Thesis AndreArsenio TUD Final
Thesis AndreArsenio TUD Final
Thesis AndreArsenio TUD Final
Push-fit Joints
Proefschrift
Samenstelling promotiecommissie:
Typeset in LATEX
Printed by Gildeprint Drukkerijen
Cover by Miguel Carvalho / www.miguelcarvalho.eu
Original photo of Pegões Acqueduct by Duarte Fernandes Pinto
ISBN 978-94-6186-217-4
Copyright © 2013 by André Marques Arsénio / andre at abrigo dot cc
Summary
Background
Despite the role played by joints in the failures occurring in drinking water networks,
scientific literature has focused on barrel failure. Therefore, it can be assumed that
the significance of these appurtenances has been disregarded.
The objective of this work is to create a procedure to predict the remaining
lifetime of PVC push-fit joints. PVC was selected due to it being the most used
material in the Netherlands and resistant to corrosion which facilitates condition
assessment due to the absence of tuberculation. Finally, working with PVC, for
example, in the laboratory, does not incur the inherent safety/health risks as with,
for example, AC.
v
vi Summary
Several NDE tools were surveyed whereby the most promising for the assessment
of PVC push-fit joints were selected and tested in the laboratory including ultra-
sound, CCTV and Panoramo® with CCTV consistently considered as the best,
delivering both accurate and reproducible results.
Finally, for a rubber ring in good condition, leakage/intrusion can only be ex-
pected at bending angles above 10° and with complete pull-out of the pipe from the
joint. Such extreme angles have not yet been detected in the field (Chapter 3).
Concluding remarks
This thesis presents a four-step procedure to predict failure in PVC push-fit joints.
The first step is selecting the best candidates for condition assessment, for example,
employing a risk map that indicates failure-prone areas. The second step is to
evaluate the joint condition through gap sizing by using non-destructive evaluation
equipment (e.g. CCTV). The third step is to utilize the IJC, having the data
retrieved from the condition assessment as input, and evaluating the condition of
the pipe and the joints. A decision can subsequently be made to either replace the
entire pipe or only a section of it or to schedule another inspection. The fourth
and final step is to recommence the process: the lifetime prediction procedure is
an iterative process, and the same pipe and joints should be inspected several times
viii Summary
throughout their lifetime. When their condition is below a certain threshold, they
should be replaced.
Samenvatting
Achtergrond
Verbindingen in drinkwaterleidingen spelen een belangrijke rol in het geregistreerd
falen van het drinkwaterleidingnet. Ondanks dit aanzienlijke aandeel, concentreert
de wetenschappelijke literatuur zich op het falen van de leidingen. Hierdoor kan
worden aangenomen dat de rol die verbindingen spelen bij storingen is onderschat.
Het doel van dit onderzoek is het opstellen van een procedure waarmee de
resterende levensduur van PVC spie-mof verbindingen kan worden voorspeld. PVC
is gekozen omdat dit het meest voorkomende materiaal is in Nederland. Daar-
naast is PVC immuun voor corrosie. Dit vergemakkelijkt de conditiebepaling van de
verbinding. Tevens vereist het werken met PVC, bijvoorbeeld in een laboratorium,
geen speciale veiligheids- of gezondheidsrisico’s in tegenstelling tot bijvoorbeeld AC.
ix
x Samenvatting
315 mm) en dicht bij de grenswaarde die is bepaald door de Nederlandse fabrikanten:
6° voor beide diameters.
Ten tweede, de rol die de insteekdiepte speelt bij de toename van de stijfheid van
de verbinding laat zien dat de leidingdelen niet volledig ingestoken moeten worden
bij installatie.
Tenslotte, als de rubberring in goede conditie is kun je alleen lekkage of indringing
verwachten bij buigingshoeken van meer dan 10° en bij complete uittrekking van de
leidingdelen van de verbinding. Zulke extreme buigingshoeken zijn niet waargenomen
in het veld (Hoofdstuk 3).
Afsluitende opmerkingen
In dit onderzoek is een faal-voorspellings-procedure voor PVC spie-mof verbindingen
gepresenteerd. De procedure kan in vier stappen worden geı̈mplementeerd. De eerste
stap is het selecteren van de beste kandidaten voor het bepalen van de conditie, door
bijvoorbeeld een risicokaart te gebruiken zoals hiervoor beschreven. De tweede stap
is het bepalen van de conditie van de verbinding door de spleetwijdte te bepalen
met behulp van de CCTV. De derde stap is het met behulp van de IJC, gebaseerd
op de gegevens van de conditiebepaling, de leiding te classificeren. Dan kan de
beslissing genomen worden om de leiding geheel of gedeeltelijk te vervangen of om
een volgende inspectie te plannen. De vierde en laatste stap is het opnieuw starten
van het proces. Het voorspellen van de restlevensduur is een iteratief proces waarin
dezelfde verbinding verscheidene keren geı̈nspecteerd wordt gedurende de levensduur.
Als de conditie een zekere grenswaarde overschrijdt, kan deze worden vervangen.
...
Yeah, do it right
And head again into space
So you can carry on and carry on
And fall all over the place
...
From early on, I indicated that I would obtain a Ph.D., but that was even before
beginning secondary school. After my first year at university, I was certain that I
did not want to otbain a Ph.D. - in fact, I was not even sure if I wanted to complete
my current degree. Now, approximately twenty years later, I am, indeed, a Ph.D.
candidate which proves that it is the first impression that matters (even when it is
ill-informed). Of course, I was not alone in this endeavour, and many people have
assisted me throughout the years.
First and foremost, I would like to thank my sponsors which include the Dutch
Drinking Water Companies that funded this project through the joint research pro-
gram (BTO) and Wetsus that funded the project through the TTIw program. With-
out their assistance, none of this would have been achievable.
I owe much to Jan Vreeburg who always believed in me and in my capacities
- many times, even more than myself. Jan certainly was not the most available
supervisor† but was, beyond any doubt, always supportive both from a technical
and, especially, from a personal perspective. I also wish to thank my promotor,
Luuk Rietveld, for all of the assistance and help provided throughout the final year
of my project. It is clear now that the arrival of Luuk as a second supervisor was a
significant encouragement to complete everything (almost) on time. I also wish to
acknowledge my first promotor, Hans van Dijk, for his role during the first years of
the project.
It was a wonderful experience to write articles with so many different co-authors.
I would like to thank them all for bringing their knowledge and opinions and helping
to enrich my work.
It has been a delight for me to work with everyone at the Water Infrastructure
group of KWR: George, Nellie, Ad, Martin, Ralph, Jos, Peter, Marcel, Mirjam,
Cláudia, Joost, Andreas and Ronald, thank you for being wonderful colleagues. Ilse:
thank you for being, from day one, the most encouraging and helpful person that has
ever been involved in the project and for pushing me to plan and schedule my work
and be more organized - I can assure you that not all was in vain and that something
did remain imprinted in my mind (please, believe me). Hendrik, thank you for all of
the help given to wire the monitoring set-up and to retrieve the data. Thank you,
† Maybe an available supervisor is the perfect oxymoron, a chimera, the Moby Dick of research.
xvii
xviii Preface
Kim, for all of the help given with GIS. One final word must to go to Irene: thank
you for being the best roomy that I have ever had and thank you for introducing me
to Mr. Fuzzy and all his fuzzyfied friends - I will never see fuzzy-logic in the same
manner as before (and thank you for that as well).
I would like to also thank the following people at the water companies: Eelco
Trietsch and Jan Pot at Vitens; Rob de Jong and Rob de Bont at Dunea; Marcel
Wielinga, Peter Schaap and Peter Horst at PWN and Henk de Kater at Evides;
and Petra Holzhaus at WMD for always being available and cooperative. To all
the pipe-fitters at the water companies that somehow helped me throughout this
project, in sunny and in rainy days: from working with you, I understand how one
builds a country such as the Netherlands.
Once again, I wish to acknowledge all of the members of the TTIw program
table for all of the meetings, the valuable input and for the thorough and insightful
reviews on my papers.
At DYKA, I would like to thank Mr. Freddie Bouma and Mr. Henk Meerman for
affording me the opportunity to use their installations. Thank you, Jarig Bangma,
for all of the support and comments derived from your deep practical knowledge.
Finally, I wish to thank all of the people at DYKA’s workshop for making me feel
welcomed at all times - it was always a pleasure to travel to Steenwijk.
I wish to thank everyone at TU Delft - Mieke, Jennifer, Anouk, Jorge, Dara, Sam,
Diana, Maria, Ran, David and Gang - and everyone at KWR - Roberto, Andreas,
Helena, Sara and Diego - for these great four years.
To everyone at Shelter Productions: Baka, Mike‡ , Sofia, Nuno, Rodrigo, António
and Seta, thank you for reminding me daily that true friendship overcomes both
barriers and distance.
Mum and Dad: thank you for always having encouraged even my strangest ideas
and having taught me to follow my dreams and to be happy. Once again: thank you
very much.
Mãe e Pai: obrigado por me terem apoiado sempre de maneira incondicional e
desde cedo me terem incentivado a seguir os meus sonhos. Uma vez mais: obrigado.
I wish to dedicate a final word to my lovely partner, Elisa, a central pillar in my
life. Without her support, love and comprehension, I would not have achieved any
of this.
Summary v
Samenvatting ix
Preface xvii
1 Introduction 1
1.1 The Dutch drinking water network . . . . . . . . . . . . . . . . . . . 2
1.1.1 Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Non-revenue water . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Failure rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 What is asset management? . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Non-destructive assessment . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Modeling remaining asset life . . . . . . . . . . . . . . . . . . . . . . 6
1.5.1 Deterministic modeling . . . . . . . . . . . . . . . . . . . . . 7
1.5.2 Statistical models . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5.3 Physical probabilistic models . . . . . . . . . . . . . . . . . . 8
1.5.4 Soft-computing or artificial intelligence-based models . . . . . 8
1.6 Mains failure databases . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.6.1 NMFD of the United Kingdom . . . . . . . . . . . . . . . . . 9
1.6.2 NMFD of the United States . . . . . . . . . . . . . . . . . . . 9
1.6.3 NMFD of the The Netherlands . . . . . . . . . . . . . . . . . 10
1.7 PVC in operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7.1 Aging & degradation of PVC . . . . . . . . . . . . . . . . . . 11
1.7.2 PVC joints in the field . . . . . . . . . . . . . . . . . . . . . . 11
1.8 Problem definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.9 Layout of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
xix
xx Contents
Bibliography 129
Glossary 139
Il Principe di Salina
These words, derived from the classic film Il Gattopardo (1963) by the Italian
master Luchino Visconti, can be roughly translated as something had to change, so
everything could stay the same. The film was localized during a troublesome period,
the times were changing and Burt Lancaster, the Prince of Salina, had to make a
decision.
Everyday, throughout the world, asset managers must also make decisions: “re-
place or repair?”, “inspect or not?”. Something must change so that everything can
remain the same, i.e., the supply of safe drinking water continues. If nothing changes,
assets will fail more frequently, and this will imply additional costs for water com-
panies. These costs, whether direct, indirect or social (Gaewski and Blaha, 2007;
Makar and Kleiner, 2000) may, in time, become financially unsustainable.
In order to assist in decision making and help utilities to minimize their expenses,
the objective behind this thesis is to develop a procedure that can be employed
to predict the remaining lifetime∗ of PVC push-fit joints. This work comprises
both condition assessment techniques for joints and the development of a lifetime
prediction model. These topics will be briefly introduced in this first chapter that
addresses the Dutch drinking network as the beginning point in order to discuss
general concepts of asset management (AM).
1
2 1 Introduction
Figure 1.1: Length of drinking water network per pipe material in the Netherlands.
In the country, the percentage of PVC has been increasing since the
1950s. The percentage of AC, after peaking in the 60s, as has been
decreasing as has that of CI. The use of polyethylene (PE) used for
house connections and ductile iron (DI) used for pipes beneath heavy-
traffic roads, have been on the rise. Adapted from Geudens (2012).
this (UKWIR, 2006). The authors concluded that the Dutch companies have a
lesser leakage percentage for the following reasons (Ofwat, 2007):
• Low operating pressures due to the flat terrain and the tall buildings being
equipped with their own pumps;
• A newer, post-war infrastructure system when compared with England and
Wales and typically made of non-corrosive PVC;
• The mains tend to be situated under footpath paving blocks and in sandy soils
which signifies that leaks cause the pavement to subside which facilitates easy
leak location and easy repair access;
• Existence of fewer joints as a single connection generally supplies a number of
buildings; and
• Quick repair of reported leaks.
f
λ= P (1.1)
L × ∆t
Vloerbergh and Blokker (2010) indicated that, for the Netherlands, the overall
failure rate (for all pipe materials together) was in the range of 0.06 #.km-1 .year-1 .
This demonstrates that the Dutch figures are among the lowest worldwide.
Both the reduced failure rate and diminished leakage figures demonstrate that,
although natural environment plays a role, proper network management is crucial,
i.e., materials used, location of the pipes and response to failures.
The EPA (2002) estimated the gap between the projected need and current
spending for clean water and drinking water infrastructure over the next 20 years
in the US. The authors indicated that the clean water capital payment could be
between $73 billion and $177 billion with a point estimate of $122 billion.
Additionally, according to Folkman (2012), over 8% of the installed water mains
in the US and Canada are beyond their expected service. The author argues that,
given the rapid increase on pipes requiring immediate replacement, improved AM
is essential so that all utilities can survive this trend. In fact, Kirby et al. (2006)
show that AM allows utilities to cut operating costs, reduce capital expenditure and
improve the level of service. These examples demonstrate the importance of AM
not only in maintaining a productive operational level but also to avoid disastrous
situations.
Nevertheless, AM is a broad field, and there is not one individual recipe for
success. Hudson et al. (1997) identified various possible approaches:
1. Select the best pipe for inspection. This can either be accomplished by em-
ploying practical knowledge - areas where the pipe-fitters are aware that the
failure rate is greater than average - or be a data-driven decision as discussed
in Kleiner and Rajani (2008). An alternative approach will be presented in
Chapter 7;
2. Inspect the pipes using the most appropriate inspection tool. This topic will
be extensively discussed in Chapter 3;
3. With the information aggregated during the inspection, either:
(a) If the condition of the asset is below a previously defined threshold, re-
pair/replace;
(b) Otherwise, predict the remaining lifetime, indicated as t, of the asset
exploiting an appropriate model (Section 1.5); and
4. Finally, schedule a new inspection after a period, t0 , shorter than t. The actual
condition of the asset at t0 can be exploited to re-calibrate the model used in
step 3b.
Figure 1.2: Scheme of a proactive AM approach. 1) pipes are selected for inspection;
2) the inspection is performed, and the condition of the pipes is evalu-
ated; 3a) if the pipes are in poor condition, they are repaired/replaced;
3b) if they are in poor condition, their expected remaining-lifetime is cal-
culated employing an appropriate model. Finally, 4) a new inspection is
scheduled, and the output of the pipe condition is utilized to calibrate the
prioritization model used in 3b.
More recently, Marlow et al. (2009) reviewed the different approaches to model
the remaining lifetime of assets. The authors distinguished the various models into
deterministic, statistical, physical and soft-computing or artificial-intelligence based
approaches.
Empirical deterministic
A classic empirical deterministic model was presented by Shamir and Howard (1979):
regression analysis was used to create a break prediction model that relates a pipe’s
breakage to the exponent of its age (Kleiner and Rajani, 2001) (Equation 1.3).
† Forexample, temperature or precipitation series; or soil properties (e.g. soil resistivity, chlo-
rides, pH) (Rajani and Kleiner, 2001)
1.6 Mains failure databases 9
Several countries around the world have attempted to create national MFDs
(NMFDs). Such a database stores all available failure data in a country to pro-
duce a larger and more significant source of data. Nevertheless, establishing such
a database is not a straightforward process as, for example, each utility utilizes its
own registration procedure.
To the author’s knowledge, worldwide, only the UK and the Netherlands have
operational NMFDs.
Initially, 20-30 utilities were contacted to participate in the project. All utilities
were from medium to larger cities (not the top-10 largest), and only two utilities
participated. The explanations for such low participation were (Hodgins, 2013):
10 1 Introduction
Nevertheless, Hodgins (2013) argues that one of the two participating utilities,
despite the necessary work to compile and upload the data, realized that the system
afforded them an improved understanding on i) the type of data and the data-quality
necessary for such project, and ii) the analysis process and overall support for the
utility in repair/replace/maintenance decisions.
A report including all of the conclusions of this project and authored by Neil
Grigg (Colorado State University) is due to be published by the Water Research
Foundation.
1. Define the failure modes of PVC joints. This will cast light on the origins of
joint failure and determine how joint condition can be assessed;
2. Develop a condition assessment method. The failure modes will be the begin-
ning point in creating the condition assessment method;
3. Choose the best NDE tool for this application. This tool will be selected from
commercially available tools; and
4. Implement a lifetime prediction procedure. Since not all pipes can be inspected,
this procedure will assist in precisely defining which pipes in a network should
be inspected first.
This chapter was the first section of this work and was initiated by Ilse Pieterse.
Nevertheless, due to being the foundation of the thesis, it was continuously updated.
The objective was to review scientific and technical literature and collect as much
information about joint failure as possible.
In fact, “joint failure” has a very broad definition. Two situations, i.e., a pipe
that is completely pulled-out from the joint or a joint that is fractured, can both
be registered as “joint failure” (Section 1.6) by a pipe-fitter. Given the difference
between both failures, it is important to register more information about the failure
- the mode. This additional information is valuable: knowing the areas inside of a
distribution network where certain failure modes occur or which type of materials
(or of what age) fail in specific ways is of the utmost importance both for scientists
and network managers.
Nevertheless, it is recognized by the author that improving failure registration
is not an easy feat: filling out a 10-20 question form on a -5°C day after having
spent 2 hours fixing a burst pipe burst is hardly a priority. Nevertheless, creating a
framework to improve failure registration is an initial step.
In this chapter, seven failure modes for joints are presented and discussed.
Arsénio, A. M., Vreeburg, J. H. G., Pieterse-Quirijns, E. J., and Rosenthal, L. (2009b). Overview of
failure mechanism of joints in water distribution networks. In Boxall, J. and Maksimović, C., editors,
Computing and Control in the Water Industry (CCWI), pages 607–612, Sheffield (UK). CRC Press
Arsénio, A. M., Pieterse-Quirijns, I., Vreeburg, J. H. G., de Bont, R., and Rietveld, L.
(2013c). Failure mechanisms and condition assessment of PVC push-fit joints in drinking water
networks. Journal of Water Supply: Research and Technology-AQUA, 62(2):78
15
16 2 Failure modes of push-fit joints
2.1 Introduction
The reason for this work was thoroughly motivated in Section 1.7.2: it was demon-
strated that joints play a major role in the failures registered in a network comprised
of PVC pipes. However, given the focus of literature on pipe barrel deterioration, it
can be assumed that the role played by PVC pipes may have been disregarded.
During an NDE inspection and/or afterwards during analysis, operators and an-
alysts should focus on detecting distress indicators on the pipes or joints. According
to Rajani and Kleiner (2004), these are forms of deterioration that have not yet
led to pipe or joint failure. Among the distress indicators presented by the authors
are cracks in cement, plastic or metallic pipes; corrosion pits in metallic pipes; and
broken prestressing wires in prestressed concrete cylinder pipes. However, no dis-
tress indicators were defined for PVC joints. This might be explained by the fact
that PVC has only been employed on a large scale since 1960 and that long-term
deterioration mechanisms are not well documented for PVC mainly because the de-
terioration occurs slowly (Rajani and Kleiner, 2001). Breen et al. (2004) also argued
that the lifetime of PVC material that is well processed, well installed, and applied
under relative mild service conditions will exceed 50 years and even possibly 100
years.
Therefore, the objectives of this chapter are identifying the appropriate distress
indicators - failure modes - for PVC joints. This chapter is also the beginning point
for the development of the inspection procedure presented in Chapter 3.
gaskets keep the system sealed, the pipes are separated inside of the joint by a ring.
For smaller diameters (< 250 mm), this ring is a loose piece that separates both pipes
inside the joint (Figure 2.2, in the center in black). For larger diameters, the ring
is molded with the joint. This ring creates a gap between the pipes, ensuring that
the pipe ends are not touching each other and are symmetrically situated inside the
joint during installation. The sizing of this gap is the principle behind the assessment
procedure presented in Chapter 3.
Figure 2.1: Schematic of a bell-and-spigot joint. In this system, the straight spigot
end of one section is inserted in the flared-out end of the adjoining sec-
tion. The rubber gasket (grey circles) seals the system.
Figure 2.2: Longitudinal cut of a double-socket DN110 PVC joint. At the center, a
ring (black) separates both of the two pipes inside the joint. For joints
below 250 mm, the ring is a stand-alone piece. For larger diameters,
the joint is molded with a ring. At each end, a rubber gasket (black)
keeps the system sealed. In this figure, on each side of the joint, two
white/light-blue rings are visible on top of the rubber gaskets. For joints
of this diameter, this is the system to fix the rubber gaskets in place; for
joints above 250 mm, the rubbers gaskets are fixed to the joint.
18 2 Failure modes of push-fit joints
• Uneven bedding support. This can result from unstable foundation materi-
als, uneven settlement due to overexcavation and non-uniform compaction or
scouring, for example, due to a leaking joint (Rajani and Tesfamariam, 2004);
• Differential settlement. This effect is especially important in wastewater net-
works when a pipe is rigidly connected to a manhole and both structures settle
at different rates; and
• Ground movement. This can occur due to earthquakes or expansive soils. Frost
heave can also play a role.
Figure 2.3: Joint bending. Contact points between the pipe and the inner-joint wall
are annotated.
Seismic movements of the soil can also induce vertical displacement of a pipe by
ground rupture (Tan and Yang, 1988). Interface between two media with different
properties can also cause vertical displacement of the pipes.
Figure 2.4: Schematic of vertical displacement where two joints subjected to joint
bending and two joints subjected to axial displacement are highlighted
and annotated.
Vertical displacement of the pipe is bound to induce joint bending at the inflection
points and axial displacement (Figure 2.4). Moreover, it is possible that torsion plays
a role due to soil movements. Vertical displacement is a composite failure mode since
various joint failure modes such as angular deviation, axial displacement and torsion
are involved depending upon the magnitude of vertical displacement. Nevertheless,
vertical displacement is included in the overview as a separate failure mode since its
cause is different. Moreover, the displacement can be easily measured, which makes
the cause for this type of failure easy to distinguish.
Figure 2.5: Alignment of a joint during axial displacement. The pipe on the right-
hand side is almost pulled-out from the rubber-gasket. The images are
not to scale.
In fact, with an axial contraction of the pipe, a complete pull-out of the pipe
from the inside of the rubber ring can occur. This leads to joint leakage. The pipes
22 2 Failure modes of push-fit joints
can also move toward each other (e.g. pipe expansion due to temperature increase),
and this may lead to contact between the pipe and joint and between the two pipes
and to joint/pipe breakage. The length of a 10 m PVC barrel can vary 8 mm over
a temperature variation of 10 ℃ to 25 ℃, considering a linear expansion coefficient
for PVC of 54 m.m-1 .K-1 (AWWA, 2002).
Torsional tests demonstrate an increase of the torque at thr failure with the
increase of the burial depth and the increase of the pipe diameter (Singhal, 1984).
No information is known on this type of failure mode of joints under the condition in
the Netherlands. It might be expected that, with a rigid connection in the pipeline,
torsion leads to a longitudinal or spiral breakage of the pipe and joint. On the other
hand, torsion of the pipe in a push-fit joint might cause the slippage and leakage of
the joint due to separation of the segments.
pipes and in the joint (Bailey and Kaufmann, 2006; Lange, 2003; Misiunas, 2005;
Rajani and Kleiner, 2001). Moreover, due to the differences in stiffness between the
pipe and the joint, hoop stress will have more influence on the pipe, which may lead
to leakage of the joint.
2.5 Conclusions
Seven failure modes of joints are derived from literature and from practice. For
the first time, to the author’s knowledge, the failure modes of joints are surveyed,
leading to a uniform nomenclature of the topic of joint failure.
This overview elucidates the importance of ground movement in the failure of
joints. The overview is essential for the water companies as well as for scientific
purposes. First, it facilitates and improves the failure registration of the water
companies which is currently incomplete and ambiguous. The improved registration
will determine the most significant joint failure mode in the Netherlands. From a
scientific perspective, the modes are crucial in modeling the failure of joints as a
function of a great number of material-related and environmental variables. The
models render insight in the failure modes which will result in limits for the joint
displacement before failure occurs.
Joints are considered to be particularly affected by bending that can originate,
for example, from differential soil movement. Bearing this in mind, an inspection
procedure for joints was developed. This procedure is presented and discussed in
the next chapter.
chapter 3
Non-destructive evaluation of PVC push-fit
joints†
This chapter addresses a very important component of this thesis: the selection of
the most accurate and reproducible non-destructive evaluation (NDE) tool for the
condition assessment of joints in the field. Following a literature review, three NDE
tools were selected: ultrasound, closed-circuit television (CCTV), and Panoramo® .
Tests were projected in order to characterize the tools on accuracy and repro-
ducibility. The work began in the laboratory at KWR. Afterwards, the best two
tools were taken to an AC research pipe of Vitens in the north of the Netherlands
and tested side-by-side. Finally, the best tool was thoroughly tested in the field in
three field inspections of PVC pipes.
The chapter is divided into two main sections. The first section presents the gen-
eral concepts and the second section discusses all of the characterization performed
to the three tested NDE tools.
Arsénio, A. M., Vreeburg, J. H. G., van Doornik, J., Dijkstra, L., and van Dijk, H. (2010).
Assessment of PVC Joints Using Ultrasound. In Water Distribution Systems Analysis (WDSA),
Tucson (Arizona, USA). ASCE
Arsénio, A. M., Vreeburg, J. H. G., de Bont, R., and van Dijk, H. (2011). Real-life inline
inspection of PVC push-fit joints using NDE equipment. In Leading Edge on Strategic Asset
Management (LESAM), Mülheim an der Ruhr (Germany)
Arsénio, A. M., Vreeburg, J. H. G., de Bont, R., and van Dijk, H. (2012b). Real-life inline
inspection of buried PVC push-fit joints. Water Asset Management International, 8(2):30–32
Arsénio, A. M., Vreeburg, J., and Rietveld, L. (2013d). Quantitative non-destructive evalu-
ation of push-fit joints. Urban Water Journal, pages 1–11
25
26 3 Non-destructive evaluation of PVC push-fit joints
3.1 Introduction
Recently, Liu et al. (2012) argued that “information on the current structural con-
dition of individual water mains [and joints], combined with a good understanding
of failure modes and deterioration models, can greatly enhance the ability of wa-
ter utilities to manage their assets in a cost-effective manner”. It could be added
that this information should be as accurate as possible, and that, with the currently
available technology, this can only be obtained from in-line inspections with NDE
equipment. During an NDE inspection and/or afterwards during analysis, opera-
tors and analysts should focus on detecting distress indicators on the pipes or joints
(Chapter 2).
Paramount work has been published on the topic of employing NDE tools for
the inspection of drinking and wastewater pipes. Beuken et al. (2011) have pre-
sented business cases regarding the inspection of drinking water pipes. Feeney et al.
(2009); Liu et al. (2012); Makar (1999) and Reed et al. (2006) have all published
work presenting and discussing state of the art equipment and an overview of their
conclusions is depicted in Table 3.1.
According to Rajani and Kleiner (2004), NDE equipment can be exploited on
two levels: first, to provide a snapshot of the pipe condition at a given time in order
to determine if immediate intervention is required. NDE tools can also be deployed
at a later date in subsequent inspections to determine the rate of deterioration.
In 2002, no vendors possessed field-ready methods to examine polymer pipes
such as PE and PVC (Dingus et al., 2002). The authors continued by stating that
polymer pipe inspection was lagging behind NDE for other piping materials. Also
NETTWORK (2002) and Grigg (2006) argued that little, if any, research had been
devoted to the development or investigation of the use of NDE tools for the inspection
of polymeric pipes or their respective jointing components. More recently, Liu and
Kleiner (2012) concluded that in-line visual tools (e.g. CCTV and Panoramo® ) and
ultrasound are those that can be used with the widest range of material types such
as metallic, cement pipes and also “might work” with polymeric pipes.
However, little information is available on the capacities of each of these NDE
tools in order to quantitatively measure the width of joint gaps and on their true
potential for inspection of polymeric pipes.
Visual inspection is usually presented as being inaccurate and prone to inter-
pretation (Dirksen et al., 2013). It is also defined as only able to deliver surface
scanning. Nevertheless, surface analysis is expected to be adequate for gap sizing,
especially since it can be performed in any pipe material. Inspections using ultra-
sound are usually presented as being able to detect a wide array of defects during
the inspection of water pipes. Even so, little information is available regarding this
type of application.
Vangdal et al. (2011) provided an example of a field inspection that was per-
formed using ultrasound. However, the results seem to lack the accuracy necessary
to measure a joint’s gap, and the tool has not yet been tested in polymeric pipes.
Reed et al. (2006) tested ultrasound to perform gap sizing at a laboratory scale using
ductile iron pipes.
Therefore, in order to select the best tool for joint-gap sizing in the field, an
3.2 Inspection procedure 27
Figure 3.1: Left: image of the interior of a 315 mm PVC joint. The gap is signaled
with an arrow. Right: underwater image of the interior of the same PVC
joint obtained with CCTV. Please note the difference between this joint
and the one presented in Figure 2.2: in this situation the ring separating
both joints (white) was molded with the joint and is not a stand-alone
piece.
The shape of the gap can be obtained through measuring its width at different
positions. If the width of a gap is obtained at four different locations, pipe crown,
12 h; invert, 6 h; and both spring-lines, 3 h and 9 h (Figure 3.2), its 3D-orientation
can be determined (Figure 3.3). The angle α between positions i and j is calculated
using Equation 3.1. Furthermore, the current shape of the gape is indicative for
the condition of the joint. The development of the gap over time, observed with
repeated inspection, is indicative for the lifetime expectancy of the joint.
gap widthi − gap widthj
αi−j = 2 × sin−1 (3.1)
2 × Do
Where Do is the outside pipe diameter [mm]. In employing this expression, it
is assumed that both pipe ends were cut perfectly perpendicular to the pipe’s axis.
28 3 Non-destructive evaluation of PVC push-fit joints
Therefore, for the pipe alignment inside of a joint, several situations are hypothe-
sized:
Various alignments of the pipes will correspond to different shapes of the gap,
for example:
1. Ideal alignment: the joint will exhibit the same gap at the four measured
points. It is assumed that the pipe is installed with the joints perfectly aligned.
This assumption is made due to the lack of information regarding the alignment
of pipes following installation;
2. Axial pull-out: the angle for the pair 12h-6h and for the pair 9h-3h is equal to
zero. The threshold distance depends on the joint’s dimensions;
3. Joint bending: the angle calculated with Equation 3.1 is nonzero. For the
12h-6h pair, a positive angle is calculated when the gap at 12 h is wider than
the gap at 6 h. For the 9h-3h, a positive angle is calculated when the gap at
3 h is wider than the gap at 9 h. Joint bending can be decomposed into pure
bending and axial pull-out; and
4. Pipe-bending: a pipe can be considered to be bent when both of the joints to
which it is connected have angles with the same sign.
Figure 3.2: Four in-pipe locations. 12 h: pipe crown; 6 h: invert invert; 3 h and 9
h both spring-lines
Figure 3.3: Downward bent joint and corresponding unfolded image of its interior
(below). This situation is identified by a negative angle α. The schematic
would be similar for a bell-and-spigot joint. The unfolded images show
the shape of the gap for the complete perimeter of the pipe.
a pipe crack) can be detected due to the changes in the travel path of the sound.
An ultrasound system is usually composed of a pulser/receiver and a transducer
(typically piezoelectric). The pulser/receiver is an electronic device that generates
high voltage electrical pulses. Whereby the transducer generates high frequency
ultrasonic energy that propagates in the form of waves. When there is a flaw in the
wave path, a portion of the energy will be reflected back from the flawed surface. The
reflected wave signal is transformed into an electrical signal by the transducer and is
displayed on a screen. Signal travel time can be directly related to the distance that
the signal traveled. With this approach, information regarding the location, size
and orientation of the flaw can be obtained. Additionally, the ultrasound requires a
couplant between the sensor and the test specimen. The couplant is usually a liquid
(e.g. water) that facilitates the transmission of the sound waves sensor into the test
specimen. The principles of the technique are thoroughly discussed by Hellier and
English (2001).
By employing ultrasound, the gap width is measured analyzing both the A and
the B-scans produced by the ultrasound equipment as presented by Reed et al.
(2006). A-scans consist of reflections from the inner and outer walls of the pipe;
B-scans are created with A-scans stacked next to each other. In Figure 3.4, part of a
B-scan is exhibited. This scan was obtained during laboratory tests (top of the pipe).
The ultrasound tool can be imagined at the bottom of the figure with the ultrasound
30 3 Non-destructive evaluation of PVC push-fit joints
waves travelling upwards, detecting the pipe (horizontal lines) after a given time of
flight (Figure 3.4, Y-axis) and reflecting back to the receiving tool. Knowing the time
of flight and the propagation speed of the ultrasound in the element that separates
the tool and the pipe (water), the distance between the pipe and the tool can be
determined. While the tool travels inside the pipe, a calliper arm determines the
in-pipe axial position. Given the information recorded by the calliper arm, the size
of this gap can be determined.
Theoretically, as long as there is good contact at the interface between the sensor and
the material to be inspected, the pipes can consist of different materials (Rajani and
Kleiner, 2004). Nevertheless, according to the same authors, an ultrasound is most
suitable for metallic pipes such as ductile and CI but is not suitable for AC pipes
as the acoustic waves are likely to significantly attenuate in a deteriorated (softened
material) pipe. The authors suggest that it “may work” with polymeric pipes such as
PE and PVC. Misiunas (2005) also argues that this technique is solely applicable to
metallic pipes. Nevertheless, it should be considered that the inspection of metallic
pipes might be problematic. The internal pipe wall must be very clean so that
all of the materials between the sensor and pipe wall have known and well-defined
acoustic properties. In corroded metallic pipes, irregular profiles of tuberculation
make it difficult to transmit ultrasonic waves which are likely to rapidly scatter and
attenuate through the much softer tubercles thus making it difficult to detect and
conduct signal processing of the reflected waves (Rajani and Kleiner, 2004).
Detected failures
It has been reported that defects greater than 3 mm can be detected as well as
pipe deflections and cracks (Feeney et al., 2009). In steel pipes, this technology is
capable to detect metal loss and cracks (Reed et al., 2006). As mentioned previously,
the detected failures are dependant on the frequency of inspection. The ultrasonic
technique can, theoretically, detect 3D geometry of corrosion pits, voids and cracks
(Rajani and Kleiner, 2004).
Possible problems
Makar (1999) argues that the technique is capable of detecting pits, voids and cracks,
although certain crack orientations are much more difficult to detect than others.
In fact, the ultrasonic wave reflects most easily when it crosses an interface between
two materials that are perpendicular to the wave. As an example, cracks that lie
perpendicular to the wave are easily detected, but cracks that lie parallel to the beam
are usually not identified by an ultrasonic examination. Therefore, if an ultrasonic
transducer is located in the center of a pipe, cracks in sewer walls would tend to lie
perpendicular to the beam and, therefore, be difficult to detect. One report suggests
that cracks as fine as 5 mm can be ascertained, if very fine, slow scans are conducted;
however, normal operation will cause such defects to go undetected (Makar, 1999).
3.3 NDE equipment 31
The influence of this downside of the technology in the assessment of plastic pipes
has yet to be studied.
Another disadvantage of using ultrasonic inspection in water filled (or flushed
pipes) is the existence of a significant difference in material properties between water
(or air) and the pipe wall. Therefore, almost all of the sound beams that hit the
wall are reflected away from it, rather than penetrating and reflecting off of defects
inside the pipe wall (Makar, 1999).
In their work, Rajani and Kleiner (2004) argued that the internal pipe wall must
be very clean so that all materials between the sensor and pipe wall have known and
well-defined acoustic properties. However, this might not be a problem with PVC
pipes. Air entrainment, at least during the assessment of sewer pipes, is another
potential problem with this technique, since the air tends to block or scatter the
ultrasonic signal (Makar, 1999).
3.3.2 CCTV
A CCTV consists of a television camera and a method of illuminating the interior of
the pipe mounted on a remotely operated vehicle (ROV) (Costello et al., 2007). An
operator controls the ROV and, on a computer screen, can see the inside of the pipe
in real-time. A CCTV for pipe inspection is usually equipped with software that
measures the distance between any two points on the screen. In the present case,
the gap widths can be calculated. This is the most common inspection technology
in wastewater networks (Figure 3.5).
3.3.3 Panoramo®
Panoramo® is also a visual tool and uses two high-resolution digital photo cameras
with 186° wide-angle lenses fit into the front and rear section of the housing as
exhibited in Figure 3.6. The system is also mounted on a ROV. The images taken
by the two cameras at 5 cm intervals result in an unwrapped 360° image of the pipes’
interior. From these, two different views are generated for analysis in the office - an
32 3 Non-destructive evaluation of PVC push-fit joints
Figure 3.4: B-scan data. The x-axis represents an in-pipe axial position. The y-axis
represents sound travel time (or travel distance). In the middle of the
image, a discontinuity in the horizontal lines between the two vertical
black lines is seen - the gap. Courtesy of Applus RTD.
Figure 3.5: Photography of a CCTV camera used during the tests. Courtesy of MJ
Oomen.
Detected failures
According to Makar (1999), laser scanning systems are capable of finer resolution
than ultrasonic systems and are better at the detection of cracking. Moreover,
the information from the laser scans is automatically recorded and analyzed by a
computer, substantially reducing operator errors. While the initial equipment may
be more expensive than the CCTV system discussed later, the reduced operator
time that is necessary to use the technique may also signify that its operation will
be more economical. It has also been claimed that laser is more effective since finer
defects can be detected and less subjective than CCTV.
Possible problems
Laser inspection can only be employed above water to inspect the sections of a pipe
wall that are, as well, above the water line. Therefore, to assess the entire internal
surface of a pipeline requires the pipe to be taken out of service (Feeney et al., 2009).
In addition to this, laser equipment is not yet commercially available as ultrasound
equipment or CCTV (Makar, 1999).
34 3 Non-destructive evaluation of PVC push-fit joints
Detected failures
According to Makar (1999) and Feeney et al. (2009), GPR has the ability to evaluate
the properties of the soil (e.g. detect subsurface voids). Moreover, with this tech-
nique, voids, rocks and regions of water saturation produced by exfiltration should
all be readily detectable by the technique. If used in-pipe, GPR can also provide in-
formation regarding the physical characteristics of the pipe wall. Maierhofer (2003)
argued that GPR is capable of locating tendon ducts (depths<50 cm), voids and
detachments, and measuring thickness of structures that are only accessible from
one side. According to Slaats et al. (2004) and Mesman and Wielen (2005), GPR is
a suitable, non-destructive method for determining the degree of internal leaching
from AC pipes. Similarly, Smolders et al. (2009) demonstrated that GPR is a ben-
eficial tool for assessing the condition of AC force mains: the authors estimated a
maximum and average deterioration speed of the wall thickness of the pipe.
Possible problems
One of the disadvantages of the technique is the difficulty of measurement interpreta-
tion. Additionally, GPR inspection does not provide extensive information regarding
the condition of the pipe (Misiunas, 2005). Moreover, in a test performed by Makar
(1999), many problems with the surface use of GPR were reported. First, a very high
3.3 NDE equipment 35
level of both false positive and false negative results was obtained. In addition, GPR
results appear to be significantly dependent on local soil conditions, thus making it
necessary to adapt the inspection approach for each location. In addition to this, the
use of GPR was not in any way recommended for locations with clay soils. Finally,
Makar (1999) concluded that surface use of GPR with the current technology would
not be able to detect actual problems in sewer pipes. What is more, the detection
of water saturated soil regions in permanently water saturated soils, for example, in
the Netherlands, might be impossible.
Detected failures
The equipment provides more reliable data and is capable of producing a continuous
profile of the pipe walls. An increased benefit/cost ratio was also anticipated by
some authors (Makar, 1999).
Possible problems
Some of the projects have been abandoned. Despite the aforementioned high ben-
efit/cost ratio, the initial investment cost is high when compared to the cost of the
36 3 Non-destructive evaluation of PVC push-fit joints
individual tools.
Figure 3.7: Ultrasound equipment being introduced inside the pipe during a labora-
tory test (left). The red foam pig is visible inside the pipe. The calliper
is outside the pipe fixed to the ultrasound tool. Schematic of the 8 ultra-
sound probes each numbered consecutively from one to eight (right).
Figure 3.8: Photograph of the ultrasound tool used in the tests attached to a foam
pig. Courtesy of Applus RTD.
Table 3.2: Operational characteristics of the ultrasound equipment used in the labo-
ratory tests. This information was supplied by Applus RTD.
Table 3.3: Overview of the equipment and set-up of the laboratory tests. The in-
spected pipe was PVC DN315 equipped with double-socket push-fit joints.
0 cm to 10 cm;
CCTV 3 33
1 cm steps
Panoramo® 11 positions 2 22
With CCTV and Panoramo® , gap sizing was accomplished at 12 h and 6 h. With
Ultrasound performed gap sizing is performed with eight different ultrasound probes
(Figure 3.7, right). Probe one is facing the 12 h position; probe five is facing the 6 h
position. The tests with the ultrasound and with CCTV were duplicated; the tests
with Panoramo® were triplicated. The total number of laboratory tests conducted
with each of the NDE tools is illustrated in Table 3.3.
A metallic loading frame was constructed for this project to test all NDE equip-
ment. A schematic of the loading frame is provided in Figure 3.9, and photos are
demonstrated in Figure 3.10. The metallic frame supports two 2-meter DN315 PVC
pipes coupled with a double-socket PVC push-fit joint. On the left hand side, the
pipe could be accessed through a flange. On the right hand side, a 45° elbow allowed
the pipe to remain filled with water while open to the exterior. This elbow was also
the way of retrieving the foam pig used during the ultrasound inspection and to grant
access to the pipe’s interior when using the CCTV and Panoramo® . The height of
the pipe support can be varied from 10 cm (0° angle) to 0 cm (5° downward angle)
in 1 cm increments (11 positions). This arrangement allows variation in the gap
width between the two barrels. Fixed points were marked on the pipe’s outer wall
at a fixed location, 0.5 m away from the beginning of the joint (dotted lines in Fig-
ure 3.9). The exact length of pipe inside the joint at the top and at the bottom was
also known. Thus, both the lengths of pipe from the fixed points until the beginning
of the joint and the length of pipe inside the joint were known. During downward
bending while the gap at the top increases, the gap at the bottom becomes smaller;
during upward bending, the opposite occurs. Using these fixed points, the gap width
at the top and at the bottom was calculated for each pipe support height.
The ultrasound and CCTV equipment used on this inspection were the same as in
the laboratory tests.
3.4 Materials & methods 41
Figure 3.9: Schematic of the loading frame. Two 2 m-barrels (1 and 2) are connected
with a double-socket push-fit joint (J). The dotted lines represent the fixed
points on the pipes’ walls used to calculate the joint’s width. These were
located 0.5 m away from the joint. The variable pipe support (below the
joint) allowed the pipes to be bent on the joint from a horizontal position
to 5° (downward) in 11 increments. On the left-hand side, the loading
frame was equipped with a flanged door while, on the right-hand side, a
45° elbow was installed. Image not to scale.
Figure 3.10: Photos of the loading frame. Top left: overview of the frame. Top
right: variable pipe support. Bottom left: 45° elbow. Bottom right:
flanged door.
Figure 3.11: Entry point with flange and hose (left). Exit point with sliding door
above ground level (right).
Equipment
The Panoramo® and CCTV equipment employed for this inspection were the same
as in the laboratory tests and in the field test.
All inspections were duplicated (run 1 and run 2). The runs were re-analyzed in
the office by a different operator (control 1 and control 2). An overview of the
characteristics of the inspected pipes and of the CCTV used in the full-scale tests is
depicted in Table 3.4.
Three inspections were performed in PVC pipes used in the Dutch drinking water
supply network.
3.5 Statistical analysis 43
Table 3.4: Overview of the equipment and set-up of the field test and of the full-
scale tests. Field test: the inspected pipe was AC DN300 equipped with
double-socket push-fit joints. Full scale tests: all inspected pipes were
PVC DN315 equipped with double-socket push-fit joints. Only CCTV was
used.
2 runs +
1st Cut 350 48 37
2 controls
Full 2 runs +
2nd 70 10 37
scale 2 controls
Flange 2 runs +
3rd 250 14 42
2 controls
Where calculated is the calculated gap width using a ruler [mm], measured is the
measured gap width using the NDE [mm], m is the number of laboratory tests [#],
and i refers to the ith test. RMSE will have the same units as the measurements.
3.5.2 Reproducibility
Consecutive runs performed with an NDE were compared employing the Wilcoxon
Rank Sum Test which is a non-parametric statistical test. The parametric equivalent
44 3 Non-destructive evaluation of PVC push-fit joints
is the Student’s t-test. Corder and Foreman (2009) have thoroughly discussed non-
parametric statistics and the Wilcoxon Rank Sum Test. The Student’s t-test and
other traditional tests follow certain assumptions called parameters - parametric
tests. One of the assumptions in traditional statistics is that samples approximately
resemble a normal distribution. If a sample breaks one of these rules, all of the
assumptions of a parametric test are violated, and parametric tests cannot be used.
To avoid violating the normality assumption, the non-parametric Wilcoxon Test was
selected.
The null hypothesis in the Wilcoxon Rank Sum Test is that two data sets which
are independent samples from identical and continuous distributions with equal me-
dians are compared against the alternative that they do not have equal medians. The
returned p-value rejects or fails to reject the null hypothesis. A high p-value (over
the significance level of 5 % for the current work) demonstrates that the probability
that the alternative hypothesis was obtained by chance is very high; impossibility
of rejecting the null hypothesis, H value equal to 0. A low p-value (< 0.05) demon-
strates that the probability that the alternative hypothesis was obtained by chance
is very low. Therefore, the null hypothesis is rejected; the H value is equal to 1. Also
in the current work, the null hypothesis was that two consecutive measurements are
statistically similar (reproducible), against the alternative hypothesis that they are
not. In this work, the p-values are not expressed as a percentage.
3.6 Results
3.6.1 Laboratory tests
In order to compare the ultrasound and the visual tools, only the measurements
made with the ultrasound at the top (probe 1, Figure 3.7, right) and the bottom
(probe 5) are presented.
The gaps measured at the top and at the bottom with each one of the three
tools are plotted together with the calculated gaps employing the marks on the pipe
exterior. The results for the ultrasound, CCTV and Panoramo® are respectively
presented in Figure 3.12, Figure 3.13, and Figure 3.14. In these figures, certain
calculated gap values are negative because, at a maximum bending angle, the pipes
overlap inside the joint. The error-bars are representations of the standard deviation
of the measurements. The results of the RMSE and of Wilcoxon Rank Sum Tests
are depicted in Table 3.5.
Ultrasound
In Figure 3.12, a valuable agreement between measured and calculated gaps can
be seen. RMSEs of 2.2 mm and 2.4 mm were obtained respectively for the set of
measurements made at 6 h and at 12 h. No statistical difference (Wilcoxon Rank Sum
Test) was ascertained between the gaps measured during each of the two runs. The
results were reproducible. Additionally, the ultrasound could perform a complete
inspection in one run without interruptions.
3.6 Results 45
CCTV
In Figure 3.13 a very good agreement between measured and calculated gaps can
be seen. RMSEs of 2 mm were obtained both for 6 h and 12 h. This demonstrates
that the CCTV was more accurate than the ultrasound. No statistical difference at
95% confidence level (Wilcoxon Rank Sum Test) was ascertained between the gaps
measured with CCTV in the different runs. The results were reproducible.
Panoramo®
In Figure 3.14, a deviation between measured and calculated gaps is shown. This
is confirmed by the RMSEs: 5.6 mm and 2.5 mm respectively for 6 h and 12 h.
these are the largest RMSE obtained during the tests. This shows that Panoramo®
struggled to deliver accurate results. No statistical difference at 95% confidence level
(Wilcoxon Rank Sum Test) was found between the gaps measured with Panoramo®
in the different runs. The results were reproducible.
From Figures 3.12, 3.13, and 3.14, it is also evident that all three tools fail to size
negative gaps (overlapping pipes), which was expected given their characteristics.
Nevertheless, this is the reason why all three tools seem more accurate for wider
gaps.
Figure 3.12: Results of the NDE inspection with the Ultrasound. The measured gap
with the ultrasound is plotted against the gap calculated using the marks
on the pipe’s outer wall. For the 12 h data points, the support height
increases from left (0 cm support) to right (10 cm support). For the 6
h data points, the support height increases from right (0 cm support) to
left (10 cm support).
46 3 Non-destructive evaluation of PVC push-fit joints
Figure 3.13: Results of the NDE inspection with the CCTV. The measured gap with
CCTV is plotted against the gap that is calculated using the marks on
the pipe’s outer wall. For the 12 h data points, the support height
increases from left (0 cm support) to right (10 cm support). For the 6
h data points, the support height increases from right (0 cm support) to
left (10 cm support).
The gaps measured by the eight ultrasound probes are provided in Figure 3.15 (left).
The gaps measured by the CCTV at 3 h, 6 h, 9 h and 12 h are depicted in Figure 3.15
(right). In Figure 3.15 (right), the error bars represent the standard deviation of
the measurements. The results of the Wilcoxon Rank Sum Tests are presented in
Table 3.6.
Ultrasound
For the AC sections of the pipe, several gaps appear to be wider than 20 mm with
many reaching 30 mm to 40 mm.
The ultrasound tested in this experiment freely flowed inside the pipe during
inspection while it rotated axially. Without a gyroscope, it was not possible to know
where each probe was focusing during the inspection. However, it is important
to note that this problem is intrinsic to this individual ultrasound and not to all
ultrasound equipment.
3.6 Results 47
Figure 3.14: Results of the NDE inspection with the Panoramo® . The measured gap
with Panoramo® is plotted against the gap calculated using the marks
on the pipe’s outer wall. For the 12 h data points, the support height
increases from left (0 cm support) to right (10 cm support). For the 6
h data points, the support height increases from right (0 cm support) to
left (10 cm support).
CCTV
For the first 20 m of pipe (PVC), wider gaps were measured - over 20 mm and
reaching 30 mm. For the AC sections, only two gaps were wider than 10 mm. This
was explained by the fact that PVC joints allow higher bending angles than AC
pipes.
The Wilcoxon Rank Sum Test demonstrated that the hypothesis whereby the
sets were different could not be rejected at the 95% confidence level - the CCTV
runs and the controls were statistically similar (Table 3.6).
Table 3.5: Results of the statistical analysis of the laboratory results. The p-values
and the H values were obtained with the Wilcoxon Rank Sum test, at a
95% confidence level. With H=0, the null hypothesis cannot be rejected
- the runs are statistically similar. With H=1, the null hypothesis is
rejected.
Table 3.6: P-values obtained with the Wilcoxon Rank Sum Test for the CCTV results
obtained from the pilot-scale test at 95% confidence level. With H=0 the
null hypothesis cannot be rejected - the runs are statistically similar. With
H=1 the null hypothesis is rejected.
Figure 3.15: Results of the NDE inspection with the ultrasound (left) and CCTV
(right). The gap measured with both the ultrasound and the CCTV is
plotted against the in-pipe axial position.
be explained with the debris (mostly sand and iron) discovered at the bottom of the
pipe (6 h) during this inspection. For some joints, the beginning and the end of the
gap was not visible, and the operator, in order to size the gap, had to estimate these
boundaries.
Figure 3.16: CCTV results of inspection one. The mean gap calculated for run 2
and control 2 is plotted against the mean gap calculated for run 1 and
control 1. The line x=y is also presented.
50 3 Non-destructive evaluation of PVC push-fit joints
Table 3.7: P-values obtained with the Wilcoxon Rank Sum Test for the CCTV results
obtained from inspections 1, 2, and 3 were at a 95% confidence level. With
H=0, the null hypothesis cannot be rejected - the runs are statistically
similar. With H=1, the null hypothesis is rejected. The results differing
statistically at a 95% significance level are highlighted in bold.
Figure 3.17: The CCTV results of inspection two. The mean gap calculated for run
2 and control 2 is plotted against the mean gap calculated for run 1 and
control 1. The line x=y is also presented.
3.7 Discussion
3.7.1 Laboratory tests
The laboratory results clearly demonstrated that both the ultrasound and the CCTV
could deliver accurate and reproducible results and performed a quantitative pipe
assessment. On the one hand, the ultrasound has a significant advantage when
compared to the CCTV; the possibility of performing a complete inspection in one
run and in a real-life inspection can be an advantage. In fact, during an inspection,
the points of interest had to be focused by the CCTV. On the other hand, in order
to operate the ultrasound, it requires a completely water filled pipe. During the
laboratory tests, two data points were lost due to the presence of air at the top of
the pipe which was only detected after the inspection was completed when the data
were analyzed.
Panoramo® also delivered reproducible results. When compared to the CCTV,
an advantage of Panoramo® is the ability to inspect a pipe in a single run without
the need to focus on points of interest. Another advantage is the production of an
unwrapped image of the pipe’s interior. Two unexpected deficiencies of Panoramo®
were identified in these tests. With this tool, it was impossible to accurately inspect
above the water level when the pipe was partially filled due to refraction. In fact,
in this situation, the Panoramo® took the image of the top wall (above the water
line) from under the water. Furthermore, light reflection on suspended matter com-
plicated being able to obtain a clear image of the pipe’s interior. It can be concluded
52 3 Non-destructive evaluation of PVC push-fit joints
Figure 3.18: CCTV results of inspection three. The mean gap calculated for run 2
and control 2 is plotted against the mean gap calculated for run 1 and
control 1. The line x=y is also presented.
that, for accurate readings, Panoramo® should preferably be operated inside empty
pipes.
Therefore, the ultrasound and the CCTV were selected for the field tests.
Figure 3.19: B-scans obtained from the ultrasound inspection. Top: sizing the gap
width is at times prone to interpretation. Bottom: loss of data due to
increased travel speed. Courtesy of Applus RTD.
as a subjective and qualitative tool which delivers results that are prone to inter-
pretation. However, it should be mentioned that the application of the CCTV in
the present work differed from the typical application of the CCTV described in
literature. The CCTV is normally employed in order to screen a multitude of pipe
and joint defects - root intrusion, pipe cracks, wall collapses, presence of debris, joint
bending, etc. - and, in fact, the detection and quantification of some of these factors
can be subjective. In the current work, the CCTV was applied as a measuring tool
for a single application - gap sizing - and this application minimized the potential
for incorrect interpretations of the CCTV inspections.
3.8 Conclusions
In this chapter, three NDE tools were systematically tested both in a laboratory and
in the field. The results delivered by the three tools were analyzed on their accuracy
and reproducibility.
In the laboratory tests, both the ultrasound and the CCTV were demonstrated
to deliver accurate and reproducible results. Panoramo® was shown to deliver
reproducible results but displayed deficiencies when compared to the other two tools,
specifically, inconsistent reproducibility and accuracy as well as problems with the
inspection of water filled pipes, especially when in the presence of suspended matter.
Both the CCTV and the ultrasound were tested in the field to measure the joint
gaps of AC and PVC pipes. The field tests were conducted on a single day, the
pipe selected for this portion of the work allowed the CCTV and the ultrasound
inspection, which is not standard in water distribution pipes in the Netherlands.
The CCTV produced reproducible results. The tested ultrasound demonstrated
certain disadvantages including the lack of a gyroscope - a gap can be sized but it is
not known to which in-pipe position it is concerned - and production of unexpected
and extreme gap values.
The full-scale tests aimed at further testing the reproducibility of the CCTV.
Thus, the CCTV was employed over three different days in order to inspect three
different pipes. In the full-scale tests the CCTV again delivered reproducible results.
These results indicated that the visual tools tested in this work (CCTV and
Panoramo® ) could produce reproducible results. In addition to this, the CCTV
could be operated to consistently deliver accurate results and was considered the
best readily available tool for the inspection of joints and detecting the failure modes
described in Chapter 2.
chapter 4
Real-time pipe monitoring†
From early on in the project, the author realized two things regarding condition
assessment. On the one hand, it provides valuable information and many recognize
its importance. On the other hand, it is very intrusive and not all utilities are willing
to have pipes closed, inspected, and disinfected.
Therefore, an alternative to condition assessment as described in Chapter 3 had to
be developed. An option is installing sensors and monitoring the pipe in real-time.
A PVC 250 mm drinking water pipe that supplies water to approximately 1,250
customers was monitored from September 2011 until June 2013. The aggregated
data comprises strain registered on pipes and joints, temperature registered adjacent
to these appurtenances and strain registered on coupons of PVC isolated and also
installed adjacent to the pipe. The data depict an expected positive correlation
between temperature and strain on PVC. It also demonstrates that such a set-up is
able to detect daily water pattern use and episodes of water-hammer.
In order to compare the monitoring results with the NDE assessments, five CCTV
inspections had been planned: one before use and the remaining four throughout
the remainder of the year. However, due to safety/public-health issues, these inspec-
tions were not performed. Additionally, some practicalities evidenced from early-on
as many sensors became inoperative. Nevertheless, the obtained data proves that
such a set-up, although complex, can become a valuable approach for important
drinking water transport mains that cannot be inspected following a standard con-
dition assessment procedure.
Arsénio, A. M., Vreeburg, J. H. G., Wielinga, M. P. C., and van Dijk, H. (2012c). Contin-
uous assessment of a drinking water PVC pipe. In Water Distribution Systems Analysis (WDSA),
Adelaide (Australia)
55
56 4 Real-time pipe monitoring
4.1 Introduction
4.1.1 Background
In Chapter 3, it was demonstrated that the 3D alignment of the pipes connected
through a joint can be known measuring the gap between the two pipes connected
through a joint at four locations (bottom, top and both springlines). This 3D align-
ment can be employed as a surrogate measurement for joint condition prior to failure.
In the field, the gap widths and the alignment of the pipes change dynamically.
Several external factors influence the change, e.g., seasonal soil-temperature vari-
ation, traffic, inner-pipe pressure, and differential soil settling. Changes in pipe
alignment can only be detected using an NDE to determine if the pipe can be ac-
cessed for inspection and only if the pipe does not require continuous monitoring.
In case of, for example, important mains that cannot be placed out-of-service to
be inspected or mains that should be monitored permanently, real-time monitoring
might be the only solution. Two examples of important mains that should be moni-
tored permanently are large mains installed adjacent to highways or mains installed
in dykes, a typical case in the Netherlands.
In several countries, real-time monitoring set-ups have already been tested and
implemented. In Canada, real-time pipe monitoring has been used to determine the
influence of frost on the behavior of a PVC pipe spanning several years. In Australia,
metallic and plastic pipes have been monitored to study the impact of expansive soils
in pipe breakage. More recently, also in Australia, pipes have also been monitored
using fibre-optics.
At the IJkdijk test location, three test dykes were collapsed in August and
September of 2009 (IJkdijk, 2009). The objective of this project was to study the
applicability of employing modern sensor technology to obtain (sub)soil information
which could act as an early warning system for dyke failure. Some of the installed
4.1 Introduction 57
• Evaluate the thermal performance and cost effectiveness of the different types
of backfill material;
• Determine the modes of failure of PVC water mains exposed to freezing ground
conditions by monitoring strains in PVC pipes; and
• Conduct measurements of vertical earth loads in selected backfills and relate
the variation of loads to climatic and soil conditions and the type of backfill.
For the project “Critical Pipes”, Rajeev et al. (2013) reviewed the use of fiber-
optics on a drinking water pipe. According to the authors, “traditional electrical
sensors∗ the sensor network becomes more complex, difficult to install and maintain,
and expensive”. The authors also argued that “the use of these discrete sensors for
large civil structures is simply impracticable and not cost effective”. Furthermore,
according to Rajeev et al. (2013), fiber-optics afford the inclusion of a significant
number of sensors in a single optical fibre line owing to the tremendous optical
bandwidth and minimal power loss. These systems also have the capability to mon-
itor local strain, temperature, and corrosion rate, etc. at thousands of locations
distributed along a single optical fiber over several tens of kilometers (Rajeev et al.,
2013). Nevertheless, substantial research efforts are required to implement such a
project and realize its full potential (Rajeev et al., 2013).
4.1.2 Objectives
The aforementioned case-studies demonstrate the added-value of monitoring set-ups
for real-time pipe monitoring. Therefore, this chapter discusses the installation of
electrical sensors for real-time monitoring of pipes and joints to be employed as an
alternative to NDE inspection as described in Chapter 3. The objective will be
to cast some light on the accuracy, usefulness, and applicability of such a set-up
to monitor pipes and joints in real-time. Electrical sensors were selected instead of
fiber-optics as it was decided that monitoring short pipe length (40 m) was sufficient.
Furthermore, for this work, a field-test was selected instead of a laboratory test for
two main reasons:
1. Investigating the feasibility of installing such sensors in the field on a real-life
pipe; and
2. Monitoring a pipe subjected to real-life dynamic loads, e.g. water-demand
patterns and seasonal temperature change.
It had also been planned to asses the condition of the monitored joints utiliz-
ing CCTV (Chapter 3). The pipe would be inspected before being in use and,
afterwards, four times in the first year. While the first assessment would indicate
the initial condition of the pipe, the subsequent inspections would characterize the
behavior pipe throughout the year. This procedure would allow comparing the re-
sults provided by the CCTV with the results obtained from the sensors and could
be interpreted as further validation of the inspection procedure. However, due to
safety/public-health reasons, these inspections were not performed.
∗ For example, strain gauges.
4.2 Materials & methods 59
a−A ∆
= = (4.1)
A A
Where is the value of strain. Although is a dimensionless parameter, it
is usually given units of µ or mm of deformation per m of initial body length.
Therefore, a 1 µ value of strain value indicates that a 1 m length body expanded
10-6 m, or 0.001 mm. A [m] and l [m] are, respectively, the initial and the final length
of a body subjected to deformation. ∆ [m] is the variation in length. Therefore, when
a body is stretched (a > A or ∆ > 0), the strain is positive (Figure 4.1, left) when
a body is compressed (a < A or ∆ < 0) the strain is negative (Figure 4.1, right).
Using Equation 4.1 and knowing the strain, the expansion/contraction (variation in
length) of a body can also be calculated.
Figure 4.1: Two bodies subjected to deformation. A is the initial body length and a
is the final body length. Left: expansion, from Equation 4.1 strain will
be positive (a > A). Right: contraction; strain will be negative (a < A).
To monitor strain on pipes and joints, strain gauges were used (Figure 4.2, top
and middle-left). Three monitoring locations were created:
1. Location I: point A, joint 1 and point B;
2. Location II: point C, joint 2 and point D; and
3. Location III: point E, joint 3 and point F.
A general schematic of the complete set-up is illustrated in Figure 4.3. A detailed
schematic description of Location I is depicted in Figure 4.4. The set-up is equivalent
for Locations II and III. As can be ascertained in Figure 4.4, at each monitoring
point, strain is measured at 3 h, 6 h, 9 h and 12 h. Thus, the average axial strain
at each monitoring point can be calculated using Equation 4.2.
3h + 6h + 9h + 12h
= (4.2)
4
60 4 Real-time pipe monitoring
Where is the average strain, and i is the strain recorded by the gauge at
position i.
In this work, hoop strains were not measured.
Figure 4.2: Photos taken during the installation of the sensors. Top left: the strain
gauges fixed on the joints. Top right: a strain gauge being fixed on a
pipe. Middle left: the dummy strain gauges inside sealed PVC canisters.
Middle right: the installation of the pipe. Bottom left: the strain gauges
fixed to a joint and two pipes. Bottom right: following the installation,
the three cabinets are visible above ground.
Dummy gauges
Strain gauges fixed to pipes and joints monitored the appurtenances responses’ to
changes in temperature, inner pipe pressure, and external loads. In order to make a
distinction between the effect of temperature and the other factors on the pipe/joint
4.2 Materials & methods 61
strain dummy strain gauges were also used. To prepare the dummies, strain gauges
were glued to small rectangular coupons (10 cm×4 cm) made of PVC similar to the
installed pipe. These coupons were then sealed in water proof-canisters to isolate
them from the surrounding soil and water (Figure 4.2, middle-right). The PVC
canisters were 50 cm sections of 110 mm PVC pipes. These dimensions minimized the
friction between the PVC coupons and the inside of the PVC canisters. Therefore,
the dummy strain gauges, while being responsive to temperature changes comparable
to those experienced by the gauges installed on the pipe and joints, are isolated from
all other factors (e.g. traffic).
bottom left), and the pipe began operating in the first week of October. Three joints
and four pipe barrels in a row are monitored - a total of 40 m. All pipes are DN250
biaxial PVC with 10 m pipe sections that were installed at 90 cm depth. The pipe
sections are connected using double-socket PVC push-fit joints (Section 2.2.2).
Pipes and joints are monitored with strain gauges and, adjacent to each joint, a
thermistor and a dummy were installed. Three cabinets were placed above ground
1 m away from the pipe (Figure 4.2, bottom right). A complete list of the installed
sensors (per monitoring location) is provided in Table 4.1. All sensors were first
checked in the laboratory.
Figure 4.3: Side view of the monitoring set-up. The pipe was buried approximately
90 cm deep. Detailed schematics of Locations I, II, and III are depicted.
Location I is composed of position A, joint 1, and position B. Location
II is composed of position C, Location 2, and position D. Location III is
composed of position E, joint 3, and position F. Image not to scale.
Figure 4.4: Schematic of the installed strain gauges on Location I. The thick, hori-
zontal black-bars represent the strain gauges. In the schematic, the strain
gauges installed at 9h are on the other side of the pipe. The strain gauges
at points A and B were installed approximately 7 cm away from the joint.
The strain gauges installed on the joint were installed approximately in
the middle. The set-up is equivalent for locations II and III. Image not
to scale.
4.2 Materials & methods 63
Figure 4.5: Top view of the monitoring set-up. Each pipe section has a length of
10 m. The dummy strain gauges (D1, D2 and D3) and the thermistors
(T1, T2 and T3) were installed approximately 50 cm away from the pipe.
Image not to scale.
64
Strain gauge Geokon VK-415 12 µ and temperature 4 gauges glued to the joint also at 12h, 3,
6h and 9h exactly at the center of the joint.
Joints 1, 2 and 3
Figure 4.6: Soil temperature recorded by Thermistor 3 (Temp 3) and maximum air
temperature. Maximum air temperature data were obtained from KNMI
(2012).
Figure 4.7: The temperature recorded by Thermistor 3 (Temp 3) are plotted collec-
tively with the temperatures recorded at point D (3h, 6h, 9h and 12h).
Strain
In Figure 4.9, the average temperature at point D (Equation 4.3) is plotted with
the strain recorded at point E (3h, 6h, 9h and 12h). The data show that, as ex-
pected, there is a correlation between soil temperature and strain, i.e., an increase
in temperature leads to an expansion of the pipe.
The spiky pattern obtained for the axial strain on point E is due to the daily
water-demand pattern. In Figure 4.10, a detail of Figure 4.9 encapsulating the
period from the 31st of May until the 2nd of June is indicated. In Figure 4.10, it is
evident that, during the period when the average temperature at point D increases,
the strain reaches throughout the day. The strain has its maximum at approximately
12:00 a.m. and a second after 8:00 a.m.. These are probably periods of low water
use (lower inner-pipe pressure), during which time the pipe contracts radially and
expands axially (Poisson Effect).
In Figure 4.11, the temperature at point D is plotted collectively with the strain
recorded at joint 2 (6h and 9h)k and calculated with Equation 4.2. Also, in the joint,
the effect of the daily water-demand pattern is obvious.
In late 2011 and beginning of 2012, two episodes of water-hammer were registered
by the set-up (Arsénio et al., 2012c). Both episodes occurred before the period of
analysis covered in this thesis. Nevertheless, to support the present data-analysis,
the first episode is presented in Figure 4.12 (annotated). In this figure, the average
strain calculated at point E is plotted together with the average temperature at point
E and the temperature recorded by Thermistor 3 (Temp 3). It is obvious that the
transient detected by the strain gauges (positive variation of more than 500 µ, or 5
mm in 10 m) was not induced by changes in the soil/pipe temperature. According
to the utility, a valve was closed when the water-hammer episode was detected.
From the analysis of Figure 4.12, it is also visible that the water temperature
(Temp E wh) increased after the water-hammer. It can be hypothesized that the
origin of the water flowing inside the pipe changed once the valve was closed, and
that the water from the new source was warmer than previously.
Figure 4.9: The average temperature recorded at position D is plotted collectively with
the strain recorded at position E (3h, 6h, 9h and 12h).
Figure 4.10: Detail of Figure 4.9 encapsulating the period from the 31st of May 2012
until the 2nd of June 2012. The effect of the daily water-demand pattern
registered by the strain on the pipe is obvious. Please notice the change
in scale from Figure 4.9.
Expansion
In Figure 4.13, the average temperature recorded at point D is plotted with the
expansion/contraction values of pipe 2 along its length (Equation 4.1). This value
4.3 Results & discussion 69
Figure 4.11: The temperature recorded by Thermistors 3 is plotted together with the
strain recorded at joint 2 (6h and 9h). The strain gauges positioned at
3h and 12h were inoperative.
Figure 4.12: One of the water hammer episodes recorded around the 14th November
of 2011. The increase in strain is indicated with a black arrow. The
water utility confirmed having closed the valves on this day. Please
notice the change in scale from Figures 4.11 and 4.9.
Figure 4.13: The average temperature recorded at position D is plotted together with
the average expansion calculated for position E. The expansion is cal-
culated for each pipe-end.
4.3.3 Dummies
Temperature
In Figure 4.14, the temperatures recorded by temperature Thermistor 3 (Temp 3) are
plotted collectively with the temperature recorded by Dummies 2 and 3∗∗ . As before,
the soil temperature is higher than the temperatures registered by the dummies.
Strain
In Figure 4.15, the strain recorded at point E is plotted with the strain recorded by
Dummy 3 where the influence of the temperature in the dummy strain is evident.
The effect of the daily water-demand pattern is, as expected, absent on the dummies.
The data demonstrate that the axial strain at joint 2 is less than the strain
recorded by the dummy gauges during summer and higher during winter. During
winter, the pipe strain is higher than the dummy strain while, during summer, the
opposite happens. Expansion due to temperature changes cannot explain this fact
since the difference between the temperature registered at position E (the highest)
and at Dummy 3 (the lowest) is consistent throughout the year (Figure 4.16).
Figure 4.15: The average strain recorded at position E is plotted together with the
strain recorded at Dummy 3.
Figure 4.16: The average temperature recorded at position E is plotted together with
the temperature recorded at Dummy 3.
4.4 Conclusions
This chapter discusses the application of thermistors and strain gauges to monitor
pipes and joints in real-time.
Throughout the monitoring period, daily water-demand patterns and episodes of
water-hammer were detected by the strain gauges which demonstrates the accuracy
of the equipment.
From the beginning of the monitoring period until the coldest period for which
data are available, the pipe contracted more than 2.5 mm at each end due to tem-
perature changes. During the summer season (June-September), the temperature
can exceed the installation temperature. At that time, contact points (pipe-pipe
and pipe-joint) can begin if one considers that the pipes were installed fully inserted
inside the joints†† . This variation in pipe length can be detected by the CCTV
following the procedure outlined in Chapter 3. Additionally, the results also demon-
strate that a joint following a winter inspection is considered to be “at risk” with
small gaps (Section 3.2) and can be in an even riskier situation during summer after
pipe expansion.
This chapter also demonstrates that a pipe monitoring set-up can be implemented
for specific situations (e.g. high-importance mains). In such cases, specific sections
of the pipe can be permanently monitored for expansion/contraction by the utility.
A final note must be dedicated to the problems encountered during the period
through which the project ran, from September 2011 until June 2013. Pipe monitor-
ing is not problem-free, and this must be acknowledge by the people projecting such
a set-up. As can be seen in all figures except 4.6, portions of the data are always
†† As
will be discussed in Chapter 5, contact points should be avoided as they have a negative
impact on a joint’s behavior and increase its stiffness.
4.4 Conclusions 73
• Of the 16 strain gauges attached to the joints, only three (one per joint) re-
mained operational;
• Monitoring points C and D only had one operational strain gauge;
• At Location 1 the dummy strain gauge and the thermistor were not opera-
tional; and
• At Location 2 the thermistor was not operational.
Most sensors stopped working during the cold period of February 2012 when the
maximum air temperature did not rose above 0 °C for more than two weeks (which
was great for ice-skating, nonetheless). Due to the cold temperatures, the batteries
became depleted, and data was not logged for approximately eight weeks. When new
batteries were installed, some of the sensors were inoperative. Two strain gauges
installed on the pipes became inoperative following episodes of water-hammer. This
further emphasizes that, although such a set-up can produce useful and valuable
data, as argued by Rajeev et al. (2013), its complexity is also its Achilles’ heel.
This chapter had a clear objective, i.e., defining the limit at which a specified joint
begins leaking during bending or pull-out tests. Such information would be an im-
portant milestone. With access to the threshold conditions of joints, it becomes
possible to accurately detect most failure modes (Chapter 2) by employing the pro-
cedure discussed in Chapter 3.
Due to safety reasons, these tests could not be conducted at KWR and were,
instead, performed at the workshop of DYKA. Since a testing standard was not
being followed, the first steps were projecting, building, and improving the testing
frame. The second step was developing a test procedure so that all specimens were
tested under the same conditions. With everything finely tuned, more than 200 PVC
pipes and joints were tested.
Arsénio, A. M., Vreeburg, J. H. G., Bouma, F., and van Dijk, H. (2012a). Destructive lab-
oratory tests with PVC push-fit joints. In Water Distribution Systems Analysis (WDSA),
Adelaide (Australia)
Arsénio, A. M., Bouma, F., Vreeburg, J. H. G., and Rietveld, L. (2013a). Characterization
of PVC joints’ behaviour during variable loading laboratory tests (submitted). Urban Water
Journal
75
76 5 Destructive laboratory tests with PVC pipes & joints
5.1 Introduction
5.1.1 Background information
In Section 1.7.2, it was demonstrated that joints play a major role in the failures
occurring in a network with PVC pipes. However, given the focus of literature on
pipe barrel deterioration, it can be assumed that this role might have been neglected
in the past.
Additionally, in order to characterize the current condition of a joint in the field,
information on its threshold conditions was nonexistant.
For joint bending, it is expected that, after a certain bending angle, the joint
begins leaking. Leakage is either followed by complete pull-out of the pipe from the
joint or by the fracture of either the pipes in the joint or the joint itself. For axial pull-
out, it is expected that leakage begins after a certain gap (Section 3.2) followed by
the complete pull-out of the pipe from the joint. Therefore, the threshold conditions
are both a limit in bending angle and in pull-out distance before leakage, before
complete pull-out, and/or before complete burst. Having access to these data, the
present condition of a joint can be assessed. Additionally, the results of inspections
conducted in different periods can be compared, and a degradation rate can be
calculated.
For joint bending, comparable information is provided by pipe manufacturers
and in installation guidelines. Some Dutch PVC pipe manufacturers define 6° as the
maximum bending angle of a joint during installation (DYKA BV, 2012; Pipelife,
2001). The American Water Works Association (AWWA) defines a maximum joint
angle that is a function of the diameter and length of the installed pipe barrels
AWWA (2002). This equation calculates the maximum bending angle that is per-
mitted when two consecutive barrels are installed. In the presented work, it is
assumed that Equation 5.1, a modified version of the equation presented in AWWA
(2002) for bell-and-spigot joints (the original equation was multiplied by a factor 2),
can be utilized to calculate the maximum bending angle for a double-socket joint.
This assumption is made so that both guidelines can be compared.
57.3 × b
AW W AL = 2 × (5.1)
300 × Do
Where AW W AL [°] is the recommended installation angle, b [m] is the length
of a pipe barrel, and Do [m] is the outside pipe diameter. The original equation
was developed for the bell-and-spigot system and, during installation, the system
will not be stressed if the bending angle is below this limit. Equation 5.1, to be
applied to a double-socket joint, was multiplied by 2. Nevertheless, these guidelines
are published with the sole purpose of defining optimal pipe-joint alignment during
installation and offers no information about allowable bending angles throughout a
pipe’s lifespan.
Additionally, it is expected that, during bending, due to changes in inner geom-
etry, the stiffness of the joint varies. These changes include: i) contact between the
pipes, the inner joint, and the joint ring (Figure 2.2) and ii) between both pipes.
In the absence of leakage and/or material fracture, joint stiffness can be exploited
5.1 Introduction 77
to characterize its condition. In this work, stiffness describes the force required to
achieve a certain bending angle in a joint.
Similar work has been presented by several authors. Singhal (1984) performed
static experiments on bell-and-spigot rubber gasketed ductile iron joints. Meijering
et al. (2004) performed destructive laboratory tests with PVC gas pipes in order
to detect leakage during radial deflection of the pipes whereby during the tests,
pipes were pressed radially next to the joint. The authors reported that the joint
begins leaking when the diameter deflection exceeded 36%. Reed et al. (2006) tested
three CI joints and Buco et al. (2008b) characterized the behavior of cement pipes
and joints subjected to bending. Both authors concluded that, during bending, the
alignment of the pipes inside the joints changes. As discussed in Section 2.4.1, with
the increase of bending angle, contact points between the pipe and the joint and
between both pipes begins∗ . Buco et al. (2008b) ascertained that stiffness is greatly
influenced by the beginning of these contact points and that these are important in
defining a joint’s condition. Thus, an increase in stiffness is considered dangerous as
the joint can become overstressed which could possibly result in failure.
Additionally, several authors have modeled the behavior of joints exploiting com-
putational methods as is apparent is certain examples in the work of Buco et al.
(2008a) with reinforced concrete sewer pipes; Jeyapalan and Abdel-Magid (1987)
with reinforced plastic mortar and Scarino (1981) for several pipe materials. Ac-
cording to Rajani and Abdel-akher (2013), O’Rourke and Trautmann (1980) pro-
posed a model to estimate the resisting moment for a lead-yarn joint, assuming that
the spigot, given sufficient rotation, mobilizes the adhesion between lead and the
cast-iron bell (inside) and spigot (outside) surfaces. Rajani and Abdel-akher (2013)
discussed bell split failure, a predominant failure mode in lead-caulked bell-spigot
joints of large-diameter cast-iron pipes installed between 1850 and the early 1960s.
The authors also presented a mechanistic model and validated it against experimen-
tal tests conducted in the mid-1930s (Prior, 1935) on lead-caulked bell-spigot joints.
This model was subsequently utilized to develop an additional model to predict the
cumulative joint response of two or more contiguous pipe segments resting on an
elastic medium and subject to overburden pressure followed by ground movement.
Nevertheless, to the knowledge of the author, a thorough characterization of of
the behavior exhibited by water-distribution PVC joints for different loading regimes
was lacking. Therefore, PVC push-fit joints and pipes were tested in the laboratory
to obtain parameters that can be used to assess the condition of a joint in the field.
For pull-out and bending tests, two threshold conditions were investigated. The first
is start of leakage and the second is material fracture. During bending tests, force
was monitored to characterize the joint’s stiffness.
Table 5.1: Values of D’ for the three levels of insertion tested in the laboratory.
D’ [mm]
Figure 5.1: Schematic of a double-socket joint. D’ is the distance between the ex-
tremity of a pipe and the ring at the center of the joint. D is the distance
between the rubber o-ring and the ring at the center of the joint. D=50
mm for 110 mm joints and 100 mm for 315 mm joints. The schematic
is not to scale.
Table 5.2: Outside diameter (Do) and wall thickness of the pipes and joints used in
the laboratory tests.
Pipe Joint
110 mm 315 mm 110 mm 315 mm
A bending frame was constructed in order to perform these tests (Figure 5.2). Pipe
barrels were cut in 1.5 m sections and connected with double-socket joints. The out-
side diameters and wall thickness of the pipes and joints employed in the laboratory
tests are provided in Table 5.2. The system allowed pipes to be bent while being
pressurized with water. Screws on each extremity of the pipe were installed to vary
the level of insertion.
During testing, the pipe on the frame was pulled sideways using a hydraulic jack
(J in Figure 5.2). Bearings installed at the central rotation point allowed the frame
to bend more easily. Each of the two pipes rotated along a fixed rotation point
(R). The lateral displacement [mm] caused by the jack was measured using a linear
variable differential transformer (LVDT) (T). The force applied by the jack to bend
the structure was quantified with a force sensor (F). This sensor measures tensile
loads or pressure. Both the LVDT and the force sensor were connected to a logger
laptop.
The frame might influence the results of the laboratory tests due to two factors.
On the one hand, the bearings create a rolling friction during bending. On the other
hand, the effect of the end caps affect the force necessary to bend the structure.
This influence will be quantified.
80 5 Destructive laboratory tests with PVC pipes & joints
Figure 5.2: Top: top-view schematic of the loading frame. The pulling and monitor-
ing equipment of the frame is composed of a hydraulic jack (J), a force
sensor (F) and an LVDT (T). L is the distance between the rotation
points (R) and the lip of the pipe inside the joint. Bottom: side-view
schematic of the loading frame. Note: images are not to scale.
Figure 5.3: Schematic of a joint being bent from an original position until a final
position with joint angle α (Equation 5.2). Note: image is not to scale.
leakage was the only monitored parameter, and the LVDT and the force sensor were
not employed.
5.3 Results
5.3.1 Bending tests
The results of the laboratory tests conducted with 110 mm and 315 mm PVC are
depicted in Figure 5.4 and Figure 5.5. The moment of the force [kN] applied by the
hydraulic jack and recorded by the force sensor (Figure 5.2) is plotted against the
bending angle [°] (Figure 5.3).
The results of six independent tests are collectively indicated (gray line) with
the 10-point average (thick black-line) and the sum of inner-pipe force and rolling
friction force (Sum Forces) (dashed black-line). In Figure 5.4 and Figure 5.5, the
tests performed out with water are given in the top row, and the tests conducted
with air are demonstrated in the bottom row. The results of the tests performed
with maximum, half-way, and minimum insertions are illustrated in the left, middle
and right column, respectively.
For the maximum insertion tests (A and D), two vertical dashed lines are illus-
trated. These lines are at 0° and 3.2° for 110 mm and at 0° and 1.4° for 315 mm.
These are the points at which pipe-ring (0°) and pipe-pipe (3.2°) contact occurs.
The angles were calculated exploiting Equation 5.2 knowing the initial gap between
both pipes inside the joint.
It is clear that stiffness increases with diameter, insertion, and inner-pipe pres-
sure; the force necessary to achieve a specified angle increases. For example, with a
110 mm joint to reach 6°, it was necessary to apply 0.8 kN with maximum insertion,
0.5 kN with half-way, and 0.4 kN with minimum (Figure 5.4 A, B and C).
5.4 Discussion
5.4.1 Destruction of PVC material, leakage & intrusion
Intrusion and leakage were only detected during minimum insertion tests for angles
above 10°. This demonstrates the high performance delivered by the PVC double-
socket used in the laboratory tests.
Throughout the laboratory tests fracture of material was never observed.
Figure 5.4: Results of the laboratory tests conducted with 110 mm PVC. In all fig-
ures, the results of six independent tests are provided (gray) together with
the moving average of 10 data points (black). Top row: water pressure.
Bottom row: air. Left column: maximum insertion. Middle column:
half-way insertion. Right column: minimum insertion.
stiffness; the joint becomes stiffer with the increase of diameter. Several factors
influence this, and some are intrinsic to the testing frame while others are intrinsic
to the joint itself, i.e., effect of water pressure (intrinsic to the frame), increased
weight of the structure (intrinsic to the frame) and increased stiffness of the pipe
(intrinsic to the joint).
F~i = P × Ai (5.3)
Where F~i is the internal force [N], P is the inner-pipe pressure [Pa], and Ai is the
inner-pipe cross-section area [m2 ]. The y component of this force F~iy has a direction
opposed to the force applied by the hydraulic jack (Figure 5.6, Equation 5.4).
5.4 Discussion 83
Figure 5.5: Results of the laboratory tests conducted with 315 mm PVC. In all fig-
ures, the results of six independent tests are provided (gray) together with
the moving average of 10 data points (black). Top row: water pressure.
Bottom row: air. Left column: maximum insertion. Middle column:
half-way insertion. Right column: minimum insertion.
α
F~iy = F~i × (5.4)
2
Where F~iy is the component in y of the internal force [N] and F~i [°] is given in
Figure 5.6. The effect of inner-pipe pressure is thoroughly discussed ahead for the
constant diameter situation.
Figure 5.6: Schematic of forces in play during bending. The force applied by the
hydraulic jack F~J to bend the joint, and the force created by the inner-
pipe pressure F~i are indicated. Note: image is not to scale.
Table 5.3: Values of rolling resistance force for the various tests.
DN Water Air
110 16.0 14.7
315 31.1 20.3
force [N], g is the standard gravity [m.s-2 ], and Mtotal is result of the summation
of the water’s mass with the structure’s mass [kg]. The mass of the structure con-
taining water is calculated (considering that the system is a perfect cylinder) using
Equation 5.6.
5.5 Conclusions
Leakage and intrusion were detected in only a few tests at extreme bending angles
or, for pull-out tests, after complete pull of the pipe from the joint. This indicates
that for PVC joints, leakage through the rubber-gasket is mostly dependent upon
the condition of the rubber. For a rubber ring in good condition, leakage can only
be expected at bending angles above 10° and with complete pull-out of the pipe from
the joint.
Joint stiffness increases with insertion, diameter, and pressure. An increase in
stiffness is considered dangerous as the joint becomes less able to bend. Nevertheless,
not all tested parameters play the same role.
Increase in diameter is the most important factor for the increase of joint stiffness.
This occurs due to three factors: increase of the structure’s weight; increase of the
inner-pipe pressure, and due to an increase in pipe stiffness bending. Both structure
weight and inner-pipe pressure played smaller roles. Pipe stiffness was ascertained
to increase more than 67 times from a 100 mm to a 315 mm pipe. Pipe stiffness
also played a role during bending both during and after the beginning of the contact
points (pipe-pipe and pipe-joint). It should also be noted that the stiffness effect is
not intrinsic to the testing frame but is a characteristic of the joint itself.
For pipes of the same diameter, insertion was ascertained to play the most im-
portant role. A pipe inserted further in the joint leads to decrease in the angle values
at which the contact points begin and the start of contact will stiffen the joint. This
situation is obvious for the maximum insertion tests. For the half-way insertion and
minimum insertion tests, the contacts points are avoided. For the minimum inser-
tion, the behavior of the joint appears to be almost independent from the bending
5.5 Conclusions 87
angle. This effect was specifically noted in the tests with 315 mm joints where a
clear decrease in joint stiffness was discovered when comparing maximum insertion
with medium insertion tests. For 110 mm pipes, the decrease was not so obvious.
Therefore, joint angle is a necessary parameter but insufficient to characterize
a joint’s condition. This work demonstrates the importance of also defining the
insertion level of the pipe inside the joint - a double parameter assessment. During
an assessment with NDE equipment for two joints with the same bending angle,
the greater risk can be allocated to the one with the greater level of insertion.
Additionally, from a practical perspective, these results demonstrate the significance
of not installing the pipes completely inserted inside the joints as this makes the
joints stiffer.
chapter 6
The index for joint condition†
Presenting results obtained with an NDE tool is not an easy task. Data should be
presented in a clear manner whereby conclusions can be derived. Additionally, it
is of the utmost importance to present the results in a way that allows comparing
different pipes - a pipe condition ranking. In fact, it might be necessary to decide
which of two pipes is in poorer condition, whether all joints in a pipe are in a deficient
condition, or if only a few joints are responsible for a certain pipe grade.
For this, the Index for Joint Condition (IJC) for PVC push-fit joints was derived
from installation guidelines and from destructive laboratory tests (Chapter 5). The
IJC is presented in a graphical framework and is a powerful tool to employ in order to
visualize and compare, in-situ, results obtained during condition assessment of PVC
joint. The graphical results can also be translated into a numerical grade that allows
comparing the conditions of various pipes and of individual joints. The applicability
of the IJC is demonstrated in the condition assessment of 222 joints inspected in 8
different sessions that encapsulate more than 2 km of older (more than 40 years) and
newer pipes (less than 2 months). While, for the new pipe, all joints were considered
to be in good condition, several joints in older pipes were considered to be “at risk”.
Arsénio, A. M., Vreeburg, J. H. G., and Rietveld, L. (2013e). Index of joint condition for
PVC push-fit joints (submitted). Water Science & Technology: Water Supply
89
90 6 The index for joint condition
6.1 Introduction
One way of preventing failures in drinking-water pipes and joints is through condition
assessment (Chapter 3). It was demonstrated that the CCTV can deliver both
accurate and reproducible results if employed in gap-sizing applications.
However, in order to accurately define a joint’s condition in the field, an NDE
is necessary but not sufficient. It is also necessary to have access to threshold con-
ditions. With the exclusion of installation guidelines, parameters to define a joint’s
condition in the field throughout its service-life were not evident in the literature.
Additionally, these installation guidelines were created to define optimal pipe-joint
alignment during installation and provide no information regarding allowable bend-
ing angles throughout a pipe’s lifetime. Dutch PVC pipe manufacturers define 6°
as the maximum bending angle at a joint during installation (DYKA BV, 2012;
Pipelife, 2001). The American Water Works Association (AWWA, 2002) defines a
maximum joint angle as a function of the diameter and length of the installed pipe
barrels.
However, these installation guidelines were created to define optimal pipe-joint
alignment during installation and offer no information about allowable bending an-
gles throughout a pipe’s lifetime. In Chapter 5, the results of a large-scale laboratory
tests with PVC pipes and joints were presented. These tests were performed in order
to obtain parameters that can be employed to characterize a joint’s condition in the
field. It was concluded that the diameter and the level of insertion play major roles
in the behavior of joints. Joints become stiffer with the increase of both parameters,
an increase in stiffness is considered a risk as the joint is less able to bend, and stress
subsequentely builds up in the pipe material. It was also indicated that a joint’s
condition should be defined utilizing two parameters including joint angle and the
level of insertion of the pipe in the joint.
Therefore, this chapter presents the IJC and uses it to characterize the condition
of the joints in eight pipes that were inspected for this project.
1. Receive input produced by any NDE tool that follows the aforementioned
inspection procedure;
2. Characterize the condition of all joints in an inspected pipe in-situ. This
characterization can be performed, for example, to evaluate a contractor’s
work (after installation) or to assess the joints’ condition of a pipe that is in
use;
3. Compare results from inspections made for various PVC pipes: aid network
managers in selecting the most appropriate pipes/joints for replacement/repair;
6.2 Materials & methods 91
and
4. Compare results of different inspections made on the same joints at different
time periods, i.e., determine joint degradation rate. It should also be con-
sidered that changes to joint alignment are a result of soil movement, poor
installation, operational factors (e.g. pressures), live-loads or a combination
of these factors. Nevertheless, these phenomena are often not continuous and
can be a one-time event. Therefore, several inspections of the same joint over
a sufficiently long period of time would be required to establish an actual
deterioration rate.
Figure 6.1: The IJC presents all of the data aggregated during a pipe inspection
(500 mm). The vertical dashed line (Bending angle = ± 6.8°; annotated
“Lab tests”) defines the maximum angle ascertained by the laboratory
tests. The dashed horizontal line (MGW = 8 mm; annotated “Ring”)
defines the width of the joint ring (Figure 2.2). Joints inside the grey
areas (limited by the aforementioned dashed lines) are considered at risk
of leaking or of fracturing. The error bars demonstrate the standard
deviation of MGW and bending angle between the several repetitions of
the inspection. The number 6 identifies the inspection.
Grading a pipe
Taking into consideration the location of a joint in the IJC (Figure 6.1), each joint
is graded separately (joint grade). The complete pipe length is also graded by
aggregating joint grades of all pipe joints within that pipe length (pipe grade). This
grading procedure is similar to the process of grading distress indicators presented
by Kleiner et al. (2006a). The grade of joint i (dimensionless and normalized) is
presented by the summation of two factors - a higher grade identifies a joint in worst
condition (Equation 6.1).
1 1
JGi = × G1i + × G2i (6.1)
2 2
Where G1i accounts for the angle and G2i accounts for the level of insertion.
Both factors are considered to have the same weight (1/2).
The factor G1i (dimensionless) is indicated by Equation 6.2 and increases with
joint angle α (Figure 3.3).
G1i = 0 if α < T A
(6.2)
α
G1i = max if α > T A
6.2 Materials & methods 93
" n # " n #
1 X1 1 X 1
P Gj = × (G1i + G2i ) + × (G3i,neg + G3i,pos ) (6.4)
3 i=1
n 3 i=1
n−1
Where G3 is a parameter that accounts for the number of contiguous joints with
α of the same sign, and n is the number of inspected joints. Each one of the three
factors is considered to have the same weight (1/3). Therefore, if two consecutive
joints achieve negative angles, G3i,neg is calculated by Equation 6.5
|αn | + |αn+1 |
G3i,neg = (6.5)
2 × max
Where |αn | and |αn+1 | are, respectively, the absolute values of the angles of con-
tiguous joints n and n+1. A similar formula is exploited in order to calculate G3i,pos ;
this parameter accounts for contiguous joints with positive α. This approach is fol-
lowed because two contiguous joints with angles with the same sign are considered
dangerous. For example, if the two joints are negative, the pipe buckles and the
joints are sagging. Thus, both values are to quantify the situation since two contigu-
ous joints with 1° can be considered less dangerous than two contiguous joints with
5°. The expression is divided by 2 × max to obtain normalized and dimensionless
results. Therefore, since all three factors range from 0 (good condition) to 1 (worst
condition), P Gj also varies in the same range.
6.3 Results
In total, 222 joints in approximately 2,100 m of PVC pipe were inspected (Table 6.1,
Figure 6.2). The pipe grades are depicted in Table 6.1, and the joint grades are
exhibited in Figure 6.3 where the joint number is demonstrated in the X-axis, and
the joint grade is given in the Y-axis. A black dot represents the grade of an inspected
joint.
Table 6.1: Overview all NDE field-inspections done with CCTV. TA is the maximum angle given in Chapter 5 which is a function
of diameter. Pipe grade is defined by Equation 3.1 and varies between 0 (good condition) and 1 (bad condition). The
grade allows ranking the pipes according to their condition.
Inspection Access Diameter [mm] TA [°] Length [m] Joints [#] Repetitions [#] Grade (× 100)
Figure 6.2: Results of the field inspections for eight inspections. A black dot identifies an inspected joint and the error bars for both
the MGW and angle are given. The error bars give the standard deviation of MGW and bending angle of the several
runs. Relevant information about the inspections is given in Table 6.1.
97
98
Figure 6.3: Grade of each inspected joint. A dot indicates an inspected joint.
6 The index for joint condition
6.4 Conclusions 99
6.4 Conclusions
The IJC has been implemented by employing installation guidelines and data from
destructive laboratory tests. When using the IJC, a grade can be assigned to both
the pipe and to each joint. The grade makes the results of the IJC less prone to
interpretation and more reproducible.
The results demonstrated that, given the degrees of freedom in a pipe system,
assessing the condition of joints cannot be performed with one parameter. For
this reason, both the individual joint grades and the overall pipe grades should be
analyzed in conjointly. This approach also provided a better perspective of the
condition of individual joints and their role in the condition of the entire pipe.
Results of eight different CCTV inspections were presented. While a great disper-
sion of angle and gap values were ascertained for the older pipes, minimal dispersion
was ascertained for the 2-month old. Two conclusions can be drawn from this. First,
the alignment of pipes inside the joints varies throughout the years. Second, per-
forming inspections on newly installed pipes can be utilized to assess the work of
contractors following pipe installation. Furthermore, the IJC has been proven to
be a powerful tool to compare the condition of different pipes. The importance of
individual joint grading is demonstrated by the possibility of a network manager to
perform selective repairs in certain joints in order to reduce the total pipe grade.
chapter 7
Correlating pipe failures & ground
movement†
In this chapter, the approach to predict pipe failure in drinking water networks
is presented. Ground movement is believed to play a crucial role in the onset of
failures in underground infrastructure. In the Netherlands, due to its geological
characteristics, this role is expected to be even more pronounced. This chapter
demonstrates a methodology that generates a risk map for underground drinking
water pipe networks employing ground movement data.
A segment of the distribution network of a Dutch drinking water company was
selected as the study-area. A USTORE (Section 1.6.3) data-set for this area was ob-
tained. The data-set was comprised of 868 failures registered throughout 40 months.
The utility also supplied geographical network data. Ground movement was esti-
mated using satellite-borne radar data.
Two types of analyzes were made: cell and pixel-based. For the cell-based anal-
ysis, AC pipes exhibited the highest failure rates. Older AC pipes were also shown
to fail more often. Conversely, failure rates for PVC were the lowest of the test.
For the pixel-based, analysis ground movement was also demonstrated a role in the
failure of all of the materials combined.
Therefore, a risk-map for AC was generated which combined ground movement
data and pipe-age data. This methodology can be a beneficial resource for the
network managers for maintenance and continuous monitoring.
Arsénio, A. M., Dheenathayalan, P., Hanssen, R., Vreeburg, J. H. G., and Rietveld, L.
(2013b). Pipe failure prediction in drinking water systems using satellite observations (submitted).
Structure and Infrastructure Engineering
101
102 7 Correlating pipe failures & ground movement
7.1 Introduction
Drinking water supply networks are part of a myriad of underground infrastructures
that underpin modern civilization. Much of this infrastructure is located under-
ground, and its fate is intimately related to that of the surrounding soil (O’Rourke,
2010). A factor expected to play a role in the failure of drinking water pipes is non-
uniform ground movement also indicated as differential ground movement (Budhu,
2010). In this process, ground movement creates stress on the pipe and may lead to
failure. Damage to one facility (e.g. water main) can culminate into damage to sur-
rounding facilities (e.g. gas or telecommunications) with system-wide consequences
(O’Rourke, 2010).
Ground movement can be especially damaging to older pipelines (Olliff et al.,
2001) and older, rigid joints (Silva et al., 2001). Hu et al. (2008) argued that
changes in soil volume can induce differential ground movement that can subse-
quently cause the development of stresses in asbestos cement (AC) pipes. The effect
of ground movement can also be potentiated in pipes whose integrity may have al-
ready been compromised by other factors, for example, chemical attacks from inside
water and/or outside soils (Hu et al., 2008). According to Breen (2006), provided
that PVC pipes are properly manufactured, installed, and free of scratches of more
than 1 mm depth, they can endure for more than 100 years in operation. However,
the same authors argued that non-uniform soil settling can cause enormous local
stresses in a PVC pipe and lead to preliminary failure, and such condition can de-
crease the lifetime of a PVC pipe to 10 years. Dingus et al. (2002) surveyed 46 of the
largest AwwaRF member utilities. For the distribution systems, according to 25%
of the survey respondents, frost heave/ground motion was the number-one problem.
Folkman (2012) surveyed 188 North American utilities. According to the authors,
55.3% of respondents identified CI as the most common failing pipe material followed
by AC at 17.0%. Furthermore, the utilities were asked to choose the most common
type of failure in their networks, and 50% answered that circumferential crack was
the main cause. In detail, the authors indicated that circumferential crack was the
primary failure mode for CI, concrete, and AC. Corrosion was the primary failure
mode of Ductile Iron and Steel. A longitudinal crack was the primary failure mode
of PVC.
It is evident that one of the causes for circumferential breaks is longitudinal
loading (Rajani and Kleiner, 2001) and that longitudinal loading can be originated
by ground movement (Moser and Folkman, 2008). According to O’Rourke (2010),
geohazards (e.g. soil subsidence, earthquakes, hurricanes) have generated substantial
interest in lifeline systems (e.g. water, gas and telecommunications). Keeping this in
mind, O’Rourke et al. (2008) examined the response of the Los Angles water supply
network during the 1994 Northridge earthquake, which was the beginning point
of developing a decision support model for the city’s distribution network. More
recently, O’Rourke (2010) expanded the previous work and analyzed and modeled
the response of three North-American networks to earthquakes (San Francisco and
Los Angeles), the effect of Katrina in New Orleans, and in the Mississippi river and
Gulf of Mexico.
In the past centuries, the Netherlands has not been affected by earthquakes or
7.2 Materials & methods 103
density) varies from area to area and depends on the satellite used. In general, a
density from 100-1000 PS-points.km-2 in urban environments can be expected.
The PSI technique requires a reference in time and space to estimate ground
movement. The reference in time is solved by selecting a reference image, referred
to as the master image and comparing other images to the master for estimating
the changes in surface. The reference in space is obtained by taking a point in the
master image as a reference point. All of the estimations are then provided with a
reference to this reference point, and the most stable point in the image is normally
chosen as a reference point.
For this analyzes, 99 TerraSAR-X∗ strip-map ascending mode images of the
study-area were processed employing PSI technique to estimate the ground move-
ment rate (expressed in mm.year-1 ) per pixel. For this satellite, a pixel, or PS-point,
indicates an area of 3 × 3 m (approximately). The satellite data used in this work
encompasses the period from February 2009 until May 2012.
∗ TerraSar-X is presented as the “most accurate high-resolution radar satellite in orbit” (Astrium,
2013).
7.2 Materials & methods 105
and latitude, and the pipe material were known. These data were obtained from US-
TORE (Section 1.6.3). USTORE provides more relevant information, for example,
regarding the origin of the failure (ground movement, inner/outer-corrosion, etc.),
diameter, type of soil, or presence of trees in the vicinity. Nevertheless, the reliability
of these data have not yet been assessed, hence these data were not exploited.
7.2.3 Study-area
The study-area is approximately half of the supply region of a Dutch drinking water
company. The total length of network in that area is more than 4,500 km, with PVC
(∼50%), AC, and cast iron (CI) (both 16%) being the most common materials. This
network supplies more than 1.2 million people. This study-area was selected for the
following reasons:
2. Anticipated ground movement above the national level (Lange et al., 2012);
and
Pixel-based approach
This approach analyzes the vicinity (< 25 m) of PS-points. The vicinity is expected
to be affected by the local differential ground movement, quantified by PDM. In
the Netherlands, drinking water pipes are usually are 10 m long. Two pipes are
connected with stand-alone pieces referred to as joints that feature two rubber-
gaskets (Section 2.2.2). These gaskets keep the system sealed and are expected to
sustain some of the ground movement. If a PS-point is located on top of a pipe, it
is hypothesized that localized ground movement will impact the pipes connected to
the two subsequent joints in each direction ( 25 m). For longer distances, the effect
of ground movement will be dissipated by the joints.
Nevertheless, it is noteworthy to mention that it can be argued that ground
movement may be the result of pipe failure (e.g. leakage) and not the cause. It
cannot be, indeed, concluded what occurred first - ground movement or pipe failure.
Therefore, in this work, it is assumed that ground movement is the cause of failure
and not the result.
For this approach, the distribution of PDM values in the vicinity of failures will
be compared with the distribution of PDM values away from failures. To produce the
106 7 Correlating pipe failures & ground movement
risk map, detailed GIS data is required, i.e., the exact coordinates of the pipe net-
work. However, these data were not available. Thus, the analysis will be conducted,
but the risk-map will not be produced.
Cell-based approach
The study-area was divided in a virtual matrix (451×320 cells). Each cell represents
an area of 100 × 100 m. This cell size was selected to allow obtaining cells with 20-30
pixels each to allow significant results in the subsequent calculations.
It can be expected that pipes installed in different backfills will respond differently
to ground movement. Nevertheless, the soil in each cell is considered to possess
homogeneous characteristics. This approach was followed due to the lack of more
detailed soil data; Dutch water companies do not register the characteristics of the
soil where the pipes are installed. In this approach, the ground movement in a cell
is hypothesized to play a role in the cell’s failure rate.
All PS-points within a cell can be collectively calculated for an average PDM
(Equation 7.1).
P
P DMj
P DMi = (7.1)
n
P
Where P DMi is the average PDM cell i [%] P DMj is the PDM of pixel j in
cell i [%], and n is the number of pixels in cell i.
Furthermore, each failure was allocated to the specific cell where it occurred and
the failure rate per cell was calculated using Equation 7.2
P
fi,j
λi,m = P (7.2)
Li,m × ∆t
Where λi,m is indicative of the failure rate [#.km-1 .year-1 ] in cell i for material
m; fi,j is the j th failure in cell i; Li,m is the pipe length [km]† of pipe material m
P
in cell i; and ∆t is the duration of the registration data [years]. This formulation for
failure rate is similar to that presented in Equation 1.1 but is adapted in this instance
to the cell situation. O’Rourke (2010) employs a similar parameter referred to as
the repair rate [# repairs.km-1 ]. This formulation, solely normalized in reference to
network length, is of use in episodic situations such as earthquakes. However, for
the long-term analysis of network performance, the normalization in respect to time
becomes necessary.
The average pipe age per cell was calculated using Equation 7.3.
P P
agek,m × Lk,m
Agei,m = P (7.3)
Li,m
P
Where Agei,m is the average age [years] of pipe material P m in cell i; agek,m is
the average of age bin k (5-year bins) of material m; and Lk,m is the length [km]
of material m of age k.
† Assuming
P
that pipe length Li,m did not change during ∆t.
7.3 Results & discussion 107
Table 7.1: Pipe length, number of failures (total and due to third party damage) and
failure rate (λ) per pipe material within the study-area.
Length [km] Failures [#] λ [#.km-1 .year-1 ] 3rd party damage [#]
In this analysis, cells for which radar estimations are unavailable or that have no
drinking water network are not considered in the analysis.
For this approach, a correlation between ground movement and pipe failures at a
cell level will be researched. To produce the cell-based risk-map GIS data, as detailed
as for the pixel-based analysis, is not necessary. In fact, knowing the total length of
pipes and the number of failures per cell is enough. These data were available, and
a risk map was produced.
the rest of the analysis, it is assumed that all of these pipes are older than 50 years.
Comparable results have been presented by Vloerbergh et al. (2012) for the complete
USTORE database (five Dutch water companies) for the failures registered in 2009
and 2010.
Figure 7.1: Failure rate per material and decade from 1900 until 2000.
It was hypothesized that, if failures are caused due to ground movement, the dis-
tribution of PDM in the vicinity of failures (< 25 m) could be skewed to positive
values (higher PDM). This analysis was conducted by including all available failures
(without third-party) collectively from all materials.
A normalized histogram of the PDM for all pixels in the study-area and for the
pixels in the vicinity of failures (distance < 25 m) is presented in Figure 7.2 (top
and bottom). While the complete histograms are presented on the left, details of
the histograms (PDM > 50%) are depicted on the right.
The complete histograms show little difference between the entire area (Fig-
ure 7.2, top left) and the vicinity of failures (Figure 7.2, bottom left).
However, it is evident from the detailed histograms that, in the vicinity of failures
(Figure 7.2, bottom right), there is a greater percentage of PS-points with a higher
PDM (above 90%) than in the entire area (Figure 7.2, bottom left), which indicates
the role of differential motion in the failures of pipes.
7.3 Results & discussion 109
Figure 7.2: Histograms of PDM for the complete study-area and in the vicinity of
registered failures (< 25 m) for all pipe materials. Top left: PDM for
all pixels in the study-area. Bottom left: maximum PDM of pixels close
to the failures. Top and bottom right: zoomed details of histograms on
the left, presented for PDM>50%.
Risk map
Despite a trend being identified, the GIS data required to produce the risk map were
not available. Thus, the pixel-based risk-map will not be produced.
Given the size of the study-area and the relative minimal number of failures, ground
movement data was divided in bins. Three bins were created including low movement
(LM), medium movement (MM) and high movement (HM).
A histogram of PDM per cell is illustrated in Figure 7.3. The boundaries between
the bins are percentiles 33rd (PDM=25.6%) and the 67th (PDM=46.4%) percentiles.
While LM is below PDM=25.6%, HM is above PDM=46.4%. The boundaries are
represented in Figure 7.3 by dashed vertical-lines. These boundaries were selected
to allow having an equal number of counts in each bin.
Age distribution
Mean age per pipe material is calculated exploiting Equation 7.3. Age distribution
indicates no variation across different areas (Figure 7.4). In Figure 7.4, the error-
bars represent the standard deviation of the age. For this area, PVC is the youngest
material (15 years) followed by AC (45 years) and CI being the oldest (70 years).
110 7 Correlating pipe failures & ground movement
Figure 7.3: Histogram PDM. The two grey dashed vertical-lines represent the bound-
aries for ground movement: percentiles 33rd (PDM=25.6%) and 67th
(PDM=46.4%). While LM is for PDM<25.6%, HM is for PDM>46.4%.
Figure 7.4: Distribution of pipe age per level of ground movement. All mat represents
all materials taken as a whole. The error-bars represent the standard
deviation of the age.
7.3 Results & discussion 111
Length distribution
Areas with increased ground movement also have a higher percentage of PVC and
lower percentages of AC and CI (Figure 7.5).
In stable areas (sand), the infrastructure was mostly installed in the late 19th
Century and early 20th Century when CI was the preferred material. Due to the
expansion of the population, unstable areas (peat) began to be urbanized beginning
in the 1960s. Since that time, the use of PVC has been increasing among Dutch
water companies (Geudens, 2012).
Given these results, it could be expected that the average pipe age in areas of
HM was lower than in the other areas, which is not the case (Figure 7.4). This
indicates that this water company does not use ground movement to prioritize pipe
replacements.
Failure rate
In Figure 7.6, the aggregated failure rate for all areas with a certain level of ground
movement is plotted against level of ground movement. In Figure 7.7, the same
data is presented in a scatter plot so that the slopes can be obtained. In Figure 7.7,
the x-coordinates are the average between the corresponding bins. A positive slope
indicates that failure rate increases with the level of ground movement. Furthermore,
the material most sensitive to ground movement will also have the highest slope.
From Figure 7.6 and Figure 7.7, is it evident that PVC experiences the lowest
failure rate values and is the least influenced by ground movement (slope=2.1x10-4 ,
lowest). This was expected because PVC pipes are both the youngest and the
112 7 Correlating pipe failures & ground movement
most ductile pipes in the network. The failure rates of both CI (slope=1.1x10-3 ,
highest) and AC (slope=8.5x10-4 ) increase with the increase of ground movement
(Figure 7.7). Also, CI has the highest slope, which indicates that, in this area, CI is
the most sensitive material to ground movement. Overall, the failure rates of AC are
the highest in the study-area, always at least four times higher than that of PVC.
Finally, no relationship between failure rate and pipe age could be ascertained since
the pipe age distribution does not vary among the various areas (Figure 7.4).
Figure 7.6: Distribution of failure rate per pipe material and per level of ground
movement. All mat represents all materials taken as a whole.
Risk map
For AC pipes, two factors exhibited a role in failure: age and ground movement.
On the one hand, a clear difference can be determined between the pipes older and
younger than 50 years (Figure 7.1). On the other hand, the failure rate for AC pipes
increases with the level of ground movement (Figures 7.6 and 7.7).
In Figure 7.8, the age distribution of AC pipes in the study-area is depicted. In
Figure 7.9, the distribution of ground movement data in the study-area is demon-
strated. In both figures, white indicates the nonexistence of network. While there
are two clusters of older pipes, 52.02°N, 4.3°E and 52.07°N, 4.37°E (Figure 7.8), the
distribution of ground movement data is random across the study-area (Figure 7.9).
Both AC pipe age and level of ground movement were employed to create an
impact matrix (Table 7.2) based on the approaches described in order to produce
risk matrices to manage wastewater networks Salman (2010) and was the foundation
of the risk-map. For this approach, a cell is prescribed age-points and area-points;
a higher grade indicates a higher risk for pipes in the cell due to ground movement.
7.3 Results & discussion 113
Figure 7.7: Failure rate per pipe material against PDM. All mat represents all ma-
terials taken collectively.
Thus, zero age-points were assigned to AC pipes younger than 50 years (lower prob-
ability of failure) and one age-point to AC pipes that are older (higher probability
of failure). Similarly, zero area-points were assigned to pipes installed in LM ar-
eas (lower probability of failure), one area-point in MM areas (higher probability
of failure), and two area-points in HM areas (highest probability of failure). The
total score per cell was determined by adding the age-points and the area-points.
Therefore, the lowest risk (0 points, green) was assigned to cells with the majority
of young AC pipes in areas of LM and the highest risk (3 points, red) was assigned
to cells with majority of old AC pipes installed in HM areas.
Using the impact matrix, the risk map was produced for AC pipes (Figure 7.10,
left). To validate the assumptions of the risk-map, the distribution of failure rate in
the study-area for AC pipes is depicted in Figure 7.10 (right). While, in Figure 7.10
(left), white indicates the nonexistence of pipe network and/or soil ground movement
data, in Figure 7.10 (right), white indicates the absence of failures and of pipe
network.
The cells that contain older pipes (Figure 7.8 in red) are also cells of higher-risk
(Figure 7.10, left in red). Furthermore, the most number of failures (Figure 7.10,
right) were registered inside or close to the cells of higher risk. It should be noted
that there are high-risk cells without failures inside or in the vicinity. This indicates
that not all high-risk areas lead to failures, but that all failures are located inside or
in the vicinity of high-risk cells. Therefore, despite needing further validation, this
work suggests that the areas that are presented in Figure 7.10 (left) in red, could
be selected by the water utility to be inspected as discussed in Chapter 3 as they
represent a greater risk for the pipes installed inside or in their vicinity.
114 7 Correlating pipe failures & ground movement
Figure 7.8: Age distribution of AC pipes in the study-area (10 × 10 km). The scale,
illustrated in years, varies from green for cells (100 × 100 m) where the
pipes are, in average, younger pipes until red for cells with older pipes
(in average). A white dot indicates the nonexistence of AC network in
that cell.
Table 7.2: Impact matrix for AC pipes created to produce the risk map. A higher
grade indicates higher probability of failures. Each cell matrix is deter-
mined by adding the area-points (top) and age-points (left).
LM MM HM
Figure 7.9: Distribution of ground movement in the study-area (10 × 10 km). The
scale, given in a percentage (PDM), varies from green for more stable
cells (100×100 m) until red for more unstable cells. A white dot indicates
the nonexistence of ground movement data within that cell.
116
Figure 7.10: Left: risk map for AC for part of the study-area (10 × 10 km). The risk varies from low (green) to high (red). Right:
failure rate distribution in the same area, the scale (#.km-1 .year-1 ) varies from green for cells with low failure rate
until red for cells with higher failure rate. In both plots, a colored dot represents a cell (100 × 100 m) in which there
is pipe network and for which ground movement data was available. A white dot indicates a cell without AC network
and/or failures.
7 Correlating pipe failures & ground movement
7.4 Conclusions 117
7.4 Conclusions
In this Chapter, it was hypothesized that ground movement leads to stresses on
the pipes which increases the number of pipe failures. Therefore, areas with more
ground movement would indicate zones that are prone to high risk for pipe failure.
The analysis of the failure registration data together with the ground movement
data clearly demonstrated that the failure rate of all materials increases with the
level of ground movement, this relationship being obvious for both CI and AC.
Additionally, it was also demonstrated that the AC pipes installed prior to the
1960s fail more frequently than younger pipes.
For the pixel-based analysis, it was discovered that, for all materials in the vicinity
of failures, there were relatively more pixels with high PDM (> 90%) than for the
entire study-area. This further demonstrates the role played by differential soil
motion on the failures in an underground infrastructure. A cell-based risk map for
AC was produced. The map takes into consideration both the age of the material
and the level of ground movement in the soil where it was installed. Employing
the risk map, pipes requiring urgent action could be identified and prioritized for
condition assessment.
For this work, a total of 589 registered failures were exploited. More robust
analysis will require having access to more failure registration data. For example,
with access to more failure registration data, analysis can also focus on the impact
of ground movement in smaller and larger diameter pipes. This emphasizes the
significance of failure data registration and its importance for utility management.
Acknowledgement
The satellite data used was provided under the collaboration with the research that
has been carried out as part of the TU Delft project “Monitoring Surface Movement
in Urban Areas Using Satellite Remote Sensing” funded by Liander. The author
would like to thank the German Aerospace Center (DLR) for their support in pro-
viding TerraSAR-X time-series images.
chapter 8
General conclusions
In the old city of Petra (ca. 300 BC), among the dramatic scenery, there are tes-
timonies to the importance of drinking-water distribution for humankind where old
versions of pipes are carved in the red sandstone. In other parts of the world, still
standing Roman aqueducts not only pay homage to the Roman technical expertise∗ ,
but also emphasize the fact that urban development is only achievable with sus-
tainable access to drinking water. Much of this knowledge was lost during the dark
ages, and many of the techniques we recognize today had to be reinvented (Dead-
man, 2010). Currently, modern drinking water networks are constructed of plastic,
ductile iron, or multi-layered cement, but several other materials have been used in
the past (Savic and Banyard, 2011). Some of these older materials are still in use
and, together, they underpin civilization.
∗ At its prime, the daily per capita water consumption in Rome, according to Brooke and Quinn
(2013), exceeded 750 L, with Russo (2006) arguing that it reached almost 950 L.
119
120 8 General conclusions
Finally, one of the personal goals for the author was that this procedure “gained
life”, avoiding that it remained only a concept.
• Joint bending;
† Typical examples are the physical based models that predict corrosion growth in metallic
pipes.
8.3 Condition assessment of PVC push-fit joints 121
• Horizontal displacement;
• Pipe bending;
• Vertical displacement;
• Axial displacement;
• Torsion by slight rotation/vibration; and
• Pipe ovalization.
All of these failure modes failures are mechanical and material independent. In
this work, failures due to physical/chemical changes in PVC (e.g. embrittlement)
are not considered. This choice was made because it was noticed that the character-
ization of PVC’s physical/chemical condition is difficult to achieve non-destructively
in the field. Therefore, the focus was on the alignment of the pipes inside the joints
to characterize joint mechanical failures.
Of the aforementioned failure modes, the most significant for PVC push-fit are
joint bending and axial pull-out. Despite pipe bending and pipe ovalization being
related to joint failure, they are not intrinsic to the joint. In spite of this, both failure
modes can be derived from the assessment of the joint’s condition. It was discovered
during the field inspections (Chapter 3) that lateral movement is minimal when
compared to vertical movement and, therefore, horizontal displacement is negligible.
PVC joints, due to not being fixed to the pipes (e.g. glued), can be considered
immune to torsion‡ .
The connection between these failure modes and the joint alignment was the
beginning point in implementing the non-destructive assessment procedure.
‡ There are mechanically restrained PVC joints (trekvast, in Dutch). Nevertheless, these are
only used in very specific situations and are not considered here.
§ Except for torsion by slight rotation/vibration, which is not a relevant failure mode for PVC
push-fit joints.
122 8 General conclusions
Several tools were surveyed, and the three most promising were selected to be
tested in the laboratory. The tools were narrowed down to visual and ultrasound,
which are typically presented as being applicable to inspect polymeric pipes. Two
visual tools were selected (CCTV and Panoramo® ) and were rented from the Dutch
company MJ Oomen. The ultrasound equipment was rented from Applus RTD, also
Dutch. Panoramo® was abandoned after the laboratory tests as the tool failed to
deliver both accurate and reproducible results and struggled to produce a clear image
under water due to the reflection of light on the suspended matter. The ultrasound
was then discarded after the field test due to the difficulties in data interpretation,
data loss, and lack of a gyroscope - throughout the inspection it was not possible to
determine which ultrasound probe was facing down or up. Nevertheless, it must be
mentioned that these problems with the ultrasound are not intrinsic to ultrasound
itself, but to the specific ultrasound that was tested in the work. Of the three tools,
the CCTV was consistently the best, delivering both accurate and reproducible
results, therefore being the chosen tool to size the gaps in PVC joints. This happened
in spite of CCTV being presented by different authors as “prone to interpretation”.
This part of the work demonstrated that CCTV can deliver the accuracy and the
reproducibility necessary if employed for gap sizing.
¶ Ultrasound requires not only a dedicated launching/entry point but also a dedicated exit.
kA classical Dutch example would be a pipe installed in a dyke or, more commonly, a pipe
installed adjacent to a highway.
8.5 Destructive laboratory tests with PVC pipes & joints 123
The data exhibit an expected positive correlation between temperature and strain
on PVC, which was expected. Daily water demand patterns could be detected with
the strain gauges connected to the pipes and joints. Two confirmed episodes of
water-hammer were also detected by the sensors. This demonstrates the accuracy of
the strain gauges that were utilized and their potential in detecting dynamic loads
that can be risky for a pipe.
The work also demonstrated that, from the beginning of the monitoring period
until the coldest period for which data are available, the pipe contracted more than 5
mm at each end due to temperature changes. During the period of June-September,
the temperature can exceed the installation temperature. At that time, contact
points (pipe-pipe and pipe-joint) can begin if one considers that the pipes were in-
stalled fully inserted inside the joints. This emphasizes the significance of registering
the water temperature (inside the pipe) during a non-destructive assessment. In ad-
dition to this, the results also demonstrate that a joint that is considered to be in a
risky situation during winter (after inspection), for example, with both pipes almost
contacting, can be in an even riskier situation during summer upon pipe expan-
sion. This work also demonstrates that a pipe can be continuously monitored for
expansion/contraction when other assessment methods are not available.
Nevertheless, some of the project practicalities were evident from early on as more
than half of the sensors stopped working. Most sensors stopped working during a
cold period registered throughout February 2012 when the maximum air temperature
did not rise above 0 °C for more than two weeks. During this period, the batteries
also became depleted. When new batteries were installed, some of the sensors were
inoperative. Two strain gauges also became inoperative following an episode of
water-hammer. This further emphasizes that, although such a set-up can produce
beneficial and valuable data, its complexity is also its Achilles’ heel. Therefore, when
projecting such a set-up, this should be taken into consideration.
conducted. The tests were planned in order to characterize the condition of joints,
keeping in mind the failure modes described above. These tests were performed
together with DYKA, a Dutch PVC pipe manufacturer. Pipes and joints of two
diameters were tested - 110 and 315 mm. The material was tested under two different
conditions - water pressure (4 bar, absolute) and air sub-pressure (0.2 bar, absolute).
Both joint bending and axial pull-out tests were performed. To monitor the behavior
of the joint, its stiffness was monitored in real-time using a force sensor.
Several conclusions were obtained from this component of the work. First, joint
stiffness increases with bending angle, insertion, diameter, and pressure. An increase
in stiffness is considered dangerous as the joint becomes less able to bend. The tests
also demonstrated that the stiffness does not increase linearly with the bending angle
(for the same diameter and pressure). The joint becomes stiffer after a threshold
angle: 3° for 110 mm and 5° for 315 mm joints. These results are, for 10 m barrel
pipes, more conservative than the limit defined by the AWWAL (34° for 110 mm;
12° for 315 mm) and close to the limit defined by the Dutch PVC manufacturers:
6° for both diameters.
Second, the significance of not installing pipes completely inside the joints cannot
be stressed enough. The difference in joint behavior is obvious if the pipes are
completely inserted or inserted half-way between the joint’s center and the rubber
gasket∗∗ .
Finally, for PVC joints, leakage through the rubber-gasket is generally depen-
dent upon the condition of the rubber. For a rubber ring in good condition, leak-
age/intrusion can only be expected at bending angles above 10° and for complete
pull-out of the pipe from the joint. Such extreme angles were never detected in the
field (Section 3).
The results demonstrated that, given the degrees of freedom in a pipe system,
assessing the condition of joints cannot be accomplished with one parameter. For
∗∗ As often happens in science, this last discovery, although not the primary objective of the
tests, was surely one of the most relevant for the utilities and for the people in the field.
8.7 Correlating pipe failures and soil movement 125
this reason, both the individual joint grades and the overall pipe grades should
be adjacently analyzed. This approach also provided a better perspective of the
condition of individual joints in their role in the condition of the entire pipe.
The field inspections clearly demonstrated that, first, the alignment of pipes
inside the joints varies throughout the years. Second, performing inspections on
newly installed pipes can be exploited in order to assess the work of contractors
following pipe installation. Furthermore, the IJC is proven to be a powerful tool to
compare the condition of various pipes. Finally, the importance of individual joint
grading is demonstrated by the possibility of a network manager to perform selective
repairs in some joints in order to reduce the total pipe grade.
Failure registration data was aggregated from the Dutch MFD, USTORE (Sec-
tion 1.6.3). These data were exploited collectively with empirical ground movement
data measured by colleagues at TU Delft. The failure registration data encompassed
a period of 40 months during which 868 failures were registered. The results clearly
demonstrated that the failure rates for PVC, CI, and AC increase with the level of
soil movement. For the study-area, while CI in the material is most affected by soil
movement, AC has the greatest failure rates. What is more, there is a clear increase
in failure rate in the AC pipes installed prior to and after the 1960s.
These conclusions were the beginning point in creating a risk map for the study-
area. This map pin-points areas inside the distribution network that are expected
to be more failure-prone. These riskier areas should either be inspected more often
(pro-active on-condition AM) or replaced more often (pro-active hard-time AM).
Either way, this map will aid water companies in managing their networks more
efficiently.
As aforementioned, AC not only has the greatest failure rates but also an age
related-failure pattern that was not detected either in PVC or CI. For this reason, AC
was the selected material in order to produce the risk map. This decision was made
in spite of PVC being the material of interest in the present work and, therefore, a
comment must be made. First, the inspection procedure described in this work can
be applied to any other pipe material as long as there is a gap between the pipes.
Second, to characterize the condition of AC pipes requires only having access to
126 8 General conclusions
• The failure modes of push-fit joints have been surveyed and presented in detail;
• An inspection procedure has been developed. It is considered that the geo-
metrical alignment of the pipes inside the joints is a surrogate measurement
for joint condition. The geometrical alignment can be obtained with an NDE
tool;
• The most accurate and reproducible NDE tool (CCTV) for inspection of PVC
push-fit joints has been ascertained and fully characterized;
• The results of various inspections can be presented and compared side-by-side
by employing the IJC. The IJC can aid the water companies in selecting those
pipes needing the soonest replacement;
• Employing the risk map, risk-prone areas inside drinking water networks can
be identified, and the inspection/replacement of these pipes can be prioritized;
and
• The final result is a procedure that can be used to predict the remaining lifetime
of PVC joints.
1. First step: selecting the best candidate pipes for condition assessment. To
begin employing the procedure, a utility must have access to data allowing the
selection of the best candidate pipes for inspection†† . For this first step, the
utility must have access to failure registration data (Section 1.6). If the utility
is also provided access to soil movement data, a risk map (Chapter 7) can
be implemented. In the case that soil movement data are not available, the
utility can select the best candidate pipes solely based on failure registration
data. The utility can further narrow down the best candidate by performing
a risk-based analysis where both of the probability of failure and the expected
cost of failure (Kleiner et al., 2009) are obtained by using, for example, GIS.
†† Itis impossible to expect all pipes belonging to a utility to be inspected - selecting a few is
the only practical option.
8.8 Concluding remarks, implementation & future prospects 127
2. Second step: inspecting the joints. Probably only part of the best candidate
pipes can be inspected either due to time or financial constraints. Exploiting
an NDE‡‡ and following the procedure described in Chapter 3, the joints can
be inspected.
3. Third step: evaluating the joints’ condition. The condition of the joints can be
evaluated employing the IJC (Chapter 6). The IJC allows detecting the most
relevant failure modes as described in Chapter 2. The results produced with
the IJC support the water utility with their decision of whether to replace the
entire pipe or only part of it.
4. Fourth and final step: re-starting the process. The lifetime prediction proce-
dure is iterative, and the same joints can§§ be inspected several times through-
out their lifetime. When the results of the third step are available, an evalua-
tion of the procedure can be made: is it actually pin-pointing joints that are
in poor condition?
• Study threshold conditions for other materials and other types of joints:
– Determine the threshold conditions for other pipe materials, for example,
AC. For this, destructive laboratory tests with AC, similar to the ones
presented here for PVC, would have to be performed. This would allow
extending the lifetime prediction procedure to other materials.
– Restrained joints (trekvast, in Dutch) are employed in very specific situ-
ations, for example, to join pipe barrels 20 m before and after 90° elbows.
Due to being restrained, these joints are expected to behave differently
and to withstand different loads. Therefore, it would be necessary to
characterize the threshold condition of these joints in order to extend the
assessment/characterization procedure.
‡‡ The selected NDE can either be a CCTV or another tool that can also deliver accurate and
reproducible results
§§ From a scientific perspective, it can be argued that the same pipes and joints (e.g. a test
installation) should be inspected several times throughout a survey period of 5-10 years.
128 8 General conclusions
– The present work exploited data retrieved from USTORE database. Nev-
ertheless, the reliability of the registered data has never been investigated.
In order to validate research produced with the database, it is of the ut-
most importance to study the significance of the registration procedure
through tests made to the pipe-fitters. This will demonstrate how reliable
each registration is.
– In USTORE, the location of a failure can be reported (e.g. pipe, joint,
valve, etc.). An investigation is also necessary on how reliable these defi-
nitions are.
• This work demonstrated that the CCTV can be employed to assess the condi-
tion of pipes in the field and provide both accurate and reproducible results.
Therefore, the feasibility of implementing a CCTV inspection as a standard
inspection method to evaluate a contractor’s work should be studied. This
application would even minimize the problems associated with a CCTV in-
spection, i.e., the need to flush and to disinfect, since these steps are standard
before a newly installed pipe can be used for the first time. Additionally,
beginning seasonal (once/twice per year) inspections on the same pipe would
elucidate on the long term behavior of underground pipes.
• The condition of the rubber o-ring is also expected to play a role in the behav-
ior of joints throughout their lifetime, and this topic was thoroughly discussed
by Warnock (1999) in his Ph.D. thesis. According to the author, until the
development of synthetic rubber, the joint’s o-rings were made of natural rub-
ber. Natural rubber not only ages but can also degrade with the action of
micro-organisms. Both of these factors lead to a decrease in the sealing ca-
pacities of the joint (Warnock, 1999). Synthetic rubbers, even being immune
to biodegradation, also suffer aging and, accordingly, a decrease in sealing ca-
pacities. Furthermore, the presence of impurities (e.g. sand from the backfill)
creates an abrasion effect that diminishes the sealing capacities of the rubber
ring. Nevertheless, to the author’s knowledge, no NDE tool, either commer-
cially available or in R&D phase, has the capacity to assess the condition of
the rubber o-rings in-situ.
Andrews, M. E. (1998). Large diameter sewer condition assessment using combined sonar and
CCTV equipment. In APWA International Public Works Congress, NRCC/CPWA Seminar
Series Innovations in Urban Infrastructures, pages 91–103, Las Vegas (Nevada, USA).
Arai, Y., Koizumi, A., Inakazu, T., Watanabe, H., and Fujiwara, M. (2010). Study on failure rate
analysis for water distribution pipelines. Journal of Water Supply: Research and Technology-
AQUA, 59(6-7):429–435.
Arsénio, A. M., Bouma, F., Vreeburg, J. H. G., and Rietveld, L. (2013a). Characterization of PVC
joints’ behaviour during variable loading laboratory tests (submitted). Urban Water Journal.
Arsénio, A. M., Dheenathayalan, P., Hanssen, R., Vreeburg, J. H. G., and Rietveld, L. (2013b). Pipe
failure prediction in drinking water systems using satellite observations (submitted). Structure
and Infrastructure Engineering.
Arsénio, A. M., Pieterse-Quirijns, I., Vreeburg, J. H. G., de Bont, R., and Rietveld, L. (2013c).
Failure mechanisms and condition assessment of PVC push-fit joints in drinking water networks.
Journal of Water Supply: Research and Technology-AQUA, 62(2):78.
Arsénio, A. M., Vreeburg, J., and Rietveld, L. (2013d). Quantitative non-destructive evaluation of
push-fit joints. Urban Water Journal, pages 1–11.
Arsénio, A. M., Vreeburg, J. H. G., Bouma, F., and van Dijk, H. (2012a). Destructive laboratory
tests with PVC push-fit joints. In Water Distribution Systems Analysis (WDSA), Adelaide
(Australia).
Arsénio, A. M., Vreeburg, J. H. G., de Bont, R., and van Dijk, H. (2011). Real-life inline inspection
of PVC push-fit joints using NDE equipment. In Leading Edge on Strategic Asset Management
(LESAM), Mülheim an der Ruhr (Germany).
Arsénio, A. M., Vreeburg, J. H. G., de Bont, R., and van Dijk, H. (2012b). Real-life inline inspection
of buried PVC push-fit joints. Water Asset Management International, 8(2):30–32.
Arsénio, A. M., Vreeburg, J. H. G., Pieterse-Quirijns, E. J., and Rosenthal, L. (2009b). Overview
of failure mechanism of joints in water distribution networks. In Boxall, J. and Maksimović, C.,
editors, Computing and Control in the Water Industry (CCWI), pages 607–612, Sheffield (UK).
CRC Press.
Arsénio, A. M., Vreeburg, J. H. G., and Rietveld, L. (2013e). Index of joint condition for PVC
push-fit joints (submitted). Water Science & Technology: Water Supply.
129
130 Bibliography
Arsénio, A. M., Vreeburg, J. H. G., van Doornik, J., Dijkstra, L., and van Dijk, H. (2010). As-
sessment of PVC Joints Using Ultrasound. In Water Distribution Systems Analysis (WDSA),
Tucson (Arizona, USA). ASCE.
Arsénio, A. M., Vreeburg, J. H. G., Wielinga, M. P. C., and van Dijk, H. (2012c). Continuous
assessment of a drinking water PVC pipe. In Water Distribution Systems Analysis (WDSA),
Adelaide (Australia).
Asnaashari, A., Mcbean, E. A., Gharabaghi, B., and Tutt, D. (2013). Forecasting watermain failure
using artificial neural network modelling. Canadian Water Resources Journal, 38(1):24–33.
AWWA (2002). PVC pipe - Design and Installation. Manual of Water Supply Practices M23.
AWWA.
Bailey, M. and Kaufmann, C. (2006). Failure Analysis of PVC Pipe Joint Separations. Materials
Science and Technology - Association for Iron and Steel Technology, 1:165–176.
Beuken, R. H. S., de Kater, H., Postmus, E. P., and Mesman, G. A. M. (2011). In-line inspection
techniques for water mains put into perspective. In Leading Edge on Strategic Asset Management
(LESAM), Mülheim an der Ruhr (Germany).
Boersma, A. (2002). Long term performance prediction of existing PVC water distribution systems
I. Literature study on relevant parameters. Technical report, TNO, Eindhoven (The Nether-
lands).
Boersma, A. and Breen, J. (2003). Long term performance prediction of existing PVC water
distribution systems - V. Experiments and procedures. Technical report, TNO, Eindhoven (The
Netherlands).
Boersma, A. and Breen, J. (2006). Long-term performance prediction of existing PVC water
distribution systems - VII. Theory and modelling. Technical report, TNO, Eindhoven.
Breen, J. (2005). Long term performance prediction of existing PVC water distribution systems -
VI. Evaluation and recommendation for extended validation. Technical report, TNO, Eindhoven
(The Netherlands).
Breen, J. (2006). Expected lifetime of existing PVC water systems - Summary. Technical report,
TNO, Eindhoven.
Breen, J., Boersma, A., Slaats, P. G. G., and Vreeburg, J. H. G. (2004). Long term performance
prediction of existing PVC water distribution systems. In XII Plastic Pipes conference, Milan,
Italy.
Breen, J. and Etsu, S. (2003). Long term performance prediction of existing PVC water distri-
bution systems III. Physical properties of aged PVC. Technical report, TNO, Eindhoven (The
Netherlands).
Buco, J., Emeriault, F., and Kastner, R. (2008a). 3D numerical analyses of the soil variability
impact on longitudinal behaviour of buried pipes. In International Association for Computer
Methods and Advances in Geomechanics (IACMAG), Goa (India).
Bibliography 131
Buco, J., Emeriault, F., and Kastner, R. (2008b). Full-Scale Experimental Determination of Con-
crete Pipe Joint Behavior and Its Modeling. Journal of Infrastructure Systems, 14(3):230.
Budhu, M. (2010). Soil Mechanics and Foundations. John Wiley & Sons, Inc., 3rd edition.
Burn, S., Davis, P., Schiller, T., Tiganis, B., Tjandraatmadja, G., Candy, M., Gould, S., Sadler,
P., and Whittle, A. K. (2005). Long-term performance prediction for PVC pipes. AwwaRF.
Chan, D., Gallage, C., Gould, S., Kodikara, J., Bouazza, A., and Cull, J. (2009). Field Instrumen-
tation of a Water reticulation pipe buried in reactive soil. In Ozwater’09: Australia’s National
Water Conference and Exhibition, Melbourne (Australia).
Chan, D., Kodikara, J. K., Gould, S., Ranjith, P. G., Choi, X. S. K., and Davis, P. (2007).
Data analysis and laboratory investigation of the behaviour of pipes buried in reactive clay.
In Common Ground Proceedings - 10th Australia New Zealand Conference on Geomechanics,
pages 206–211, Brisbane (Australia).
Cima, J. and Peters, D. (2009). National Sewers and water mains failure database - Report Ref.
No. 08/RG/05/26. UKWIR, London.
Costello, S., Chapman, D., Rogers, C., and Mietje, N. (2007). Underground asset location and
condition assessment technologies. Tunnelling and Underground Space Technology, 22(5-6):524–
542.
Davis, P., Burn, S., Moglia, M., and Gould, S. (2007). A physical probabilistic model to predict
failure rates in buried PVC pipelines. Reliability Engineering & System Safety, 92(9):1258–1266.
Deadman, C. (2010). Strategic Asset Management: The Quest For Utility Excellence. Troubador
Publishing.
Demcenko, A., Akkerman, R., Nagy, P., and Loendersloot, R. (2012a). Non-collinear wave mixing
for non-linear ultrasonic detection of physical ageing in PVC. NDT & E International, 49:34–39.
Demcenko, A., Ravanan, M., Visser, H., Loendersloot, R., and Akkerman, R. (2012b). Investigation
of PVC physical ageing in field test specimens using ultrasonic and dielectric measurements. In
IEEE international ultrasonics symposium (IUS), Dresden (Germany).
Demma, A., Cawley, P., Lowe, M., Roosenbrand, A., and Pavlakovic, B. (2004). The reflection of
guided waves from notches in pipes: a guide for interpreting corrosion measurements. NDT &
E International, 37(3):167–180.
Devarajan, S., Miller, M. J., and Swanson, E. V. (2002). Goals for Development: History, Prospects
and Costs. Technical report, The World Bank, Washington (DC, USA).
Dheenathayalan, P., Cuenca, M. C., and Hanssen, R. (2011). Different approaches for PSI tar-
get characterization for monitoring urban infrastructure. In 8th International Workshop on
Advances in the Science and Applications of SAR Interferometry “FRINGE 2011”, Frascati
(Italy).
Dheenathayalan, P. and Hanssen, R. F. (2013). Radar target type classification and validation. In
IEEE International Geoscience and Remote Sensing Symposium, Melbourne (Australia).
Dingus, M., Haven, J., and Austin, R. (2002). Nondestructive, noninvasive assessment of under-
ground pipelines. AwwaRF, Denver (USA).
132 Bibliography
Dirksen, J. (2013). Monitoring ground settlement to guide sewer asset management. Ph.d., TU
Delft (Delft).
Dirksen, J., Clemens, F. H., Korving, H., Cherqui, F., Le Gauffre, P., Ertl, T., Plihal, H., Müller,
K., and Snaterse, C. T. (2013). The consistency of visual sewer inspection data. Structure and
Infrastructure Engineering, 9(3):214–228.
Dirksen, J., Veldhuis, J. A. E. T., Clemens, F. H. L. R., and Baars, E. J. (2010). Sensible sewer
system rehabilitation using information on sewer system settlement. In Novatech 2010, pages
1–9.
Doleac, M. L., Lackey, S. L., and Bratton, G. (1980). Prediction of time-to-failure for buried cast-
iron pipe. In Proceedings of the Annual Conference of the American Water Works Association,
pages 31–38, Denver (USA).
DYKA BV (2012). Price list March 2012 (in Dutch). Technical report, DYKA BV, Steenwijk.
Edwards, R. (2010). Asset Management - Whole-life Management of Physical Assets. ICE Pub-
lishing, London (UK).
Eiswirth, M., Frey, C., Herbst, J., Jacubasch, A., Held, I., Heske, C., Kramp, J., Munser, R., and
Wolf, L. (2001). Sewer assessment by multi-sensor systems. In IWA World Water Congress.
EPA (2002). The Clean Water and Drinking Water Infrastructure Gap Analysis. Technical report,
US EPA.
Farley, M., Trow, S., and Lossleakage, D. W. (2003). Losses in Water Distribution Networks - A
Practitioner’s Guide to Assessment, Monitoring and Control. IWA Publishing, London.
Farley, M., Wyeth, G., Ghazali, Z. B. M., Istandar, A., and Singh, S. (2008). The Manager’ s Non-
Revenue Water Handbook. Technical report, USAID and Ranhill Utilities Berhad, Bangkok
(Thailand).
Feeney, C. S., Thayer, S., Bonomo, M., and Martel, K. (2009). Condition Assessment of Wastewater
Collection Systems. Technical Report May, US EPA, Washington, DC (USA).
Ferretti, A., Prati, C., and Rocca, F. (2001). Permanent scatterers in SAR interferometry. IEEE
Transactions on Geoscience and Remote Sensing, 39(1):8–20.
Folkman, S. (2012). Water main breaks rates in the USA and Canada: a comprehensive study.
Technical report, Utah State University Buried Structures Laboratory.
Fukushima, Y., Cayol, V., and Durand, P. (2005). Finding realistic dike models from
interferometric synthetic aperture radar data: The February 2000 eruption at Piton
de la Fournaise. Journal of Geophysical Research, 110(B3):B03206. Available from:
http://doi.wiley.com/10.1029/2004JB003268.
Gaewski, P. E. and Blaha, F. J. (2007). Analysis of Total Cost of Large Diameter Pipe Failures.
Technical report, AwwaRF.
Gallage, C. P. K., Chan, D., Gould, S., and Kodikara, J. (2009). Behaviour of an in-service Cast
Iron water reticulation pipe buried in expansive soil. In Ozwater’09: Australia’s National Water
Conference and Exhibition.
Gallage, C. P. K., Kodikara, J., and Chan, D. (2011). Response of a plastic pipe buried in expansive
clay. Proceedings of the ICE - Geotechnical Engineering, 165(1):45–57.
Gallage, C. P. K., Kodikara, J. K., Chan, D., and Davis, P. (2008). A comparison of the results of
the numerical analysis and the physical behavior of a pipe buried in reactive clay. In International
Association for Computer Methods and Advances in Geomechanics (IACMAG), Goa (India).
Bibliography 133
Geudens, P. (2012). Dutch Drinking Water Statistics 2012. Technical report, VEWIN, Rijswijk
(The Netherlands).
Grigg, N. S. (2009). National Mains Failure Database Project. In World Environmental and Water
Resources Congress, pages 694–699, Kansas City (Missouri, USA).
Hanssen, R. F. (2001). Radar Interferometry: Data Interpretation and Error Analysis. Vol. 2.
Kluwer Academic Publishers, Dordrecht.
Herz, R. K. (1996). Aging processes and rehabilitation needs of drinking water distribution net-
works. Journal of Water Supply: Research and Technology-AQUA, 45:221–231.
Howard, A. K. (1977). Modulus of Soil Reaction Values for Buried Flexible Pipe. Journal of the
Geotechnical Engineering Division, 103(GT1):33–43.
Hu, Y., Vu, H. Q., and Lotfian, K. (2008). Instrumentation of asbestos cement pipe in expansive
soil. In American Society of Civil Engineering International Pipelines Conference, pages 1–10,
Atlanta (Georgia, USA).
Hudson, W. R., Haas, R., and Uddin, W. (1997). Infrastructure management: Design, construction,
maintenance, rehabilitation, renovation. McGraw-Hill, New York.
IAM (2012). Asset Management - an anatomy v1.1. Technical Report February, The Institute of
Asset Management, Bristol (UK).
IJkdijk (2009). Press release: Dike breaches predictable thanks to sensor systems. Available from:
http://www.ijkdijk.nl/en/experiments/piping.
ISO 13783 (1997). Plastics piping systems - Unplasticized PVC end-load-bearing double-socket
joints - Test method for leak tightness and strength while subjected to bending and internal
pressure.
Kirby, R., van der Walt, T., Abbott, D., Jones, P., and Stewart, D. (2006). International Infras-
tructure Management. Technical report, NAMS Group Ltd.
Kirkham, R., Kearney, P. D., Rogers, K. J., and Mashford, J. (2000). PIRAT - A System for Quan-
titative Sewer Pipe Assessment. The International Journal of Robotics Research, 19(11):1033–
1053.
KIWA (2007). Piping systems of PVC for the transport of drinking water and raw water. Technical
report, KIWA, Rijswijk.
Kleiner, Y. and Rajani, B. (2001). Comprehensive review of structural deterioration of water mains:
statistical models. Urban Water, 3:131–150.
Kleiner, Y. and Rajani, B. (2008). Prioritising individual water mains for renewal. In World Envi-
ronmental and Water Resources Congress, pages 1–10, Honolulu (Hawaii, US). ASCE/EWRI.
134 Bibliography
Kleiner, Y., Rajani, B., and Sadiq, R. (2009). Drinking Water Infrastructure Assessment: The
National Research Council of Canada Perspective. World Environmental and Water Resources
Congress 2009: Great Rivers, pages 67–67.
Kleiner, Y., Rajani, B., and Wang, S. (2007). Consideration of static and dynamic effects to plan
water main renewal. In Middle East Water 2007, 4th International Exhibition and Conference
for Water Technology, pages 1–13, Manama, Bahrain.
Kleiner, Y., Sadiq, R., and Rajani, B. (2006a). Modeling Deterioration and Managing Failure Risk
of Buried Critical Infrastructure. In Building Science Insight.
Kleiner, Y., Sadiq, R., and Rajani, B. (2006b). Modelling the deterioration of buried infrastructure
as a fuzzy Markov process. Water Supply Research and Technology - AQUA, 55(2):67–80.
KNMI (2012). Hourly weather data in the Netherlands (in Dutch). Available from:
http://www.knmi.nl/klimatologie/uurgegevens/#no.
Kuraoka, S., Rajani, B., and Zhan, C. (1996). Pipe-soil Interaction Analysis of Field Tests of
Buried PVC Pipe. Journal of Infrastructure Systems, 2(4):162–170.
Lange, G. D. (2003). Long term performance prediction of existing PVC water distribution systems
- IV. Literature study on external load factors. Netherlands Institute of Applied Geosciences
TNO, Utrecht (The Netherlands).
Lange, G. D., Bakr, M., Gunnink, J. L., and Huisman, D. J. (2012). A predictive Map of
Compression-Sensitivity of the Dutch Archaeological Soil Archive. Conservation and Manage-
ment of Archaeological Sites, 14(1-4):284–293.
Liu, Z. and Kleiner, Y. (2012). State of the art review of inspection technologies for condition
assessment of water pipes. Measurement, 46(1):1–15.
Liu, Z., Kleiner, Y., Rajani, B., Wang, L., and Condit, W. (2012). Condition Assessment Technolo-
gies for Water Transmission and Distribution Systems. Technical report, US EPA, Cincinnati
(US), US EPA.
Mackellar, S. and Pearson, D. (2003). Nationally Agred Failure Data and Analysis Metholody
for Water Mains. Volume 1: Overviw and Finding. Rep. No. 03/RG/05/7. Technical report,
UKWIR, London (UK).
Makar, J. M. (1999). Diagnostic techniques for sewer systems. Journal of Infrastructure Systems,
5(2):69–78.
Makar, J. M. and Kleiner, Y. (2000). Maintaining water pipeline integrity. In AWWA Infrastructure
Conference and Exhibition, Baltimore (Maryland, US).
Marlow, D., Davis, P., Trans, D., Beale, D., and Burn, S. (2009). Remaining Asset Life: A State
of the Art Review. Technical report, WERF, Alexandria (US).
Meijering, T. G., Wolters, M., and Hermkens, R. J. M. (2004). The durability of a low-pressure
gas distribution system of ductile PVC. In Plastics Pipes XII, Milan (Italy).
Mesman, G. and Wielen, J. V. D. (2005). Georadar: adequate to assess the condition of AC pipes?
(in Dutch). Technical report, Kiwa Water Research, Nieuwegein.
Methods, H., Walski, T. M., Chase, D. V., Savic, D. A., Grayman, W., Beckwith, S., and Koelle,
E. (2003). Advanced Water Distribution Modelling and Management. Haestad Methods.
Bibliography 135
Misiunas, D. (2005). Failure Monitoring and Asset Condition Assessment in Water Supply Sys-
tems. Ph.d., Lund University (Lund).
Moglia, M., Davis, P., and Burn, S. (2008). Strong exploration of a cast iron pipe failure model.
Reliability Engineering & System Safety, 93(6):885–896.
Moser, A. P. and Folkman, S. (2008). Buried Pipe Design. McGraw-Hill Professional, 3 edition.
Najjaran, H., Sadiq, R., and Rajani, B. (2004). Modeling pipe deterioration using soil properties
- an application of fuzzy logic expert sytem. In ASCE International Conference on Pipeline
Engineering and Construction, pages 1–10, San Diego (California, US).
NRC-CNRC (1997). Acquisition of Field Data on the Performance of Buried PVC Pipes - Thermal
Performance of Trench Backfills and Mechanical Performance of Buried PVC Water Mains for
1993-96. Technical report, NRC-NCRC.
Ofwat (2007). International comparison of water and sewerage service 2007 report. Technical
report, OFWAT.
Olliff, J. L., Rolfe, S. J., Wijeyesekera, D. C., and Reginold, J. T. (2001). Soil-Structure-Pipe
Interaction with Particular Reference to Ground Movement Induced Failures. In Proceedings of
Plastic Pipes XI, pages 1127–1140, Munich (Germany).
O’Rourke, T. D., Jezerski, J. M., Olson, N. a., Bonneau, a. L., Palmer, M. C., Stewart, H. E.,
O’Rourke, M. J., and Abdoun, T. (2008). Geotechnics of Pipeline System Response to Earth-
quakes. In Geotechnical Earthquake Engineering and Soil Dynamics IV, pages 1–38, Reston,
VA. American Society of Civil Engineers.
PBL (2007). Correction wording flood risks for the Netherlands in IPCC report. Available from:
http://www.pbl.nl/en/dossiers/Climatechange/content/correction-wording-flood-risks.
Pipelife (2001). Technical documentation for PVC pressure networks (in Dutch). Technical report,
Pipelife, Enkhuizen.
Prior, J. (1935). Investigation of bell and spigot joints in cast-iron water pipes. Part I-Pullout
strength. Part II-Bell strength. Part III-Harness strength.”. Technical report, College of Engi-
neering, Ohio State University, Columbus (OI, USA).
Rajani, B. and Kleiner, Y. (2001). Comprehensive review of structural deterioration of water mains:
physically based models. Urban Water, 3:151–164.
136 Bibliography
Rajani, B. and Kleiner, Y. (2002). Towards pro-active rehabilitation planning of water supply
systems. In International Conference on Computer Rehabilitation of Water Networks CARE-
W, pages 29–38, Dresden (Germany).
Rajani, B. and Kuraoka, S. (1995). Field Performance of PVC Water Mains Buried in Different
Backfills. In International Conference on Advances in Underground Pipeline Engineering II,
pages 138–149, New York (NY, USA). NRC, ASCE.
Rajani, B., McDonald, S., and Félio, G. (1993). Water Mains Break Data on Different Pipe
Materials for 1992 and 1993 (A-7019.1 Final). Technical report, NRC-CNRC, Ottawa (Ontario,
Canada).
Rajani, B. and Tesfamariam, S. (2004). Uncoupled axial, flexural, and circumferential pipe-soil
interaction analyses of partially supported jointed water mains. Canadian Geotechnical Journal,
1010:997–1010.
Rajani, B. and Zhan, C. (1996). On the estimation of frost loads. Canadian Geotechnical Journal,
33:629–641.
Rajani, B., Zhan, C., and Kuraoka, S. (1996). Pipe-soil interaction analysis of jointed water mains.
Canadian Geotechnical Journal, 33:393–404.
Rajeev, P. and Kodikara, J. (2011). Numerical analysis of an experimental pipe buried in swelling
soil. Computers and Geotechnics, 38(7):897–904.
Rajeev, P., Kodikara, J., Chiu, W. K., and Kuen, T. (2013). Distributed Optical Fibre Sensors
and their Applications in Pipeline Monitoring. Key Engineering Materials, 558:424–434.
Randall-Smith, M., Russell, A., and Oliphant, R. (1992). Guidance manual for the Research,
structural condition assessment of the trunk mains. Technical report, Water Research Centre,
Swindon (UK).
Reed, C., Robinson, A. J., and Smart, D. (2006). Potential techniques for the assessment of joints
in water distribution pipelines. AwwaRF, Denver (USA).
Reissner, E. (1959). On Final Bending of Pressurized Tubes. Journal of Applied Mechanics, pages
386–392.
Rietveld, H. (1984). Land subsidence in The Netherlands. In Johnson, A. I., Carbognin, L., and
Ubertini, L., editors, Proceedings of the Third International Symposium on Land Subsidence,
pages 455–464, Venice (Italy). IAHS.
Russo, R. (2006). Aqueduct Architecture: Moving Water to the Masses in Ancient Rome. Available
from: http://www.yale.edu/ynhti/curriculum/units/2006/4/06.04.04.x.html.
Savic, D. A. and Banyard, J. K. (2011). Water Distribution Systems. Thomas Telford Ltd, 1st
edition.
Sellmeijer, H., López, J. d. l. C., van Beek, V. M., and Knoeff, H. (2012). Fine-tuning of the back-
ward erosion piping model through small-scale, medium-scale and IJkdijk experiments. European
Journal of Environmental and Civil Engineering, 15(8):1139–1158.
Bibliography 137
Selvadurai, A. P. S., Hu, J., and Konuk, I. (1999). Computational modelling of frost heave induced
soil-pipeline interaction II. Modelling of experiments at the Caen test facility. Cold Regions
Science and Technology, 29(3):229–257.
Sepehr, K. (1994). Frost protection of buried PVC water mains in western Canada. Canadian
Geotechnical Journal, 31(4):491–501.
Shamir, U. and Howard, C. (1979). Analytic approach to scheduling pipe replacement. Journal of
American Water Works Association, 71(5):248–258.
Silva, D. D., Burn, L. S., and Eiswirth, M. (2001). Joints in Water Supply and Sewer Pipelines:
An Australian Perspective. In Wagga Wagga.
Slaats, P. G. G., Mesman, G. A. M., Rosenthal, L. P. M., and Brink, H. (2004). Tools to monitor
corrosion of cement-containing water mains. Water science and technology, 49(2):33–39.
Smolders, S., Verhoest, L., De Gueldre, G., and Van De Steene, B. (2009). Inspection of deteri-
orating asbestos cement force mains with georadar technique. Water science and technology,
60(4):995–1001.
Stiglitz, J. E. and Bilmes, L. (2008). The Three Trillion Dollar War. W. W. Norton.
Tabesh, M., Soltani, J., Farmani, R., and Savic, D. (2009). Assessing pipe failure rate and mechan-
ical reliability of water distribution networks using data-driven modeling. Journal of Hydroin-
formatics, 11(1):1–17.
UKWIR (2006). A comparison of leakage practice and leakage levels in UK and Netherlands (Rep.
ref. no. 06/WM/08/34). Technical report, UKWIR.
UKWIR (2008). National sewers and water mains failure database water data protocol. Technical
Report 1, UKWIR.
Van, M. A., Zwanenburg, C., Koelewijn, A. R., and Lottum, H. V. (2009). Evaluation of Full Scale
Levee Stability Tests at Booneschans and Corresponding Centrifuge Tests. In M. Hamza et al.,
editor, 17th International Conference on Soil Mechanics and Geotechnical Engineering, pages
2048–2051.
van der Helm, A. (2007). Integrated modeling of ozonation for optimization of drinking water
treatment. Ph.d., TU Delft (Delft).
Vangdal, A. C., Reksten, K., Saegrov, S., and Ugarelli, R. (2011). Experience from inspection
of water pipes - hard fact based AM - opportunities and consequences. In Leading Edge on
Strategic Asset Management (LESAM), Mülheim an der Ruhr (Germany).
Visser, R. (2009). Residual lifetime assessment of uPVC gas pipes. Ph.d., University of Twente.
Vloerbergh, I. N. and Blokker, E. J. M. (2010). Sharing failure data to gain insight into network
deterioration. Water Asset Management International, 6(2):9–14.
Vloerbergh, I. N., Schipdam, R., van Thienen, P., and Beuken, R. (2012). Pipe fitters assist to
predict investment needs for water main. Water Asset Management International, 8(3):12–18.
Warnock, J. S. (1999). A study of the factors influencing mechanical joint performance in water
pipelines. Ph.d., Imperial College of Science, Technology and Medicine.
Zhan, C., Goodrich, L., and Rajani, B. (1995). Thermal Performance of Trench Backfills for Buried
Water Mains. In Jeyapalan, J. K. and Jeyapalan, M., editors, Advances in Underground Pipeline
Engineering II, pages 650–661, New York (NY, USA). ASCE.
Glossary
139
140 Glossary
List of abbreviations
The following abbreviations are used in this thesis:
AC Asbestos cement
AM Asset management
AWWAL American Water Works Association Limit
CCTV Closed-circuit television
CI Cast iron
DI Ductile iron
GPR Ground-penetrating radar
IJC Index for joint condition
LVDT Linear variable differential transformer
MSS Multi-sensor system
(N)MFD (National) mains failure database
NDE Non-destructive evaluation
PE Polyethylene
RMSE Root mean squared error
ROV Remotely operated vehicle
(u)PVC (Unplasticized) polyvinyl chloride
2D Two dimensional
3D Three dimensional
List of publications
Peer-reviewed journals
• Arsénio, A. M., Vreeburg, J. H. G., de Bont, R., and van Dijk, H. (2012b).
Real-life inline inspection of buried PVC push-fit joints. Water Asset Manage-
ment International, 8(2):30–32
• Arsénio, A. M., Bouma, F., Vreeburg, J. H. G., and Rietveld, L. (2013a). Char-
acterization of PVC joints’ behaviour during variable loading laboratory tests
(submitted). Urban Water Journal
• Arsénio, A. M., Dheenathayalan, P., Hanssen, R., Vreeburg, J. H. G., and Ri-
etveld, L. (2013b). Pipe failure prediction in drinking water systems using
satellite observations (submitted). Structure and Infrastructure Engineering
143
144 List of publications
Conference proceedings
• Arsénio, A. M., Vreeburg, J. H. G., Pieterse-Quirijns, E. J., and Rosenthal,
L. (2009b). Overview of failure mechanism of joints in water distribution net-
works. In Boxall, J. and Maksimović, C., editors, Computing and Control in
the Water Industry (CCWI), pages 607–612, Sheffield (UK). CRC Press
• Arsénio, A. M., Vreeburg, J. H. G., van Doornik, J., Dijkstra, L., and van Dijk,
H. (2010). Assessment of PVC Joints Using Ultrasound. In Water Distribution
Systems Analysis (WDSA), Tucson (Arizona, USA). ASCE
• Arsénio, A. M., Vreeburg, J. H. G., de Bont, R., and van Dijk, H. (2011). Real-
life inline inspection of PVC push-fit joints using NDE equipment. In Leading
Edge on Strategic Asset Management (LESAM), Mülheim an der Ruhr (Ger-
many)
• Arsénio, A. M., Vreeburg, J. H. G., Bouma, F., and van Dijk, H. (2012a).
Destructive laboratory tests with PVC push-fit joints. In Water Distribution
Systems Analysis (WDSA), Adelaide (Australia)
Posters in conferences
• Arsénio, A. M., Pieterse-Quirijns, E. J., and Vreeburg, J. H. G. (2009a). Failure
mechanisms of joints in water distribution networks and its application on as-
set management. In Leading Edge on Strategic Asset Management (LESAM),
Miami (Florida, USA)