Infrastructure
Evaluating the Risks of Water
Distribution System Failure
Effective asset management starts with accurate knowledge of system
components. Condition assessment modeling is helping utilities forecast
distribution system frailties. BY ROBERT M. CLARK AND ROBERT C. THURNAU
W
ater treatment
equipment can be
evaluated by mea­
suring performance
and inspecting the
maintenance paper trail. Assessing a
distribution system is more complex,
because it’s mostly buried and unavailable
for routine visual and physical inspection.
Water utilities have become
increasingly interested in using condition
assessment (CA) modeling to manage
their infrastructure assets. An important
aspect of CA modeling is the ability to
quantitatively estimate risks associated
with failure of distribution system
pipe segments. The US Environmental
Protection Agency (USEPA) has identified
the repair, replacement, and rehabilitation
of distribution system assets as important
economically and in the vital interest of
public health. Because of the national
Filter Effluent Particle Count Comparison
Filter effluent particle counts were consistently below the 20-particles/mL goal.
Inspection Technologies
Pipe Type
Acoustic Emission Monitoring
PCCP
Acoustic Leak Detection
All
Acoustical Leak Detection/Digital Correlator
All
Acoustical Monitoring Soundprint
PCCP
Broadband Electromagnetic
Steel
Closed-Circuit TV
All
Electromagnetic Inspection
PCCP
Long-Range Ultrasonic
Steel
Magnetic Flux Leakage
Steel, Cast Iron, and Ductile Iron
Remote Field Eddy Current
Steel, Cast Iron and Ductile Iron
Remote Field Transformer Coupling
PCCP
Ball Leak Detection
PPCP
Transient Pressure Monitoring
All
Ultrasonic
PCCP , Asbestos Cement
24 Opflow August 2011
2011 © American Water Works Association
importance of this issue, USEPA initiated
a project to develop a quantitative risk
model for drinking water infrastructure
repair and replacement.
PROJECT SCOPE
The project focused on four aspects of
drinking water infrastructure condition
and quantitative risk assessment. The
first part of the project focused on devel­
oping a quantitative, statistically based
risk model for pipe-segment failure. The
idea was to combine a technique called
frailty analysis with a Cox Proportional
Hazards Model (CPHM). The study team
collaborated with the Laramie (Wyo.)
Utility Division (Laramie Water) to quan­
tify the risk of failure associated with
individual pipe sections and complete
pipelines or runs of steel, cast-iron,
ductile-iron, and polyvinyl chloride
(PVC) pipe and to identify factors influ­
encing the risks.
The second aspect of the project
was to develop a model for estimating
costs and benefits, including secondary
and societal costs, of pipe repair,
rehabilitation, and replacement using
the Inspection Value Method evaluation
technique. The model was used to
examine the advantages of using
inspection technology to anticipate
needed replacement of pipe-run sections.
www.awwa.org/opflow
Condition assessment modeling has become
an increasingly important tool to help utilities
manage their infrastructure assets.
Robert M. Clark is an environmental engineering
and public health consultant based in Cincinnati.
Robert C. Thurnau is a project manager and
principal investigator with Eastern Research Group
(www.erg.com), Lexington, Mass.
components is presented in the table at
left.
One such technique used in water and
wastewater networks is a self-contained
electromagnetic device that rolls along
a pipeline to detect leaks. The device
was used as part of the cost–benefit
analysis.
photograph: AWWA
QUANTITATIVE RISK MODEL
The third aspect of the study assessed
prestressed cylindrical concrete pipe
(PCCP), which is of particular interest
because it’s frequently the means of
transmitting large volumes of water.
Therefore, when failure occurs, significant
damage likely results. Commonly used
in the 1960s and early 1970s, PCCP
was thought to solve water-related
transmission and collection pipe
problems. However, use of large-diameter
PCCP has led to increased financial
and product-loss risks associated with
pipeline failure.
www.awwa.org/opflow
The fourth part of the project was to
identify and assess current, state-of-the
art pipe inspection technology. Various
types of inspection technology have
been used to identify failure caused
by corrosion or stress in storage
tanks, pipes, and pressure vessels. To
conduct a CA of distribution system
mains, utilities must have access to
technologies that are nondisruptive to
normal service, accurate and reliable,
and affordable. A summary of currently
available inspection technologies and
their application to various pipeline
2011 © American Water Works Association
The model used data provided by Lara­
mie Water, which provides water and
sewer services to Laramie residents.
Pipe-break data provided by the city of
Laramie were analyzed using the CPHM,
which was modified by use of frailty
modeling. Frailty analysis is a technique
used by medical epidemiologists to iden­
tify subpopulations with a common sus­
ceptibility to a disease or weakness. In
terms of pipe failure, this means identi­
fying a group of pipes that seem to be
especially susceptible to failure from
specific causes. Frailty is defined as hav­
ing an expected (mean) value of one,
so it doesn’t alter conditional expected
values from the model but provides an
estimate of random factors that might
influence pipe breaks.
The Laramie dataset contains
information about individual pipe
sections, pipe runs, and pipelines.
Therefore, the first step was to develop
a model based on pipe sections. Next,
the models were aggregated into piperun models. Analysis was limited to pipe
diameters between 6 in. and 36 in. and
to pipes installed after 1940. Individual
pipe-run lengths varied from 1 ft to
9,241 ft.
Pipe Section Model. Risk models
were developed for metallic and PVC
pipe. According to the model, the frailty
variance for PVC pipe and metallic pipe
was determined to be 0.94 and 11.43,
respectively. Therefore, the PVC frailty
variance is an order of magnitude less
than the metallic pipe frailty, which
means that PVC pipe is much less
susceptible to local random external
August 2011 Opflow 25
Infrastructure
Figure 1. Mean Survival Probabilities for 24-in. Pipe Sections
Survival rates are identical for steel and cast-iron sections, but PVC and ductile-iron
sections have shorter survival rates.
1
Cast Iron and Steel
0.9
Survival Probability
0.8
Ductile Iron
PVC
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
10
20
30
40
50
60
Time (years)
70
80
90
100
Figure 2 illustrates the effect of
diameter on survival probability for DIP
sections. The same pattern was also
typical for other types of pipe. Figure
3 illustrates the effect of frailty on PVC
and DIP sections. For DIP, 5 percent of
the pipe had a much shorter survival rate
than an equivalent population of PVC
pipe. Several implications can be drawn
from this analysis. One interpretation is
that a fraction of DIP is more vulnerable
factors than those that affect metallic
pipes.
Figure 1 illustrates applying the model
to 24-in.-diameter pipe sections. The
mean survival probabilities for steel and
cast-iron pipe sections are identical, but
PVC and ductile-iron pipe (DIP) sections
have shorter survival rates. In Laramie,
larger-diameter pipes of all types had
better survival characteristics than
smaller-diameter pipes.
Figure 2. Survival Probabilities for Ductile-Iron Pipe Sections
Regardless of material used, large-diameter pipes fail less frequently than smallerdiameter pipes.
to unspecified environmental effects then
PVC pipe. In addition, it appears that,
when DIP begins to break, it may continue
to break at a faster rate than PVC.
Cost–Benefit Analysis. The pipe-section
model was integrated into a pipe-run
model and incorporated into a cost–
benefit model. Following is an example:
Assume a 7,000-ft, 24-in. DIP pipe
run consisting of 350 sections of 20-ft
pipe. Using the model, the pipe-run
replacement cost was calculated to be
$1.12 million ($3,200 per pipe section).
The Producers Price Index (http://
data.bls.gov/PDQ/servlet/Survey
OutputServlet) was used to upgrade
these costs to current costs (2008). These
data yielded a current total replacement
cost for the described DIP pipe run of
$1,568,000. Therefore, it’s assumed that
replacing a single section will cost $4,480
and the inspection technology cost would
be $20,000/mile/year. So, for a 7,000-ft
pipe, the inspection cost is estimated at
$30,000/yr. If a break actually occurs, it’s
assumed that two sections of pipe would
need to be replaced, so the repair cost
for a 24-in. DIP pipe would be $8,960.
It’s further assumed that, if a pipe
section doesn’t break but is identified
by the inspection technology device as
a candidate for replacement, only one
section of pipe is assumed to be replaced.
A cost–benefit model was developed
that includes the following factors:
1
Survival Probability
0.9
0.8
12in.
0.7
24in.
36in.
0.6
0.5
0.4
0.3
0
10
26 Opflow August 2011
20
30
40
50
60
Time (years)
70
80
90
100
2011 © American Water Works Association
V = value added by the inspection
AIC = annual cost of applying the
inspection technology
C = total cost of inspection, including
all direct and indirect costs associated
with restoration
E(t) = expected number of breaks for
a pipe-run at time t
F = percentage of the pipe covered
FR = cost of forced repair
POD = probability of detecting a
failure (technology specific)
R = preemptive cost of repair identified
by the application of inspection technology
www.awwa.org/opflow
The study shows that large-diameter pipes fail
less frequently than smaller-diameter pipes,
and all pipes fail more frequently with age.
Illustrating the cost–benefit model,
Figure 4 shows the relationship between
the benefit and the cost of inspection for
a 24-in. ductile iron pipe over time. For
the first 10 years, costs exceed benefits.
However, after 15 years, the use of
inspection technology yields a positive
benefit. Analysis shows the largest gains
are derived from larger-diameter pipes.
A possible asset management strategy
might be to delay inspection until the
benefits are clear. Analysis of 12-in.and 36-in.-diameter pipe indicates
benefits increase dramatically with
diameter. Although this analysis, based
on data compiled by Laramie Water,
shows a positive benefit for inspection
technology, selected results might be
dramatically different for another type
of device or material.
PCCP ANALYSIS
In a recent survey, USEPA found that, of
202,128 mi of reported distribution sys­
tem pipes, 4,774 mi (2.3 percent) were
PCCP. If that number is extrapolated
on a national basis, it’s estimated that
about 1 million mi of pipe are in ser­
vice, of which 23,000 miles are PCCP.
Use of large-diameter PCCPs increased
the financial and product-loss risks asso­
ciated with failure. Some catastrophic
drinking water transmission line rup­
tures have illustrated this problem. For
example, in 2006, the San Diego Water
Authority responded quickly to a rup­
tured 96-in. line. Although response was
rapid, an estimated 2 mil gal of water
were lost before the break was brought
under control. When direct and indirect
repair costs of the break were totaled,
several millions of dollars were spent
in repairs. Reducing the failure risk by
www.awwa.org/opflow
1
0.9
Median Survival
0.8
Survival Probability
V = [(E(t) * POD * F * FR) + (E(t) * POD
* F * SC)] – [(E(t) * POD * F * R) + E(t) *
(1 – POD) * F * (FR + SC) + AIC)]
Figure 3. Frailty Effect on Mean Survival Rate
Some ductile-iron pipe sections (top) had a much shorter survival rate than an
equivalent population of PVC pipe sections (bottom).
95% Quantile Curve
0.7
0.6
5% Quantile Curve
0.5
0.4
Average Survival
0.3
0.2
0.1
0
0
10
20
30
40
50
60
Time (years)
70
1
80
90
100
95% Quantile Curve
0.9
Median Survival
0.8
Survival Probability
SC = cost to society of water loss and
other indirect costs
Average Survival
0.7
0.6
0.5
5% Quantile Curve
0.4
0.3
0.2
0.1
0
0
10
20
30
40
50
60
Time (years)
knowing the breakage probabilities of
large-diameter pipes can facilitate tar­
geted maintenance, reduce costs, and
allow more efficient use of existing and
future assets.
A comprehensive dataset for largediameter PCCPs was identified and used
to analye PCCP breakage rates. The
database contained 588 entries from 153
utilities and included pipe diameter and
type, wire class and size, age at ruptures
and other failures, failure condition,
length, pressure, and manufacturer.
Thirty-one pipe diameters, ranging from
16 in. to 252 in., were represented in the
database.
2011 © American Water Works Association
70
80
90
100
The database was examined for
specific pipelines that had undergone
multiple maintenance activities over
time. The pipelines that met this
requirement were 24 in., 30 in., 36 in.,
48 in., 60 in., 72 in., 84 in., 90 in., and
96 in. in diameter. At least one pipeline
was selected for analysis from each of
the diameters. A total of 15 pipelines was
selected, which yielded 226 individual
maintenance activities, representing
more than 112 service observations. A
regression analysis determined that,
of the three variables studied affecting
breakage rate/mi, only age and diameter
were significant.
August 2011 Opflow 27
Infrastructure
Figure 4. Inspection Benefit and Cost Over Time
Total Cost (millions USD)
For 24-in. ductile-iron pipe, inspection technology yields a positive benefit after 15 yr.
8
7
6
5
4
3
2
1
0
Inspection Benefit
Inspection Cost
0
20
40
60
Time (years)
80
100
120
STUDY RESULTS PROMISING
Figure 5. Predicted Break Rate per Mile With Regard to
Diameter and Age
Break rate decreased with diameter (top) and increased with age (bottom).
4.5
Calculated Break Rate
4
24
30
36
48
60
72
84
90
96
3.5
3
2.5
2
1.5
1
0.5
0
0
10
20
30
Age (years)
40
50
in.
in.
in.
in.
in.
in.
in.
in.
in.
60
4.5
0 years
Calculated Break Rate
4
5 years
3.5
10 years
3
15 years
2.5
20 years
25 years
2
30 years
1.5
35 years
1
40 years
0.5
45 years
0
50 years
0
20
28 Opflow August 2011
40
60
Diameter (in.)
80
Break rate decreased with diameter
and increased with age. Although age and
diameter are probably the most important
variables, considerable work is required
before PCCP break rates can be correlated
with a high degree of confidence. The
results of those calculations are shown in
Figure 5.
The study shows that large-diameter
pipes fail less frequently than smallerdiameter pipes, and all pipes fail more
frequently with age.
100
120
2011 © American Water Works Association
In summary, a model incorporating
CPHM with shared frailty (by pipe sec­
tion) for metallic and PVC pipe was
developed to estimate pipe-break risks.
The Inspection Value Method, which
incorporated CPHM, was applied to
Laramie Water’s pipeline data based on
a hypothetical 24-in. DIP to illustrate rel­
ative costs and benefits of using inspec­
tion technology. A separate pipe-break
analysis for PCCP revealed that largerdiameter PCCP had lower break rates
than smaller-diameter PCCP pipes. Var­
ious inspection technologies were
examined, evaluated, and found to be
promising.
Authors’ Note: The authors acknowledge USEPA’s Office of Research and
Development, which funded, managed,
and collaborated in this research, which
was subjected to USEPA’s administrative review and approved for external
publication. Opinions expressed in this
article are those of the authors; no official endorsement by USEPA should be
inferred. The authors acknowledge the
assistance of Peter D. Rogers, University of Texas–Tyler; Neil S. Grigg, Colorado State University; and Cal Van Zee,
Laramie Water. In addition, the authors
would like to acknowledge John Carson, E.R. Krishnan, and Sri Panguluri
of Shaw Environmental & Infrastructure
for their assistance in completing this
project.
www.awwa.org/opflow