HomeMy WebLinkAbout6 - Update on watermain break data analysis CITY OF BOULDER
WATER RESOURCES ADVISORY BOARD
AGENDA ITEM
MEETING DATE: February 23, 2009
AGENDA TITLE: Update on Water Pipe Replacement Analysis
PRESENTERS: Nick Worley, Temporary Civil Engineer
Bob Harberg, Utilities Planning and Project Management Coordinator
EXECUTIVE SUMMARY:
Staff has recently completed a water pipe replacement analysis that considers an
extensive GIS based inventory of the water distribution system pipes as well as break
records from the previous 20 years. The analysis is documented in the attached Technical
Memorandum regarding Water Main Replacement Analysis and CIP Budget Estimate.
STAFF RECOMMENDATION:
The technical memo provides recommendations for funding of the future water main
replacement program.
BOARD AND COMMISSION FEEDBACK: None to date
PUBLIC FEEDBACK: None to date
ANALYSIS:
Please refer the attached technical memo.
MATRIX OF OPTIONS:
Please refer the attached technical memo.
ATTACHMENTS:
Technical Memorandum regarding Water Main Replacement Analysis and CIP Budget
Estimate
AGENDA ITEM # 6 PAGE 1
City of Boulder Technical Memorandum
To: Bob Harberg
From: Nick Worley
CC: Utilities Project Management Staff
Date: 2-9-09
RE: Water main replacement analysis and CIP budget estimate
Introduction
The water distribution infrastructure in cities around the US has begun to come to the forefront
of municipality's focus as the piping begins to fail more frequently while costs of repair continue
to increase. Dependable water is such a critical part of society and is necessary for the
wellbeing of the public. The city of Boulder has over 400 miles of water distribution system
piping much of which was installed during the 1950's through the 1970'x. The majority of this
aging pipe is still functioning properly although some parts of the city are experiencing more
frequent breaks1.
The purposes of this technical memorandum are to:
1. evaluate and summarize available data,
2. create a model that predicts future water main break frequency,
3, evaluate new model, and
4. create a capital improvement program budget for water main rehabilitation and
replacement
-1- Pipe Replacement Technical Memo
Table of Contents
1 Literature Search 3
2 Water Distribution System Characteristics 5
2.1 Data Sources 5
2.2 Data Summary 5
3 Water Distribution System Analysis History .........................................................................14
4 Current Water Distribution System Analysis ........................................................................16
5 Model Results .......................................................................................................................16
5.1 Classification 1-one break allowed before replacement ............................................17
5.2 Classification 2-two breaks allowed before replacement ...........................:...............19
5.3 Classification 3-Three breaks allowed before replacement 20
5.4 Classification 4- Four breaks allowed before replacement 21
5.5 Summary of Classifications 22
6 Plan Budget ...........................................................................................................................23
7 Recommendation ..................................................................................................................25
-2- Pipe Replacement Technical Memo
1 Literature Search
Over the past few decades more and more research has been conducted on the
deterioration of our water distribution system infrastructure. The need for replacement has
caused companies and researchers to begin to develop computer models that can help with the
decision making process of when and which pipes need to be replaced. These models can be
used to either help with the decision making process of how much money should be allocated
for future pipe rehabilitation or to assist in the selection process of which pipes to replace. The
purpose of this memo is to summarize an innovative method that can be used to help with the
budgetary planning process of future pipe replacement.
The methods that have been used by others in the pipe rehabilitation selection process can
be categorized into the followingz:
• Deterioration point assignment methods
• Break even analysis
• Failure probability and regression methods
• Mechanistic methods
The first method is the point system that the City of Boulder currently uses for the
prioritization process. This method determines which pipes have the highest priority for
replacement depending on their characteristics as well as environment conditions. The method
gives a certain quantity of points for each characteristic of a segment of pipe. Categories that
are typically used are pipe age, pipe material, pipe diameter, soil conditions, water pressure,
ambient temperature, etc.
The break even analysis is a cost based method that compares repair costs and replacement
~ costs. This method must have a predictive model that can generate a quantity of future
breaks/repairs for comparison with the replacement cost. Pipe replacement is necessary when
the cost for pipe repair reaches an unacceptable level in comparison with the present worth of
the pipe.
The failure probability and regression method is the method that uses statistical functions to
predict future failures and is the method discussed in this technical memo. This method uses
past break data along with the distribution system pipe data to determine a failure curve. The
failure curve can be used to find the cost of pipe replacement necessary to keep up with pipe
failures. The disadvantage of this method is that it only provides a quantity of pipes that will fail
for any given time period and does not predict which ones. A separate analysis must be used in
conjunction with this method to assist in the selection of pipes to be replaced.
Mechanistic methods are those that gather data and determine failure rate by observing
physical specimens that have been subjected to a similar set of conditions. An example of this
experiment would be burying a number of sticks of pipe and monitoring the corrosion over time.
A failure rate can be then determined based on how the specimens fail. These methods use
parameters such as measured corrosion pit depth, soil type, age of pipe, temperature induced
stresses, frost load, and other failure processes using underlying physical principlesz.
-3- Pipe Replacement Technical Memo
Other local municipalities are beginning to analyze their distribution systems and take action
to replace deteriorating pipe. The city of Westminster in Colorado is one of the nearby cities
that have begun to take action. In 2005 they contracted URS corporation to do a
comprehensive study on their system and produce some recommendations for future action3.
Much of the metallic pipe in their system is corroding and failing at a faster rate and they have
had to become more aggressive in pipe replacement. URS used the city of Westminster GIS data
as well as their pipe break records to create a new system improvement plan. The final
recommendations made to the city were to increase pipe replacement as the rate of failure is
expected to increase over the next few years.
Denver Water also is beginning to analyze their water distribution system more intensely.
They have had an aggressive rehabilitation plan in action for years that has lined many of their
corroding cast iron pipes with cement mortar. Recently they have begun to work with some
professors at the University of Texas to create a predictive break model and a new prioritization
process that will take into account more criteria than their current point system prioritization
process.
Corrosion and tuberculation has been the dominant problem in the ®enver water
distribution system. They have begun to test soil resistance at every pipe segment that breaks
and document it as one of the criteria in the point system analysis. In conjunction with soil
resistance, they have begun to incorporate their GIS data into the MWH Infowater modeling
program in order to produce pressure and flow output for use in the point system pipe selection
method.ls
From the literature search, it is apparent that the prediction of future pipe failures is a best
estimate only due to hack of data. Record keeping in city public works has not been sufficient
over the past 50 years to accurately determine future break patterns. There are many unknown
pier_es of data that must be accounted Fr~r by best guess assumptions which brings in
uncertainty.
Many of the documented pipe break studies have incorporated into the final
recommendation several scenarios of how the system wi11 potentially fail as time goes on, The
final recommendation comes down to a choice of which scenario is the most likely the one that
represents the system most nearly. As more data becomes available as time goes on, the
scenarios can be reevaluated in comparison with the system and the budgetary decisions could
change.
Although there is uncertainty in all the documented analysis, the consensus among
municipalities, researchers, and consultants is that more money must be allocated for pipe
replacement in the near future. Otherwise the clean, dependable water we now enjoy will be
compromised in the future.
-4- Pipe Replacement Technical Memo
2 Water Distribution System Characteristics
The City of Boulder water distribution system is comprised of many different sizes and
materials of pipe totaling over 400 miles in length. The majority of the system was installed
during the 1950's, 1960's, and the 1970's along with the sanitary sewer and storm sewer
systems. Metal pipes such as Cast Iron Pipe (CIP) and Ductile Iron Pipe (DIP) were the primary
materials of choice during installation and many pipes have since failed due to corrosion, joint
failure, pressure, and settlement. Other materials such as ACP and PVC were also used starting
in the 1970's and 1980's respectively.
The scope of this memo is to evaluate the distribution pipes 16 inch in diameter and
smaller. Pipes in the system larger than 16 inch will be considered transmission lines and will be
excluded from the analysis. The reason for this narrowed scope is because 99% of the breaks
occur on pipes 16 inr_h and smaller. The transmission main system accounts for 8% of the total
length of installed pipe and has a very low break rate in comparison. These pipes may be
considered for rehabilitation or replacement as well but due to the more reliable nature, they
have been excluded from the analysis.
2.1 Data Sources
The data used in the analysis summarized in this memo was taken from the City of Boulder
_ GIS information system database and the utilities maintenance pipe break database. The GIS
database is a graphical summary of all the eity's records of water main as-builts and provides a
complete plan view of the underground water main infrastructure. It contains information
.y about the pipe characteristics and date of installation as well as whether or not the pipe is
private or owned by the city. For the purpose of this Technical Memo, only the pipes owned by
the city will be evaluated. The utilities maintenance pipe break database was initiated in the
early 1990's and contains all the information about the individual pipe breaks that have
occurred since that time.
Both databases have been significantly filtered for completeness and correctness providing
a strong basis for the accuracy of the analysis.
2.~ Data Summary
The raw data used in the analysis will be briefly summarized in this section for a complete
picture of the distribution system characteristics. This summary of data only includes the pipes
16 inch diameter and less as defined by the scope of the TM.
As shown in the following tables, the decade with the most installed pipe was in the 1970's.
The piping previous to the 1970's was almost entirely CIP which has been the pipe material with
the most breaks. The construction industry began to use DIP in the 1970's and PVC in the
1980's. -
-5- Pipe Replacement Technical Memo
Table 1 Summary of pipe installed"
1950's 1960's 1970's 1980's 1990's 2000's Total
Total Pipes Installed 1245 1412 2205 545 886 580 6873
Total Length of Pipes (mile) 85 86 124 35 52 29 412
Table 2 Pipe material distributions
Pipe Length % of.
_ Material (mile) Total
ACP 1% 4%
Cl P 162 40%
DIP 124 30%
PCCP 2 0.4°l0
PVC 96 23%
ST 8 .2%
Table 3 Pipe diameter distributions
Pipe Length % of
_ Diameter (mile) Total-
4 13 3%
6 98 24%
8 2f?3 50%
12 65 16%,
14 4 1
16 26 6%
-6- Pipe Replacement Technical Memo
Water Pressure Distribution
60
50
t
40 - ~
i ~ i
s
rn 30 '
J 20 _ ~ _~}_t _
m
0 I. -
^o ~O ~O ,~O ~O ^,~O ^r~j0 ~4j0 ^.~O ^~O
Pressure (psi)
Figure 1 Water pressure distribution6
The above tables and figure were taken from the GIS database. The following tables and
figures were compiled from the maintenance pipe break database.
Total l3reaks per Year
140
120 -
100 -
Y ~0 -
f0
~ - - -
m 60
40 - - - - -
20 - - -
0
0~ 0~ X01 .~O 'l0 1°~ ~~L ~5 °~R 01 ~o ~'3 00
Year
Figure 2 Total breaks per yeara
Figure 2 shows the total breaks per year that have been recorded. It is important to note
that most pipes will break multiple times during their life before being replaced. As mentioned
before, the utilities maintenance database was initiated in 1990 and all the pipe breaks previous
were found in annual reports. The data showing breaks per year from 1950 until 1960 and 1962
to 1981 is currently unknown.
-7- Pipe Replacement Technical Memo
Pipes with First Break
so
40 ,'41
N 'I f ~
x
~ 30 - - - fi - - ~ n
'
m -
D L. P - `'r W 6~
~.a 1y~\ 9.~ 9: ~~1' g4Q' ic"~h 94J °~*g ,SyS~ ~b1~.dS~,dyS~ .dyf~ ,y0l, .~~PP•JO~
Figure ~ Pipes with first break
Figure 3 shows the number of pipes that have their first break each year. Many of the
annual breaks occur on pipes as repeat breaks. No trend can be seen in either the total annual
breaks nor the annual first time breaks.
Over the years city staff have discussed the recorded data and come up with different ideas
as to what data is representative of the system and what data is not. Some staff have noted
during the years that a significant number of breaks have occurred during a problem "event"
when the distribution system was stressed by external conditions. An example of this is when
maintenance crews open and close valve incorrectly causing a spike in pressure. !n 2U02, work
was being done oi~ the Gunbarrel tank when valves were incorrectly operated. A spike in
pressure immediately caused 13 breaks in the surrounding distribution system. These breaks
could be looked at in two different ways:
1. The pipe that failed was the weakest point in the system and would be the first to
eventually break
2. These pipes broke prematurely due to a mistake by the maintenance crew and
therefore do not represent true pipe failures
These arguments must be kept in mind when viewing the data. The years with particularly
large numbers of breaks usually had a number of events that caused an abnormal number of
pipe failures. In the following analysis all pipe breaks have been used to model the system as it
functions with the seldom operational mistakes that cause event failures. The following figures
summarize the entire break data as recorded.
Table 4 Recorded pipe break totals6
Total % of
Break type Breaks Total
Corrosion 292 29%
Pressure 116 12%
Settlement 534 53%
Other 59 6%
-8- Pipe Replacement Technical Memo
Breaks by Type
50
45 '
40
35 - ` -Corrosion
Y 30 - Pressure
~ 25 -Settlement
~ 20 -
15 f j -Other
10
1990 1995 2000 2005
Year ,
- - - rs,
Figure 4 Recorded pipe breaks6
As shown in the above figure, settlement breaks have dominated in total annual breaks for
the past 18 years. A possible theory for this is the use of native soil for pipe bedding causing
pipe movement during soil expansion and contraction. Other breaks include joint failure, water
system events causing failure, or other abnormal reasons.
Weighted Breaks by Material
0.60
~ ~ 0.50
~ 0.40 ' - ~ -ACP'
L
n, 0.30 CtP
~ y - DIP
~ 0.20 -PVC
m
pp 0.10
0.00 ~ - - - - `1
1990 1995 2000 2005
Year
Figure 5 Pipe breaks classified by material6
-9- Pipe Replacement Technical Memo
Weighted Breaks by Diameter
0.7
0.6
d 0.5 4 inch
~ 0 4 6 inch
e. - 8 inch
0.3 12 inch
m 0.2 -16 inch
F _
0.1 - - _
_
0
1990 1995 2000 2005
L - Year
Figure 5 Pipe breaks classified by diameter6
Figure 5 above shows a general relationship that CIP is typically the pipe that has the most
annual breaks where PVC being the newest pipe has the least annual breaks. An explanation for
the spike in ACP breaks/mile in 2002 was the above mentioned Gunbarrel tank event that
caused many breaks around Gunbarrel. ACP was the main pipe material used in the water
distribution system in Gunbarrel in the 1970's. Because of the relatively high number of breaks
due to the event, the ACP shows a spike in breaks/mile.
Figure 6 shows a slight correlation in the pipe diameter where the srnaller pipes tend to
have more annual breaks,
Weighted Breaks vs Pressure
4.5
4
3.5 -
~ 3
2.5
2
N 1.5
m
1
a~
m 0.5 - - -
0- r----
0 50 100 150 200
Pressure (psi)
Figure 7 Relationship between water pressure and breaks6
Figure 7 shows little correlation between the water pressure and number of pipe breaks per
mile of pipe. Intuitively the higher pressure would create more stress on the pipe and more
frequent breaks could be expected. This however is not strongly supported by the data.
-10- Pipe Replacement Technical Memo
The pipe break records from the maintenance department have been integrated into a GIS
spatial mapping file so the different breaks can be viewed spatially. This is helpful to locate
areas where similar type breaks have happened repeatedly and to understand how pipe
replacement can most efficiently be used. The following three figures show the locations of the
recorded pipe breaks by type.
~ ~ C
f- - I
O
_
1,'~~~ I ~II
a
- J ~ 1J
N
WE
S
Legend
• Corrosion Breaks ~
Figure 8 Corrosion break locations'
-11- Pipe Replacement Technical Memo
L~
r-
. -
~ - L
' ~ II
~C
•
•
~ r
N-CJ..1~
WE
S
Legend '
• Pressure Breaks. 'm]
Figure 9 Pressure break locations8
-12- Pipe Replacement Technical Memo
L r
h
T - •
.,d.
i
f
t
r~
i
-T
ii
i lL
I ~i
l
s
Legend
• Settlement Breaks ~
Figure 10 Settlement break locations9
-13- Pipe Replacement Technical Memo
3 Water Distribution System Analysis History
The City of Boulder utilities department began a more intense analysis of the water
distribution system in 2002 when Integra consulting was hired to create a new system for pipe
replacement selection. This initial report has been evaluated and critiqued by staff as well as
other consultants in order to produce additional strategies and methods for pipe cost analysis
and selection for replacement.
The report that Integra produced summarized a point system that the City of Boulder staff
could use to prioritize which pipes are most in need of replacement. Each pipe segment would
receive points based on its age, material, size, soil type, and relative location in the city. The
points would be summed up for each pipe and the one with the most points has the highest
priority for replacement. This point system and methodology was integrated into the water
main and main breaks databases and has been used for the prioritization of pipe replacement
since 2002.
In 2007, the City of Boulder decided to take this analysis a step further and create a
prediction model using Integra's methodology. The model utilized a statistical function called
the Weibull function which was originally created to model the failure rate of a given
population. The water main pipe segments had previously been assigned point values by the
point system methodology and the points were used to symbolize failure. Segments with higher
point values were assumed to have more chance of failure and thus the data was used to create
a Weibull curve. A cut off point value was assumed which signified the point value above which
the pipe segments have failed. This cut off point value was chosen based on the estimated total
failure of the water main distribution system. The results of this analysis produced pipe
replacerrient costs ranging from $6 million to $10 million per year.
In April of 2008, the City of Boulder hired HDR to critique the original Integra analysis and
the Weibull function results. Their comments on the Integra analysis dealt mostly with the lack
of methodic determination of how many points each pipe characteristic should receive, They
also took the analysis a step further and suggested that the point system should be used on a
per foot basis in order to correctly evaluate break potential between long and short pipe
lengths. Little was mentioned concerning the city staff produced technical memo on the
Weibull function analysis.
HDR continued the analysis of the Boulder water distribution system by trying to find a
correlation between pipe breaks and other pipe characteristics. They evaluated several
scenarios by grouping characteristics together such as Breaks per Mile vs. Diameter, Breaks per
Mile vs. Age-Material groups, and Breaks per Mile vs. Age-Material-Sail groups. A correlation
was found when the Age-Material-Soil groups were plotted with the Breaks per Mile and they
decided to use this correlation to create a model to show which pipes should be replaced first.
The model they created grouped pipes into similar groups and the assumption was made that
they will all fail at a similar time relative to other groups. A cut off value was chosen which they
called a service level that would determine which "groups" of pipe should be replaced based on
the acceptable breaks per year. The model was used to evaluate three different service levels
based on the current level of service the city provides. The results of the model produced
necessary pipe replacement costs ranging from $4.8 million to $10.6 million per year.
-14- Pipe Replacement Technical Memo
The prediction model presented by HDR wasn't considered an accurate representation of
the distribution system because of the limited data available to use. City staff began to look at
additional methods of failure prediction and pipe replacement selection. The critical break area
method utilized the GIS spatial mapping to locate hot spot areas where multiple breaks have
occurred. An assumption was made that all pipe within the hot spot areas should be priority for
replacement within the next 20 years. The resulting cost of replacement would be $4.7 million
per year.
City staff also looked for other correlations in Breaks vs. operating pipe pressure. The city
was able to finalize and run the city water distribution system model in MWH's InfoWater
program in 2008 which provided additional data for analysis. Little correlation was found
between breaks and main pressure so the city continued to look at other prediction methods.
The pursuit of a more correct analysis continued with the use of the Weibull function again in an
analysis that used several scenarios to test the sensitivity of the Weibull function. The different
scenarios included and excluded different pipe breaks based on different conditions and a
prediction model and budgetary CIP was created for each. The analysis produced a wide range
of results that were not entirely useful according to the city's desires. The idea of using
different scenarios in this analysis is what eventually led to the analysis described in the
following section which wa~,chosen as a more system representative analysis.
I.
I
-15- Pipe Replacement Technical Memo
4 Current Water Distribution System Analysis
The above mentioned analyses produced relatively conservative values of annual pipe
replacement cost that are not possible for the city budget to meet. All the previous models
made the assumption that one break on a pipe segment should be considered failure. This
assumption has been questioned due to the fact that many of the pipe segments in the city that
have broken once haven't broken again since. Many of these breaks were due to distribution
system events or settlement in the surrounding soil and don't necessarily reflect a failing pipe.
The definition of what constitutes a failure has now been evaluated for a more realistic and less
conservative approach to the analysis.
The analysis of the water distribution system has continued onward by changing the
definition of pipe failure. In this new model, the Weibull function will be used to specifically
model the number of first, second, third, and fourth breaks that any given pipe will have. Four
separate Weibull functions have been created to model ear_h break classification. These Weibull
functions will be calibrated to actual data and used to predict the number of expected breaks in
the future. A decision will have to be made as to how many annual breaks will be considered
acceptable and from this one of the break classifications or prediction models will be chosen as
the desired model to follow. Each prediction model will have different financial implications as
less breaks allowed means more replacement casts.
The break classifications can be looked at as different paths that can be taken and a
predicted number of future breaks will be calculated for each path. 'T'he choice of the best path
will d2pend on how many breaks per year are considered acceptable. There will be a balance
that must be met between the amount of annual budget spent on break repairs and pipe
replacement. The number of breaks that any given pipe is allowed to have will determine how
long into the future the replacement is delayed.
5 Model Results
The results of the model have been broken down by each break "classification" which
correspond to one, two, three, or four breaks that are allowed on a pipe before replacement is
necessary. Each classification has a specific curve generated from the Weibull function and are
shown on a common graph below.
-16- Pipe Replacement Technical Memo
Pipe Failure vs Time
12o°i°
100% - _
~ 80%
-First Break
i ur i
N 60% J I Second Break
a
a ---Third Break
40% - -Fourth Break
20% -
0%
0 50 1D0 150 200 250 300 350
Pipe Age (years)
Figure 11 Pipe failure curveslo
5.1 Classification 1- one break allowed before replacement
Typically one break on a pipe is considered failure for most analyses. In this analysis "one
break equals failure" is considered the baseline model and most conservative. Many pipes that
break once will not break again for many years and the repair can be justified to extend the use
of the pipe until further breaks begin to occur.
In this analysis all recorded pipe breaks were used and an unrecorded failure factor was
used to extrapolate breaks that occurred before the recording system was started. The breaks
and total numbers of pipe installed were broken down by decade and the percentage of failed
pipes was calculated for use in the Weibull function. Table 5 below shows the data used for the
first classification.
Table 5 Data used for first break Weibull function°
Summary of Pipes and Failures
1950's 1960's 1970's 1980's 1990's 2000's Total
Total Pipes Installed 1325 1462 2242 548 886 580 7043
Recorded Pipe Failures 209 217 154 8 17 11 616
Unrecorded Failure Factor 1.48 1.27 1.13 1.03 1 1
Total Pipe Failures 310 278 175 9 17 11 800
Percentage of Failed Pipes 23.4% 19.0% 7.8% 1.6% 1.9% 1.9% 11.4%
The Weibull function used the percentage of failed pipes to create the curve and the curve
was then used to calculate the number of annual breaks that have occurred and will occur in the
future. The total number of breaks recorded was used to constrain the Weibull function and
produce a curve that represented the current system. The unrecorded failure factor was then
found by iteration.
-17- Pipe Replacement Technical Memo
The following table shows the comparison of the results of the extrapolated actual data and
the predicted data as given by the model. The Weibull curve was fit mainly to the older pipe
failures as they had more data points and provide stronger statistical accuracy. The newer pipe
showed larger variance from the model.
Table 6 Actual and modeled first break comparisonlo
Pipe installed Pipe installed Pipe installed Pipe installed Pipe installed
in 1950's in 1960's in 1970's in 1980's in 1990's
Extrapolated
Actual First 310 278 175 9 17
Time Breaks
Model
Predicted First 343 223 171 16 7
Time Breaks
Model Error 11% -20% -2% 78% -59%
Pipe Failure vs Time
120%
-a 1 p00%
c0 80%
lJ..
a
40%
20% - - - -
0% ~ ---T- -
0 100 200 300
-Weibull
Function Pipe Age (years)
• Actual Data
Figure 12 Weibull fit for first break
The Weibull function created a model that closely fit the actual data for first time breaks.
The city however does not replace pipe based on a first time break but typically waits until there
are more breaks to justify the replacement. For this reason more break classifir_ations were
created using filtered actual break data.
-18- Pipe Replacement Technical Memo
5.2 Classification 2- two breaks allowed before replacement
As more breaks are allowed to occur before replacement, less pipe replacement is necessary
in the near future but more will be required in the medium term. In the second classification
the Weibull function will be used to model the frequency of future second breaks. The model is
based on the recorded data showing pipes with two or more breaks. This classification is less
conservative than the first classification because many of the one time breaks were filtered out
and the result shows a system with a Power percentage of failed pipes.
Table 7 Data used for second break Weibull functionll
_ Summary of Pipes and Failures
1950's 1960's 1970's 1980's 1990's 2000's Total
Total Pipes Installed 1325 1462 2242 548 886 580 7043
Recorded Pipe Failures 75 84 43 1 3 2 208
Unrecorded Failure Factor 1.37 1.23 1.11 1.02 1 1
Total Pipe Failures 103 104 48 2 3 2 262
Percentage of Failed Pipes 7.8% 7.1% 2.1% 0.4% 0.3% 0.3% 3.7%
Table 8 Actual and modeled second time break comparisonli
Pipe installed Pipe installed Pipe installed Pipe installed Pipe installed
in 1950's in 1960's in 1970's in 1980's in 1990's
Extrapolated
Actual Second 103 104 48 2 3
Time Breaks
Model
Predicted 123 73 51 4 2
Second Time
Breaks
Model Error 19% -30% 6% 100% -33%
-19- Pipe Replacement Technical Memo
Pipe Failure vs Time
120%
100%
'a
m
80% -
v_
~ 60% -
a~
a 40% -
0
20% -
w
0 50 100 150 200 250
-weibuil Function Pipe Age (years)
¦ Actual Data
Figure 13 Weibull fit for second break
5.3 Classification 3- Three breaks allowed before replacement
The third classification is similar to two except more pipes are filtered out of the analysis as
only pipes with three or more oreaks are used.
Table 9 Oata used for third break Weibull functionlZ
Summary of Pipes and Failures
1950'5 1960'5 197Q's 1980'5 1990'5 2000'5 Total
Total Pipes Installed 1325 1462 2242 548 886 580 7043
Recorded Pipe Failures 36 40 23 1 0 0 100
Unrecorded Failure Factor 1:44 :1.28 1:14 1.03 1 1
Total Pipe Failures 52 52 27 2 0 0 133
Percentage of Failed Pipes 3.9% 3.6% 1.2% 0.4% 0.0% 0.0% 1.9 %
Table 10 Actual and modeled third time break comparison1z
Pipe installed Pipe installed Pipe installed Pipe installed Pipe installed
in 1950'5 in 1960'5 in 1970'5 in 1980'5 in 1990'5
Extrapolated - - --i-
Actual Third 52 52 27 l 0
Time Breaks
Model
Predicted Third 68 43 33 3 1
Time Breaks
Model Error 31% -17% 22% 50%
-20- Pipe Replacement Technical Memo
Pipe Failure vs Time
100%
i ~ 80% -
~ 60%
I y
40%
a
0 20% - - - -
0 50 100 150 200 250
WeibullFunction~ Pipe Age (years)
~ ¦ Actual Data ~
Figure 14 Weibull fit for third break
5.4 Classification 4- Four breaks allowed before replacement
The fourth classification is similar to the second and third where the number of failed pipes
has been filtered further. Only pipes with four breaks or more will be considered a failure
resulting in less total failures than the previous classifications.
Table 11 Data used for fourth break Weibull functionls
Summary of Pipes and Failures
1950's 1960's 1970's 1980's 1990's 2000's Total
Total Pipes Installed 1325 1462 2242 548 886 580 7043
Recorded Pipe Failures 14 23 8 0 0 0 45
Unrecorded Failure Factor 1.47 1.30 1.15 1:03 1 1
Total Pipe Failures 21 31 10 0 0 0 62
Percentage of Failed Pipes 1.6% 2.1% 0.4% 0.0% 0.0% 0.0% 0.9%
Table 12 Actual and modeled fourth time break comparisonls
Pipe installed Pipe installed Pipe installed Pipe installed Pipe installed
in 1950's in 1960's in 1970's in 1980's in 1990's
Extrapolated
Actual Third 21 31 10 0 0
Time Breaks
Model
Predicted Third 37 24 19 2 1
Time Breaks
Model Error 76% -23% 90%
-21- Pipe Replacement Technical Memo
Pipe Failure vs Time
70%
60%
~ 50%
~ 40%
a 30%
a 20%
0
10% `,7 -
0 50 100 150 200 250
•---Weibull Function Pipe Age (years)
¦ Actual Data
Figure 15 Weibull fit for fourth break
5.5 Summary of Classifications
Each of the classifications described above are different paths that can be taken financially
in the budgeting of pipe replacement. There are advantages and disadvantages of each and
therefore a balanced decision must be made in the selection. The balance that must be made is
between the money spent now on replacement and the money spent later on repair and
replacement, Each of the classification has a unique Weibull curve and failure response over
time. From these mathematical functions, annual pipe break curves were created to help
evaluate which financial "path" should be taken by the city,
The following table shows the predicted number of breaks for each classification.
Breaks per Year
180
160
lam.
140
120 -
-1 Break=Failure
z 100 - - - -~=1- -2 Breaks=Failure
A
60 - - 3Breaks-Failure
- d Breaks=Failure
60 - -
40 -
20 - - -
O N ~ (D CO O N 7 (D 00 O N e} fD N O N d- O O O
Ql O O m 01 O O O O O N N N N N
m O~ O~ m O O O O O O O O O O O O O O O O O
~ r r r N N N N N N N N N N N N N N N N
Year
Figure 16 Predicted annual breaks10
-22- Pipe Replacement Technical Memo
One of the most important factors in allocating sufficient funding for pipe replacement is
the public perception of breaks. Water main breaks can cause the public to distrust the utility's
ability to plan for and provide a stable water supply. According to the AWWA benchmarking
utility performance, the City of Boulder water distribution system is currently operating with less
annual breaks than other municipalities in the region. Above 120 annual breaks puts the city at
the average mark and more than 220 annual breaks causes the city to operate at the worst level
in the region.
In order to keep annual breaks below the 120 threshold in the medium term, a budgetary
plan similar to the results of the second classification must be taken. This means that 3 breaks
or more would be considered unacceptable and any pipes that have 3 breaks or more will need
to be replaced in conjunction with the new failing pipe. From the recorded break data, 135
pipes have 3 or more breaks and will have to be caught up to the current system. This accounts
for over $7 million that will have to be invested along with the predicted budgetary amount.
6 Plan Budget
Currently the City of Boulder spends around $2 million each year on pipe replacement. The
goal for pipe replacement is to keep the distribution system functioning properly without
excessive pipe breaks. From the results of the modeled scenarios, a predicted decade by decade
cost of pipe replacement has been generated for the City of Boulder to use in budgeting for
effective pips replacement in the future. This budgetary information is the goal for the all the
modeling as it gives a best estimate for pipe replacement schedule of costs in the future.
The following table summarizes the budgetary costs necessary for keeping the distribution
system consistent with the classifications discussed in the previous section. The actual average
annual pipe replacement during the decade of the 2000's is shown in the first column.
Table 13 Rehabilitation cost per decade in 2007 dollars14
Avera e Annual Re lacement Cost and Pi a Breaks
2000 2010 2020 2030
Classification $ 2,177,900 $ 4,121,600 $ 5,702,200 $ 6,575,600
1 39 57 73 85
Classification ~ 2,177,900 8 1,391,200 ~ 2,377,500 $ 3,547,900
2 53 80 107 130
Classification $ 2,177,900 $ 805,200 $ 1,351,300 $ 2,048,200
3 61 92 123 _ 152
Classification $ 2,177,900 S 450,300 ~ 752,400 $ 1,151,800
4 67 101 137 170
The projected annual pipe replacement costs and annual pipe repair costs are plotted in the
following charts for all four classifications. These charts visually show at what classification pipe
replacement becomes less than annual break repair. A $10,000 repair cost was assumed per
break which takes into account an estimated $5,000 direct cost of repair and $5,000 cost to
account for potential damages to private property and public inconvenience. The annual cost of
a break repair in 2008 was around $5,000. All values are shown in 2007 dollar value.
-23- Pipe Replacement Technical Memo
One Break Equals Replacement
$~,ooo,ooo
H $6,000,000
Y
d $5,000,000
m $4,000,000
$3,000,000
~ $2,000,000
U $1,000,000
---T--
1995 2000 2005 2010 2015 2020 2025 2030 2035 2040
Cost of Annual Break Repair Year
~ Cost of Annual Replacement
Figure 17 Classification 1 break repair vs replacementlo
Two Breaks Equals Replacement
$4,000,000
w $3,500,000
Y
~ $3,000,000
m $2,500,000 - _ -
$2,000,000
$1,500,000
~ $1,000,000 -
V $500,000
1995 2000 2005 2010 2015 2020 2025 2030 2035 2040
Cost of Annual Break Repair Year
-a- Cost of Annual Replacement
Figure 18 Classification 2 break repair vs replacement''0
-24- Pipe Replacement Technical Memo
- - -i
Three Breaks Equals Replacement
$2,500,000
w
~ $2,000,000 '
m
m $1,500,000
m $1,000,000 -
E
U $500,000
i
1995 2000 2005 2010 2015 2020 2025 2030 2035 2040
Cost of Annual Break Repair ~ Year
-~Costof Annual Replacement
- - -
Figure 19 Classification 3 break repair vs replacementlo
Four Breaks Equals Replacement
I
$2,500, 000
tll
~ $2.000.000
L
m $1,500,000 -
m $1,000,000
~ $500,000
U
1995 2000 2005 2010 2015 2020 2025 2x30 2035 2040
Cost of Annual Break Repair Year
Cost of Annual Replacement
Figure 20 Classification 4 break repair vs replacementlo
7 Recommendation
The analysis of the different classifications presented above produced results that can give
the City of Bou{der an advantage in preparing for the future. The water distribution system is a
critical part of any society and should be well maintained and cared for property. Classification 2
presented above is the best cost estimate for the city to use for planning the future CIP. When
considering the annual replacement cost based on classification 2 and the $7 million backlog of
pipe replacement needed to reach classification 2, it is recommended that the average annual
funding of $2.1 million be maintained during the next deeade.
-25- Pipe Replacement Technical Memo
8 Discussion on Confidence
The Weibull function is a statistical function that uses current data to produce a future
prediction model. There is always uncertainty when trying to make predictions about the future
especially when limited data exists. Because of the predictions made in this analysis, the
confidence in the accuracy of the results must be discussed. Statistically speaking, the two
factors that limit confidence in an analysis are lack of sufficient data and data that doesn't
support a common trend.
In order to better understand the results produced from the above analysis, a quantification
of the confidence was calculated. Quantifying the confidence in the results is highly dependent
on the quality of the raw data as well as the goodness of fit it has to a statistical function.
As seen in section 2 of this memo, pipe break data has anly been recorded for one third of
the current pipe lifetime. This lack of data created the need to extrapolate for unknown data
backwards in time before making a prediction about the future. Much of the pipe ages in the
system had to be estimated as well due to insufficient recorded data. Approximate pipe age
was estimated and grouped by decade which didn't provide many data points to create a curve
from. This can be seen in figures 12 through 15.
The data in this analysis had to be used to create the Weibull curve and fitting the curve to
the data wasn't an exact fit. This ability to fit the data is taken into account when quantifying
the confidence level of the results. The better the function can be fit to the data points the
higher the confidence will be in the final results. Due to both a low number of data points and
the quality of fit to the Weibull function, the greatest confidence that was realistic tc~ use was
80%.
The following figure shows the range of values of percent failure by year for the
recommended Classification 2. As seen in the figure, there is an 80% confidence that the actual
pipe will fail at a rate between the High and Low curves. The Weibull curve in the middle is the
best estimate and the curve used in the cost calculations above.
Percent Failure by Year
100%
~ 80% -
~ Weibull
LL 60% High
~ 40% - - Low
d 20% -
~o~ ~~ti ~o~
Year
Figure 21 Curves in the 80% confidence range
-26- Pipe Replacement Technical Memo
Using the values plotted in the above figure, a new cost estimate was created to show the
range of values that could be spent within the 80% confidence range. The following table shows
the results.
Table 14 80% confidence yearly cost estimate
Classification 2 Yearly Replacement Cost Range
2010 2020 2030 2040 2050
High Values $5,578,712 $9,343,575 $12,240,255 $12,834,276 $11,543,135
Weibull Values $1,391,200 $2,377,500 $ 3,547,900 $ 4,650,100 $ 5,368,900
Low Values $ 292,152 $ 427,950 $ 567,664 $ 651,014 $ 697,957
As seen in the above table, the possible necessary yearly cost can vary greatly from the
recommended estimate. This is where the decision and reevaluation process will need to be
used to make use of more available data in the future. As more data is used in creating the
Weibull curve, more potential for a higher confidence is possible.
9 " Conclusion
"
~~`Jncertainty will exist-whenever the use of a statistical function is used in a future prediction
" 'mbclel. We are also uncertain at this point why some pipes will fail and others will not in similar
conditions. The use of a statistical failure rate model such as the Weibull function captures the
~~unown failure mechanisms and predicts the future failure rate based on what the water
dim
~ribution system has historically experienced. The uncertainty can create doubt in ones mind
as to the relevance or importance of the use of statistical functions. The fact is that the Weibull
function has been used to create some certainty in the very uncertain pipe failure patterns.
City of Boulder staff has made the recommendation for future pipe replacement funding
with the understanding that it is the best estimate that can be made at the present time with
the available data. The City of Boulder currently provides excellent service to its water
customers and the results of the above analysis will help guide the city to continue providing
excellent service.
-27- Pipe Replacement Technical Memo
References and Notes
1. Much of the water main break data is unknown from the 1960's and 1970's although
the 1961 COB annual report stated 35 breaks. There has-been a significant increase in
annual water main breaks from the 1960's and the models predict even more annual
breaks in the future.
2. Arun K. Deb, et al. Prioritizing water main replacement and rehabilitation, AWWA
Research Foundation. 2002.
3. URS comprehensive review and assessment of the city of Westminster's water and
wastewater infrastructure, 2005.
All the following files are located in S:\PIN\ParkCentral\Utilities\Engineering\Nick
Worley\Pipe Replacement\Revision 2 Fall 2008
4. New Scenario 1.xls (Point System Basis tab)
5. New Scenario 1.xls (Pipe Data Summary tab)
6: New Scenario 1.xls (Break Data Summary tab)
7, Corrosion Breaks Map.jpg
8. Pressure Breaks Map.jpg
9. Settlement Breaks Map.jpg
10. New Scenario 1.xls (Weibull Function(New) tab)
11. New Scenario 2.xls
12. New Scenario 3.xls
13. New Scenario 4.xls
14. IUew Scenario 1.xls (Cost Summary tab)
1S. Denver Water- Conversation with Alan Crouch 1-27-09
-28- Pipe Replacement Technical Memo