EPA/600/R-13/080 I December 2013 I www.epa.gov/research
United States
Environmental Protection
Agency
Primer on Condition Curves
for Water Mains
Office of Research and Development
-------
EPA/600/R-13/080
December 2013
Final Report
Primer on Condition Curves for Water Mains
by
James Thomson
and
Stephanie Flamberg and Wendy Condit
Battelle
Columbus, OH 43201
Contract No. EP-C-05-057
Task Order No. 0062
for
Michael Royer
Task Order Manager
Water Supply and Water Resources Division
Urban Watershed Management Branch
2890 Woodbridge Avenue (MS-104)
Edison, NJ 08837
National Risk Management Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH 45268
-------
DISCLAIMER
The U.S. Environmental Protection Agency (U.S. EPA), through its Office of Research and Development,
funded and managed, or partially funded and collaborated in, the research described herein under Task
Order (TO) 0062 of Contract No. EP-C-05-057 to Battelle. It has been subjected to the Agency's peer
and administrative review and has been approved for publication. Any opinions expressed in this report
are those of the author(s) and do not necessarily reflect the views of the Agency, therefore, no official
endorsement should be inferred. Any mention of trade names or commercial products does not constitute
endorsement or recommendation for use. The quality of secondary data referenced in this document was
not independently evaluated by EPA and Battelle.
-------
ABSTRACT
The development of economical tools to prioritize pipe renewal based upon structural condition and
remaining asset life is essential to effectively manage water infrastructure assets for both large and small
diameter pipes. One tool that may facilitate asset management is a condition curve. A condition curve is
a graphical representation of the condition of a pipeline versus time. This report provides a review of the
state-of-the-technology for structural/physical condition curves for water mains. Various models are
summarized such as break frequency curves, deterioration/decay/survival curves, condition rating curves
and condition rating indices, and serviceability/performance curves. This report also provides new case
study information on how condition curves are used by utilities for managing their water infrastructure
based upon a survey of nine utilities. The utilities that were surveyed for these case studies used methods
that ranged from very detailed asset management programs that combine inspection, monitoring, and test
data with their pipeline condition assessment program to simple analyses of pipe break history to
prioritize pipeline renewal activities. The review also discusses short-term and long-term research needs
for further development of a performance-based buried infrastructure asset management approach to
improve the quality and quantity of data used by all utilities.
in
-------
ACKNOWLEDGMENTS
The authors would like to thank Frank Blaha, Yehuda Kleiner, Walter Graf, Anthony Tafuri, and
Ariamalar Selvakumar for providing written comments. Sincere appreciation is extended to all of the
water utilities who took the time to provide information to Virginia Tech on their condition assessment
programs (EPCOR Water Services Inc., Las Vegas Valley Water District, Newport News Waterworks,
Seattle Public Utilities, Sydney Water, Washington Suburban Sanitary Commission, City of Hamilton
Public Works Department, Louisville Water Company, and Philadelphia Water Department).
IV
-------
EXECUTIVE SUMMARY
A key requirement for asset management is to understand the condition of pipelines in a system. The
development of economical tools to prioritize pipe renewal based upon structural condition and remaining
asset life is essential to effectively manage water infrastructure assets for both large and small diameter
pipes. One tool that may facilitate asset management is a condition curve. A condition curve is a
graphical representation of the condition of a pipeline versus time. If the appropriate curve can be
matched to the pipe system of interest, then condition curves can be used to estimate the current
condition, remaining asset life, and failure rate of a pipeline. These estimates can be very useful for both
short-term and long-term maintenance and capital improvement planning.
The term "condition" can take on many different meanings for water utilities. To some it means the
structural condition of the pipe, and to others it may mean pipe serviceability (i.e., the ability of the pipe
to provide the type of service expected by its customers). For most utilities, the term "condition" often
involves several different factors such as structural condition, water quality, hydraulic capacity,
serviceability, location, and economics. Therefore, to develop condition curves, utilities must first start
by defining what "condition" means for their particular concern, and what constitutes unacceptable
conditions that require action (e.g., inspection, repair, rehabilitation, or replacement).
"Structural condition" of the pipeline is narrowly defined in this review as the presence/absence of holes,
cracks, breaks, or circumstances leading to their formation, in the transmission or distribution pipe wall,
lining, coating, and joints. Structural condition does not, as defined here, generally include occlusion of
the pipe bore by tuberculation, scale, or other deposits.
The term "condition curve" can be defined in several ways. To avoid confusion, it is important to specify
the particular definition in use. The general definition for condition curve is a graphical representation of
condition versus time. The condition of a pipeline can refer to its hydraulic, water quality, economic, or
structural condition. However, in this report, the focus is on the structural/physical condition curves of
pipelines or cohorts of pipes. Condition curves are most often generated for a pipeline (e.g., a contiguous
section of pipes) or pipe cohorts (e.g., a relatively homogenous population of pipes expected to have
similar physical, environmental, and operational characteristics and therefore similar performance).
This report provides a review of the state-of-the-technology for structural/physical condition curves for
water mains. Various classes of models are summarized such as break frequency curves,
deterioration/decay/survival curves, condition rating curves and condition rating indices, and
serviceability/performance curves. In order to define and document the use of condition curves and
deterioration models, a comprehensive literature review was performed including an examination of
research efforts undertaken by organizations such as the Water Research Foundation (WaterRF), Water
Environment Research Foundation (WERF), National Research Council Canada (NRC), Commonwealth
Scientific and Industrial Research Organization (CSIRO), U.S. EPA, and others. The information on
condition curves obtained from this state-of-the-technology review is summarized, along with selection
factors, advantages, and limitations for the use of condition curves.
This report also provides new case study information on how condition curves are used by utilities for
managing their water infrastructure based upon a survey of nine utilities (along with three additional case
studies presented in the literature). The nine utilities surveyed operate atotal of over 32,000 mi. of water
pipe. The pipe networks were overwhelmingly less than 21 in. in diameter. There was considerable
variance in the pipe types used, but cast iron and ductile iron have the greatest length for most utilities.
The utilities that were surveyed for these case studies used methods that ranged from very detailed asset
-------
management programs that combine inspection, monitoring, and test data with their pipeline condition
assessment program to simple analyses of pipe break history to prioritize pipeline renewal activities.
It was found that the most widely used approach was the break frequency curve. In the literature, over 20
models based on break frequency have been developed in the last 30 yr. Out of the twelve utilities
examined, nine of them made use of break frequency approaches within their condition assessment
programs. The next most common practice was the use of a condition rating curve and/or condition rating
index with five utilities reporting use of this approach. Because of the difficulties and cost of modeling,
some utilities use these rating systems that are based largely upon judgment, expert opinion and/or
performance indicator data to determine criticality and assign priorities. Although this approach is reliant
upon expert opinion, the condition rating curve or index approach does provide a logical and documented
framework for determining pipeline renewal priorities. Four examples were identified of utilities that
have reported using some form of deterioration, decay, or survival curves.
Any asset management program must start with a thorough review of available historical data about pipe
performance and failure. Once the necessary data are gathered, condition curves and/or deterioration
models can go a long way in providing insight into the condition of these assets. In general,
empirical/statistical models are an economically viable approach for smaller distribution water mains.
Currently, only large water mains with costly consequences of failure may justify the cost of
accumulation of data that are required for physical model application. For these high risk pipelines,
where failures are catastrophic and unacceptable at any time, more extensive and complex approaches
may be warranted and cost effective. Condition curves for larger diameter pipes should generally be
generated from hard data based on non destructive testing inspections, investigations, and/or laboratory
testing to define pipe condition and obtain more accurate predictions of remaining life.
There are several short-term and long-term needs for the development of a performance-based buried
infrastructure asset management approach that could yield major improvements in the quality and
quantity of data used by all utilities.
A general consensus is that any condition curve should be simple to understand, transparent to the users,
and easy to implement. There is need for the development of standardized methodologies for data
collection and standardized protocols to generate and calibrate condition curves for the site-specific data
collected. There is also very little subsequent validation of the condition curves and/or deterioration
models with "real-world" case studies, which is also an important need. Piloting existing and/or new
models at various utilities could be conducted to define their practical use and ease of adoption by
utilities.
Additional research is needed on how to design more efficient and cost-effective data collection
strategies, how to extract information from existing datasets, and how to standardize names and
definitions for water utility assets, which subsequently will allow the data to be shared and compared
across utilities. The lack of data for many utilities is a major limitation in using anything, but the most
basic approaches to condition assessment. However, in many cases, the reality is that only partial data
exist. A robust model or condition curve should be able to deal with partial data, but it should be clear
that in general the results will be less accurate and less precise compared to those results obtained with a
more complete dataset.
Further research can help to continue to refine critical inferential parameters that affect water pipe
performance based on pipe material, diameter, joint type, external and internal environmental factors, and
more. A longer-term goal should be to develop a national database of assets and failures with common
terminology and methods of data collection and analysis for assets and breaks. WaterRF, in partnership
with UKWIR, is currently assessing the feasibility of such a national database. Ultimately, this will allow
VI
-------
identification of the most vulnerable pipes, reduce failures, and improve understanding of the type and
distribution of failure modes and indicators.
Additional guidance should also be developed for identifying and quantifying the high risk scenarios,
which requires characterizing both the likelihood and consequence of failure. With limited funds, it is
necessary for utilities to focus on the highest risk situations to limit the impact of failures on consumers
and the public.
vn
-------
CONTENTS
DISCLAIMER ii
ABSTRACT iii
ACKNOWLEDGMENTS iiv
EXECUTIVE SUMMARY v
FIGURES x
TABLES x
ACRONYMS AND ABBREVIATIONS xi
1.0: INTRODUCTION 1
1.1 Background 1
1.2 Project Objective and Scope 1
1.3 Target Audience 1
1.4 Basic Terminology 2
1.5 Use of Condition Curves and Reasons for Using Condition Curves 2
1.6 Project Approach 3
1.7 Report Organization 3
2.0: ROLE OF CONDITION CURVES IN ASSET MANAGEMENT AND ISSUES IN THEIR
DEVELOPMENT 4
2.1 Defining Asset Life 4
2.2 Defining End of Asset Life 4
2.3 Defining Remaining Asset Life 5
2.4 Assessing Asset Condition 5
2.5 Pipeline Functional Life 6
2.6 Condition Assessment Tool 7
2.7 Pipe Networks, Characteristics, and Behavior 8
2.7.1 The Distribution and Transmission Water Network 8
2.7.2 Ferrous Pipe 8
2.7.3 Asbestos Cement Pipes 11
2.7.4 Prestressed Concrete Cylinder Pipes 12
2.7.5 Polyvinyl Chloride Pipes 14
2.8 Key Issues for Condition Curves 15
3.0: REVIEW OF METHODS USED TO GENERATE CONDITION CURVES AND TYPES
OF CONDITION CURVES 17
3.1 Development of Condition Models 17
3.2 Types of Condition Curves 20
3.2.1 Break Frequency Curves 20
3.2.2 Deterioration, Decay and Survival Curves 23
3.2.3 Condition Rating Curves and Condition Rating Indices 25
3.2.4 Serviceability/Performance Methods 29
3.2.5 Economic Models - Prediction of Annual Replacement Curves 30
3.3 Structural Condition Curve Selection Factors, Benefits, and Limitations 31
3.3.1 Benefits of Condition Curves 32
3.3.2 Limitations of Condition Curves 33
4.0: CURRENT USE OF CONDITION CURVES 35
4.1 Who is Using Condition Curves 35
4.1.1 Types of Condition Curves Being Used by Utilities 35
Vlll
-------
4.1.2 Primary Pipe Types and Sizes for Condition Curves 35
4.2 Case Studies from Various Utilities 36
4.2.1 Development of Survey 36
4.2.2 Participating Utilities 36
4.2.3 Water Pipe Inventories for Case Study Utilities 36
4.2.4 Inspection, Monitoring and Condition Assessment Programs 36
4.3 Survey Results: State-of-the-Practice in Pipe Condition Curves for Renewal
Prioritization 39
4.3.1 EPCOR Water Services Inc 39
4.3.1.1 Reactive Water Pipe Renewal Program (Break Frequency Curves) 39
4.3.1.2 Proactive Water Pipe Renewal Program (Condition Rating Index) 40
4.3.1.3 Validation of Condition Assessment Models and Associated Costs
for Generating Models 41
4.3.2 Las Vegas Valley Water District 41
4.3.2.1 CARE-W and Casses Software (Break Frequency, Deterioration,
and Economic Models) 41
4.3.2.2 Validation of Condition Assessment Models and Associated Costs
for Generating Models 43
4.3.3 Newport News Waterworks 43
4.3.3.1 Pipe Replacement Prioritization Program (Break Frequency) 43
4.3.3.2 Validation of Condition Assessment Models and Associated Costs
for Generating Models 43
4.3.4 Seattle Public Utilities 44
4.3.4.1 Wave Rider (Economic Model) 44
4.3.4.2 Water Main Replacement Model (Condition Rating Index) 44
4.3.4.3 Validation of Condition Assessment Models and Associated Costs
for Generating Models 44
4.3.5 Sydney Water 46
4.3.5.1 KANEW (Deterioration, Decay, and Survival Curves) 46
4.3.5.2 Pipeline Asset and Risk Management System (Deterioration,
Decay, and Survival Curves and Condition Rating Index) 46
4.3.5.3 Validation of Condition Assessment Models and Associated Costs
for Generating Models 47
4.3.6 Washington Suburban Sanitary Commission 47
4.3.6.1 Water Pipe Condition Rating System (Condition Rating Index) 48
4.3.6.2 Validation of Condition Assessment Models and Associated Costs
for Generating Models 48
4.3.7 City of Hamilton Public Works Department 49
4.3.7.1 Infor Hansen Asset Management System (Economic Model) 49
4.3.8 Louisville Water Company 49
4.3.8.1 Pipe Evaluation Model (Condition Rating Index) 49
4.3.9 Philadelphia Water Department 50
4.3.9.1 Structural Condition Model (Safety Factor and Deterioration,
Decay, and Survival Curves) 50
4.3.9.2 Point System (Break Frequency) 51
4.3.10 The Metropolitan District 51
4.3.10.1 Break Frequency and Deterioration, Decay, Survival Curves 51
4.3.11 The Los Angeles Department of Water and Power 53
4.3.11.1 Break Frequency Curves 54
4.4 Summary of State-of-the-Practice Review 54
IX
-------
5.0: FINDINGS AND RECOMMENDATIONS 57
5.1 Selecting the Approach 57
5.2 Application of Condition Assessment Models 58
5.3 Snort-Term Needs and Improvements 59
5.4 Long-Term Needs and Improvements 60
5.5 Conclusions 61
6.0: REFERENCES 62
FIGURES
Figure 2-1. Functional Life of a Pipeline 6
Figure 2-2. The "Bathtub" Curve 6
Figure 2-3. Condition Curves for a Pipeline with and without Rehabilitation 7
Figure 3 -1. Optimal Renewal Frequency for Distribution Mains versus Transmission Mains 18
Figure 3-2. Optimal Timing of Intervention 19
Figure 3-3. Example of Break Frequency Curves by Pipe Class 22
Figure 3-4. Typical Condition Curve Based on Broad Performance Indicators 26
Figure 3-5. Overall Belief Network - Likelihood and Consequences 27
Figure 3-6. Asset Replacement Cost Profile 31
Figure 4-1. Example of Area Criteria Ranking and Weights for Cast Iron Water Pipe 40
Figure 4-2. Area Ranking Threshold Values and Graphical Prioritization for Cast Iron Water Pipe 41
Figure 4-3. CARE-W Data Flowchart for LVVWD 42
Figure 4-4. Sydney Water's Risk Ranking Matrix 47
Figure 4-5. WSSC Risk Model Range 48
Figure 4-6. Example of LWC's 2007 PEM Criteria and Scoring 50
Figure 4-7. Deterioration Curve for Cast Iron Pipe Greater than 10-in. Diameter 52
TABLES
Table 2-1. U.S. Pipeline Mileage by Pipe Material and Pipe Diameter 8
Table 2-2. Summary of Factors Leading to Failure for Ferrous Pipes 9
Table 2-3. Ferrous Pipe Break Frequency/100 miles/yr 10
Table 2-4. Cast, Spun, and Ductile Iron- Forms, Causes, and Indicators of Failure 10
Table 2-5. Asbestos Cement - Forms, Causes, and Indicators of Failure 12
Table 2-6. PCCP - Forms, Causes and Indicators of Failure 13
Table 2-7. PVC Pipes - Forms, Causes, and Indicators of Failure 15
Table 4-1. Water Pipe Inventories for Case Study Utilities 37
Table 4-2. SPU Input Parameters Included in the Water Main Replacement Model 45
Table 4-3. Outputs of the Water Main Replacement Model 46
Table 4-4. The PWD Points Assigned for Year of Installation and Break Frequency 51
Table 4-5. Pipe Class Replacement Ages 53
Table 4-6. Summary of Utility Inspection, Condition Index/Performance Measures, and Models 55
-------
ACRONYMS AND ABBREVIATIONS
AC asbestos cement
ANN artificial neural networks
AWWA American Water Works Association
AwwaRF American Water Works Association Research Foundation
CARE-W Computer Aided Rehabilitation of Water Networks
CIP capital improvement program
CP cathodic protection
CSIRO Commonwealth Scientific and Industrial Research Organization
DIPRA Ductile Iron Pipe Research Association
D-WARP Distribution - Water Mains Renewal Planner
GIS geographic information system
I-WARP Individual Water Main Renewal Planner
LADWP Los Angeles Department of Water and Power
LEFM linear elastic fracture mechanics
LEYP Linearly Extended Yule Process
LVVWD Las Vegas Valley Water District
LWC Louisville Water Company
NOT nondestructive testing
NPV net present value
NRC National Research Council - Canada
PCCP prestressed concrete cylinder pipe
PE polyethylene
PEM Pipe Evaluation Model
ppm parts per million
PV present value
PVC polyvinyl chloride
PWD Philadelphia Water Department
RPV replacement priority value
SIF stress intensity factor
SPU Seattle Public Utilities
T-WARP Transmission - Water Mains Renewal Planner
UKWIR U.K. Water Industry Research
UMP Utility Master Plan
U.S. EPA United States Environmental Protection Agency
WaterRF Water Research Foundation (formerly known as AwwaRF)
WERF Water Environment Research Foundation
WRc Water Research Center
WSSC Washington Suburban Sanitary Commission
xi
-------
1.0: INTRODUCTION
1.1 Background
Effective management of water supply networks is a major challenge for all water utilities. It is complex
not only because of the vast array of components, pipe sizes, vintages, and pipe types that exist, but also
because most pipelines are installed underground (i.e., out of sight), in variable soil conditions, and
subject to various environmental factors.
A key requirement for asset management is to understand the condition of pipelines in a system.
Although on-site, real-time inspection of a pipeline is the ideal method to analyze and understand its
condition, this approach can be expensive and usually cannot be cost effectively applied to smaller
diameter distribution lines, which make up the majority of water systems. Therefore, the development of
economical tools to prioritize pipe renewal is essential to effectively manage water infrastructure assets
for both large and small diameter pipes.
One tool that may facilitate asset management is a condition curve. A condition curve is a graphical
representation of the condition of a pipeline versus time. If the curve is accurate, and if the appropriate
curve can be matched to the pipe system of interest, then condition curves can be used to estimate the
condition, remaining asset life, and failure rate of a pipeline. These estimates can be very useful for both
short-term and long-term maintenance and capital improvement planning. Although the pipe condition
curve concept is fairly straightforward, the state-of-the-technology is not so obvious.
1.2 Project Objective and Scope
The objective of this project was to review the state-of-the-technology for structural/physical condition
curves for water mains and to produce a document that consolidates, analyzes, and clearly presents this
information as a guideline for utilities. This report documents the review of various classes of models
such as break frequency curves, deterioration/decay/survival curves, condition rating curves and condition
rating indices, and serviceability/performance curves. The report does not provide an exhaustive list of
models in each category, but rather provides a few examples to illustrate the concept of each type of
curve.
The scope of this report is limited to structural/physical condition curves as they have the greatest
potential for prediction/management of water main breaks and serve as a tool for pipe renewal planning.
Pipe investment and renewal decisions are also based on other considerations (e.g., hydraulics, water
quality, economics, and related infrastructure), but these factors are not the focus of this report.
1.3 Target Audience
This report is aimed at relatively new practitioners and utility managers of medium (3,300 to 10,000
customers) to large (10,000 to 100,000 customers) community water systems. The report is intended to
provide a basic definition for structural/physical condition curves and to present a brief overview of the
components of a condition curve, condition curve variations, and how condition curves are used as tools
for making asset management decisions. Experienced asset management practitioners looking for more
in-depth information should refer to detailed reports published by the Water Environment Research
Foundation (WERF) and the Water Research Foundation (WaterRF).
-------
1.4 Basic Terminology
The term "condition" can take on many different meanings for water utilities. To some it means the
structural condition of the pipe, and to others it may mean pipe serviceability (i.e., the ability of the pipe
to provide the type of service expected by its customers). For most utilities, the term "condition" often
involves several different factors such as structural condition, water quality, hydraulic capacity,
serviceability, location, and economics. Therefore, to develop condition curves, utilities must first start
by defining what "condition" means for their particular concern, and what constitutes unacceptable
conditions that require action (e.g., inspection, repair, rehabilitation, or replacement).
"Structural condition" of the pipeline is narrowly defined here as the presence/absence of holes, cracks,
breaks, or circumstances leading to their formation, in the transmission or distribution pipe wall, lining,
coating, and joints. Structural condition does not, as defined here, generally include occlusion of the pipe
bore by tuberculation, scale, or other deposits.
"Structural condition assessment" involves: (1) development of a formal or informal structural condition-
rating approach that links pipeline parameter data to the likelihood of structural failure (i.e., holes, cracks,
and breaks) for the time period of interest; (2) collection of data (e.g., physical, environmental, and
operational characteristics; failure history, processes, and associated indicators) by applicable direct
and/or indirect methods; and (3) analysis of the pipeline data and information to categorize the pipe's
current and future structural condition as input for pipe renewal decisions. This analysis could be based
upon empirically derived statistical models or physical/mechanistic modeling, which is based on pipe-
intrinsic properties, internal/external loading, and other factors (United States Environmental Protection
Agency [U.S. EPA], 201 la).
The term "condition curve" can be defined in several ways. To avoid confusion, it is important to specify
the particular definition in use. The generic definition for condition curve is a graphical representation of
condition versus time. In general, the condition of a pipeline can refer to its hydraulic, water quality,
economic, or structural condition. However, in this report, the focus is on the structural/physical
condition curves of pipelines or cohorts of pipes. Condition curves are most often generated for a
pipeline (e.g., a contiguous section of pipes) or pipe cohorts (e.g., a relatively homogenous population of
pipes expected to have similar physical, environmental, and operational characteristics and therefore
similar performance). Unless it is a high consequence scenario, it is not cost effective to generate a
condition curve for a single pipe (e.g., a pipe segment from bell to spigot).
Structural condition can be determined by various approaches (e.g., by historical data, by inspection, by a
point system, by a model, or some combination of these approaches). The time axis on the structural
condition curve does not necessarily have to be the age of the pipe. It could, for example, be pressure
cycles or temperature cycles, but one must be able to convert the independent variable into a time
measurement to use the condition curve to support decisions about when to take corrective action.
1.5 Use of Condition Curves and Reasons for Using Condition Curves
The main reason for using condition curves is to plan and prioritize renewal projects. Condition curves
can be used as a tool to predict the remaining asset life of a pipe and therefore to plan for the overall
timing of renewal activities. Condition curves can also be useful tools in rating or scoring the pipeline
condition and therefore assist in quantifying the probability of failure. Ultimately, this information, when
combined with other considerations (such as hydraulics, water quality, economics, and related
infrastructure), can be used to prioritize renewal activities for more efficient expenditure of utilities'
annual capital improvement budget.
-------
The different techniques and methods currently used by utilities to develop condition curves include break
frequency curves, deterioration, decay, and survival curves, condition rating curves and condition rating
indices, and serviceability curves. These curves are discussed in greater detail in Sections 2.0 and 3.0 of
this report with examples of their application by utilities presented in Section 4.0.
There are also economic-based curves, such as Nessie curves, that are developed and used to plan for the
future cost of network replacement. These curves are not based on structural condition, but strictly on the
time of installation and design life. These types of curves are not covered in great detail in this report;
however, references are provided should the reader want to seek additional information on the
development and use of Nessie curves.
1.6 Project Approach
This report was developed based upon a comprehensive literature review including an examination of
extensive research undertaken by organizations such as the WaterRF, WERF, National Research Council
Canada (NRC), Commonwealth Scientific and Industrial Research Organization (CSIRO), and U.S. EPA.
In addition to a literature review, this report summarizes new case study information on how condition
curves are used by utilities for managing their water infrastructure based upon a survey of nine utilities
conducted by Virginia Tech.
In its basic form, this report is meant to serve as a primer to provide utilities or new practitioners of asset
management with a concise overview of the various types of condition curves available. The key feature
of this report distinguishing it from previous research efforts is that it offers a concise overview within a
single document that covers the types of condition curves, their benefits and limitations, who is using
them, and for what purpose. It enables readers to understand the general nature and various classes of
condition curves with reference to more detailed documents for the specifics on data requirements and
model/curve development. A comprehensive list of references is provided as a valuable resource for
those interested in finding out more details on how to implement condition curves and their associated
data requirements. The report also highlights gaps between the state-of-the-art in the literature and state-
of-practice in the field.
1.7 Report Organization
This report is organized into five sections that include introductory material (Section 1.0), the role of
conditions curves in asset management and issues in their development (Section 2.0), review of methods
used to generate condition curves and types of condition curves (Section 3.0), current use of condition
curves (Section 4.0), and findings and recommendations on condition curve gaps and research needs
(Section 5.0).
-------
2.0: ROLE OF CONDITION CURVES IN ASSET MANAGEMENT AND ISSUES IN THEIR
DEVELOPMENT
The main goal for determining pipeline condition is to gain a better understanding of remaining asset life.
By understanding the remaining asset life, utilities will be able to better prioritize and optimize
operations, maintenance, and capital improvement decisions. This will help to reduce pipeline failures
and their adverse effects and minimize life-cycle costs.
One tool for predicting remaining asset life is the condition curve. This section sets up the role of
condition curves in asset management by first providing basic definitions of asset life, end of asset life,
and remaining asset life as defined in the WERF report "Remaining Asset Life: A State of the Art
Review" (Marlow et al, 2009). This is followed by a discussion of how condition curves can be used as a
tool for predicting remaining asset life, along with an overview of distress indicators for the predominant
pipe types (ferrous, asbestos cement [AC], prestressed concrete cylinder pipes [PCCP], and plastic).
Lastly, key issues are discussed related to the development and use of condition curves.
2.1 Defining Asset Life
Asset life could be defined as the time between installation and reaching an "end of life" criterion
(defined in Section 2.2). Additional terminology related to the life of an asset from Marlow et al. (2009)
is presented below with some modification to focus on water pipelines as follows:
Design life, the period of time over which the pipeline is designed to be available for use and
able to provide the required level of service at an acceptable risk of failure (e.g., the product
of failure likelihood and failure consequence).
Service life, the period of time over which the pipeline is actually available for use and able
to provide the required level of service at an acceptable risk of failure (e.g., without
unforeseen costs of disruption for maintenance and repair).
Operational life, some pipeline assets may be operated past the point where they provide the
required level of service at an acceptable risk of failure. As such, the operational life is taken
to be the time over which the asset remains operational irrespective of its serviceability or
performance. This situation tends to occur where the expected cost of a failure is lower than
the cost of mitigating or preventing the failure via renovation or replacement.
2.2 Defining End of Asset Life
Ultimately, the end of asset life occurs when the asset has to be replaced or is taken out of service.
However, given the maintenance options available, predicting when this event will occur is a complex
issue. Various definitions relevant to the end of asset life were compiled from Marlow et al. (2009) as
follows:
End of physical asset life, when the pipe is physically derelict and non-functioning.
End of technical service life, when the pipe is failing to provide required functionality,
service levels and/or reliability.
End of economic asset life, when the pipe is physically able to provide a service, but ceases
to be the lowest cost alternative to satisfy a particular level of service (Institute of Public
-------
Works Engineering Australia, 2006). In practice, this often reduces the time when the
risk/cost associated with retaining an asset exceeds the cost of rehabilitating the asset.
End of financial life, when the pipeline's initial capital value is fully depreciated.
Obsolescence, when the pipeline is obsolete because of changes in technology, regulatory
requirements, or performance criteria.
2.3 Defining Remaining Asset Life
The remaining asset life could then be defined as the time remaining until one of the "end of life"
criterion described above is reached. A more detailed assessment of remaining asset life could include
the following considerations from Marlow et al. (2009):
"Since an asset passes through a range of condition and/or performance states as it
deteriorates, condition and performance assessment can be used to understand
remaining asset life. In particular, acceptable asset condition states can be defined
that characterize the threshold above which risk is deemed to be unacceptable.
Condition assessment can then be used to determine whether the current state of an
asset, expressed in terms of failure likelihood, is acceptable. Acceptability criteria for
assessing degraded condition can be obtained from a number of sources including
the use of expert opinion, condition grading, detection of critical defects and
performance monitoring."
Based upon the condition of the pipeline, whether or not the asset can be renovated also has a significant
influence on the concept of remaining asset life. Loss of function does not necessarily imply the end of
asset life as pipelines are repairable or can be rehabilitated. Maintenance activities such as repair,
cleaning, and relining can remedy defects and restore condition, which will slow the overall deterioration
of an asset and extend its remaining asset life.
2.4 Assessing Asset Condition
A key consideration in asset management of water distribution and transmission systems is determining
current condition and predicting future condition. Condition can be assessed in terms of a number of
functions including:
1. Leakage
2. Hydraulic capacity
3. Water quality
4. Serviceability (e.g., defined as the capability of a system of assets to deliver a reference level
of service to customers and to the environment now and into the future)
5. Structural/physical integrity
6. Economic
In reality, these conditions are interrelated. For example, leakage can be an important contributor to
physical failure by eroding the pipeline bedding. While on the other hand, structural deterioration by
corrosion, such as a through-the-pipe-wall hole, can be a cause of leakage.
In the most basic of terms, the aim of condition assessment is to provide insight into the nature of possible
root causes of pipeline failure, the pattern of the pipeline deterioration curve, and the timing of possible
failure, which will determine its remaining asset life.
-------
2.5
Pipeline Functional Life
The functional life of a pipeline is illustrated in Figure 2-1 and takes into account the pipeline design,
construction, operation, maintenance, and ultimately the end of the pipeline's functional or service life.
The concept of a pipeline functional life is presented by Rose (2008) and is a similar concept to the
"service life" terminology presented by Marlow et al. (2009) and summarized in Section 2.1.
Performance parameters * \ure ^38.
-whauo monitor [ : rฑ:ฑr
s
Function
(It works)
\
\ \
Function
Defined by
performance
standards
V J
Functional
Failure
End state or
potential end
State,
Evidence
what you see
^ Failure ^
Cause
Contributing
& root causes;
reasons why
failure
occurred
V ^
1
Failure end state
* Failure consequences
Functional Failure "]
(Ir doesn't work)
^ Failure ^
Mode
Mechanism
of failure
^ >
' Failure ^
Behavior
^ J
Failure
Con sequent ex
Cost, safety,
environmental
V J
Figure 2-1. Functional Life of a Pipeline (after Rose, 2008)
A "bathtub" curve can also be used to illustrate this concept of the pipeline functional life. The bathtub
theory is a function of the probability of failure over time. The name "bathtub" comes from the shape of
the line commonly produced by the conditional failure probability curve overtime. The bathtub curve
generally consists of three periods: 1) a premature failure period with a decreasing failure rate, 2)
followed by a normal service life period with a low, relatively constant failure rate, and 3) concluding
with a wear-out period that exhibits an increasing failure rate. A graphical representation of the "bathtub'
process is illustrated in Figure 2-2.
cc
a
Decreasing
Failure
Rate
Early
' Failure
Constant
Failure
Rate
Observed Failure
Rate
Constant (Random)
Failures
Increasing
Failure
Rate
.* Wear Out
Failures
''...
'"
Time
Figure 2-2. The "Bathtub" Curve (Source: Wikipedia)
-------
2.6
Condition Assessment Tool
Condition curves are tools to graphically represent a pipeline's condition versus time. If the curve is
developed with sufficient data for the particular pipe type under consideration, then the pipeline's
functional life can be projected to predict its remaining life or when intervention will be required. The
condition curve will not be precise to the month or even year as too many factors are involved. However,
it is a useful tool to provide a priority rating based on likelihood of failure. There is a demonstrated need
for such tools to aid utilities in prioritizing and allocating their limited capital improvement funding and
to provide a technically defensible and repeatable process to evaluate pipeline structural condition and the
likelihood of failure overtime.
Condition curves usually are based on a horizontal axis representing the age of the pipeline starting with
the installation date and finishing at some estimated maximum life beyond the "design life." The vertical
axis represents the state, or sometimes the performance, of the pipe ranging from 100% when newly
installed to 0% when the asset reaches the end of its service life. Figure 2-3 illustrates the concept with
and without appropriate rehabilitation that can increase a pipeline's functional life.
The likelihood of failure increases as the pipeline ages and deteriorates. Given sufficient data, it is
possible to estimate the likelihood of failure for a given pipeline. However, in some cases, acquiring
sufficient data will be impractical or not cost effective; therefore, alternative approaches for asset
condition assessment should be employed. All of these alternative approaches involve defining condition
states based on beliefs about the types and levels of distress indicators that characterize increasing and
finally unacceptable levels of failure risk.
It is important to stress that condition curves are tools that come in a number of types and varying degrees
of sophistication. As such, the chosen tool must match the problem or the need. For example, the
problem may be a lack of data due to cost constraints for acquiring the data (e.g., asset not worth the
expense to employ inspection tools or to develop detailed data sets). In this case, more simplistic
condition curves should be chosen that can be applied using a minimum amount of data and resources to
construct. Section 3 provides a more detailed review of the types of condition curves, their development,
and applications.
Maintenance Maintenance Maintenance
Rehabilitate Rehabilitate
Figure 2-3. Condition Curves for a Pipeline with and without Rehabilitation
-------
2.7
Pipe Networks, Characteristics, and Behavior
To determine pipe condition, utilities need to have a good understanding of the pipe materials of
construction and their behavior over time under the anticipated environmental and operational conditions.
This basic understanding of pipe behavior provides the focus on key issues in the development and use of
condition curves.
2.7.1 The Distribution and Transmission Water Network. Estimates of the total current length
of water distribution and transmission pipelines in the U.S. exceed one million miles. In round figures, as
best as can be estimated, this mileage breaks down into the percentages shown in Table 2-1 (Stone et al.,
2002):
Table 2-1. U.S. Pipeline Mileage by Pipe Material and Pipe Diameter
Classifying by Pipe Material
Cast/spun Iron
Ductile Iron
Steel
Asbestos Cement
PVC
PCCP
Other
38%
23%
4%
15%
15%
3%
2%
Classifying by Pipe Diameter
Diameters of 10 in. and less
Diameters 12 to 24 in.
Diameters greater than 24 in.
73%
20%
7%
Research indicates that between 250,000 and 300,000 breaks occur every year in the U.S., which
corresponds to a rate of 25 to 30 breaks/100/miles/year (Grigg, 2007; Deb et al., 2002).
In undertaking condition assessment and remaining life prediction for all forms of pipe, it is important to
understand the initial form of failure (distress indicators), the pattern of their development (failure causes
and distribution), and the ultimate form of the failure (failure mode).
Distress indicators, the indicators that can be used to infer that a pipe is degrading or will
degrade.
Failure causes, the factors that contribute to the failure mode. There can be multiple paths
leading to a particular failure mode.
Failure mode, the ultimate form of the failure.
2.7.2 Ferrous Pipe. Ferrous type pipe materials account for around 65% of the U.S. water
network and consist of some of the oldest pipes in the network. By 1930, vertically cast iron pipe was
superseded by spun cast iron pipe. Since the mid-1960s, spun cast iron pipe began to be replaced in the
U.S. by ductile iron pipe. Spun cast iron use decreases substantially after the early 1970s, and production
ceased in the 1980s (US. EPA, 2002). The approximately 38% of the network that is pit cast and spun
cast iron pipe (Stone et al., 2002) is therefore at least 30 years old and most of it is much older. This
percentage is gradually decreasing due to pipe replacement. The oldest cast iron pipes, dating to the late
-------
1800s, have an average functional life of about 120 years. The average life of these older cast iron pipes
ranged from 90 to 150 years before they needed to be replaced. There is some evidence that pipes laid in
the post World War II boom have a reduced life averaging 75 years, which could be due to factors like the
reduced wall thickness compared to earlier specifications and/or to sub-standard site installation
(AWWA, 2001).
Forms, Causes, and Modes of Failures
The causes of failure in ferrous pipes are often complex and due to a combination of factors. Corrosion
causing pipe wall loss is a primary factor that is often combined with external and internal loadings,
which ultimately lead to pipe failure. Table 2-2 provides a summary of the main factors that lead to
pipeline failure for ferrous pipes.
Table 2-2. Summary of Factors Leading to Failure for Ferrous Pipes
Factor
Chemical
stressors
Physical
stressors
Other factors
Description
Internal and external corrosion caused by factors such as aggressive water or soil,
microbes, stray currents, oxygen gradients, and bi-metallic connections.
Damage during transport, unloading, storage and installation
Traffic loads
Soil loads from differential settlement caused by soil movement
Point loads (impingement)
Internal radial loads from internal pressure fluctuations
Axial loads from seismic activity, soil movement and water hammer
Thermal stress from temperature differences between water, pipe and soil
Damage by third parties - dig-ins
Damage to external coatings or internal linings
Aging - the accumulation over time of chemical and physical stressors
Pipe flaws - inadequacies in design, raw materials or manufacturing
Installation defects - incorrect bedding, backfill, jointing, encapsulation and coatings
Extensive literature provides a comprehensive review of forms, causes and modes of failure of steel, cast,
spun, and ductile iron water mains. The WaterRF report "Main Break Prediction, Prevention and
Control" (Grigg, 2007) devotes an entire chapter to reviewing this research. Additional sources of
literature for failure modes and mechanisms for ferrous pipe types include: "Investigation of Grey Cast
Iron Water Mains to Develop a Methodology for Estimating Service Life" (Rajani, 2000), "Long-Term
Performance of Ductile-Iron Pipe" (Project #3036; WaterRF, in progress a), "Fracture Failure of Large
Diameter Cast Iron Water Mains"(Project #4035; WaterRF, in progress b), and "Long-Term Performance
Prediction of Steel Pipe" (Project #4318; WaterRF, in progress c_).
In considering failure modes of ferrous pipe, they should be allocated into groups by material, diameter,
and age.
By far the greatest number of failures due to perforation or fracture occurs in pipe diameters
less than 10 inches. A U.K. survey of failures by pipe size and failure type showed that 77%
of failures were due to pipe fractures and 97% of these were circumferential for pipes with
diameters less than 10 inches. Some 14.3% of failures were due to pipe perforations with
94% of the failures in pipe diameters of less than 10 inches (U.K. Water Industry Research
[UKWIR], 2001).
-------
The main mechanism for smaller diameter failures is circumferential cracking.
Random and localized corrosion occurs in cast iron pipes and the failure of one pipe is no
indication as to the condition of adjacent pipes.
In terms of break frequency, Table 2-3 shows that the majority of pipe breaks occur in cast
iron pipes which, based on their manufacturing history, are usually much older.
Table 2-3. Ferrous Pipe Break Frequency/100 mi/yr
Source Cast Iron Ductile Iron Steel
NRC(1995)
Pelletier et al. (2003)
Weimer(2001)
58
93
44
15.5
32.5
5
n/a
n/a
53.5
n/a = not available
Vintage and pipe class determine the wall thickness. There has been a substantial reduction
in wall thickness over the years due to improvements in manufacture and material strength. It
is also common to find that in times past the cheapest, lowest class pipe available with the
thinnest wall would be installed. The corrosion rate for all ferrous pipes is considered to be
similar so that a thinner pipe will corrode and perforate quicker when subjected to the same
conditions.
The majority of water pipes are internally protected with cement mortar lining. It is
estimated by Ductile Iron Pipe Research Association ([DIPRA]) that 30% to 40% of ductile
iron pipe has been protected externally by polythene wrap. DIPRA has investigated the
current condition of some of the oldest installations, which were installed in the 1960s and
found no evidence of significant corrosion in corrosive soils (Horn, 2010). In addition, some
larger utilities have fitted or retrofitted ferrous transmission pipe with cathodic protection
(CP) systems. If installed correctly and maintained, these systems will greatly inhibit
corrosion.
The forms, causes and indicators of distress of cast and ductile iron are summarized in Table 2-4.
Table 2-4. Cast, Spun, and Ductile Iron - Forms, Causes, and Indicators of Failure
Form of Failure
Burst failure
Burst failure
Burst failure
Cause of Failure
External pitting and
graphitization
corrosion weakening
the pipe wall; often
combined with
induced strains.
Internal pitting and
graphitization
corrosion weakening
the pipe wall; often
combined with
induced strains.
Third party damage.
Indicators of Distress
Damaged protection - wall loss
from external pitting, graphitization
(hard to detect), leaks.
External loads, pressure variations.
Aggressive/polluted soils.
Galvanic/electrolytic conditions.
Damaged lining - wall loss from
internal pitting, graphitization (hard
to detect), leaks.
External loads, pressure variations.
Construction activity - impact
Comments
Corrosion is principal
contributor to failure.
Specific vintages; unprotected
pipe and diameters <8 in. are
more vulnerable.
Cathodic and external
protection mitigates.
Low pH water.
Unlined pipe mainly cast and
some spun.
Ductile iron pipe mainly
cement mortar lined.
Unpredictable.
10
-------
Table 2-4. Cast, Spun, and Ductile Iron - Forms, Causes, and Indicators of Failure (Continued)
Form of Failure
Structural failure
- circumferential
cracking
Structural failure
- longitudinal
cracking
Structural failure
-joint split and
sheared bells
Leaks
Cause of Failure
Thermal stresses, poor
support leading to
bending, internal
pressure, excess
external loads.
Internal transient
pressures, high
external loadings.
Excessive stress at
joint, fatigue, loss of
support.
Wall perforations,
cracks and defective
joints.
Indicators of Distress
damage to pipe or protection.
Circumferential cracks, loss of
bedding, joint movement, high
traffic loads, frost regions.
Longitudinal cracks, increasing
external/internal loads, frost
regions.
Leadite joints (cast iron), joint
rotation, leakage.
Wet areas, leak noise.
Comments
Most common cause of failure
in small diameters. Often
combination of loss of wall
through corrosion and
internal/external loads.
Mostly in diameters >12 in.
Often combination of loss of
wall through corrosion and
internal/external loads.
Can be due to manufacturing
and installation defects.
Many small leaks not detected.
Leaks contributory cause due
to loss of bedding.
2.7.3 Asbestos Cement Pipes. The bulk of AC pipes were installed for water mains initially in the
1930s and extensively in the 1950s and 1960s. The recognition of the health hazards in manufacture and
use of asbestos products led to a reduction in use and phasing out of their manufacture in 1983. It is
therefore likely that most AC pipe was installed 50 or more years ago. The degradation of AC pipe can
release asbestos fibers into the water supply and limits have been set on the amount of these free fibers
(U.S. EPA, 201 Ib).
Forms, Causes, and Modes of Failures
Failures can be classified into:
Corrosion failures: AC pipe corrosion failures are the result of the degradation by leaching
resulting in wall thinning, loss of strength, and through-the-wall holes. The process of
deterioration is through decomposition of hydrated silicates in the cement mortar brought
about by the leaching of calcium hydroxide. AC pipe is vulnerable to this form of attack
internally and externally.
Mechanical failures: AC pipe is classified as being rigid and susceptible to mechanical
failure. Poor handling and installation can damage the pipe. In service, higher external and
internal loads can lead to breaks. Pipes of diameters less than 8 in. have low beam strength
and are prone to circumferential failure. An NRC study showed that circumferential breaks
accounted for nearly 67% of all recorded AC pipe failures and longitudinal cracking some
10% of failures (Huy et al., 2010).
A New Zealand study based on 400 samples indicated that the average asset life for AC pipe increases
with diameter (Opus Consultants, 2001). AC pipes with diameters of 15 in. had an average life twice that
of AC pipes 4 in. in diameter (i.e., 80 to 85 yr. versus 35 to 40 yr.). Table 2-5 summarizes the forms,
causes, and distress indicators for AC pipe. Additional information on AC pipe will be reported in
"Long-Term Performance of Asbestos Cement Pipe" (Project # 4093; WaterRF, in progress d) ongoing
WaterRF project).
11
-------
Table 2-5. Asbestos Cement - Forms, Causes, and Indicators of Failure
Form of
Failure
Burst failure
Burst failure
Burst failure
Structural
failure -
Joint
Leaks
Cause of Failure
External leaching corrosion
with loss of wall strength.
Internal leaching corrosion
with loss of wall strength.
Excessive operating and
transient pressures.
Fatigue.
Third party damage.
Excessive stress at joint,
fatigue, leakage causing
loss of bedding.
Defective joints or through
wall corrosion.
Indicators of Distress
Aggressive soil/groundwater.
Damage to external coating.
Reduced wall thickness.
Softening of wall.
Circumferential cracks.
Longitudinal cracks.
Inadequate or excessive depth
of cover.
Low pH water.
Operational records.
Construction activity - impact
damage to pipe or protection.
Change in pipe alignment, joint
rotation, crack in external
diaper.
Wet areas, leak noise.
Comments
Circumferential breaks common in
small diameters.
Inadequate design, manufacture, or
installation.
Bending due to loss of support or
high operational loads.
Often a combination of loss of wall
strength with other defects
Longitudinal cracks due to low
hoop resistance or high operational
loads often combined with reduced
wall strength.
Unpredictable.
Can be due to manufacturing and
installation defects.
Many small leaks not detected.
Leaks contributory cause to failure
due to loss of bedding.
2.7.4 Prestressed Concrete Cylinder Pipes. PCCP is used for transmission mains with diameters
of 16 in. and larger and includes some of the largest transmission mains. Although PCCP is only around
3% of the total network length, it plays a critical role due to its large diameters and therefore greater
consequences of failure. There are two forms of PCCP:
Lined cylinder pipe manufactured in diameters from 16 in. to 60 in..
Embedded cylinder pipe manufactured in diameters from 48 in. and upwards.
Both forms use a steel internal cylinder wrapped with high strength stressed steel wire and then are coated
with a protective mortar coating.
Forms, Causes, and Modes of Failures
Wire Breaks: The most common defect for PCCP is wire breaks due to corrosion and/or hydrogen
embrittlement. Breaks release the core compression and give rise to a number of distress indicators which
can be identified by investigation. PCCP is designed to have a factor of safety against failure so some
wire breakage can be tolerated. In some cases, it takes more than 50% wire loss to result in a failure. The
early 1970s saw pipes manufactured with a Class IV wire, which has proved to be very prone to early
failure.
It has been found that the extent and speed of corrosion of the underlying steel wire is a direct function of
the quality of mortar and its application (Price et al, 1990). The extent and rapidity of corrosion of the
mortar are a function of:
12
-------
The quality of the mortar
Volume of permeable voids
Aggressiveness of the environment
Operational Factors: One of the major causes of PCCP failure, second to wire corrosion, is operating
conditions that greatly exceed the pipes' original design. Water hammer or surge pressure is one of the
more commonly encountered problems. It has been shown that total submersion of PCCP does not
necessarily increase the risk of corrosion, but that cycling from a wet to dry environment does
significantly increase corrosion risks.
Soil Considerations: Bianchetti (1993) reports the following as being favorable environments for
corrosion:
Soil acidity, pH less than 5
Sulfate >6,000 parts per million (ppm) or >2000 mg/L of sulfate ions
Magnesium >5 0,000 ppm
Groundwater with a negative Langelier Index
Table 2-6 summarizes the forms, causes, and distress indicators for PCCP. Additional information on
PCCP can be found in "Performance of Prestressed Concrete Pipe" (American Water Works Association
Research Foundation [AwwaRF], 1993), "Failure of Pre-Stressed Concrete Cylinder Pipe" (Romer et al,
2008), and several U.S. Bureau of Reclamation reports on PCCP failures (Travers, 1994; Hartwell, 1994;
Von Fay and Peabody, 1994; Uyeda et al., 1994).
Table 2-6. PCCP - Forms, Causes and Indicators of Failure
Form of
Failure
Burst
Failure
Burst failure
Burst failure
Structural
failure -
Joint
Leaks
Cause of Failure
Wire breaks beyond
critical number.
Excessive operating
or transient
pressures.
Greater cover depths
than design.
Fatigue.
Third party damage.
Excessive stress at
joint, fatigue,
leakage causing loss
of bedding.
Defective joints.
Indicators of Distress
Class III and IV wire.
Mortar coating - spalling, damage,
aggressive soil conditions, external
circumferential cracks.
Circumferential cracks to inner
core.
Out of roundness.
Delamination and hollow areas.
Galvanic/electrolytic conditions.
Low pH water.
Operational records.
Construction activity - impact
damage to mortar coating and pipe.
Change in pipe alignment, joint
rotation, crack in external diaper.
Wet areas, leak noise.
Comments
Class III and IV wire susceptible to
hydrogen embrittlement.
Mortar exposure to wet/dry ground
water conditions.
Frost loads.
Pre 1970 cast coating used which has
greater tendency to spall.
Shorting straps mitigate corrosion.
None.
Unpredictable.
Can also be due to manufacturing and
installation defects.
Many small leaks not detected. Leaks
contributory cause due to loss of
bedding and possible catalyst for
chemical attack in dry soils.
13
-------
2.7.5 Polyvinyl Chloride Pipes. Since their introduction in the 1950s, polyvinyl chloride (PVC)
pipes have established a major market share for new installations of small diameter mains (although
diameters up to 36 inches are available). The technically correct designation is PVC-U (unplasticized), as
there are other modified forms including PVC-M (manufactured by mass polymerization) and PVC-O
(molecularly oriented).
Forms, Causes and Modes of Failures
Failures in plastic pipelines take the following forms:
Leaks
Physical failure
- Chemical breakdown of the physical structure
- Crack growth mechanism
- Fatigue
- Buckling
Leaks: The evidence suggests that the majority of leaks in PVC pressure mains arise at joints and fittings
and is predominantly caused by improper or poor installation.
Physical failure: The cause of a physical failure is often not immediately clear from observation and thus
field descriptions can be misleading. As plastic pipes are not subject to aging by corrosion, modes of
failures are frequently related to inherent defects in the pipe, damage during installation, and impingement
damage. Closer examination and laboratory investigation indicate that, for many failures, there is a
combination of factors. Physical failure can also occur due to a chemical breakdown of the polymer
structure or by structural breakdown.
Most failures in the field are attributed to slow crack growth followed by brittle fracture rather than
ductile mechanisms. Cracks are initiated from concentrations in the pipe wall. These concentrations can
be defects built in during manufacture such as air bubbles, dust or particles or they can be created by
impingement on the wall from stones or sharp objects in the backfill surround.
Buckling (e.g., instability causing excessive deformations) is created by static and dynamic compressive
forces. Buckling behavior is affected by the interaction of the pipe and its surrounding soil. The relative
stiffness of the pipe and the surrounding soil and the external loading determine the deformation mode of
the pipe.
Although failure probability is related to age for PVC, the statistical evidence may not provide a sound
basis for extrapolating likelihood of failure. PVC was first used some 50 years ago and has been
upgraded and improved since.
Failure of PVC pipes is a complex subject and for a more in-depth understanding reference should be
made to "Plastic Pipe Systems: Failure Investigation and Diagnosis" (Farshad, 2006), "Long-Term
Performance Prediction of Polyethylene (PE) Pipe" (Davis et al., 2007), "Long-Term Performance
Prediction for PVC Pipe" (Burn et al., 2006), and "Evaluation of PVC Pipe Performance" (Moser and
Kellogg, 1994).
Table 2-7 summarizes the forms, causes and distress indicators for PVC pipe.
14
-------
Table 2-7. PVC Pipes - Forms, Causes, and Indicators of Failure
Form of
Failure
Burst failure
Burst failure
Burst failure
Structural
failure -
Joint
Leaks
Cause of Failure
Pipe splitting by slow
crack growth.
High pressure, pressure
changes, frequency and
external loadings in
excess of design.
Third party damage.
Excess stress at joint.
Defective joints.
Indicators of Distress
Manufacturing defect.
Scratch >10% wall
thickness.
Stones in backfill.
Operational records.
Construction activity -
impact damage to pipe or
protection.
Change in pipe alignment,
joint rotation.
Wet areas, leak noise.
Comments
PVC not subject to corrosion.
Manufacturing defects such as air bubbles
and inclusions can initiate crack growth.
Scratch resulting from transport or
installation mishandling.
Impingement from stones in backfill. The
vulnerability to failure due to these
defects increases due to creep as the pipe
ages.
The indicators of stress are not
identifiable by current site investigation.
Tapping can initiate and lead to splits.
PVC not suitable for frost loads.
Unpredictable.
Initial leak may create bedding loss.
Poor workmanship a frequent cause.
Many small leaks not detected. Leaks
contributory cause to failure due to loss of
bedding.
2.8
Key Issues for Condition Curves
It is important to understand the types of decisions that condition curves can be used to support and the
circumstances under which they are best applied. Some of the key issues that need to be considered in the
development and use of condition curves are as follows:
Condition curves have greater potential application in the prediction of condition and failure
for smaller diameter pipelines.
- For many pipelines with a relatively low replacement cost, it is not cost effective to
obtain direct condition data by investigation or laboratory testing.
- Smaller diameter pipelines account for the majority of the total length of the network and
also for the majority of failures.
- There is a much greater body of historical experience of failures and defects that can be
used in the development of condition curves and management of these parts of the
network.
- The direct and indirect consequences of failure are much less for small diameters.
- For small diameter pipelines (with low consequences of failure), condition curves can be
used as a means to monitor and keep the annual breakage rate below an acceptable
threshold through proactive renewal.
15
-------
Larger diameter pipelines represent a much smaller percentage of the network and failures are
much less frequent.
- The direct and indirect consequences of failure are much greater.
- The forms of failure and distress indicators tend to differ from smaller diameters.
- There are often insufficient historical data to develop the trend or shape of condition
curves for large diameter pipelines.
- Condition curves for larger diameter pipelines generally need to be generated based on
hard data from inspection and investigation or laboratory testing. This is due to the fact
that there is often minimal historical data for large diameter pipelines on which to base
models or curves and therefore field data are needed to verify the rate of deterioration.
- For large diameter pipelines (with high consequences of failure), condition curves can be
used to monitor trends in pre-failure indicators over time in order to provide a warning of
either an impending failure or an increased probability of failure. This enables a response
to be implemented in order to prevent a catastrophic failure.
It is important to understand how to develop condition curves for a given network.
- Condition curves are very site specific and a clear understanding is needed of the basic
assumptions that form the basis of the curve.
- Various pipeline assets may have differing rates of deterioration. Developing a single
condition curve that works for an entire network is unlikely. A condition curve scenario
should be generated for each pipe cohort of interest. The accuracy will be determined by
how well the pipe cohorts are grouped in terms of common factors.
- Curves will need to be generated for various scenarios and to take into account the
appropriate factors (e.g., environmental conditions, operational and maintenance
practices, pipe types, diameters, vintages, installation practices, etc.).
It is important to understand how to apply condition curves for improved decision-making.
- Condition curves can be used to evaluate and quantify current asset condition and to
predict future condition and remaining asset life.
- Condition curves can be used to support the development of short-term and long-term
maintenance program priorities and schedules. With sufficient data and understanding of
failure modes and causes of distress, a utility can effectively use condition curves to
support rehabilitation or replacement decisions.
- Condition curves can be used for long-term economic planning. For a utility just starting
an asset management program, the primary benefit of the use of condition curves is to
create a repeatable and technically defensible process for prioritizing and allocating
funding for capital improvement plans based upon remaining asset life.
16
-------
3.0: REVIEW OF METHODS USED TO GENERATE CONDITION CURVES AND TYPES OF
CONDITION CURVES
This section presents an overview of available literature on the methods used to generate condition curves
and the types of curves being used by utilities. The intent of this section is to provide guidance to utilities
on available condition assessment tools and where to find more detailed information for specific
applications.
3.1 Development of Condition Models
A great deal of work has been undertaken by researchers to understand the distress indicators and
mechanisms for the various types of pipe used in the water system. There is a large amount of
information on the numerous combinations of mechanisms leading to failure and in some cases
conflicting findings across research projects. This makes it difficult to synthesize this research into a
single methodology that has broad and practical application to the industry. It can be defined as a
multivariate statistical problem with uncertainty because of lack and variability of data on pipe condition,
operating conditions, and environmental conditions.
It is not possible to develop a universal, reliable condition/performance prediction model for use with all
types of pipes and conditions. Instead, various prediction models have been developed for specific
implementations of decision type, pipe material, diameter, vintage, structural design, environmental, and
operational factors. Such models are constrained by data availability and other factors.
As defined in the WERF report on "Remaining Asset Life: A State of the Art Review" (Marlow et al.,
2009), existing condition/performance prediction models can be classified under the following
approaches:
1. Deterministic models - where relationships between external factors and asset performance
are assumed to be certain. Deterministic models are relatively simple to develop and apply.
However, they usually rely on a number of simplifying assumptions. Furthermore,
deterministic models do not account for the uncertainty that is associated with asset
deterioration and failure.
2. Statistical models - based on analysis of historical failure rate or service lifetime and other
data. Statistical models attempt to capture this inherent uncertainty and use historical data
describing failure rates or service lifetimes in asset cohorts. Statistical models work for assets
where historical data are readily available for analysis. Bayesian analysis is a robust way of
supplementing historical data with beliefs (or opinions) concerning asset failure rates or
lifetimes, which are based on engineering knowledge or related observations.
3. Physical probabilistic models - based on an understanding of the physical processes that
lead to asset failure, while accounting for realistic uncertainty. These models are
underpinned by a robust understanding of the degradation and failure processes that occur for
an asset in service (corrosion, fracture, etc.). They also attempt to account for realistic
uncertainty by using appropriate probability distributions for model variables. However, they
can be data intensive and, in the event that insufficient data exist to adequately describe
model variables, simplifying assumptions are required.
4. Soft computing or artificial intelligence models - There are a number of approaches such
as neural networks where model structure is determined by the data and no prior relationships
17
-------
are assumed. Other models use complex mathematics, but are otherwise transparent and the
underlying methodologies and assumptions are fully published (e.g., fuzzy-based models).
In addition, Kleiner and Rajani have developed a comprehensive review of structural deterioration models
to predict the probability of water main failures that were presented in two papers, one on physical
(mechanistic) models (Rajani and Kleiner, 2001) and the other on statistical (empirical) models (Kleiner
and Rajani, 2001).
A key objective in asset management is to balance system performance and cost. The balance differs for
small distribution mains compared to large transmission mains, and this difference leads to different
forms of management for the two classes of assets. Figure 3-1 illustrates these differences qualitatively.
Cost
(present
value)
min.
cost
Failure frequency (#/km)
1234
-i 1-
Failure risk
Cost of
renewal
Time of renewal
Failure frequency (#/km)
Cost
(present
value}
min.
cost
Time of renewal
Figure 3-1. Optimal Renewal Frequency for Distribution Mains (Top) versus Transmission Mains
(bottom) (Rajani and Kleiner, 2002)
18
-------
As a pipe ages and deteriorates (without renewal), the likelihood of failure increases along with the risk.
The risk can be expressed as the present value (PV) of expected cost (or consequences) of failure. At the
same time, the discounted (or PV) renewal cost declines as pipe renewal is deferred. The total expected
life-cycle cost typically forms a convex shape, where the minimum point depicts the optimal time of
renewal (t*). The top part of Figure 3-1 illustrates this situation for small distribution mains, where the
cost of failure is relatively low. Therefore, the optimal time of renewal strategy allows a relatively higher
failure frequency. In contrast, the bottom part of Figure 3-1 illustrates this concept for large transmission
mains, where the cost of failure is typically high. Therefore, the optimal strategy is to avoid failure
altogether (i.e., failure prevention rather than failure frequency management). The time of renewal scales
are not the same for the two cases.
Currently, most nondestructive testing (NDT) technologies that can identify distress indicators are not
cost effective for small diameter water distribution mains (the exception being where the consequences of
failure are high). The predominant approach for assessing condition of small diameter distribution mains
is based on the observation of historical failure frequency. On the other hand, condition assessment of
transmission mains frequently requires the use of NDT to identify distress indicators due to the lack of
inferential indicators and the high consequences of failure. Some of the technologies intended to identify
inferential indicators (e.g., those related to soil properties) are used for both small and large pipes.
It is possible to balance maintenance expenditure against risk-cost such that the overall cost of asset
ownership is minimized. Importantly, increased reliability can be achieved through higher maintenance
expenditure with reduced capital expenditure. In financial terms, the increase in expenditure to achieve
this level of asset performance needs to be justified. While in practice it may be difficult to undertake
analysis to develop the type of curve shown in Figure 3-2, reducing risk to an economic/acceptable level
through a judicious combination of maintenance and renewal activity is still a viable approach to asset
management.
ฃ looo.
TOTAL IMPACT
Risk Exposure!
Not the optimum
this if wbซre CC55S art risks
are EQUAL, not mhtt* ft*
impict it itjit)
Preventive
Costs
2 3 4 5 G 7 9 9 10 11 12 13 14 15 IB IT IB 13 20
Ins pe c iionrMamten an ce'Re placement interval (montnti
Figure 3-2. Optimal Timing of Intervention (Woodhouse, 1999)
The above is a limited overview of condition assessment tools. A comprehensive review and evaluation
of tools can be found in "Remaining Asset Life: A State of the Art Review" (Marlow et al., 2009),
"Comprehensive Review of Structural Deterioration of Water Mains: Physically Based Models" (Rajani
and Kleiner, 2001), "Comprehensive Review of Structural Deterioration of Water Mains: Statistical
19
-------
Models (Kleiner and Rajani, 2001), and a U.S. EPA report prepared by NRC "Condition Assessment
Technologies for Water Transmission and Distribution Systems" (U.S. EPA, 201 la).
3.2 Types of Condition Curves
Developing condition-based and performance-based deterioration curves appropriate to the risk of failure,
benchmarking of deterioration curves to real-life asset failures, and the examination of the underlying
factors has generated a large body of literature. This subsection attempts to provide an overview with
references where more detailed information can be found.
Condition curves are of particular interest because of their usefulness for long-term water asset planning.
This subsection focuses on structural/physical condition curves rather than investment and renewal
curves, which are based on other considerations. For condition curves to be useful to a utility, they must
recognize the following:
Trend or shape of the failure or decay (or deterioration) curve,
Where on the curve is an asset's current condition, and
Asset's remaining useful life.
Moreover, condition curve predictions need to recognize that the original pipes may have been modified
by repair, rehabilitation, or replacement.
Types of condition curves that have been developed and applied by various utilities include:
Break frequency curve
Deterioration, decay, and survival curves
Condition rating curve and condition rating index
Serviceability curve
In addition, other types of curves are available, but are not the focus of this report:
Structural P-F curve: These to date have not been used by utilities for predicting failure of
pipelines. They find greatest use in situations where the time between where failure
indicators are noted (P) and failure (F) are short such as electric motors, tires, and other
applications.
Nessie curve: The prediction of annual replacement curves are also known as Nessie curves.
Nessie curves find significant use by utilities for forecasting capital expenditure needs, but
are based on the design life of the pipelines and not on structural condition or failure
considerations.
3.2.1 Break Frequency Curves. The most widely used statistical approach is the break frequency
curve. Most utilities keep records of pipe breaks, so that the basic data to populate a curve are available
and it can be applied to all types of pipes. The application of these data can range from simple direct
analysis of break rates (e.g., breaks per mile per year) to sophisticated mathematical models incorporating
a variety of factors (i.e., soil condition, operating conditions, pipe types, location, etc.) to predict
remaining life. This method is best suited to applications where there is an adequate amount of historic
break data over a sufficient time period on which to develop the curve(s). After the break frequency
curves have been developed, the results of the analysis can be used to facilitate proactive renewal by
replacing pipelines based on projected breakage rates that exceed an acceptable threshold.
20
-------
Break rates can be calculated for different pipe cohorts and used to prioritize replacement or
rehabilitation. Typically, a utility will determine a level of break where it becomes unacceptable on cost
or serviceability grounds. A WaterRF study of 20 cities found that 0.1 to 0.2 breaks per mile per year was
considered to be an "acceptable benchmark" (American Water Works Association [AWWA], 2001).
The most common method for generating break frequency curves is to allocate pipelines into relatively
homogenous groups or pipe cohorts based upon specific pipe characteristics (e.g., pipe type, pipe
diameter, age). The next step is to generate curves based on these characteristics. Several utilities also
collect valuable information on the nature of the pipe failure, which allows them to identify possible
systemic issues.
With this information, a utility can undertake a location or spatial analysis by plotting the location of the
breaks on a map to discern the patterns and the potential causes of breaks. If break data are collected over
a sufficiently long period of time, the information can be used as a temporal analysis to determine if there
are any factors other than age (such as weather conditions, soil conditions, loading factors, etc.) that are
impacting the break rate. The same approaches can be used to evaluate leak frequency.
An NRC review of condition assessment technologies for the U.S. EPA lists approximately 20 models
based on break frequency that have been developed over the last 30 years (U.S. EPA, 201 la). Out of the
12 utilities described in Section 4.0, nine of them made use of break frequency approaches within their
condition assessment programs. The discussion below provides several examples of the use of break
frequency curves as discussed in the literature.
Main Break Modeling
Grigg (2007) discussed the link between main break models and contributing risk factors (predictor
variables) in a WaterRF project. Examples of these predictor variables include pipe age (minor factor in
later stages of failure), material, diameter, soil corrosivity, and operational pressures that vary by utility.
It was found that different coefficients were required to model different scenarios and pipe variables to
estimate break frequency for each utility.
Grigg describes an approach to develop break predictions and prioritized segments for replacement
where, if the PV of a future break cost (consequence) is greater than the current cost of pipe replacement,
the pipe should be replaced. The model was applied and customized for two water utilities. The inputs
required are:
At least six years of coded data on all past breaks showing location, diameter, installation
year, material, type of break, and some other break qualifiers.
Categorical descriptions of mains mileage by diameter, installation year, and material.
Unit costs of several break consequences and replacement by diameter.
Grigg was not able to develop coefficients for variables that can be transferred from one utility to another
because of too much variation in data. Grigg found that main break modeling is complex and requires
trained modelers to develop and run models. In addition, internal data are used to calibrate the models
that require statistical expertise and data quality checks (Grigg, 2007).
21
-------
Pipe Class Analysis to Predict Main Breaks
Another example of the use of breakage rates and pipe categorization into groups is described by Sekuler
and Banciulescu (2009). This work was undertaken for the Metropolitan District in Hartford, CT. The
stages of development were:
Data management by developing a structured database
Pipe class analysis: 21 pipe classes were identified by pipe material, soil type, and pipe
diameter (breakage rates were plotted for each class)
Field sampling and testing (to obtain extent of corrosion and wall thickness for coupons)
Asset modeling and development of deterioration curves
Figure 3-3 illustrates six different break frequency curves for 4 to 6 in. cast iron pipes of different ages,
materials, and soil types. The graph also shows the condition index and the breakage rate level where
replacement is deemed to be cost effective. What is significant is that although 4 to 6 in. cast iron could
be considered as one class of pipe, the curves predict a range of replacement times from 63 to 133 yr.
depending on the soil type and date of installation for the pipe. This wide variation is attributed to the
fact that some of the older pipes had greater wall thicknesses at the time of installation compared to the
newer pipes.
4in - Sin Cast iron curves
o 30% Pipe Replacement
CD
3
a>
7T
CD
ca
(D
33
ฃB
CD
o
o
3
20
40
Age
Family 1 -Cast lron:<1925:4-6:Other Soil
Famity 2 -Cast lron:1925-1949:4-6:Other Soil
Family 4 -Cast Iron: 1925-1949:4-6 :Other Firw Soil
. Family 3 - Cast lfon:<1925:4-6:Fine Soil
Family 13-Cast lron:1950-19S9:<10
. Family 4-Cast lron:1959:4-6
Figure 3-3. Example of Break Frequency Curves by Pipe Class (Sekuler and Banciulescu, 2009)
22
-------
Regression Models
Regression models have proved to be a popular approach with modelers. They identify relationships
between cumulative historical breakage patterns in space and time. The assumption is that the patterns
will continue into the future, which allows for the forecasting of break rates (breaks per distance at some
year). Researchers applied linear and exponential regression techniques to obtain a relationship for the
breakage rate of a pipe as a function of time. Based on the costs associated with pipe repairs and
forecasted breakage rates, an economic break-even analysis can be developed to determine the optimal
year of pipe replacement.
Distribution - Water Mains Renewal Planner (D-WARP)
A development has been the NRC's D-WARP model, which analyzes the deterioration of water
distribution pipe cohorts (in terms of the increase of their breakage rates) as an exponential function of
age. The analysis of water main breakage patterns takes into consideration time-dependent factors such
as temperature, soil moisture and rainfall deficit, and CP strategies. D-WARP allows the user to see the
"optimal" time of pipe replacement, as well as to generate, examine, and compare complex scenarios that
include combinations of replacement and CP strategies. D-WARP is currently a standalone program,
available for free download at the NRC Web site (Kleiner and Rajani, 2004).
Individual Water Main Renewal Planner (I-WARP) Model
The most recent development from NRC is I-WARP, which uses a statistical method to provide an
effective way to estimate deterioration for individual pipelines or pipe cohorts. The model takes into
account both dynamic and static factors. I-WARP requires inventory and breakage data about individual
pipes. Pipes are divided into homogenous groups - material, diameter, vintage - or any other grouping for
which data are available. The model is calibrated to discern historical breakage patterns for each group
providing group-specific parameters, which can be used to forecast future breaks. A minimum of 5 years
of break data are required. The model allows the consideration of time-dependent factors such as
temperature, soil moisture and rainfall deficit, CP strategies, as well as user defined qualitative and
quantitative factors (e.g., changes in operational conditions, leak-detection campaigns, etc.). I-WARP is
currently a software prototype, available through WaterRF (Kleiner and Rajani, 2009).
3.2.2 Deterioration, Decay and Survival Curves. Deterioration, decay, and survival are different
names for curves that mirror the decay process and the deterioration of the pipe. These curves are
primarily used for ferrous pipe, PCCP, and AC pipe where a time-dependent factor, such as corrosion, is a
primary cause of failure. These types of curves are not typically appropriate for polymer pipes, which are
more vulnerable to failures from inclusions, scratches, or impingements.
Such curves are developed in a number of ways using a range of information and levels of acceptability.
These types of curves are designed to provide estimates of:
Trend and/or pattern of the deterioration
The current condition of the pipeline on the curve
A prediction of the remaining life of the pipeline.
Direct inspection of the pipeline with measurement of defects will provide data on the progress of the
deterioration and allow a curve to be plotted. Alternatively, laboratory testing and analysis of test
specimens taken from the pipeline will provide condition data. As the pipe diameter increases and the
consequences of failure increase, the use of inspection to provide data becomes more cost effective.
23
-------
The last 30 years has seen the development of many models although not many have become established
as everyday tools used by the water industry. Section 4.0 provides four examples of utilities (Las Vegas
Valley Water District [LVVWD], Sydney Water, PWD, and the Hartford Metropolitan District) that have
reported using some form of deterioration, decay, and survival curves. The discussion below provides
several examples of the use of deterioration, decay, or survival curves as discussed in the literature.
Deterministic Models
Deterministic models aim to predict corrosion rates and to estimate the remaining wall thickness and,
consequently, the service life of the pipe. The deterministic models typically use two or three parameter
equations to model pipe breakage patterns. These models are best applied to pipe cohorts or groups of
water mains that are relatively homogeneous with respect to factors that might influence their breakage
patterns. These models are relatively simple to apply, but require careful consideration of water main
grouping schemes.
For ferrous pipe, the type of data that are required for different methods is similar, including pipe age, soil
parameters, wall thickness, and current extent of corrosion.
The corrosion pit measurement can be acquired using NOT techniques or by examining exhumed pipe
samples. Assessment based on pit depth measurement has limitations. It requires an assumption that
measurement or number of pits at one location will be representative of the whole pipeline. For structural
failure, it requires a significant grouping of pits to reduce the wall strength. The widely used assumption
that failure occurs when through wall penetration occurs is overly conservative in terms of ongoing
service life particularly with cast iron pipes.
Probabilistic Models
Probabilistic models are designed to calculate the probability of the survival of the pipeline over a certain
period of time, predict the remaining life time, or estimate the probability of failure. The main difference
between the probabilistic and deterministic models is that probabilistic models incorporate an uncertainty
component, which is ignored in the deterministic models. Probabilistic models are used where direct data
on condition are not available, limited or not cost effective to obtain. Probabilistic models use explicit
assumptions about the probability distribution of the modeled event.
The probabilistic "multi-variate" models can consider many of the factors that influence breakage
patterns, thus reducing the need for the model operator to partition the water mains into homogeneous
groups. The advantage is that the model can incorporate multiple factors such as environmental
conditions (e.g., soil corrosivity), operational stress factors, pipe age at certain breakage rates, number of
previous breaks in pipe, period of installation, and other relevant covariates. These models provide
outputs such as the instantaneous rate of failure (e.g., hazard function) or estimates of the probability of
the time duration between consecutive breaks. These models require significant technical expertise and
sufficient data available that cover multiple variables (Ugarelli and Bruaset, 2010).
The probabilistic "single-variate" group-processing models include models that use probabilistic
processes on grouped data to derive probabilities of pipeline life expectancy, probability of breakage, and
probabilistic analysis of break frequency.
24
-------
T-WARP
Transmission - Water Mains Renewal Planner (T-WARP) models the deterioration of larger diameter
mains using a so-called fuzzy rule based on the Markov deterioration process. It requires that the pipe be
investigated at least once to establish the current condition rating, while future pipe deterioration is
predicted through the analysis of failure likelihood. The owner is required to rate the consequences of
failure on a fuzzy scale. Given likelihood and consequences of failure, a fuzzy risk of failure is computed
and a rehabilitation strategy formulated. This program developed by Kleiner et al. (2006) is available
through WaterRF.
CARE-W
Computer Aided Rehabilitation of Water Networks (CARE-W) is a suite of tools developed in Europe by
a collaborative research effort. It contains several independent decision support tools, including CARE-
W Fail, which has five different modules to forecast pipe failure. These five modules predict failures
based upon statistical, probabilistic, or physical means. The tools include the: (1) Markov model, based
on Asset-mapl (Malandain et al., 1999); (2) Poisson model, based on Asset-map2 (Malandain et al.,
1999); (3) Proportional Hazard Model (PHM), based on Failnet-Stat (Le Gat and Eisenbeis, 2000); (4)
UTILNETS (Hadzilacos et al., 2000); and (5) Non-Homogeneous Poisson Process (NHPP) model, based
on Winroc (Rostum, 2000; Eisenbeis et al., 2002). LWWD uses CARE-W software combined with the
Casses software to evaluate its water pipe condition for renewal planning as discussed in Section 4.0.
Safety Factor Curves
The safety factor concept allows for a quantitative comparison of the anticipated stresses on a pipe and its
residual strength. One approach is the safety factor curve, which allows for an assessment of the residual
strength as the wall thins from deterioration over time. A safety factor of "1" indicates that the pipe is
likely to fail. For large PCCP transmission mains, this approach has been successfully used to prioritize
which individual pipes need to be replaced or rehabilitated. As discussed in Section 2, the main distress
indicator for PCCP is wire breaks that can be determined from inspection techniques and from these the
safety factor curve can be developed.
An example of the safety factor approach is provided in Deb et al. (2002) where a mechanistic model for
cast iron pipe was developed involving four modules:
Pipe Load Module - the loads to which the pipe is subjected
Pipe Deterioration Module - the corrosion process and loss of strength
Statistical Correlation Module - calculates the residual strength based upon the reduced wall
strength
Pipe Break Module - compares the stresses on the pipe and the residual strength. The ratio
between these represents a safety factor.
3.2.3 Condition Rating Curves and Condition Rating Indices. Because of the difficulties and cost
of modeling, some utilities use systems based on judgment, expert opinion, and/or performance indicator
data to determine criticality and assign priorities. Although it is reliant upon expert opinion, the condition
rating curve or condition rating index approach does provide a logical and documented framework for
determining pipeline renewal priorities. This approach is best applied when the deterioration can be
classified into discrete states to define a structural condition rating or score (such as the ratings from
25
-------
excellent to very poor shown in Figure 3-4). This rating or score can then be assigned to corresponding
action levels such as no action, increased frequency of inspection, repair, rehabilitate, or replace).
This approach is also well suited to incorporating consequence of failure ratings, as well as likelihood of
failure ratings, to give an estimate of the overall failure risk. Another advantage of such an approach is
that local conditions can easily be incorporated. Figure 3-4 graphically shows an example of the
condition rating curve approach. It should be noted that the condition rating output does not have to be
developed into a curve, but can be implemented on a score or rating index basis only.
This type of curve is widely used for wastewater and stormwater gravity systems, but is less common for
water networks. WERF has a project "Best Practices in Water Infrastructure Asset Management" based
on condition ratings for all types of utility assets including pipelines (Bhagwan, 2009)
Using a formalized approach, it is possible to develop relative criticality based on expert judgment
combined with weightings of defects to arrive at a condition rating. This approach is used in prioritizing
gravity pipelines using mainly information from closed-circuit television inspections, historical, and
environmental data. Methodologies developed by the Water Research Center (WRc) and National
Association of Sewer Service Companies provide the ratings for evaluating the condition (WRc, 2003).
Decay curve after
rehabilitation
o
o
O
FAILURE
Maintain
Nominated minimum
service standard
Rehabilitate
100%
% Useful life
Figure 3-4. Typical Condition Curve Based on Broad Performance Indicators (DERM, 2001)
A few examples of this approach as applied to water mains and/or pressure mains are provided below. As
discussed in Section 4.0, several utilities employ a condition rating curve or condition rating index
approach including EPCOR Water Services, Seattle Public Utilities (SPU), Sydney Water, Washington
Suburban Sanitary Commission (WSSC), and Louisville Water Company (LWC).
26
-------
Bayesian Belief Networks
The application of this approach to pressure mains is more complex because of the wide range of
variables. One approach for pressure mains is to use what have been termed "Tendency Trees" or "Belief
Networks" (Thomson, 2010). This method can be used when failure or investigation data are limited.
They use physical, operational and environmental factors that have a bearing on the likelihood of failure.
Each factor is assessed on a given scale indicating its condition. Each factor is given a relative weighting
relating to its importance in the failure process. Weightings are based on observation, data analysis, and
expert opinion and are usually specific to a particular utility.
This approach is a structured application of expert knowledge and experience using available information
and observation. Many engineers and operators use this approach instinctively in making their
assessments. A basic version of this approach has been developed for small to medium utilities that have
need for a tool that can be used by their own staff. The technique is used to provide a rating both for
likelihood and consequence of failure and an example of the overall rating approach is shown in Figure 3-
5. For both likelihood and consequences of failure detailed "Belief Networks" are developed to provide
the inputs for this summary network to reach a criticality rating.
Belief Network - Evaluation of Likelihood and Consequences
Consequences of failure
Internal
corrosion
Figure 3-5. Overall Belief Network - Likelihood and Consequences (Thomson, 2010)
27
-------
FARMS-PRIORITY
FARMS is a suite of computer applications based on models that have been developed by CSIRO of
Australia. FARMS-PRIORITY is a commercially available decision support system designed to
prioritize pipeline renewal, mainly for low to medium cost failures. The software uses asset and failure
data to develop deterioration curves through statistical analysis and the use of physical/probabilistic
models. The curves take into consideration factors such as the pipe age, material type, pipe diameter,
operating pressure, length of pipe, pipe failure history, and soil conditions. The FARMS-PRIORITY
model is primarily based on five key tasks: risk calculation, failure prediction, cost assessment, data
exploration (asset and failure records), and scenario evaluation. In Section 4.0, the use of a risk score or
matrix along with the FARMS model by Sydney Water is described. This risk matrix assigns a rating
based on the consequence of failure defined in monetary terms (cost of pipe renewal, customers affected
by the loss of water supply, etc.), and the probability of a failure as calculated using the FARMS model
output. More information on model parameters for FARMS can be found in Moglia et al. (2006).
Condition Rating Approaches for AC Water Mains
Failures for AC water mains can be broadly divided into corrosion and mechanical failures. Corrosion
failures are the result of the deterioration of the pipe wall resulting in wall thinning and through holes.
AC degradation can occur at both the internal and external surfaces of a pipe and if sufficiently rapid, can
decrease residual strength to a level where structural integrity is lost before the design life of a pipe is
reached.
One method of approximately representing degradation and creating a condition curve is to empirically
determine the rate of decrease in tensile strength (e.g., assuming the rate of decrease in tensile strength is
constant over time). Since NDT technologies for assessing the condition of AC pipes are currently
unavailable, degradation rates are measured using small coupons extracted from the pipe wall. A second
approach to measuring the amount of wall degradation is treating core samples with phenolphthalein,
which gives rise to a marked change in color between the degraded and non-degraded portions that can be
measured.
"Condition Assessment of an Asbestos Cement Pipeline" (Ojdrovic et al., 2007) describes a condition
study of 16.5 mi. of 12 and 14 in. diameter AC transmission line to determine if the pipeline could
continue to provide service under proposed higher flow conditions. The work involved analysis of
available data, structural evaluation, determination of corrosivity of soil/groundwater and internal water.
Samples of the pipe were laboratory tested by three methods: (1) edge bearing test, (2) flexural test and
(3) petrographic examination.
Condition Rating Approaches for Plastic Water Mains
A number of researchers have developed approaches for predicting failure in plastic water mains. Some of
this work is set out below. However, to date no viable cost-effective approach has been adopted by
utilities.
For plastic materials such as PVC, only limited failure data are available. Lifetime prediction for brittle
polymers is largely based on the linear elastic fracture mechanics (LEFM) theory. LEFM uses the
concept of a single parameter known as the applied stress intensity factor (SIF), which characterizes the
stress field at the tip of a crack in a plastic material. The dependence of crack growth rate on the applied
SIF can be determined using small laboratory scale tests on coupon samples, which can then be applied to
the geometry of a flawed asset in service via a geometrical correction factor.
28
-------
A WaterRF study (Burn et al., 2005) provides an in-depth review on the causes of failure and concludes
that manufacturing and installation practices may be the largest factor in determining the likelihood of
failure. The report sets out tests which concentrated on brittle and ductile failure and from this develops a
fracture mechanics failure model to predict time of failure under combined internal pressure and
deflection loads. This is then further developed with the use of Monte Carlo simulation models to predict
the performance of PVC pipelines.
Farshad (2006) developed a methodology based on an expert system for failure diagnosis of plastic pipe.
His system uses communication between the user and a software system. The system has two levels: a
core program that is a knowledge base and a user interaction level where the user can apply his/her own
experience and know how.
There is no literature describing use of these approaches by utilities.
3.2.4 Serviceability/Performance Methods. An NRC study "Measuring and Improving
Performance" (NRC, 1995) states "performance was the degree to which infrastructure provided the
services that the community expects of the infrastructure and can be defined as a function of
effectiveness, reliability, and cost." The primary function of a distribution system is to reliably deliver a
sufficient quantity of good quality water under adequate pressures to its customers (Deb, 1994). Deb
suggests that there is no single performance indicator that will define the performance of a distribution
system and proposes four performance indicators that should be analyzed over time: adequacy,
dependability, efficiency, and quality. These four indicators essentially define pipeline "serviceability."
There is a need to address deterioration modeling in terms of service, and not just asset deterioration.
Aging water pipes present other problems including decreasing hydraulic capacity, degradation of water
quality, increasing customer complaints, and increased liability resulting from direct and indirect
economic consequences of service disruption. The concept of serviceability rather than failure is gaining
favor in a number of countries and particularly in the U.K. where serviceability methods have operated
successfully for several years covering water and wastewater services for around 53 million people.
"Serviceability" is defined as the capability of a system of assets to deliver a reference level of service to
customers and to the environment now and into the future. It is on this basis that the utilities are judged
by the U.K. Water Services Regulation Authority (Office of Water [OFWAT], 2000). The basic
assumption underlying the approach is that a water network's life can be extended infinitely if properly
maintained (including replacement).
OFWAT requires that water mains and sewers be maintained in perpetuity. In practice, this means that
network (pipe) assets are not depreciated in the same way as discrete assets (such as pumps, tanks or
treatment works). Instead, an infrastructure renewal charge that reflects the level of investment needed to
maintain the network of assets is calculated.
One consequence of this approach is that the concept of "remaining life" is not explicitly used for
network assets. Furthermore, OFWAT does not accept investment planning approaches that are based
solely on estimations of remaining asset life. For example, a recent OFWAT ruling was that a particular
utility should reduce its replacement proposal of 1% mains replacement with only 0.3% mains
replacement per year.
The key performance indicators that OFWAT uses to decide if a utility is maintaining serviceability to its
customers and which are used as the basis of curves are:
29
-------
Number of bursts
Assessment of extent of unacceptably low pressure
Scale of interruptions
Water quality compliance
Although break frequency as described earlier is a widely used basis for pipe replacement, in this case, it
is also a measure of the scale of interruptions.
The utilities' charges to the consumer are regulated by OFWAT so gaining approval is fundamental to
management and provides every incentive to maintain and improve the key indicators. OFWAT collects
standardized data from all companies on an annual basis and for 5-year license reviews in order to assess
the structural condition and performance of the networks over time. OFWAT examines overall trends for
the performance indicators to determine whether the renewal activities carried out by the utility have
resulted in stable, improving, or deteriorating services to customers. It assigns each company a combined
performance score based upon the performance indicators summarized above. Collecting these data on an
annual basis allows OFWAT to track the overall structural condition of the U.K. network over time and to
determine the percentage of the network that remains in poor condition. It is important to assess trends in
structural condition overtime to determine if operation/maintenance and rehabilitation practices are
having a positive impact on system costs and overall serviceability in meeting customer expectations.
Interestingly, all of these performance indicators have shown an improving trend over the last 15 years
since the serviceability concept has been applied and tracked (Stone et al., 2002).
The OFWAT serviceability approach is a unique model that has been in wide use for a number of years.
It has been successfully used by a number of different utilities serving large populations over many years.
It is worth noting that it also takes a less pessimistic approach to remaining service life, which has
resulted in a reduction in replacement capital expenditures. The U.S. EPA report "Decision-Support
Tools for Predicting the Performance of Water Distribution and Wastewater Collection Systems" (Stone
et al., 2002) provides a fuller review of the OFWAT approach. This report also provides a large amount
of useful information on the use of performance indicators in Europe.
The CARE-W suite of software has already been noted. One of the modules CARE-W PI is used to
estimate the current and future condition of a water network against a range of key performance
indicators. It is based on the International Water Association list of performance indicators. There are a
total of 49 performance indicators in five groups, including operational, quality of service, financial,
water resources, and physical indicators. It is noted that 153 single pieces of utility information are
required to assess the 49 performance indicators. In addition, 29 external indicators, not under utility
control, such as climate, soil, and topography, are considered in the evaluation (Batista and Alegre, 2002).
3.2.5 Economic Models - Prediction of Annual Replacement Curves. Economic models to
predict annual pipe replacement are not condition curves as they are based on theoretical life expectancy
or design life and not on structural condition. Therefore, the details on how these curves are generated are
outside the scope of this review. However, as economic models that are often closely allied with
condition curves in asset management, a brief overview is provided.
Economic models are used to determine the present worth of future pipe repair and eventual replacement
costs. An example of such a model is the Nessie curve, which is an aggregate prediction of replacement
capital needs projected over time to forecast reinvestment needs. The humps in the cumulative
reinvestment shown in a Nessie curve are due to the echo effect where pipe reinvestment needs mirror, at
a projected future date, the original installation date of pipes. The name comes from the belief that the
30
-------
curve with its humps and troughs resembles the legendary Loch Ness monster. A typical curve showing
the echo effect of pipe construction is shown in Figure 3-6.
The decision-making process for determining the end of economic life is developed with varying degrees
of complexity in a paper by Buckland and Hastings (2001). Some researchers have attempted to
incorporate condition predictions into the decision-making process. To forecast the number of breaks in
future years these models are augmented with either a regression or probability-based predictive model.
This information is then used to perform a break-even analysis in which a total cost curve is derived as
the sum of the pipe repair and replacement curves. The low point of the total cost curve represents the
optimal replacement time (Grablutz and Hanneken, 2000).
10,000
Water Supply
50 Year Replacement Profile
3.3
Figure 3-6. Asset Replacement Cost Profile (DERM, 2001)
Structural Condition Curve Selection Factors, Benefits, and Limitations
The intended use of a condition curve should be to develop a repeatable and technically defensible
process for prioritizing the renewal of water mains based upon an evaluation of the structural
condition/performance of the pipeline or pipe cohorts in a given network. Any structural
condition/performance evaluation process should ideally possess the following characteristics:
Applicability the model is technically feasible and applies to the decision(s) and pipe
scenarios of interest;
Repeatability the same condition rating will be determined if a different person performs
the rating;
Objectivity the rating can be measured on the basis of some physical characteristics such
as cracking;
Simplicity simple systems are easier to understand and use; and
Cost effectiveness it is economically feasible to gather the data and the value of the
information provided (in terms of improved decision-making) exceeds the cost to the utility to
gather the data.
31
-------
The selection of tools for developing structural condition curves will depend upon the level of asset
management sophistication particularly in relation to data, as well as the capacity of the tools themselves.
Additional considerations in tool selection are listed below:
1. Data requirements: Are there high levels of data requirements?
2. Inclusion of expert opinion: Does the tool allow expert opinion to be utilized?
3. Treatment of uncertainty: Does the tool explicitly consider uncertainty?
4. Treatment of economic risk factors: Does the tool consider economic risk factors, or does it
simply consider physical asset life?
5. Blackbox: Does the tool provide a framework within which to understand the problem or is it
more of a blackbox approach, providing answers with no specific insights into the
mechanisms of the decision-making?
6. Availability: Is the tool available packaged and off the shelf, while allowing for
customization to meet utility needs and site-specific conditions?
7. Population or single asset: Does the tool apply to a single asset, population, or both?
8. Scenario testing: Does the tool provide the capacity to test different scenarios and thus
understand the results of different strategies?
The approach of utilities can be guided by the stages of network development. Utilities with relatively
young networks need to develop good practices in the early life-cycle stages of asset management and
specifically in data collection. Utilities with aging networks will be more concerned with rehabilitation
planning and delivery. This is likely to call for more comprehensive approaches for determining
remaining asset life.
3.3.1 Benefits of Condition Curves
Condition curves can contribute to better prioritization, scheduling, and funding of inspection
and renewal activities, which in turn helps to meet service goals and reduces failures.
Ultimately, this could lead to an improved asset management program to help to reduce
premature replacement, to enable better financial planning and rate determinations, and to
avoid large maintenance backlogs.
Condition curves provide a systematic and repeatable process to improve the decision-making
process for water main renewal based upon remaining asset life. The process should be
transparent and technically defensible, so that utility stakeholders are confident in the
usefulness of the results.
Improving the understanding of risk can help utilities to anticipate main breaks and reduce
their direct costs, as well as those suffered by the public. Utilities need to have this
information to build reliable cases for renewal and obtain the required funding. By using
condition curves appropriate to the situation, utilities become more aware of risk factors and
any information from new breaks adds to their understanding. In-house employees can be
trained to use this information in main break programs.
The selection of the appropriate rehabilitation/proactive failure management technique
depends on the cost-benefit ratio. The cost is represented by the cost of inspection,
assessment, and renewal, and the benefit is proportional to the losses avoided by prevention
of failure. The AWWA 2002 Water Utility Distribution Survey (AWWA, 2004) found that
the average reported cost per main break was $1,320, although other studies indicate this
32
-------
figure may be higher. Assuming this average cost is reasonably accurate, this would favor -
especially for small diameter, low risk pipes where the cost of inspection may not be
justified - less expensive statistical or expert-opinion -based approaches for predicting pipe
failures and developing renewal prioritization mechanisms.
The cost of investigation and building physical models is more easily justified for the large
transmission pipelines where the consequences of the failure can be large. Gaewski and
Blaha (2007) determined the cost of failure for larger diameter pipes (20 inches and greater)
is much greater in direct and indirect costs than it is for distribution network pipelines. The
average mean cost for 30 large diameter failures was $500,000 split about evenly between
direct and indirect costs.
The Governmental Accounting Standards Board requires reporting of all major capital assets
in the utility's financial statement using the historical cost approach or the approved modified
approach. Certain benefits to a utility are available by using the modified approach. Because
the modified approach defines the level of service to be maintained and maintains the value
of assets on the utilities' balance sheets, there is an improvement in the standing of the utility
with the customers, regulators and financial community. The utility can expect to have a
better bond rating and get more favorable bond rates for its capital improvement projects if
the modified approach is used. For the modified approach, condition assessment of assets
must be performed every three years. Statistical approaches, such as structural condition
curves, can be used to fulfill this requirement. The modified approach also requires
development of a complete inventory of existing assets. In addition, the condition of the
assets should be documented and a program established to maintain the condition of the
assets at the desired or specified performance level (Nelson, 2005).
3.3.2 Limitations of Condition Curves
Value is often the limiting factor when it comes to selection of failure management strategies.
Water supply utilities do not usually have a budget that would allow large investment and
therefore less advanced techniques are chosen. Utilities need to have a clear understanding of
the investment and effort required to implement the various methods described in this report.
Information on the cost to develop and implement these curves/models is limited and not
widely available in the literature as it resides with individual utilities and/or consultants.
Although the information exists, it is difficult to document as noted in the survey work
discussed later in Section 4.0.
Without an adequate level of data, models cannot operate successfully. It can be very costly
to collect and manage the level of data that are required.
The cost of acquiring pipe condition information that is necessary for some models must be
balanced against the critical nature of the pipeline. Proactive failure management is
approached in different ways for transmission and distribution systems.
For distribution systems, the need is for less expensive options that are likely to have a lower
performance standard. The relatively low cost associated with "fail and fix" for small
diameter pipes in low risk situations make it difficult to justify significant investment in data
collection and more sophisticated modeling. The greater volume of basic data for breaks in
small diameters makes simpler break frequency approaches the most feasible.
For some models, the underlying assumptions in terms of factors and their weighting are not
clear and may not be applicable to all utilities. One size does not fit all. Validity of methods
33
-------
depends on validity of data. Even with valid data there is a need for the right objectives,
criteria, alternatives and the cost benefit in what is being proposed.
It is recognized that many models are mathematically complex and will require special skills
to operate. Reluctance on the part of utilities to use "black box" models is acknowledged and
even understandable. However, the complexity that underlies more transparent models is not
(and should not be) a deterrent for its use. Further efforts are needed for utilities to become
familiar with and confident in the output of these models.
Some may consider these approaches as an imprecise science, which miss some failures and
replace some pipe that still has an economic life. Despite modeling advances, the application
of models is constrained by limited below-ground data and the complexity caused by the
interaction between factors that contribute to water pipe failure (Grigg, 2004).
34
-------
4.0: CURRENT USE OF CONDITION CURVES
This section provides an overview of current utility practices related to condition curves for water mains
through examination of published case studies and results from a survey of nine major utilities conducted
by Virginia Tech.
4.1 Who is Using Condition Curves?
As discussed in the previous sections, numerous methods, including various forms of condition curves,
are used by water utilities to assess the condition of their water transmission and distribution systems.
The utilities that were surveyed for these case studies used methods that ranged from very detailed asset
management programs that combine inspection, monitoring, and test data with their pipeline condition
assessment program to simple analyses of pipe break history to prioritize pipeline renewal activities.
Few utilities surveyed are using deterioration, decay, and survival curves likely due to the more extensive
data required to generate meaningful curves for the range of pipe types and diameters. In addition, there
is a need to clearly demonstrate a net benefit for the utility in order to warrant the extra cost of the data
collection efforts. However, there has been increasing recognition that development of quality data is a
major element in managing water networks. Such data have to be structured using a standard system for
naming and defining water utility assets, which subsequently allows the data to be shared and compared
across utilities. The WaterRF and WERF have a project underway related to standardizing terminology
which could be a major advance specifically for improving condition curve implementation (WaterRF
Project #4187).
4.1.1 Types of Condition Curves Being Used by Utilities. Each utility has developed its own
methodology for tracking and analyzing the condition of its network over time based on available data,
resources, and ultimate goals of the utility. Condition curve methods that were found to be in use include
pipe break frequency curves, condition rating curves or condition rating indices, and economic
forecasting models, such as Nessie curves. The only commonality found was that many utilities use pipe
break frequency as a main indicator for pipe renewal decisions. This finding is supported by a U.S. EPA
report (Stone et al., 2002) and a WaterRF report (Grigg, 2007), which indicated that the number of pipe
breaks is one of the most commonly used indicators of water pipe effectiveness and that the break rate is
the most important factor for prioritizing renewal activities.
4.1.2 Primary Pipe Types and Sizes for Condition Curves. The "condition curves" generated by
each utility focus on the pipe types that represent a majority in their water network. For those utilities that
participated in the survey, the focus of their condition assessment programs was primarily on ductile iron
and cast iron with slightly less focus on AC, PCCP, and plastic pipe types. Development of condition
curves for these pipe types (especially ductile iron, cast iron, and PCCP) can be somewhat easier because
utilities have a fairly good understanding of the failure modes and mechanisms and can collect data on
failure indicators. Plastic pipes, on the other hand, have very different failure mechanisms that are far
more difficult to detect, making it complex to develop accurate condition curves and failure predictions.
Focus is also placed on smaller diameter distribution mains as these dominate both the total water pipeline
mileage and the number of failures. Because of the significant quantity of pipeline mileage, utilities need
methods to prioritize their renewal activities so that the most critical infrastructure is replaced in a timely
and cost-effective manner. Therefore, break frequency curves combined with economic models appear to
be the most common type of analysis conducted. These methods are easier and less costly to develop, yet
provide sufficient information to do a fair job at prioritizing renewal of distribution pipes.
35
-------
4.2 Case Studies from Various Utilities
To better understand who is using condition curves and how they are being used, utilities with significant
activities in water pipe infrastructure management in the U.S., Australia, and Canada were contacted by
Virginia Tech to participate in a survey. Only nine utilities were selected due to restrictions of the
Paperwork Reduction Act. All participating utilities provided detailed information related to current
practices/programs for pipe inspection, condition assessment, and renewal planning. In addition, several
papers were reviewed that detailed case studies of how utilities are managing water pipe assets with
specific emphasis on the use of condition curves. All sources of data have been combined into this
section to provide a general overview of who is using condition curves and for what purpose in an effort
to provide guidance to users of this document and to highlight potential gaps and research needs.
4.2.1 Development of Survey. A survey of the nine utilities was conducted by Virginia Tech to
identify the factors that influence water main deterioration and state-of-the-art in condition
curves/deterioration modeling. Focus was placed on each utility's practices related to the use of pipe
deterioration models for generating pipe condition curves. The type of information requested included:
Types of inspection and condition assessment techniques used;
Prioritization of inspection, maintenance, and renewal;
Methods used for condition deterioration prediction;
Methods used to generate condition curves;
Factors included within the condition curves;
Software used;
Associated costs in generating condition curves; and
Type of pipe condition and/or performance index.
4.2.2 Participating Utilities. A total of nine utilities across U.S., Canada, and Australia
participated in the survey, providing an overview of the current best practices available (see Table 4-1 for
the list of utilities). These utilities were selected because they are fairly sizeable utilities representing a
range of pipe types and diameters and they expressed a willingness to provide information for this
research. To supplement the information obtained from the nine utility surveys, additional case studies
were reviewed, including:
The Metropolitan District - Hartford, CT (Sekuler and Banciulescu, 2009)
The Los Angeles Department of Water and Power (LADWP, 2010)
U.K. Water Services Regulation Authority OFWAT (discussed in Section 3.2.4)
4.2.3 Water Pipe Inventories for Case Study Utilities. The nine case study utilities operate a
total of over 32,000 miles of water pipe. The largest of these utilities is Sydney Water, operating nearly
13,000 miles of water pipe. The pipe is overwhelmingly <21 inches in diameter. There is considerable
variance in the pipe types used, but cast iron and ductile iron have the greatest length for most utilities. A
summary of the water pipe inventories for each utility is provided in Table 4-1.
4.2.4 Inspection, Monitoring and Condition Assessment Programs. Each of the utilities that
participated in the survey indicated varying degrees of inspection, monitoring, and condition evaluations
for their water transmission and distribution systems. Some utilities, such as EPCOR in Edmonton,
Alberta, Canada and LWC, responded to the survey with greater detail on their condition assessment
programs. These two utilities have comprehensive programs that generate a large amount of data for
managing the condition of their assets, including techniques such as CP, hydraulic modeling, geographic
information system (GIS) integration, soil and/or pipe sampling and testing, and non-invasive inspection.
36
-------
Table 4-1. Water Pipe Inventories for Case Study Utilities
Utility
EPCOR Water
Services Inc.
Las Vegas Valley
Water District
(LWWD)
Newport News
Waterworks
Seattle Public
Utilities (SPU)
Sydney Water
Washington
Suburban Sanitary
Commission
(WSSC)
City of Hamilton
Public Works
Department
Pipe
Diameter
(inches)
<21
21 to 36
>36
<21
21 to 36
>36
<21
21 to 36
>36
<21
21 to 36
>36
<6
6 to 12
12 to 24
>24
<16
>16
<18
>18
Percentage of Pipe Miles
Asbestos
Cement
(AC)
33%
1%
40%
6%
0.4%
0.5%
1%
0.4%
0.0%
Concrete
Cylinder'1'
0%
40%
70%
0%
27%
13%
25%
Concrete'2'
0.3%
46%
93%
4%
6%
0.0%
0.1%
0.4%
0.3%
Cast
Iron
(CI)(3)
22%
1%
2%
1%
33%
31%
83%
31%
1%
66%
64%
67%
17%
51%
43%
-%8
-%8
Ductile
Iron
(DI)(4)
0.1%
2%
12%
0.4%
55%
22%
6%
14%
8%
2%
22%
24%
23%
6%
49%
28%
-%(8)
-%(8)
Galvanized
Iron
7%
2%
0.0%
1%
pyC(5)
43%
18%
55%
5%
4%
0%
10%
7%
Steel<6)
2%
35%
30%
0.4%
75%
99%
1%
1%
28%
77%
0.2%
2%
8%
74%
3%
Other<7)
0.1%
5%
0%
0%
1%
0%
1%
1%
1%
2%
1%
3%
Total
(miles)
2,044
110
33
4,190
293
227
1,680
89
81
1,600
95
159
9,658
1,871
1,054
413
3,899
821
1,079
96
-------
Table 4-1. Water Pipe Inventories for Case Study Utilities (Continued)
Utility
Louisville Water
Company
(LWC)(9)
Philadelphia
Water Department
(PWD)
Pipe
Diameter
(inches)
Unknown
<6
6 to 12
12 to 20
24 to 30
36 to 48
60
3 to 93
Percentage of Pipe Miles
Asbestos
Cement
(AC)
4%
2%
0.9%
Concrete
Cylinder'1'
0.9%
0.4%
Concrete'2'
0.1%
73%
49%
99%
0.2%
Cast
Iron
(CI)(3)
5%
9%
42%
21%
17%
40%
0.1%
81%
Ductile
Iron
(DI)(4)
1%
5%
27%
72%
10%
8%
16%
Galvanized
Iron
0.4%
pyC(5)
58%
75%
28%
1%
Steel<6)
0.2%
3%
Other<7)
36%
7%
1%
5%
0.1%
2%
0.1%
Total
(miles)
o
6
200
3,503
213
89
45
37
3,278
OJ
oo
(1) The concrete cylinder category can include concrete cylinder and PCCP.
(2) The concrete category can include concrete and reinforced concrete pipe.
(3) The cast iron category can include cast iron, cement lined cast iron pipe, and unlined cast iron.
(4) The ductile iron category can include ductile iron and cement lined ductile iron pipe.
(5) The PVC category can include plastic, mPolyvinyl chloride, oPolyvinyl chloride, and uPolyvinyl chloride pipe
(6) The steel category can include steel, mortar lined and coated steel pipe, galvanized steel, lock bar steel, riveted steel, stainless steel, welded steel, wrapped
steel, cement lined steel.
(7) The other category can include fiberglass, high-density polyethylene (PE), copper, Kalamein, iron, wood stave, FL Bar, glass reinforced pipe, PE, vitrified
clay, wrought iron.
(8) Actual percentages were not provided. The information that was received indicated that pipe types owned and operated by the City of Hamilton Public
Works Department are predominately cast iron and ductile iron.
(9) LWC has approximately 105 miles of PCCP; however the GIS system on which the numbers in Table 4-1 are based did not note whether or not it was
PCCP.
-------
Several utilities use some of the same techniques as EPCOR and LWC, but reported their condition
assessment activities on a more limited scale in their survey response. For example, the LVVWD
primarily assesses AC pipes greater than 4 inches and steel pipes greater than 12 inches in diameter using
non-invasive leak and condition assessment technologies, CP for steel pipes, and forensic analyses of
failed pipes. Newport News also varies its condition assessment techniques based on pipe diameter.
Sydney Water prioritizes inspection and condition assessment of critical water mains using a quantified
risk-based prioritization tool. Sydney Water mainly focuses on the condition assessment of cast iron
water mains using non-invasive inspection techniques such as linear polarization resistance, magnetic flux
leakage, and ultrasonics.
Similarly, WSSC selects transmission main condition assessment methods based on pipe material type
and diameter. The techniques used include internal visual/sounding inspection (PCCP only), NDT
techniques (remote field eddy current, sonic/ultrasonic pulse echo), acoustic fiber optics (PCCP only), and
electrochemical potential surveys. It does not actively inspect distribution mains with the use of NDT
tools, but prioritizes renewal activities by analyzing work order histories, pipe physical properties, defect
types, and soil corrosivity.
Some utilities surveyed, such as SPU, do not conduct routine inspections of transmission and distribution
water mains. Rather, SPU uses a combination of CP and spot checks during maintenance to assess the
condition of its transmission mains in combination with leak and break data as indicators of structural
condition. During the spot checks, they record data for several variables including physical pipe
properties (material, joint type, wall thickness, diameter, etc.), soil properties, and pipe condition. The
only time internal inspections are performed is when the pipe is out of service. SPU did have a condition
assessment program (primarily consisting of pipe sample collection during tapping or repairs) for several
years in the 1990s; however, it was discontinued because costs exceeded the value of the information
obtained.
4.3 Survey Results: State-of-the-Practice in Pipe Condition Curves for Renewal
Prioritization
4.3.1 EPCOR Water Services Inc. EPCOR Water Services Inc. is a corporatized public utility
located in Edmonton, Alberta, Canada. EPCOR prioritizes its water main renewal activities through both
a reactive and proactive renewal program. EPCOR's reactive and proactive renewal programs primarily
use break frequency and a condition rating index using a GIS platform to prioritize renewal activities.
EPCOR has also experimented using other deterioration models including the development of pipe failure
prediction curves to allocate capital funds and artificial neural networks (ANN) to predict upcoming
spikes in main breaks, which was intended as a tool to assist in arranging staff schedules and minimizing
overtime. However, EPCOR found that efficiently modeling failure prediction curves to allocate capital
funds was largely unsuccessful and it was unable to utilize the ANN to predict main breaks with ample
notice. In the end, EPCOR plans to continue using its reactive and proactive renewal programs rather
than further pursue the development of deterioration type condition curves.
4.3.1.1 Reactive Water Pipe Renewal Program (Break Frequency Curves). EPCOR's reactive
renewal program began in 1985. The "reactive" terminology used here refers to the fact that EPCOR is
reacting to the increasing frequency of breaks along a given pipeline (this is the utility's own terminology
and somewhat different than the traditional "reactive" approach of only repairing a main after it is
unserviceable and replacing it). The pipeline sections for replacement are identified using a GIS
application designed to calculate break frequencies for candidate pipeline sections between valves. The
break frequency is classified by the replacement priority value (RPV), which is calculated by the total
39
-------
number of main breaks over five years divided by the total length of the pipe section between the valves.
Pipeline candidates with frequencies of five or more breaks/km/year are identified for replacement. Used
in conjunction with the main criteria, alternate criteria that were evaluated by EPCOR for pipe
replacement decision-making included replacement at a minimum of at least two breaks in five years,
more than six breaks in five years, and 12 or more breaks since 1982. In general, the deteriorating
pipelines requiring replacement are replaced with PVC pipe.
4.3.1.2 Proactive Water Pipe Renewal Program (Condition Rating Index). The proactive renewal
program started in 2002 and was designed to target water mains that do not qualify for reactive renewal
based on break frequencies, but are performing below EPCOR's design standards (e.g., fire flow, water
quality, etc.). The first part of the proactive renewal program consists of area prioritization through
evaluation of pipe condition, hydraulic deficiencies, and water quality from several different data sources
all imported into the GIS database. Specific data used to rank areas include pipe material, pipe length,
flow, unidirectional flushing frequency, leak and/or break history, available flow, and hydrant data (i.e.,
hydrants with long flushing times and total number of hydrants). Current conditions, standards, and/or
common industry practices influence the threshold values for each criterion. An example of EPCOR's
area prioritization system for cast iron pipe is shown in Figure 4-1 in which 45% of the pipe renewal
score is based on structural factors, 30% is based on hydraulic factors, and 25% is based on water quality.
The total score for a particular pipe type is compared against threshold values set by EPCOR to prioritize
renewal of specific areas. An example of the area ranking threshold values for CI pipe and a graphical
representation of the area prioritization is shown in Figure 4-2. Using this information, EPCOR engineers
pinpoint critical areas for more detailed evaluation.
Candidate prioritization is then used to down-select areas to one pipeline section. EPCOR establishes
candidate criteria rankings for pipeline sections by evaluating condition/break history, demographics,
hydraulics, water quality, and economies of scale. EPCOR uses sensitivity analysis to determine the
candidate weightings for the various pipe types.
Condition/ Material/
Maintenance History
Components
(weight 45%)
%Castlron(CI)
(weight 15%)
Rank
>80% 5
>60% 4
>40% 3
>20% 2
Hydraulic Components
(weight 30%)
Low Flow
Water Quality (WQ)
Components
(weight 25%)
%100mmCI
(weight 5%)
Rank
Breaks/km/Syr
(weight 15%)
Rank
Brks/km/5yr> 0.0025 5
Brks/km/5yr> 0.002 4
Brks/km/5yr> 0.0015 3
Brks/km/5yr> 0.001 2
Hydroscope
(weight 10%)
Rank
Avg Holes per 10m > 15 5
Avg Holes per 10m > 10 4
Avg Holes per 10m > 7 3
Avg Holes per 10m > 4 2
Else 1
Figure 4-1. Example of Area Criteria Ranking and Weights for Cast Iron Water Pipe
40
-------
Area Cl Total Score Rank
> 60 points 5
>45 points 4 i__p*j_ii 1-1
-*"-"' Map Legeud
i i-^ *^r ^^^^^^
> 30 points 3 _
D ? (41)
> 15 points 2
Else 1 ~1J I OJ
Figure 4-2. Area Ranking Threshold Values and Graphical Prioritization for
Cast Iron Water Pipe
4.3.1.3 Validation of Condition Assessment Models and Associated Costs for Generating Models.
EPCOR's main focus is minimizing the impacts and response times to breaks, improving tools for
selecting candidate pipes for renewal, and reducing the construction impact during renewal. Even though
validation of its program with respect to predictive effectiveness has not been a main focus, EPCOR did
evaluate the RPV renewal qualification criteria and found that if a pipe was not renewed once the renewal
criteria were reached, its break rate would increase.
EPCOR estimated that it spends approximately $50,000 to $100,000 per year on pipe inspections, data
collection, data management, modeling software, and interpreting results to identify pipes at risk.
4.3.2 Las Vegas Valley Water District. The LVVWD is a public utility located in Las Vegas,
Nevada. LVVWD uses CARE-W and Casses software modules to analyze and prioritize renewal of its
water infrastructure system.
4.3.2.1 CARE-W and Casses Software (Break Frequency, Deterioration, and Economic Models).
LVVWD uses the CARE-W software combined with the Casses software to evaluate its water pipe
condition for renewal planning (see Figure 4-3). CARE-W contains modules for estimating the current
and future condition of water networks and includes routines for estimating long-term planning and
investment needs as well as selection and ranking of annual rehabilitation projects (using the CARE-W
ARP tool). These tools are integrated and operated jointly in the CARE-W software.
The Casses software (based on the LEYP model) is then used to analyze pipe physical and environmental
data together with break history to predict pipe break rates. LEYP can assign a probability of failure to a
pipe based on time (Weibul module), failure factors (Cox module), and past breaks (Yule module).
LVVWD only uses the Casses model for predicting break failure rates for AC pipes since there is an
insufficient amount of break data for steel pipes to obtain reasonable results. Instead, LVVWD uses
corrosion data and the potential consequence of failure (e.g., business disruption) to prioritize renewal
activities for steel water mains.
More information on the CARE-W1 and Casses2 software packages are provided on their Web sites.
1 SINTEF project Web site: http://www.sintef.no/Projectweb/CARE-W/Project-Results/ARP-Multicriteria-decision-
of-rehab-projects-/.
2 Cemagref Casses Web site: https://casses.cemagref.fr/.
41
-------
CARE-W Data Flowchart
PRIMARY DATA
INTERMEDIATE DATA
CARE-W MODULES
For display on digital map, data originated from another source could be imported to GIS (at the zone level)
ฉ
*
ฉ
Data: water
resources, physical
condition, financial
factors, operations,
quality of service.
Manually entered in
the PI databases.
i
Main Failure Data
(from Avantis & GIS)
i
GIS Pipe Data
(H2O Map)
ฉ
(ID Features tor pr
zone to be evaluat
Request trom Plan
months in advance
Utility databases (pipe
environment, surface
type, population
supplied, etc)
Data: Unit costs,
budget restrictions,
leaks, degradation,
inflation rates,
replacement
scenarios, etc.
OJ
3
Slat
_ O
@^ ^
^ Sum
O
ority
2dOnly.
n ng 2-3
.)
ฎ
ฉ;,
1
ฉ1 1ฉ
|ปl Pipes.csv* pปซ
sties
ly
nary
iy
ฉ
i*cess PI (Performance Indicators) $ป
[city/zone level] Jaซ
Output helps managers decide which zone to act upon
first. Also keeps track of past actions and consequences.
^ j
PI-UI-EI WPlMDBl
1^ Fail (Failure Forcasting): LEYP** ^
[pipe level] e"&>e
City
1 Generates table of Break Rate predictions for each pipe.
The program (Casses) uses the LEYP engine. Requires
USB Protection Key to Calibrate data. Licensed.
^ J j
KSP file |
RelNet (Hydraulic Criticality) Plsnni-ng
[pipe level] rtjr!sftis
Ouput data (HCI) goes to ARP database.
Program has a limit of 1 0,000 pipes.
^ ARP (Annual Rehabilitation Planning)
[pipe level] 0/7e^0
Output helps manager decide which pipes to replace first
(given a set budget).
^ ^
rules, weights, hotspots, etc I ฉ
Color-coded overlay of high-, medium-, and low-threat pipes
ฉ>
|oe\ฐw LTP (Long Term Planning): Kanew***
[system level]
Session results allow managers to decide on the
sequence of replacement and how much money should be
assigned for next budget. Licensed.
kadata.gdb I
ฉ
* Fail's pipe data input file includes deleted and abandoned pipes.
** Fail's engine was formerly PHM.
*** Lip's components are Kanew (the primary application), WRaP (Weighing & Ranking Procedure), and Scenario Writer (seldom used).
File: Data Flow v1.4.vsd
Date September 1, 2010
Figure 4-3. CARE-W Data Flowchart for LVVWD
42
-------
4.3.2.2 Validation of Condition Assessment Models and Associated Costs for Generating Models.
The LVVWD has not performed an evaluation of the CARE-W software results primarily because there is
not sufficient pipe break data to be statistically valid; however, LVVWD is confident that the models are
practical. LVVWD has performed back-casting on the Casses break prediction model that indicated the
software produces reasonable results. Some limitations of these models identified by the LVVWD
include the need to acquire reliable data and costly in-house analyses.
Data collection and management, modeling, and identifying pipes at risk are done in house by LVVWD
personnel while some assessment work is performed by specialized consultants as needed. LVVWD did
incur the one-time cost for implementing the CARE-W program; however, there are no annual fees for
continued use of the software. If any software/programs are needed in the future, these costs would be
included or accounted for as part of LVVWD's operating budget.
4.3.3 Newport News Waterworks. Newport News Waterworks is a public utility located in
Newport News, Virginia. Prioritization of renewal activities at Newport News Waterworks varies based
on pipe diameter. Small diameter pipes (2 to 4 inches) are replaced based on failure rates using a
prioritization program focused on the economics of repair versus replacement. For pipes 6 to 12 inches in
diameter, replacement is based primarily on their failure rate, as well as experience with similar pipe
materials. Large diameter pipes (>12 inches) are replaced when there is a known propensity for
manufacturing and/or installation issues (e.g., pre-1985 unwrapped ductile iron pipe is often
recommended for early replacement in areas of corrosive soil). To date, Newport News has not
experienced a significant number of failures in large diameter pipes since most have been installed within
the last 40 years.
Historically, Newport News used a point system (condition rating index) in combination with economic
models (Nessie curves) to analyze water pipe condition and prioritize renewal activities. However, the
point system has evolved over time to primarily account for break frequency as described below. The
focus of the discussion below is on Newport News' pipe replacement prioritization program, which
assigns points based on pipe breaks and their potential costs.
4.3.3.1 Pipe Replacement Prioritization Program (Break Frequency). Newport News' pipe
replacement prioritization program was established in the 1980s and has been updated and revised over
time. The initial program was a point system. A score was assigned to pipe sections based on evaluations
of 10 different pipe criteria: break history, pipe size, pipe depth, grid pattern in area, pipe material, soil
corrosiveness, system pressure, available fire flow, work in area by others, and cost comparison between
repair and replace/rehabilitation.
Over time, Newport News found that many of these pipe categories did not affect the priority ranking.
After several iterations, the replacement prioritization program evolved into what it is today, which
involves prioritizing pipe replacement based only on the number of pipe breaks, pipe life expectancy
(based on Nessie curves), and maintenance costs. In particular, up to five points are assigned for each
pipe break. Points assigned for life expectancy are based on the equation: % of life reached/25. Points
assigned for cost are based on the formula: (100 * cost of breaks)/(cost to replace). The Arc View GIS
program stores this information and is used to identify projects and to update the priority rankings.
4.3.3.2 Validation of Condition Assessment Models and Associated Costs for Generating Models.
Newport News has not performed statistical analyses of its models. The Nessie curve analysis was
performed by an outside consultant for an estimated cost of $20,000 to $40,000. However, Newport
News reports spending in the magnitude of millions of dollars to create and maintain its customized
geodatabase for asset management, which supplies the pipe data upon which their prioritization
43
-------
program/point system is based. All associated costs have been included within Newport News' annual
budget.
4.3.4 Seattle Public Utilities. The SPU is a private utility located in Seattle, Washington. The
SPU prioritizes water transmission main renewal based on failure probabilities for critical pipes generated
by a Water Main Replacement model combined with economic forecasts from the Wave Rider model.
The definition of "critical" pipes is still being refined by SPU but includes factors that lead to increased
risk (e.g., transmission mains under body of water, railroad track, hospital, pipe material, soil corrosivity,
leak history, etc.). The methods used to generate the models for critical pipe failure probability and
economic forecasting are statistical and empirical and were developed in house based on industry
accepted best practices for infrastructure asset management.
For distribution pipes, a prioritization process is used that compares the expected future cost of repair to
the life-cycle cost of replacement to determine whether distribution pipe leaks/breaks should be repaired
or replaced. This method is the same for all distribution pipes regardless of the type of material, size,
and/or location.
4.3.4.1 Wave Rider (Economic Model). The SPU's economic model, Wave Rider, is used for long-
term capital planning. Wave Rider forecasts renewal expenditures by year for nine pipe classes (ductile
iron, cast iron divided into four subcategories by size and vintage, steel, concrete, galvanized, and other)
and is based on Weibull distribution curves for each class. The model results have been compared and
calibrated to the actual break history/repair rate data (collected since 1990). Although Wave Rider helps
SPU forecast renewal expenditures for groups of water pipes, it does not address individual pipes.
4.3.4.2 Water Main Replacement Model (Condition Rating Index). The SPU also uses a Water
Main Replacement Model for individual transmission pipe renewal decisions. It is primarily a cost-
benefit model that compares the net present value (NPV) of the replacement cost to the potential
economic and social disruption costs of a failure event. The analytical framework for the model is the
same across pipe materials, sizes, locations, and other parameters and is based primarily on a Weibull
distribution curve generated from the leak history data of a given pipe.
The model is used to determine whether or not the water pipe is at the end of its economic life and should
be replaced. The pipe has reached the end of its economic life when the replacement cost of a new pipe is
lower than the "marginal risk cost" of the pipe failure. The marginal risk cost is defined as the product of
the probability of failure and the total repair cost. The probability of failure is determined by leakage rate
data collected by SPU for the individual pipe. As shown in Table 4-2, the total repair cost consists of
construction costs plus various social costs associated with service loss, traffic disruption, lost water,
property damage, fire risk, and water quality issues. The output of the model, shown in Table 4-3,
includes the NPV of a replacement, total leaks per year over time, and a break-out of costs for repair and
replacement options.
4.3.4.3 Validation of Condition Assessment Models and Associated Costs for Generating Models.
No statistical analysis has been completed to evaluate the validity of the Water Main Replacement Model.
It has been difficult for SPU to validate the predictive effectiveness of the failure curves since the
majority of its pipes have remained in the flat part of the curve. If and when SPU's break rates begin to
rise, it may need to re-assess its current pipe replacement program. SPU does not document or keep track
of the costs of these activities. In a 2008 estimate, SPU spent approximately $43,000 on asset data,
decision models, and related support. In general, most of the work in generating the failure curves is
conducted in house, with the occasional use of consultants. SPU also did not incur the costs for
purchasing specialized software because the analysis relies upon software applications already in place
such as Microsoftฎ Excel and Arc View GIS.
44
-------
Table 4-2. SPU Input Parameters Included in the Water Main Replacement Model
Option
Leak Repair
Replacement
Data Class
Pipe
Construction
Service
Traffic
Lost Water
Damage
Fire Risk
Water Quality
Pipe
Construction
Service
Traffic
Water Quality
Benefits
Input Variables
Pipe Length Miles
Leaks per Mile per Year in Year 1
Pipe Age
Leak Repair Hours
Persons per Repair
Cost per Person per Hour
Equipment Pieces per Repair
Cost per Equipment Piece per Hour
Material Cost
Total Cost per Leak
Hours Service Interruption per Leak
Customers Impacted per Leak
% Leak Repairs w/ Water Shutoff
Cost per Customer per Hour
Hours Traffic Interruption
Traffic Flow Cars per Hour
Cost per Car
Hours of Water Loss per Leak
Gallons Lost per Hour
Cost per Gallon Lost
Number of Damage Claims per Leak
Settlement Cost per Claim
Customers Impacted Fire Flow
Property Value per Customer
Probability each Year Fire w/ Inadequate Fire
Flow
Damage % Property Value
Customers Impacted Low Water Quality
Cost per Customer per Leak Low Water
Quality
New Pipe Economic Life Years
Replacement Construction Cost per Foot
Hours Service Interruption During
Construction
Customers Impacted Construction
Cost per Customer per Hour
Hours/Project Traffic Interrupt During
Construction
Feet per Project
Traffic Flow Cars per Hour
Cost per Car
Customers Impacted Water Quality
Construction
Cost per Customer Low Water Quality
Customers Gain Improved Service Levels
Annual Benefit per Customer Improved
Service
Input Values
0.088
11.4
60.0
5
3
$50
3
$75
$625
$ 2,500
3
15
50%
$5
5
40
$2
168
25
$ 0.002
0.167
$ 2,000
15
$500,000
0.00001
100%
15
$25
175
$300
5
15
$5
72
300
40
$2
15
$10
15
$-
45
-------
Table 4-3. Outputs of the Water Main Replacement Model
Outputs
Net Present Value of Replacement @ 3%
Net Present Value of Replacement @ 5%
Net Present Value of Replacement @ 7%
Total Leaks per Year in Year 1
Total Leaks per Year in Year 20
Construction % of Leak Repair Option Cost
Service % of Leak Repair Option Cost
Traffic % of Leak Repair Option Cost
Lost Water % of Leak Repair Option Cost
Damage % of Leak Repair Option Cost
Fire Risk % of Leak Repair Option Cost
Water Quality % of Leak Repair Option Cost
Construction % of Replacement Cost
Service % of Replacement Cost
Traffic % of Replacement Cost
Water Quality % of Replacement Cost
Value
$48,298
$276
$34,570
1.00
4.00
66%
3%
11%
0%
9%
1%
10%
94%
0%
6%
0%
4.3.5 Sydney Water. Sydney Water is a public utility located in Sydney, Australia. Sydney Water
has used KANEW and is currently implementing the FARMS model in cooperation with the Water
Service Association of Australia as described below.
4.3.5.1 KANEW (Deterioration, Decay, and Survival Curves). KANEW is a cohort survival model
for infrastructure assets. KANEW predicts when selected pipe sections will reach the end of their service
lives, differentiated by date of installation and by type of pipe sections with distinctive life spans.
The current version of the KANEW software includes a module to manage pipe inventory, a module to
perform the cohort survival calculations, a failure and break forecasting module, a module to perform cost
calculations, a module to support decision-making by running and comparing various scenarios, an
economic data module, and a strategy comparison module. Sydney Water uses KANEW to analyze its
long-term capital investments. KANEW analyzes data according to the year of pipe installation or
rehabilitation against the pipes' aging behavior. KANEW is limited in that the failure curves produced
represent a cohort of pipes and not an individual pipeline asset. Furthermore, Sydney Water feels that
there is no explicit relationship between the assets' performance versus the deterioration curve. Sydney
Water has concluded that KANEW is not suitable by itself for its critical water mains because it does not
take risk into account.
4.3.5.2 Pipeline Asset and Risk Management System (Deterioration, Decay, and Survival Curves
and Condition Rating Index). Working with the Water Service Association of Australia, Sydney Water
is currently implementing the FARMS3 software to manage its water assets. FARMS is a suite of
computer applications based on models that have been developed by CSIRO of Australia. Currently, two
FARMS applications are publicly available as commercial products (Marlow et al., 2007). PARMS-
Planning forecasts the number of pipe failures and assesses cost implications of various high-level, long-
term pipe renewal scenarios. PARMS-Priority allows prioritization of individual pipes for renewal and
facilitates low-level planning of pipe replacement and some aspects of network operations (Moglia et al.,
! For more information on FARMS-PRIORITY, refer to http://www.csiro.au/files/files/pt41.pdf.
46
-------
2006). The PARMS-Priority software uses asset and failure data from Sydney Water to develop
deterioration curves through rigorous statistical analysis and the use of physical/probabilistic models. The
PARMS-Priority model is primarily based on five key tasks: risk calculation, failure prediction (using a
statistical non-homogeneous Poisson model and a physical/probabilistic model based on fracture
mechanics), cost assessment, data exploration (asset and failure records), and scenario evaluation.
Sydney Water then generates a risk score or matrix with the output from the FARMS model, which
considers the consequence of failure defined in monetary terms (cost of pipe renewal, customers affected
by the loss of water supply, etc.) and probability of a failure. Figure 4-4 illustrates the risk matrix used by
Sydney Water where pipelines that fall into the red area have the highest risk and those that fall in the
green area have the lowest risk.
Probab
Consequence of
Failure
5
4
3
2
1
ility of failure
>$2M
$1M-$2M
$0.75M-$1M
$0.35M - $0.75M
$0 - $0.35M
4
>50%
3
20% - 50%
2
5% - 20%
1
0% - 5%
Figure 4-4. Sydney Water's Risk Ranking Matrix
4.3.5.3 Validation of Condition Assessment Models and Associated Costs for Generating Models.
Sydney Water continually validates and calibrates the deterioration curves based on analyses and failure
history. It is currently working on refining the statistical model basis within PARMS-PRIORITY that
estimates the likelihood of failure based upon functions of pipe length, time, and other covariates. It is
also looking into how to measure the likelihood of failure (i.e., using either an annual versus a cumulative
probability). Because the FARMS model is developed and calibrated for each participant utility based on
its failure data, forecasted performance tends to mirror the actual asset performance. Sydney Water has
funded the Water Services Association of Australia for $100,000 to redevelop the FARMS model to a
new language and customize it for its use.
4.3.6 Washington Suburban Sanitary Commission. The WSSC is a public utility located in
Maryland. The WSSC prioritizes the inspection, maintenance, and renewal of water pipes using methods
that vary for different pipe types and diameters.
Currently, the WSSC prioritizes renewal activities (i.e., inspection, maintenance, repair, rehabilitation,
and replacement) for distribution water pipes (diameter < 16 inch) based on break frequency through an
analysis of work order history, pipe physical property, defect type, and soil corrosivity. In the future, the
WSSC plans to implement a water pipe condition rating system based on these parameters (see
Section 4.3.6.1). A pipeline is typically replaced if it has experienced three or more breaks within a five-
year period.
Condition assessment and renewal activities for transmission mains (diameter >16 inch) are dependent
upon pipe material. WSSC prioritizes condition assessment for cast iron and ductile iron pipes with
diameters ranging from 16 to 48 inches by data mining. Replacement decisions for ferrous pipes are
47
-------
primarily based on the analysis of the work order history, the age or class of pipe, and the presence of lead
or leadite joints. Alternatively, WSSC has developed a specific program for condition assessment
activities for PCCP transmission pipe (48 inch diameter or greater), which is based on pipe design, pipe
materials, size, manufacturer, reliability, and consequence of pipe failure. The program generates a risk
rating that WSSC uses to make decisions regarding inspection, repair, rehabilitation, or replacement.
Replacement decisions for PCCP are based on findings from pipe inspections.
4.3.6.1 Water Pipe Condition Rating System (Condition Rating Index). WSSC's Utility Master
Plan (UMP) is aimed at establishing a baseline condition rating for WSSC's transmission and distribution
water pipes. The UMP rating system consists of six risk factors that aid WSSC in prioritizing pipe
inspection needs: Land Use Factor (LF), Repair History (RH), Operational Needs (ON), Known
Manufacturing Defects (KD), Last Inspected (LI), and Pipe Diameter (DI). Each risk factor is further
subdivided into several rating factors to generate a score. For example, the last inspected risk factor is
subdivided into less than 5 years ago, 5 to 8 years ago, 9 to 12 years ago, 13 to 15 years ago, and 16 to
never giving scores of one, two, three, four, and five, respectively. Similar scoring is used for the other
risk factors. Overall, the empirical formula defining risk is given by:
Risk = (RH + DI + KD) * (ON * 4 + 17) (LF)
The use of the risk terminology here by WSSC does not follow the traditional definition of risk equal to
the likelihood of failure times the consequence of failure, although factors related to likelihood and
consequences of failure are included. The combined risk factor formula does not have an upper limit;
however, the WSSC considers a risk value of approximately 80 to be at risk and above 100 to be at
greater risk as illustrated in Figure 4-5. At the time this report was written, the rating system did not yet
include hydrology or hydraulic factors.
No Risk or Good
Condition
Slight Risk or
Average Condition
Medium Risk or
Be low Average
Condition
High Risk or Poor
Condition
Very High Risk or
Bad Condition
50
80
100
150
WSSC Risk Model
Figure 4-5. WSSC Risk Model Range
4.3.6.2 Validation of Condition Assessment Models and Associated Costs for Generating Models.
The water pipe condition rating system was developed in house, however cost details were not provided.
48
-------
4.3.7 City of Hamilton Public Works Department. The City of Hamilton Public Works
Department Water and Wastewater Division is a public utility located in Hamilton, Ontario, Canada. The
City of Hamilton has been working towards a goal of the development of a GIS-based Enterprise Asset
Management Plan across its transportation, water, and wastewater assets.
4.3.7.1 In/or Hansen Asset Management System (Economic Model). Hamilton has developed a
replacement profile for water mains utilizing the Infor Hansen asset management system,4 which is a key
component for estimating the timing of renewal activities. The replacement profile for water mains is
primarily based on the age of the asset and is allocated by year (either presented as km of pipe to be
replaced per year or revenue requirements per year). The average annual revenue requirements to sustain
the City's water network are derived through analysis of capital and operating revenues; however, the
exact calculation methods were not presented. This type of analysis appears to be similar to a Nessie
curve approach, which is used primarily for long-term capital planning and is not applicable for the
annual prioritization of pipeline renewal projects.
4.3.8 Louisville Water Company. LWC has a number of programs in place for managing its
water infrastructure assets. In particular, LWC has developed a Pipe Evaluation Model (PEM) for
prioritizing water pipe renewal. Additionally, in 1992, LWC established a $150 million, 15-year program
designed to replace and rehabilitate approximately 500 miles of unlined cast iron pipe installed between
1860 and 1935, numerous hydrants and valves, and 40,000 lead service lines.
The LWC pipe assessment and replacement decisions have been featured in WaterRF reports (O'Day et
al., 1986) and its pipe work has been the subject of an asset management case study as part of the larger
Global Water Research Coalition (Bhagwan, 2009).
4.3.8.1 Pipe Evaluation Model (Condition Rating Index). The LWC's PEM is a comprehensive
planning and decision support tool designed to assess priorities for the replacement and rehabilitation of
water pipes. The PEM is a detailed scoring system that assigns points based on over 23 assessment
factors that can take on different weighting schemes to allow LWC to adjust its model based on annual
priorities. The main assessment factors included within this model are categorized as follows:
Geographical (central business district, redevelopment areas, and roadway classifications)
Hydraulic (main size, fire flow availability, number of parallel mains, high pressure
frequency, and low pressure frequency)
Maintenance (main break frequency, joint leak frequency, material samples, corrosive soil
data, installation date, pipe type, joint type, and maintenance record)
Quality of service (taste and odor complaints, discolored water complaints, water quality
data, number of domestic/fire services, lead service frequency, dead-end water mains, and
paving age).
The renewal projects are scored according to all of these criteria and then the projects are ranked based
upon their degree of importance. LWC uses a criterion of two breaks per mile per year as the threshold
for replacement. Additional information about LWC's PEM approach can be found in Bhagwan (2009).
An example of LWC's PEM criteria and scoring for physical pipe aspects is shown in Figure 4-6.
1 For more information, refer to http://www.infor.com/hansen/solutions/asset-management/.
49
-------
Frequency 25 yr
(MBF) breaks/100
mi/yr
(weight 60%)
Age
(weight 5%)
Rating
1862 1865 & 1926 1942 5
1866 1925 0
Fypeof Pipe
weight 10%)
Main Break Split
(weight 15%)
Lead Service Freq
(LSF) per 500ft block
(weight 10%)
Rat in
Unlined,DCI,AC 10
SCI, UCI 0
CLD, Dl 0
other 0
Rating
#(splits)>50%oftotal 15
#(splits)>50%oftotal 0
Figure 4-6. Example of LWC's 2007 PEM Criteria and Scoring
4.3.9 Philadelphia Water Department. The PWD serves approximately 1.5 million residents in
the Philadelphia, Pennsylvania area. PWD has a long track record of monitoring and recording water
main failures. This historical information has served as a useful basis for its asset management decision-
making over the years. Since the 1960s, PWD has used a point-score system on a full-scale basis to
screen and prioritize water mains for renewal work, as presented in O'Day et al. (1986). This point-score
system is still in use today in a modified form as described below, but PWD is researching a more
comprehensive approach to asset management.
PWD has taken several steps to advance its overall asset management approach. To date, PWD has
established a GIS system to track its network assets and also developed a new hydraulic model that assists
in scenario testing to determine the criticality and failure consequence for specific water mains. It is also
currently deploying CityWorksฎ, which has improved maintenance recordkeeping to document more
detailed information from the field on each water main break experienced. It is researching the use of the
LEYP model to assign a likelihood of failure to a pipeline based on past breaks and other failure factors.
PWD has funded the trial of LEYP for approximately $100,000 and the trial is expected to last for a one-
year period. Its overall goal is to integrate the resulting information on failure likelihood and failure
consequence into a single asset management software program that also takes into account scheduling
considerations from street paving schedules in order to optimize the timing of renewal projects.
4.3.9.1 Structural Condition Model (Safety Factor and Deterioration, Decay, and Survival
Curves). PWD tested a structural condition model on a research basis as described in Deb et al. (2002),
but these models were not deployed on a long-term basis. At the time, PWD undertook a sampling
program to assess the physical condition of its water mains. For this sampling program, PWD conducted
testing to document the pipe's structural deterioration, inspected water mains that had failed, conducted a
pipe-wall analysis of the mains, and classified the surrounding soils. PWD then developed a computer-
based model to assess the structural condition of cast iron water mains using inputs such as pipe
characteristics, structural properties (tensile strengths), internal and external forces acting on the pipe, age
factors (manufacturing techniques, construction practices, etc.), corrosion rates, and soil characteristics.
This model was designed to generate a "structural condition rating" where values greater than 1.0 were
classified as satisfactory and values less than 1.0 were classified as questionable. The model was found to
be sensitive to several factors such as beam span space, corrosion rates, diameter, temperature range, and
working pressure and thus concluded that the success of the model depended on accurately determining
these difficult to quantify factors.
50
-------
4.3.9.2 Point System (Break Frequency). PWD has developed a point system for prioritization of
water main renewal activities. This point system comprises a combination of the age of the water main
and its break frequency. Overall, the goal of the PWD is to replace mains with seven or more points.
While this system is currently in use, PWD is testing a more comprehensive approach as described above
and expects its use of the point-score system to be phased out over the next few years. Limitations cited
were that the point system is heavily weighted toward the age of the pipe, it does not allow for
quantification of failure likelihood or consequence in a systematic manner, and it does not allow the City
to track trends in breaks from ambient temperature changes, increased road traffic, or other failure factors.
The PWD point system is a series of integrated databases. The PWD evaluates the maintenance history,
date and location of main breaks, installation year, size of main, and other information compiled in a
database. Each break is tagged with a status code and contract number of the replacement. Status codes
include active, on hold, planning, design, etc. Breaks entered each month are printed out and contain all
of the break information and the status code of each break. The points assigned for the year of installation
and break frequency can be seen in Table 4-4. The points are assigned on a block-by-block basis (which
is typically about 500 ft in length for the City of Philadelphia).
Table 4-4. The PWD Points Assigned for Year of Installation and Break Frequency
Year of Installation
pre 1854
1854-1877
1878-1900
1901-1938
1939-1966
1967-present
Break Frequency
A. Two or more breaks in the most recent year OR
B. Three or more breaks within the past 5 years AND
C. Each break not accounted for in A or B above
Points Assigned
5
4
3
2
1
0
Points Assigned
2 per break
2 per break
1 per break
There are some exceptions when assessing the assigned points. For example, cast iron water mains
constructed between 1946 and 1954 were installed using leadite joint compound, which has been found to
become brittle with age, causing premature joint failures. For the areas with leadite joints, replacement is
scheduled after the third break on a block.
Using the point system, PWD has witnessed a decreasing break trend and the target point total for
scheduling replacement has decreased from 10 to 7 over time.
4.3.10 The Metropolitan District. The Metropolitan Water District provides water and sewer
service to 12 communities, and more than 400,000 people in the Hartford, CT area. Its water distribution
system includes 1,600 miles of water mains and 100,000 customer connections. Over 92% of the water
mains are cast iron and ductile iron pipes with 67% of the pipes 10 in. diameter or less. The District
worked with Malcolm Pirnie to develop its capital improvement program (CIP), which utilizes a
combination of break frequency and deterioration curves to ascertain pipe condition and prioritize capital
improvement projects (Sekuler and Banciulescu, 2009).
4.3.10.1 Break Frequency and Deterioration, Decay, Survival Curves. According to Sekuler and
Banciulescu (2009), the Metropolitan District set out to accomplish three goals for its CIP:
51
-------
Create a process for prioritizing and allocating funds for water infrastructure needs on an
annual basis;
Employ a computer model to simulate water main deterioration and its impact on CIP
expenditures;
Establish a defensible, repeatable CIP process.
To achieve these goals, the consultant first compiled an asset inventory to ensure the District had a
complete listing of all of its assets and used this data as input into Harfan's Integrated Decision Support
System (IDSS) model to simulate asset deterioration and calculate anticipated CIP needs. The District's
database already included information on pipe installation year, material, and diameter. Additional data
that was added included soil type and operating pressure extrapolated from the District's hydraulic model
corresponding to the maximum daily demand. A statistical evaluation was then performed using
historical data on water main breaks to develop pipe classes and deterioration curves for use in the
computer model. Through this effort, deterioration curves were generated for 21 pipe classes that were
identified as having different breakage rate patterns. Each pipe class included a different combination of
material, age, soil type, and pipe diameter. Figure 3-3 in Section 3 shows the deterioration curves for 4-
in. to 6-in. diameter cast iron. Figure 4-7 below shows an example deterioration curve for cast iron pipes
greater than 10-in. diameter, along with the historic break data and end of life for that pipe class predicted
by a field sampling program. The pipe replacement threshold was set at a breakage frequency of 80
breaks/100 miles/year and defined as equivalent to a condition index of 0.25 (or 25%).
100%-
90%-
80%
70%
X
-a 60%
g 50%
1 40%-
o
0 30%-
20%
10*
0%
1
Sample deterioration curve:
n
" -^^^ Cast iron, 1925 - 1949, > 10" diameter, in Fine soil
^""^ *
^xซ
XT-TXn 0
o" (^^
* S
X
\
\
V
99 years ~
} 20 40 GO 80 100 I20
CD
10 3
Q>
7?
20
-------
methods. Data was recorded on the extent of corrosion, remaining wall thickness, types(s) of corrosion,
failure mechanisms, and then used to estimate remaining asset life. The remaining asset life was
incorporated into the deterioration curve as the age at which the condition index was expected to reach
0% (see Figure 4-7). The remaining asset life that was determined for the various pipe classes based on
this effort are shown in Table 4-5.
Table 4-5. Pipe Class Replacement Ages
Pipe Class
Replacement
Age fyrsT
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
^^M
Cast lron:<1 925:4-E:oFine
Cast lron:1 925-1 949:4-6:<>Fine
Castlron:<1925:4-6:Fine
Cast lron:1 925-1 949:4-6:Fine
Castlron:<1925:8:oFine
Cast Ironl 925-1 929:8:<>Fine
Cast lron:1 930-1 949:3:<>Fine
Castlron:<1925:8:Rne
Cast lron:1 925-1 949:8:Fine
Castlron:<1925:>-1G:Fine
Cast lron:1 925-1 949:>=10:Fme
Cast lron:<1 950:>=1 0:<>Fine
Cast Ironl 950-1959x10
Cast lron:1 950-1 959:>=10
Cast lron:>1 359:4-6
Cast lron:>1 959:8
Castlron:>1959:>-10
Ductile Iron
Other
Reinforced Concrete
Concrete
130
93
133
72
135
94
91
12S
87
139
99
115
80
102
63
74
90
87
73
140
130
The consultant configured the asset model, performed various model scenarios, and worked with the
District to develop a 45-year capital improvement program for water mains, valves, pump stations, and
storage tanks. The timing for the water main replacement was based upon a pipeline's physical condition
falling below the 25% condition index (e.g., 80 breaks/100 miles/year). A pipeline would not be placed
onto the CIP schedule until this criteria was met. Each year, the highest priority projects can be identified
based upon physical condition (via break history and field condition assessment), functional condition
(pressure, fire flow, capacity), and socio-economic (criticality, flow, traffic, critical customers, and access
problems) factors. Ultimately, the project benefitted the District by allowing it to identify the highest
priority projects that would fit within each year's annual capital improvement budget. The City has the
capability to continue to update the model input and output based upon the latest information stored in its
GIS and maintenance records (e.g., pipe repairs and break history). The City was able to save funding
resources by using the process to optimize the timing of renewal projects to coincide with the town's
paving schedules and combined sewer overflow separation plan (Sekuler and Banciulescu, 2009).
4.3.11 The Los Angeles Department of Water and Power. In the summer of 2009, LADWP
conducted an investigation of its water main leaks to determine the cause(s) for the increase in
distribution main breaks on cast iron pipe (LADWP, 2010). In its preliminary investigative report, plans
for transitioning to a more formalized and proactive pipe replacement program were presented as detailed
below.
53
-------
4.3.11.1 Break Frequency Curves. Prior to 2007, the LADWP primarily used a reactive strategy for
water main replacement with the primary drivers being failure modes of repeated leaks or insufficient
capacity. However, according to a 2009 water main preliminary leak investigation report (LADWP,
2010), LADWP is transitioning to a more proactive replacement program to better prioritize and target the
appropriate water mains for renewal. LADWP is working to develop a predictive model that uses
existing data for factors that contribute to water main deterioration (pipe material, soil corrosivity,
construction methods, and other internal/external conditions) to determine replacement priority for all
pipe segments in the system. The results of the model, combined with pipe criticality assessments and
leak history, will be used to focus resources on pipe segments that are more likely to fail and disrupt
service.
4.4 Summary of State-of-the-Practice Review
The municipalities and utilities that participated in the survey have very different inspection, condition
assessment, and renewal prioritization techniques. Most utilities apply some form of break frequency
curve often combined with an economic model to prioritize pipeline renewal activities. Some utilities use
more sophisticated deterioration curves; however, they typically required the assistance of external
consultants and/or computerized modeling programs. A table illustrating the utilities condition and/or
economic models can be seen in Table 4-5.
Based on the survey results and review of other case studies in the literature, a select number of medium
to large size utilities are using a formal condition assessment process involving either break frequency
and/or condition rating curve/index. This approach is often combined with an economic model in order to
both prioritize pipe renewal in the short term and capital improvement budgets over the long term. Input
for these models is generated from a number of different inspection, monitoring, and condition
assessment techniques. In general:
Inspection, monitoring and condition assessment techniques are dependent on the pipe
properties and differ significantly across utilities.
The types of "condition curves" and the factors used to generate the curves vary significantly
across utilities and are dependent on the utility's asset management objectives. These
objectives included using the condition curves to drive renewal timing decisions for pipelines
and/or for long-term capital investment planning.
Based on the survey results, break frequency and condition rating curve/index (with threshold
values) are the most common approaches used by the utilities to prioritize water pipe renewal.
Condition curve development is dependent on site-specific factors such as pipe material; pipe
diameter; maintenance practices; leakage and main break history; operational conditions; pipe
criticality; environmental conditions, etc. which vary significantly across utilities.
Six of the nine utilities surveyed are combining the results of long-term economic forecasting
models with their "condition curve" models to facilitate asset management decision-making.
Validation of condition curves is not a priority for most utilities surveyed and the validation
process could take years or even decades. Therefore, few utilities have validated their
models; however, most of the utilities surveyed felt confident about the results generated
from their models.
Utilities incur exceptionally different costs for their condition assessment programs which are
dependent on the extent of inspections, monitoring, testing, software development, data
54
-------
management, resources, etc. The software and programs utilized by each utility vary based
on their needs and budget.
All of the utilities surveyed use some form of GIS platform to integrate their water
infrastructure and condition assessment data.
Table 4-6. Summary of Utility Inspection, Condition Index/Performance Measures, and Models
Utility
EPCOR Water Services
Inc. (Canada)
Las Vegas Valley Water
District (LWWD)
(Nevada)
Newport News
Waterworks (Virginia)
Seattle Public Utilities
(SPU) (Washington)
Sydney Water
(Australia)
Washington Suburban
Sanitary Commission
(WSSC) (Maryland)
City of Hamilton Public
Works Department
(Canada)
Louisville Water
Company (LWC)
(Kentucky)
Philadelphia Water
Department (PWD)
(Pennsylvania)
The Metropolitan
District
Hartford, Connecticut)
Los Angeles
Department of Water
and Power (LADWP)
(California)
United Kingdom Office
of Water (OFW AT)
(United Kingdom)
Types of Condition Curves
Break
Freq.
X
X
X
X
X
X
X
X
X
Deter.,
Decay,
Survival
X
X
X
X
Cond.
Rating
/Index
X
X
X
X
X
Service
-ability
X
Econ.
X
X
X
X
X
X
Models
Reactive Renewal Program
Proactive Renewal Program
Computer Aided
Rehabilitation of Water
Networks (CARE-W)
Casses Linearly Extended
Yule Process (LEYP)
Pipe Prioritization
Replacement Model
Nessie Curve Economic
Model1
Water Main Replacement
Model
Wave Rider Economic Model
FARMS-PRIORITY (Water
Main Prediction Model)
KANEW Economic Model
UMP Condition Rating
System
Nessie Curve Economic
Model1
Hansen Asset Management
System Economic Model
Pipe Evaluation Model
(PEM)
Structural Condition Rating
System
Point System
Break Frequency and
Deterioration Curves
Reactive Renewal Program
Proactive Renewal Program
Serviceability based on
Performance Indicators
(Break Frequency)
(1) The use of Nessie curves by these utilities is not discussed in the text because the scope of the report does not include this
type of curve, which is primarily an economic forecasting model.
55
-------
As evidenced by the above summary, a number of different approaches are being implemented by
utilities, yet validation of the approaches is limited or non-existent. The various users indicated that they
were satisfied with their particular mix of tools, but reported large differences in the costs for developing
and implementing such tools. Without data on the accuracy and value of the various pipe condition
curves, there is a limited basis for the selection of one approach over another.
No one curve or index will meet all utility needs. The approaches described here are flexible enough to
be customized to meet the site-specific conditions and needs of other utilities. The approximately 52,000
community water systems in the U.S. might benefit from recommendations, guidance, and training on the
best practices related to asset management tools, models, curves, and indices currently available as they
move toward adopting more formalized asset management programs.
56
-------
5.0: FINDINGS AND RECOMMENDATIONS
5.1 Selecting the Approach
Condition curves are very site specific and a clear understanding is needed of the basic assumptions built
into the curves. With sufficient data and an understanding of failure modes and causes of distress, a
utility can effectively use condition curves to make replacement decisions; however, it is unlikely that a
single condition curve will work for an entire network. Curves will need to be generated for various pipe
classes and take into account the appropriate factors (e.g., environmental conditions, operational and
maintenance practices, pipe types, diameters, vintages, installation practices, etc.).
Like all tools, condition curves need to be appropriate to the task and the conditions being evaluated. It is
important to accept that like any tool they don't provide the answer, but assist the user in reaching one.
Curves and the methods used to develop the curves range from simplistic to complex. The more
extensive the factors that are taken into consideration, the more data hungry and complex the analysis
becomes for utilities.
The current short-term approach to condition assessment is to use readily available data, rather than
collect and/or integrate the appropriate data for making predictions and setting priorities. Every utility
has a wealth of environmental, historical, and operational information that has been or can be developed
into an asset database.
Those utilities that use condition curves tend to focus on the majority of pipe types in their systems,
divided into specific pipe classes. For those utilities that participated in the survey, the focus of their
condition assessment programs was primarily on ductile iron and cast iron, with slightly less focus on
AC, PCCP, and plastic pipe types. Development of condition curves for ductile iron, cast iron, and PCCP
can be somewhat easier because utilities have a fairly good understanding of the failure modes and
mechanisms and can collect data on failure indicators. Plastic pipes, on the other hand, have very
different failure mechanisms that are far more difficult to detect, making it more complex to develop
accurate condition curves and failure predictions.
Condition curves have greater potential application for smaller diameter distribution mains, which
dominate both the total water pipe mileage and the number of failures. There is a much greater body of
historical experience of failures and defects for smaller diameter pipes that can be used in the
development of condition curves and management of the network. For many smaller diameter pipes with
a relatively low replacement cost, it is not cost effective to obtain direct condition data by investigation or
laboratory testing, particularly because the consequence of failure from small diameter pipes is much less.
For these reasons, utilities need simple methods to prioritize their renewal activities for small diameter
pipes. For small diameter pipes break frequency curves combined with economic models appear to be the
most common approach. These methods are easier and less costly to develop, yet provide sufficient
information to do a fair job at supporting renewal scheduling of smaller distribution pipes. Utilities prefer
to set limits on the number of breaks where it becomes more cost effective to replace or rehabilitate than
to repair low risk pipelines. Break frequency curves are not meant to identify the date when the next
failure will occur; rather, they are used to forecast future trends in breakage rates.
Although larger diameter pipes represent a small percentage of the network and failures are much less
frequent, the consequences of failure are much greater. In addition, the forms of defects and modes of
failure for large diameter pipes also tend to differ from smaller diameters and there is relatively sparse
historical data on which to develop the condition curves. For these high risk pipelines, where failures are
catastrophic and unacceptable at any time, more extensive and complex approaches may be warranted and
57
-------
cost effective. Condition curves for larger diameter pipes should generally be generated from hard data
based on NDT inspections, investigations, and/or laboratory testing to define pipe condition and obtain
more accurate predictions of remaining life.
5.2 Application of Condition Assessment Models
Many different condition assessment approaches, some of which were described in Section 3, have been
developed by researchers and consultants. Many models have yet to be validated or verified because of a
shortfall in the quantity and quality of available data. In addition, the level of training and expertise in
statistics required to operate some of the techniques are not always available within a utility. There is a
gap in enthusiasm between those who develop condition assessment models and those who need to use
them. The model developers need to show how they use the data in their models and prove the business
case to utility employees and managers.
A general consensus is that any condition curve should be simple to understand, transparent to the users,
and easy to implement. Validation of the model with "real-world" case studies is also important. While
the model should be transparent and easy to use, the person implementing the model should be an
engineer or technical staff member with sufficient experience to understand the inherent complexities and
the correctness of assumptions used in the model.
A number of mathematical models have been proposed, yet few utilities have translated these
models and/or methods into a practical asset management tool. It is recognized that many
models are mathematically complex and will require special skills to operate. Most utilities
are uncomfortable with inputting data and leaving the model to provide the answer.
Limitations of model capabilities are not in mathematics, but in lack of fundamental
understanding of pipe performance based on environmental factors combined with failure
modes and mechanisms.
For some models, the underlying assumptions for factors and their weighting are not clear
and may not be applicable to all utilities.
Without adequate data, models cannot make informed predictions and it can be very costly to
collect and manage the level of data that are required. The cost of acquiring the necessary
data must be balanced against the pipe criticality.
What is not readily available and is needed by utilities contemplating the use of models to construct
condition curves is information on the varying approaches. For example:
How proven is a particular piece of software in providing the required level of performance
and decision-making help?
What is the basis for deciding which of the competing approaches is the most suitable for a
utility and its particular conditions?
How well do predictions correspond with actual events for a range of pipe types and
conditions?
Can the utility operate the model using its own staff? Does it require specialists or consultants
to undertake the work?
What level of data and in what format is needed to operate these methods? Is the model
capable of dealing with partial data when complete data are not available?
58
-------
What kind of costs will a utility incur in setting up and effectively operating any particular
approach?
How transparent is the method in terms of the assumptions used and their relevance to a
particular need of a utility? Has the underlying model been published and received peer-
acceptance?
Can the model be modified to take into account differing conditions and their relative
importance?
How much calibration of the model is required to make it work with a utility's data?
5.3 Short-Term Needs and Improvements
Overall, significant challenges still remain in predicting the remaining asset life of water mains using
condition curves including:
A range of factors influence the condition and performance of water pipe infrastructure,
which makes it nearly impossible to develop universally applicable condition curves.
There is a lack of historical databases, standard data collection and analysis protocols for the
development of robust condition curves capable of accurately predicting asset life.
There is no clear definition of pipe 'failure' some utilities consider a water pipe leak as a
failure, while other utilities consider water pipe burst/rupture as a failure.
Development costs are often the limiting factor when it comes to selection of condition
assessment strategies. Utilities generally do not have a budget that would allow large
investment and therefore less advanced techniques are usually chosen.
For distribution systems, the need is for less expensive options to estimate remaining life that
are likely to have a lower performance standard.
Based on the above gaps, several recommendations are provided for improving condition curve methods
for predicting the condition/performance of water pipes.
Development of condition curve methodologies that use standardized data collection and
reporting and can then be calibrated to site-specific data and conditions and then tested for a
range of utilities.
Continue to refine critical inferential parameters that affect water pipe performance based on
pipe material, diameter, joint type, external and internal environmental factors, etc.
Develop methods to improve the quality, quantity, and accessibility of condition/performance
assessment data. Additional research is needed on how to design more efficient and cost-
effective data collection strategies, how to extract information from existing datasets, and
how to standardize names and definitions for water utility assets, which subsequently will
allow the data to be shared and compared across utilities.
Due to restrictions imposed by the Paperwork Reduction Act, only surveys from nine utilities
are reported. It may be of value to conduct a larger scale utility survey to develop a more
extensive database on who is using condition curves/deterioration models and for what
purposes.
59
-------
Pilot existing and/or new models at various utilities to define their practical use and ease of
adoption by utilities.
New technologies with the potential for improving the pipe condition/performance
assessment process continue to emerge, and utilities need help in their technical and
economic evaluation.
5.4 Long-Term Needs and Improvements
Beyond the short-term needs, there is a fundamental long-term need to develop a performance-based
buried infrastructure asset management approach that involves a major improvement in the quality and
quantity of data used by all utilities.
The lack of data for many utilities is a major limitation in using anything, but the most basic approaches
to condition assessment. Ideally, pipe material, size, age, type of bedding, soil characteristics, operating
pressures, water temperatures, time, place and type of historical breaks should be available. However, in
many cases, only partial sets of data exist. A robust model or condition curve should be able to deal with
partial data, but it should be clear that in general the results will be less accurate and less precise
compared to those results obtained with a more complete dataset.
A second serious limitation is the lack of consistency or thoroughness in terminology and data collection,
which inhibits comparison of data amongst utilities or for research. For example, identifying critical
trends in main failures is complicated with different definitions of "water main break." Some consider
any leak, big or small, from a joint or a small perforation, to be a "water main break," while others
consider only large breaks where the line is no longer functioning to be classified as a "water main
break."
WaterRF and WERF have suggested data sets for asset management, both at the higher level (strategic)
and lower levels (operational or tactical). WaterRF projects have especially considered lower level data
related to condition assessment, pipeline renewal, optimizing the distribution system, and performance of
pipe materials. A broad overview of data obtained during condition assessment is included in a WERF-
WaterRF report titled "Condition Assessment Strategies and Protocols for Water and Wastewater Utility
Assets" (WERF, 2007). Examples of lower level data are the field data that may be collected after a pipe
is exposed for repair or maintenance work (Grigg, 2004; Matichich et al., 2006).
Some guidance should also be developed for identifying and quantifying the high risk scenarios, which
requires characterizing both the likelihood and consequence of failure. With limited funds, it is necessary
for utilities to focus on the highest risk situations to limit the impact of failures on consumers and the
public.
The longer-term goal should be to develop a national database of assets and failures with common
terminology and methods of data collection and analysis for assets and breaks. Relating breaks and leaks
to specific pipe materials and environments would support the decision-making processes of utilities and
allow them to benchmark their own experience against other utilities to provide more realistic life
expectancy predictions. Ultimately, this will allow identification of the most vulnerable pipes, reduce
failures, and improve understanding of the type and distribution of failure modes and mechanisms. It will
also facilitate improved allocation of funds in long-term replacement or rehabilitation programs.
An example of the successful implementation of a national asset and failure database is the UKWIR
mains break and asset databases in the U.K. The UKWIR's "Nationally Agreed Failure Data Base and
Analysis Methodology for Water Mains" has provided statistically robust information on mains failure
60
-------
together with basic data on the assets for more than 95 % of the U.K. water network (UKWIR, 2004). It
has more than 500,000 records and is increasing the number and quality of records in a staged
development. This database has provided valuable insights into both national and regional patterns of
failure for all types of pipe material (Trew and Mills, 2011). WaterRF Project #4195, in partnership with
UKWIR, is currently assessing the adaptability of the UKWIR database to the U.S.
Further understanding is needed on the distress factors and their combination that lead to failure. A great
deal of valuable research has been undertaken and further projects are currently underway. What is
apparent is that the combinations of distress factors vary in their importance according to location and
local conditions. For example, the role of freezing soils and low temperatures is not a concern in Florida.
Understanding and guidance are needed not just on the distress indicators, but their combinations that
become critical in differing situations.
Design life is often stated as some definitive number of years under normal operating conditions. The
reality is that the operational life of pipes is highly variable in relation to the calculated design life. Many
pipes have already doubled their design lives and others have failed in less than half that time. Design life
needs to go further than just specifying a pipe that meets the capacity and pressure requirements by
looking it up in the manufacturer's design guide and then developing layouts covering alignment, grade
and depth with appropriate appurtenances. To obtain longer service lives and reduce premature failures,
other factors that lead to failure need to be taken into account. Investigation of failures indicates that
many problems arise from improper handling, installation and operation. In addition, pipe performance
depends on a number of local factors such as soil conditions, temperature, and installation practices.
Current design guides for water mains make no reference to corrosion.
An example of this wider approach is AltPipe, a Web-based tool that can be used to assist designers in the
appropriate selection of pipe materials for culvert and storm drain applications. The AltPipe tool was
developed by the California Department of Transportation and takes into account various combinations of
pipe materials, dimensions, fluid compositions, operating conditions, liners, and coatings in the design of
culverts. The AltPipe experience, although not directly applicable to pressure pipes, provides a useful
example for developing a more realistic approach to design life. The AltPipe tool is available at
http://dap 1 .dot.ca.gov/design/altpipe/.
5.5 Conclusions
Any asset management program must start with a thorough review of available historical data about pipe
performance and failure. Once the necessary data are gathered, condition curves and/or deterioration
models can go a long way in providing insight into the condition of these assets. Currently, only large
water mains with costly consequences of failure may justify the cost of accumulation of data that are
required for physical model application, whereas the empirical/statistical models are an economically
viable approach for the smaller distribution water mains. For both large and small diameter pipes,
condition curves and deterioration models can help to evaluate current asset condition, assess the rate of
deterioration, and help to predict future condition and remaining asset life. This information can benefit
utilities by creating a repeatable and technically defensible process for prioritizing and allocating funding
for capital improvement plans. With sufficient data and understanding of failure modes and causes of
distress, a utility can also use condition curves to make renewal decisions based upon remaining asset life.
61
-------
6.0: REFERENCES
American Water Works Association (AWWA). 2004. Water://Stats 2002 Distribution Survey. American
Waterworks Association, Denver, CO.
American Waterworks Association (AWWA). 2001. Dawn of the Replacement Era, Reinvesting in
Drinking Water Infrastructure, American Waterworks Association, Denver, CO.
AwwaRF. 1993. Performance ofPrestressed Concrete Pipe, AwwaRF Project No. 724 (unpublished; see
references for U.S. Bureau of Reclamation reports).
Batista, J., and H. Alegre. 2002. "CARE-W WP1 D2 - Validation of the Rehabilitation Performance
Indicators System." National Civil Engineering Laboratory, Lisbon, Portugal.
Bhagwan, J.N. 2009. "Compendium of Best Practices in Water Infrastructure Asset Management." Global
Water Research Coalition, London, U.K.
Bianchetti, R. 1993. "Corrosion and Corrosion Control ofPrestressed Concrete Cylinder Pipes."
Corrosion Control Conference. New Orleans.
Buckland, P. and N. Hastings. 2001. "The Replacement Decision for 'Linear Assets'", International
Conference of Maintenance Societies - ICOMS 2001.
Burn, S., P. Davis, T. Schiller, B. Tiganis, G. Tjandraatmjadja, M. Cardy, S. Gould, P. Sadler, and A.
Whittle. 2006. Long-Term Performance Prediction for PVC Pipes, prepared for CSIRO,
Victoria, Australia, IWA Publishing.
Burn, S., P. Davis, T. Schiller, B. Tiganis, G. Tjandraatmadja, M. Cardy, S. Gould, P. Sadler, and A.J.
Whittle. 2005. Long-Term Performance Prediction for Polyvinyl Chloride Pipe. AWWA
Research Foundation, Denver, CO.
Davis, P., S. Burn, S. Gould, M. Cardy, G. Tjandraatmadja, and P. Sadler. 2007. Long-Term
Performance Prediction of PEPipe. AWWA Research Foundation, Denver, CO.
Deb, A.K., Grablutz, P.M., Hasit, Y.J., Snyder, J.K., Loganathan, G.V. and N. Agbenowsi. 2002.
Prioritizing Water Main Replacement and Rehabilitation. AWWA Research Foundation, Denver,
CO.
Deb, A.K. 1994. Water Distribution System Performance Indicators, Water Supply IWSA (Zurich)
12(3/4), pp. 11-20.
DERM. 2001. Guidelines for Implementing Total Management Planning- Asset Management: Asset
Evaluation and Renewal Implementation Guide. (Australia), 27 pp.
62
-------
Eisenbeis, P., Y. Le Gat, K. Laffrechine, P. Le Gauffre, A. Konig, J. Rostum, L. Tuhovak and P.
Valkovic. 2002. "Report No. 2.1: Report on Models Description," WP2 - Description
and Validation of Technical Tools, Aug.
Farshad, M. 2006. Plastic Pipe Systems: Failure Investigation and Diagnosis, Elsevier Science.
Gaewski, P. and F. Blaha. 2007. Analysis of Total Cost of Large Diameter Pipe Failures. AWWA
Research Foundation, Denver, CO.
Grigg, N.S. 2007. Main Break Prediction, Prevention and Control. AWWA Research Foundation,
Denver, CO.
Grigg, N.S. 2004. Assessment and Renewal of Water Distribution Systems. AWWA Research
Foundation, Denver, CO.
Grablutz, F. and S. Hanneken. 2000. "Economic Modeling for Prioritizing Pipe Replacement Programs,"
presented at the AWWA Infrastructure Conference and Exhibition, Baltimore, MD, 14 March.
Hadzilacos, T., D. Kalles, N. Preston, P. Melbourne, L. Camarinopoulos, M. Eimermacher, V.
Kallidromitis, S. Frondistou-Yannas and S. Saegrov. 2000. "UTILNETS: A Water Mains
Rehabilitation Decision Support System," Comp., Envir. and Urb. Syst., 24(3), 215-232.
Hartwell, J. 1994. Findings ofPetrographic Investigations of Embedded Cylinder Prestressed Concrete
Pipe, prepared for U.S. Bureau of Reclamation (USER), Denver, CO, September.
Horn, G. 2010. Personal Communication between James Thomson and Gregg Horn of DIPRA.
Hu, Y., Wang, D., and R. Chowdhury. 2010. "Condition Assessment Methods for AC Pipe and Current
Practices." Centre for Sustainable Infrastructure Research, National Research Council Canada,
Regina, SK. September.
Institute of Public Works Engineering Australia (IPWEA). 2006. The International Infrastructure
Management Manual. NAMS Group, New Zealand.
Kleiner, Y., and B. Rajani. 2004. "Quantifying Effectiveness of Cathodic Protection in Watermains:
Theory." ASCE Journal of Infrastructure Systems, 10(2), 43-51.
Kleiner, Y., R. Sadiq, and B.B. Rajani. 2006. "Modeling the Deterioration of Buried Infrastructure as a
Fuzzy Markov Process." Journal of Water Supply Research and Technology, 55(2), 67-80.
Kleiner, Y. and B.B. Rajani. 2001 . "Comprehensive Review of Structural Deterioration of Water Mains:
Statistical Models." Urban Water, 3, (3), pp. 157-176.
Kleiner, Y., and B. Rajani. 2009. "I-WARP: Individual Water Main Renewal Planner." Computing and
Control in the Water Industry 2009: Integrating Water Systems, Sheffield, U.K.
Le Gat, Y. and P. Eisenbeis. 2000. "Using Maintenance Records to Forecast Failures in Water Networks,"
Urban Water, 2(3), 173-181.
Los Angeles Department of Water and Power (LADWP). 2010. Summer 2009 Water Main Leaks
Preliminary Investigation Report. 14pp. LADWP, Los Angeles, CA. 14pp.
63
-------
Malandain, J., P. Le Gauffre and M. Miramond. 1999. "Organizing a Decision Support System for
Infrastructure Maintenance: Application to Water Supply Systems," J. of Dec. Syt, 8(2), 203-
222.
Marlow, D., S. Heart, S. Burn, A. Urquhart, S. Gould, M. Anderson, S. Cook, M. Ambrose, B. Madin, A.
and Fitzgerald. 2007. "Condition Assessment Strategies and Protocols for Water and Wastewater
Utility Assets." Water Environment Research Foundation, Alexandria, VA.
Marlow, D., P. Davis, D. Trans, D. Beale, S. Burn, and A. Urquhart. 2009. Remaining Asset Life: A
State of the Art Review, prepared for WERF, WERF Project No. SAMlR06d, Alexandria, VA.
Matichich, M., R. Booth, J. Rogers, E. Rothstein, E. Speranza, C. Stanger, E. Wagner, and P. Gruenwald.
2006. Asset Management Planning and Reporting Options for Water Utilities. AWWA
Research Foundation, Denver, CO.
Moglia, M., S. Burn, and S. Meddings. 2006. "Decision Support System for Water Pipeline Renewal
Prioritization," Journal of Information Technology in Construction 2: 237-256.
Moser, A.P. and K.G. Kellogg. 1994. Evaluation ofPolyvinyl Chloride (PVC) Pipe Performance.
AWWA Research Foundation, Denver, CO.
National Research Council of the National Academies (NRC). 1995. Measuring and Improving
Performance. National Academies Press, Washington, D.C.
Nelson, R. 2005. "Evaluating Pipeline Reinvestment Needs & GASB 34," Presented at the North
American Society for Trenchless Technology (NASTT) NO-DIG 2005 Conference, Orlando, FL,
24-27 April.
O'Day, O.K., R. Weiss, S. Chiavari, and D. Blair. 1986. Water Main Evaluation for
Rehabilitation/Replacement. AWWA Research Foundation, Denver, CO.
OFWAT. 2000. "Serviceability of the Water Main and Sewer Networks in England and Wales up to
March 1999." OFWAT Information Note No 35A, London, England.
Ojdrovic, R.P., C. Moody, M.T. Schroeder, M.S. Zarghamee, and M. Scali. 2007. "Condition
Assessment of an Asbestos Cement Pipeline," Pipelines 2007 Conference Proceedings: Advances
and Experiences with Trenchless Pipeline Projects.
Opus Consultants. 2001. "New Zealand Asbestos Cement Water Mains Manual" Wellington, New
Zealand.
Pelletier, G., Mailhot, A., and J.P. Villeneuve. 2003. "Modeling Water Main Breaks." Journal of Water
Resource and Management, 129(2), 115-123.
Price, R., R. Lewis, and B. Erlin. 1990. "Effects of Environment on the Durability of Prestressed
Concrete Cylinder Pipe." ASCE Conference Pipelines in the Constructed Environment.
64
-------
Rajani, B.B. 2000. Investigation of Grey Cast Iron Water Mains to Develop a Methodology for
Estimating Service Life. AWWA Research Foundation, Denver, CO.
Rajani, B. and Y. Kleiner. 2001. "Comprehensive Review of Structural Deterioration of Water Mains:
Physically Based Models," Urban Water, 3(3), 151-164.
Rajani, B. and Y. Kleiner. 2002. "Towards Pro-active Rehabilitation Planning of Water Supply
Systems," Proceedings of the International Conference on Computer Rehabilitation of Water
Networks CARE-W, Dresden, Germany, November, pp. 29-38.
Romer, A.E., G.E.C. Bell, D. Ellison, and B. Clark. 2008. Failure ofPre-Stressed Concrete Cylinder
Pipe. AWWA Research Foundation, Denver, CO.
Rose, D. 2008. Power Point Presentation. EPA Technical Forum. Washington, D.C.
Rostum, J., S. Saegrov, J. Vatn and G.K. Hanson. 2000. "Aquarel - A Computer Program for Water
Network Reliability Analysis Combining Hydraulic, Reliability and Failure Time Models," Water
Network Modeling for Optimal Design and Management, Woodbury Park, UK, Sept. 11-12.
Sekuler, L. and C. Banciulescu. 2009. "Developing a CIP Using Deterioration Modeling and Field
Sampling Approach," presented at the AWWA Third Leading-Edge Conference on Strategic
Asset Management, Miami, FL, November 11-13.
Stone, S., E. Dzuray, D. Meisegeier, A. Dahlborg, and M. Erickson. 2002. Decision-Support Tools for
Predicting the Performance of Water Distribution and Wastewater Collection Systems, prepared
for U.S. EPA National Risk Management Research Laboratory, EPA/600/R-02/029, Edison, NJ.
Thomson, J. 2010. Inspection Guidelines for Wastewater Force Mains, prepared for WERF, WERF
Project No. 04-CTS-6URa, May.
Travers, F. 1994. Acoustic Monitoring ofPrestressed Concrete Pipe at the Agua Fria River Siphon,
prepared for U.S. Bureau of Reclamation (USER), Denver, CO, December.
Trew, J. and D. Mills. 2011. "Case Study No.l Polythene Pipe Failures on Water Mains." UKWIR
National Sewers and Water Mains Failure Database. January.
Ugarelli, R. and S. Bruaset. 2010. Review of Deterioration Modelling Approaches for Ageing
Infrastructure. SINTEF Building and Infrastructure. Norway. August.
U.K. Water Industry Research (UKWIR). 2001. Understanding Burst Rate Patterns of Water Pipe.,
Report Ol/WM/02/14, U.K. Water Industry Research, Queen Anne's Gate, London, U.K.
U.K. Water Industry Research (UKWIR). 2004. Nationally Agreed Failure Data and Analysis
Methodology for Water Mains. Vol. 1 (03/RG/05/7) and Vol. 2 (03/RG/05/8). U.K. Water
Industry Research, Queen Anne's Gate, London, U.K.
United States Environmental Protection Agency (U.S. EPA). 2002. Deteriorating Buried Infrastructure
Management Challenges and Strategies. Prepared by American Water Works Service Company
for U.S. EPA, Office of Water. Washington, DC.
65
-------
United States Environmental Protection Agency (U.S. EPA). 201 la. Evaluation of Condition Assessment
Technologies for Water Transmission and Distribution Systems. Office of Research and
Development. Cincinnati, OH.
United States Environmental Protection Agency (U.S. EPA). 201 Ib. "National Primary Drinking Water
Regulations."
Uyeda, H.K., M.T. Peabody, and D.T. Johnson. 1994. Concrete Pipe Failure Jordan Aqueduct, Reach 3,
prepared for U.S. Bureau of Reclamation (USER), Denver, CO, July.
Von Fay, K.F. and M.T. Peabody. 1994. Historical Performance of Buried Water Pipe Lines, prepared
for U.S. Bureau of Reclamation (USER), Denver, CO, September.
Water Research Center (WRc). 2003. Manual of Sewer Condition Classification, fourth edition,
December.
Water Environmental Research Foundation (WERF). 2007. Condition Assessment Strategies and
Protocols for Water and Wastewater Utility Assets. Water Environment Research Foundation,
Alexandria, VA.
Water Research Foundation (WaterRF). In progress a Long-Term Performance of Ductile-Iron Pipe,
research being conducted by the National Research Council of Canada, WaterRF Project No.
3036.
Water Research Foundation (WaterRF). In progress b_. Fracture Failure of Large Diameter Cast Iron
Water Mains, research being conducted by the National Research Council of Canada, WaterRF
Project No. 4035.
Water Research Foundation (WaterRF). In progress c^ Long-Term Performance Prediction of Steel Pipe,
partnership with Commonwealth Scientific & Industrial Research Organization, Australia,
WaterRF Project No. 4318.
Water Research Foundation (WaterRF). In progress d. Long Term Performance of Asbestos Cement
Pipe, research being conducted by the National Research Council of Canada, WaterRF Project
No. 4093.
Water Research Foundation (WaterRF). In progress e. Access to UKWIR Failure Data for Analysis
Methodology for Water Mains. WaterRF Project No. 4195.
Weimer, D. 2001. "Water Loss Management and Techniques." DVGW Technical Report, Germany.
Woodhouse, J. 1999."Innovative Approaches to Decision Making for the Management of Aging Physical
Assets." The SALVO Project.
66
------- |