a displaced volume of approximately 3mL, therefore to achieve a <1% error
would require at least a 300mL reference chamber. Prioritizing the use of standard
components, a 2in PVC pipe was used as an outer shell for the device. It provided a
reference volume of 218mL, which led to a 1.3% error, equivalent to the valve having
a set point of -0.14m of water pressure.
70
4.3.4 Safe for drinking water contact
The final additional requirement for the alpha prototype was that it be safe for drinking water contact. NSF 61 certification is a stringent standard for materials that will
be in contact with drinking water. The certification is only mandatory in the US for
the testing of lead leeching in devices covered in NSF 61 §9, which applies devices
that are "typically installed within the last liter of the distribution system" [41] and
that intend to "dispense water for human ingestion" [42, p.
46].
In a house with
half-inch piping, this 1L extents about 8m back from the tap, far enough that the
valve might have qualified for the 1L range, but as it was not a water dispensing
product, compliance was presumed not to be mandatory.
To ensure the safety of the valve, efforts were made to source materials that
had been used in other NSF-61-compliant products.
approvable
However, the supply of NSF
materials was extremely limited. Instead, all wetted materials used were
selected to be made of FDA
approvable
materials; and none had lead components.
The FDA and the NSF only certify products and not materials; therefore it was not
possible to use NSF or FDA certified materials, only material that had been certified
in other products.
For 3D printing or casting, two good options were found: Steralloy FDG Rigid
for urethane casting was FDA approvable, and NWRapid’s Nylon 12 for 3D printing
had been used in products approved for food-contact by the FDA. Ultimately, the
cast steralloy was used because of long-term swelling concerns with a submerged 3D
printed part. While the urethane casting process was outsourced, a process photograph is included for completeness in Figure 4-20.
4.3.5 Laboratory fatigue test
To maximize the fatigue life of the valve, the effect of the soft tube’s material and
the insert’s inner fillet radius were investigated using a full-factorial experiment. The
test matrix is outlined in Table 4.6.
The first unit failed at 3970 cycles, but because of Christmas holidays the test was
71
Figure 4-20: Process steps in the outsourced casting of the urethane inserts for the
alpha prototype
Table 4.6: Alpha fatigue test matrix: tube material and inner fillet radius
0.05" inner fillet radius
0.005" inner fillet radius
Latex
N=3
N=3
Silicone
N=3
N=3
72
Figure 4-21: Identifying where fatigued prototypes leaked
Mesh seam line
Mesh seam line
crack
Figure 4-22: Detail of failure location and suggested cause in fatigue test
not turned off until 20,000 cycles. A post-mortem analysis revealed four prototypes
had catastrophically failed; one from each quadrant of the text matrix. Leak locations
were identified as in Figure 4-21. In all four failed prototypes, the wire mesh seam
aligned with the edge of the stabilizing insert’s opening, creating a pinch point; cracks
occurred along this pinch point, as shown in Figure 4-22. Examination of the other
prototypes showed some signs of wear, but none of them had the seam alignment
found in the four failed prototypes.
While the fatigue test did reveal some very useful information about a previously
unconsidered variable in the manufacturing of the alpha prototype, it did not answer
either question about fillet radius or soft tube material.
Since the opportunity for
a field trial in January was rapidly approaching, further experimentation was not
possible and so the larger fillet radius was used with a silicone tube.
73
Figure 4-23: Alpha prototypes after field testing
4.3.6 Field test
The field performance of both the alpha and beta prototypes are combined and discussed in Chapter 6. Seven of the alpha prototypes that were installed during the
January 2014 field test are pictured in Figure 4-23. While the pressure performance
of these prototypes was very successful, five of eleven of the units failed in shear
fracture at the threads. After the first few valve breaks, the remaining valves were
installed using a flexible hose connection on the booster pump side of them, as in
Figure 4-24. This flexible hose allowed for minor misalignments in the pipe network,
but had higher frictional losses through it.
4.3.7 Improvements needed
Two improvements were needed after field testing because five of the eleven field units
failed in shear fracture at the threads, and several of the units had their reference pressure drift. It was hypothesized that the shear failures were caused by a combination
of the cast urethane being brittle and large residual stresses from the misalignment of
metal pipes in households. Additionally, Indian plumbers typically only use one pipe
wrench and rely on the pipe network to provide the counter force. Together these two
large loads applied to a cast material caused frequent shear failure in the prototypes,
74
Figure 4-24: Alpha prototypes’ field test required a flexible hose to reduce the shear
load on the valve; from bottom to top, pressure logger, alpha prototype, check valve
to air, flexible hose, booster pump
as seen in Figure 4-25.
The drift in reference pressure was likely caused by a slow leak from the valve
into the reference chamber. Once the reference pressure was above atmospheric pressure, the unpressurized valve would stay partially closed when there was atmospheric
pressure in the pipe, as in Figure 4-26a. The hypothesized leak pathway is detailed
in Figure 4-26b. Several prototypes were observed to have changes in their reference
pressure within a few days of being connected to the water system. This slow leak only
occurred when there was a large pressure difference between the reference pressure
and the pipe pressure. Therefore, the leak initially acted as a one-way leak, allowing
only leakage into the reference pressure chamber, not out of it. As such, even a very
small would eventually cause a significant pressure drift. The beta prototype sought
to maintain the positive field performance of the alpha prototype, while eliminating
these two issues.
4.4 Beta prototype
While Delhi’s water utility was happy with the field test results of the alpha prototype, they were concerned that the test had only been conducted on older connections
75
Figure 4-25: 55% of alpha field units failed in shear at the threads
(a) Several units had the reference set point
drift, causing the default position of the valve
to be partially closed
(b) The hypothesized leak path was under the
crimped seal; highlighted in red
Figure 4-26: Alpha prototype’s set-point drift because of a slow leak into the reference
chamber
76
and focused mostly on continuously supplied connections. To address this concern, a
second field test was planned and the beta prototype was built to use in this subsequent test. Additionally, the beta prototype sought to address the two failure modes
observed in the alpha prototype: shear failure and drift in the set-point pressure.
Addressing these required making compromises in the functional requirements, summarized in Table 4.7. The functional requirement that the valve should be able to
withstand axial and radial moments (F4d in Table 4.7) was added to account for
the loading forces of plumbers and plumbing. To address the drift in the set-point
pressure, a vent was added to the reference chamber, which required a reduction in
the tamper-resistance of the valve (F6 in Table 4.7). The following two subsections
elaborate on these design decisions and trade-offs. The third subsection, 4.4.3, describes the detailed design of the beta prototype, and finally this section concludes
with a summary of the valve’s fatigue and field performance and the next steps for
further improvements.
4.4.1 Increasing the valve’s toughness
In the alpha prototype, the pipe threads of the cast urethane insert transfered the
external load through the threads into the outer PVC shell; this proved not to be
strong enough. To maintain the functionality of the valve’s collapse mechanism, but
increase the valve’s ability to withstand loading, three options were considered:
1. strengthen the insert,
2. carry the load through the middle reinforcing layer (previously the wire mesh
in the alpha prototype), or
3. carry the load through the outer protective shell (previously the PVC tube in
the alpha prototype).
To simplify the design, separate components of the valve were designed for separate
functions. Since the insert’s primary function was stabilizing the collapsing tube, the
77
Table 4.7: Beta prototype’s functional requirements and design parameters
F1
Functional Requirement
Design Parameter
Sustain positive
Back-pressure regulating valve, with a set
back-pressure
point of atmospheric pressure, i.e. if P
then close more; if P
F2
Minimize frictional losses
>
<
0,
0, then open.
Use a full-bore valve design
through the valve when
open
F3
F4a
Inexpensive
Simple and self-actuated, low part count
Durable: minimize
Simple and self-actuated, low part count
complexity
F4b
Durable: throttle flow
Use the stabilizing insert strategy
stably
F4c
F4d
F5
Durable: minimize strain
Use a gradually opening insert and have strain
mismatches and stress
transition sections at the tube’s attachment
concentrations
points
Durable: able to
Separate structure from function; carry the
withstand large moments
load through the metal outside; no loads on
axially and radially
the plastic insert
Withstand cyclic loading
Soft tube balloons under high pressure,
of 17m of positive
provide support on the outside of the tube
pressure and full vacuum
F6
Be difficult to tamper
Limited direct access from outside; breathing
with
holes in outer shell and middle shell do not
align
F7
Safe for drinking water
As a first prototype use NSF or FDA
contact
approvable materials
78
load-bearing requirement was left to other components. This left either Option 2 or
3; the strategy for eliminating set-point drift differentiated between them.
4.4.2 Preventing set-point drift
Two design parameters were adopted simultaneously to address the issue of the reference pressure drifting: the strategy of using a sealed reference chamber was abandoned, and the quality of the sealing mechanism was increased. If a sealed reference
chamber had been used, the seal would have had to be extremely reliable as even
an very slow leak, one milliliter per month for example, would lead to a significant
set-point drift over the course of a year, in this example +0.4m. Therefore, it was
determined instead that the air chamber should be vented to the atmosphere. This
required, however, a sacrifice in the design’s ability to withstand tampering as the air
vent could provide access to the soft membrane. Further, if the vent were plugged,
the valve’s performance would suffer. One possible future strategy to explore for increasing tamper resistance would be the addition of a large vented chamber such that
if the vent were plugged, the chamber’s dead volume would act as a sealed reference
chamber and the valve’s functionality would not be significantly compromised.
The robust sealing mechanisms of a glued joint and an O-ring were preferred
over a crimp.
While getting a strong circumferential seal from a crimped joint is
possible, as is frequently done with hydraulic hoses, with a very soft tube, assuring the
even distribution of the compressive force and accounting for creep was prohibitively
challenging.
Sealing the soft tube to either the insert, the middle layer, or the outermost layer
was required. Both an O-ring seal or an adhesive seal could have been used, but to
account for the possibility of creep — which would change the O-ring’s compression
— it was determined that the soft tube should be sealed using an adhesive.
The
only available NSF-certified silicone adhesives had relatively poor peel strength. To
minimize the loads applied to the glued joint (using the principle of reciprocity), it
was determined that the soft tube should be adhered to either the middle layer or the
outermost layer. In this configuration, the larger loads of positive pressure would act
79
Figure 4-27: The peel force was dramatically reduced by turning the membrane insideout at each end of the middle layer
to hold the joint sealed, while the smaller negative pressure loads would be applied
to the joint.
To further strengthen this seal, it was desired to minimize the peel forces acting
on the glued joint. To accomplish this, the soft tube was turned inside-out at each
end, as shown in Figure 4-27, where the centerline is horizontal and below the figure,
the rigid layer is shown in blue, the soft tube in black, and the adhesive in red. In this
configuration, when the soft tube is pulled by a negative pressure load down into the
flow path (towards the bottom right of Figure 4-27), the glued joint experiences only
a shear force. The only mechanism by which a peeling force could be applied to the
glued joint would be if the soft tube were loose enough to allow the end sections to
balloon out. Even in this scenario however, the surface area of the collapsing section
in the center of the valve (not shown in Figure 4-27) would be much larger than the
surface area of the potentially ballooning end sections; therefore, the collapsing force
would act to limit the ballooning and restore purely shear loading.
The inside-out glued-joint design meant that it would be complicated to have the
soft tube adhered to the load-bearing member in the structure and therefore the outer
layer was taken as the load-bearing element while the soft tube was adhered to the
middle layer. In an effort to utilize off-the-shelf components, it was determined that
the outer load-bearing shell could be made of standard metal pipe and a simple seal
between the outer layer and the middle layer could be accomplished with an O-ring,
as in Figure 4-28a, where the black circle represents the O-ring.
80
(a) Sketched sealing mechanism
(b) CAD section view
(c) CAD doubly sectioned view
Figure 4-28: Section views of the beta prototype
81
ITEM NO.
PART NUMBER
DESCRIPTION
QTY.
Seamless steel
pipe
Reducing
Coupling
1.25" Seamless Steel pipe;
Reamed ID
1.25"NPT female to 0.5"NPT
female
3
Stabilizing Insert
Cast urethane, linear-side-profile
style
1
4
Middle layer
Custom turned part; aluminum
1
5
Soft tube
A-35 Silicone rubber; t=1/16";
ID=0.75"
1
6
O-ring
Viton O-ring Dash 025
2
1
2
1
2
1
2
6
5
4
3
SolidWorks Student Edition.
For Academic Use Only.
Figure 4-29: Beta prototype’s exploded view
4.4.3 Detailed design of the beta prototype
The section and exploded views of the valve assembly are shown in Figures 4-28 and
4-29. The assembly stages of the physical valve are shown in Figure 4-30.
The outermost component of the beta prototype was a 1.25 inch seamless, steel
pipe (black in Figure 4-28b) with a reamed and sanded inside face, shown in the part
drawing in Figure 4-31. Breathing holes were added to the center of this pipe. This
pipe connected to two reducers which ended in 1/2" NPT female ends, thus forming
the outer and load bearing shell of the valve.
Sealing against the reamed and sanded inner surface of the pipe, two O-rings joined
the outer shell to a middle supporting layer.
The O-rings, grooves, and clearance
spacing were designed according to specifications in the Parker O-ring Handbook [43,
82
Figure 4-30: Beta prototype’s assembly stages, left to right: the cast stabilizing insert;
the middle layer with the white soft tube glued in place and the O-rings attached; the
middle layer inserted into the outer shell; and finally the outer shell with the reducing
end-pieces and the set screw installed
4.00
2.00
1.66
1.317
1.25" Male NPT
1.25" Male NPT
Figure 4-31: Beta prototype’s outer pipe
UNLESS OTHERWISE SPECIFIED:
DIMENSIONS ARE IN INCHES
TOLERANCES:
FRACTIONAL
ANGULAR: MACH
BEND
TWO PLACE DECIMAL
THREE PLACE DECIMAL
INTERPRET GEOMETRIC
TOLERANCING PER:
PROPRIETARY AND CONFIDENTIAL
THE INFORMATION CONTAINED IN THIS
DRAWING IS THE SOLE PROPERTY OF
<INSERT COMPANY NAME HERE>. ANY
REPRODUCTION IN PART OR AS A WHOLE
WITHOUT THE WRITTEN PERMISSION OF
<INSERT COMPANY NAME HERE> IS
PROHIBITED.
MATERIAL
SolidWorks Student Edition.
For Academic Use Only.
NEXT ASSY
5
USED ON
APPLICATION
4
NAME
DATE
DRAWN
TITLE:
CHECKED
1.25" Seamless Steel pipe; Reamed ID
ENG APPR.
MFG APPR.
Q.A.
COMMENTS:
SIZE DWG. NO.
83Dimensioned
A
FINISH
SHEET 1 OF 1
SCALE: 1:1 WEIGHT:
DO NOT SCALE DRAWING
3
REV
Reamed pipe
2
1
0.041
#59 drill
spaced at 45 deg +/- 5 deg all around
Debur both sides!
A
0.246
0.142
DETAIL B
SCALE 5 : 1
R0.025
R0.025
Both sides
R0.010 ±0.005
SECTION A-A
32
D
+0.002
0.050 0.000
Approx R0.005
C
B
+0.005
0.093 0.000
+0.003
1.307 - 0.003
R0.050
3.950
3.357
2.538 ±0.020
2.288 ±0.020
2.038 ±0.020
1.218
0.625
1.133
R0.050
1.008
Both sides
4.575
4.433
4.329
R0.050
+0.007
0.875 0.000
R0.050
A
Both sides
16
DETAIL C
SCALE 2 : 1
Both sides
0.593
0.250
DETAIL D
SCALE 5 : 1
Both sides
Figure 4-32: Beta prototype’s aluminum middle layer; units of inches
pp. 4-9] and then the clearance was adjusted according to the pressure and durometer
ratings in Oberg et al. [44, p. 2589]. An FDA approved Viton O-ring of size Dash 025
was selected for its resistance to ozone, weather, water, and abrasion. Its hardness
was A60.
For ease of manufacturing, the middle layer of the beta prototype (gray in Figure 4-28b) was made of aluminum. The mixing of metals in a water pipe network
must typically be avoided in order to minimize corrosion, but to quickly develop this
prototype, its ease of manufacturing was prioritized over its longevity. The middle
layer was made on the CNC lathe and then breathing holes were drilled using a rotary
encoder mounted on a CNC milling machine. The complete part drawing is specified
in Figure 4-32. Axial translation and rotation of the middle layer with respect to the
outer pipe were prevented using a setscrew.
84
The same silicone A-35 tube as in the alpha prototype was used (pink in Figure 4-28b) and bonded to the middle layer using Dow-Corning 732 silicone adhesive.
Contact adhesives were considered, but for an underwater bond of two dissimilar materials, their durability would be questionable at best. The 732 was NSF 61 certified
and was able to bond aluminum and silicone. The shape of the fold-over was changed
between Figure 4-28a and Figure 4-28b so that the elasticity of the soft tube could
be leveraged to hold itself in place during the drying of the glue.
The innermost component of the beta prototype was the stabilizing insert. The
cutaway geometry was the same as that of the alpha prototype, except it was cut short
because it no longer needed to interface with the pipe network. No simple method was
found in time for March field trial to secure the orientation of the insert with respect
to the middle layer. The middle layer’s breathing holes were made axisymmetric so
that rotation of the insert would not be problematic. The insert was, however, still
free to translate axially.
4.4.4 Laboratory fatigue test
The same apparatus as the alpha prototype was used to cyclically test the beta.
Because of the small breathing holes in the middle layer of the beta prototype, the
test procedure was modified to use asymmetric pressurization cycles of +27m of air
pressure on the inside of the valve for 7 seconds, and of +8m of air pressure on the
outside of the valve (simulating an internal vacuum pressure) for 5 seconds.
A two-by-two full-factorial experiment was done to examine the effect of the insert’s fillet radius and the size of the clearance gap for the soft tube (i.e. the difference
in diameters between the insert’s OD and the middle supporting layer’s ID, less twice
the thickness of the soft tube); the parameter variation is summarized in Table 4.8.
The soft tube’s thickness was 1/16" and therefore the middle layer’s ID=0.875" left no
clearance gap. The larger ID was selected so that it could be reamed with a 22.5mm
metric reamer.
The twelve fatigue tested valves were built according to the design of Section
4.4.3, except the valves did not have a set screw to ensure alignment.
85
After 3000
Table 4.8: Beta fatigue test matrix: clearance for membrane and the insert’s inner
fillet radius
0.05" inner fillet
0.005" inner fillet
radius
radius
ID=0.875" (no clearance)
N=3
N=3
ID=0.886" (0.011"
N=3
N=3
clearance)
cycles, one of the middle layers translated axially far enough that the O-ring became
un-compressed and leaked. This occurred because the McMaster-Carr reducers used
in the fatigue test were longer than the original McMaster-Carr ones measured and
designed for (despite having the same part number).
To correct this failure mode, spacers were inserted in all of the valves to limit the
range of translation and even the “failed” prototype was realigned and testing was
resumed. With the spacers, no failures were observed over 26500 cycles of -8m and
+27m of pressure on twelve valves. The test was stopped because while new middle
and inner layers had been ordered for the March field trial, extra outside shells had
not been. Therefore, the true fatigue life was not discovered, it was only determined
that the life was greater than 26500.
Since no valves failed, the fatigue test did not indicate which clearance gap to use.
Instead, since the soft tubes in the beta valves had been qualitatively observed not
to collapse as easily as tubes in the alpha prototypes, the larger clearance gap was
selected to try to improve the tubes’ ease of collapse. Further, it was hypothesized
that breaking in the valves might loosen the membrane and improve this problem.
Therefore, six of the twelve newly assembled and unfatigued beta valves were tested
for 500 cycles before the March field test. No performance difference between the six
un-fatigued and the six fatigued valves was noticed in the subsequent field trial.
86
Figure 4-33: Field modification of the beta prototype
4.4.5 Field test
The field performance of the beta prototypes was tested in detail in the March field
trial.
For this field trial, the spacers used in the fatigue test could not be used
because they were not NSF or FDA approved and their metallic content was uncertain.
Instead, the valves were fitted with a setscrew through the outside shell that would
contact the middle layer. These setscrews were installed at a machine shop in New
Delhi, pictured in Figure 4-33.
4.4.6 Improvements needed
Without fail, whenever someone in India — engineer, social scientist, or curious bystander — examined the beta prototype, their feedback was that it was “heavy” and
therefore was likely to get stolen. This weight arose from the explicit design choice
to overdesign the load-bearing structure.
An important component of future work
would be to better quantify the required strength, so that the valve’s structure could
87
be optimally strengthened. To further reduce the possibility of theft, wherever possible, moving from metal to plastic parts would be desirable as it would minimize the
valve’s recycling value.
The beta prototype used a set screw to prevent relative motion between the outer
shell and the middle reinforcing layer. No such alignment feature was present between
the stabilizing insert and the middle layer. After several field tests, it was observed
that some of the inserts had migrated to the downstream end of the valve assembly,
out of their optimal positioning.
A feature to align the insert with respect to the
middle layer should be included in future designs.
One of the core design requirements of the valve was to prevent negative pressure.
To this end the valve’s set point was chosen to be atmospheric pressure. However, as
will be explored more thoroughly in Chapter 6, this logic left no room for variation
or error in the set point — equivalent to a zero sigma control process where half of
the variation would automatically fall outside of specifications.
Future prototypes
instead could either attempt to have a mildly variable set point (perhaps by changing
the axial tension in the membrane), or could intentionally have a set point several
standard deviations (sigma) away from the target pressure.
Finally, bench-level experiments indicated that the beta prototype’s soft tube did
not collapse as easily as the alpha’s. Three hypothesis exist for why this might have
been the case: first, the assembly procedure might have left residual axial tension in
the membrane; second, the smooth surface area of the middle layer might have been
sticking to the membrane; or third, the limited number of small breathing holes in
the middle layer might not have provided enough area for the pressure difference to
cause a force large enough to create the local deformations in the soft tube required to
get larger surface areas exposed to the pressure difference. Since the delayed collapse
seemed to be independent of lubrication and constant across prototypes with axially
loose membranes, the third hypothesis should be prioritized as the most needed area
for improvement in future prototypes.
88
Chapter 5
Experimental and numerical methods
5.1 Experimental methods
5.1.1 Laboratory setup and experimental method
A test setup was designed and built to simulate the typical residential booster pump
configuration observed in many houses in New Delhi, where the pump is connected
directly to the service connection pipe without a sump, as in Figure 2-7.
The key
features the test setup attempted to imitate were:
1. a static pressure head upstream from the pump,
2. a moderate friction loss between the supply pressure and the pump,
3. a pump that was able to induce negative pressure on its suction side, and
4. frictional losses after the pump.
Additionally, for ease of experimentation, it was also required that the setup have the
ability to:
1. insert a prototype valve upstream from the pump;
2. measure the pressure before the valve, between the valve and the pump, and
after the pump;
89
1” pipe
FM
0.5” pipe
0.5” Al. pipe
0.5” Clear pipe
RBP
Full-bore valve
SP
P
Flow meter
Pressure transducer
Booster pump
Sump pump
P
P
P
FM
RBP
Valve
SP
Filter
Figure 5-1: Laboratory test setup
3. bypass the valve to get baseline data; and,
4. bypass the pump to get baseline data.
A schematic of the experimental apparatus which met these requirements is shown
in Figure 5-1. The actual setup is pictured in Figure 5-2.
The pressure head upstream from the pump was created with an elevated supply
reservoir made of a 27x17x15 inch tote bin, filled with water, and placed on top of a
9 foot scaffolding. From there the water flowed out through a port in the bottom of
the bin, through a one-inch plastic pipe down to the experimental setup, and into a
three-foot straight section of half-inch aluminum pipe. Along this aluminum pipe was
an Omega FDT-31 ultrasonic flow meter. The water continued through a throttling
gate valve, past a pressure transducer, and into the valve test section where, using
full-bore ball valves, the flow was either directed through the valve to be tested, or
around it. The parallel sections merged again and the water continued to flow past
another pressure transducer. Next, the water flowed through a 1.5 horsepower pump,
past a final pressure transducer, and through a section of clear plastic pipe to make
cavitation more visible. Finally, the water discharged as a free stream into the air
and spilled into the discharge reservoir.
90
Flow Meter
Valve
Throttling valve
Booster pump
Discharge reservoir
Filter
Figure 5-2: Top view of the valve testing area; booster pump is on the right, discharge
reservoir is the white tote bin
Table 5.1: Laboratory measurement equipment
Item
Range
Accuracy
Supplier
Ultrasonic Flow Meter
0.5 - 25 GPM
reading
Omega
Discharge-side Pressure
0-100 psia
±0.25%
±0.25%
full scale
Omega
Full vacuum-45 psig
±0.25%
full scale
Omega
Transducers
Suction-side Pressure
Transducer
The discharge reservoir had a submersible pump with a float switch that would
automatically pump water to refill the supply reservoir. The set points of the submersible pump were 1.5 inches apart, ensuring the consistency of the pressure head
at the top of the experiment. Before reaching the supply reservoir, the water passed
through a filter. The accuracy of the measurement equipment is summarized in Table
5.1.
The tendency of early prototypes to flutter, documented in Section 4.2.1, was
investigated using this test setup. The testing procedure was that for a given upstream
resistance, the throttling gate valve on the discharge side of the pump was slowly
opened and then slowly closed over the course of three minutes.
Then, the valve
was throttled back and forth around positions that caused the valve to open or close
91
for another three minutes. All prototypes were tested with high and low upstream
resistance.
5.1.2 Fatigue test setup
To bound the estimates of valve life, it was desired to force the valves to cyclically
experience positive system pressure and then negative booster-pump pressure. After
two failed attempts to create an apparatus that used water as the internal fluid,
a third test apparatus was successfully built.
It cyclically pressurized air on the
inside, and then air on the outside of the valve (to simulate negative pressure on the
inside of the valve).
This apparatus used shop air, a 4-way solenoid, two pressure
regulators, and push-to-connect tubes to connect the test units to the air pressure.
The experimental schematic for a single valve is shown in Figure 5-3, in which the
pictured solenoid control logic was reproduced from [45]. For the actual test, twelve
units were connected in parallel and the setup is pictured in Figure 5-4. A limitation
of this apparatus design was that if one unit failed catastrophically, air would escape
through that unit and prevent the other units from being tested until the failed unit
had been removed; the apparatus could therefore only autonomously test until the
first unit failed.
5.1.3 Field testing setup and method
The goal of the field test was to evaluate the claim that the designed valve prevented
booster-pump-induced negative pressure. A Hydreka pressure logger, shown in Figure
5-5, was installed above the ground and immediately upstream from where the valve
would be placed.
Pressure measurements were made with and without the valve
installed. A schematic of the desired layout and its predicted effect on pipe pressure
is shown in Figure 5-6.
Informed consent from participating houses was verified through the use of an
independent translator. Additionally, the valve was installed with an ‘opt-out’ bypass
whose use was demonstrated to each participating house; by turning a quarter-turn
92
Solenoid Driven at 1/12 Hz
Shop air
H=10m
R
H=17m
R
Figure 5-3:
Vent
Layout of the fatigue test; ‘R’ are the two regulators; pressures were
typically set to 10m and 17m
Figure 5-4: Built fatigue testing apparatus; blue tubes carry 17m pressure for the
inside of the valves; pink tubes carry 10m pressure for the outside of the valve
93
Figure 5-5: Pressure loggers from Hydreka, reproduced from [46]
valve, the functionality of the valve would be eliminated.
The field setup varied slightly between the alpha and beta prototypes because the
alpha prototype required a stress-relieving, flexible hose connection. Alpha and beta
test setups are shown in Figures 5-7 and 5-8. In the setup for the beta prototype,
the control (baseline) data was taken by turning on the opt-out bypass valve, thus
measuring the pressure in the pipe without the valve. Because of a strength issue in
the alpha prototype, its opt-out bypass had to be made with a flexible hose. Using
this bypass as the control was not possible, because its added friction would have
skewed the baseline data.
Therefore in the alpha prototypes’ test, a separate test
unit with the same fittings, except with a straight plastic pipe instead of the valve,
was used to gather baseline data; it is pictured in Figure 5-7b.
In the design of the field study, it was assumed that the prevalence of negative
pressure would vary more between houses, because of booster pump sizes and household demand patterns, than between days at any given house.
A crossover study
was therefore designed, where the prevalence of negative pressure at a house would
be measured when the valve was not installed, and then this baseline would be compared to the prevalence of negative pressure with the valve installed at that same
house. The order of control and treatment was varied between houses.
In January 2014, Delhi’s water utility, the Delhi Jal Board (DJB), agreed to test
prototype valves on 10 houses within their staffing quarters in Pitampura. The staffing
quarters were located adjacent to a large reservoir and pumping station, and so the
houses in these quarters had exceptionally good pressure. Only apartments three or
more stories above the ground used booster pumps. Even among these apartments,
94
BP
Supply Pressure
M
V
Pressure
P
= Booster Pump
= Water Meter
= Valve
=Pressure Logger
Tank
2 Story
Home
P
V
Ground Level
Distribution
Main
M BP
P<0
Service Connection Pipe
Figure 5-6: Field test schematic and predicted pressure performance
Stress
Relieving
Hose
Stress Relieving Hose
Booster Pump
Control Pipe
Valve
Pressure
Sampling Point
Bypass
Pressure
Logger
Pressure
Logger
(a) Alpha prototype during installation (b) Control measurement for alpha prototype
Figure 5-7: Three prototypes attempting to remove instability
95
Pipe to overhead tank
Booster Pump
Valve
Pressure Logger
Bypass is open to get control measurement
DJB Supply Pipe
Pressure Sampling Point
Figure 5-8: Field equipment arrangement for beta prototype; currently taking control
measurements — valve is bypassed
negative pressure lasting more than a few seconds was only observed at one service
connection.
valve.
As such, these quarters were not an effective testing location for the
Two more connections were identified by the DJB in Pitampura which had
more representative supply pressures. To supplement these data points, a convenience
sample of houses in Azad Market was taken.
neighbors.
All five Azad Market houses were
The final January connection was also a convenience sample and the
house was that of a potentially interested industrial partner’s home. The quantitative
characteristics of these houses are summarized in Table 5.2.
After reviewing the gathered data, the DJB was concerned that the data was not
representative of Delhi because it did not sample enough intermittently connected
houses. Further, the DJB wanted to know if the baseline problem of negative pressure
was present only on old connections with high frictional losses (like those sampled),
or if it would also be a problem on newly connected houses. The latter question is
of particular importance because were the valve to be purchased by a water utility it
would likely be installed as part of a service connection replacement project. However,
would that service connection replacement negate the valve’s importance?
To address both of these concerns, a second round of field trials was done in March
2014 with the beta prototype. Finding new connections meant working with some
96
Table 5.2: House connection characteristics, where high pressure is >6m, and medium
pressure is between 2 & 6m, and low pressure is between 0.5 & 2m; C/I indicates if
the connection was Continuous or Intermittent; O/N indicates if the connection was
Older than 2 years or Newer than 1 year
Data Label
Location
High
Medium
Low
C
N
Pressure
Pressure
Pressure
or
or
(hrs/day)
(hrs/day)
(hrs/day)
I
O
J01
Azad Market
5
16.5
1.9
C
O
J02
Azad Market
4.6
17.3
0.5
C
O
J03
Azad Market
4.7
17.1
0.6
C
O
J04
Azad Market
4.7
18.8
0.3
C
O
J05
Azad Market
0
12.9
9.9
C
O
J06
Pitampura
10.8
10.1
1
C
O
J07
Pitampura
4
11.7
4.1
C
O
J08
Vivek Vihar
1.1
2.9
0.7
I
O
J09
Pitampura
23.4
0.1
0.1
C
O
J10
Pitampura
23.9
0
0
C
O
J11
Pitampura
23.8
0
0.1
C
O
J12
Pitampura
23.9
0
0
C
O
M01
Vasant Vihar
0.9
7.6
7.4
C
N
M02
Vasant Vihar
0
0.3
3.3
I
N
M03
Shivalik
0
0.3
3.3
I
N
M04
Shivalik
2
0.4
4.2
I
N
M05
Lado Sarai
0.4
0.4
2
I
O
M06
Lado Sarai
0.3
0.4
0.8
I
O
M07
Lado Sarai
0.7
0.3
2.1
I
O
of the private water companies in New Delhi, which had recently been upgrading
service connections.
A total of four new house connections were included in the
trial. The private companies selected which houses would be included in the study
based on the request that houses be selected where: booster pumps were used, system
pressure was positive during supply hours, and where new house connections had been
recently made. An additional three houses were included in the author’s plumber’s
neighborhood.
97
5.2 Hydraulic notation and units
In the field of municipal water systems, one often accounts for pressure, velocity, and
frictional losses in units of
head,
and
𝑔
𝐻,
is defined as
pressure head, velocity head,
𝐻 =
𝑝
where
πœŒπ‘”
𝑝
is the pressure,
and
𝜌
head losses.
Pressure
is the density of water,
is the gravitational acceleration. Similarly, the velocity head is
𝑉2
, where
2πœŒπ‘”
𝑉
is the velocity [47, p.8]. In this thesis, unless otherwise specified, pressure will refer
to pressure head and will be measured in units of meters of water.
Additionally,
the pressures specified herein will always be given in gauge pressure where 0m is
atmospheric pressure.
One commonly used empirical equation for determining frictional losses
𝐻loss
in
meters, is the Hazen-Williams equation [47, p.14]:
𝐻loss =
where
𝑄
10.67𝐿𝑃 𝑄1.852
1.852 4.87
πΆπ»π‘Š
𝐷
3
is the volumetric flow rate [π‘š /𝑠],
length of pipe [m], and
πΆπ»π‘Š
𝐷
is the pipe diameter [m],
(5.1)
𝐿𝑃
is the
is a tabulated empirical constant for a given pipe material
and age. Another is the Darcy-Weisbach equation [47, p.10]:
𝐻loss = 𝑓
where
𝑓
8𝐿𝑄2
πœ‹ 2 𝐷5 𝑔
(5.2)
is the friction factor, independent of Reynolds number for wholly turbulent
flow.
5.3 Quantifying the intrusion risk
Contaminant intrusion describes the phenomenon of a contaminating substance entering into the potable water supply. A model to quantify the relative risk of contaminant
intrusion was required so that the risks associated with different pipe-pressure histories could be compared. Such a method could, for example, determine which scenario
had the higher risk of contaminant intrusion: four hours where a water system was
98
unpressurized and its pipes were empty, or 15 minutes where the same system was
filled with water that had a pressure of -3m. This section develops such a method.
Contaminant intrusion in a water distribution system can only occur when three
requirements are met: when there is a pathway for contamination, when contaminants
are present, and when there is a pressure gradient that allows contaminants into
the pipe.
A perfect contaminant intrusion model would consider all three of these
requirements and their interdependence in a given distribution system.
In continuously operated systems, negative pressure occurs only in extreme circumstances or during transient events. Modelers of such systems, therefore, typically
focus on carefully quantifying the location, duration, and magnitude of these infrequent occurrences of negative pressure. They then approximate the size of the pathway for contamination by the physical leakage rates, and finally, they conservatively
assume that contaminants are omnipresent [48, 29].
In intermittently operated systems, an alternative modeling approach was proposed by Vairavamoorthy et al. [29]. They argued that in intermittent water systems,
because of the long stretches with no pressure, intrusion is not pressure limited, but
limited by the presence of contaminants and contamination pathways. As such, their
contaminant intrusion model focused on mapping and predicting the presence and
overlap of contaminants and contamination pathways.
Applying their method to a city in India which had only one hour of water supply
per day, they created a 3D spacial model for the relative location of every water
pipe in the system with respect to every possible contamination source, as in Figure
5-9, reproduced from [29]. In areas where water pipes were close to contamination
sources, Vairavamoorthy et al. then used coupled models for pipe condition to assess
the likelihood of a contamination pathway and models for contaminant flow through
the soil around the contamination sources. Together these coupled models assessed
the likelihood of contaminant intrusion.
The advantage of Vairavamoorthy et al.’s method was that it identified the highest
risk areas in the existing system that should be addressed. The recommended course
of action was therefore very accessible to the partner water utility as it focused on
99
Figure 5-9: The spacial method used by Vairavamoorthy et al. to model where along
a distribution system contaminants would be present, reproduced from [29]
incremental improvements such as replacing certain sections of pipe and upgrading
certain sections of sewer.
Unfortunately, the model was extremely information in-
tensive. It relied not only on 3D GIS information about the location and depth of
water pipes, sewers, and standing bodies of water, but it also required estimates of
the material and age of these water pipes. Further, the model neglected the intrusion
potential of saturated soil during or after rains, found to have measurably reduce
water quality in a different Indian city [4].
Vairavamoorthy et al.’s model, while theoretically interesting, could not be applied to the data considered later in this thesis for the same reason it has not been
used by most Indian utilities: it requires information that is not readily available
to utilities and which would be very expensive to gather. Therefore, in this thesis,
the more conventional approach to modeling intrusion, which assumes omnipresent
contaminants, was used.
5.3.1 Water leakage and its pressure dependency
Historically, leaks of water out of pipes were modeled as a free jet leaving a sharp
edged orifice with the equation [49, 50]:
𝑉 =
where
𝑉
√οΈ€
was the exit velocity of the jet and
2𝑔𝐻
𝐻
(5.3)
the pipe pressure. To convert this to a
volumetric leakage rate, the orifice area was assumed to be constant and the leakage
100
flow rate,
𝑄𝐿 ,
was found to be [9, p.101]:
𝑄𝐿 = 𝐴eq
where
𝐴eq
√οΈ€
2𝑔𝐻
(5.4)
was an equivalent area of leakage which summed all of the leaks in the
system. It was therefore common practice to assume that the leakage flow rate would
scale as
√
𝐻
and to lump the equivalent leak area under the empirical coefficient
𝐿𝐢
[48]:
𝑄 𝐿 = 𝐿𝐢 ×
where
𝐻𝑝
√οΈ€
𝐻𝑝
for
𝐻𝑝 > 0
(5.5)
was the pressure in the pipe.
The industry standard changed, however, after it was shown that the pipe’s pressure affected the size and shape of the leaking orifice; for example, a higher pressure
may act to widen cracks. To account for this effect, the American Water Works Association records that common practice is to use a coefficient of about one:
𝑄∝𝐻
[9].
5.3.2 Quantifying the potential for contaminant intrusion
This section uses the traditional method for modeling contaminant intrusion in continuously supplied water systems; intrusion is modeled as pressure and pathway dependent and contaminants are assumed to be omnipresent. To account for pathway
dependence, this section builds on the methodology of Ebacher et al. [48], which used
a pipe’s leakage rate as a proxy for the size of that pipe’s contamination pathway.
In the historical model for clean water leaking out of a pipe, the leakage coefficient,
𝐿𝐢 ,
was used to account for the size of the available pathway for leakage, as in
Equation 5.5. Contaminant intrusion can be modeled as a leak flowing in the opposite
direction; instead of a potable water jet exiting through a crack, a jet of sewage might
enter through that same crack. This section develops parallel sets of equations for
leakage rates and contaminant-intrusion rates. If pressure had no effect on the size
of the leakage area, then the leakage coefficient
101
𝐿𝐢
would be shared between these
parallel equations.
The similarities between leakage and intrusion equations, however, breakdown
when the pressure-dependent size of the leakage area is considered.
For example,
increasing the magnitude of positive pressure in a pipe will cause leaks to flow out of
the pipe faster and will increase the size of some of the cracks; similarly, increasing
the magnitude of negative pressure in a pipe will cause contaminants to flow into the
pipe faster, but it will not increase the size of any of the cracks and might act to close
them. Quantifying the difference between these two scenarios was beyond the scope
of this thesis and so the pressure dependent size of orifices was neglected, allowing the
use of the historical leakage equation (Equation 5.5).
This simplifying assumption
was also made by Ebacher et al. [48].
Taking
and
π»π‘œ
𝑄𝐼
as the volumetric flow rate of contaminants flowing into the pipe,
𝐻𝑝
as the respective pressures of water inside the pipe and of contaminants
outside the pipe, and using the same leakage coefficient
𝐿𝐢
as in Equation 5.5, we
find Equation 5.6:
⎧
⎨ 𝐿 × √︀𝐻 − 𝐻 : (𝐻 − 𝐻 ) > 0
𝐢
π‘œ
𝑝
π‘œ
𝑝
𝑄𝐼 =
⎩ 0
: (π»π‘œ − 𝐻𝑝 ) ≤ 0
(5.6)
The pressure of contaminants is typically less than, or equal to, their depth underground. In the field setup described in Section 5.1.3, the reference for the pipe’s
pressure,
𝐻𝑝 , was taken at or just above ground level.
In this pressure reference plane,
the pressure of contaminates at the underground pipe minus the depth of the pipe is
π»π‘œ .
Since
π»π‘œ
will not typically be above zero, it is taken to be zero and neglected
from the equation. This assumption overestimates the amount of contamination.
Adapting Ebacher et al.’s methodology to account for temporal and spacial variation in pressure, a daily averaged intrusion volume
of pipe of length
time,
𝑑,
𝐿𝑝 ,
and observation window
π‘₯,
integral of Equation 5.6:
and space,
102
𝑇
𝑄𝐼
(defined over a defined section
in days) was found by taking the
Where
∫︁ 𝑇 ∫︁ √︁
𝐿𝐢
𝑄𝐼 =
(−𝐻𝑝* (𝑑, π‘₯))𝑑π‘₯𝑑𝑑
𝑇 𝐿𝑝 0
⎧
⎨ 𝐻 (𝑑, π‘₯) : 𝐻 (𝑑, π‘₯) < 0
𝑝
𝑝
𝐻𝑝* (𝑑, π‘₯) =
⎩ 0
: 𝐻 (𝑑, π‘₯) ≥ 0
(5.7)
(5.8)
𝑝
Similarly, taking the time integral for the daily averaged outward leakage rate,
𝑄𝐿 ,
but assuming constant pressure along the length of the pipe, allows Equation 5.5 to
be rearranged to :
Where
∫︁
𝐿𝐢 𝑇 √︁ **
𝑄𝐿 =
(𝐻𝑝 (𝑑))𝑑𝑑
𝑇 0
⎧
⎨ 𝐻 (𝑑) : 𝐻 (𝑑) > 0
𝑝
𝑝
𝐻𝑝** (𝑑) =
⎩ 0
: 𝐻𝑝 (𝑑) ≤ 0
(5.9)
Flow-rate changes at a single house in isolation do not affect the pressure in the
distribution main because head loss scales as pipe diameter to approximately the fifth
power, as in Equations 5.1 and 5.2. When a booster pump turns on and increases the
flow rate in the service connection pipe, the pressure drops from the system pressure
in the distribution main,
booster pump
𝐻𝐡𝑃 (𝑑).
𝐻𝑆 (𝑑),
to the measured pressure just upstream from the
This leads to the pressure profile shown in Figure 5-10, where
the horizontal axis is the effective length of the pipe between the distribution main
and the suction inlet of the RBP. All local losses, e.g. elbows, have been converted
to their effective length of straight pipe. To proceed with the integration in Equation
5.7, the simplifying assumption was made that the leakage coefficient was evenly
distributed along the effective length of the service connection. Figure 5-10 defines
point
π‘Ž
as the x-coordinate of zero pressure and
103
𝐿𝑝
as the length of the pipe. Using
Figure 5-10: Pressure drop along the service connection, from the distribution main
to the suction inlet of the RBP
the definition of
𝐻𝑝* (𝑑, π‘₯)
in Equation 5.8, yields:
⎧
⎨ 𝐻 (𝑑, π‘₯) : π‘₯ > π‘Ž
𝑝
𝐻𝑝 * (𝑑, π‘₯) =
⎩ 0
:π‘₯≤π‘Ž
(5.10)
Combining Equations 5.7 and 5.10,
𝐿𝐢
𝑄𝐼 =
𝑇 𝐿𝑝
∫︁
0
𝑇
∫︁
𝐿𝑝
√︁
(−𝐻𝑝 (𝑑, π‘₯))𝑑π‘₯𝑑𝑑
π‘Ž
104
Linearizing
𝐻𝑝 (𝑑, π‘₯)
based on Figure 5-10
*
𝐻𝑝 (𝑑, π‘₯) = 𝐻𝑆 (𝑑) − 𝐿π‘₯𝑝 (𝐻𝑆 (𝑑) − 𝐻𝐡𝑃
(𝑑))
√︁
∫οΈ€
∫οΈ€
𝑇 𝐿
*
(𝑑))𝑑π‘₯𝑑𝑑
∴ 𝑄𝐼 = 𝑇𝐿𝐿𝐢𝑝 0 π‘Ž 𝑝 −(𝐻𝑆 (𝑑) − 𝐿π‘₯𝑝 (𝐻𝑆 (𝑑) − 𝐻𝐡𝑃
⎧
⎨ 𝐻 (𝑑) : 𝐻 (𝑑) < 0
𝐡𝑃
𝐡𝑃
*
Where 𝐻𝐡𝑃 (𝑑) =
⎩ 0
: 𝐻𝐡𝑃 (𝑑) ≥ 0
[οΈƒ
]︃𝐿𝑝
* (𝑑)
𝐻𝑆 (𝑑)−𝐻𝐡𝑃
3/2
∫οΈ€
2(
π‘₯−𝐻
(𝑑))
𝑆
𝑇
𝐿𝑝
𝑑𝑑
∴ 𝑄𝐼 = 𝑇𝐿𝐿𝐢𝑝 0
𝐻𝑆 (𝑑)−𝐻 * (𝑑)
𝐡𝑃
3(
)
𝐿𝑝
π‘Ž
But from Figure 5-10
π‘Ž=
⎧
⎨
𝐻𝑆 (𝑑)𝐿𝑝
* (𝑑)
𝐻𝑆 (𝑑)−𝐻𝐡𝑃
: 𝐻𝑆 (𝑑) ≤ 0
⎩ 0
2𝐿𝐢
∴ 𝑄𝐼 =
3𝑇
: 𝐻𝑆 (𝑑) > 0
∫︁
0
𝑇
⎧
βŽͺ
βŽͺ
βŽͺ
⎨
* (𝑑))3/2
(−𝐻𝐡𝑃
* (𝑑)
𝐻𝑆 (𝑑)−𝐻𝐡𝑃
*
(−𝐻𝐡𝑃 (𝑑))3/2 −(−𝐻𝑆 (𝑑))3/2
* (𝑑)
𝐻𝑆 (𝑑)−𝐻𝐡𝑃
βŽͺ
βŽͺ
βŽͺ
⎩ √οΈ€−𝐻 * (𝑑)
𝐡𝑃
⎫
βŽͺ
: 𝐻𝑆 (𝑑) > 0&𝐻𝑆 (𝑑) ΜΈ= 𝐻𝐡𝑃 (𝑑) βŽͺ
βŽͺ
⎬
: 𝐻𝑆 (𝑑) ≤ 0&𝐻𝑆 (𝑑) ΜΈ= 𝐻𝐡𝑃 (𝑑) 𝑑𝑑
βŽͺ
βŽͺ
βŽͺ
: 𝐻 (𝑑) ≤ 0&𝐻 (𝑑) = 𝐻 (𝑑) ⎭
𝑆
𝑆
𝐡𝑃
(5.11)
To better visualize the connection between the pressure history and the intrusion
and leakage rates, the un-normalized forms of Equations 5.9 and 5.11 were used in
Section 6.2. These un-normalized forms were defined as the relative total volume of
intrusion and leakage,
𝑉𝐼′ (𝑑)
and
𝑉𝐿′ (𝑑),
𝑉𝐿′ (𝑑)
𝑉𝐿 (𝑑)
=
=
𝐿𝐢
as calculated in Equations 5.12 and 5.13.
∫︁ 𝑑 √︁
(𝐻𝑝** (𝑑))𝑑𝑑
⎧0
⎨ 𝐻 (𝑑) : 𝐻 (𝑑) > 0
𝑝
𝑝
**
𝐻𝑝 (𝑑) =
⎩ 0
: 𝐻 (𝑑) ≤ 0
𝑝
105
(5.12)
⎧
* (𝑑))3/2
βŽͺ (−𝐻𝐡𝑃
: 𝐻𝑆 (𝑑) > 0&𝐻𝑆 (𝑑) ΜΈ= 𝐻𝐡𝑃 (𝑑)
* (𝑑)
βŽͺ
𝐻
(𝑑)−𝐻
∫︁ 𝑑 βŽͺ
⎨ 𝑆 * 𝐡𝑃
𝑉
(𝑑)
2
𝐼
(−𝐻𝐡𝑃 (𝑑))3/2 −(−𝐻𝑆 (𝑑))3/2
𝑉𝐼′ (𝑑) =
=
: 𝐻𝑆 (𝑑) ≤ 0&𝐻𝑆 (𝑑) ΜΈ= 𝐻𝐡𝑃 (𝑑)
* (𝑑)
βŽͺ
𝐿𝐢
3 0 βŽͺ √οΈ€ 𝐻𝑆 (𝑑)−𝐻𝐡𝑃
βŽͺ
⎩ −𝐻 * (𝑑)
: 𝐻𝑆 (𝑑) ≤ 0&𝐻𝑆 (𝑑) = 𝐻𝐡𝑃 (𝑑)
𝐡𝑃
⎫
βŽͺ
βŽͺ
βŽͺ
⎬
𝑑𝑑
βŽͺ
βŽͺ
βŽͺ
⎭
(5.13)
5.3.3 Defining the contaminant intrusion ratio
The Contaminant Intrusion Ratio (CIR), is here defined as the ratio of the volume
𝑉𝐼 ,
of potentially intruded contaminants,
to the volume of clean water leakage,
Over a common time interval, the ratio of volumes
of averaged intrusion and leakage rates
𝑄𝐼
and
𝑄𝐿
𝑉𝐼
and
𝑉𝐿
𝑉𝐿 .
is the same as the ratio
defined in Equations 5.9 and 5.11.
The CIR is therefore defined as:
2
3𝑇 ′
CIR
=
𝑉𝐼
𝑄𝐼
=
=
𝑉𝐿
𝑄𝐿
⎧
* (𝑑))3/2
(−𝐻𝐡𝑃
βŽͺ
βŽͺ
: 𝐻𝑆 (𝑑) > 0
* (𝑑)
βŽͺ 𝐻𝑆 (𝑑)−𝐻𝐡𝑃
∫οΈ€ 𝑇 ′ ⎨ (−𝐻
* (𝑑))3/2 −(−𝐻 (𝑑))3/2
𝑆
𝐡𝑃
: 𝐻𝑆 (𝑑) ≤ 0
* (𝑑)
0 βŽͺ
𝐻𝑆 (𝑑)−𝐻𝐡𝑃
βŽͺ
√οΈ€
βŽͺ
⎩ −𝐻 * (𝑑)
: 𝐻𝑆 (𝑑) = 𝐻𝐡𝑃 (𝑑)
𝐡𝑃
√οΈ€
∫οΈ€
𝑇𝑇
1
𝐻𝑆** (𝑑)𝑑𝑑
0
𝑇
⎫
βŽͺ
βŽͺ
βŽͺ
⎬
𝑑𝑑
βŽͺ
βŽͺ
βŽͺ
⎭
𝑇
(5.14)
⎧
⎨ 0
: 𝐻𝐡𝑃 (𝑑) > 0
*
where 𝐻𝐡𝑃 (𝑑) =
⎩ 𝐻 (𝑑) : 𝐻 (𝑑) ≤ 0
𝐡𝑃
𝐡𝑃
⎧
⎨ 𝐻 (𝑑) : 𝐻 (𝑑) > 0
𝑆
𝑆
**
and where 𝐻𝑆 (𝑑) =
⎩ 0
: 𝐻 (𝑑) ≤ 0
𝑆
Care must be taken in the definitions of the averaging periods,
denominator and numerator of Equation 5.14.
𝑇𝑇
𝑇𝑇
and
𝑇 ′,
in the
is taken as the total duration of the
control and treatment tests at a particular connection so that the leakage rate will
be the average outward leakage rate of both the control and treatment observations.
Since the ultimate goal was to compare the CIR with and without the valve, recorded
pressures during field studies were substituted for
106
𝐻𝐡𝑃 (𝑑).
To calculate the baseline
CIR of a customer’s connection, the numerator of Equation 5.14 used control data
from that connection and therefore
𝑇′
was the duration of the control measurement.
To calculate the treatment CIR, treatment pressure data was used in the numerator
and
𝑇′
was the duration of the treatment study at that house.
The assumptions made in the derivation of Equation 5.14 are listed in Table 5.3;
their effects are summarized in Table 5.4.
5.3.4 Contaminant concentration at the consumer’s tap
The Contaminant Intrusion Ratio (CIR) is the ratio of the volume of potentially
intruded contaminants to the volume of water leakage on that same pipe over the same
time interval. Where the physical losses from leakage are known as a percentage of
clean water delivered,
can be found.
𝐿𝐿, the ratio of intruded contaminants to clean water delivered
This ratio will be referred to as the Intrusion Concentration (IC),
defined by Equation 5.15:
𝐼𝐢 = 𝐢𝐼𝑅 × (𝐿𝐿)along the service connection
(5.15)
This thesis focuses specifically on the contamination occurring along the service
connection pipe.
Therefore to calculate the IC, the leakage rate along this pipe
specifically must be estimated. The American Water Works Association’s manual on
water loss states that the majority of real losses (see Table 2.1 for the definition) occur
along household connections [9, p.101]. The Delhi Jal Board estimated real losses of
20-25% throughout their system [11]. Combining these two numbers resulted in an
estimated 10% real losses along the service connection.
As an example, consider a customer connection measured to have a CIR of 5%.
Assuming a 10% leakage level along this pipe, the final water delivered to the customer would have an IC =0.5%; their drinking water could be 0.5% rainwater, or even
sewage! While qualitatively, this number is much higher than desirable, the quantitative significance of this metric is still unclear. The following section considers how
107
Table 5.3: The assumptions behind Equation 5.14
Assumption
Assumption
Number
1
Contaminants are always present in the vicinity of the pipe
2
Area for leakage out of the pipe does not increase with pressure
3
Area for intrusion into the pipe does not decrease with larger
magnitude negative pressure
4
Areas for leakage and intrusion are the same
5
Pressure of contaminants is the same as their depth underground
6
Driving pressure for leaks is constant along the service connection
7
Pressure in the distribution main is not dependent on the service
pipe
connection flow rate
8
Leakage and intrusion rates are evenly distributed along the
service connection’s effective pipe length
Table 5.4: The effects of Equation 5.14’s assumptions
Assumption
Effect on calculated
Effect on calculated
leakage rate
intrusion rate
1
Greatly
overestimates
2
Underestimates
3
Overestimates
overestimates
Overestimates
Unknown
5
Usually
Usually
overestimates
overestimates
significantly
significantly
Overestimates
7
8
Greatly
Overestimates
4
6
Effect on CIR
Underestimates
Slightly
Slightly
underestimates
underestimates
Varies, but mean
Varies, but mean
Varies, but mean
effect is zero
effect is zero
effect is zero
108
this metric relates to human health risks.
5.3.5 Connecting contaminant intrusion to health risks
A more common metric for water contamination in distribution systems is the concentration of indicator organisms, either total coliforms or
E. coli, measured in colony
forming units (CFU) per 100mL. The two most likely sources of contamination are
sewage and rain or surface water. Assuming the potable water coming into the service
connection has zero CFU and further assuming perfect mixing such that the intrusion
is spaced out over the whole supply time yields the following expected water quality
at the customer’s connection:
𝐢𝐹 π‘ˆπ»π» = 𝐼𝐢 × πΆπΉ π‘ˆcontaminant
(5.16)
= 𝐿𝐿 × πΆπΌπ‘… × πΆπΉ π‘ˆcontaminant
The contamination concentration at the household,
𝐢𝐹 π‘ˆπ»π» ,
is therefore linearly
dependent on the contaminant concentration in the intruding volume,
𝐢𝐹 π‘ˆcontaminant .
No information was found regarding the contamination of rain water in soils;
instead, contamination levels in surface water was used as an estimator. Sargaonkar
and Deshpande classified excellent, acceptable, and slightly polluted surface water in
India as having a total coliform count of less than 50, 500, and 5000 CFU/100mL,
respectively [51].
Quantifying sewage contaminants, Miescier found contamination
levels flowing into nine municipal waste water facilities ranged from: total coliforms
106
to
108
CFU/100mL and
E.coli 105
to
108
CFU/100mL.
The World Health Organization’s guidelines for surveillance and control of community water supplies give an example risk categorization for water supplies based
on
E. coli
concentration in drinking water, shown in Table 5.5 [52, p.78]. Connecting
these risks to the calculated CIR values requires four assumptions, summarized in
Table 5.6. Notably, these assumptions conservatively overestimate the health risk associated with negative pressure because they represent a worst case scenario. These
109
Table 5.5: WHO’s example thresholds for health risks from
E. coli
contamination,
from [52, p.78]
Concentration of
E. coli
CFU/100mL
Risk Level
0
Safe: meets requirements
1-10
Low
10-100
Intermediate
100-1000
High
>1000
Very high
Table 5.6: Assumptions required to connect CIR values to health risks
Assumption
Health risk level
Improvement in
health risk
Water entering the service connection
Underestimates
Overestimates
Overestimates
Overestimates
Overestimates
Overestimates
The physical leakage level along the
Unknown
Unknown
service connection pipe is 10%
accuracy; linear
accuracy; linear
effect
effect
pipe is free of contamination
The entire length of the service
connection pipe is exposed to
contamination
The contaminant is sewage with E.
5
concentration of 10 CFU/100mL
coli
assumptions allow Table 5.5 to be rearranged to provide maximum allowable CIR
levels, as in Table 5.7. Notably, the risk levels increase with the order of magnitude
of contaminant concentration.
For intrusion reduction to have a significant health
impact therefore, it must reduce the CIR value by an order of magnitude.
5.4 Other numerical methods
5.4.1 Monte Carlo sensitivity study
A Monte Carlo simulation with N=1000 was conducted to study the sensitivity of the
results of Section 6 to variation in the baseline pressure history of the connections,
110
5
Table 5.7: Allowable IC and CIR levels for sewage (10 CFU/100mL) as the contaminant; assumptions are detailed in Table 5.6
Targeted Risk Level
Maximum IC
Maximum CIR @ 10% leakage
Safe: meets requirements
<0.001%
<0.01%
Low
0.001-0.01%
0.01-0.1%
Intermediate
0.01-0.1%
0.1-1%
High
0.1-1%
1-10%
Very high
>1%
>10%
sensor error, and incorrect assumptions about contaminant pressure.
A normally
distributed, zero-mean, mean shift was applied to each half of the crossover data.
I.e. for a specific house, the control and the treatment pressure histories each received
a different random error.
The variance in the simulated error was determined by the precision of the pressure
loggers measured in Appendix A and was taken to have an approximate standard
deviation of
𝜎=
0.45π‘š
. To summarize the range of possible values determined by the
3
simulation, x and y error bars that spanned, unless otherwise noted, the 1st and 99th
percentile were included with the data.
5.4.2 Determining upstream system pressure during booster
pump events
The method for quantifying the intrusion risk caused by negative pressure developed
in Section 5.3 requires that the upstream system pressure be known.
During field
trials, only one pressure measurement was taken and so an algorithm was developed
to estimate the upstream pressure.
It identified sudden pressure drops caused by
booster pumps and stored the pressure before the pump turned on as the upstream
system.
The upstream pressure was then assumed to be constant until a sudden
pressure rise was observed, signaling the turning off of the booster pump. A crude
threshold and delay algorithm was used; if the pressure changed by more than 0.75m
over two minutes, a pump start or a pump stop was identified. Further, the maximum
111
J09 Upstream Pressure Estimate
Upstream Estimate
Control
Upstream Estimate
Valve
M01 Upstream Pressure Estimate
12
4
10
3
8
Pressure (m)
Pressure (m)
2
6
1
0
4
−1
2
−2
0
01/18/14−09:30
09:40
Date
09:50
04/06/14−15:00 18:00
21:00
04/07−00:00
Date
Figure 5-11: Two examples of upstream pressure estimation; the algorithm worked
well for cases, like M01, where the pressure drop and rise caused by a booster pump
was abrupt, but worked less well for cases, like J09, where the action of a booster
pump was less clear
pump action was assumed to be 5 hours.
Two examples of the algorithm’s mixed accuracy are shown in Figure 5-11. Where
the upstream estimate prematurely declines, as in J09, the contamination risk at that
house is overestimated. In future work, finding a more robust method of identifying
booster pump cycles would make the contamination estimate more accurate.
112
Table 5.8: Measured supply times
House Identifier
First Supply Time
Second Supply Time
J08
5:50AM - 9:05AM
6:55PM - 9:05PM
M02
2:50AM - 11:25AM
4:10PM - 11:35PM
M03
3:45AM - 7:45AM
3:45PM - 7:05PM
M04
3:40AM - 7:40AM
3:50PM - 7:05PM
M05
2AM - 4:35AM
5:15AM - 6:40AM
M06
2:05AM - 4:35AM
5:15AM - 6:30AM
M07
2AM - 4:40AM
5:05AM - 6:35AM
5.4.3 Determining the supply hours of intermittently supplied
houses
To determine the valve’s impact on water quality during supply hours, the supply
hours for each intermittently supplied house were measured and are summarized in
Table 5.8. The measurement was conducted by overlaying the control and treatment
data from each house and visually selecting the two periods in the day where the
most pressure variation occurred. Three examples of this are shown in Figure 5-12.
113
Pressure History Overlaid J08
10
Pressure (m)
Control
Valve
5
0
−5
01/00/00−00:00 03:00
06:00
09:00
12:00
Time
15:00
18:00
21:00
01/01−00:00
Pressure History Overlaid M02
Pressure (m)
15
Control
Valve
10
5
X: 11:36 PM
Y: 0.2038
0
−5
01/00/00−00:00 03:00
06:00
09:00
12:00
Time
15:00
18:00
21:00
01/01−00:00
Pressure History Overlaid M07
20
Control
Valve
Pressure (m)
15
10
5
0
−5
01/00/00−00:00 03:00
Figure 5-12:
06:00
09:00
12:00
Time
15:00
18:00
21:00
01/01−00:00
Supply times were determined manually for each intermittently con-
nected house; results are summarized in Table 5.8
114
Chapter 6
Results
The key design objective of the valve designed in Chapter 4 was to limit the ability
of booster pumps to create negative pressure in the service connection pipe, thereby
decreasing contamination and increasing distribution equality. This chapter evaluates
the valve’s performance in two rounds of a crossover study conducted in New Delhi
in January and March of 2014. In this study, the pressure upstream from a particular
house was observed for at least 24 hours both with and without the valve at a sampling
frequency of 0.1Hz.
Because of the limited scope of this trial, it was not possible to evaluate the valve’s
impact on distribution equity. Instead, three metrics were used to measure the valve’s
ability to reduce negative pressure and contamination risk;
1. Negative pressure duration: the average number of minutes per day a customer
had negative pressure in their service connection pipe;
2. Contaminant Intrusion Ratio (CIR): the volume ratio of possibly intruded contaminants to leaked water in the customer’s service connection pipe (more formally defined in Section 5.3.3); and,
3. Health risk level: the World Health Organization’s example health risk level
associated with a given CIR value — the assumptions behind this conversion
are outlined in Section 5.3.5.
115
Before comprehensively evaluating the valve’s performance, this chapter first examines the valve’s impact on booster pump cycles in Section 6.1. Second, the performance metrics are introduced by means of a case study.
Third, the sources of
uncertainty are introduced with a second case study.
Next, in Sections 6.4 through 6.7, the valve’s effect on these metrics is shown
to vary significantly between different types of customer connections.
The valve’s
performance is evaluated using the four categories:
1. overall performance;
2. performance on continuously-supplied houses;
3. performance on intermittently-supplied houses, particularly while the system is
operating; and,
4. overall performance while the system is operating.
Finally, the chapter ends with a discussion about how the valve’s effect on distribution
equality could be tested in future work.
The results in this chapter demonstrate that:
βˆ™
the valve was most effective at limiting severe negative pressure, preventing 96%
of all pressure less than -1 meter;
βˆ™
the signal-to-noise ratio in the data gathered from intermittently-connected
houses was low;
βˆ™
the valve reduced the duration of negative pressure by an average of 53 minutes
per day, per connection while the system was operating;
βˆ™
the valve reduced the risk of contamination by a median of 80% while the system
was operating;
βˆ™
the negative pressure caused by booster pumps poses a significant health risk
while the system was operating; and
116
βˆ™
the valve not only took seven of the fourteen houses out of the
very high
high
safe
health risk levels, but also brought six of those seven into the
and
category.
6.1 Effect of the valve on a single pump cycle
The valve’s effect on customers’ residential booster pumps (RBPs) depended on the
water supply pressure. Figure 6-1 shows examples of single OFF-ON-OFF cycles for
RBPs installed on houses with high, medium, and low pressure. On most tested highpressure connections (above 6m of pressure), when a RBP was turned on, no negative
pressure was induced, as shown in the top panels of Figure 6-1. The valve was neither
supposed to nor did have any effect in this scenario. Conversely, on medium-pressure
connections (pressure between 2 & 6m), RBPs did induce negative pressure.
In
these medium-pressure cases, the valve effectively maintained non-negative pressure
upstream from the valve when the RBP was turned on, as shown in the middle panels
of Figure 6-1.
On the lowest-pressure connections, the upstream supply pressure was negative
during most of the time that the ‘system was pressurized.’ This meant that to extract
any water from the system, houses required an operating RBP. On these connections,
the upstream pressure was not positive, and therefore while the valve’s throttling
effect did reduce the magnitude of the negative pressure, it did not restore positive
upstream pressure, as seen in the bottom two panels of Figure 6-1.
Despite not
eliminating all of the negative pressure, the valve reduced much of the suction capacity
of the RBPs and therefore caused significant — often greater than 40% — reductions
in flow rates. Therefore finding consenting customers with low-pressure connections
who would allow a 24-hour study was understandably not possible. The rest of the
data presented herein was taken at medium- and high-pressure connections.
117
Figure 6-1: Effect of the valve on different types of house connections
118
6.2 Case study 1: applying the metrics to a house in
Old Delhi (J02)
To demonstrate the use and meaning of the three performance metrics, a house in Old
Delhi (J02) is used, whose pressure history is summarized in Figure 6-2. The top panel
shows the pressure history of the house throughout the January field study. Green is
the pressure history while the valve was installed; blue is the pressure history without
the valve. Each vertical line extending down from the pressure history represents a
booster pump cycle, such as those in the middle two panels of Figure 6-1.
6.2.1 Duration of negative pressure
To better demonstrate how the duration of negative pressure was calculated, a histogram of the pressure distribution with and without the valve is shown in the middle
panel of Figure 6-2; the bin width was 0.5m. Notably, 100% of the negative pressure
below -0.75m was eliminated by the valve. The bottom panel shows the difference
between the two histograms, control data minus valve data. The valve prevented 93
minutes per day of negative pressure at this connection.
Unfortunately, this simple histogram method obscured what happened within the
bins.
The actual metric used for performance evaluation was the average number
of minutes per day at any magnitude of negative pressure. Without the valve, the
house averaged 101 minutes per day; with the valve, 63 minutes per day.
Despite
the visually significant impact of the valve on the pressure history of J02, the valve’s
impact was only a moderate 38% reduction in the average daily duration of negative
pressure. While this improvement is significant in and of itself, it neglects to account
for the magnitude of negative pressure.
6.2.2 CIR
The contamination risk, as discussed at length in Section 5.3, is a function of the
pressure difference between contaminants outside the pipe and the pressure inside the
119
Pressure Profile
15
Control
Valve
Pressure (m)
10
5
0
−5
Avg. Min/Day at this Pressure
−10
01/22/14
01/23
01/24
01/25
01/26
Date
01/27
01/28
01/29
01/30
Pressure Histogram
300
Control
Valve
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure 6-2: Negative pressure reduction example from J02 in Old Delhi
120
pipe. The contaminant intrusion ratio (CIR) accounts for this pressure dependency.
Contamination also depends on the integrity of the pipe. The CIR accounts for this by
normalizing the calculated intrusion rate by the leakage rate, which is also a function
of the pipe’s integrity.
The CIR therefore calculates the volume ratio of potentially intruded contaminants to water leakage in the service connection pipe. Neither the absolute value of
intruded contaminants nor leakage could be calculated from the pressure alone without an assessment of the pipe’s condition. The pipe’s condition is accounted for by
an empirical constant,
𝐿𝐢 ,
assumed (details of this derivation are in Section 5.3.3) to
be the same for leakage and intrusion.
The connection between the pressure history, the relative volume of leaked water,
the relative volume of intruded contaminants, and the CIR, is shown in Figure 63.
The top panel again shows the pressure history of J02.
The middle two panels
show the relative accumulated volume of leakage and intrusion during each half of the
trial. These numbers are calculated using Equations 5.12 and 5.13, and have units of
π‘š3
. In these panels, it can be seen that while the leakage rates accumulate similarly
𝐿𝐢
between the valve and control, the contaminant intrusion volume accumulates much
more quickly without the valve.
Finally, the bottom panel graphs the instantaneous CIR during each half of the
trial. At the beginning of the control, the pressure was negative and therefore the CIR
started at 100%, but this was not shown so as to focus on the more representative
CIR range.
valve.
The total CIR without the valve was 3.2%, but only 0.03% with the
This represents contamination reduction of two orders of magnitude.
The
instantaneous CIR was shown in this section, but elsewhere in this thesis the CIR
will always refer to the average CIR, which is to the ratio of daily-averaged intruded
contaminants to daily-averaged leaked water.
6.2.3 Health risk
Assuming the service connection pipe was exposed to sewage and had a leakage rate
of 10%, then using Table 5.7’s correlation of CIR values to health risk levels, house
121
Pressure [m]
Pressure History
10
0
−10
01/22/14
01/23
01/24
01/25
01/26
01/27
01/28
01/29
01/30
01/23
01/24
01/25
01/26
01/27
01/28
01/29
01/30
01/23
01/24
01/25
01/26
01/27
01/28
01/29
01/30
01/23
01/24
01/25
01/26
Date
01/27
01/28
01/29
01/30
15000
10000
5000
3
Contaminates Intruded [ LmC ]
3
Water Leaked [ m
LC ]
Control
Valve
20
0
01/22/14
600
400
200
0
01/22/14
CIR
0.04
0.02
0
01/22/14
Figure 6-3: Contamination risk associated with the pressure history of J02
122
J02 had a health risk level of
high
without the valve and
low
with the valve.
6.3 Sources of uncertainty in the data
While the performance metrics applied to J02 in Old Delhi clearly showed the valve’s
impact, the results from other connections showed more variability.
Two major
sources of error are hypothesized: variability in the baseline pressure history of the
house and sensor accuracy. These sources of error are more prominent in the data
gathered from intermittently-supplied houses. This section briefly discusses baseline
variability and then focuses on sensor accuracy with a case study.
6.3.1 Variability in baseline characteristics of customer connections
The crossover design of the field study sought to isolate the impact of the valve on customer connections by comparing the performance metrics with and without the valve.
The crossover design relied on the impact of the treatment (the installation of a valve)
being larger than the random variation in the control data. For the intermittentlysupplied houses in this sample, the baseline variation was, unfortunately, much larger
than the treatment impact. This is discussed more in Section 6.6.
6.3.2 Case study 2: the effects of variability on a house in
Lado Sarai (M05)
The pressure loggers, used to gather the presented results, measured pressure with
a standard deviation of 0.15 meters, as measured in Appendix A.
To consider the
effects of this variability, a connection from Lado Sarai, M05, is taken as an extreme
example of noisy data. The measured data showed that the house experienced more
than 10 hours per day of additional negative pressure with the valve installed. Since
there is no physical mechanism by which the valve could have made the duration of
negative pressure 10 hours longer, the data from this connection serves as an excellent
123
example of what types of uncertainty were in the data and how they could affect the
results presented in this chapter.
Figure 6-4 shows the pressure profile of M05 over time in the top panel.
This
house received one hour of high pressure water once per day, spent more than 17.5
hours a day between -0.25 and 0.25 meters of pressure, and had two, two-hour periods
per day, where pressure rose to around 0.5 meters.
Despite the middle panel of Figure 6-4 showing that the valve made the duration
of negative pressure 10 hours longer in the range of -0.35 to -0.05 meters, the bottom
panel highlights that the valve made the duration of negative pressure below -1.5
meters shorter by eight minutes per day. This reduction in severe negative pressure,
when compared to the baseline negative pressure of approximately four hours per
day, results in the low signal-to-noise ratio of eight minutes of signal to four hours of
noise.
The low signal-to-noise ratio arose because the valve was never intended to prevent
all negative pressure, only negative pressure induced by booster pumps. Across nine
days of observations at M05, the booster pump’s pressure signature was observed
only six times, and only once trying to draw water outside of the high-pressure time
period. Instead of just a low signal-to-noise ratio however, the data suggests that the
presence of the valve made the duration of negative pressure longer.
Overlaying the pressure history of each day and considering in detail the observed
pressures around zero, Figure 6-5 shows a mean shift between control and treatment.
Since it is unlikely that valve would have caused an increase in the duration of negative
pressure, two alternative causes are hypothesized: drift in the pressure logger, which
would have been within 2 standard deviations (±0.3m) of the observed variance in the
pressure loggers detailed in Appendix A; or, the system pressure during the second
five days of observation was higher than during the first five days. Since the pressure
uniformly shifted even during times when the system was turned off, it is likely that
this difference was caused by drift in the pressure logger.
Whatever the reason for the effect, connections such as M05 were very sensitive
to minor changes in either the actual or the measured pressure while the system was
124
Pressure Profile
20
Control
Valve
Pressure (m)
15
10
5
0
Avg. Mins/Day at this Pressure
Avg. Mins/Day at this Pressure
−5
03/30/14 03/31
04/01
04/02
04/03
04/04 04/05 04/06 04/07
Date
Pressure Histogram Detailed at Zero Pressure
04/08
04/09
Control
Valve
400
300
200
100
0
−0.5
−0.4
−0.3
−0.2
−0.1
0
0.1
0.2
Pressure Bin (m)
Severe Negative Pressure Histogram
0.3
0.4
0.5
Control
Valve
1.2
1
0.8
0.6
0.4
0.2
−4.5
−4
−3.5
−3
−2.5
−2
Pressure Bin (m)
−1.5
−1
−0.5
0
Figure 6-4: How the valve caused a longer duration of negative pressure at M05
125
Control Pressure Profile
Control
Valve
1.5
Pressure (m)
1
0.5
0
−0.5
−1
01/00/00−00:00 03:00
06:00
09:00
12:00
Date
15:00
18:00
21:00
Figure 6-5: Evidence of a mean shift in the sensor at M05
turned off. This sensitivity is not a mathematical quirk, but arises out of the physics of
the system; systems that spend many hours per day unpressurized can easily become
contaminated if the contaminants are slightly pressurized, or if the system becomes
negatively pressurized. This is a severe risk associated with intermittently-operated
systems that the valve was never intended to correct; the valve can only reduce the
negative pressure caused by RBPs. The root cause of this sensitivity is therefore the
large portion of each day during which the pressure is very close to zero.
To account for this sensitivity in the rest of the results presented in this chapter, a
Monte Carlo simulation was conducted with N=1000. A normally distributed, zeromean, mean shift was applied to each half of the crossover data, with a standard
deviation of
𝜎 = 0.15m.
I.e. for a specific house, the control and the treatment
pressure histories each received a different random error.
More details about the
simulation method are included in Section 5.4.1. The results presented in the rest of
this chapter are the mean values of this Monte Carlo simulation; error bars spanning
the 1st and 99th percentile of the simulation are shown in red and error bars spanning
the 5th and 95th percentile are shown in blue.
126
6.4 Total aggregate impact of the valve
The valve’s function was to reduce (or eliminate) the magnitude and duration of
upstream negative pressure. Seventeen of the 19 observed houses averaged more than
one minute per day of pressure less than zero.
As was in the case study of M05,
several of the intermittently-connected houses had long durations of pressure close
to atmospheric pressure, which were very sensitive to small shifts in the data.
To
highlight the smaller, but important changes in severe negative pressure, this section
presents the duration of negative pressure below several thresholds. The valve most
consistently eliminated negative pressure below -1 meter.
6.4.1 Duration of negative pressure
In the top half of Figure 6-6, each data point represents one house; its x coordinate
is the length of time that the house averaged below -2m without the valve, and the
y coordinate is the length of time that the house spent below -2m with the valve
installed.
The neighborhood location and connection characteristics of each data
point are tabulated by label in Table 5.2. For confidentiality reasons, the addresses
of these houses were not published, but more information is available upon request
of the author.
The valve eliminated 100% of observed upstream negative pressure less than -2m,
which occurred for an average (across all 19 houses) of 38 minutes per house, per day.
Similarly, in the bottom half of Figure 6-6, the valve prevented 96% of all negative
pressure less than -1m, which occurred for an average of 52 minutes per house, per
day.
The valve’s reduction of pressure between -1m and 0m was less clear, and the
impact on all negative pressure (below 0m), was statistically inconclusive. As shown
in Figure 6-7, eight of the 19 data points suggest we must accept the null hypothesis
(blue dashed line) that the valve did not reduce negative pressure. Data points with
magenta and black error bars are continuously-supplied houses, while the standard
blue and red error bars are intermittently-supplied houses.
127
Because of the stark
Duration with Valve
(Avg. Min/Day)
Sensitivity of Negative Pressure Duration <−2m
80
60
40
20
Duration with Valve
(Avg. Min/Day)
0
J08 M04
M05J05
J03 J01 J07 J06 J02 M03
0
50
M01
100
150
200
Duration without Valve (Avg. Min/Day)
Sensitivity of Negative Pressure Duration <−1m
250
100
50
M01
M07 J08
J06
M03
J09 J03J01 J05 J02 J07
0
0
50
100
150
200
250
300
350
Duration without Valve (Avg. Min/Day)
Figure 6-6: Severe negative pressure reduction: valve consistently reduced negative
pressure less than -2m and -1m
128
contrast between the continuously and intermittently supplied connections, they are
considered independently in Sections 6.5 and 6.6. While it must be concluded that
the overall impact of the valve on the total duration of all negative pressure was
inconclusive, the valve did reduce 96% of negative pressure less than -1 meter — an
average of 52 minutes per house per day.
6.4.2 CIR
Figure 6-8 shows the total impact the valve had on the CIR; the x-axis was used as the
control direction (the CIR without the valve), the y-axis was used as the treatment
direction (the CIR with the valve).
While the uncertainty was still very large, 13
of 19 data points had mean values that favored rejecting the null hypothesis. This
suggests that the valve did likely improve contamination, but the actual impact was
vividly different between continuously and intermittently-supplied connections; they
are therefore treated separately in Sections 6.5 and 6.6.
6.5 Aggregate impact of the valve on continuouslysupplied houses
As seen in the previous section, the performance of the valve differed significantly
from continuously to intermittently-supplied houses. The mean impact of the valve
on the 12 continuously-supplied houses was: a reduction in the duration of negative
pressure by 82 minutes per house, per day; a 74% reduction in CIR values; and, the
shifting of seven customers out of the
the
low
and
safe
high
and
very high
health risk categories into
categories. This section elaborates on the valve’s effect on each of
these three metrics.
The effect of the valve on continuously-supplied houses was clearer than intermittentlysupplied houses for two related reasons: first, the control measurements of negative
pressure more effectively identify the booster pump problem, as without booster
pumps, the supply pressure would never be negative; and, second, were the valve
129
Null Hypothesis
1200
M03
J08
1000
M06
M05
Duration with Valve
(Avg. Min/Day)
M04
M07
800
600
400
M02
200
M01
J01
0
0
Figure 6-7:
J05 J02
J06
J03
200
J07
400
600
800
1000
Duration without Valve (Avg. Min/Day)
1200
Inconclusive impact of duration of negative pressure; clear differences
between intermittent connections (magenta and black) and continuous connections
(blue and red); null hypothesis that the valve has no impact is the blue dashed line
130
3.5
Null Hypothesis
3
M03
CIR with Valve
2.5
2
M06
1.5
1
J08
M05
M04
0.5
M02
0
M07
M01
0
Figure 6-8:
0.5
1
1.5
2
2.5
CIR without Valve
3
3.5
4
4.5
Inconclusive impact on CIR; intermittent connections (magenta and
black) are dwarfed by the continuous connections (blue and red); null hypothesis
that the valve has no impact is blue dashed line
131
450
Null Hypothesis
400
Duration with Valve
(Avg. Min/Day)
350
300
250
200
M01
150
100
J02
J05
J06
J07
J03
J01
0 J09
0
100
200
300
400
Duration without Valve (Avg. Min/Day)
50
Figure 6-9:
500
Time spent at negative pressure on continuous connections; red error
bars span the 1st and 99th percentile of results from the Monte Carlo simulation;
blue error bars, the 5th and 95th
to work perfectly on these houses, the total intrusion risk could be brought to zero.
6.5.1 Reduction in the duration of negative pressure
The valve’s impact on the duration of negative pressure is shown in Figure 6-9. Nine
of twelve data points have mean values which favor rejecting the null hypothesis. All
three that support the null hypothesis are very close to zero; the valve increased the
duration of negative pressure for only a total of eight minutes between these three
houses. The mean values therefore clearly indicate that overall the valve significantly
reduced the duration of negative pressure.
To better understand the certainty of these results, two observations should be
made about the span of the error bars in Figure 6-9: first, the range in the x or control
direction is smaller than the y or treatment direction; and second, the range in the
y or treatment direction frequently spans up to the null hypothesis line. A narrow
132
range of error bars in the control directions indicates that most of the negative pressure
experienced at these connections was below -0.45 meters. Adding a random noise with
three standard deviations equal to 0.45m still did not shift most of these observed
negative pressures into the positive region. This allows the confident conclusion that
negative pressure was a problem at continuously-supplied houses.
If the valve had functioned perfectly, it would have throttled all negative pressure
to exactly zero. Adding noise to this expected pressure signal would result in half of
the data points being shifted back to negative pressure values, leaving the duration
of negative pressure unchanged. Therefore, it was expected that the y-error bars in
Figure 6-9 would stretch up from the x-axis to the null hypothesis line. Where this
was not the case, the data may have been skewed (in either direction) by a negative
pressure event not caused by booster pumps or the valve’s set point may have drifted
up to a pressure higher than +0.45m. Both these possibilities were seen in the data
and are discussed below.
The y-error bar of J07 is significantly shorter than expected. Figure 6-10 shows
J07’s pressure history in detail; during the first two days of the control measurements,
there was a significant drop in system pressure not related to booster pumps, and
therefore this data should be excluded. This type of event in the control data inflates
the perceived effect of the valve. Beginning the control data on January 22 instead
of January 20, 2014, resulted in Figure 6-11, which was closer to expectations. For
completeness, the pressure histories of all 19 connections are included in Appendix C.
Another non-booster pump event skewed the results of M01, whose pressure history is shown in Figure 6-12. Whereas it was typical for the system pressure to drop
to almost zero at midday, on April 3rd (04/03), it dropped to -1m. This non-booster
pump event created negative pressure that the valve could not prevent and deflated
its perceived effectiveness. This explains in part why the data point was elevated and
the y-bar did not reach down to the x-axis. However, another explanation is needed
for why the y-error bar did not reach the null hypothesis line.
On several occasions at M01, the valve throttled the booster pump to a set point
of between positive 0.5 and 1m. One such example is shown in Figure 6-13. While
133
Pressure Profile
20
Control
Valve
15
Pressure (m)
10
5
0
−5
−10
01/20/14
Figure 6-10:
01/21
01/22
01/23
01/24
Date
01/25
01/26
01/27
01/28
Detail of pressure history at J07; highlights the anomaly during the
control measurement
Null Hypothesis
300
Duration with Valve
(Avg. Min/Day)
250
200
150
100
M01
J02
50
J05
J03
J06
J09 J01 J07
0
0
100
200
300
400
Duration without Valve (Avg. Min/Day)
500
600
Figure 6-11: Time spent at negative pressure on continuous connections; adjusted for
J07
134
Pressure Profile
10
Control
Valve
8
Pressure (m)
6
4
2
0
−2
−4
03/31/14
04/01
04/02
04/03
04/04
04/05
04/06
04/07
04/08
04/09
Date
Figure 6-12: Detail of pressure history at M01; highlights the anomaly during the
valve measurement on midday 04/03
this higher set point made the reduced duration of negative pressure more robust
to the magnitude of error used in the Monte Carlo simulation, it also caused higher
reductions in flow rate.
This higher set-point was observed in several of the alpha
prototypes as was discussed in Section 4.3.6.
Summarizing the observed effect of the valve on continuously-supplied houses,
Table 6.1 quantifies the mean, median, maximum, and minimum impact of the valve
on the duration of negative pressure.
From the mean values, it can be confidently
concluded that the null hypothesis should be rejected and that the valve does indeed
reduce the duration of negative pressure.
6.5.2 Reduction in CIR
Figure 6-14 compares the CIR without the valve to the CIR with the valve at each
continuously-supplied connection and demonstrates that the valve caused a very significant reduction in the contamination risk across every continuously-supplied house.
Even at connections where the valve did not prevent all the time spent at negative
pressures, such as J02, J05 and J06, the valve forced the negative pressures close
enough to zero to significantly limit their associated contamination risk. The range
135
Pressure Profile
4
Control
Valve
3.5
3
Pressure (m)
2.5
2
1.5
1
0.5
0
−0.5
−1
04/01/14−14:00
15:00
16:00
17:00
Date
18:00
19:00
20:00
Figure 6-13: Extreme detail of pressure history at M01; highlights the valve throttled
to a higher pressure than expected during booster pump events
Table 6.1: Summary of negative pressure reduction on continuously supplied houses;
-∞ is treated as -100% for mean calculation
Duration of Negative
Mean
Median
Maximum
Minimum
(mins/day)
(mins/day)
(mins/day)
(mins/day)
109
64
481
0
Valve
27
13
145
0.1
Single house difference
82
44
351
-3.2
37%
70.5%
99.4%
-∞
Pressure; N=12
Control (No valve)
(𝐢𝑖
− 𝑉𝑖 )
Single house percent
𝑉
difference (1 − 𝑖 )
𝐢𝑖
136
CIR with Valve
M01
0.04
0.02
0
J06J05J02
0
0.05
Null Hypothesis
J07
0.1
0.15
0.2
CIR without Valve
0.25
0.3
Figure 6-14: Valve’s ability to change CIR values for continuously pressurized houses;
CIR is extremely robust to noise in this data set
Table 6.2: Summary of contamination reduction on continuously supplied houses; a
negative value means the CIR increased when the valve was installed
CIR Values; N=12
Control (No valve)
Valve
Single house difference
(𝐢𝐼𝑅control (𝑖)
Median
Maximum
Minimum
0.04
0.02
0.2
0
0.004
1 × 10−5
0.05
0
0.03
0.02
0.2
−5 × 10−6
74%
97%
100%
-68%
− 𝐢𝐼𝑅valve (𝑖))
Percentage Difference
(1
Mean
− 𝐢𝐼𝑅valve (𝑖)/𝐢𝐼𝑅control (𝑖))
of calculated CIR values for continuously supplied houses is summarized in Table 6.2.
Notably, the mean effect was a 74% reduction in contamination risk and the maximum
observed effect was a CIR reduction of difference of 0.2 at M01.
The error bars in Figure 6-14 demonstrate that the valve’s reduction in contamination was robust to noise. More generally, the CIR metric was always less sensitive
to noise than the duration of negative pressure metric because it took into account
the magnitude of negative pressure. For example, a small shift in the pressure signal
from +0.01m to -0.01m could significantly skew the duration of negative pressure
metric, but would have little effect on the CIR metric.
137
Table 6.3: Reduction in health risk at continuously supplied houses
Health Risk
Maximum CIR with 10%
Houses in
Houses
leakage & sewage exposure
without Valve
with Valve
<0.01%
3
8
0.01-0.1%
2
3
Intermediate
0.1-1%
0
0
High
1-10%
5
1
Very high
>10%
2
0
Level
Safe: meets
requirements
Low
6.5.3 Reduction in health risk
The worst-case assumptions and method detailed in Section 5.3.5 were used to determine the health risks associated with the CIR values found; the result was Table 6.3.
Six of seven connections with
safe
very high
and
high
risk levels improved to
levels. The one connection that remained in the
high
low
and
risk category even with
the valve was M01, whose non-booster pump negative pressure was already discussed
in Section 6.5.1. During the control, M01 had a CIR of 25% and with the valve it
had a CIR of 5%. Therefore, even at M01, the valve still reduced contamination by
almost an order of magnitude and the health risk by a risk category.
The valve’s
clear and significant impact on negative pressure, contamination, and health risks at
continuously supplied houses urgently impels the further development of this valve.
6.6 Aggregate impact of the valve on intermittently
supplied houses
As discussed in Section 2.1.3, one of the major risks associated with the intermittent
supply of water is contaminant intrusion during un-pressurized times. When the system is first turned on, the contaminants which have been percolating into the pipe
while the system was turned off are flushed out; during this flushing, the water has
measurably lower quality [4].
The tested valve cannot prevent this contamination
138
source; however, the metrics used to evaluate the valve’s performance do not differentiate between negative pressure during times when the system is turned off, and
booster-pump-induced negative pressure. Because of this, the intermittently-supplied
connections in this study had very low signal-to-noise ratios, which prevented any conclusion from being made about the valve’s overall impact on intermittently-supplied
houses. This signal-to-noise ratio is discussed more in Section 6.6.1. To improve the
signal-to-noise ratio, the pressure and contamination at these houses are considered
only when the system is turned on (during supply hours).
Even after isolating for supply hours, the signal-to-noise ratio was still low because
only three of seven houses had booster-pump-induced negative pressure lasting more
than ten minutes per day.
Accordingly, the valve’s reduction in the duration of
negative pressure during supply hours was statistically inconclusive, despite having
an average reduction of 22%.
of contaminant intrusion.
The valve did, however, conclusively reduce the risk
With the valve, the CIR was reduced by an average of
26%; unfortunately this improvement only correlated to an improvement in the risk
threshold of one household, which went from
high
to
intermediate
risk.
6.6.1 Intermittent connections require data filtering
Filtering the data after the trial was required to improve the signal-to-noise ratio. The
three reasons the intermittently-supplied houses in this trial had low signal-to-noise
ratios were:
1. particular to this trial, four of seven of the intermittently-supplied customers
used their booster pumps for less than 10 minutes per day; the total booster
pump usage is summarized in Table 6.4;
2. particular to this trial, the duration of the control and treatment observation
windows was not long enough to account for the baseline variability in negative
pressure and contamination risk; and,
3. all intermittently-supply customers had stretches of time where negative pressure and contamination risk were not from booster pumps, but from the system’s
139
Table 6.4: Booster pump induced pressure characteristics on intermittent houses
House
Duration of Negative Pressure
RBP’s Suction
Supply
(minutes / day)
Pressure (m)
Pressure (m)
≈ 20
J08
≈
-2.5
≈
3
M02
None
N/A
N/A
M03
≈ 100
≈ 25
≈5
≈2
≈3
≈
≈
≈
≈
≈
≈1
≈0
≈ 10
≈7
≈ 12
M04
M05
M06
M07
-3
-3
-4
-4
-4
depressurization.
The design of the crossover field study was based on the assumption that a given
house would have a similar pressure history from one day to another and therefore the
treatment effect could be found by considering the difference between the control and
treatment measurements. To verify this assumption, Figure 6-15 plots the duration of
negative pressure during the first half of the control measurements for intermittentlysupplied houses against the second half. The figure demonstrates that the baseline
variation was extremely high. To improve upon this, future field work must strive to
lengthen the treatment and control observation windows to average out this source
of noise.
To visualize the signal-to-noise ratio without isolating for supply hours, Figure 616 plots the expected result from a perfect valve as a red dot on top of the duration
of negative pressure as measured with and without the valve. The valve’s ability to
reduce the duration of negative pressure could not be verified. Even if the valve had
performed perfectly, its impact would not have been significant compared to other
sources of negative pressure. Using the CIR metric instead resulted in the similarly
inconclusive Figure 6-17. Together, these figures force the conclusion that the null
hypothesis must be accepted; the valve did not significantly affect contamination or
the duration of negative pressure on intermittently-supplied houses. The rest of this
section focuses on the valve’s ability to reduce negative pressure and contamination
140
Duration during the second half of the control
(Avg. Min/Day)
1200
M03
M06
1000
M07
800
M05
J08
600
M04
400
M02
200
0
−200
0
200
400
600
800 1000 1200
Duration during the first half of the control
(Avg. Min/Day)
1400
Figure 6-15: The duration of negative pressure in the first half of the control observation compared to the second half shows that the variability in the supply pressure
to intermittent houses was larger than the effect of the valve; consistent data would
sit on the null hypothesis line
during the time when the system is turned on —
during supply hours.
6.6.2 Duration of negative pressure during supply hours
By isolating the valve’s performance during supply hours, using the method described
in Section 5.4.3, the signal-to-noise ratio was slightly improved. Figure 6-18 shows
the valve’s impact on the duration of negative pressure; the theoretically expected
performance of a perfect valve (see Table 6.4) was superimposed as red dots. While
four of seven mean values suggest the null hypothesis must be accepted, only one of
the three connections (M04) which had more than ten minutes per day of negative
pressure suggests accepting the null hypothesis. Both J08 and M03 have mean values
that suggest that the null hypothesis must be rejected; no part of J08’s uncertainty
bars span the null hypothesis line. It is therefore likely that while the valve did not
have any overall impact on the duration of negative pressure during supply hours at
intermittently-connected houses, the valve did improve the negative pressure duration on connections that had at least ten minutes per day of booster-pump-induced
141
1200
M03
J08
Duration with Valve
(Avg. Min/Day)
1000
M05
M06
M04
800
M07
600
Null Hypothesis
Theoretically Expected Results
400
M02
200
0
0
200
400
600
800
1000
Duration without Valve (Avg. Min/Day)
1200
Figure 6-16: Valve’s impact on negative pressure reduction on intermittently-supplied
houses is inconclusive because of very low signal-to-noise ratios; a perfect valve’s
expected result is overlaid in red
142
5
Null Hypothesis
Theoretically Expected Performance
4.5
4
CIR with Valve
(Avg. Min/Day)
3.5
3
2.5
M03
2
1.5
M06
J08
1
M04
M05
M07
0.5
M02
0
0
0.5
1
1.5
2
2.5
3
3.5
CIR without Valve (Avg. Min/Day)
4
4.5
5
Figure 6-17: Valve’s impact on CIR on intermittently-supplied houses is inconclusive
because of very low signal-to-noise ratios; a perfect valve’s expected result is overlaid
in red
143
negative pressure.
Figure 6-19a highlights why the valve was not able to reduce all of the time spent
at negative pressure at M04: after the characteristic sudden drop and then sudden
rise in pressure associated with a booster pump cycle, the pressure at the customer
connection continued to be negative. M04 used booster pump Configuration 2 where
the booster pump is placed upstream from the sump.
If the outlet into the sump
had been below the elevation of the distribution main, after the booster pump had
turned off, the water would have continued to be pulled into the sump. If this was
the case, the beta prototype used at that connection was either not sensitive enough,
or did not seal tightly enough to prevent this small magnitude (-0.5m to 0m) negative
pressure.
Figure 6-19b overlays the evening pressure recordings from M04. While the valve
did clearly prevent the severe negative pressure caused by the booster pump action,
variability in the starting time and duration of the positive pressure cycles are also
visible. For several days when the valve was installed, there was a delayed start to
supply hours, which was the most likely cause for the measured increase in negative
pressure duration. This variation in the baseline behavior of the system also affected
the CIR values.
The quantitative performance of the valve was equally inconclusive, as summarized
in Table 6.5.
The only positive metric was that the valve did reduce the negative
pressure duration by an average of 22%.
6.6.3 CIR during supply hours
The valve reduced the contamination risk during supply hours at intermittentlysupplied connections. Figure 6-20 shows the valve’s impact on the CIR; the expected
results from a perfect valve are superimposed as red dots.
Six out of seven mean
CIR values support rejecting the null hypothesis; therefore, the valve did reduce the
risk of contamination during supply hours. As in the case of the duration of negative
pressure, J08 and M03 showed significant benefits from the valve.
Table 6.6 quantifies that the presence of the valve did improve the CIR. The results
144
Null Hypothesis
Perfect Valve Performance
200
M03
M04
Duration with Valve
(Avg. Min/Day)
150
100
M02
M06
50
M07
M05
J08
0
50
100
150
Duration without Valve (Avg. Min/Day)
200
Figure 6-18: Negative pressure reduction during supply hours on intermittent houses;
red dots indicate theoretically ideal performance; red bars are the 99% CI, blue the
95% CI
145
2
Pressure (m)
1
0
−1
−2
−3
−4
04/03/14−16:00
16:30
17:00
17:30
18:00
18:30
Date
(a) Detail of negative pressure at M04; booster pump is followed by small magnitude negative
pressure of unknown origin
Pressure History Overlaid M04
1.5
1
0.5
Pressure (m)
0
−0.5
−1
−1.5
−2
−2.5
−3
−3.5
Control
Valve
01/00/00−16:00
16:30
17:00
17:30
Time
18:00
18:30
19:00
(b) For several days when the valve was installed, there was a delayed start to supply hours, which
increased the duration of the non-booster-pump negative pressure and skewed the results
Figure 6-19: Investigating the unexpected results form M04
146
0.9
Null Hypothesis
Theoretically Expected Performance
0.8
0.7
CIR with Valve
(Avg. Min/Day)
0.6
M03
0.5
0.4
0.3
M04
0.2
0.1 M05 M07
M06
M02
0
J08
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
CIR without Valve (Avg. Min/Day)
0.8
0.9
1
(a) All seven connections
Null Hypothesis
Theoretically Expected Performance
0.3
CIR with Valve
(Avg. Min/Day)
0.25
0.2
0.15
0.1
M07
M05
0.05
M06
M02
J08
0
0
0.05
0.1
0.15
0.2
0.25
CIR without Valve (Avg. Min/Day)
0.3
(b) Detail of lower five CIR values
Figure 6-20: CIR reduction during supply hours on intermittent houses; red dots are
expected values, red error bars are 99% range and blue error bars are the 95% range
147
Table 6.5:
Duration of negative pressure at intermittently-supplied houses during
supply hours; a negative difference implies the valve made negative pressure worse
Duration of Negative
Pressure; N=7
Mean
Median
Maximum
Minimum
(mins/day)
(mins/day)
(mins/day)
(mins/day)
Control (No valve)
81
50
212
31
Valve
77
48
174
10
4
-5
39
-38
22%
22%
80%
-47%
Single house difference
(𝐢𝑖
− 𝑉𝑖 )
Single house percent
𝑉
difference (1 − 𝑖 )
𝐢𝑖
Table 6.6: Summary of contamination reduction on intermittently supplied houses,
during supply hours; the negative difference in CIR means the presence of the valve
increased the CIR value
N=7
Mean
Median
Maximum
Minimum
Control (No valve)
0.19
0.08
0.78
0.02
Valve
0.13
0.07
0.50
0.01
Single house difference
0.05
0.01
0.28
−2 × 10−3
26%
12%
82%
-3%
(𝐢𝐼𝑅control (𝑖)
− 𝐢𝐼𝑅valve (𝑖))
Percentage Difference
(1
− 𝐢𝐼𝑅valve (𝑖)/𝐢𝐼𝑅control (𝑖))
are not as large as those from continuously-supplied houses, but they are nevertheless
significant. The mean improvement in CIR because of the valve was 0.05, representing
a 26% improvement in the CIR. In the best example of the valve’s benefit, at M03,
the absolute CIR was lowered by 0.28.
6.6.4 Health risk during supply hours
While the presence of the valve did improve the CIR, in order to correspond to decreases in the health risks, the CIR must decrease by an order of magnitude. Table 6.7
shows that only one of the seven houses had an improvement in their health risk category. While a project which would decrease the health risks of one in seven houses
might still be worth pursuing, it is the author’s strong hypothesis that the incon-
148
Table 6.7: Reduction in health risk at intermittently supplied houses, during supply
hours
Health Risk
Maximum CIR with 10% leakage
Houses in
Houses
& sewage exposure
Control
with Valve
<0.01%
0
0
0.01-0.1%
0
0
Intermediate
0.1-1%
0
1
High
1-10%
5
4
Very high
>10%
2
2
Level
Safe: meets
requirements
Low
clusive data in this section is misleadingly underwhelming. Because of the sampling
bias that houses needed to give informed consent for an experiment on their house,
and because houses could opt out at anytime, the intermittently supplied houses that
frequently used their booster pumps did not enter or stay in the trial long enough to
qualify for inclusion. Had such houses been part of the study, the fraction of negative
pressure problems caused by booster pumps would have been larger and, therefore,
the valve’s impact would have been more significant.
The top priority for future field work is to better characterize the pressure histories
of representative intermittently-supplied connections. If the connections summarized
in this section are representative of the typical intermittently-supplied connections,
the valve’s utility to such connections is dubious because a perfect valve would only be
able to solve a small fraction of the pressure and contamination problem. Conversely,
if the sampled intermittently-connected houses are not representative (as the author
has hypothesized), then the valve would still be able to address one significant source
of the contamination problem and would therefore be worth pursuing.
149
6.7 Valve’s aggregate impact whenever water was
being supplied
This section combines the results from Sections 6.6 and 6.5 to present the valve’s aggregate impact on all tested connections whenever water was being supplied; continuouslysupplied houses were examined in their entirety, while observations from intermittentlysupplied houses were cropped using the method described in Section 5.4.3.
6.7.1 Duration of negative pressure
The combined effect on the duration of negative pressure is shown in Figure 6-21. The
error bars span the 1st and 99th percentile, black represents continuously-supplied
connections, while magenta represents intermittently-supplied ones. Despite the inconclusive subset of intermittently-supplied houses, when taken as a whole, the data
suggests rejecting the null hypothesis.
Table 6.8 quantifies the valve’s average im-
pact on negative pressure; notably, the presence of the valve prevented an average of
53 minutes per day, per connection of negative pressure, with a maximum observed
benefit of almost six hours per day of negative pressure prevented at J07 and M01.
While the valve significantly improved negative pressure, it must be noted that
the valve’s impact on intermittently-supplied connections was much less conclusive
and requires immediate further study.
6.7.2 CIR
Figure 6-22 shows the valve’s combined impact on the CIR across all connections
while water was being supplied. While large uncertainty was still associated with the
intermittently-supplied connections, the overall results were clear: the valve significantly reduced the contamination risk. For a more detailed view of the continuously
supplied connections in Figure 6-22, the reader is directed to Figure 6-14. Table 6.9
quantifies the significant benefits of the valve: the contamination risk was reduced by
a mean of 43% and a median of 80%.
150
Null Hypothesis
350
Duration with Valve
(Avg. Min/Day)
300
250
200
M03
M04
150
M01
100
M02
M06M07 J02
M05
J06
J05
J08 J03
J09
J01
0
0
100
50
J07
200
300
400
Duration without Valve (Avg. Min/Day)
500
Figure 6-21: During supply times, the valve decreased the duration of negative pressure, but the intermittent connections are an inconclusive subset of this data; in pink
are intermittently-supplied connections, in black are continuously-supplied connections; the error bars span the 1st and 99th percentile of the Monte Carlo simulation
Table 6.8: Summary of negative pressure reduction during supply hours; a negative
difference means an increase in the pressure duration with the valve;
−∞
is taken as
-100% for the mean calculation
Negative Pressure
Mean
Median
Maximum
Minimum
Duration (N=19)
(mins/day)
(mins/day)
(mins/day)
(mins/day)
Control (No valve)
99
57
484
0
Valve
45
35
175
0.1
Single house difference
53
22
351
-38
23%
49%
99%
−∞
(𝐢𝑖
− 𝑉𝑖 )
Single house percent
𝑉
difference (1 − 𝑖 )
𝐢𝑖
151
1
Null Hypothesis
0.9
0.8
CIR with Valve
(Avg. Min/Day)
0.7
0.6
M03
0.5
0.4
0.3
M04
0.2
M05 M06
M07
0.1
0
Figure 6-22:
M02
J02
0
M01
J07
0.1
0.2
0.3
0.4
0.5
0.6
0.7
CIR without Valve (Avg. Min/Day)
0.8
0.9
During supply times, the valve reduced the CIR, but the intermit-
tent connections are a less conclusive subset of this data; in pink are intermittentlysupplied connections; in black are continuously-supplied connections; the error bars
span the 1st and 99th percentile of the Monte Carlo simulation; for a closer look at
the continuously supplied data points, see Figure 6-14
152
Table 6.9:
Summary of contamination reduction during supply hours; a negative
difference implies an increase in CIR with the valve;
−∞
is taken as -100% for the
mean calculation
CIR Values (N=19)
Mean
Median
Maximum
Minimum
Control (No valve)
0.10
0.02
0.78
0
Valve
0.06
5 × 10−4
0.50
0
Single house difference
0.04
0.02
0.28
−3 × 10−3
43%
80%
100%
−∞
(𝐢𝐼𝑅control (𝑖)
− 𝐢𝐼𝑅valve (𝑖))
Percentage Difference
(1
− 𝐢𝐼𝑅valve (𝑖)/𝐢𝐼𝑅control (𝑖))
Table 6.10: Reduction in health risk during supply hours
Health Risk
Maximum CIR with 10% leakage
Houses in
Houses
& sewage exposure
Control
with Valve
<0.01%
3
9
0.01-0.1%
2
2
Level
Safe: meets
requirements
Low
Intermediate
0.1-1%
0
1
High
1-10%
10
5
Very high
>10%
4
2
6.7.3 Health risk
To reduce the health risks associated with contamination, contamination has to be
reduced by an order of magnitude. Table 6.10 documents the number of houses in
each risk threshold, calculated according to the method in Section 5.3.5. Notably, the
valve removed seven of fourteen houses from the
rendered six of those seven in the
safe
very high
and
high
risk category and
category. Therefore, even amidst the baseline
uncertainty and the infrequent booster-pump usage in the intermittently-supplied
connections, the reduction in CIR and health risks clearly demonstrates the valve’s
utility and importance.
153
Table 6.11: Given an assumed booster pump suction capacity, this table summarizes
the minimum required supply pressure so that when the valve is installed, customer
flow rates are not reduced more than the amount in the column heading
Typical Booster Pump
10% Flow
25% Flow
40% Flow
Suction Capacity
Reduction
Reduction
Reduction
-0.5m
2.1m
0.6m
0.3m
-1m
4.3m
1.3m
0.6m
-2m
8.5m
2.6m
1.1m
-3m
12.8m
3.9m
1.7m
-4m
17.1m
5.1m
2.3m
-5m
21.3m
6.4m
2.8m
6.8 The necessity of a pilot study and recommendations for its scoping
The crossover study of individual house connections presented in this results section,
was not able to verify the valve’s aggregate impact on water and pressure distribution
equality.
To investigate this impact, a pilot study, in which one valve would be
installed at each connection in a small hydraulically-isolated area, is required.
Executing such a pilot study, or even a longer crossover study would only be feasible where consumer responses to the valve’s installation could be managed. Whenever
negative pressure is prevented, flow rates are reduced, and customers are sometimes
unhappy. If a water utility were able to determine what reduction in flow rate would
be acceptable to their customers, then given an assumption about the strength of
customers’ booster pumps, Table 6.11 summarizes the minimum supply pressure that
must be provided by the utility to ensure that flow rates are not limited more than
the acceptable amount. If for example, booster pumps are presumed to induce -2m
of negative pressure and customers will accept a 25% flow reduction, then the valve
should be installed only in areas with a minimum supply pressure of +2.6m.
The flow rate reductions could be further mitigated if the valves were installed in a
isolated neighborhood, where the system pressure could be independently controlled
and temporarily increased to compensate for the flow reduction caused by the valves.
154
In such a pilot study, the valve’s benefits would be most evident where the supply
pressure is below +6m of pressure for at least half of the supply time, thereby allowing
booster pumps to induce negative pressure. An ideal pilot study neighborhood would
therefore be one which:
1. is hydraulically isolated,
2. has good relationships between the residents and the water utility,
3. has a supply pressure below +6m for at least 50% of the supply hours,
4. has independently controllable pressure,
5. always has a minimum pressure of +2.5m during supply hours, and
6. is as small as possible.
155
THIS PAGE INTENTIONALLY LEFT BLANK
156
Chapter 7
Conclusions and future work
This thesis sought to improve water quality in intermittently-supplied water networks
by preventing residential booster pumps from inducing negative pressure in the water
supply pipe, which can cause contaminant intrusion. To effectively do so, a patentpending, self-actuating, full-bore, back-pressure regulating valve was designed and
tested in New Delhi, India. A crossover study was conducted to isolate and verify the
valve’s ability to reduce the magnitude and duration of negative pressure, as well as
decrease the associated water contamination risks at intermittently- and continuouslysupplied connections.
The valve proved most effective at continuously-supplied connections, preventing
an average of 82 minutes per day, per house of negative pressure.
The valve also
reduced the contamination risk at each connection by a median of 97%, enough to
cause significant improvements in the contamination-associated health risks.
The results from intermittently-supplied connections were much less clear because:
four of the seven intermittently-supplied connections in the study did not have instances of booster-pump-induced negative pressure longer than 10 minutes per day;
the booster-pump-induced risk at the remaining three connections was still only a
small fraction of the total risk; and high variability in the baseline pressure history of
these connections obscured the effects of the valve. Therefore these seven connections
had a low signal-to-noise ratio, and most conclusions needed to be qualified by ‘while
the system was pressurized.’
157
Combining the results from continuously- and intermittently-supplied connections,
this study showed that:
βˆ™
the valve was most effective at limiting severe negative pressure, preventing 96%
of all pressure less than -1 meter;
βˆ™
the valve reduced the duration of negative pressure by an average of 53 minutes
per day, per connection while the system was pressurized;
βˆ™
the valve reduced the risk of contamination by a median of 80% while the system
was pressurized;
βˆ™
the negative pressure caused by booster pumps posed a significant health risk
while the system was pressurized; and
βˆ™
the valve not only took seven of the fourteen houses out of the
very high
high
safe
health risk levels, but also brought six of those seven into the
and
category.
During the process of scoping, designing, and testing the valve, several opportunities to improve the design, to improve the test method, and for alternate projects
were encountered; they are summarized briefly in Appendix B. Three key future steps
are:
1. The valve was designed to prevent negative pressure by throttling to exactly
atmospheric pressure; this left no room for variability, drift, or error in this set
point.
Future iterations of the valve must compensate for this variability by
throttling to a higher set point.
2. The seven tested, intermittently-supplied connections had the majority of their
contamination risk arise from non-booster-pump events, even while the system
was pressurized. The author has hypothesized that this resulted from a sampling
bias, and that a representative sample would show booster-pump events playing
a larger role in contamination. This hypothesis must be verified immediately.
158
3. The aggregate benefits of the valve must be verified in a pilot study where
one valve would be connected to each house in a hydraulically-isolated neighborhood.
Such a study would gather the requisite proof that the valve not
only reduces contamination but also increases supply pressure and distribution
equity, justifying its application on a broader scale.
159
THIS PAGE INTENTIONALLY LEFT BLANK
160
Bibliography
[1] World Health Organization and UNICEF, “Data resources and estimates of the
WHO/UNICEF joint monitoring programme for water supply and sanitation,”
2010. [Online]. Available: http://www.wssinfo.org/data-estimates/table/
[2] D. McKenzie and I. Ray, “Urban water supply in india: status, reform options
and possible lessons,”
Water Policy,
vol. 11, no. 4, pp. 442–460, 2009. [Online].
Available: http://dx.doi.org/10.2166/wp.2009.056
[3] Asian Development Bank and India. Ministry of Urban Development,
Benchmarking and data book of water utilities in India.
Manila:
2007
Asian De-
velopment Bank, 2007.
[4] E.
an
vol.
Kumpel
and
intermittent
47,
no.
14,
K.
L.
and
Nelson,
“Comparing
continuous
pp.
microbial
water
quality
in
piped
water
supply,”
Water Research,
Sep.
2013.
[Online].
Available:
5176–5188,
http:
//dx.doi.org/10.1016/j.watres.2013.05.058
[5] P. S. Kelkar, S. P. Andey, S. K. Pathak, and K. G. Nimbalkar, “Evaluation of water distribution system for water consumption, flow pattern and pressure survey
during intermittant vis-à-vis continuous water supply in panaji city,”
Indian Water Works Association,
Journal of
vol. 34, pp. 27–36, 2002.
[6] K. Vairavamoorthy and K. Elango, “Guidelines for the design and control of
intermittent water distribution systems,”
Waterlines,
vol. 21, no. 1, pp. 19–21,
Jul. 2002. [Online]. Available: http://dx.doi.org/10.3362/0262-8104.2002.041
[7] E. Lee and K. Schwab, “Deficiencies in drinking water distribution systems
in developing countries,”
J water health,
vol. 3, pp. 109–127, 2005. [Online].
Available: http://www.iwaponline.com/jwh/003/jwh0030109.htm
[8] T. R. Lindley and S. G. Buchberger, “ Assessing intrusion susceptibility in
distribution systems,”
Journal of American Water Works Association,
vol. 94,
no. 6, p. 66, Jun. 2002. [Online]. Available: http://www.awwa.org/publications/
journal-awwa/abstract/articleid/14600.aspx
[9] American Water Works Association,
3rd ed., ser. AWWA manual.
Water audits and loss control programs,
Denver, Colorado: American Water Works Asso-
ciation, 2009, no. M36.
161
[10] World Health Organization and United Nation’s Children’s Fund Joint Water
Supply and Sanitation Monitoring Programme for Water Supply and Sanitation,
Progress on drinking water and sanitation: special focus on sanitation.
World
Health Organization and UNICEF, 2008.
[11] V. K. Babbar and Vikram Singh, “Towards rational capital water supply system,”
DJB Planning Meeting, 2011. [Online]. Available: http://www.urbanindia.nic.
in/programme/uwss/uiww/PPT_4th_Meeting/DJB_Water_PPT.pdf
[12] World Health Organization and United Nation’s Children’s Fund Joint Water
Supply and Sanitation Monitoring Programme for Water Supply and Sanitation,
Global water supply and sanitation assessment 2000 report.
World Health
Organization and UNICEF, 2000.
[13] H. W. King,
Handbook of Hydraulics for the Solution of Hydraulic Problems.
McGraw-Hill Book Company, Incorporated, 1918.
[14] India Central Public Health and Environmental Engineering Organisation,
ual on water supply and treatment.
Man-
Central Public Health and Environmental
Engineering Organisation, Ministry of Urban Development, 1999.
[15] M. Echenique and R. Seshagiri, “Attribute-based willingness to pay for improved
water services: a developing country application,”
B: Planning and Design,
Environment and Planning
vol. 36, no. 3, pp. 384–397, 2009. [Online]. Available:
http://dx.doi.org/10.1068/b33126
[16] E.
E.
Ameyaw,
F.
A.
Memon,
intermittent water supply systems,”
Technology—AQUA,
and
J.
Bicik,
“ Improving
equity
in
Journal of Water Supply: Research and
vol. 62, no. 8, p. 552, Dec. 2013. [Online]. Available:
http://www.iwaponline.com/jws/062/jws0620552.htm
[17] P. F. Boulos, B. W. Karney, D. J. Wood, and S. Lingireddy, “Hydraulic transient
guidelines,”
Journal of American Water Works Association,
pp. 111–124,
2005. [Online]. Available:
vol. 97, no. 5,
http://www.awwa.org/publications/
journal-awwa/abstract/articleid/15239.aspx
[18] D. H. Axworthy and B. W. Karney, “Discussion:
undulating elevation profiles,”
Filling of pipelines with
Journal of Hydraulic Engineering,
no. 12, pp. 1170–1174, 1997. [Online]. Available:
vol. 123,
http://dx.doi.org/10.1061/
(ASCE)0733-9429(1996)122:10(534)
[19] Bryan Karney, “Personal correspondence,” Mar. 2013.
[20] M.-H. Zérah,
Water, unreliable supply in Delhi.
Manohar Publishers, 2000.
[21] A. C. McIntosh, Asian Development Bank, and International Water Association,
Asian water supplies: reaching the urban poor: a guide and sourcebook on urban
162
water supplies in Asia for governments, utilities, consultants, development agencies, and nongovernment organizations. London: Asian Development Bank,
2003.
[22] B. E. Evans,
pay
studies
“Willing to pay but unwilling to charge:
make
a
difference?”
UNDP,
World
Do willingness to
Bank,
and
Water
and
Sanitation Program — South Asia, Tech. Rep., 1999. [Online]. Available: https:
//www.wsp.org/sites/wsp.org/files/publications/94200711902_sawilling.pdf
[23] S. K. Pattanayak, J.-C. Yang, D. Whittington, and K. C. Bal Kumar, “ Coping
with unreliable public water supplies: averting expenditures by households in
Kathmandu, Nepal,”
Water Resources Research,
vol. 41, no. 2, Feb. 2005.
[Online]. Available: http://dx.doi.org/10.1029/2003WR002443
[24] M.
A.
Altaf,
“ Household
a large secondary city:
Habitat International,
demand
for
improved
water
and
sanitation
findings from a study in Gujranwala,
vol. 18,
no. 1,
pp. 45–55,
in
Pakistan,”
1994. [Online]. Available:
http://dx.doi.org/10.1016/0197-3975(94)90038-8
[25] P.
Klingel,
“Von
intermittierender
zu
kontinuierlicher
wasserverteilung
in
entwicklungsländern (from intermittent to continuous water distribution in
developing countries),” Ph.D. dissertation, Institute for Water and River Basin
Management, Karlsruhe Institute of Technology, Karlsruhe, Germany, 2010.
[Online]. Available: http://digbib.ubka.uni-karlsruhe.de/volltexte/1000019357
[26] Legistlative Assembly of the National Capital Territory of Delhi, “ The Delhi
Jal Board Act 1998,” Apr. 1998. [Online]. Available: http://ielrc.org/content/
e9802.pdf
[27] K. Vairavamoorthy and M. Mansoor, “ Demand management in developing countries,” in
Water Demand Management,
D. Butler and F. A. Memon, Eds.
Lon-
don: IWA Publishing, 2006.
[28] K. Vairavamoorthy, J. Yan, H. M. Galgale, and S. D. Gorantiwar, “ IRA-WDS:
a GIS-based risk analysis tool for water distribution systems,”
Modelling & Software,
Environmental
vol. 22, no. 7, pp. 951–965, 2007. [Online]. Available:
http://dx.doi.org/10.1016/j.envsoft.2006.05.027
[29] K. Vairavamoorthy, J. Yan, and S. D. Gorantiwar, “ Modelling the risk of
contaminant
Management,
intrusion
vol. 160,
in
water
no. 2,
mains,”
Proceedings of the ICE - Water
pp. 123–132,
Jan. 2007. [Online]. Available:
http://dx.doi.org/10.1680/wama.2007.160.2.123
[30] R. Franceys and A. Jalakam, “The karnataka urban water sector improvement
project; 24x7 water supply is achievable,” Water and Sanitation Program, The
World Bank, Tech. Rep., Sep. 2010. [Online]. Available: https://www.wsp.org/
sites/wsp.org/files/publications/WSP_Karnataka-water-supply.pdf
163
[31] US EPA,
[32] US
Cross-connection Control Manual.
EPA,
“Methods
devices
for
the
prevention
Cross-Connection Control Manual,
backsiphonage,” in
[Online].
and
Office of Water, 2003.
Available:
of
backflow
and
Feb. 2003, pp. 12–16.
http://water.epa.gov/infrastructure/drinkingwater/pws/
crossconnectioncontrol/upload/2003_04_09_crossconnection_chapter04.pdf
[33] P. L. Skousen,
Valve handbook.
New York: McGraw-Hill, 1998.
[34] Jordan Valve, “Back pressure regulators: Mark 50 self-operated back pressure
regulators.”
[Online].
Available:
http://www.jordanvalve.com/sites/default/
files/mk50.pdf
[35] Spirax
Sarco,
able:
“Types
of
safety
valve.”
[Online].
Avail-
http://www.spiraxsarco.com/resources/steam-engineering-tutorials/
safety-valves/types-of-safety-valve.asp
[36] W. A. Conrad, “Pressure-flow relationships in collapsible tubes,”
tions on Biomedical Engineering,
IEEE Transac-
vol. BME-16, no. 4, pp. 284–295, Oct. 1969.
[37] F. P. Knowlton and E. H. Starling, “The influence of variations in temperature
and blood-pressure on the performance of the isolated mammalian heart,”
Journal of Physiology,
The
vol. 44, no. 3, pp. 206–219, May 1912. [Online]. Available:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1512817/
[38] C.
and
D.
Bertram
flow-induced
Structures,
and
J.
Tscherry,
oscillations
vol. 22,
no. 8,
in
“The
collapsible
pp. 1029–1045,
onset
tubes,”
of
flow-rate
limitation
Journal of Fluids and
Nov. 2006. [Online]. Available:
http://dx.doi.org/10.1016/j.jfluidstructs.2006.07.005
[39] R. E. Grandage, “ Development of a variable pressure reducing valve to minimise
Papers presented at the International
Conference on Developments in Valves & Actuators for Fluid Control, Oxford,
England 10-12 September 1985, R. Pickford, Ed. Oxford: BHRA, Sep. 1985,
leakage in water distribution systems,” in
pp. 251–276.
[40] D. McCloy and H. R. Martin,
Control of fluid power: analysis and design.
New
York: E. Horwood, 1980.
[41] US EPA, “Commonly asked questions:
water
act
and
the
NSF
standard,”
Section 1417 of the safe drinking
2012.
[Online].
Available:
http:
//water.epa.gov/scitech/drinkingwater/dws/plumbing.cfm
[42] NSF International Standard and American National Standard,
system components - Health effects: NSF/ANSI 61 – 2007a.
gan: NSF International, 2007.
164
Drinking water
Ann Arbor, Michi-
[43] “Parker o-ring handbook (ORD 5700),” Parker Hannifin Corporation, Cleveland,
Tech.
Rep.,
2007.
[Online].
Available:
http://www.parker.com/literature/
ORD%205700%20Parker_O-Ring_Handbook.pdf
[44] E. Oberg,
F. D. Jones,
H. L. Horton,
P. B. Schubert,
and G. Garratt,
Machinery’s handbook: a reference book for the mechanical
engineer, draftsman, toolmaker and machinist, 29th ed., C. McCauley,
“O-rings,”
Ed.
in
New York:
Industrial Press, 2012, pp. 2587–2592. [Online]. Available:
http://cds.cern.ch/record/230198
[45] “National
fluid
power
association
(NFPA)
diagrams
-
62475KAC,”
2010.
[Online]. Available: http://www.mcmaster.com/#62475kac/=rzoxie
[46] Hydreka,
able:
“Data
logger
vista
+
datasheet.”
[Online].
Avail-
http://www.hydreka.fr/biblioweb/documents/UK/Data%20sheet/Data%
20Loggers/DOC_Vista+_0710UK.pdf
[47] B. E. Larock, R. W. Jeppson, and G. Z. Watters,
Hydraulics of pipeline systems.
Boca Raton, FL: CRC Press, 2000.
[48] G. Ebacher, M. C. Besner, B. Clément, and M. Prévost, “Sensitivity analysis
of some critical factors affecting simulated intrusion volumes during a low
pressure
Research,
transient
vol.
46,
event
no.
in
13,
a
full-scale
pp.
water
4017–4030,
distribution
Sep.
2012.
system,”
[Online].
Water
Available:
http://dx.doi.org/10.1016/j.watres.2012.05.006
[49] J. H. Lienhard (V) and J. H. Lienhard (IV), “Velocity coefficients for free jets
from sharp-edged orifices,”
Journal of Fluids Engineering,
vol. 106, no. 1, pp.
13–17, Mar. 1984. [Online]. Available: http://dx.doi.org/10.1115/1.3242391
[50] R. L. Panton,
Incompressible flow,
4th ed.
Hoboken, New Jersey: Wiley, 2013.
[51] A. Sargaonkar and V. Deshpande, “ Development of an overall index of pollution
for surface water based on a general classification scheme in Indian context,”
Environmental Monitoring and Assessment,
vol. 89, no. 1, pp. 43–67, 2003.
[Online]. Available: http://dx.doi.org/10.1023/A:1025886025137
[52] World Health Organization,
Guidelines for drinking-water quality,
Geneva: World Health Organization, 1997, vol. 3.
165
2nd ed.
THIS PAGE INTENTIONALLY LEFT BLANK
166
Appendix A
Measuring the precision and accuracy
of the pressure loggers
The accuracy of the pressure loggers was quoted at
for the range 0 to 102m.
±0.1%
Best Fit Straight Line
A calibration check of their ability to measure negative
pressures showed less accuracy and more variability in this range.
A.1 Measurement method
Using a set of standard gym weights suspended from a hydraulic cylinder a series of
known negative pressures were created.
These pressures were measured both with
10 Hydreka loggers and a reference calibrated pressure sensor (with an accuracy of
±0.25psi).
A box plot for each logger at each pressure point is shown in Figure A-1.
Measuring the total variation in the loggers gave the detailed variation data summarized in Table A.1. 99.5% of the measurements between 0 & -4m of pressure were
within
±0.45m
of the reference pressure. Measurements below -4m inflated the mag-
nitude of negative pressure. Since only two data points had pressure in this range:
J01 and J02, all measurements below -4m were either taken as -4m so as not to inflate
the contamination risk. To test the conclusions made in the results section for their
robustness, a MonteCarlo simulation was applied where the variation in the normal
error is
𝜎=
0.45π‘š
.
3
167
168
Pressure (m)
Pressure (m)
−4.1
−4
−3.9
−3.8
−3.7
−1.7
−1.6
−1.5
−1.4
−1.3
−1.2
−0.4
−0.2
0
0.2
0.4
0.6
1
1
1
2
2
2
3
3
3
4
4
4
5
6
7
Logger Unit
Piston + 15lbs
5
6
7
Logger Unit
Piston + 5lbs
5
6
7
Logger Unit
Atmospheric Pressure
8
8
8
9
9
9
10
10
10
Pressure (m)
Pressure (m)
Pressure (m)
shown as a dashed black line.
Pressure (m)
Figure A-1: Accuracy and precision or pressure loggers.
The reference pressure is
−6.2
−6
−5.8
−5.6
−5.4
−5.2
−5
−3.1
−3
−2.9
−2.8
−2.7
−2.6
−2.5
−0.8
−0.6
−0.4
−0.2
1
1
1
2
2
2
3
3
3
5
6
7
Logger Unit
5
6
7
Logger Unit
4
5
6
7
Logger Unit
Piston + 20lbs
4
Piston + 10lbs
4
Self−Weight of Piston
8
8
8
9
9
9
10
10
10
Table A.1: Total variation in the logging equipment
169
THIS PAGE INTENTIONALLY LEFT BLANK
170
Appendix B
Other ideas for future work
In the development and scoping of this project, several areas of supporting knowledge
were not explored as fully as they could have been. Future explorations into these
would allow for a stronger foundation and likely a more robust design. These include:
1. This thesis used primitive statistical methods to assess the variability, uncertainty, and confidence intervals associated with field-measured pressure and
contamination risk.
Adding more formal methods would improve the credi-
bility and accuracy of the presented conclusions. Future papers published on
this data will attempt to incorporate more formal methods of quantifying and
bounding the variability and uncertainty.
2. In moving beyond the Beta prototype, the theft risk will be minimized by a
design that shifts from metal to plastic and is as lightweight as possible.
A
more detailed accounting for the predicted loads will be required to achieve
such an optimization.
Over the course of scoping, executing, and analyzing this project, several divergent
paths for alternative projects were identified. These included:
1. Providing a two-tiered storage system at the household level would allow households to gather water during the full duration of supply time, but the lower
quality water initially provided by the water company would not be mixed in
171
with the higher quality water that is provided after the system has flushed the
contaminants that intruded while it was turned off.
2. As briefly discussed in the background section, the degradation of a distribution
system is accelerated by the severe pressure transients associated with filling
and draining pipe networks.
Quantifying this damage, and exploring options
for mitigating it could be an important contribution to the body of knowledge
surrounding how to incrementally improve intermittently supplied water.
172
Appendix C
Impact of the valve on each house
While the results presented in Chapter 6 focused on the aggregated performance of
the valve on all houses, for reference, this appendix documents the valve’s impact on
each house included in the Results chapter.
All nineteen houses that met the requirement that experimental data was gathered
for at least 24 hours with the valve and without it were included in Chapter 6 and are
documented herein. A single exception was a twentieth house that had played with
the bypass valve during the control period; because the pressure at this connection
was high, it could not be determined when and for how long the valve had been
activated. This connection was, therefore, neither included in Chapter 6 nor in this
appendix.
173
Pressure Profile
15
Control
Valve
Pressure (m)
10
5
0
−5
Avg. Min/Day at this Pressure
−10
01/19/14
01/20
01/21
01/22
01/23
Date
01/24
01/25
01/26
01/27
Pressure Histogram
250
Control
Valve
200
150
100
50
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
100
50
0
−50
−100
−10
−5
0
5
Pressure Bin (m)
10
Figure C-1: Pressure summary: Azad Market: J01
174
Pressure Profile
15
Control
Valve
Pressure (m)
10
5
0
−5
Avg. Min/Day at this Pressure
−10
01/22/14
01/23
01/24
01/25
01/26
Date
01/27
01/28
01/29
01/30
Pressure Histogram
300
Control
Valve
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure C-2: Pressure summary: Azad Market: J02
175
Pressure Profile
Pressure (m)
15
Control
Valve
10
5
0
−5
01/22/14
01/23
01/24
01/25
01/26
01/27
01/28
01/29
Avg. Min/Day at this Pressure
Date
Pressure Histogram
300
Control
Valve
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure C-3: Pressure summary: Azad Market: J03
176
Pressure Profile
Pressure (m)
15
Control
Valve
10
5
Avg. Min/Day at this Pressure
0
01/24/14
01/25
01/26
01/27
Date
01/28
01/29
01/30
Pressure Histogram
300
Control
Valve
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
50
0
−50
−100
−10
−5
0
5
Pressure Bin (m)
10
Figure C-4: Pressure summary: Azad Market: J04
177
Pressure Profile
6
Control
Valve
Pressure (m)
4
2
0
−2
Avg. Min/Day at this Pressure
−4
01/22/14
01/23
01/24
01/25
01/26
Date
01/27
01/28
01/29
01/30
Pressure Histogram
400
Control
Valve
300
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
100
50
0
−50
−100
−150
−10
−5
0
5
Pressure Bin (m)
10
Figure C-5: Pressure summary: Azad Market: J05
178
Pressure Profile
Pressure (m)
30
Control
Valve
20
10
0
−10
01/23/14
01/24
01/25
01/26
01/27
01/28
01/29
01/30
Avg. Min/Day at this Pressure
Date
Pressure Histogram
300
Control
Valve
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure C-6: Pressure summary: Pitampura: J06
179
Pressure Profile
Pressure (m)
20
Control
Valve
10
0
Avg. Min/Day at this Pressure
−10
01/20/14
01/21
01/22
01/23
01/24
Date
01/25
01/26
01/27
01/28
Pressure Histogram
500
Control
Valve
400
300
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure C-7: Pressure summary: Pitampura: J07
180
Pressure Profile
Pressure (m)
10
Control
Valve
5
0
Avg. Min/Day at this Pressure
−5
01/25/14
01/26
01/27
01/28
Date
Pressure Histogram
01/29
01/30
1500
Control
Valve
1000
500
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
1000
500
0
−500
−1000
−10
−5
0
5
Pressure Bin (m)
10
Figure C-8: Pressure summary: Vivek Vihar: J08
181
Pressure Profile
20
Control
Valve
Pressure (m)
15
10
5
0
Avg. Min/Day at this Pressure
−5
01/16/14 01/17
01/18
01/19
01/20
01/21 01/22
Date
01/23
01/24
01/25
01/26
01/27
Pressure Histogram
400
Control
Valve
300
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
100
50
0
−50
−100
−10
−5
0
5
Pressure Bin (m)
10
Figure C-9: Pressure summary: Pitampura: J09
182
Pressure Profile
20
Control
Valve
Pressure (m)
15
10
5
0
−5
01/16/14
01/17
01/18
01/19
01/20
01/21
01/22
01/23
Avg. Min/Day at this Pressure
Date
Pressure Histogram
300
Control
Valve
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
150
100
50
0
−50
−100
−10
−5
0
5
Pressure Bin (m)
10
Figure C-10: Pressure summary: Pitampura: J10
183
Pressure Profile
Pressure (m)
15
10
5
0
−5
01/16/14
Avg. Min/Day at this Pressure
Control
Valve
01/17
01/18
01/19
01/20
Date
01/21
01/22
01/23
01/24
Pressure Histogram
400
Control
Valve
300
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
300
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure C-11: Pressure summary: Pitampura: J11
184
Pressure Profile
20
Control
Valve
Pressure (m)
15
10
5
0
Avg. Min/Day at this Pressure
−5
01/16/14
01/17
01/18
01/19
01/20
Date
01/21
01/22
01/23
01/24
Pressure Histogram
400
Control
Valve
300
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
100
50
0
−50
−100
−10
−5
0
5
Pressure Bin (m)
10
Figure C-12: Pressure summary: Pitampura: J12
185
Pressure Profile
Pressure (m)
10
Control
Valve
5
0
Avg. Min/Day at this Pressure
−5
03/31/14
04/01
04/02
04/03
04/04
04/05
Date
04/06
04/07
04/08
04/09
Pressure Histogram
300
Control
Valve
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure C-13: Pressure summary: Vasant Vihar: M01
186
Pressure Profile
Pressure (m)
15
10
5
0
−5
03/31/14
Avg. Min/Day at this Pressure
Control
Valve
04/01
04/02
04/03
04/04
04/05
Date
Pressure Histogram
04/06
04/07
04/08
04/09
500
Control
Valve
400
300
200
100
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
100
50
0
−50
−10
−5
0
5
Pressure Bin (m)
10
Figure C-14: Pressure summary: Vasant Vihar: M02
187
Pressure Profile
Pressure (m)
10
Control
Valve
5
0
−5
04/02/14
04/03
04/04
04/05
04/06
04/07
04/08
04/09
Avg. Min/Day at this Pressure
Date
Pressure Histogram
800
Control
Valve
600
400
200
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
0
−200
−400
−10
−5
0
5
Pressure Bin (m)
10
Figure C-15: Pressure summary: Shivalik: M03
188
Pressure Profile
Pressure (m)
4
Control
Valve
2
0
−2
−4
04/02/14
04/03
04/04
04/05
04/06
04/07
04/08
04/09
Avg. Min/Day at this Pressure
Date
Pressure Histogram
800
Control
Valve
600
400
200
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
0
−200
−400
−10
−5
0
5
Pressure Bin (m)
10
Figure C-16: Pressure summary: Shivalik: M04
189
Pressure Profile
20
Control
Valve
Pressure (m)
15
10
5
0
Avg. Min/Day at this Pressure
−5
03/30/14 03/31
04/01
04/02
04/03
04/04
Date
04/05
04/06
04/07
04/08
04/09
Pressure Histogram
1500
Control
Valve
1000
500
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
200
100
0
−100
−200
−10
−5
0
5
Pressure Bin (m)
10
Figure C-17: Pressure summary: Lado Sarai: M05
190
Pressure Profile
20
Control
Valve
Pressure (m)
15
10
5
0
Avg. Min/Day at this Pressure
−5
03/30/14 03/31
04/01
04/02
04/03
04/04
Date
04/05
04/06
04/07
04/08
04/09
Pressure Histogram
1500
Control
Valve
1000
500
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
150
100
50
0
−50
−100
−10
−5
0
5
Pressure Bin (m)
10
Figure C-18: Pressure summary: Lado Sarai: M06
191
Pressure Profile
20
Control
Valve
Pressure (m)
15
10
5
0
Avg. Min/Day at this Pressure
−5
03/30/14 03/31
04/01
04/02
04/03
04/04
Date
04/05
04/06
04/07
04/08
04/09
Pressure Histogram
1500
Control
Valve
1000
500
0
−10
−5
0
5
10
Pressure Bin (m)
Avoided Pressure Histogram (Control−Valve)
15
20
15
20
Avg Min/Day Avoided
100
50
0
−50
−100
−10
−5
0
5
Pressure Bin (m)
10
Figure C-19: Pressure summary: Lado Sarai: M07
192
Download