Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 119 (2015) 253 – 262

13th Computer Control for Water Industry Conference, CCWI 2015

Pipe failure analysis and impact of dynamic hydraulic conditions in

water supply networks
Hossein Rezaeia*, Bernadette Ryanb, Ivan Stoianovc
a,c
InfraSense Labs, Department of Civil and Environmental Engineering, Imperial College London, SW7 2AZ, United Kingdom
b
Severn Trent Water, 2 St John's Street, Coventry CV1 2LZ, United Kingdom

Abstract

Water pipe failures are associated with pipe characteristics, material properties, environmental and loading conditions.
Understanding the impact of individual factors, their dynamics and interactions is extremely complex. The steady and unsteady
state hydraulic conditions may impose excessive loading on assets. The resulting cyclic loadings when acting upon degraded (e.g.
corroded) pipes could contribute to the development and acceleration of failures. This paper investigates the impact of the
dynamic hydraulic conditions on pipe failures by analysing historical burst records, designing and implementing an extensive
experimental programme and investigating case studies by correlating dynamic pressure and pipe failures.

© 2015
© 2015The TheAuthors.
Authors. Published
Published by Elsevier
by Elsevier Ltd.
Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Scientific Committee
Peer-review under responsibility of the Scientific Committee of CCWI 2015.
of CCWI 2015

Keywords: Pressure Transient; Pressure Flactuations; Pipe Failure; Failure Mechanism; Deterioration Models

1. Introduction

Pipes and material properties, environmental conditions, and internal and external loads are critical factors
affecting pipe failure mechanisms. Understanding the interactions of individual factors and their impact on pipe
failures is extremely complex. While most of previously published work has focused upon pipe, material properties
and environmental conditions, there is little understanding of the extent operational conditions such as sudden and
gradual pressure fluctuations are a contributing factor for pipe deterioration and failures. It is an industry recognised
fact that a reduction in the average zone pressure lead to a reduction in pipe failures. However, the impact of gradual

*
Corresponding author. Tel.: +44-079-4222-1302; fax: .: +44-020-7594-5934
E-mail address: h.rezaei12@imperial.ac.uk

1877-7058 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Scientific Committee of CCWI 2015
doi:10.1016/j.proeng.2015.08.883
254 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262

and sudden pressure fluctuations on mains failures is not well understood. Hydraulic instabilities will have both
short and long term effects on the deterioration rate and failures of assets. The risk of occurrence of hydraulic
instabilities in water supply systems is increasing. Regulatory drivers for improving the operational efficiency has
encouraged a wide scale implementation of control in water distribution systems both in terms of pump scheduling
and pressure minimisation. The complex hydraulic interactions and poor understanding of the dynamic hydraulic
conditions may accelerate deterioration mechanisms, thus limiting the benefits of pressure control. Water utilities
operate an ageing infrastructure with diminishing pipe pressure ratings. The initiation and development of micro
cracks, corrosion and fluctuating pipe stresses due to operational conditions and/or environmental factors have
lowered the tensile strength of pipelines. Cyclic pressure loading can be a major contributing factor in the
deterioration and failure of mains.
The research in pressure transients has mainly focused upon the derivation of design criteria for the selection of
maximum pipe pressure rating and the sizing of surge protection devices. The short and long term impacts of the
dynamic fluid conditions on the deterioration of pipelines has not been given adequate attention. The main reason
for this is the lack of understanding of the dynamic hydraulic behaviour, available high resolution data, detailed
records and forensic investigations documenting failure mechanisms. Over the last few years, advances in data
logging tools and technologies have allowed the investigation of this phenomenon in a more rigorous approach [1].
As a continuation of this work, an extensive experimental programme has been designed and executed by the
authors to study the occurrence and severity of the dynamic flow conditions in operational networks (both
transmission and distribution systems). This unique investigation has provided a detailed insight into the dynamic
hydraulic behaviour of water supply systems
In this paper, we present an initial investigation of the impact of the dynamic hydraulic conditions on pipe
failures. We start with the analysis of historic pipeline failures from a UK water utility and discuss the relationship
between pipe failures and commonly perceived contributing factors [2–5]. Various pipe failure models are then
reviewed and their limitations with regards to non-inclusion of the hydraulic conditions is highlighted. It is proposed
that flow dynamic conditions should be included amongst the factors impacting mains failures. The development of
cracks and failure mechanisms initiated and/or aggravated by stresses from pressure fluctuations might be
significant. Various elements contributing to these mechanisms, including fatigue failure, pipe corrosion, and
pressure fluctuations are outlined and combinations of these factors resulting in pipe degradation and failure is
discussed. In order to validate to what extent the hydraulic instabilities in a system have an effect on pipe
deterioration and failure, an extensive experimental programme has been setup. We present details of the design and
execution of the experimental programme which includes the time-synchronised high-frequency (128S/s)
acquisition of pressure data at more than 200 locations across the supply network of a UK water company. Finally a
case study is presented discussing the correlation of pressure variations and historic pipeline failures.

2. Analyses of historic pipe failures

Pipe failures are caused by applied forces exceeding the residual strength of the pipe material. Pipe breakage
occurs when the stresses (operational and environmental) act on pipes where corrosion, degradation, inadequate
installation or manufacturing defects has impacted the pipes’ structural integrity. The physical mechanisms of
failures in mains are usually a complex function of many contributing factors. This includes pipe properties (e.g.
material, size, etc.); internal and external loadings; and environmental factors (e.g. chemical properties of
surrounding soil, etc.) [4]. Consequently, various failure modes are observed, including: joint/ connection failure,
brittle failure, split pipe, transverse break, graphitisation, pitting holes, longitudinal, circumferential failures spiral
cracking, blowout hole, etc.
Amongst the factors contributing to mains failures, physical characteristics (material and size), surrounding soil
properties (corrosivity and fracture potential) and temperature have been identified as most important factors. An
analysis of the relationship between these contributing factors and historic bursts records is presented in Section 2.1.
The analysis aims to validate whether these factors can explain all failures or there might be other underlying factors
contributing to development of some of these failures.
 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262 255

2.1. Historical failure analyses

The analysis of historic burst records that was carried out is based on nearly 78,000 failure records over a 10-year
period (2003-2013) from a water supply network of approximately 48,000 kilometres of mains. The pipe material
consists of around 50% metallic pipes, mainly Cast Iron (CI) and Ductile Iron (DI), 10% unknown material and the
remaining being plastic, concrete, and steel. Nearly 70% of the pipes have diameter less than 150mm, 90% less than
300mm, and only 7% greater than 300mm (the rest are unknown). In the presented analysis, the term ‘mains burst’
indicates a ‘mains repair’, either as result of a visible burst, a detected leak resulting in a repair activity.
Pipes internal diameter: Figure 1(a) shows a reduction in burst rate per length of mains with the increase of
internal pipe diameter. Larger diameter pipes have thicker walls, enabling them to withstand higher circumferential
loads, but also they are more resistant to failure following loss of wall thickness as a result of corrosion or other
degradations.
250
5% 3%
No of bursts /100km

200 6%
150 CI
mains

100 7% AC
PVC
50 10% DI
0 PE
<100 mm 100-150 150-250 250-400 400-600 >600 mm 69% Unknown/others
mm mm mm mm
Mains size ranges
(a) (b)

Figure 1 Distribution of failures: across mains diameter range (a); and different mains material (b)

Mains material: While only one third of pipe length is cast iron (CI, ~34%), nearly two thirds of the historic
mains bursts have occurred on cast iron pipes, Figure 1(b). This results in 326 bursts/km for CI, 187 bursts/km for
asbestos cement (AC), 143 bursts/km for polyvinyl chloride (PVC), 99 bursts/km for polyethylene (PE) and 75
bursts/km for DI over a 10-year period.
The brittleness of cast iron and its lower tensile strength (~ 200MPa) are probably the main reasons for the higher
density of failures. Furthermore, age combined with varied manufacturing and installation methods, temperature and
cyclic pipe stresses due to pressure variations further increase the occurrence of bursts in cast iron pipes.
Temperature: A reduction in temperature can contribute to pipe failures through volume expansion of water, frost
heave in the soil surrounding the pipe, and increased brittleness. Various factors such as intensity and duration of
cold weather, rate of freezing, surface cover, mains depth, etc. might affect the development of frost-heave. As
shown in Figure 2(a), nearly 46 thousands bursts occur in the Autumn and Winter seasons which is 60 percent of all
bursts. Approximately 30,000 of bursts occur in the Spring and Summer which is 40 percent of all bursts. This
summary shows that while temperature is an important factor, it is not a predominant factor.

30000 10000
No fo bursts

8000
No. of bursts

20000
6000
25179 4000
10000 20917
14228 16006
2000
0 0
Spring Summer Autumn Winter
1-5
6-10

96-100

Unknown
101-105
106-110
11-15
16-20
21-25
26-30
31-35
36-40
41-45
46-50
51-55
56-60
61-65
66-70
71-75
76-80
81-85
86-90
91-95

>111

Pipe Age
(a) (b)

Figure 2 Distribution of pipe failure seasonal (a); and for various pipe age (b)
256 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262

Pipe age: Intuitively, there should be a strong positive correlation between pipe age and the number of pipe
bursts. Figure 2(b) shows that this correlation is not strong, as expected. This could be due to the combined effect
of multiple factors associated with age such as method of pipe manufacturing, construction methods, corrosion rate,
etc. Previous studies, such as [6], also suggest a complex relationship between age and burst rate.
Surrounding soil: Surrounding ground conditions have been identified as the most important factors contributing
to pipe failures [7] and [8]. The impact of this is generally in the form of external loads on pipes as the result of
ground movement as well as soil corrosivity causing physical damage to the pipe structure. Two indices of fracture
potential and corrosivity are used to represent the soil behaviour. Developed by the Soil Survey and Land Research
Centre (SSLRC), maps detailing the extent of soil fracture potential and corrosivity across the UK [9] are used to
analyse soil-burst relationships. As illustrated in figure 3(a), approximately half of the historical bursts have
occurred in the areas where soil fracture potential is considered very low to low. This is compared with 41% of
bursts occurring in areas with a moderate level of fracture potential. Only less than 15% of all bursts have been on
mains surrounded by soil with a fracture potential level of high to very high.
2% 0%
10% 4%
8% Non-aggressive
27% 17% 25% Slightly aggressive
Very low
Low Moderately aggressive
Moderate Slightly to highly aggressive
41% High 12% Highly aggressive
20% Very high Moderately to very highly aggressive
34% Very highly aggressive / rock

(a) (b)

Figure 3 Distribution of historical bursts vs. soil fracture potential (a); and soil corrosivity (b)

With regard to the corrosivity characteristics of the surrounding soil, as presented in figure 3(b), around one
quarter of all bursts have occurred in non-aggressive soils and approximately three quarters of bursts have occurred
in mains where the surrounding soil is either non-aggressive, slightly or moderately aggressive. Only 25% of all
bursts occurred in mains with surrounding soil of aggressive (sulphate rich). In agreement with published studies
(e.g. [7] and [8]), many bursts occurred in aggressive soils, indicating the impact of soil corrosivity in initiating and
accelerating pipes/joins corrosion and consequently resulting in failure. However, the result above highlights that
these factors cannot explain large number of bursts across the system where soil is either not aggressive or slightly
aggressive.
Conclusion: The analysis presented a weak and in some cases no correlation between mains failures and the
factors conventionally considered as the key impacting factors. For the condition on the surrounding soil as well as
temperature, there is a lack of strong correlation between a large proportion of the mains failure and these factors.
Numerous failure events have occurred on mains surrounded by low or non-aggressive soil with very low fracture
potential level. Similarly, high proportion of historical failures have occurred at times of mild to warm temperature.
This analysis has limitations due to lack of consideration for spatial distribution of the failure records as well as
interaction of various factors in promoting failures. Despite this, it cannot be ignored that factors traditionally
considered to be the main causes of failures fail to explain a large proportion of the historical burst records. Thus, in
causing mains failure, there seems to be impacts from other potential factors; that so far have been largely ignored.
Stresses imposed by hydraulic instabilities in the system are suggested to be such factors.

2.2. Deterioration models

There are generally two approaches used to estimate pipe breakage vulnerabilities in water pipelines: physical-
based and statistical-based modes [4]. Physical-based models are used for the prediction of the remaining service
life of mains; statistical-based models are employed for interpolation of the whole area. By analysing the loads that
the pipe are exposed to, and the pipe capacity to resist these loads, physical models attempt to predict pipe failure
[3]. This approach requires understanding of degradation and corrosion processes in pipelines. Combined with
 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262 257

information on the external loadings, pipe residual structural resistance is obtained. Then a soil-structure interaction
analysis is undertaken and the remaining service life of the asset is calculated. Physical models are categorised into
deterministic and probabilistic, where deterministic models estimate the pipe failure without use of probability; and
probabilistic models incorporate uncertainties in the modelling of physical pipe conditions [10]. In physical
deterministic models, input variables have specified values and these values do not change temporarily or spatially.
Such values are determined by lab tests or design standards. Various models used in this category are presented in
[11], [12], etc. Physical probabilistic models consider various temporal and spatial variability which can cause a
level of uncertainty during estimation. Various models based on this method include: [3], [13], etc. Physically based
methods require significant data and usually costly investigations on the pipe’s deterioration processes. To avoid
these limitations, statistical-based models are developed to predict break patterns and the level of deterioration from
historical data.
Statistical-based models utilise the environmental, operational and historical data in order to develop a pipe
failure rate. Statistical models are generally categorised in deterministic and probabilistic groups. Statistical
deterministic models express the pipe failure as a time dependant phenomenon. These could be categorised time
exponential and time-linear models. Examples of time exponential modes are: [14], [15], etc. and deterministic time
linear models are: [16], [17], etc. Statistical probabilistic models can be divided to single-variant and multi-variant
models. Multi-variant models are useful when various covariates are considered in the analysis, examples of this
approach include: [18], [19], etc. Single-variant models determine burst probabilities on grouped or clustered data.
Examples of single-variant model are: [20], [21], etc. Moreover, [10] adds an additional category of soft computing
methods that includes fuzzy sets, artificial neural networks and generalised multivariate exponential models.
There has been a number of failure models developed over the last four decades, as listed in [3], [4] and [10].
Depending on the approach and method used, these models may use a variety of factors that contribute to the system
deterioration. [22] provides a list of factors often used in order to predict risk of failure in water mains. A review of
the data requirement for the existing failure models confirms that only a few of these models consider hydraulic
conditions of the system (pressure and/or velocity) as an impacting factor; and all ignored impact of pressure
fluctuations.

3. Impact of pressure fluctuation on pipeline failure

The term pressure fluctuation is used to express any alteration in pressure level inside the pipelines, whether it is
a gradual change as exists in the diurnal pressure profile or sudden changes as happens in pressure transient events.
These tend to have a repeating pattern, even though there might be much variation in the frequency, magnitude and
shape of each event. In addition, separation of gradual and sudden pressure fluctuation profiles from other profile
‘noise’ might be rather complex. Existing theories and some practical methods may consider maximum pressure
when studying mains failure. However, there have been limited studies looking into failure mechanism of mains
when exposed to frequent cycles of pressure fluctuations. In practice, most pipes across a distribution network will
be exposed to varying pressure during daily and seasonal variation of water use as well as due to changes in
operation of the system. Similar to sudden pressure changes (i.e. pressure transient events), the relationship between
(gradual) pressure variation and mains failure is not well understood.
The design of water pipes takes into consideration the working pressures, and as such, maximum water pressure
across the network is unlikely to be sufficient to cause failure. However, where pipe wall thickness is degraded due
to corrosion or other factors, high pressure could cause failure. Considering the cyclic nature of pressure
fluctuations, fatigue mechanisms might be an underlying factor in asset deterioration (and eventually failure). In the
case of water pipelines, the initiation phase of fatigue might be of little concern. It might be rare for such pipes to
experience fatigue failure in the absence of initial flaws, damage or other fractures that concentrates stresses locally.
However, existence of such flaws can dramatically change the component behaviour. Such factors may have the
effect of concentrating stresses by factors of 2 to 5 [23]. Therefore, locally concentrated stresses may be great
enough to cause failure in much lower number of cycles, possibly hundreds to thousands of cycles. In such cases,
the initiation phase is bypassed and fatigue failure is governed by the fatigue crack propagation life, according to
Paris Law. Cyclic loading could lead to sub-critical crack growth from corrosion-induced defects and hence
258 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262

eventually promote failure. Existence of cracks, as result of these factors, substantially reduces the number of cycles
required for development of fatigue failure.
Shown in Figure 4, hoop stresses imposed by cyclic loadings of pressure fluctuation in the pipe continue to
enlarge the cracks and expose a larger surface of the material within the crack. In addition, when corrosion is
developed in such an opening, corroded deposit expansion forces additional loads on the inside of the crack,
enlarging the opening even further. This in turn exposes more area to the corrosive environment therefore
developing additional corrosion.

Figure 4 Schematic of crack growth on pipe wall

This cyclic phenomenon combined with the crack development as a result of hoop stress will accelerate crack
growth, in comparison with each factors impacting in isolation. Crack development will continue until the residual
wall thickness (shown as af in Figure 4) is low enough not to resist the next loading stress. If an individual crack is
not sufficiently large to cause pipe failure, multiples of such cracks, when aligned, can join to develop larger cracks.
In addition to the pipe’s material, wall thickness and pipe structural condition, the main factors impacting
development of this failure mechanism and/or degradation of the pipelines exposed, are pipe internal and external
corrosion, various cyclic loadings, and fatigue failure characteristics, as described below.
Types of corrosion include internal, external, Stress Corrosion Cracking (SCC) and corrosion fatigue. Pipe
properties, internal and external environmental characteristics and stresses imposed on the pipe result in the
development of various corrosion forms.
Fatigue involves the initiation and evolution of cracks by loads that are too low to cause failure under monotonic
loading. Loading profiles as a result of diurnal and seasonal pressure variations, in most cases, could be
approximated to an asymmetric profile (non-zero mean stress) and pressure transient events to a random profile
(varying stress ration). The overall effect of stress (or strain) range is often characterized by stress-life (S-N) curves,
S is the amplitude of cyclical stresses and N is the number of cycles. This consists of crack initiation and crack
propagation curves, and the combined effect of both creating total life curve. With a sufficiently large number of
cyclic loadings, fatigue can occur eventually even if the material surface is free of stress-concentrating features. This
however requires a very large number of cycles that might not be the common scenario for water pipes. Initiation of
fatigue occurs at microstructure-scale nucleation sites within the material. Such sites could be inclusions, pores, soft
grain regions or micro-voids [24]. The presence of macro-scale stress concentrators, such as manufacturing flaws,
sites of corrosions pits and stress-corrosion cracks, etc. enhances this process. The crack growth rate could be
approximated to increase exponentially, as per Paris Law, as load cycles accumulate. This means that a small crack
may remain small for a long time, and suddenly expand to a failure limit. There is also impact of mean stress, Haigh
diagram [24] suggesting a lower stress amplitude could cause fatigue failure provided a higher imposed mean stress.
The effect of fatigue is often assumed to be cumulative. Cumulative damage from various cyclic loadings can be
estimated by using a linear damage model, double linear damage rule, damage curve approach, inverse power law
model, etc. [24].
Despite various attempts to operate ‘calm networks’, there are numerous factors causing unsteady flow
conditions across the network. These might be associated with system operation, pressure control activities, rapid
flow changes, customers’ behaviour (especially industrial customers), and operational variations (planned or
accidental) such as pumping, valve status changes, operational activities, etc. There are various methods of pressure
transient suppression including direct (e.g. rerouting, eliminating valve slam, etc.) and diversionary (e.g. air vessels,
surge shafts, etc.) methods [25]. In addition, various mechanisms of pressure transient wave energy dissipation (such
as leakage, unsteady friction, network demand, etc.) might result in a smaller amount of energy travelling across the
 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262 259

system. Despite such reductions, the dynamic hydraulic behaviour could have various impacts on the distribution
networks; namely increased stress on pipes resulting in bursts, discolouration and water quality issues, as well as
increased corrosion rates accelerating the deterioration of the system. The unacceptable conditions occurring due to
pressure transients depend on magnitude of pressure changes, speed of occurrence and length of time the system
components are exposed to them [25].
While dynamic hydraulic conditions have been researched for a while, considerable attention has been devoted to
impacts from sudden collapse of (large) pipes. However, short and long term impacts of dynamic hydraulic
conditions on deterioration of pipelines, specifically effect of pressure transients on pipes gradual degradation, have
not been given adequate attention. The main reason for this could be attributed to a lack of understanding of
dynamic hydraulic behaviour and its long term impact in the system. This is mainly due to limited insight into
dynamic system behaviour and its impact and also lack of adequate instrumentations to gather operational high
frequency dynamic data.

3.1. Experimental programme

With recent development in tools and technologies, capturing, storing, reviewing and analysing such high
frequency data seem more practical. ‘InfraSense’, developed at Imperial College London, provides a sophisticated
tool and platform for capturing and analysing high frequency pressure data. This platform is used to develop and
execute capturing and analysing a large scale high frequency pressure logging across the live network of a water
company (at more than 200 locations). An experimental programme was designed and executed to gather data
across various assets and locations with each location logged for between 4-10 weeks. A multi-stage selection
algorithm is used in order to select the logging points. This included selection of locations where hydraulic
instabilities are expected, for instance, downstream of large pumps switching frequently. This also investigated the
locations with both historically low and high densities of pipe failures per pipe length. In addition, various random
points were selected by randomly selecting DMAs with an average rate of failure and then selecting random points
within these.
An important initial finding of the programme is the extent of hydraulic instabilities in the system. While logging
was undertaken at various locations expected to exhibit instabilities, the results presented a (much) higher level of
instabilities than expected. These were for both frequency and magnitude of the numerous events captured; for
instance as illustrated in Figure 5. Furthermore, in addition to the impact of operational activities, such as pump
switches and valve operations, sources of pressure transient events have been found to be, in many cases, large
commercial customers as well as manual operation of control assets.

45 90
Pressure (mH2O)

Pressure (mH2O)

70
35
50
25
30
15 Time
10
Time
5 -10
28/5 28/5 28/5 28/5 29/6 29/6 29/6
10:00 10:15 10:30 10:45 13:21 13:23 13:25
(a) (b)

Figure 5 Frequency of pressure transient events on a 500mm main (a); amplitude of a pressure transient event (b)

Design of water mains generally includes a safety factor that takes into account pressure transient events. In
practice, consideration of pressure transients in the system is usually limited, due to various factors. Firstly, transient
analysis of a network is only undertaken in cases with either high risk/consequence of potential failure or existence
of large pumping stations. Even where transient analysis is undertaken, in the majority of cases, surge protection
schemes are not fully validated, both during and after commissioning. This is in part due to lack of adequate
260 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262

monitoring equipment to provide the required information and data for validation. In addition, lack of adequate
maintenance of surge protection devices (e.g. air vessels, air valves, etc.) is an important factor in poor performance
of these schemes. Moreover, pressure transient analyses are only valid for the system configurations for which the
analyses are carried out. While changes in system operation and configuration are common across the water supply
network, many surge protection schemes might not be updated following such changes.

3.2. Pressure fluctuation vs. mains failure – case study

Analyses were undertaken to investigate the correlation between daily/weekly pressure variations and historical
bursts records across a case study network. The case study area consists of 48 DMAs feeding a total population of
approximately 100,000 people. Total mains length in the area is approximately 1090km, with plastic, metallic and
other materials accounting for around 54, 27 and 19 percent of the total mains length, respectively. Historical bursts
records for year 2003-2013 were used for the analyses. Pressure variation for each point in the area was obtained by
running the hydraulic model of the area for an extended period of time (normally around 1 week) and then
calculating the range of pressure from the minimum to the maximum level of pressure the pipe has been exposed to.
Table 1 shows the distribution of system exposure (for all material types and CI mains) to various ranges of
pressure variation against the percentage of bursts recorded on the mains exposed to a similar pressure variation
range. For all mains material, there are just above 15% of pipes exposed to a pressure variation of 10 mH2O or more.
These, however, are accountable for just below 30% of bursts in the area.

Table 1 System exposure vs. distribution of pipe failure for various ranges of pressure variation in all pipe material and CI mains

Pressure Variation % of system % of bursts in % of system % of bursts

range (mH2O) exposure-all material all material exposure-CI mains in CI mains
0-5 56.5% 43.0% 40.1% 28.0%
5-10 24.7% 28.5% 26.4% 30.5%
10-15 8.4% 10.4% 15.7% 19.5%
15-20 5.2% 10.3% 13.8% 17.7%
20-25 3.5% 7.6% 2.3% 4.3%
A normalized distribution (number of bursts per length of main) across various ranges of pressure variation for
metallic pipes is presented in Figure 6. This represents a positive correlation between rate of failure and exposure of
pipes to higher ranges of pressure variation.
5.0
Mains bursts / km

4.0
y = 0.6628x + 0.285
3.0 R² = 0.9424
length

2.0
1.0
0.0
0-5 5-10 10-15 15-20 20-25
Pressure Variation range (mH2O)

Figure 6 Normalized distribution of bursts vs. pressure variation for metallic pipes

Pressure variation vs. failure mechanisms: Loadings imposed through hydraulic dynamic condition and varying
pressures will cause hoop stresses which in turn contribute to the development of longitudinal failures. The case
study presented below investigated the potential correlation between the varying pressure and mains failure
mechanisms.
The case study covers an area with approximately 4600km length of main. Table 2 shows the distribution of
mains length exposed to various ranges of pressure and the corresponding number of longitudinal or circumferential
failures.
 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262 261

Table 2 Distribution of historical burst records and failure mechanisms vs. pressure variation range

Pressure variation Mains length Longitudinal Circumferential

range (m H2O) (m) No of bursts Burst/ mains length No of bursts Burst/ mains length
0-5 2,210,027 123 0.056 373 0.169
5-10 1,165,037 69 0.059 235 0.202
10-15 686,881 19 0.028 124 0.181
15-20 277,848 16 0.058 48 0.173
20-25 132,454 9 0.068 21 0.159
>25 212,366 25 0.118 33 0.155
As shown in Figure 7(a), for longitudinal failures, there is a positive correlation between pressure variation
ranges and rate of failure. This is in line with expectations that pressure fluctuations impose additional hoop stresses
causing longitudinal cracks growth and eventually causing mains failure.
0.15 0.25
y = 0.0105x + 0.0277 y = -0.0058x + 0.1934
R² = 0.4404 R² = 0.4201
0.10 0.20

0.05 0.15

0.00 0.10
0-5 5-10 10-15 15-20 20-25 >25 0-5 5-10 10-15 15-20 20-25 >25
Pressure Variation range (mH2O) Pressure Variation range (mH2O)
(a) (b)

Figure 7 Failure rate vs. pressure variation range for longitudinal (a) and circumferential (b) bursts

4. Conclusions

There have been various studies on the failure mechanisms in pipes and their root causes. While some of these
factors have been investigated to a reasonable level, there is generally a lack of understanding of failure mechanisms
at a fundamental level and impacting factors. Review and analyses of historical burst records for a 10-year period in
a large water company in the UK highlighted that there are numerous mains failure events that cannot be explained
by factors traditionally believed to be the main causes of failure. Due to the impact from various factors and
uncertainty in the nature of these factors, failure mechanisms are very complex. Thus, failures may not be easily
attributed to a single or a combination of factors. As shown in the analyses above, there have been a large number of
mains failures that have occurred in conditions where usually mains failures are not expected (e.g. moderate climate,
low/non-corrosive surrounding soil, soil with low fracture potential factors, etc.). The findings suggest that there
might be other impacting factors that, in certain circumstances and/or when combined with other factors, could
contribute to and/or cause pipe failures.
Previous studies, industry investigation and the extended experimental programme undertaken in this work have
highlighted the extent of hydraulic instabilities in the system. While occasional pressure transient events
downstream of certain operational assets with no surge protection measures might not be unusual, it is found that
hydraulic instabilities could be much more common across the network. In addition to usual sources of pressure
transient events, such as pumps and valves, large commercial consumers and operational activities on the network
could also be important sources of such events.
The cyclic loadings due to such events, with sufficient frequency and magnitude, could result in development of
fatigue failure. Existing flaws and cracks will significantly increase the likelihood of such development.
Manufacturing flaws, cracks developed during installation, areas with corrosions and other imperfections could act
as existing cracks, hence bypassing the crack initiation stage in the fatigue process. In addition, these result in areas
of stress concentration that can speed up the crack propagation stage, therefore reaching the failure stage in a
reduced number of cycles. The study of various cases highlights that while many of the pipes might still have some
remaining life, localised development of cracks could fail the main well before the end of their service life.
262 Hossein Rezaei et al. / Procedia Engineering 119 (2015) 253 – 262

Analyses shown in above sections illustrate some positive correlation between pressure variation in the pipelines
and the rate of historical failures. Additional studies are required to investigate the impact of loadings imposed by
pressure transient events on the network and the impact these might have on the development of mains failure.
Currently, none of the deterioration modelling concepts consider dynamic hydraulic condition. With recent
technological development in logging technologies and data management, measuring the operational factors more
accurately and obtaining insight into hydraulic behaviour of the networks become more feasible. This will result in
further development in understanding of the fundamentals of failure mechanisms and eventually in reducing the
frequency and cost of mains failures.

Acknowledgments

The authors would like to thank the EPSRC and Severn Trent Water for their support.

References
[1] A. Hoskins, I. Stoianov, InfraSense: A Distributed System for the Continuous Analysis of Hydraulic Transients, Procedia Eng. 70 (2014)
823–832. doi:10.1016/j.proeng.2014.02.090.
[2] J.M. Makar, R. Desnoyers, S.E. McDonald, Failure modes and mechanisms in gray cast iron pipe, in: Proc Int Conf Undergr. Infrastruct.
Res., Citeseer, 2001: pp. 303–312. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.4.2771&rep=rep1&type=pdf (accessed June
1, 2015).
[3] B. Rajani, Y. Kleiner, Comprehensive review of structural deterioration of water mains: physically based models, Urban Water. 3 (2001)
151–164. doi:10.1016/S1462-0758(01)00032-2.
[4] Y. Kleiner, B. Rajani, Comprehensive review of structural deterioration of water mains: statistical models, Urban Water. 3 (2001) 131–
150. doi:10.1016/S1462-0758(01)00033-4.
[5] A. Paradkar Brahmanand, An Evaluation of Failure Modes for Cast Iron and Ductile Iron Water Pipes, University of Texas at Arlington,
2012. https://uta-ir.tdl.org/uta-ir/handle/10106/11660.
[6] J.B. Boxall, A. O’Hagan, S. Pooladsaz, A.J. Saul, Unwin, Estimation of burst rates in water distribution mains, Proc. ICE - Water Manag.
160 (2007) 73–82. doi:10.1680/wama.2007.160.2.73.
[7] O.G. Pritchard, S.H. Hallett, T.S. Farewell, Soil Corrosivity in the UK–Impacts on Critical Infrastructure, (2013).
[8] O.V. Hembara, O.E. Andreikiv, Effect of hydrogenation of the walls of oil-and-gas pipelines on their soil corrosion and service life,
Mater. Sci. 47 (2012) 598–607. doi:10.1007/s11003-012-9433-x.
[9] N. Jarvis, J. Hollis, P. Nicholls, T. Mayer, S. Evans, MACRO—DB: a decision-support tool for assessing pesticide fate and mobility in
soils, Environ. Model. Amp Softw. (1997) 251–265. doi:10.1016/S1364-8152(97)00147-3.
[10] M. Nishiyama, Y. Filion, Review of statistical water main break prediction models, Can. J. Civ. Eng. 40 (2013) 972–979.
doi:10.1139/cjce-2012-0424.
[11] M. Doleac, S. Lackey, G. Bratton, Prediction of time-to failure for buried cast iron pipe, in: AWWA Annu. Conf., 1980.
[12] B. Rajani, J. Makar, A methodology to estimate remaining service life of grey cast iron water mains, Can. J. Civ. Eng. 27 (2000) 1259–
1272.
[13] D. Linkens, N.K. Shetty, M. Bilo, A probabilistic approach to fracture assessment of onshore gas-transmission pipelines, Pipes Pipelines
Int. 43 (1998) 5–16.
[14] C.D.D.H. Uri Shamir, An Analytic Approach to Scheduling Pipe Replacement, J. Am. Water Works Assoc. 71 (1979) 248–258.
[15] A.P. Thomas M. Walski, Economic analysis of water main breaks, J. Am. Water Works Assoc. 74 (1982) 140–147.
[16] L.D. McMullen, Advanced concepts in soil evaluation for exterior pipeline corrosion, in: Proc. AWWA Annu. Conf., 1982.
[17] P. Jacobs, B. Karney, GIS development with application to cast iron water main breakage rate, in: 2nd Int. Conf. Water Pipeline Syst.,
1994.
[18] L.A. Jeffrey, Predicting urban water distribution maintenance strategies: a case study of New Haven, Connecticut, Massachusetts Institute
of Technology, 1985.
[19] A. Constantine, J. Darroch, R. Miller, Predicting Underground Pipe Failure, Australian Water Works Association, 1996.
[20] A.K. Deb, Hasit,Y, Grablutz J. F. M., Herz, R. K, Quantifying future rehabilitation and replacement needs of water mains, AWWA
Research Foundation, Denver, CO, 1998.
[21] K. Mavin, Predicting the failure performance of individual water mains, Urban Water Research Association of Australia, 1996.
[22] G. Kabir, S. Tesfamariam, A. Francisque, R. Sadiq, Evaluating risk of water mains failure using a Bayesian belief network model, Eur. J.
Oper. Res. 240 (2015) 220–234. doi:10.1016/j.ejor.2014.06.033.
[23] M. Rosenfeld, J. Kiefner, Basics of Metal Fatigue in Natural Gas Pipeline Systems—A Primer for Gas Pipeline Operators, 2006.
[24] Y.-L. Lee, R. Hathaway, J. Pan, Fatigue Testing and Analysis: Theory and Practice, Butterworth-Heinemann, Amsterdam ; Boston, 2004.
[25] A.R.D. Thorley, Fluid Transients in Pipeline Systems: A Guide to the Control and Suppression of Fluid Transients in Liquids in Closed
Conduits, Second edition, Professional Engineering Publishing, 2004.