You are on page 1of 186

Expert Review of Water

System Pipeline Breaks in


the City of Los Angeles
during Summer 2009

Report to the Steering Committee on Water


Pipeline Breaks of the City of Los Angeles

J.P. Bardet (Chair), D. Ballantyne, G.E.C. Bell, A. Donnellan, S. Fos-


ter, T.S. Fu, J. List, R.G. Little, T.D. ORourke, and M.C. Palmer

April 9, 2010
Page | ii
EXECUTIVE SUMMARY
Background

During the summer of 2009, several high-profile water main breaks occurred in Los Angeles,
raising concerns from both the media and the public with respect to the safety and overall
condition of the Citys water distribution system. Several questions were raised, including:

What caused a sudden increase in water pipeline breaks?


Were the breaks related to pressure changes as a result of the Citys water-rationing
program?
Did the Monday/Thursday watering schedule cause particular damage to the Citys
older pipelines?
Were the Citys water pipelines being replaced in a timely manner?

As a consequence of the public outcry for an explanation, the Mayor and City Council re-
quested an independent investigation to study the cause of the increase in water pipeline
breaks and the overall vulnerability of the Citys water supply and distribution network. This
Investigation Team staffed by multidisciplinary experts in water distribution systems as well
as other related engineering fields, conducted its work apart from the Los Angeles Depart-
ment of Water and Power (LADWP).

Independent Investigation/Study

The Investigation Team analyzed data collected by LADWP and identified the main factors
contributing to the pipeline breaks that occurred during the summer of 2009. Specifically,
the Investigation Team considered: (1) material sciences, (2) geotechnical engineering, and
(3) statistics and geo-statistics. Apart from those considerations, the Investigation Team
also initiated a pilot study in hydraulics to measure pressure transients with digital sensors,
as well as to audit water pressures at key locations within the distribution system. Addition-
ally, the Investigation Team initiated a second pilot study in remote sensing, which was lead
by Jet Propulsion Laboratory (JPL), using state-of-the-art remote sensing and GPS tech-
niques to assess the effects of tectonic deformation, subsidence, and earthquakes on water
pipeline breaks. Finally, the Investigation Team conducted a review of asset management
practices for the Citys water distribution systems.

LADWP Water Distribution Systems

LADWP operates a complex water distribution system, extending over 7200 miles and is
comprised of pipelines made of various materials. Approximately 65% of LADWP water
pipelines are cast iron, a brittle material that becomes even more brittle when it is embedded
in corrosive soils.

Aging Infrastructure Cast Iron Pipes

Over 90% of blowouts observed since July 1, 2009 have occurred in cast iron pipes that
were externally corroded. A blowout is a term used by LADWP to describe a pipeline break

Page | iii
that heaves and undermines the street and requires pavement replacement for 100 ft2 or
more. Corrosion contributed to the pipe wall failure. Pipe walls failed both circumferentially
and longitudinally, resulting in the release of a substantial amount of water. This undermined
the structural integrity of overlying streets above, creating the blowouts and sink-holes that
attracted public attention.

In addition to cast iron pipe corrosion, two significant factors water temperature and inter-
nal pressure changes contribute to cast iron pipe failures. Temperature and pressure
changes increase stresses in the pipe walls and (coupled with external corrosion) diminish
pipe integrity.

Investigation Teams Findings

The Investigation Teams findings reveal a connection between the Citys water-rationing
program and the increase in pipe breaks during the summer of 2009, especially with cast
iron pipes,. At various locations in the LADWP water distribution system, the water pressure
dropped significantly on Mondays and Thursdays after the beginning of the water rationing
program on June 1, 2009. Those water pressure drops on these days were caused by an
increased water flow during the watering of lawns. As a result, the cyclic levels of water
pressure increased and accelerated the metal fatigue failures of aged and corroded cast
iron pipes.

These findings conclude that that the sudden changes of water pressure in the system, at-
tributable to the water-rationing program, had a negative impact on cast iron pipes with low-
er fatigue resistance (i.e. especially corroded cast iron pipes).

Investigation Team Recommendations

For the summer of 2010, the Investigation Team recommends avoiding abrupt variations
in water pressure as much as possible. To avoid increased levels of water main breaks,
it is recommended that LADWP alters its existing water-rationing program so that it
evenly distributes the variations of water pressure over time, and avoids sudden drops of
water pressure. For instance, LADWP could devise water rationing so that properties
with odd and even street numbers engage in watering lawns on different days. This solu-
tion would reduce sudden drops of water pressure and would impose less stress on cor-
roded cast iron pipelines.

Starting in 2010, the Investigation Team recommends that LADWP invests in research
aimed at:
o exploring and further refining our understanding of the factors affecting pipeline
failure
o developing tools that better quantify, understand, and predict system failures
o improving risk-based asset management of LADWP facilities

The Investigation Team also recommends that LADWP considers:


o a more efficient pipe replacement program as part of its asset management plan,
and improved field inspection techniques

Page | iv
o an aggressive pipeline replacement program that reduces the effects of aging
and increased vulnerability over the long term

Page | v
ACKNOWLEDGEMENTS
The authors thank the Los Angeles Department of Water and Power (LADWP) for their
complete collaboration during the analysis. They recognize in particular the contribution of
the following individuals:

Albert G. Gastelum, P.E., Manager of Project Engineering


Andrew L. Linard, P.E., Managing Water Utility Engineer
Craig Davis Ph.D., P.E., Geotechnical Engineer
Daniel Resch
Fred S. Barker, P.E., Waterworks Engineer
Greg Ammon, Water GIS Systems Manager
Gregory A. Loveland, P.E., Manager
George Mavrakis, Asset Management Group
Jamal H. Safa, P.E., Corrosion Engineer
Jianping Hu Ph.D., P.E.
Joe A. Castruita, Assistant Director
Jonathan K. Leung, Managing Water Utility Engineer
Jose C.E. Martinez, P.E., Associate Water Control Engineer
Joseph Ramallo, Director of Public Affairs
Julie M. Spacht, P.E., Water Executive Managing Engineer
Kevin J. Brown, P.E., Manager of Construction Support
Liping Yan
Lydia Alegria, Corrosion Engineering Staff
Stephan D. Tucker, Senior Asset Manager

The authors also thank the Jet Propulsion Laboratory (JPL), California Institute of Technolo-
gy for their contribution to the project. A portion of this research was carried out at the JPL
under a contract with the National Aeronautics and Space Administration. The authors rec-
ognize in particular the contribution of the following individuals:

Dr. Andrea Donnellan, QuakeSim Principal Investigator


Dr. Robert Granat,
Dr. Paul Lundgren
Dr. Jay Parker
Dr. Frank Webb,
Dr. Nevin A. Bryant, Principal
Margaret T. Glasscoe, Technical Staff
Dr. Cathleen E. Jones Senior Systems Engineer
Dr. Raymond J. Bambery, Principal Investigator

Page | vi
The authors thank the contribution of Hushmand Associates, Inc. (HAI), the project prime
contractor, in particular Ben Hushmand, Ph.D., P.E., President and Principal Engineer, HAI;
Ali Bastani, Ph.D., G.E., Vice President and Principal Engineer, HAI; and Nazila Mokarram,
Ph.D. Candidate, USC and Senior Staff Engineer, HAI.

The authors also thank the following contractors for their point of view on pipe installations,
in particular Richard G. Frye, Owner, Golden Bears Enterprises; and Paul E. Warnick, Own-
er, Golden Bears Enterprises.

Additionally, the authors thank the contribution of Mike Woodcock, Principal Engineer,
WSSC; and the engineers from Seattle, and other municipalities who have provided statis-
tics on water pipeline breaks.

Finally, the authors are thankful for the advice of:

Louis Herremans, Technical Director, Veolia Water


Lionel Gommery, Veolia Water
James Good, Veolia Water
J. Erick Heath, Vice President, MWH Soft
Donald Smith, Chairman, MWH
Paul Boulos, P.E., President, MWH Soft
Philip E. Smith, P.E., Vice President, MWH

Page | vii
CONTENTS

1 INTRODUCTION ................................................................................................................... 1
1.1 Background ............................................................................................................ 1
1.2 Objectives and scope of work ................................................................................ 1
1.3 Team of experts ..................................................................................................... 3
1.4 Organization of report............................................................................................. 4

2 MAIN EVENTS, LADWP SYSTEM, AND LITERATURE REVIEW ........................................ 5


2.1 Observations on water pipeline breaks .................................................................. 5
2.2 LADWP water distribution system ........................................................................ 14
2.2.1 Pressure Zones ........................................................................................ 14
2.2.2 Pipeline Properties.................................................................................... 15
2.3 Literature review on water pipeline breaks ........................................................... 19

3 MATERIAL ANALYSIS AND RESULTS .............................................................................. 21


3.1 History and nature of cast iron pipe...................................................................... 21
3.2 Failure modes and material properties of cast iron pipe ...................................... 23
3.2.1 Circular or beam breaks ........................................................................... 23
3.2.2 Longitudinal splits or breaks ..................................................................... 25
3.3 Brittle fracture, fracture toughness, flaws and damage accumulation .................. 28
3.4 Time dependent damage accumulation mechanisms for cast iron pipe: Corrosion
and fatigue ........................................................................................................... 29
3.4.1 Corrosion .................................................................................................. 29
3.4.2 Fatigue ...................................................................................................... 35

Page | viii
4 GEOTECHNICAL ANALYSIS OF BLOWOUTS .................................................................. 37

5 STATISTICAL ANALYSIS AND RESULTS ......................................................................... 41


5.1 Data collection practices at LADWP..................................................................... 41
5.1.1 Data sets used in analyses ....................................................................... 43
5.2 Statistical analysis of data collected ..................................................................... 44
5.3 Time series analysis ............................................................................................. 46
5.4 Comparison with other cities ................................................................................ 52
5.4.1 AWWA comparative study ........................................................................ 53
5.4.2 Present comparative study ....................................................................... 54
5.5 Spatial analysis .................................................................................................... 56
5.5.1 Spatial Distribution of Pipeline Breaks ...................................................... 56
5.5.2 Geo-statistics of pipeline break locations ................................................. 58
5.6 Variation of water pressure in LADWP system .................................................... 65
5.7 Variation of water temperature in the LADWP system ......................................... 69
5.8 Relations between pipeline breaks, temperature and pressure variations ........... 71
5.9 Effects of Corrosion Damage and Soil Corrosivity ............................................... 73
5.10 Explanation for Summer 2009 pipeline breaks..................................................... 83
5.10.1 Model assumptions ................................................................................... 83
5.10.2 Model formulation ..................................................................................... 84
5.10.3 Illustrative example ................................................................................... 87
5.11 Summary .............................................................................................................. 91
5.11.1 Probable explanations and recommendations .......................................... 92

6 HYDRAULIC PILOT STUDY ............................................................................................... 94


6.1 Water hammers and pressure surges .................................................................. 94
6.2 Pressure zones selected ...................................................................................... 96
6.3 Measurement of transients ................................................................................... 98
6.4 Additional Monitoring .......................................................................................... 106

7 REMOTE SENSING PILOT STUDY .................................................................................. 116


7.1 Background ........................................................................................................ 116
7.2 Pipeline break data............................................................................................. 117
7.3 SCIGN GPS time series ..................................................................................... 121
7.4 Covariance descriptor based analysis of transient deformation ......................... 126

Page | ix
7.5 InSAR data ......................................................................................................... 131
7.6 Summary ............................................................................................................ 135

8 ASSET MANAGEMENT .................................................................................................... 136


8.1 Best practices from AWWARF ........................................................................... 136
8.2 Review of asset management practices at LADWP ........................................... 137
8.3 Asset management as a risk management problem .......................................... 138
8.4 The LADWP Asset Management Program......................................................... 140
8.5 Recommendations ............................................................................................. 141

9 CONCLUSIONS ................................................................................................................ 143


9.1 Main findings ...................................................................................................... 143
9.2 Recommendations ............................................................................................. 145

10 REFERENCES .................................................................................................................. 147

11 APPENDICES .................................................................................................................... 152


11.1 Summary of literature review.............................................................................. 152
11.2 The team of experts ........................................................................................... 156
11.3 Seattle Water System (by Don Ballantyne, MMI) ............................................... 160

Page | x
TABLE OF FIGURES
Figure 2-1. The fire engine drove over an unsupported layer of asphalt pavement and fell
into to a hole that was created by an underlying water pipe leak at 12212
Hartsook Street at Bellingham Avenue on September 8, 2009. ........................ 6
Figure 2-2. Closer view of the transverse failure of the water pipe break at 12212 Hartsook
Street on September 8, 2009. This mode of failurewhen the pipeline breaks
due to bendingis also referred to as a beam break.The two pieces of the
pipes may have separated as the fire engine fell into the hole. ......................... 6
Figure 2-3. Break of a large-diameter riveted steel pipe along Coldwater Canyon Avenue at
Dickens Avenue on September 8, 2009. This water pipe break created
extensive damage to the street pavement. ........................................................ 7
Figure 2-4. Longitudinal failure of 8-inch diameter cast iron pipe at the intersection of Corbin
and Kittridge on September 9, 2009. The pipe displayed graphitic corrosion.... 7
Figure 2-5. Repaired section of asphalt pavement at the intersection of Corbin and Kittridge.
Damaged was caused by the longitudinal failure of an 8-inch diameter cast iron
pipe on September 9, 2009. .............................................................................. 8
Figure 2-6. Segment of a cast iron pipe that broke on Melrose Avenue on September 14,
2009. The pipe segment was removed and brought to the LADWP yard for
analysis. This mode of pipeline failure, which is usually caused by a water
surge or water hammer, results in a sudden and energetic release of water
pressure. ............................................................................................................ 8
Figure 2-7. Longitudinal failure of 6inch cast iron pipe located at 20741 Burbank Boulevard,
on September 16, 2009. The pipe displayed graphitic corrosion. The crack
length was about 7 feet in length. This pipe break resulted in about 400 square
feet of damage to street asphalt pavement. ...................................................... 9
Figure 2-8. Comparison of total number of blowouts during summer months (July, August,
and September) between 2001 and 2009. ...................................................... 10
Figure 2-9. Comparison of ratio of blowouts to leaks during summer months (July, August,
and September) between 2001 and 2009. ...................................................... 11
Figure 2-10. Cumulative numbers of breaks (leaks and blowouts) and blowouts between
2001 and 2009. The annual number of breaks and blowouts corresponds to
constant slopes as indicated on graph. ........................................................... 11
Figure 2-11. Location of LADWP water pipeline breaks (circles) and blowouts (triangles) in
Summer 2009 (July, August and September). ................................................. 12
Figure 2-12. Location of LADWP water pipeline breaks (circles) and blowouts (triangles)
between July 2001 and October 2009. ............................................................ 13
Figure 2-13. Geographic location of pressure zones in LADWP water distribution system.
Gravity-driven pressure zones are shown in green, while the pump-regulated
pressure zones are shown in gray. Each zone number is identified by the
highest elevation in that zone. ......................................................................... 15
Figure 2-14. Distribution of LADWP pipe lengths in terms of main material types (top left),
age (top right) and diameters (bottom). ........................................................... 16

Page | xi
Figure 2-15. Distribution of pipelines by main material types in LADWP water distribution
system. ............................................................................................................ 18
Figure 2-16. Three-dimensional representation of pipe length distribution in terms of main
material types and pipe diameters. .................................................................. 18
Figure 2-17. Three-dimensional representation of pipe length distribution in terms of main
material types and pipe ages. .......................................................................... 19
Figure 3-1. Right: ductile iron microstructure (NOT Brittle); left: cast iron microstructure
(Brittle) ............................................................................................................. 22
Figure 3-2. Low carbon steel microstructure very ductile and high fracture toughness ........ 23
Figure 3-3. "Circular Break" (Circumferential Crack Propagation) due to beam loading at
12212 Hartsook Street (Figure 2-2) ................................................................. 24
Figure 3-4. Failure morphology along edge of rapid circumferential crack propagation due to
beam loading from circular blowout at 12212 Hartsook Street. ....................... 24
Figure 3-5. Axial View of cast ironlongitudinal break at Corbin Avenue and Kittridge Street 25
Figure 3-6. Longitudinal propagation of crack due to radial stress at Corbin Avenue and
Kittridge Street ................................................................................................. 26
Figure 3-7. Surface indication of underground break Corbin Avenue and Kittridge Street:
Note support has not been lost so the break is less likely to be beam or circular
type break. ....................................................................................................... 26
Figure 3-8. Repair to riveted steel pipe at Coldwater Canyon Avenue and Dickens Street.
Note repair steel plate on bottom runs back to riveted joint. ............................ 27
Figure 3-9. Riveted steel pipe section removed at Coldewater Canyon Avenue and Dickens
Street. Note longitudinal split and deformation of steel due to internal pressure
forces and possibly removal. ........................................................................... 28
Figure 3-10. Bimodal distribution of failure of historical service CI Pipe (Atkinson, et al. 2002)
......................................................................................................................... 29
Figure 3-11. Example of a corrosion cell .............................................................................. 31
Figure 3-12. Graphitized layer on external surface of iron pipe sample ............................... 32
Figure 3-13. Relationship between soil resistivity, soil pH and corrosion rates on ferrous
metals (King 1977) ........................................................................................... 33
Figure 3-14. LADWP trend for cast iron leak rate vs. soil resistivity range (LADWP 2010a).34
Figure 3-15. LADWP trend for steel pipe leak rate vs. soil resistivity range (LADWP 2010b)
......................................................................................................................... 34
Figure 3-16. LADWP data for leak rates by diameter of CI (LADWP 2010a)........................ 35
Figure 4-1. Typical profile of a trench, bedding and surface pavement for water pipes for
analysis of blowouts. ........................................................................................ 38
Figure 4-2. Illustration of various types of pavement failures caused by water pipe
leaks/breaks: (a) a small water leak from pipe collected in a drain under the
pavement surface or emerging onto the street pavement; (b) a large pipe
breaks releasing large quantity of water under pressure, and breaking up the
street pavement; and (c) medium-size water leak washing away the soil

Page | xii
bedding, gradually creating a cavity under the street pavement due to soil
erosion, and resulting in collapse of surface pavement when vehicles pass
over it. .............................................................................................................. 40
Figure 5-1. Information flow for collecting pipeline break data .............................................. 42
Figure 5-2. Information flow for collecting pipeline repair data ............................................. 43
Figure 5-3. Pipe material compositions (left); average ages by pipe materials (right); data
source: File E. .................................................................................................. 45
Figure 5-4. Leaks per 100 miles of pipeline (left); leaks per 100 miles of pipeline and
average age (right); data source: File E. ......................................................... 45
Figure 5-5. Leaks per 100 miles of pipeline in different pressure zones. Only pressure zones
with more than 30 miles of pipelines are included. Zone 0 is a virtual pressure
zone for pipes not assigned to a specific zones. ............................................. 46
Figure 5-6. LADWP analysis (note: a different data source on Summer 2009) .................... 47
Figure 5-7. Independent analysis from this report (note: same data source for all summers:
File D) .............................................................................................................. 47
Figure 5-8. Distribution of leaks (July 2001- October 2009) in term of pipe material types
using (top) total number of leaks, (leftmost bottom) percentage of pipe length in
various pipe materials, and (bottom) percentage of leaks in various pipe
materials; data source: File D. ......................................................................... 48
Figure 5-9. Distribution of blowouts (July 2001- October 2009) in terms of pipe material type
using (top) total number of blowouts, (leftmost bottom) percentage of pipe
length in various pipe materials, and (bottom) percentage of blowouts in
various pipe materials; data source: File D. Note: only 51% of all blowout data
is used, and 2001 and 2009 are results based on partial data sets. ............... 49
Figure 5-10. Distribution of annual leaks (July 2001- October 2009) in terms of pipe diameter
using (top) total number of leaks, (leftmost bottom) percentage of pipe length in
various pipe diameters, and (bottom) percentage of annual leaks in various
pipe diameters; data source: File D. Note: 2001 and 2009 results are based on
partial data sets. .............................................................................................. 50
Figure 5-11. Distribution of annual blowouts (July 2001- October 2009) in terms of pipe
diameters using (top) total number of annual blowouts, (leftmost bottom)
percentage of pipe length in various pipe diameters, and (bottom) percentage
of annual blowouts in various pipe diameters; data source: File D. Note: only
51% of all blowout data is used and2001 and 2009 results are based on partial
data sets. ......................................................................................................... 50
Figure 5-12. Zone 579: Distribution of annual leaks (July 2001- October 2009) in terms of
pipe material types using (top) total number of leaks, (leftmost bottom)
percentage of pipe length in various pipe materials, and (bottom) percentage of
leaks in various pipe materials; data source: File D. Note: 2001 results are
based on a partial data set, and there are no leak data in 2009 for Zone 579. 51
Figure 5-13. Zone 579: Distribution of annual blowouts (July 2001- October 2009) in terms
of pipe material type using (top) total number of blowouts, (leftmost bottom)
percentage of pipe length in various pipe materials, and (bottom) percentage of
blowouts in various pipe materials; data source: File D. Note: only 51% of all

Page | xiii
blowout data are used; 2001 is a partial data set and there is no blowout data
in 2009 for Zone 579. ....................................................................................... 52
Figure 5-14. Breakdown of total pipe length per pipe in terms of material type in six
municipal systems labeled AF (AWWA, 2007) ). System C is LADWP. ........ 53
Figure 5-15. Breakdown of water pipeline breaks per total length in terms of material type in
six municipal systems labeled AF (AWWA, 2007). System C is LADWP. ..... 54
Figure 5-16. Breakdown of total pipe length in terms of by pipe material for seven municipal
water utilities; LA data source: File E. .............................................................. 54
Figure 5-17. Two-dimensional breakdowns of total pipe length in terms of pipe material type
and pipe diameter for Los Angeles, and cities U, V, and X. LA data source: File
E. ..................................................................................................................... 55
Figure 5-18. Breaks per 100 miles of pipeline for LA, and for cities W, X, Y, Z, plus
boundaries for national averages; LA leak data sources: File A; LA pipe data
source: File E. .................................................................................................. 56
Figure 5-19. Contours showing annual breaks per 100 miles of pipeline in all pipes (left) and
cast iron pipes (right). ...................................................................................... 57
Figure 5-20. Definition of geo-statistical variables used for pipeline leaks, blowouts, and
properties: (1) a square area A is drawn centered at a leak point; (2) number of
leaks (or blowouts) is added up within Area A; (3) the length of each pipe is
recorded and summed up with Area A; and (4) the average properties of pipes
are calculated based on a weighted average that uses the length of individual
pipe as weight factors. ..................................................................................... 58
Figure 5-21. Number of pipeline leaks per square mile (July 2001 to October 2009); data
source: File B. .................................................................................................. 60
Figure 5-22. Spatial distribution of blowouts per square mile (July 2001 to October 2009);
data source: File B. .......................................................................................... 61
Figure 5-23. Spatial distribution of ratio between blowouts and leaks (July 2001 to October
2009); data source: File B. ............................................................................... 62
Figure 5-24. Spatial distribution of pipe density (length in miles) per square mile; data
source: File F. .................................................................................................. 63
Figure 5-25. Spatial distribution of annual leaks per 100 mile of pipelines (July 2001 to
October 2009); leak data source: File B; pipe length data source: File F. ....... 64
Figure 5-26. Spatial distribution of blowouts per 100 miles of pipeline (July 2001 to October
2009); leak data source: File B; pipe length data source: File F. ..................... 65
Figure 5-27. Time histories of water pressures at eight locations during 2009. .................... 66
Figure 5-28. Locations of pressure measurements. ............................................................. 66
Figure 5-29. Time histories of water pressures (minute intervals) in June 2009................... 67
Figure 5-30. Variation of water pressures (daily means, maxima and minima) in 2009. ...... 68
Figure 5-31. Lowering of water pressures on Mondays and Thursdays. .............................. 69
Figure 5-32. 2009 water temperature variations at three locations. ...................................... 70
Figure 5-33. Seasonal water temperature variations between 1996 and 2009. ................... 70

Page | xiv
Figure 5-34. Time histories of pipeline breaks, water temperatures (Figure 5-32) and
pressures (Figure 5-27); leak data source: File A. ........................................... 72
Figure 5-35. Time histories of pipeline breaks and water temperatures; leak data source: File
A. ..................................................................................................................... 73
Figure 5-36. Soil corrosivity map; corrosivity data source: File F.......................................... 74
Figure 5-37. Distribution of the number of pipeline breaks and blowouts in term of soil
corrosivity; leak data source: File B; pipe data source: File F. ........................ 75
Figure 5-38. Variation of pipeline breaks and blowouts as a function of soil resistivity; leak
data source: File B; pipe data source: File F. .................................................. 76
Figure 5-39. Variation of pipeline breaks and blowouts as a function of soil resistivity for
different pipe materials; leak data source: File B; pipe data source: File F. .... 77
Figure 5-40. Variation of pipeline breaks and blowouts as a function of soil resistivity for
different pipe diameters; leak data source: File B; pipe data source: File F. ... 78
Figure 5-41. Variation of pipeline breaks and blowouts as a function of soil resistivity for
different pipe ages; leak data source: File B; pipe data source: File F. ........... 79
Figure 5-42. Variation of pipeline breaks and blowouts as a function of soil corrosivity and
pipe age; leak data source: File B; pipe data source: File F. ........................... 80
Figure 5-43. Pipeline system in terms of pipe age and soil resistivity. .................................. 81
Figure 5-44. Spatial distribution of soil corrosivity (top) and pipeline leaks per 100 miles of
pipeline (bottom); leak data source: File B; pipe data source: File F. .............. 82
Figure 5-45. Ranges of fatigue behaviors for materials in general. ...................................... 84
Figure 5-46. Due to material fatigue, the number of pressure cycles before a pipe fails
decreases as the amplitude of cyclic pressure increases. ............................... 84
Figure 5-47. Assumed uniform distribution of pipe quantity as a function of number N of
pressure cycles. ............................................................................................... 85
Figure 5-48. Illustration of increase in number of fatigueinduced failures due to an increase
in amplitude of pressure cycles from S1 to S2 during time interval t. The
number of cycles to reach pipe failure decreases abruptly from N1 to N2 during
t, and returns to N1 when the amplitude returns to S1. The break rate, which
was initially constant, jumps suddenly, then becomes constant again during t.
When the amplitude decreases back to S1, the breaks stop during N1-N2 days,
then the breaks start again at the original constant rate. ................................. 86
Figure 5-49. Distribution of cast iron pipe length in term of pipe age in LADWP water
distribution system ........................................................................................... 87
Figure 5-50. Distribution of pipe quantity as a function of N number of daily pressure cycles
assuming an average pipe length of 20 feet. ................................................... 88
Figure 5-51. Assumed distribution of pipe quantity as a function of number of cycles to
failures. ............................................................................................................ 88
Figure 5-52. S(N) fatigue curves obtained for gray cast iron and different imperfection size
a/w (after Mohebbi et al., 2009), and estimated ranges of cyclic pressure
amplitudes and number of pressure cycles (red area)..................................... 90

Page | xv
Figure 5-53. Simplified explanation for increase of pipe breaks caused by fatigue and an
increase in the amplitude of pressure cycles. .................................................. 91
Figure 6-1. 579 pressure zone with leaks and breaks .......................................................... 97
Figure 6-2. Pressure recorder at intersection of Van Ness and Sunset (11/20/2009
11/26/2009)...................................................................................................... 99
Figure 6-3. Pressure recorder at intersection of Van Ness and Sunset (11/27/2009
12/3/2009)........................................................................................................ 99
Figure 6-4. Pressure recorder at intersection of Van Ness and Sunset (12/4/2009
12/10/2009).................................................................................................... 100
Figure 6-5. Pressure recorder at intersection of Van Ness and Sunset (12/11/2009
12/17/2009).................................................................................................... 100
Figure 6-6. Pressure recorder at intersection of Van Ness and 3rd (11/20/2009
11/26/2009).................................................................................................... 101
Figure 6-7. Pressure recorder at intersection of Van Ness and 3rd (11/27/2009
12/3/2009)...................................................................................................... 101
Figure 6-8. Pressure recorder at intersection of Van Ness and 3rd (12/4/2009
12/10/2009).................................................................................................... 102
Figure 6-9. Pressure recorder at intersection of Van Ness and 3rd (12/11/2009
12/17/2009).................................................................................................... 102
Figure 6-10. Pressure recorder at intersection of Van Ness and Sunset. Approximate time of
pipeline break in Hancock Park. .................................................................... 103
Figure 6-11. Pressure recorder at intersection of Van Ness and 3rd. Approximate time of
pipeline break in Hancock Park is noted on the figure. .................................. 104
Figure 6-12. Monitored pump stations. ............................................................................... 106
Figure 6-13. Pressure recorder on the suction side of Winnetka pump station. ................. 108
Figure 6-14. Pressure recorder on the discharge side of Winnetka pump station. ............. 109
Figure 6-15. Pressure recorder on the suction side of Encino pump station. ..................... 109
Figure 6-16. Pressure recorder on the discharge side of Encino pump station. ................. 110
Figure 6-17. Pressure recorder on the suction side of Beverly Glen pump station. ............ 110
Figure 6-18. Pressure recorder on the discharge side of Beverly Glen pump station......... 111
Figure 6-19. Pressure recorder on the suction side of Coldwater Canyon pump station. ... 111
Figure 6-20. Pressure recorder on the discharge side of Coldwater Canyon pump station.
....................................................................................................................... 112
Figure 6-21. Pressure recorder on the suction side of Roxbury pump station. ................... 112
Figure 6-22. Pressure recorder on the discharge side of Encino pump station. ................. 113
Figure 6-23. Pressure recorder on the discharge side of Roxbury pump station on
2/17/2010. ...................................................................................................... 114
Figure 7-1. Area serviced by the LADWP. Regions and GPS stations presented in this
analysis are marked. ...................................................................................... 118

Page | xvi
Figure 7-2. Cumulative total leaks including blowouts for separate regions within the LADP
water district. .................................................................................................. 119
Figure 7-3. Cumulative blowouts for regions within the LADWP jurisdiction....................... 119
Figure 7-4. Total leaks including blowouts binned by month for July 1, 2001 through
September 30, 2009. Red indicates leaks and blue indicates blowouts. ...... 120
Figure 7-5. Total number of blowouts plotted by mouth for the period July 2001 end of
September 2009. ........................................................................................... 120
Figure 7-6. Locations of the continuous GPS stations in the Los Angeles region. Stations
considered in this study are shown in Figure 7-1. ......................................... 121
Figure 7-7. Typical high-precision continuous GPS station. The pipes are drilled 10 meters
into the Earth. A ray dome protects the choke ring antenna, which is designed
to minimize noise and ensure a clean radio signal from the GPS satellites. . 122
Figure 7-8. GPS time series for six stations distributed across the water district. Top time
series in blue is north component, middle green time series is east component,
and bottom time series in yellow is vertical or up component ..................... 123
Figure 7-9. GPS time series for six stations distributed across the water district plotted in a
larger format. The time period is consistent with the water pipe data, which
range from July 2001 October 2009. Top time series in blue shows north
component, middle green time series is east component, and bottom time
series in yellow is vertical or up component ................................................ 124
Figure 7-10. Map view of Fairfax High School, located south of the Hollywood Hills. Red
dates indicate blowouts and blue dates indicate pipeline breaks. Locations are
marked by red triangles or dots respectively. FXHS GPS station is indicated by
the yellow box and the time series for the period May 19, 2009 through
November 3, 2009. Blue vertical lines indicate clear offsets in the time series,
with the dates noted. The red vertical lines show the times of the pipeline
breaks and blowouts ...................................................................................... 125
Figure 7-11. Long-term time series for FXHS. Time series are for north, east, and up from
top to bottom .................................................................................................. 126
Figure 7-12. Divergence time series for all stations for 2009.167.001253. The vertical axis
displays time (in days), while the horizontal axis displays stations (in
alphabetical order by station name) ............................................................... 128
Figure 7-13. The most anomalous GPS stations as determined by average covariance
divergence for each of the Phase II, Group A data sets ................................ 129
Figure 7-14. Top 20 most divergent GPS stations for the time period November 2008
November 2009 ............................................................................................. 131
Figure 7-15. Still image from a movie of deformation in California as observed by the
European radar satellite. The motion is clearly bounded by the Newport-
Inglewood fault (Data provided by Paul Lundgren, animation created by Vince
Realmuto, JPL). ............................................................................................. 133
Figure 7-16. Interferogram from the Canadian RADARSAT for the time period August 24,
2002 September 30, 2004. Different colors indicate different amounts of
motion in the direction of the satellite. The change in color indicates a
difference in motion across a region. ............................................................. 133

Page | xvii
Figure 7-17. Pipeline break data for the same time period as the interferogram in Figure
7-16. The data suggest a concentration of breaks south of the Hollywood Hills,
consistent with the substantial deformation in that area. There is a weaker
suggestion of increased breaks along the southern San Fernando Valley .... 134
Figure 7-18. Interferogram for the time period September 12, 2003 September 30, 2004.
Different colors indicate different amounts of motion in the direction of the
satellite. The change in color indicates a difference in motion across a region
....................................................................................................................... 134
Figure 7-19. Pipeline break data for the same time period as those shown in Figure 7-18.
....................................................................................................................... 135
Figure 8-1. Main replacement strategies should balance risk and cost .............................. 138

Page | xviii
TABLE OF TABLES

Table 1-1. List of experts who participated in analysis and their primary role......................... 3
Table 2-1. Main material types used in LADWP water distribution system. .......................... 17
Table 2-2. Factors affecting water pipeline breaks classified in terms of demand and
capacity............................................................................................................ 20
Table 5-1. Data files collected............................................................................................... 44
Table 5-2. Explanations of statistical variables. .................................................................... 59
Table 5-3. Parameters and values used for simplified fatigue analysis. ............................... 90
Table 5-4. Parameters and values used for estimating the increase in pipe breaks due to
fatigue. ............................................................................................................. 91
Table 6-1. Pump Station supply and delivery zones. .......................................................... 107
Table 7-1. Offsets for each component of motion for FXHS. .............................................. 126
Table 7-2. Clusters of stations found using the covariance descriptor method. ................. 130
Table 7-3. Means and covariances that describe the locations of the transient signals ..... 131
Table 8-1. Risk Options. ..................................................................................................... 140
Table 11-1. Seattle Water System Pipe Material ................................................................ 160
Table 11-2. Pipe Joint Type ................................................................................................ 161
Table 11-3. Pipe Length by Material and Diameter (miles) ................................................. 162
Table 11-4. Number and Percentage of Pipe Failures by Category for All Materials.......... 163
Table 11-5. Number and Percentage of Pipe Failures by Category for Cast Iron Pipe ...... 164
Table 11-6. Pipe Break Rates/100 miles of pipeline/Year for the Period of 1990-2002 ...... 165
Table 11-7. Break Rates Using Modified Method for the Period of 2000-2008 .................. 165
Table 11-8. Water Main Repairs on 4-inch and Larger Pipe Requiring Pipe Segment
Replacement .................................................................................................. 166

Page | xix
1 INTRODUCTION

1.1 Background
During the summer of 2009, Los Angeles experienced several high-profile water trunk line and
main breaks that raised concerns about the performance of the Los Angeles Department of Wa-
ter and Power (LADWP) water distribution system. Some of the water main and trunk line
breaks attracted significant media attention. For example, a pipeline break on September 8,
2009 caused pavement collapse with attendant undermining and damage to a fire engine. The
disruptive water pipeline breaks resulted in questions from the media and public concerning the
water distribution system.

Why there was a sudden increase in water pipeline breaks?


Were they related to pressure changes caused by the City's new water rationing sys-
tem?
Was the watering on Mondays and Thursdays creating sudden stresses and damage to
older pipelines?
Were the old water pipelines in Los Angeles being replaced in a timely manner?

On September 23, 2009, the Mayors Office and the Los Angeles City Council (by motion) re-
quested a thorough and independent investigation of the increase in high-profile pipeline breaks
by a multidisciplinary team of experts. The City appointed the Summer 2009 Pipe Leak Investi-
gation Steering Committee (Steering Committee) to monitor the investigation. The Steering
Committee consists of a representative from the LADWP, Bureau of Engineering, Chief Admin-
istrative Officer, and Chief Legislative Analyst.

1.2 Objectives and scope of work


This report summarizes the findings of the external experts who were assembled to provide a
third-party assessment for the Steering Committee of (1) potential causes of the LADWP water
main breaks during the summer of 2009 and (2) overall vulnerability of the water supply and dis-
tribution system of buried pipeline networks in Los Angeles. The assessment of this team of ex-
perts was carried out independently of the ongoing pipeline break investigation performed by
LADWP.

The scope of work includes the formation, implementation, and administration of a team of ex-
perts to perform an external third-party assessment of the potential causes of pipeline breaks in

1
1. Introduction
the LADWP water distribution network. The external and expert assessment is independent of
internal investigations by LADWP.

The scope of work is derived from the last paragraph of the Los Angeles City Council motion
that requires the DWP and the Bureau of Engineering, in conjunction with the City Administra-
tive Officer and the Chief Legislative Analyst, initiate an independent third party assessment to
review the Departments water system and determine the cause(s) of the recent water main
breaks, including the impact of water use restrictions, and provide recommendations for ad-
dressing the problem.

The scope of work includes the following:

1. Develop a team of experts to review concerns relating to the buried water supply and distri-
bution system, that may include, but are not limited to:
Soil corrosivity
Heavy traffic loading
Geodetic movements, soil settlement, and ground subsidence
Temperature, pressure, and other pertinent physical properties
Pipe materials, size, type, and dates of construction
Soil backfill
Water quality
Metal fatigue
System operations (and operational changes)
Water use changes
Other environmental factors and data (as deemed important and necessary to meet the
purpose of this task)
2. Provide expert review and opinion of the water system infrastructure program and compari-
son with industry standards including a statistical analysis of the water pipeline breaks in re-
cent years
3. Review pertinent documents and data pertaining to the water supply and distribution system
including any recent information on testing soil samples and pieces of pipeline in the areas
of the main breaks
4. Evaluate and determine potential causes of water pipeline breaks
5. Advise on advanced technology research and data analysis needs for water system infra-
structure including:
Evaluation of stressors that can lead to system component damages (e.g., pipeline
leaks)
Identification of any potential weak links
Potential operational and community impacts related to damage
6. Develop recommendations of possible improvement and mitigation strategies

Page | 2
1. Introduction

1.3 Team of experts


The team of experts was made up of nine experts covering various aspects of water distribution
systems. Table 1-1 lists the team members and their affiliation, and identifies their main contri-
butions to the project. The team was assembled progressively by J. P. Bardet in order to ad-
dress the various fields of expertise required to analyze the relevant information that was re-
leased by LADWP. During this analysis, the team of experts had access to LADWP information
and interacted with LADWP personnel as listed in Acknowledgements.

Table 1-1. List of experts who participated in analysis and their primary role.

Name and Affiliation Primary Role


Jean-Pierre Bardet, Professor Assemble and lead the team of experts
Chair, Sonny Astani Department of Civil and Envi- Supervise the overall evaluation and peer re-
ronmental Engineering view work
Director, Center on Megacities Assist on all aspects of the work described be-
University of Southern California low.
Report to the Steering Committee
Tat Fu Perform statistical analysis and GIS analysis
Postdoctoral Fellow
Sonny Astani Department of Civil and Environmen-
tal Engineering, University of Southern California
Thomas D. ORourke Perform engineering review and evaluation of
Thomas R. Briggs Professor of Engineering the pipeline system
School of Civil and Environmental Engineering Provide engineering evaluation of the LADWP
Cornell University large, geographically distributed water supply
Michael C. Palmer, PH.D., P.E. systems and underground construction tech-
Instructor/Research Associate nologies.
School of Civil and Environmental Engineering
Cornell University
E. John List, Ph.D., P.E. Investigate the effect of transient pressures in
Principal Consultant water distribution system
Flow Science Incorporated Supervise deployment of sensors for measur-
R. Scott Foster, P.E. ing transients
Vice President Interpret sensor reading
Flow Science Incorporated
Graham E.C. Bell, Ph.D., P.E. Examine the effects of external corrosion and
Schiff Associates metal fatigue
Provide additional data sets from other cities
Donald Ballantyne Provide and analyze data sets on water pipe-
Senior Consultant line breaks from Seattle
MMI Engineering
Richard G. Little, Director Analyze assets management
Keston Institute for Public Finance and Infrastruc-
ture Policy, University of Southern California
Andrea Donnellan Conduct remote sensing pilot study
QuakeSim Principal Investigator
Jet Propulsion Laboratory (JPL)

Page | 3
1. Introduction

1.4 Organization of report


The report is organized in nine sections that describe the multi-disciplinary study of the water
pipeline breaks and LADWP system. The fields of expertise involved in this study include statis-
tics, geo-statistics, geographic information systems (GIS), material sciences, geotechnical engi-
neering, hydraulics, finance and economics, and remote sensing.

Following the introduction, the second section of the report summarizes the principal characte-
ristics of the water pipeline breaks in Los Angeles during the summer of 2009, and describes
previous work and associated technical literature relevant to this study.

The third, fourth and fifth sections summarize three different analyses that look at water pipeline
breaks from different perspectives including (1) material sciences, (2) geotechnical engineering,
and (3) statistics and geo-statistics respectively. The influence of materials and corrosion on
the pipeline degradation process is discussed in the third section. The fourth section attempts
to explain blowout mechanisms by modeling street pavement collapse caused by pipeline
breaks. The fifth section provides a probable explanation for the increase of water pipeline
breaks in the summer of 2009, referring to statistics about LADWP pipeline performance, which
was collected and analyzed by the external expert team.

The sixth section summarizes the hydraulic pilot study that measured pressure transients using
highly accurate sensors deployed at several locations within the water distribution system.

The seventh section summarizes the remote sensing pilot study that the Jet Propulsion Labora-
tory performed using state-of-the-art remote sensing and GPS techniques.

The eighth section addresses some basic principles of asset management for water distribution
systems.

Finally the last section summarizes the main findings of each section, and makes recommenda-
tions that should help reduce pipeline breaks in the summer of 2010. The last section provides
both short-term and long-term recommendations, including measures to reduce cyclic loading of
corroded cast iron pipelines and guidance for research on integrated remote-sensing inspection
and asset management.

Page | 4
2 MAIN EVENTS, LADWP SYSTEM,
AND LITERATURE REVIEW

This section reviews a few examples illustrating the rash of water pipeline break in Los Angeles
during the summer of 2009, which prompted this analysis. This section also summarizes the lite-
rature relevant to the analysis, and describes the main characteristics of the LADWP water dis-
tribution system.

2.1 Observations on water pipeline breaks


The incidents started when a water main broke in the San Fernando Valley on September 5,
2009. Another spectacular pipeline break occurred at 12212 Hartsook Street, Los Angeles, on
September 8, 2009. During that incident, a fire engine fell into the hole that suddenly opened up
when the unsupported asphalt pavement collapsed under the fire engines weight. This dramatic
incident attracted the attention of media, which covered the water pipeline breaks as an unprec-
edented series of events in Los Angeles, and raised the public concerns to new heights.

Figure 2-1 to Figure 2-7 illustrate a few examples of incidents that took place in Los Angeles
between September 8 and 16, 2009. As shown in Figure 2-1 and Figure 2-2, the asphalt pave-
ment collapsed under the fire engine weight. In this case, the cast iron pipe failed in bending,
and the water it released washed out the bedding sand and removed the support to the asphalt
pavement. As shown in Figure 2-3, the failure of a large-diameter riveted steel pipe caused ex-
tensive damage in view of the large volume of water it suddenly released. As shown in Figure
2-4 and Figure 2-7, many cast iron pipes failed longitudinally during the summer of 2009, and
this mode of failure was often associated with extensive damage to street pavements (Figure
2-5). As shown in Figure 2-6, in a few extreme cases, large pieces of cast iron pipes failed cata-
strophically, releasing large amounts of water.

5
2. Main Events, LADWP System and Literature Review

Figure 2-1. The fire engine drove over an unsupported layer of asphalt pavement and fell
into to a hole that was created by an underlying water pipe leak at 12212 Hartsook Street
at Bellingham Avenue on September 8, 2009.

Figure 2-2. Closer view of the transverse failure of the water pipe break at 12212 Hartsook
Street on September 8, 2009. This mode of failurewhen the pipeline breaks due to
bendingis also referred to as a beam break.The two pieces of the pipes may have se-
parated as the fire engine fell into the hole.

Page | 6
2. Main Events, LADWP System and Literature Review

Figure 2-3. Break of a large-diameter riveted steel pipe along Coldwater Canyon Avenue
at Dickens Avenue on September 8, 2009. This water pipe break created extensive dam-
age to the street pavement.

Figure 2-4. Longitudinal failure of 8-inch diameter cast iron pipe at the intersection of
Corbin and Kittridge on September 9, 2009. The pipe displayed graphitic corrosion.

Page | 7
2. Main Events, LADWP System and Literature Review

Figure 2-5. Repaired section of asphalt pavement at the intersection of Corbin and Kit-
tridge. Damaged was caused by the longitudinal failure of an 8-inch diameter cast iron
pipe on September 9, 2009.

Figure 2-6. Segment of a cast iron pipe that broke on Melrose Avenue on September 14,
2009. The pipe segment was removed and brought to the LADWP yard for analysis. This
mode of pipeline failure, which is usually caused by a water surge or water hammer, re-
sults in a sudden and energetic release of water pressure.

Page | 8
2. Main Events, LADWP System and Literature Review

Figure 2-7. Longitudinal failure of 6inch cast iron pipe located at 20741 Burbank Boule-
vard, on September 16, 2009. The pipe displayed graphitic corrosion. The crack length
was about 7 feet in length. This pipe break resulted in about 400 square feet of damage to
street asphalt pavement.

LADWP generally defines blowouts as the water pipeline breaks that generate damage of sur-
face pavement in excess of 100 square feet. There are no real scientific/engineering reasons for
this particular area size; it is used because that level of damage requires the LADWP to interact
with the Bureau of Street Services to resurface the streets. Blowouts are usually associated with
potential hazards to traffic. They can also cause extensive property damage due to the intensity
of water flow, including interruption of services, impact to adjacent business activities, flooding
of neighboring properties, and subsequent water damage to properties.

As shown in Figure 2-8, there were 101 blowouts in Los Angeles in the summer of 2009, twice
as many as during the summer of 2008, and much more than during previous summers (sum-
mers for this report are defined as the months of July, August and September). A large number
of blowouts also occurred in the summer of 2004 but did not receive comparable attention from
the media. Figure 2-9 compares the increase in blowouts in relation to the total number of leaks
plus blowouts. This graph indicates a definite and unprecedented increase in disruptive water
pipeline breaks during the summer of 2009. This rash of blowouts prompted this analysis.

Figure 2-10 shows the variations of the cumulative numbers of breaks (leaks and blowouts) and
blowouts between 2001 and 2009. These variations are nearly linear over a 10-year period (the
coefficient r2 of linear regression is equal to 0.998.), and the slope of these variations corres-
ponds to the average annual number of breaks (1,512) and blowouts (205). From the perspec-
tive of Figure 2-10, which contrasts sharply with Figure 2-8, one may argue that the blowouts
that occurred during Summer 2009 were within expected ranges of failure for the LADWP sys-
tem.

Page | 9
2. Main Events, LADWP System and Literature Review
Figure 2-11 shows the location of blowouts that occurred in Summer 2009 (between July and
September). The blowouts occur at various locations in various pressure zones of the LADWP
water distribution system. As shown in Figure 2-11, there is a noticeable concentration of blo-
wouts in the Southern part of the San Fernando Valley, in comparison to other water pipeline
leaks, and a concentration of blowouts and water leaks in the Hollywood area. For reference,
Figure 2-12 shows the location of water leaks and blowouts between 2001 and 2009.

120
TotalNumberofBlowoutsDuringSummer

101
100

80
80
69

60 57
47 48 49
42
38
40

20

0
2001 2002 2003 2004 2005 2006 2007 2008 2009

Figure 2-8. Comparison of total number of blowouts during summer months (July, Au-
gust, and September) between 2001 and 2009.

Page | 10
2. Main Events, LADWP System and Literature Review

25%
23%

RtaioofBlowoutstoLeaksDuringSummer
21%
20%
17%
15% 16%
14%
15% 13%
11% 11%

10%

5%

0%
2001 2002 2003 2004 2005 2006 2007 2008 2009

Figure 2-9. Comparison of ratio of blowouts to leaks during summer months (July, Au-
gust, and September) between 2001 and 2009.

14000
Cumulativenumberofbreaks

12000

10000
Breaks:1512peryear
andblowouts

8000

6000

4000

2000 Blowouts:205peryear
0
2001 2003 2005 2007 2009
Year
Figure 2-10. Cumulative numbers of breaks (leaks and blowouts) and blowouts between
2001 and 2009. The annual number of breaks and blowouts corresponds to constant
slopes as indicated on graph.

Page | 11
2. Main Events, LADWP System and Literature Review

Figure 2-11. Location of LADWP water pipeline breaks (circles) and blowouts (triangles)
in Summer 2009 (July, August and September).

Page | 12
2. Main Events, LADWP System and Literature Review

Figure 2-12. Location of LADWP water pipeline breaks (circles) and blowouts (triangles)
between July 2001 and October 2009.

Page | 13
2. Main Events, LADWP System and Literature Review

2.2 LADWP water distribution system


Founded in 1902, the LADWP is the largest municipal utility in the United States, serving more
than four million residents. The LA water distribution system consists of 7,238 miles of pipeline,
and it operates with an annual budget of $1 billion. In the fiscal year 2007-2008, 199 billion gal-
lons of water were delivered with an average use of 144 gallons per person per day. The water
supply comes from the LA Aqueduct (35%), Metro Water District (53%), groundwater (11%) and
recycled water (1%) (LADWP, 2009).

2.2.1 Pressure Zones


The LADWP water distribution system is divided into 114 pressure zones. Figure 2-13 shows
the pressure zones operated by LADWP with each zone being identified by its highest elevation
(in feet above mean sea level). These pressure zones are either gravity driven or pump driven.

Page | 14
2. Main Events, LADWP System and Literature Review

Figure 2-13. Geographic location of pressure zones in LADWP water distribution system.
Gravity-driven pressure zones are shown in green, while the pump-regulated pressure
zones are shown in gray. Each zone number is identified by the highest elevation in that
zone.

2.2.2 Pipeline Properties


Figure 2-14 breaks down the 7,238 miles of LADWP pipelines in terms of materials, diameters,
and age. The pipe materials are divided into five main categories asbestos cement (AC),
cast iron (CI), ductile iron (DI), steel (STL) and others (other and unidentifiable materials). As
shown in Table 2-1, these five categories are groups of pipe material types. Each category con-
tains pipelines that were manufactured by different sources and were put into service at different
times. As shown in Figure 2-14, cast iron pipes makes up most of the LADWP water distribution

Page | 15
2. Main Events, LADWP System and Literature Review
system; 65% of the total length of the LADWP system is made of cast iron (Figure 2-15). As
shown in Figure 2-14, about 200 miles of pipelines are older than 100 years, but the majority of
the system is about 60 years old. Most of the pipelines have a diameter between 4 and 12 inch-
es.

5000 1600

1400
4000
1200
Pipe length (miles)

Pipe length (miles)


1000
3000
800

2000
600

400
1000
200

0 0
AC CI DI STL others 10 20 30 40 50 60 70 80 90 100
Pipe materials Pipe age (years)

10000

3000

1000
Pipe length (miles)

100

10

2 4 6 8 12 16 20 25 30 35 40 45 50 75 100 125 144


Pipe diameters (in)

Figure 2-14. Distribution of LADWP pipe lengths in terms of main material types (top left),
age (top right) and diameters (bottom).

Page | 16
2. Main Events, LADWP System and Literature Review
Table 2-1. Main material types used in LADWP water distribution system.

PIPE MATERIAL PIPE TYPE DESCRIPTION


AC AC Asbestos Cement
Asbestos Cement TRANS Transite; Asbestos Cement
AM American; Cast Iron
BEL Belgian; Cast Iron
C Cast Iron
Cast Iron; Manufacture and Process Un-
CI known
CI DEL Delavaud Cast Iron
Cast Iron FR French; Cast Iron
GR German; Cast Iron
MCW McWane; Cast Iron
MONO Monolith; Cast Iron
RDW R.D. Wood; Cast Iron
US United States; Cast Iron
DI DI Ductile Iron
Ductile Iron DIA Ductile Iron (AWWA)
DRIV Double Riveted; Steel Pipe
M Matheson - Milled; Steel Pipe
MANN Mannesman - Milled; Steel Pipe
MATH Matheson - Milled; Steel Pipe
STL
RIV Riveted; Steel Pipe
Steel
STD Standard; Steel Pipe
STL Steel Pipe
STL GALV Galvanized; Steel Pipe
TRIV Triple Riveted; Steel Pipe
W Steel Pipe
CMP Corrugated Metal Pipe
CONC Concrete
COP Copper
HDPE High Density Polyethylene
Others
MPCC Modified Prestressed Concrete Cylinder
PCC Prestressed Concrete Cylinder
PVC Polyvinyl Chloride
RC Reinforced Concrete
RCC Reinforced Concrete Cylinder

Page | 17
2. Main Events, LADWP System and Literature Review

% by length
8% 2%
14%
AC
CI
DI
STL
others 11%

65%

Figure 2-15. Distribution of pipelines by main material types in LADWP water distribution
system.

2500

2000
Pipe length (miles)

1500

1000

500

4
6
8
12
16
20
25
30
35
4045
Pipe <= diameters (in) 50
75 CI
STL
100 DI
125 AC
144 others Pipe materials

Figure 2-16. Three-dimensional representation of pipe length distribution in terms of


main material types and pipe diameters.

Page | 18
2. Main Events, LADWP System and Literature Review
Figure 2-16 shows the three-dimensional distribution of pipe length in terms of pipe materials
and diameters. Most pipelines are 4-12 inches in diameters and cast iron is the dominant ma-
terial in this diameter range. Steel is used more often for large diameters (25 inches and
above). Figure 2-17 demonstrates the pipe age range compared to materials and lengths.
Cast iron is the oldest material, followed by steel and asbestos cement; ductile iron pipes are
installed most recently.

1500
Pipe length (miles)

1000

500

10
20
30
40
50
60
70
Pipe age (years) 80 CI
90 STL
DI
100 AC
others Pipe materials

Figure 2-17. Three-dimensional representation of pipe length distribution in terms of


main material types and pipe ages.

2.3 Literature review on water pipeline breaks


The American Water Works Association (AWWA, 2007) reports that water utilities in the United
States suffer 250,000 to 300,000 main breaks per year. These cause total annual damages and
indirect consequences of about $3 billion. AWWA (2007) conducted an in-depth analysis and
survey, the results of which are summarized in Appendix 11.1. The AWWA survey yielded the
following findings:

Average annual break rate (for all materials) is 25 breaks per 100 miles of pipeline per
year
Cast iron break rates are significantly higher than other materials

Page | 19
2. Main Events, LADWP System and Literature Review
A utilities break rate is the most important criterion in prioritizing replacements
Utilities vary in the methods they use to select mains for replacement
Utilities see assessing each mains life as a clear issue but consider cost assessment to
be a lesser problem
Utilities prefer that research focus on segments rather than categories of mains
Only a few utilities have tried to apply break prediction methods
Availability of consequences data by utilities is very limited
Utilities are renewing their mains very slowly, once every 200 years or less.

Water pipelines break in many different ways and for many reasons. Based on AWWA (2007),
Table 2-2 summarizes the main factors that may contribute to water pipeline breaks, sorted in
terms of capacity and demand. Capacity encompasses various pipe characteristics that allow
the pipes to resist the demands to which they are exposed.

Table 2-2. Factors affecting water pipeline breaks classified in terms of demand and ca-
pacity.

Demand Capacity
System operations and operational Manufacturing defects
changes Degradation of material due to
Water hammer o internal corrosion
Pressure surges and transients o external corrosion
Changes in water use due to water ration- o fatigue
ing Aging
Changes in water temperature Loss of ductility and increase in brittleness
Changes in surrounding temperature, in- Poor quality control and inspection
cluding freeze thaw Soil corrosivity
Heavy traffic loading Thrust blocks
Nearby construction activities Pipe materials, size, type, and dates of
Geodetic movements construction
Transient ground motion from earthquakes Backfill
Permanent ground motion from earth- Lining
quakes Weak joints
Ground subsidence
Internal pressure (static and cycles)
Water quality
Soil corrosivity
Construction accidents

The objectives of this study are to differentiate among these factors to identify those that most
affected the increase in water pipeline breaks in Los Angeles during the summer of 2009.

Page | 20
3 MATERIAL ANALYSIS AND RE-
SULTS1

Pipelines leak, blowout, or catastrophically fail when the mechanical/structural demands of the
system exceed the mechanical/structural capacity of the pipeline at the time when the leak or
blowout occurs. When a pipeline is new, catastrophic failures are not common because design
practices use known design conditions along with engineering safety factors to insure that ca-
pacity exceeds demand when the pipeline is fabricated and installed. By definition, prior to the
failure, the capacity always exceeded demand. Over time, both demand and capacity change.
Damage to the pipeline during installation, operation, and interaction with the internal and exter-
nal environments accumulates over time and reduces pipe capacity. In general, demand in-
creases with community development. Capacity decreases as material properties and condition
deteriorate due to environmental and operational degradation mechanisms. For example, from
the soil environment, corrosion removes sound metal and reduces the mechanical/structural
capacity of the pipeline where the corrosion occurs. From the operation standpoint(depending
on the pipe material), pressure transients, spikes, or surges can result in damage accumulation
and reduction of mechanical/structural capacity.

Data and analysis both from this study and the Department indicate that leaks and blowouts are
dominated (90+%) by cast iron (CI) pipes 8inches in diameter and smaller. Because of this, the
influence of materials and corrosion on the process will focus on CI. Understanding the material
characteristics and degradation mechanisms of CI material and pipes are important to under-
standing failures and the pattern of failures.

3.1 History and nature of cast iron pipe


A cast iron pipe is a historical metal that found widespread use as a pressure pipe for transmis-
sion of water, gas and sewage, and as a water drainage pipe, during the 19th and 20th centu-
ries. CI has been used for potable water for centuries. CI pipe comprises predominantly a gray

1
This section of the report is authored by Graham Bell, Schiff Associates, Claremont.

21
3. Material Analysis and Results
cast iron tube and was frequently used uncoated, however, later developments did result in var-
ious coatings and linings to reduce corrosion and improve hydraulics. The earliest recorded in-
stallation of cast-iron pipe was in 1455 at the Dillenburg Castle in Germany (DIPRA 1984).
Cast-iron pipe was installed in the United States as early as 1804 in Philadelphia (Stroud 1985).
At least ten utilities in the United States have cast iron mains that have served continuously for
150 years or more (DIPRA 1990). Cast iron pipe was gradually superseded by ductile-iron pipe
with most existing manufacturing plants transitioning to the new material during the 1970s and
1980s. Currently, there is almost no new manufacture of cast iron pressure pipe. The Depart-
ments inventory of pipes reflects this transition in iron pipe materials as a function of time.

Both cast and ductile iron are ferrous alloys consisting primarily of iron (93%+), carbon (~3%)
and silicon (~2%). Both ductile-iron and cast iron pipe have approximately the same percent
carbon by weight. In cast iron, the carbon is present in the form of flakes interconnected
throughout the metal matrix. These flakes create planes of weakness within the metal that limit
the mechanical properties that can be achieved, and thus are responsible for the characteristic
low ductility of gray iron. In ductile-iron pipe, carbon is present in the form of discrete nodules or
spheroids that do not create planes of weakness. The result is better continuity of the ferritic
structure (see Figure 3-1). The spheroidal structure of the graphite in ductile iron is accom-
plished by an additional treatment of the molten iron, usually with magnesium, and by exercising
greater control over certain embrittling agents during the manufacturing process. The result is a
form of cast iron with much greater strength and ductility, which is called ductile iron. Steel is
also a ferrous alloy but contains much less carbon (less than 0.2%). Carbon in steel alloys is
much more finely distributed and carbon phases do not interconnect (See Figure 3-2), which
results in much greater ductility (strain or extension to failure) and toughness (energy to failure).

Figure 3-1. Right: ductile iron microstructure (NOT Brittle); left: cast iron microstructure
(Brittle)

Page | 22
3. Material Analysis and Results

Figure 3-2. Low carbon steel microstructure very ductile and high fracture toughness

Over the 150+ years of manufacturing, cast iron pipe has been manufactured in sizes ranging
from 2 inches to 84 inches in diameter, and in various laying lengths from 3- to 20-feet. The first
cast iron pipes were statically cast in horizontal molds. The position of casting the molds
changed from horizontal to a slope, and finally to a vertical position around the year 1850. Cen-
trifugal casting methods have been in the process of commercial development and refinement
since 1925. The Department has both vertically cast and centrifugally cast iron in their pipe in-
ventory.

3.2 Failure modes and material properties of cast iron pipe


With that background and understanding, the most important aspect to understand about CI is
that it is by its nature, microstructure with interconnected brittle carbon phases. Cl is a more brit-
tle material compared to steel or ductile-iron pipe materials. Having said that, brittle materials
are not necessarily weak. That is to say that when uniaxial tension is applied in a uniform fa-
shion, the stress required to failure the material is not necessarily small by comparison to other
materials. However, the amount of energy (force times displacement) required is much less
than other more ductile/less brittle materials.

There are two primary types of breaks that occur on CI pipe, so called beam or circular breaks
and longitudinal splits or breaks. The type and morphology of each type of break is consistent
with the loading and brittle failure characteristics of the material. Fundamentally, cracks propa-
gate perpendicular to the applied stress.

3.2.1 Circular or beam breaks


For a circular break, the applied stress is vertical (beam loading) and the crack will propagate
circumferentially around the pipe creating a circular profile. The now infamous fire engine blo-
wout failure at 12212 Hartsook Street (Figure 2-2) was an example of a beam or circular break.
Beam deflection implies lack of vertical pipe support at the point of failure.

For steel pipe, the material properties are such that beam breaks are much less likely, since the
thin and flexible steel pipe wall tends to locally buckle (rather than cause a circumferential crack

Page | 23
3. Material Analysis and Results
to form and propagate). When steel pipe loses vertical support, the pipe defects from a circle to
an oval and then eventually locally buckles and may pull away from rivets or push on type
joints. In general, steel pipes do not suffer from circular beam breaks.

Vertical Beam
Loading

Figure 3-3. "Circular Break" (Circumferential Crack Propagation) due to beam loading at
12212 Hartsook Street (Figure 2-2)

Radial Crack
Growth

Figure 3-4. Failure morphology along edge of rapid circumferential crack propagation
due to beam loading from circular blowout at 12212 Hartsook Street.

Page | 24
3. Material Analysis and Results
3.2.2 Longitudinal splits or breaks
Longitudinal splits or breaks are the result of internal radial pressure resulting in longitudinal
crack growth for an initiating flaw location. The failure at Corbin Avenue and Kittridge Street is
an example of a longitudinal break. Internal pressure generates radial stresses, which result in
longitudinal crack growth.

Radial Pres-
sure/Stress

Figure 3-5. Axial View of cast ironlongitudinal break at Corbin Avenue and Kittridge
Street

Page | 25
3. Material Analysis and Results

Longitudinal
Crack Propa-
gation

Likely Point of Initia-


tion/critical Flaw

Figure 3-6. Longitudinal propagation of crack due to radial stress at Corbin Avenue and
Kittridge Street

Figure 3-7. Surface indication of underground break Corbin Avenue and Kittridge Street:
Note support has not been lost so the break is less likely to be beam or circular type
break.

Page | 26
3. Material Analysis and Results
The same morphology of longitudinal splits applies to steel pipe breaks. In general, thinning of
the steel usually due to external or internal corrosion creates an initial flaw. Internal pressure
provides radial stress that propagates along the longitudinal crack. However, steel is usually
not a brittle material so the crack propagates a short distance and then runs out of energy. The
process is repeated over time until stresses build up locally or, due to operations or surges, al-
low further propagation.

An example of a riveted steel pipe longitudinal break was the failure at Coldwater Canyon Ave-
nue and Dickens Street.

Figure 3-8. Repair to riveted steel pipe at Coldwater Canyon Avenue and Dickens Street.
Note repair steel plate on bottom runs back to riveted joint.

Page | 27
3. Material Analysis and Results

Longitudinal Split
Propagation

Figure 3-9. Riveted steel pipe section removed at Coldewater Canyon Avenue and Dick-
ens Street. Note longitudinal split and deformation of steel due to internal pressure
forces and possibly removal.

3.3 Brittle fracture, fracture toughness, flaws and damage accumula-


tion
Fracture toughness is a property of a material that describes the ability of a material containing
a crack or flaw to resist fracture. Brittle materials such as CI are said to have low fracture
toughness or flaw tolerance (Atkinson et al. 2002, Seica and Packer 2004, Ivanova, 1978).
Generally, at a given stress, there is a critical flaw size above which the material exhibits brittle
fracture and below which cracks do not grow.

Atkinson et al (2002) found that failure of CI pipes taken from historical underground service in
water distribution systems exhibited a bimodal distribution of flaw size. It was postulated that
manufacturing flaws, inherent in the casting and fabrication process, tended to be the lower
(smaller) flaw size population and time dependent flaws (corrosion, fatigue, etc.) accumulated
while in service were the larger in size.

Page | 28
3. Material Analysis and Results
Casting Flaws

Corrosion Flaws

Figure 3-10. Bimodal distribution of failure of historical service CI Pipe (Atkinson, et al.
2002)

The implication of this result is that over time, CI pipes which had previously been capable of
withstanding internal pressure and external loadings without failing in a brittle manner, will ac-
cumulate flaws due to corrosion and other time dependent processes and eventually will reach
the critical flaw size and fail in a brittle manner. Due to the accumulation of damage from corro-
sion and fatigue, the capacity of the material handle demand always decreases with time.

3.4 Time dependent damage accumulation mechanisms for cast iron


pipe: Corrosion and fatigue
There are two primary time dependent damage mechanisms that accumulate overtime on bu-
ried pipes: corrosion and fatigue. Corrosion is an environmental mechanism; fatigue is general-
ly related to operational factors.

3.4.1 Corrosion
Corrosion is the deterioration of a material as a result of a chemical reaction with its environ-
ment. Common examples of metallic corrosion are the rusting of iron, the tarnishing of silver,
the dissolution of metals in acid solutions, and the growth of patina on copper. In water distribu-
tion systems, corrosion of facilities occurs under a variety of conditions and for different rea-
sons. For example, pipelines can corrode due to galvanic corrosion, soil variations, stray cur-
rent-induced corrosion, and environmental contamination.

Corrosion is a natural process. Metals are generally most stable when in the form of their natu-
rally occurring ores. For example, iron ore, which is used to produce various steel alloys, is es-
sentially a form of iron oxide, magnetite (Fe3O4) and hematite (Fe2O3). Energy in the form of
heat is added to these stable ores to produce metal. This added energy is gradually dissipated
via a nonreversible electrochemical process (known as the corrosion reaction) as the metal al-

Page | 29
3. Material Analysis and Results
loys revert back to their naturally occurring state. In the specific case of iron-based metals, the
corrosion byproduct is the compound commonly referred to as rust.

In order for the electrochemical process or corrosion to occur, a complete electrical circuit is re-
quired made up of the following components: the anode; the cathode; an electrolyte; and a me-
tallic path.

Anode (or anodic area), the electrode of an electrochemical cell at which oxida-
tion occurs. At the anode, the metal atoms give up one or more electrons and be-
come metal ions. In chemical shorthand the general formula for this reaction is
written:

M0 -> M+n+ + ne-

M0 represents a metal atom such as iron or copper in a metallic structure. The ar-
row represents the direction in which the reaction is occurring. The symbol M+
represents a metal ion. Metal ions formed in the corrosion reaction leave the metal
structure (corrosion) and enter the environment (electrolyte), destroying the anode.
The symbol e- represents the negatively charged electron released by the forma-
tion of the metal ion. The free electron that is formed in the corrosion reaction re-
mains within the metal structure. Note that there is no change in total charge (0 =
+2 + -2). Electrons flow away from the anode via the external circuit.

Cathode (or cathodic area), the electrode of an electrochemical cell at which re-
duction is the principal reaction. At the cathode there are many possible reactions.
The simplest common cathodic reaction is the reaction of hydrogen ions, which are
present in water solutions, with electrons to form hydrogen gas. In chemical short-
hand this reaction is written:

2 H+ + 2e- -> H2

This represents the reaction of two hydrogen ions (2H+) with two electrons (e-) to
form two hydrogen atoms, which then combine to form one molecule of hydrogen
(H2) gas. As in the case of anodic reactions, there is no change in net charge in
this reaction (+2 + -2 = 0). Electrons flow toward the cathode in the external circuit.
An example of this current flow is shown in Figure 2-1.

Electrolyte, a chemical substance containing ions that migrate in an electric field.


This substance, such as the soil or water surrounding a pipeline, is capable of
transmitting electricity.

A metallic path electrically connecting the anode and cathode or anodic and ca-
thodic areas. For example, the steel in a pipeline serves as a conductive path link-
ing cathodic and anodic areas together.

Page | 30
3. Material Analysis and Results

CURRENT FLOW
ELECTRON FLOW
EXTERNAL METALLIC PATH

H 2 GAS

+
2H +2e H2

++
Fe +2OH Fe(OH)
2
+++
++ Fe +3OH Fe(OH)
Fe Fe + 2e 3
++ +++
Fe Fe + 2e
Fe(OH)2 and Fe(OH) 3 CATHODE
ANODE
CORROSION ( COPPER-Cu)
(IRON-Fe)

ELECTROLYTE
+
(H2O H +OH )

Figure 3-11. Example of a corrosion cell

These four parts constitute what is called the corrosion cell. The corrosion cell is driven by the
potential energy difference between the anode and cathode. The anode is at a higher energy
level, and corrodes as it releases energy to the cathode.

When corrosion of either cast or ductile iron pipe occurs, the graphite present remains as an
integral part of the corrosion by-products that adhere firmly to the unattacked metal substrate.
The mechanism is called selective leaching or dealloying and known specifically for CI and DI
as graphitization. La Que (Stroud 1989, LaQue 1958) states: The firmness of the attachment
and compactness or permeability of the graphitic residue are influenced by the strengthening
effect of other insoluble constituents of the iron such as carbides, silicious compounds, and
phosphide eutectic stringers or insoluble corrosion products that may be precipitated within the
graphitic corrosion product layer. The graphitic layer with its intermixed insoluble corrosion
products, etc., can become so impermeable to the further penetration of corrosive liquids that it
will form an excellent protective coating that can stifle further attack.

The by-products provide a barrier against further corrosion attack, which, if left undisturbed, can
slow or even stop the corrosion process in many soil environments. Although the graphitization
products are brittle and weaker than the original iron structure, they nevertheless have consi-
derable mechanical strength, which is evidenced by the long service life of many unprotected
iron pipes in corrosive environments (LaQue 1958). However, once the iron is removed, the
graphitized area represents a flaw or weak point in the wall of the pipe.

Page | 31
3. Material Analysis and Results

Figure 3-12. Graphitized layer on external surface of iron pipe sample

Soil Corrosivity and corrosion rates on ferrous materials including CI, DI and steel are fairly well
characterized by two soil chemical characteristics: soil resistivity and soil pH. The electrical re-
sistivity of a soil is a measure of its resistance to the flow of electrical current. Corrosion of bu-
ried metal is an electrochemical process in which the amount of metal loss due to corrosion is
directly proportional to the flow of electrical current (DC) from the metal into the soil. Corrosion
currents, following Ohm's Law, are inversely proportional to soil resistivity. Lower electrical re-
sistivities result from higher moisture and soluble salt contents and indicate corrosive soil. Insitu
measured soil resistivities in the Los Angeles basin vary from less than 100 ohm-cm to more
than 5,000 ohm cm, depending on location, geology and depth. Soil pH is a measure of the
acidity of the soil. Typical soil pHs are between 6 and 8 which are slightly acidic to alkaline.

Other factors such as ground water, presence of soil bacteria, stray electrical direct current and
construction details such as electrical isolation from unintentional galvanic cells influence but
are extensive rather than intensive properties of the alignment and not a characteristic of the
soil per se.

King (1977) developed a relatively simple monograph, which depicts the relationship between
these parameters and soil corrosion rates. Others have published tables of data (Romanoff
1957) and guidelines for design (AWWA 2005, Schiff 1993), but the relative impact of these two
soil parameters is described by Kings figure.

Page | 32
3. Material Analysis and Results

Figure 3-13. Relationship between soil resistivity, soil pH and corrosion rates on ferrous
metals (King 1977)

The LADWP has collected in situ soil resistivity data at intervals of less than 1000 feet along
most of their pipeline alignments. In addition, soil samples are collected and analyzed when
leak or blowouts occur. The LADWP data for CI leak rate versus soil resistivity shown below
confirms that lower resistivity soils result in higher leak rates for CI pipe (LADWP 2010a). The
effect is most pronounced below about 3,000 ohm-cm and increases exponentially as soil resis-
tivity decreases. The data are consistent with the premise stated above that as cast iron pipe
ages and corrodes, corrosion flaw size increases until finally the flaw tolerance capacity of the
pipe can no longer meet the applied stress for operation. A similar analysis for steel pipes
(LADWP 2010b) is not as dramatic and most likely reflects the difference in fracture and flaw

Page | 33
3. Material Analysis and Results
tolerance between CI and steel pipe materials. Steel pipe has greater flaw tolerance so creation
of flaws post installation does not dominate the failure mechanism or behavior.

Figure 3-14. LADWP trend for cast iron leak rate vs. soil resistivity range (LADWP 2010a).

Figure 3-15. LADWP trend for steel pipe leak rate vs. soil resistivity range (LADWP 2010b)

Page | 34
3. Material Analysis and Results
3.4.2 Fatigue
Fatigue is the progressive and localized accumulation of structural damage that occurs when a
material is subjected to cyclic loading. It can occur even when cyclic and total stresses are less
than the yield and ultimate stress limits for the material.

It has been postulated that pressure surges, spikes, and changes observed in the system are
contributing to the increase in leaks and blowouts in the system as a function of time. Fatigue
begins with dislocation movement, which eventually form slip bands that nucleate to form short
cracks. Particularly for brittle materials, fatigue damage over time results in increasing flaw siz-
es in a similar manner to corrosion. Like corrosion, fatigue damage does not recover even
when the stresses and cycling are removed or stop. The material is permanently degraded by
the mechanism.

From a predictive standpoint, it is difficult to use laboratory data and assumed cycle rates to
prognosticate failure of any particular piece of pipe. This is because:

1) Fatigue is a stochastic process and there is considerable scatter in the data even
under controlled environments.
2) Typical fatigue data assumes consistent changes in stress or strain. In reality,
stresses and strains are not constant in each cycle, further complicating the analysis.
We do know that higher stresses result in shorter fatigue life or cycles to failure.
3) Mohebbi et al. (2009) showed that there is substantial variation in measured fatigue
behavior between different CI pipes.

We can use some deductive reasoning along with the leak and blow out data to investigate the
effect that fatigue may have on CI pipe behavior. LADWP (2010a) data for CI leaks by diameter
are shown below.

Figure 3-16. LADWP data for leak rates by diameter of CI (LADWP 2010a)

The data indicate that leak rate increases as diameter decreases below 8 inches in diameter.
Wall thicknesses for CI pipe vary by approximately 20 % between 8 inch (0.47 inches and 2 inch
(0.38 inches) within the same pressure class. Since hoop stress in the pipe decreases with the
pipe diameterhoop stresses in the smaller diameter pipes are less 25% and 50% that of the 8
inch pipeoverall, stress is lower in the smaller diameter CI pipes, yet leak rates are higher. If

Page | 35
3. Material Analysis and Results
fatigue were the only mechanism of long-term degradation, leak rates should increase with di-
ameter.

Based on this simplistic analysis, fatigue does not appear to be the main contributing factor to
CI leaks or blowouts. However, the analysis assumes all other things are equal and does not
take into account the possible and likely synergistic effect of corrosion damage locally increas-
ing stresses and that stresses are not purely radial.

Page | 36
4 GEOTECHNICAL ANALYSIS OF
BLOWOUTS

The present study on water pipeline breaks resulted from the rash of catastrophic failures of
street pavements caused by water pipeline breaks. These dramatic and highly visible pavement
failures received unusual attention from the media after a fire engine fell into a large hole that
occurred in the asphalt due to the failure of a underground water main (Figure 2-1). By defini-
tion, these catastrophic failures, which LADWP refers to as blowouts, are water main breaks
damaged by water pressure and/or subgrade erosion and requires the repair of at least 100
square feet of asphalt or concrete pavements. This definition is based on practical considera-
tions for repairing the street pavements, and not on specific engineering or scientific criteria.

In contrast to water main breaks, the catastrophic failures of the ground and asphalt pavement
resulting from water main breaks remained less investigated. The scarcity of studies suggests it
is more efficient to avoid the causes of these failures, i.e., regulating pressure and avoid tran-
sient pressure surges (e.g., Flemming et al., 2006; Wood et al., 2005), rather than analyzing
their immediate manifestations and consequences. There may also be complexities arising from
the multitude of geometries of pipe trenches and street pavements, and nonlinearities leading to
subsurface soil erosion and void formation.

In view of the absence of past models for explaining blowouts, a simplified model was devel-
oped to identify the conditions required for blowouts to occur. Figure 4-1 represents the two-
dimensional geometry of the simplified model. The model geometry is inspired from typical con-
struction geometries. The pipe of diameter is laid within a trench that has a width T and depth
H. The soil trench is typically about 1-2 feet wider than the pipe diameter for facilitating pipe in-
stallation. The pipe is embedded in soil bedding, which is usually made of sandy materials, and
in some instances native soils. The top surface layer (AC), which is usually made of asphalt
concrete has a thickness Ht, and is underlain by a layer of Coarse Aggregate Base (CAB) and
Crushed Miscellaneous Base (CMB). The asphalt layer is usually impervious to prevent water
from infiltrating and damaging road subgrades. In contrast, the CAB/CMB layer is very porous
and diffuses the water laterally under the impervious layer of surface pavement.

As shown in Figure 4-1, the simplified model accounts for the strength of pipe leaks (Point P).
The amount q of water released per unit of time and unit of pipe length through a leak crack or

37
4. Geotechnical Analysis of Blowouts
hole is assumed to be proportional to the pressure difference across the pipe wall of thickness t,
i.e.:

Q = K pw p P
(1)
Where pw is the water pressure inside the pipe; pP is the water pressure right outside the pipe at
the soil-pipe interface; and K is a leak coefficient characterizing the size of the crack or hole in
the leaking pipe. K has the dimension of L3T-1F-1/2 where L, F and T denote the dimension of
length, force and time, respectively. K is often expressed in term of hole area A and shape fac-
tor Cd as follows:

K Cd A
(2)
The values of K can be selected based on various references.

As shown in Figure 4-1, the simplified model for blowouts also accounts for drainage conditions
that influence the distribution of water pressure in the soil, the hydraulic forces, and the amount
of soil erosion. The pipe bedding is assumed to be saturated, which is a reasonable assumption
as small leaks gradually saturate bedding and infiltrate pavement subgrades before they culmi-
nate resulting in pipe breaks. The bedding has for unit weight sat, and permeability kB. The wa-
ter from the pipe is assumed to drain first upward to the top surface and then laterally under the
impervious asphalt pavement. The water pressure is assumed to become zero at some points O
and O in the pavement subgrade. Point O and O are positioned using the horizontal drainage
length L.

Figure 4-1. Typical profile of a trench, bedding and surface pavement for water pipes for
analysis of blowouts.

Page | 38
4. Geotechnical Analysis of Blowouts
Without going deeply into the mathematical derivations and numerical applications for various
leak sizes and drainage conditions, the simplified model captures the three basic types of water
leaks/breaks. In case of small water leaks: (a) the hydraulic gradient may be too weak to erode
the soil, and the water leak may be invisible from the surface when the water drains to hidden
locations. In the case of very large pipe breaks: (b) the sudden release of high water pressure is
likely to liquefy the bedding and heave most pavements, resulting in a geyser as the form of
failure. In the case of medium-size water leaks: (c) the hydraulic gradient may be large enough
to erode the soil creating a cavity under the street pavement.

The simplified model for blowouts (although it is still in progress) concludes that blowouts of
types (b) or (c) may only occur in the case of sufficiently large water pipe breaks as those
shown in Figures 2-1 to 2-7. In other words, only the largest water leaks are likely to create blo-
wouts. This suggests that large-diameter pipes and pipes that exhibit brittle failure (e.g., cast
iron) are more likely to generate blowouts.

Page | 39
4. Geotechnical Analysis of Blowouts

(a)

(b) (c)

Figure 4-2. Illustration of various types of pavement failures caused by water pipe
leaks/breaks: (a) a small water leak from pipe collected in a drain under the pavement
surface or emerging onto the street pavement; (b) a large pipe breaks releasing large
quantity of water under pressure, and breaking up the street pavement; and (c) medium-
size water leak washing away the soil bedding, gradually creating a cavity under the
street pavement due to soil erosion, and resulting in collapse of surface pavement when
vehicles pass over it.

Page | 40
5 STATISTICAL ANALYSIS AND RE-
SULTS

This section, which is the core of the present study, analyzes the vast amount of data that
LADWP collects on a regular basis from its water distribution system. The section examines
how LADWP collects data, and extracts from the collected data the main characteristics of water
pipeline breaks. It also compares LADWP data with those of a few other municipal systems of
comparable sizes in an effort to understand how LADWP systems perform compared to those of
other municipalities. The section analyzes not only the time series of collected data, but also
their geo-spatial distribution. It relates data on water pipeline breaks to other relevant factors
such as water pressure, water temperature, and soil corrosivity. Finally, the section attempts to
address the concerns raised about the effects of water rationing on water pipeline breaks, and
proposes a simplified model to explain the unusual water pipeline breaks during the summer of
2009.

5.1 Data collection practices at LADWP


Figure 5-1 illustrates the flow and collection of information on LADWP pipeline breaks. After a
leak is reported to the Water Trouble System, a technician is usually dispatched to assess the
situation before a crew is sent to repair the leak. As the crew repairs the leaking pipeline, it iden-
tifies the pipelines location using various maps, websites, and reports as shown in Figure 5-1.
As shown in Figure 5-2, the leak repair is reported into two systems Construction Productivity
System (CPS) and LADWP Geographic Information System (GIS). Mainly used for accounting
purposes, CPS contains the most complete record of leaks, but it lacks detailed information
about individual leaks, such as the type of breaks. On the other hand, the LADWP GIS contains
a comprehensive GIS model about LADWP pipelines and distribution systems. Although it con-
tains the most detailed information on leaks, the GIS model is incomplete; it is missing many
CPS leak records. Only 60% of the leaks reported in CPS were entered in the GIS model be-
tween 2001-2009 due to backlog and insufficient CPS data entry. This study analyzes both CPS
and GIS data. Depending on the data types, the analyses use either one data set or a combina-
tion of the data sets to produce the most accurate results the data sources are identified in the
figure captions.

41
5. Statistical Analysis and Results
As shown in Figure 5-2, the LADWP GIS model contains only about 60% of the water pipeline
repairs recorded in CPS. In other words, approximately 40% of the data on water pipeline re-
pairs are ignored by the LADWP GIS model and is, therefore, not available to help in under-
standing how to improve the performance of LADWP systems and to better manage LADWP
assets. However, LADWP is to be commended for establishing a new program for thoroughly
documenting repair characteristics, such as the type, length and position of pipeline failures. In
order to minimize data loss for the LADWP GIS master model, it is recommended that these
additional details be acquired using modern data collection techniques such as smart phones or
PDAs with GPS positioning data and geo-tagged photos.

Figure 5-1. Information flow for collecting pipeline break data

Page | 42
5. Statistical Analysis and Results

Figure 5-2. Information flow for collecting pipeline repair data

5.1.1 Data sets used in analyses


Table 5-1 summarizes the characteristics of the six data files used in this report. Hereafter, the
data files are referred to as Files A, B, C, D, E and F. Files A, B, C, and D contain data for leaks
since 2001, while Files E and F contain data about the LADWP pipeline system. Files A and B
originate from the CPS system and reflect the largest number of leaks and blowouts. Files C
and D originate from the GIS system and include more data on pipe properties and leak charac-
teristics, but they are missing at least 40% of leaks and blowouts that are included in Files A
and B. Files E and F have very little data on leaks as they describe mainly the LADWP pipeline
system in a GIS format. Since the multiple data files of different origins and versions often with
overlapping and inconsistent information, made it difficult to select the right data set for analys-
es, File B was created to remedy some of the inconsistencies between data files. File B geo-
referenced all leaks and blowouts contained in the CPS system between July 2001 and October
2009. It matched leaks and blowouts to pipe properties of the LADWP GIS model using the
closest distance between pipe centers and leak locations (see Figure 2-12 and Section 5.5.2 for
details).

Page | 43
5. Statistical Analysis and Results
Table 5-1. Data files collected

Leakdata Pipedata
FileA FileB FileC FileD FileE FileF
GregAllLeaks GregAllLeaks LeaksJuly2001 WGS_Leaks_C PIPESYSTEM Leak_Study
20091014.xls georefe toCurrentwith PS_Blowout_ DATA _2009.map
Filename
renced.xls PipeInfo.xls Match_Queri BASE.mdb
es12345.xls
12/22/2009
Datereceived 11/6/2009 10/24/2009 11/22/2009 11/9/2009 11/9/2009
createdbyUSC
Database CPS CPS GIS GIS GIS GIS
#ofleaks 12,359 12,359 5,120 7142 7326 NA
#ofblowouts 1,618 1,618 88 792 NA NA
Startingdate 7/1/2001 7/1/2001 7/1/2001 7/1/2001 1900 1900
Endingdate 10/13/2009 10/13/2009 12/31/2008 10/22/2009 2007 2007
5months 5months
Dategap none none (11/2003 (11/2003 none NA
3/2004) 3/2004)
EXCELcolumns 8 13 34 35 36 NA
Georeference no yes yes no no yes
PipeMaterials organized* organized* unorganized* organized* organized* organized*
Pipesize no no yes yes yes yes
Soiltype no no yes yes yes yes
LeakID no no yes yes no yes
Repair/manhours no no yes yes no yes
Yearinstalled no no yes yes yes yes
Pipematerialson
organized* organized* unorganized* organized* NA NA
leaks
Pipematerialson
missing* missing* organized* organized* NA NA
blowouts
*organized: categorized material type inputs;
*unorganized: uncategorized material type inputs (e.g., different names for same material types,
misspells, conflicting inputs, unknown material types)

5.2 Statistical analysis of data collected


Figure 5-3 breaks down the LADWP pipeline system by material and age. By definition, a pipe
age refers to the number of years since the pipe was installed. Using the recorded leak data
over the lifespan of pipes, Figure 5-4 shows annual leaks per 100 miles of pipeline for different
types of pipe material. This representation standardizes pipeline leaks and allows analysts to
compare the performance of various pressure zones and municipal systems independently from
the pipe length. When leaks are expressed in terms of pipeline unit length ( per 100 miles) , the
results are as follows: steel pipes leak the most, cast iron pipes are a close second, and ductile
iron pipes leak the least.

Page | 44
5. Statistical Analysis and Results
When the age of the pipes is also factored into the assessment, the results are significantly dif-
ferent. Figure 5-4 also shows leaks per length and ages. As one might expect, older pipes have
typically been repaired several times over their lifespan, and the number of repairs can affect
pipe performance. Therefore, an analysis that includes both leaks per length and the age of the
pipes lead to a fair is a more accurate assessment of comparison of pipe performance. As
shown in Figure 5-4 (right), over their lifespan, ductile iron pipes leak more than pipes made up
of asbestos cement and other materials. Cast iron pipes are no longer a close second to steel
pipes; instead, steel pipes leak significantly leak more than the other materials.

5000 80

Pipe avg age (years)


4000
Pipe length (miles)

60
3000
40
2000

20
1000

0 0
AC CI DI STL others AC CI DI STL others

Figure 5-3. Pipe material compositions (left); average ages by pipe materials (right); data
source: File E.

150 3
Leaks per 100 miles / avg age
Leaks per 100 miles of pipes

2.5

100 2

1.5

50 1

0.5

0 0
AC CI DI STL others AC CI DI STL others

Figure 5-4. Leaks per 100 miles of pipeline (left); leaks per 100 miles of pipeline and aver-
age age (right); data source: File E.

Pressure can also influence the number of pipe leaks. The LADWP system has 114 pressure
zones; Figure 5-5 compares the number of pipeline leaks in those pressure zones with more
than 30 miles of pipelines (Figure 2-13). As shown in the top Figure 5-5, Zone 386 (with 1000+
miles) contains the longest length of pipelines; Zone 1134 is the second largest with about 900
miles. As shown in the middle Figure 5-5, Zone 1096 leaks the most when the CPS leak data is
normalized by the lengths of pipelines in each pressure zone. The bottom Figure 5-5 shows a
similar trend on leak data collected from July 2001 to October 2009. The comparison of pres-
sure zone performances using the number of leaks per 100 miles of pipeline is useful to identify
potentially troublesome zones.

Page | 45
5. Statistical Analysis and Results
Length data: File E
1500
Pipe length (miles)

1000

500

0
0 205 240 275 295 325 386 400 426 448 462 477 517 579 610 619 737 757 769 778 830 847 944 947 950100010751096112311341216124013051424144514491597163617201750

Leak data: File E


500
Leaks per 100 miles

400

300

200

100

0
0 205 240 275 295 325 386 400 426 448 462 477 517 579 610 619 737 757 769 778 830 847 944 947 950100010751096112311341216124013051424144514491597163617201750

Leak data: File D


600
Leaks per 100 miles

400

200

0
0 205 240 275 295 325 386 400 426 448 462 477 517 579 610 619 737 757 769 778 830 847 944 947 950100010751096112311341216124013051424144514491597163617201750
Pressure zones (with pipes > 30 miles)

Figure 5-5. Leaks per 100 miles of pipeline in different pressure zones. Only pressure
zones with more than 30 miles of pipelines are included. Zone 0 is a virtual pressure zone
for pipes not assigned to a specific zones.

The analysis in this section did not include blowouts because the pipe data set (File E) lacks
blowout information.

5.3 Time series analysis


Extracted from an LADWP presentation on October 2009, Figure 5-6 underscores that leaks in
the cast iron pipes steadily and dramatically increased from 54% to 77% in the summers of
2007, 2008 and 2009. This result can be simply interpreted as the number of cast iron pipes
that fail each summer is increasing, cast iron pipes are failing more and more and they are be-
coming responsible the primary source for an increase in leaks in the summer of 2009. Howev-
er, this interpretation is based on using different data sources for 2009 than for 2007 and 2008.
As illustrated by the differences between Figure 5-7 and Figure 5-6, when the analysis is based
on a consistent data source (File D), it does not reveal a clear leak increase in cast iron pipes in
Summer 2009. This difference in results illustrates the challenges in performing statistical anal-
ysis on incomplete and complex data sets.

Page | 46
5. Statistical Analysis and Results

Figure 5-6. LADWP analysis (note: a different data source on Summer 2009)
Leaks Summer 2009 (n=160) Leaks Summer 2008 (n=129) Leaks Summer 2007 (n=170)
9% 9% 9%
2%
4% 5%
others
AC
STL
DI 22%
20% CI
26% 54%

64% 2% 67%
3%

6%

Figure 5-7. Independent analysis from this report (note: same data source for all sum-
mers: File D)

Figure 5-8 breaks down the number of leaks from July 2001 to October 2009 in terms of pipe
material types. The lower Figure 5-8 normalizes (on a 0-to-100% scale) both the breakdown of
leak numbers and the LADWP pipeline system (in terms of length) and materials type. This re-
presentation displays the portions of leaks that occur in each material type. The leftmost histo-
gram bar shows the distribution of pipe length in term of pipe material type. If all material types
leak uniformly, all the leak distributions will coincide with the leftmost histogram. If a material
type leaks more than others, then its area in the leak distribution becomes larger than its area in
the material distribution. For example, in 2001, less than 1% of leaks were in ductile iron pipes,
which accounted for about 10% of all pipes per length; this implies that ductile iron pipes leak
less than pipes made of other materials. Comparing the normalized breakdown of material types
and leaks is useful to determine which pipe material leaks the most. Cast iron pipes make up
65% of LADWP pipes and account for around 65% of all leaks, with some variations yearly. In
other words, cast iron pipes do not leak more (or less) than pipes made up of other material
types, as suggested in Figure 5-6. Compared to other material types, steel pipes leak the most,

Page | 47
5. Statistical Analysis and Results
asbestos cement pipes leak slightly, and ductile iron pipes leak significantly less than any other
type. Ductile iron is the material used in the newest portions of the LADWP system.

Using the same representation as Figure 5-8, Figure 5-9 shows the breakdown of total and
normalized number of blowouts in terms of material types between July 2001 and October 2009.
Figure 5-9 strikingly shows that blowouts mostly happen in cast iron pipes.
1200
AC
1000 CI
Number of leaks

DI
800 STL
others
600

400

200

0
2001 2002 2003 2004 2005 2006 2007 2008 2009
Year

100
Percentage in various

80
pipe materials (%)

60

40

20

0
pipe length 2001 leaks 2002 2003 2004 2005 2006 2007 2008 2009 2001-09 lifespan leaks

Figure 5-8. Distribution of leaks (July 2001- October 2009) in term of pipe material types
using (top) total number of leaks, (leftmost bottom) percentage of pipe length in various
pipe materials, and (bottom) percentage of leaks in various pipe materials; data source:
File D.

Page | 48
5. Statistical Analysis and Results
200

Number of blowouts AC
CI
150 DI
STL
others
100

50

0
2001 2002 2003 2004 2005 2006 2007 2008 2009
Year

100
Percentage in various

80
pipe materials (%)

60

40

20

0
pipe length 2001 leaks 2002 2003 2004 2005 2006 2007 2008 2009 2001-09

Figure 5-9. Distribution of blowouts (July 2001- October 2009) in terms of pipe material
type using (top) total number of blowouts, (leftmost bottom) percentage of pipe length in
various pipe materials, and (bottom) percentage of blowouts in various pipe materials;
data source: File D. Note: only 51% of all blowout data is used, and 2001 and 2009 are re-
sults based on partial data sets.

Figure 5-10 and Figure 5-11 break down leaks and blowouts between July 2001 and October
2009 in relation as a function of pipe diameters. As shown in the lower Figure 5-10 and Figure
5-11, the leftmost histogram bar represents the distribution of pipe diameters in the LADWP sys-
tem. If leaks occur independently of pipe diameters, the normalized leak distributions will coin-
cide with the leftmost histogram bar. As shown in Figure 5-10, pipes with smaller diameters
(less than or equal to 4inches) leak more than other pipes, and large pipes (10inches and larg-
er) leak much less compared to other pipes. Figure 5-11 shows that 8inch pipes blowout more
per unit length than pipes of other diameters. Small pipes (smaller than 8-inch) blowout at a rate
comparable to their material distributions. Large pipes (10-inch and larger) blowout and leak
significantly less.

One should note that 8inch pipes are mostly made up of cast iron (Figure 2-16). Figure 5-9 and
Figure 5-11 show that 8inch cast iron pipes blowout the most per unit length compared to other
pipes.

Page | 49
5. Statistical Analysis and Results
1200
<4" pipe diameter
Number of leaks 1000 6"
8"
800 10-12"
>16"
600

400

200

0
2001 2002 2003 2004 2005 2006 2007 2008 2009
Year

100
Percentage in various
pipe diameters (%)

80

60

40

20

0
pipe length 2001 leaks 2002 2003 2004 2005 2006 2007 2008 2009 2001-09 lifespan leaks

Figure 5-10. Distribution of annual leaks (July 2001- October 2009) in terms of pipe di-
ameter using (top) total number of leaks, (leftmost bottom) percentage of pipe length in
various pipe diameters, and (bottom) percentage of annual leaks in various pipe diame-
ters; data source: File D. Note: 2001 and 2009 results are based on partial data sets.
160

140 <4" pipe diameter


Number of blowouts

6"
120 8"
10-12"
100
>16"
80

60

40

20

0
2001 2002 2003 2004 2005 2006 2007 2008 2009
Year

100
Percentage in various
pipe diameters (%)

80

60

40

20

0
pipe length 2001 blowouts 2002 2003 2004 2005 2006 2007 2008 2009 2001-09

Figure 5-11. Distribution of annual blowouts (July 2001- October 2009) in terms of pipe
diameters using (top) total number of annual blowouts, (leftmost bottom) percentage of
pipe length in various pipe diameters, and (bottom) percentage of annual blowouts in

Page | 50
5. Statistical Analysis and Results
various pipe diameters; data source: File D. Note: only 51% of all blowout data is used
and2001 and 2009 results are based on partial data sets.
100
AC
CI
80
Number of leaks

DI
STL
60 others

40

20

0
2001 2002 2003 2004 2005 2006 2007 2008 2009
Year

100
Percentage in various

80
pipe materials (%)

60

40

20

0
pipe length 2001 leaks 2002 2003 2004 2005 2006 2007 2008 2009 2001-09

Figure 5-12. Zone 579: Distribution of annual leaks (July 2001- October 2009) in terms of
pipe material types using (top) total number of leaks, (leftmost bottom) percentage of
pipe length in various pipe materials, and (bottom) percentage of leaks in various pipe
materials; data source: File D. Note: 2001 results are based on a partial data set, and
there are no leak data in 2009 for Zone 579.

Page | 51
5. Statistical Analysis and Results
25
AC

Number of blowouts
20 CI
DI
STL
15 others

10

0
2001 2002 2003 2004 2005 2006 2007 2008 2009
Year

100
Percentage in various

80
pipe materials (%)

60

40

20

0
pipe length 2001 leaks 2002 2003 2004 2005 2006 2007 2008 2009 2001-09

Figure 5-13. Zone 579: Distribution of annual blowouts (July 2001- October 2009) in terms
of pipe material type using (top) total number of blowouts, (leftmost bottom) percentage
of pipe length in various pipe materials, and (bottom) percentage of blowouts in various
pipe materials; data source: File D. Note: only 51% of all blowout data are used; 2001 is a
partial data set and there is no blowout data in 2009 for Zone 579.

Figure 5-12 and Figure 5-13 are similar to Figure 5-9 and Figure 5-10. They break down the
numbers of leaks from July 2001 to October 2009 in terms of pipe material, but only for Pres-
sure Zone 579 instead of the complete LADWP system. Zone 579 was selected because it har-
bored a relatively high numbers of leaks and blowouts (shown in Figure 5-5). Figure 5-12 shows
that, except for 2003, there are cast iron pipes leak more in Zone 579 (except for 2003) com-
pared to the entire LADWP system. As shown in Figure 5-13, most blowouts occur in cast iron
pipes a result that is similar to Figure 5-13. In Pressure Zone 579, cast iron pipes leak more
than any pipes in any other pressure zones.

5.4 Comparison with other cities


Water pipeline breaks are by no means exclusive to Los Angeles; they are a problem occurring
with various frequencies in all cities throughout the world. These breaks are largely the result of
aging water distribution infrastructures, as reported by the American Society of Civil Engineers
(ASCE) 2009; Boulos et al., 2006; Committee on Public Water Supply Distribution Systems,
2006; Kinneyer et al., 1994; and Runge and Mann, 2005.

Various aspects of water main breaks have been extensively studied (e.g., Andreou et al., 1987;
Andreou, 1986; Arnold et al., 1960; AwwaRF, 1997; Cullinane et al., 1989; Dolson, 1955; Hosoi

Page | 52
5. Sta
atistical Analysis
A s and Re
esults
et al. 198
89; Kettler and
a Goulter, 1985; Klein ner and Rajaani, 2002; Makar, 2000; Morris Jr., 1967;
1
Pelletier et al., 2003 a 1986; Wolfe, 1946; Kleiner and Balvantl 20
3; Rajani et al, 001). Many water
w
main breeaks are relaated to corro
osion (e.g., De
D Rose and d Parkinson,, 1985; Fitzg
gerald, 1968; Ma-
kar et al., 2005) andd have been n observed in i cast iron mains (e.g., Rajani et al.,
a 2000; Ta albot,
1926).

5.4.1 AWWA
A comp parative stu udy
The Ame erican Wateer Works Association (AWWA,( 2007) estimattes the total value of water w
supply networks in the
t US at $4 4.8 billion peer million pe eople. The nation's
n wate er supply syystem
serves soome 300 million Americcans, using 408 4 billion gallons of wa ater each da ay. The Ame erican
drinking water infras
structure nettwork spanss over 700,0 000 miles, more
m than four
f times lo
onger
than the National Highway
H Sysstem. There e are 240,00 00 water piipeline brea aks occur accross
America each year. AWWA
A estimmates that water
w utilitiess will have to
o invest $2550 billion oveer the
next 30 years
y to rep
place aging pipelines.
p Thhe cost of pipelines
p for new develo opments, seccurity
upgradess, advanced d treatment methods,
m annd other nee eds may incrrease this biill to $500 billion.
In 2008, the US Env vironmental Protection Agency
A (EPA) and othe ers estimated that waterr and
wastewater infrastruccture repair costs may be b as much as a $745 billiion to $1 trilllion over the
e next
20 yearss. USGS rep ports 6 billio
on gallons ofo withdrawa als per yearr are lossess or public use-
enough to
t supply thee 10 largest US cities in each year.

AWWA (2007) comp pares the perrformance of o the water distribution


d s
system in sixx major utilities in
the US. Figure
F 5-14 and Figure 5-15 break down per material
m type the total pip
pe length and the
total num
mber of breaaks for thesee six cities, which
w are labeled A to F.
F Utility C re efers to LAD DWP.
Except foor system B, most waterr distribution n systems co ontain a larg
ge number of o cast iron pipes.
p
Figure 5--15 shows thhat cast iron
n pipes do no ot break more than othe er types of pipes.
p Howevver, it
also shows that a la arge numbe er of pipelinee breaks are e not assocciated to anyy material tyypes,
which suuggests that pipeline breeaks are sim mply poorly documented
d d in many citties. In this pers-
pective, LADWP
L is collecting bettter data on pipeline breaks than the e other five utilities.
u

100%
PipeLength/TotalLength
g

80%
Others
60% Steel
g /

DI
40%
AC
20% CIP
p

0%
A B C D E F

Figure 5-14.
5 Breakd
down of tota
al pipe leng
gth per pipe in terms off material ty
ype in six mu-
m
nicipal systems
s lab
beled AF (A
AWWA, 200 07) ). System
m C is LADWWP.

Page | 53
3
5. Sta
atistical Analysis
A s and Re
esults

NumberofBreaks/TotalLength
100%

80% Otheers
60% Steel
DI
40%
AC
20%
CIP
0%
A B C D E F

Figure 5-15.
5 Breakd down of watter pipeline breaks perr total length in terms of
o material type
in six muunicipal sys
stems labelled AF (AWWWA, 2007)). System C is LADWP..

5.4.2 Present
P com
mparative sttudy
In additio
on to AWWA w LADWP and other utilities,
A and in colllaboration with u the authors
a colle
ected
data fromm the water distribution systems of sixs major mu hereafter referred as cities
unicipalities
U, V, W, X, Y, and Zand
Z to co
ompared the e performancce of each citys
c water pipelines
p to those
t
in Los An ngeles (LA).. Figure 5-1
16 breaks do own the pipe e length in terms
t of material type fo
or LA
and citiess U, V, W, and
a X. LADW WP is the one e of the largest systemss, with more than 7,200 miles
of pipelin
nes. Figure 5-17
5 breaks down the to otal length of pipelines inn terms of pipe
p materiall type
and pipe diameter fo or LA, and citties U, V, an
nd X.

7000

AC
6000 CI
DI
5000
STL
Pipe length (miles)

others

4000

3000

2000

1000

0
L (7238 miles)
LA City U (4039) City V (4286) City W (1656)) City X (1827
7) City Y (512
20) City Z (80
072)

Figure 5-16.
5 Breakd down of tota
al pipe leng
gth in terms of by pipe material for seven munic-
ipal wate
er utilities; LA data sou
urce: File E.
E

Page | 54
4
5. Statistical Analysis and Results
LA Utility U

2500 800

Pipe length (miles)


2000 600
Pipe length (miles)

1500 400

1000 200

500 0

0 <4
<4 6
6 8 CI
8
CI 10-12 STL
10-12
STL Pipe diameter (in) DI
Pipe diameter (in) 16 DI 16
AC AC
>20 others >20 others

Utility V Utility X
1000
800
800
Pipe length (miles)
Pipe length (miles)

600
600

400 400

200 200

0
0
<4
<5
6
5-15 8
15-25 CI CI
STL 10-12
STL
25-35 DI Pipe diameter (in) 16 DI
Pipe diameter (in) AC AC
>35 >20 others
others

Figure 5-17. Two-dimensional breakdowns of total pipe length in terms of pipe material
type and pipe diameter for Los Angeles, and cities U, V, and X. LA data source: File E.

Page | 55
5. Statistical Analysis and Results
45
LA
City U
40 City V
Number of leaks per 100 miles of pipes

City W
City X
35
City Y
City Z
30 national average limits

25

20

15

10

0
1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009

Year
Figure 5-18. Breaks per 100 miles of pipeline for LA, and for cities W, X, Y, Z, plus boun-
daries for national averages; LA leak data sources: File A; LA pipe data source: File E.

Figure 5-18 shows the annual number of breaks per 100 miles of pipeline for seven municipal
utilities including LA. The seeming decreases in pipeline breaks for 2001 and 2009 for the
LADWP system are apparent; the data in 2001 and 2009 were simply incomplete. The LADWP
system performs near the national average in terms of pipeline breaks per 100 miles of pipeline.
As one of the largest water distribution systems, LADWP performs similarly to most other mu-
nicipal systems. There are, however, a few municipal systems, such as Seattle (or City X in Fig-
ure 5-18), that over the years perform systematically better than LADWP. Seattle system pipe-
line performance is summarized in Appendix 11.3.

5.5 Spatial analysis


5.5.1 Spatial Distribution of Pipeline Breaks
The spatial distributions of pipeline breaks across the LADWP system were examined via con-
tour maps created using the LADWP-provided Manifold GIS file (File F),the GIS breaks file (File
C), and Google Earth. Two of the contour maps indicated the number of annual breaks per 100
miles of pipeline. This is a common way of representing pipeline breaks. Figure 5-19 (left)
shows contours based on all pipelines in the LADWP system and all breaks included in the File
C for the period of 2002-2007. As mentioned previously for File C, there is a 6-month gap in the

Page | 56
5. Statistical Analysis and Results
records between late 2003 and early 2004. To accommodate this, the average number of an-
nual breaks was calculated using 5.5 years. The contour interval was set at the overall average
break rate. Where no contours are shown, the break rate in that area matches the overall aver-
age. This allows one to see where the break rate is above average.

Figure 5-19. Contours showing annual breaks per 100 miles of pipeline in all pipes (left)
and cast iron pipes (right).

As shown in Figure 5-19(left), there is a higher break rate in the area surrounding the Santa
Monica Mountains. Break rates appear to be higher in areas where pumped pressure zones ad-
join gravity pressure zones. The Santa Monica Mountains are largely pumped pressure zones,
as are areas in the San Gabriel Valley foothills, Baldwin Hills, and the Harbor District.

Figure 5-19 (right) shows a similar map of breaks based on cast iron breaks and cast iron pipes
alone. As stated previously, cast iron pipes make up 65% of the system and cast iron pipes ac-
count for a similar percentage of the breaks over the same 2002-2007 time period. As shown,
the contours highlight a somewhat similar concentration of the breaks at the base of the Santa
Monica Mountains.

A review of pipeline breaks versus pressure zones reveals that many of the zones with the
highest break rates are located in the Santa Monica Mountains and in zones with or near pump
stations. Pressure zones 1096, 1216, and 1424 have extremely high break rates. These three
zones are located in the Santa Monica Mountains north of Beverly Hills and Hollywood. Each of

Page | 57
5. Statistical Analysis and Results
these zones is a pumped pressure zone and each contains both storage tanks and pump sta-
tions.

5.5.2 Geo-statistics of pipeline break locations


Figure 5-19 shows the distribution of pipeline breaks in the LADWP system based on calcula-
tions that use a uniform grid. This approach, although very informative, extrapolates contour
maps beyond the irregular boundaries of the LADWP system (Figure 2-13).

Therefore, another spatial analysis method was applied to account for the irregular spatial cov-
erage of the LADWP system. The spatial variables were calculated using scattered points in-
stead of uniformly spaced grid points. The scattered points are located at pipeline breaks. The
average spatial variables were calculated within a square mile centered at pipeline breaks. Spa-
tial gaps, such as other municipalities, in the LADWP system, were automatically excluded as
there are no data on breaks in these areas for which LADWP is responsible. This method pro-
duces gaps for LADWP regions without pipeline breaks, which must differentiated from non-
LADWP area coverage. Figure 5-20 illustrates this geostatistical method geographically. The
analysis used a 1 mile 1 mile square area, but it yielded similar results for both a finer (0.5
mile 0.5 mile) and a coarser (2 mile 2 mile) scale. Following Figure 5-20 is a discussion of
the weighted average determination along with many different other LADWP system statistics,
defined in regardless if part of the pipe may be outside the area.

Table 5-2.

(1) Draw a square Area A (2) Identify and count


centered at a leak leaks or blowouts inside
Area A
1 mile

(3) Identify pipes in-


(4) Calculate weighted aver-
side Area A and cal- 1 mile age of pipe properties using
culate individual and
individual pipe length inside
total length of pipes
Area A

Figure 5-20. Definition of geo-statistical variables used for pipeline leaks, blowouts, and
properties: (1) a square area A is drawn centered at a leak point; (2) number of leaks (or
blowouts) is added up within Area A; (3) the length of each pipe is recorded and summed

Page | 58
5. Statistical Analysis and Results
up with Area A; and (4) the average properties of pipes are calculated based on a
weighted average that uses the length of individual pipe as weight factors.

The weighted averages of pipe properties, Pavg, are calculated using


P
i = 1, 2, , n

where Pi and Li is the property and length, respectively, of Pipe i inside a square area; n is the
number of pipes inside the square area; a pipe is considered inside this area if its center is lo-
cated within this area regardless if part of the pipe may be outside the area.

Table 5-2. Explanations of statistical variables.

Figure Variables Statistics Units Explanations


Figure Dl Number of leaks 1/mile2 Dl represents the leak density,
5-21 i.e., the amount of leaks inside
a square mile area (Figure
5-20) Figure 5-20
Figure Db Number of blowouts 1/mile2 Dl represents the blowout den-
5-22 sity, i.e., the amount of blowout
inside a square mile area
Figure Rbl Ratio of blowouts / No unit Rbl represents the ratio of blo-
5-23 leaks wout density to leak density,
i.e., Rbl = Db / Dl
Figure Dpl Pipe length 1/mile Dpl represents the pipe density,
5-24 i.e., the total length of pipe in-
side a square mile area
Figure Nl100 Annual leaks per 1/(100mile Nl100 represents the number of
5-25 100 miles of pipeline year) annual leaks for 100 miles of
pipeline, i.e.,
Nl100 = Dl / Dpl 100.
Figure Nb100 Annual blowouts per 1/(100mile Nb100 represents the number of
5-26 100 miles of pipeline year) annual blowouts for 100 miles
of pipeline, i.e.,
Nb100 = Db / Dpl 100.

The pipeline leaks were analyzed spatially using File A originating from CPS (July 2001 till Oc-
tober 2009). File B is created using pipeline leaks geo-referenced with their addresses as rec-
orded in CPS. Not all leaks could be geo-referenced due to misspelling and other entry errors.
Figure 2-12 shows all the geo-referenced pipeline leaks. As shown in Figure 5-21, the number

Page | 59
5. Statistical Analysis and Results
of pipeline leaks between 2001 and 2009 are counted per square mile. In West Holly-
wood/Hollywood Hills, there are up to 430 leaks per square mile. Figure 5-22 shows a similar
density map for pipeline blowouts. Unlike Figure 5-21, there are more regions with dense con-
centrations of blowouts with up to 38 blowouts per square mile. The regions of high-density blo-
wouts are located close to the regions of high-density leaks. Using Figure 5-21 and Figure 5-22,
Figure 5-23 shows the ratios between the numbers of blowouts and leaks. These ratios indicate
regions in which pipelines blowout more than they break. The regions are located far from the
high-density regions in Figure 5-21 and Figure 5-22. In the San Fernando Valley, the number
blowouts are comparable to the number of leaks.

434

34.3

372

34.2

310

Dl : Number of leaks
34.1
248
Latitude

34
187

33.9
125

33.8
63

33.7
1
-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1
Longitude

Figure 5-21. Number of pipeline leaks per square mile (July 2001 to October 2009); data
source: File B.

Page | 60
5. Statistical Analysis and Results
38

34.3

32

34.2

27

Db : Number of blowouts
34.1
22
Latitude

34
16

33.9
11

33.8
5

33.7
0
-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1
Longitude

Figure 5-22. Spatial distribution of blowouts per square mile (July 2001 to October 2009);
data source: File B.

Page | 61
5. Statistical Analysis and Results
0.71

34.3

0.61

34.2

Rbl : Ratio of blowouts / leaks


0.51

34.1
0.4
Latitude

34
0.3

33.9
0.2

33.8
0.1

33.7
0
-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1
Longitude

Figure 5-23. Spatial distribution of ratio between blowouts and leaks (July 2001 to Octo-
ber 2009); data source: File B.

Figure 5-24 shows the pipe length per square mile throughout LADWP system. The spatial dis-
tribution of leaks per 100 miles of pipeline is also represented, in addition to leaks and blowouts,
to account for different pipe densities in LADWP system. The data on pipe length are from
LADWP GIS model (File F). As shown in Figure 5-24, there are 30 miles of pipelines per square
mile in Los Angeles; central LA, south LA, east LA, Marina Del Ray and parts of San Pedro
have the densest concentrations of pipelines.

Based on Figure 5-21 and Figure 5-24, Figure 5-25 shows how the number of pipeline leaks per
100 miles of pipeline is distributed over Los Angeles. Leaks are counted between July 2001 and
October 2009. The Hollywood Hills area harbors the largest number of annual leaks per 100
miles of pipeline (over 200). The pipes in these pressure zones are mostly old steel pipes, which
are the most leaky pipe materials (Figure 5-8). Zones 1096, 1216 and 1424 are pressure zones
in the Hollywood Hills area; they perform the worst in terms of number of leaks per mile (Figure
5-5).

Page | 62
5. Statistical Analysis and Results
Figure 5-26 shows the spatial distribution for the number of annual blowouts per 100 miles of
pipeline. West Hollywood is most prone to blowouts (close to 20 blowouts per 100 miles of pipe-
line). This agrees to the large number of blowout incidents in Zone 579 (a gravity drive zone).
Compared to Figure 5-25, the blowouts per mile are more spatially spread out; central LA and
San Fernando Valley both have multiple concentrations of high blowouts per mile.

30

34.3

26

34.2

22

Dpl : Pipe density (length)


34.1
17
Latitude

34
13

33.9
9

33.8
4

33.7
0
-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1
Longitude

Figure 5-24. Spatial distribution of pipe density (length in miles) per square mile; data
source: File F.

Page | 63
5. Statistical Analysis and Results
216.4

34.3

185.5

Nl100: Annual leaks per 100 miles of pipes


34.2

154.6

34.1
123.7
Latitude

34
92.8

33.9
61.8

33.8
30.9

33.7
0
-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1
Longitude

Figure 5-25. Spatial distribution of annual leaks per 100 mile of pipelines (July 2001 to
October 2009); leak data source: File B; pipe length data source: File F.

Page | 64
5. Statistical Analysis and Results
19.63

34.3

Nb100: Annual blowouts per 100 miles of pipes


16.83

34.2

14.02

34.1
11.22
Latitude

34
8.41

33.9
5.61

33.8
2.8

33.7
0
-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1
Longitude

Figure 5-26. Spatial distribution of blowouts per 100 miles of pipeline (July 2001 to Octo-
ber 2009); leak data source: File B; pipe length data source: File F.

5.6 Variation of water pressure in LADWP system


Figure 5-27 shows the time histories of water pressures measured at eight locations as identi-
fied in Figure 5-28. As shown in Figure 5-27, there are daily variations as well as long-term vari-
ations in the pressure measurements in 2009. In some cases, pressures vary as much as 40 psi
each day. Figure 5-29 represents the variations of pressure during June 2009, which revealed
both daily and weekly pressure cycles. For two stations (Zone 386s 43rd and Figueroa and
Zones 386s Manhattan), there are sudden changes in water pressure near June 1, 2009, which
coincides with the beginning of Los Angeles water rationing program on June 1, 2009, and re-
stricted watering lawns to Mondays and Thursdays.

Page | 65
5. Statistical Analysis and Results
386 - 43rd and Figueroa Regulator Station Pressure
250 386 - 99th Street PS Discharge Pressure
386 - Manhattan PS Discharge Pressure
579 - 4th and Detroit Regulator Station Pressure
579 - Franklin Reservoir Elevation
579 - Roxbury PS Suction Pressure
579 - Venice and Overland Regulator Station Pressure
579 - Wooster and Pico Regulator Station Pressure
200
Water pressure (psi)

150

100

50

0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov
2009

Figure 5-27. Time histories of water pressures at eight locations during 2009.

Figure 5-28. Locations of pressure measurements.

Page | 66
5. Statistical Analysis and Results

220

200

180

160
Water pressure (psi)

140

120

100

80

60

40
Jun Jul
2009

Figure 5-29. Time histories of water pressures (minute intervals) in June 2009.

Figure 5-30 shows the daily average of water pressure measured during 2009, and the corres-
ponding pressure envelop built using maxima and minima of pressure. As shown in Figure 5-30,
more than two stations suddenly display changes in water pressure after June 1. After June 1,
while the water pressure maxima remain consistent for most stations, their minima drop notably
and the variations of minima increase considerably. In other words, the daily variations in water
pressure changed after June 1, especially on the lower end of the pressures. These changes
can be explained with the water rationing program that started on June 1 and restricted watering
on lawns on two days of the week, namely Mondays and Thursdays. The watering of lawns on
two specific days increased water usage these particular days, causing a substantial drop in
water pressure in the LADWP system. Figure 5-31 illustrates the increase in the amplitudes of
pressure cycles in the LADWP system on Mondays and Thursdays. The effects of such an in-
crease in cyclic pressure on pipeline failures are discussed in Section 5.10.

Page | 67
386 - 43rd and Figueroa Regulator Station Pressure
386 - 99th Street PS Discharge Pressure
386 - Manhattan PS Discharge Pressure
579 - 4th and Detroit Regulator Station Pressure max
200 mean
Water pressure (psi)

579 - Franklin Reservoir Elevation


min
579 - Roxbury PS Suction Pressure
579 - Venice and Overland Regulator Station Pressure
579 - Wooster and Pico Regulator Station Pressure
150

100

50
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov
2009

200
Water pressure (psi)

150

100

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov
2009

Figure 5-30. Variation of water pressures (daily means, maxima and minima) in 2009.

Page | 68
5. Statistical Analysis and Results

579 - Venice and Overland Regulator Station Pressure


230

220
daily max
210
water pressure (psi)

200 daily mean

190

180 daily min

170

160

150 Tue Wed Fri Sat Sun


Mon Thur
140
Jun Jul Aug
2009

Figure 5-31. Lowering of water pressures on Mondays and Thursdays.

5.7 Variation of water temperature in the LADWP system


Figure 5-32 shows the variation of water temperatures at three locations during 2009. At Man-
hattan, temperatures do not vary throughout the year while temperatures at the other two loca-
tions peak in September/October and decrease to become the lowest in March. There are sub-
stantial drops in temperatures from October to December in 2009. Figure 5-33 shows the tem-
perature history at the Los Angeles Aqueduct Filtration Plant (LAAFP), over a longer time win-
dow. The annual variations of water temperature are similar between 1996 and 2009.

Page | 69
5. Statistical Analysis and Results

75
LAAFP
HWD45
70 Manhattan

65
Temperatures ( F)
o

60

55

50

45
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Daily intervals

Figure 5-32. 2009 water temperature variations at three locations.

LAAFP
85

80

75
Water temperature (oF)

70

65

60

55

50

45

40

35
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Daily intervals

Figure 5-33. Seasonal water temperature variations between 1996 and 2009.

Page | 70
5. Statistical Analysis and Results

5.8 Relations between pipeline breaks, temperature and pressure var-


iations
Figure 5-34 superimposes the variations of pipeline breaks with water temperatures (Figure
5-32) and pressures (Figure 5-27) from January 2009 to October 2009. As indicated in the up-
per portion, the coldest temperatures from January to March occur at only two locations, which
coincide with a large number of breaks. The slow increase in temperature from March to Octo-
ber seems to correspond to a gradual increase in pipeline breaks. This figure is misleading as it
implies that a pipeline is more likely to break when the temperature increases slowly. In fact, this
figure only shows that the increase in temperature coincides with the summer month during
which the water-rationing program was initiated. It is however not ruled out that the temperature
and pressure variations may have had effects on water pipeline breaks in Summer 2009.

Figure 5-35 shows the variations of water temperatures and pipeline breaks between 2001 and
2009. In the past ten years, the sudden decrease in water temperature during December and
January correlates well with a sudden increase in pipeline breaks. The effects of decreasing
temperatures on water main breaks are well documented (e.g. Habibian, 1994). Between 2001
and 2009, the worst events in the LADWP system happened in the winter of 2006, which was
the coldest winter during that period. However, this seasonal effect during winter cannot explain
the rash in water pipeline breaks in the summer of 2009.

Page | 71
5. Statistical Analysis and Results

temperatures
blowouts
leaks

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov
2009

water pressure

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov
2009

Figure 5-34. Time histories of pipeline breaks, water temperatures (Figure 5-32) and pres-
sures (Figure 5-27); leak data source: File A.

Page | 72
5. Statistical Analysis and Results

Blowouts
Leaks
400
Water temperature
Number of leaks

300

200

100

0
2002 2003 2004 2005 2006 2007 2008 2009
Monthly intervals

50
Number of leaks

40

30

20

10

0
Jul01 Jan02 Jul02 Jan03 Jul03 Jan04 Jul04 Jan05 Jul05 Jan06 Jul06 Jan07 Jul07 Jan08 Jul08 Jan09 Jul09
Daily intervals

Figure 5-35. Time histories of pipeline breaks and water temperatures; leak data source:
File A.

5.9 Effects of Corrosion Damage and Soil Corrosivity


Soil corrosivity affects pipe conditions and therefore pipeline breaks. As shown in Figure 5-36,
soil corrosivity varies spatially in Los Angeles. Soil corrosivity is estimated using the soil resistiv-
ity; the lower the soil resistivity, the higher the soil corrosivity. The regions with the most corro-
sive soils happen to coincide with the regions with larger numbers of pipeline failures. Figure
5-38 breaks down pipeline breaks and blowouts in seven levels of soil corrostivity, i.e., soil re-
sistivity. There are significantly more pipeline breaks in the two most corrosive levels (0-3,000
ohm-cm). The number of pipeline breaks and blowouts, when normalized per mile of pipelines in
different corrosive levels, increases monotonically with soil corrosivity. In addition, the ratio be-
tween the number of blowouts and breaks increase monotonically with soil corrosivity. These
results provide strong evidence that corrosive soils increase the frequency of pipeline failures.
Figure 5-38 agrees with the LADWP corrosivity and break analysis shown in Figure 3-14 and
Figure 3-15.

Page | 73
5. Statistical Analysis and Results

more corrosive

less corrosive

Figure 5-36. Soil corrosivity map; corrosivity data source: File F.

Page | 74
5. Statistical Analysis and Results
8000
# of pipe breaks Leaks
6000 Blowouts

4000

2000

0
0 - 1000 1000 - 3000 3000 - 5000 5000 - 10000 10000 - 20000 20000 +
extremly corrosive highly corrosive corrosive moderately corro. mildly corrosive non-corrosive

25
100 miles of pipes

Leaks
Annul breaks per

20
Blowouts
15

10

0
0 - 1000 1000 - 3000 3000 - 5000 5000 - 10000 10000 - 20000 20000 +

0.2
Ratio of blowout / leak

0.15

0.1

0.05

0
0 - 1000 1000 - 3000 3000 - 5000 5000 - 10000 10000 - 20000 20000 + overall
Resistivity (ohm-cm)

Figure 5-37. Distribution of the number of pipeline breaks and blowouts in term of soil
corrosivity; leak data source: File B; pipe data source: File F.

The strong correlation between soil corrosivity and pipeline breaks per mile was further ana-
lyzed. Instead of using six discrete levels of soil resistivity as shown in Figure 5-37, Figure
5-38displays a continuous variation of soil resistivity and examines its relationship to breaks and
blowouts. Most breaks and blowouts happen when the soil resistivity is about 1000 ohm-cm (top
of Figure 5-38). This is because most pipelines are located in soils that have a resistivity of 1000
ohm-cm (middle of Figure 5-38). When the pipeline breaks are normalized per mile of pipelines,
the number of pipeline breaks per mile increases when soil resistivity decreases (i.e., corrosive
soils), as lower Figure 5-38 clearly shows.

Page | 75
5. Statistical Analysis and Results
Pipe breaks Blowouts
600 80
# of pipe breaks

60

# of blowouts
400
40
200
20

0 0
3 4 3 4
10 10 10 10
Pipe system
250
Pipe length (miles)

200

150

100

50
3 4
10 10

3 0.8
Blowouts per mile
Breaks per mile

0.6
2
0.4
1
0.2

0 0
3 4 3 4
10 10 10 10
Resistivity (ohm-cm) Resistivity (ohm-cm)

Figure 5-38. Variation of pipeline breaks and blowouts as a function of soil resistivity;
leak data source: File B; pipe data source: File F.

Based on Figure 5-38, Figure 5-39, Figure 5-40, and Figure 5-41 show how pipeline breaks are
affected by pipe material, pipe diameter, and pipe age. As shown in Figure 5-39, when pipes are
cast iron, asbestos cement and ductile iron, they are more likely to break and blowout if they are
located in corrosive soils. However pipes of steel and other materials do not follow such a trend;
in fact, steel pipes leak more per mile in less corrosive soils. Cast iron pipes blowout the most in
corrosive soils, implying that they are more affected by corrosive soils than other pipe materials.

Pipes of all diameters break more when they are in corrosive soils. Figure 5-40 shows the ef-
fects of soil corrosivity on breaks per mile for different pipe sizes (smaller than 4-inch, 6, 8, and
12-inch and above). Small pipes are more affected by soil corrosion. The trend for blowouts per
mile of pipelines is less obvious than the trend for breaks per mile of pipelines, but blowouts
generally do increase with soil corrosivity.

Figure 5-41 breaks down pipelines by age and shows that the oldest pipelines break and blow
out more than younger ones. There are more breaks in the oldest (100-year) pipelines though

Page | 76
5. Statistical Analysis and Results
there are fewer of them than 60-year old pipelines. In all age ranges, the numbers of breaks and
blowouts per mile increases when soils are more corrosive.

Pipe breaks by pipe materials Blowouts by pipe materials


600
100
# of pipe breaks

# of blowouts
400
50
200

0 0
3 4 3 4
10 10 10 10
Pipe system
300
Pipe length (miles)

AC
200 CI
DI
100 STL
others
0
3 4
10 10

10 1

8 0.8
Blowouts per mile
Breaks per mile

6 0.6

4 0.4

2 0.2

0 0
3 4 3 4
10 10 10 10
Resistivity (ohm-cm) Resistivity (ohm-cm)

Figure 5-39. Variation of pipeline breaks and blowouts as a function of soil resistivity for
different pipe materials; leak data source: File B; pipe data source: File F.

Page | 77
5. Statistical Analysis and Results

Pipe breaks by pipe diameters Blowouts by pipe diameters


400
60
# of pipe breaks

# of blowouts
40
200
20

0 0
3 4 3 4
10 10 10 10
Pipe system
200
Pipe length (miles)

<4 in pipe
6 in
100
8 in
12+ in

0
3 4
10 10

7 0.8

6 0.7

0.6
5
Blowouts per mile
Breaks per mile

0.5
4
0.4
3
0.3
2
0.2

1 0.1

0 0
3 4 3 4
10 10 10 10
Resistivity (ohm-cm) Resistivity (ohm-cm)

Figure 5-40. Variation of pipeline breaks and blowouts as a function of soil resistivity for
different pipe diameters; leak data source: File B; pipe data source: File F.

Page | 78
5. Statistical Analysis and Results

Pipe breaks by pipe ages Blowouts by pipe ages


40
# of pipe breaks

# of blowouts
200

20
100

0 0
3 4 3 4
10 10 10 10
Pipe system
200
Pipe length (miles)

pipe age<20 years old


20 ~ 40 years old
100 40 ~ 60 years old
60 ~ 80 years old
80 ~ 100 years old
0
3 4
10 10

5 1
Blowouts per mile

4 0.8
Breaks per mile

3 0.6

2 0.4

1 0.2

0 0
3 4 3 4
10 10 10 10
Resistivity (ohm-cm) Resistivity (ohm-cm)

Figure 5-41. Variation of pipeline breaks and blowouts as a function of soil resistivity for
different pipe ages; leak data source: File B; pipe data source: File F.

Corrosion is a chemical process that slowly degrades material properties. The longer a pipeis
exposed to corrosive soils, more likely the pipe is to break (Figure 5-41).Figure 5-42 illustrates
how breaks and blowouts per mile of pipeline are affected by pipe corrosion. It introduces a new
variable (bottom Figure 5-42), which is characterized using the product of pipe age and soil

Page | 79
5. Statistical Analysis and Results
conductivity (conductivity being the inverse of resistivity and a measure of soil corrosivity). At
the top of Figure 5-42, pipeline breaks and blowouts per mile are clearly shown to increase with
soil conductivity. In the middle Figure 5-42, the relationship between breaks or blowouts and
pipe age is not apparent. Breaks and blowouts are most strongly correlated with a combination
of the effects of soil conductivity and pipe age (bottom Figure 5-42), breaks and blowouts show
the strongest correlations in Figure 5-42.The red circles in Figure 5-42(bottom of figure) com-
pared to the top figure for older pipelines (top of figure).Figure 5-43depicts the LADWP pipeline
system in terms of pipe age and soil resistivity. The figure shows that for the most part, pipes in
each age group are located in soils with similar corrosivity. On average, 90 year-old pipelines
are located in the most corrosive soils, while 60-70 year-old pipelines are located in the least
corrosive soils.

Pipe breaks Blowouts


3

Blowouts per mile


Breaks per mile

2 0.4

1 0.2

0 0
-3 -3
10 10
1/resistivity (1/ohm-cm) 1/resistivity (1/ohm-cm)

4 0.5
Blowouts per mile
Breaks per mile

3 0.4

2 0.3

1 0.2

0 0.1
20 40 60 80 100 20 40 60 80 100
Pipe age (year) Pipe age (year)

4 0.8
Blowouts per mile
Breaks per mile

3 0.6

2 0.4

1 0.2

0 0
-2 -1 -2 -1
10 10 10 10
Pipe age * 1/resistivity (year/ohm-cm) Pipe age * 1/resistivity (year/ohm-cm)

Figure 5-42. Variation of pipeline breaks and blowouts as a function of soil corrosivity
and pipe age; leak data source: File B; pipe data source: File F.

Page | 80
5. Statistical Analysis and Results

pipe age<10 years old


10 ~ 20 years old
90 20 ~ 30 years old
30 ~ 40 years old
80
40 ~ 50 years old
70 50 ~ 60 years old
Pipe length (miles)

60 ~ 70 years old
60
70 ~ 80 years old
50 80 ~ 90 years old
90 ~ 100 years old
40
30
20
10
0 3 4
10 10
Soil resistivity (ohm-cm)

3000
Average soil resistivity (ohm-cm)

2500

2000

1500

1000

500

0
10 20 30 40 50 60 70 80 90 100
Pipe age (years)

Figure 5-43. Pipeline system in terms of pipe age and soil resistivity.

As shown in Figure 5-37 to Figure 5-42, pipelines break more per unit length as the soil be-
comes more corrosive. However, as shown in Figure 5-44, a comparison of density maps of soil
corrosivity and pipeline break per mile of pipeline indicates that the regions of high corrosivity do
not coincide with regions of high breaks per mile. This result suggests that pipeline breaks and
soil corrosivity are correlated on a global scale and that other factors may contribute to breaking
pipelines. On average, pipelines break more in regions of high soil corrosivity (Figure 5-37 to
Figure 5-42), but break rates are not necessarily high in these regions. Soil corrosivity is only
one of the many factors influencing pipeline breaks.

Page | 81
5. Statistical Analysis and Results

more corros.
34.3

34.2

(Corrosivity per sq. mile)


34.1
Latitude

34

33.9

less corros.
33.8

33.7

-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1


Longitude
216.4

34.3

185.5

Nl100: Annual leaks per 100 miles of pipes


34.2

154.6

34.1
123.7
Latitude

34
92.8

33.9
61.8

33.8
30.9

33.7
0
-118.7 -118.6 -118.5 -118.4 -118.3 -118.2 -118.1
Longitude

Figure 5-44. Spatial distribution of soil corrosivity (top) and pipeline leaks per 100 miles
of pipeline (bottom); leak data source: File B; pipe data source: File F.

Page | 82
5. Statistical Analysis and Results

5.10 Explanation for Summer 2009 pipeline breaks


It has been suggested that the increase in water pipeline breaks in Summer 2009 may have
been due to a concurrent water-rationing program. A simplified model supporting that hypothe-
sis is presented below

5.10.1 Model assumptions


A simplified model was developed to explain the increase in pipeline breaks that may have re-
sulted from changes in the daily cyclic amplitude of water pressure, during the summer of 2009,
when lawn watering was restricted to i.e., an effect of watering lawns during two specific days
per week. This simplified model only explains a temporary increase of water pipeline breaks
during the summer of 2009, and does not cover all aspects of pipeline failure in a water distribu-
tion system.

The two main variables of the simplified model are (1) number of cycles N and (2) amplitude of
cyclic pressure S. Selected for the sake of simplicity, these variables may not represent all the
types of loads applied on pipelines.

The model assumes that pipes are subjected to different numbers of cycles N, but to consistent
cyclic pressure variation, and that the pipes break according to the same fatigue curve S(N).
The model also assumes that the pressure cycles have constant amplitude, which is another
drastic assumption in view of the pressure variability across the LADWP system. Therefore, the
simplified model only illustrates the effects of operational changes on a generic aging water dis-
tribution system. A more realistic model would require many more variables and much more de-
tailed calibration to account for the complexities inherent in the LADWP system.

As shown in Figure 5-45, materials in general exhibit different fatigue behaviors. Materials fati-
gue when they are subjected to a large number of stress cycles, and eventually fail after a par-
ticular number of cycles. For the cast iron pipes used in water distribution systems, the range of
fatigue behaviors corresponds to high stress cycles and finite life. In other words, pipes do not
have an infinite life.

Page | 83
5. Statistical Analysis and Results

Figure 5-45. Ranges of fatigue behaviors for materials in general.

5.10.2 Model formulation


The service life expectancy of the pipes of a water distribution system is assumed to be con-
trolled by a exponentially decaying fatigue curve S(N) (shown in Figure 5-46). This curve S(N)
indicates that it takes N cycles of amplitude S for a pipe to break due to fatigue.

S
amplitude
of cyclic
pressure

S2

S1

N2 N1 N number of cycles

Figure 5-46. Due to material fatigue, the number of pressure cycles before a pipe fails de-
creases as the amplitude of cyclic pressure increases.

Page | 84
5. Statistical Analysis and Results
As shown in Figure 5-47, the water distribution system is made of M pipes that have been sub-
jected to N cycles during their service lives. The relation between pipe quantity and number of
cycles can be described using a statistical distribution m(N), which is unfortunately unknown.
Fortunately, the analysis needs only the right tail of the distribution m(N); only the pipes that
have been long enough in service are candidates for fatigue failures. The analysis does not
need the rest of the distribution m(N),which corresponds to younger pipes not yet subject to fati-
gue.

Based on the constant rate of pipeline breaks in the LADWP system (Figure 2-10), it is reason-
able to assume that the right tail of the distribution m(N) is flat; i.e., the quantity of pipes is inde-
pendent of the number of cycles. As shown in Figure 5-47, this assumption corresponds to m(N)
= Q where Q is the number of pipeline breaks for a given number of cycles.

As shown in Figure 5-48, the right tail of the uniform distribution is cut off at N1, an upper bound
value for Nthat is estimated from the S(N) fatigue curve, assuming that the amplitude S1 of
stress cycles is known and remains constant. No pipes can survive more than N1 stress cycles.
When all the pipes are cycled once a day, N1varies at a constant rate; i.e., each day Q pipes fail
due to fatigue. The corresponding annual rate of pipeline failure is R = Q 360.

Pipe
Quantity
N2 N1
Q

N increase in number of pipe failures due to an


increase in cyclic amplitude from S1 to S2

Number of Cycles N

Figure 5-47. Assumed uniform distribution of pipe quantity as a function of number N of


pressure cycles.

Page | 85
5. Statistical Analysis and Results

Amplitude ofcyclicstress
S2

S1
N1
Nf Numberofcyclestofailure

N1N2 t

N2 Time(year)

N1N2
NumberofBreaks

Rx(N1N2)

Time(year)

Figure 5-48. Illustration of increase in number of fatigueinduced failures due to an in-


crease in amplitude of pressure cycles from S1 to S2 during time interval t. The number
of cycles to reach pipe failure decreases abruptly from N1 to N2 during t, and returns to
N1 when the amplitude returns to S1. The break rate, which was initially constant, jumps
suddenly, then becomes constant again during t. When the amplitude decreases back
to S1, the breaks stop during N1-N2 days, then the breaks start again at the original con-
stant rate.

As shown in Figure 5-48, when the amplitude of stress cycles increases suddenly from S1 to S2,
then N1 decreases abruptly to N2. The sudden increase in pipeline failure is therefore
N = Q (N1 N2) where N1 and N2 are expressed in days (one cycle per day), or
N = R (N1 N2) where N1 and N2 are expressed in years. During the time interval t, the cyc-
lic amplitude is maintained constant, and the break rate is again equal to R. When the amplitude
of stress cycles decreases back to S1, then N2 increases back to N1. There is no pipeline failure
during N1N2 days. All the pipes that were scheduled to fail for the amplitude S1 have already
failed for the amplitude S2. This quiet period of no-pipeline failure stops after N1N2 days, and
the break rate is again equal to R.

The model only illustrates the effect of an increase of cyclic stress amplitude on pipeline breaks.
It is based on a single S(N) curve and oversimplifies how materials respond to fatigue cycles of
irregular and asymmetric amplitudes. It also implies that the number of pipeline failures increas-
es immediately with cyclic amplitude, although this is likely to spread over a longer time period
for the complex LADWP system.

Page | 86
5. Statistical Analysis and Results
5.10.3 Illustrative example
Figure 5-49 shows the distribution of cast iron pipe length as a function of pipe age. This distri-
bution can be converted into a distribution of pipe quantity as a function of N daily pressure
cycles. The conversion from length to quantity assumes an average pipe length of 20 ft. As pre-
viously mentioned, the simplified model only considers the right tail of the distribution in Figure
5-50, which shows that many LADWP pipes have been subjected to more than 30,000 daily
stress cycles. The distribution of Figure 5-50 is useful to define an upper bound that limits the
uniform distribution of Figure 5-51.

200
180
160
140
Pipelength(mile)

120
100
80
60
40
20
0
1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96
PipeAge(year)

Figure 5-49. Distribution of cast iron pipe length in term of pipe age in LADWP water dis-
tribution system

Page | 87
5. Statistical Analysis and Results

60000

50000
PipeQuantity
40000

30000

20000

10000

0
2190
4015
5840
7665
9490
365

11315
13140
14965
16790
18615
20440
22265
24090
25915
27740
29565
31390
33215
35040
NNumberofDailyPressureCycles

Figure 5-50. Distribution of pipe quantity as a function of N number of daily pressure


cycles assuming an average pipe length of 20 feet.

1.0

0.8
Quantityofpipes

0.6

0.4

0.2

0.0
20000 30000 40000 50000
Numberofcyclestofailure

Figure 5-51. Assumed distribution of pipe quantity as a function of number of cycles to


failures.

For the sake of illustration, the numbers of cycles N1 and N2 are calculated using a fatigue curve
obtained by Mohebbi et al. (2009) for gray cast iron in the United Kingdom (Figure 5-52). No
specific fatigue curves were obtained for LADWP materials as they usually require a few months

Page | 88
5. Statistical Analysis and Results
to perform, unfortunately too long for this short-duration project. The peak stress was calculated
as detailed in Table 5-3. The hoop stress in the pipe wall was estimated as follows:

pD
= (5)
2t

where p is the water pressure inside the pipe; D is the internal pipe diameter; and t is the pipe
wall thickness. For the values in Table 5-3, the calculated points (N, peak stress) fall close to
the lowest fatigue curves of Figure 5-52, which indicates that fatigue is a possible cause of fail-
ure for LADWP systems. However, the calculations assume (1) large operating pressures, (2)
reduction of the pipe wall thickness t due to corrosion, and (3) existence of imperfections in the
cast iron. In any case, Figure 5-52 justifies the simplified model that assumes fatigue failure as a
possible source of pipeline breaks.

Table 5-4 and Figure 5-53 illustrate how to apply the simplified model to quantitatively estimate
the increase in pipeline failures resulting from an increase of cyclic pressure amplitude. It is as-
sumed (1) that the increase in cyclic pressure amplitude is large enough to decrease by 100 the
number of cycles to reach failure, and (2) that the pressure returns to normal after the 3 months
of summer. Removing 100 cycles from N1, the number of cycles to failure is a small variation in
the context of fatigue analysis; it implies that the pipe service life is reduced by 100 days, or
0.27 year. The break rate R is calibrated from the blowout rate, which is 205 blowouts per year.
Figure 5-53 shows how the pipeline breaks vary with time. They suddenly increase by 55 when
the cyclic pressure changes, increase at the usual rate during the summer, flatten out during
100 days, and then increase again to the usual rate.

Page | 89
5. Statistical Analysis and Results
175
Pipe3
150
Pipe2
PeakStress(Mpa) 125
Pipe1

100 LADWP

75

50

25
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Numberofcycles
Figure 5-52. S(N) fatigue curves obtained for gray cast iron and different imperfection
size a/w (after Mohebbi et al., 2009), and estimated ranges of cyclic pressure amplitudes
and number of pressure cycles (red area).

Table 5-3. Parameters and values used for simplified fatigue analysis.

Variable Values
Pipeinternaldiameter 8 inch 20.3 cm
Wallthicknessofintactpipe 0.5 inch 1.3 cm
Pipeinternalpressure 150 psi 1.0 MPa
Hoopstressofintactpipe 1200 psi 8.3 MPa
Wallthicknessofcorrodedpipe 0.1 inch 0.04 cm
Hoopstressofcorrodedpipe 6000 psi 41.4 MPa
Ultimatetensilestrengthforgreycastiron 30 ksi 206.9 MPa
Endurancelimit(40%strength) 12 ksi 82.8 MPa
Maximuminternalpressure 260 psi 1.8 MPa
Maximumhoopstress 10400 psi 71.7 MPa
Minimuminternalpressure 30 psi 0.2 MPa
Minimumhoopstress 1200 psi 8.3 MPa
Midrangestress 5800 psi 40.0 MPa
Amplitudeofcyclicstress 9200 psi 63.4 MPa

Page | 90
5. Statistical Analysis and Results
Table 5-4. Parameters and values used for estimating the increase in pipe breaks due to
fatigue.

Definition Variable Value Unit


NumberofcyclestofailureforstressamplitudeS1 N1 40000
NumberofcyclestofailureforstressamplitudeS2 N2 39900
Changeinnumberofcyclestofailure N 100
Shorteningofoperationallifeassumingonecycleperday L 0.27 Year
Dailypipefailuresinsystemforanumberofcycles Q 0.57 1/Day
Timeperiodduringchangeofamplitude t 0.25 Year
Annualpipefailuresinsystem R 205 1/Year

800
Cumulativenumbersofbreaks

700
600
500
400
300
200
100
0
0 1 2 3 4
Time(year)

Figure 5-53. Simplified explanation for increase of pipe breaks caused by fatigue and an
increase in the amplitude of pressure cycles.

The model, although it is useful to understand the combined effects of operational changes and
fatigue on aging pipes, is still to be calibrated in greater detail for the complex LADWP water
system.

5.11 Summary
The pipeline break data provided by LADWP have been extensively analyzed. Data collection
and management are not trivial for the LADWP water system, which is one of the largest in the
U.S. Statistical analyses were significantly complicated by missing data. (Over 45% of CPS data
are missing in the GIS model) LADWP should, however, be commended for starting a new data

Page | 91
5. Statistical Analysis and Results
collection program on pipe repairs and building a GIS master model of its complete water distri-
bution system.

The analysis shows that, in terms of total numbers of breaks, cast iron pipes break more than
other pipes because they are more numerous, and that, in term of breaks per unit length, steel
pipes break more than pipes made of any other material. When normalized by unit length, cast
iron is second in breaks compared to steel pipes. More blowouts occur in cast iron pipes than in
other types. Smaller pipes (4-inch and smaller) break more per unit length than other sizes,
while 8-inch pipes blowout the most per unit length.

The performance of the LADWP system was compared with those of six other water distribution
systems. LADWP is near the national average on pipeline breaks per 100 miles of pipeline.

Since 2001 pipelines have broken in larger numbers in the areas surrounding the Santa Monica
Mountains, and in the West Hollywood/Hollywood Hills area. In the same areas, the pipelines
have also broken and blown out in greater number in relation to miles of pipeline. The pipelines
in the San Fernando Valley displayed the highest blowout-to-break ratio.

In the LADWP water distribution system, pipeline breaks are affected by water temperatures,
water pressures, and soil corrosivity. Pipeline breaks usually increase when temperature de-
creases sharply during the coldest months; they are also affected by pressure cycles, especially
those with large amplitudes. Pipelines break and blowout more in highly corrosive soils, and
among all pipe materials, cast iron pipes blowout even more due to soil corrosivity. The effects
of soil corrosivity are accentuated by pipe age because older pipelines are subject to the corro-
sivity for a longer period.

5.11.1 Probable explanations and recommendations


As pressure measurements indicated at several locations of the LADWP water distribution sys-
tem, the water pressure dropped significantly on Mondays and Thursdays starting June 1, 2009,
the beginning of the water-rationing program in Los Angeles. These temporary water pressure
drops coincided an unusual increase in water flow, which was caused by the restriction that
lawns be watered only on watering lawns during these two days of the week. As the result of the
pressure drop, the amplitude of the daily pressure cycles was augmented. This affected nega-
tively the pipe materials with lower fatigue resistance. Low resistance to fatigue is a common
characteristic of corroded cast iron pipes, and can result in sudden brittle failures after a large
number of load cycles. The fatigue effects on cast iron pipes were not felt immediately after
June 1, but became gradually evident during the month of July, August, and September, as ag-
ing cast iron pipes exceeded their fatigue resistance to pressure cycles. The increase in fatigue-
related pipeline breaks resulted in an increase in dramatic pipeline failures referred to as blo-
wouts, which raised media attention and public concern in view of the large damage to street
pavements and properties.

This explanation, although it is the most probable at this stage of the study, requires further in-
vestigation to be fully validated for the complex LADWP water distribution system. Nevertheless,
based on this explanation, it is recommended to avoid abrupt drops of water pressure as much
as possible. Before the summer of 2010, it is recommended that LADWP alters its water-

Page | 92
5. Statistical Analysis and Results
rationing program so that variations of water pressure will be distributed more evenly over time,
and thus avoid increases in the amplitude of water pressure cycles. For instance, the water ra-
tioning could be devised so that residence having odd street numbers would water lawns on
Mondays and Thursdays, while those with even street numbers would be allowed to water
lawns only on Tuesday and Fridays. This recommendation does not rule out devising other solu-
tions for reducing the sudden drops of water pressures and negative effects on corroded cast
iron pipes.

But this solutionwhatever the details may beis a short-term expediency, intended only to
avoid another rash of blowouts in the summer of 2010. It does not eliminate the need for
LADWP to undertake a more profound upgrading of the water distribution system and to invest
in long-term research that supports development and implementation of better field inspection
techniques and more efficient pipe replacement programs and integrated asset management.

Page | 93
6 HYDRAULIC PILOT STUDY2

LADWP monitors water pressure using mostly analog pressure dial gauges that are time-
consuming to process and unresponsive during times of crisis. In addition, these devices do not
measure high-frequency pressure waves, which are known to damage water pipelines. The in-
vestigation of water pipeline breaks created a pressing need for real-time monitoring of water
pressure at various locations of the LADWP water distribution systems. It undertook a pilot
study to deploy portable accurate instruments and audit what was happening in the water distri-
bution system.

6.1 Water hammers and pressure surges


Water hammers and pressure surges in piping systems are created when a change in the pipe-
line flow rate occurs. The source of the change in flow rate may be a result of normal opera-
tions, such as the starting or stopping of a pump, or the opening or closing of a valve. In addi-
tion, sudden and unplanned changes in flow can occur as a consequence of loss of power to
pumps or a pipeline break.

A major source of water hammers in pipelines is valve operations. If a controlled valve is


opened too quickly, the pipe pressure will drop suddenly. The sudden pressure drop propa-
gates upstream from the valve site as a pressure downsurge wave that may cause the hydraulic
grade line (HGL) to drop far enough below the pipeline crown to form a vapor cavity. On the
other hand, closing an open valve too quickly can create a sudden pressure rise as the flow of
kinetic energy is converted to pressure energy.

Control of water hammers induced by valve operations is simplethe rate of valve motion is
adjusted to an appropriate speed. Prevention of inadvertent rapid control valve motion is at-
tained by using the proper control mechanisms. In addition, pressure relief valves may be in-
stalled to release any untoward rise in pressure.

2
This section of the report is co-authored by Scott Foster and John List, Flow Science, Pasade-
na.

Page | 94
6. Hydraulic Pilot Study
When a pumping system is shut down as part of normal operations, or by power failure, the hy-
draulic grade line (HGL) downstream of the pump station falls very rapidly. The rapidity of the
pressure drop is controlled primarily by the polar moment of the inertia of the pump/motor sys-
tem. If the inertia is high the HGL falls slowly, but for most small pumping units it drops to the
suction water elevation, or below, in a second or so. This rapid pressure drop on the down-
stream side (created by loss of power to the pump) travels out along the downstream pipelines
as a pressure drop wave (i.e., low pressure wave) moving at a speed of 500-4500 ft/sec, de-
pending upon the pipe material and dimensions and the fluid being pumped, which in this case
is water.

Since the steady flow HGL slopes down toward the pipeline discharge point, and in many cases
the pipeline profile may rise toward the discharge point, at some location along the profile the
dropping HGL may fall below the invert of the pipe, thereby creating a vacuum in the pipe. If the
HGL falls one atmospheric pressure head (33-34 feet) below the pipe crown, the pressure in the
pipeline will be less than the vapor pressure of the fluid and it will begin to boil at ambient tem-
perature. Once boiling occurs a vapor cavity will form at the crown of the pipe and the pressure
downsurge wave will continue propagating along the pipeline leaving behind a pipeline under
vacuum and filled with boiling water. When the downsurge wave reaches the discharge point,
or other more or less constant pressure point, it is reflected as a re-pressurization wave. This
wave travels back up the pipe, removing the vacuum and stops the boiling. When there is an
extensive vapor cavity it will tend to accumulate at some point in the pipeline (usually at a break
in the slope or local high point) and collapse explosively. The net result is a localized region in
the pipe that is subjected to an extremely high impulsive pressurea water hammer. As the re-
pressurization wave finally returns to the pump station it may close the pump check valve sud-
denly and create an additional water hammer. If pump control valves are installed the reverse
flow may accelerate through these valves and lead to high water hammer pressures as the
valves close on the reverse flow.

There are, therefore, two sources of water hammers associated with power failure to pumps
one from vapor cavity formation, the other from the return flow reaching the pump check valve
or flow control valve. Maximum pressures generated by the first mechanism (vapor cavity col-
lapse) cannot be predicted for two reasons. First, it is almost impossible to predict where the
vapor cavity collapse will occur. Second, the speed of collapse cannot be accurately predicted.
Water hammers resulting from pump check valves closing can be predicted quite accurately
provided vapor cavity formation does not occur in the pipeline. Accurate predictions for the
closing of pump flow control valves are dependent on knowing both the exact closing time for
the valves and the valve characteristics.

For in-line (i.e. booster) pump stations that are the primary type of pump station utilized by
LADWP, the fall in downstream HGL is usually accompanied by a concomitant rise in the up-
stream HGL as the upstream flow is brought to rest. This high-pressure wave on the upstream
side of the booster pump station behaves identically to the low-pressure wave on the down-
stream side of the booster pump station. It will propagate out into the upstream side pipelines
as a pressure upsurge wave, potentially increasing the HGL elevation sufficiently so that the
pipelines become over pressurized.

Page | 95
6. Hydraulic Pilot Study
For LADWPs system, valve and booster pump station operations, along with past pipeline rup-
tures, are likely the primary cause of pressure transients in the system.

6.2 Pressure zones selected


The purpose of the pilot study is to take pressure measurements within the particular zone of
interest, in this case the 579 zone, and see if there are any sudden changes in pressure that
would indicate the presence of a water hammer or pressure surge event within the zone.

For the pilot study of the LADWP distribution system, the 579 pressure zone located south of
the Hollywood Hills was selected based on the number of breaks and leaks that had been rec-
orded between July and October 2009. This pressure zone is a gravity fed zone from the 777
zone via pressure regulators set at the zone boundaries. This pressure zone also gravity feeds
the 477, 448, and 386 pressure zones, along with an emergency connection to the Beverly Hills
system. Since it is a gravity fed zone, valves would be the primary source of transients, aside
from any actual pipeline leaks or ruptures. This could be the result of the rapid operation of
pressure regulating station valves or a sudden change in flow demand from a large user, such
as a studio, school, the Wilshire Country Club, or the Hollywood Forever Cemetery. There is
one pump station, Roxbury, which pumps from the 579 zone and into the 1096 zone near the
western edge of the pressure zone. From the information provided, this would appear to be the
only pump station that is directly connected to this zone.

The 579 zone consists of many pipes in a grid style plan layout. Almost all the pipes appear to
be 12 inches in diameter or smaller, with a majority in the 6 inch to 8 inch diameter range. One
would expect pipes of this size to be constructed of cast iron, however, this needs to be verified.

Page | 96
6. Hydraulic Pilot Study

Figure 6-1. 579 pressure zone with leaks and breaks

Figure 6-1 shows the blowouts (red triangles) and breaks (red circles) in the 579 zone between
July and October 2009. TP-1 pressure recorders were installed at zonal boundaries at the
pressure regulating station at Van Ness and Sunset and at the pressure regulating station at
Van Ness and 3rd (see Figure 6-1). Both of these stations reduce the pressure head from the
777 zone to the 579 zone. These locations were chosen based on that they were secure loca-
tions and had an available power supply.

To record water hammer and pressure surge events, pressure recorders capable of recording
pressure every 1/100th of a second are required. With water hammer waves propagating
throughout the system at potentially more than 4000 feet per second, standard chart recorders
would not be capable of picking up the sudden changes in pressure associated with these
events.

A system that is capable of monitoring over extended periods of time in a snoozing mode, re-
cording background pressure at any interval desired by the user, yet detecting and recording

Page | 97
6. Hydraulic Pilot Study
any transient in detail is ideal for this application. The system chosen for application here is
from Pipetech International, the TP-1 Transient Pressure Monitoring System. This system con-
sists of a controller or signal processor, a pressure transducer, a hand-held PDA (Personal Digi-
tal Assistant) communication device, a Wi-Fi signal for wireless data upload and download, and
a GPS antenna. Pressure data is recorded in digital form within the 23 megabyte memory of
the TP-1, which at optimum recording settings will hold several months of data. The system is
very busy even in the snoozing mode, statistically analyzing the pressure data, which has
been digitized at a frequency of up to 1 kHz. It continuously computes a running average, and
when a pressure is detected that differs significantly from the average in other words when a
transient is detected an alarm clock goes off that wakes up and records all data at another
user-set rate up to 100 Hz. This continues until the transient has passed, at which time the sys-
tem goes back to the snoozing mode.

6.3 Measurement of transients


Figure 6-2 through Figure 6-5 show the pressure fluctuations in the system between approx-
imately 3:10 p.m. November 20 and 10:00 a.m. December 17 at Van Ness and Sunset. Normal
background pressures, indicated as the blue line, are prevalent throughout the recording period.
In the last week of recording, there are some red recordings that normally would indicate a wa-
ter hammer or pressure surge event. However, these spikes are simply background noise in
the electronics (this is apparent from the single-sidedness of the apparent pulses). These fig-
ures show that the pressure remains between a low of 80 psi and a high of 96 psi, indicating
that at this location, no sudden water hammers or pressure surges worthy of causing the re-
corders to begin recording data every 1/100th of a second occurred during the recording period.
The same is true for the pressure recordings at Van Ness and 3rd, as shown in Figure 6-6
through Figure 6-9. Again, the normal pressure oscillations are within a tight band, 136 psi to
155 psi. These figures show considerably more spikes than those for the other recorder.
Again, upon analysis of these spikes, they were determined to be background noise. By chang-
ing the sensitivity of the recorders between the third and fourth weeks or recordings, much of
this noise was eliminated.

Page | 98
6. Hydraulic Pilot Study

Figure 6-2. Pressure recorder at intersection of Van Ness and Sunset (11/20/2009
11/26/2009).

Figure 6-3. Pressure recorder at intersection of Van Ness and Sunset (11/27/2009
12/3/2009).

Page | 99
6. Hydraulic Pilot Study

Figure 6-4. Pressure recorder at intersection of Van Ness and Sunset (12/4/2009
12/10/2009).

Figure 6-5. Pressure recorder at intersection of Van Ness and Sunset (12/11/2009
12/17/2009).

Page | 100
6. Hydraulic Pilot Study

Figure 6-6. Pressure recorder at intersection of Van Ness and 3rd (11/20/2009
11/26/2009).

Figure 6-7. Pressure recorder at intersection of Van Ness and 3rd (11/27/2009
12/3/2009).

Page | 101
6. Hydraulic Pilot Study

Figure 6-8. Pressure recorder at intersection of Van Ness and 3rd (12/4/2009
12/10/2009).

Figure 6-9. Pressure recorder at intersection of Van Ness and 3rd (12/11/2009
12/17/2009).

Page | 102
6. Hydraulic Pilot Study
Even though the recorders did not pick up any large magnitude water hammers or pressure
surges, it does not mean that they do not occur in the pressure zone. It simply means they
have not occurred during the monitoring period. In fact, a 6 inch diameter pipeline break did
occur within this pressure zone in the early morning of November 28 in the 300 block of North
Spaulding Avenue. Figure 6-1 shows the approximate location of the break. By zooming in on
Figure 6-3 and Figure 6-7 to between 2:00 a.m. and 5:00 a.m., as depicted in Figure 6-10 and
Figure 6-11, respectively, it can be seen that there is a fairly sudden drop in the pressure at the
pressure recorders around 2:56 a.m. This would be the pressure drop associated with the sud-
den rupturing of the pipe on Spaulding Avenue. Particularly evident in Figure 6-10 is that the
pressure stays lower until around 4:15 a.m., when the pressure fairly suddenly recovers back to
the pre-rupture pressure.

Figure 6-10. Pressure recorder at intersection of Van Ness and Sunset. Approximate time
of pipeline break in Hancock Park.

Page | 103
6. Hydraulic Pilot Study

Figure 6-11. Pressure recorder at intersection of Van Ness and 3rd. Approximate time of
pipeline break in Hancock Park is noted on the figure.

Although the magnitudes of the pressure changes associated with this pipeline rupture are rela-
tively minor, it needs to be pointed out that the recorders are located more than 2.5 miles from
the rupture location and, as shown in Figure 6-1, there are a multitude of pipes between the
break location and the recorders, each of which will help to diminish the impact of the rupture
the farther from the rupture one moves. Therefore, if a recorder were to be located in the im-
mediate vicinity of a rupture, one would expect to see a much larger magnitude pressure drop
and subsequent pressure fluctuations.

Something that is not evident in either Figure 6-3 and Figure 6-7, or Figure 6-10 and Figure
6-11, is any kind of sudden pressure increase immediately prior to the drop in pressure that in-
dicates the pipeline rupture. This does not mean there was not a pressure spike prior to failure.
As was previously explained, the pressure recorders are quite distant from the failure location.
However, one would expect to see something, even a pulse of a few psi, such as the pressure
drop recorded for the failure, at the recorder locations. Possible explanations range from:

a) a leak already there that finally blew out

b) a pipe failure due to corrosion

c) a sudden temperature change caused a stress fracture

These are just a few possible failure mechanisms. However, it does not appear that there was
any sudden increase in pressure prior to the rupture that would indicate that a pressure tran-
sient event was responsible for the failure.

Page | 104
6. Hydraulic Pilot Study
Although from a transient perspective these pressure recordings do not show anything overly
significant happening within the system, the data collected does allow for the determination of
the normal pressures within the system at the time of year the recordings were made. This
information can be compared to pressure recordings elsewhere in the system during other parts
of the year to determine if any changes have been made to the operation of the zone. See-
mingly small changes in static pressure from summer to winter or larger fluctuations in pressure
over the course of the day could cause cyclical stresses on piping that may already be at risk
(due to corrosion, or overpressure). These cyclical stresses, when applied over the 100-year
life of the piping would result in more than 35,000 stress cycles, which in cast iron pipe could
result in crack propagation in the castings.

It can also be seen that the use of pressure recorders capable of recording pressures at a very
small time step can be used to determine detailed timings of breaks or unusual operations with-
in systems. These pressure recorders will be particularly useful in pumped zones where the
everyday operation of pumps will create pressure surges in a system.

Page | 105
6. Hydraulic Pilot Study

6.4 Additional Monitoring

Figure 6-12. Monitored pump stations.

As a follow up to the pilot study of monitoring transients in the 579 zone, an additional monitor-
ing program at 5 pump stations was implemented. This monitoring program required the instal-
lation of 2 pressure recorders at each pump station, 1 on the suction side and 1 on the dis-
charge side. The 2 recorders that were installed at the pressure regulating stations at Van Ness
and Sunset and at Van Ness and 3rd were relocated to the suction and discharge sides of the
Roxbury Pump Station at the western edge of the 579 zone. Another 8 pressure recorders were
installed at pump stations throughout the southern San Fernando Valley. The pump stations
selected for monitoring are Winnetka, Encino, Beverly Glen, and Coldwater Canyon. Figure
6-12 shows a general layout of the location of these five pump stations. All the push-pin sym-
bols indicate pump stations within this portion of LADWPs service area. The 5 pump stations
that were monitored are called out with red stars.

Page | 106
6. Hydraulic Pilot Study
Each pump station is supplied flow from a lower zone and boosts this flow to a higher zone.
Table 6-1 shows the pump stations with their supply and delivery pressure zones.

Table 6-1. Pump Station supply and delivery zones.

Pump Station Supply Zone Delivery Zone

Winnetka 1123 1240

Encino 1134 1240

Beverly Glen 1134 1424

Coldwater 1000 1424

Canyon

Roxbury 579 1096

Every time a pump station starts or stops, it creates very distinct and measurable pressure tran-
sients. By installing recorders at pump stations, it is more likely that measurable pressure tran-
sients will be recorded. As to whether or not these transients are of sufficient magnitude to re-
sult in ruptures within the system will depend on the:

original pressure class of the pipe

condition of the pipe

current allowable pressure of the pipe

The purpose behind installing the additional recorders at the pump stations was to perform a
kind of mini audit of LADWPs system. By setting up the recorders on both the suction and
discharge sides of the pump stations, it would be possible not only to see the pressure surges
created by the normal or unplanned operation of the pump stations, but also to detect any pres-
sure surges that may be created elsewhere in the system. Any mystery pressure surges
would propagate throughout the pressure zones, eventually being picked up by the recorders at
the pump stations.

The recorders were installed on January 5 and 6. Due to minor programming errors, data for all
the recorders did not start being collected until January 22. The initial intention of the monitoring
program was to record pressure fluctuations for approximately 3 months to see if any events,
besides the normal starting and stopping of the pumps. However, the recordings were ended
after approximately 4 weeks.

Page | 107
6. Hydraulic Pilot Study
Upon retrieval of the data, it was noticed that the recorder on the suction side of the Roxbury
Pump Station was not operating correctly. And although replacement instrumentation was ob-
tained there was insufficient time to install it before the recording program was halted. Also,
during data retrieval, it was brought to attention that Winnetka did not operate at this time of the
year. This means that any pressure surges recorded on either the suction or the discharge
sides of the Winnetka Pump Station were all created somewhere else in either the 1123 zone
(suction side) or the 1240 zone (discharge side). However, Figure 6-13 and Figure 6-14 below
show, the pump station did appear to start in the early morning of February 18.

During this time, numerous pump starts and stops were recorded. Figure 6-13 through Figure
6-21 show the recordings made at each location, both suction and discharge (discharge only for
Roxbury), during the time period from 12:00:00 a.m. on February 14 to 23:59:59 on February
20.

Figure 6-13. Pressure recorder on the suction side of Winnetka pump station.

Page | 108
6. Hydraulic Pilot Study

Figure 6-14. Pressure recorder on the discharge side of Winnetka pump station.

Figure 6-15. Pressure recorder on the suction side of Encino pump station.

Page | 109
6. Hydraulic Pilot Study

Figure 6-16. Pressure recorder on the discharge side of Encino pump station.

Figure 6-17. Pressure recorder on the suction side of Beverly Glen pump station.

Page | 110
6. Hydraulic Pilot Study

Figure 6-18. Pressure recorder on the discharge side of Beverly Glen pump station.

Figure 6-19. Pressure recorder on the suction side of Coldwater Canyon pump station.

Page | 111
6. Hydraulic Pilot Study

Figure 6-20. Pressure recorder on the discharge side of Coldwater Canyon pump station.

Figure 6-21. Pressure recorder on the suction side of Roxbury pump station.

Page | 112
6. Hydraulic Pilot Study
It is not evident from the data gathered that any unplanned emergency shutdowns or power fail-
ure events occurred. What can be seen from these figures is that the recorded maximum pres-
sures (pump start up), minimum pressures (pump shut down), and steady state pressures
(pumps either operating or turned off) generally remained within tight bands for all the recorders.
However, there are a few items of interest that show how useful the recordings of the pressure
fluctuations can be.

In some of these figures, particularly Figure 6-16 and Figure 6-21, there are numerous red
transient recordings that occurred during this time period that can be directly tied to a specific
event. For Encino (Figure 6-16), they appear to be almost exclusively associated with the
planned stopping of the pumps. A blowup of the pump shutdown occurring prior to midnight on
February 17th is shown in Figure 6-22, which is evident by the gradual increase in pressure im-
mediately prior to one of these transient events occurring and then the reduction in pressure
immediately following one of these events. Although these transients occur very suddenly, thus
the red plotting color, their magnitude is small, on the order of less than 20 psi deviation from
the pumping pressure. Since there are no red lines associated with the starting of the pumps,
this indicates that the starting of the pumps results in a more gradual increase in pressure.

Figure 6-22. Pressure recorder on the discharge side of Encino pump station.

In Figure 6-21, it is interesting to note that starting in the morning of February 15 and ending on
the afternoon of the 18th, both pumps at the pump station were used. This can be seen in Fig-
ure 6-23, which shows the stepping up and down of the pressure during this time period after
each red transient line. This actually corresponds to a mini-heat wave in Southern California
when temperatures approached or exceeded 80 degrees. Prior to and after this time period,
only one pump was operated, which corresponded to times of cooler, rainy weather.

Page | 113
6. Hydraulic Pilot Study

Figure 6-23. Pressure recorder on the discharge side of Roxbury pump station on
2/17/2010.

Also of particular note in the suction side pressure records (shown in Figure 6-13, Figure 6-15,
Figure 6-17 and Figure 6-19) is the reduction in pressure that occurs beginning around 5:00 am
on February 18th and which continues to drop until just before 8:00 am on the 18th. While this
reduction is evident in all four of these figures, it is particularly pronounced in Figure 6-15 and
Figure 6-17. Both the Encino and Beverly Glen pump stations are supplied flow from the same
1134 pressure zone and as such would see very similar pressure fluctuations, as shown in
these two figures. This drop in pressure within the San Fernando Valley portion of the LADWP
service could be indicative of a sudden large demand, such as a major fire or more likely anoth-
er pipeline rupture.

Overall, the pressure monitoring did exactly as was anticipated. It allowed monitoring of the
pressure changes at locations within the system at a fine enough time scale that would allow
any pressure transients generated within the system to be picked up and recorded. However,
during the time of monitoring no out-of-the-ordinary pressure transients appeared to have oc-
curred.

It should be noted that the failure of this monitoring program to record any severe pressure tran-
sients does not mean that adverse pressure transients or pressure surges do not occur in the
LADWP system. It simply means that none occurred during the monitoring period. Further-
more, it does not mean that the pressure transients created as a part of normal operations will
not adversely affect the system. Factors such as the condition of the pipe, the pipe material, the

Page | 114
6. Hydraulic Pilot Study
range of pressures within the pipes, all could contribute to the failure of a pipeline and the range
of pressures occurring every day may be sufficient to contribute to their eventual failure.

Another unfortunate issue with the monitoring program was its timing. The months from approx-
imately November through April fall within the rainy season in Los Angeles when demand is
lowest and the lowest range in everyday pressures can be expected. Greater variations in the
pressure are more likely to occur during the higher use season, which could increase the fre-
quency of failure during summer months. An extension of the monitoring program to the high
demand season may be more revealing.

Page | 115
7 REMOTE SENSING PILOT STUDY3

The Jet Propulsion Laboratory (JPL) analyzed crustal deformation data for the Los Angeles ba-
sin and San Fernando Valley areas to determine if there is any evidence for tectonic drivers for
the water pipeline breaks. Examination of the data shows little evidence for any tectonic drives
of the pipeline breaks; however, there is evidence in the crustal deformation data that the
ground water in the possibly shows motions associated with water leaking from the pipes. There
is a suggestion in the data that a large-scale right-lateral creep event may have occurred rough-
ly along the northern extent of the Newport-Inglewood fault in late summer 2009, during the
timeframe of the increase in pipeline breaks and blowouts. This could be due to stress on the
fault being relieved due to ground water changes (Hamilton and Meehan, 1971).

7.1 Background
To measure crustal deformation, JPL uses spaceborne data in the form of GPS time series and
velocities and Interferometric Synthetic Aperture Radar (InSAR). The primary goal is to improve
understanding of geophysical processes. In Los Angeles, the goal is to understand which faults
are active and, therefore, to improve earthquake hazard assessments in the region. The mea-
surements reveal crustal deformation of the region, which can be due to tectonic signals or
changes in ground water in the region.

JPL used the QuakeSim computational environment to examine and model various data that
might provide insight into the LA area pipe breaks. QuakeSim is a cooperative effort for study-
ing, modeling, and forecasting earthquake faults from a system perspective. The focus of Qua-
keSim is on the interseismic earthquake cycle of strain accumulation and release. It is used to
study earthquake processes through the integration of multiple data types and models. The key
data sources used by QuakeSim are spaceborne observations from GPS navigation and radar
satellites, earthquake locations from seismicity, and fault data. QuakeSim can be used to look
for anomalies related to both crustal deformation time series data and seismicity. While the re-

3
This section of the report is co-authored by Andrea Donnellan, Robert Granat, Paul Lundgren,
Jay Parker, Greg Lyzenga, and Frank Webb from the Jet Propulsion Laboratory, California Insti-
tute of Technology.

Page | 116
7. Remote Sensing Pilot Study
sults suggest that there may be some anomalous tectonic activity, JPL finds no indication that
this correlates with the pipe breaks. Any observed changes in time series appear to be an effect
of water changes, rather than a cause of the breaks.

7.2 Pipeline break data


JPL used the LADWP pipe-break data provided by the University of Southern California. Plots of
cumulative pipe breaks are analyzed as are data binned by month. The cumulative total pipeline
breaks (Figure 7-2) and blowouts (Figure 7-3) are plotted by district and compared to GPS time
series. There are no events in the GPS time series that appear to correlate with an increase or
decrease in leaks and blowouts by district.

Page | 117
7. Rem
mote Se
ensing Pilot
P Stud
dy

Figure 7-1. Area serrviced by th


he LADWP. Regions and GPS sta
ations prese
ented in this
s
s are marked
analysis d.

Page | 11
18
7. Remote Sensing Pilot Study

Figure 7-2. Cumulative total leaks including blowouts for separate regions within the
LADP water district.

Figure 7-3. Cumulative blowouts for regions within the LADWP jurisdiction.

Page | 119
7. Rem
mote Se
ensing Pilot
P Stud
dy
JPL plottted, interdep
pendently, th
he leak data for the entirre water disttrict separate
ed by leaks,, total
leaks andd blowouts (Figure
( 7-4), and blowou uts (Figure 7-5).
7 The datta show a large spike arround
the 2006
6-2007 year change, an increase arround summ mer 2004 and d another sp pike in leakss late
summer 2009 (which h prompted this
t study).

Figure 7-4. Total lea


aks includin
ng blowouts
s binned by
y month forr July 1, 200
01 through Sep-
S
tember 30,
3 2009. Red R indicates leaks andd blue indicates blowouts.

Figure 7-5. Total nu


umber of blo
owouts plottted by mou
uth for the period
p July 2001 end
d of
Septembber 2009.

Page | 12
20
7. Rem
mote Se
ensing Pilot
P Stud
dy
The resuults roughly mimic
m the tottal leaks plus blowout da
ata, showingg a spike in the
t late 20066
early 200
07 timeframe e, the end off summer 20 009, and in July
J 2004. There is also an annual signal
s
that manifests itself in the data.

7.3 SC
CIGN GPS
S time se
eries
In the latte 1990s NA ASA, the U.S S. Geologicaal Survey, thhe National Science
S undation, and the
Fou
Keck Fou undation established a continuously operating,, high-precission GPS ne etwork called the
Southern n California Integrated GPS
G Networkk (SCIGN). The
T networkk was established to mea asure
crustal deformation and a to discrriminate mottions betwee en faults in order
o to imp
prove earthq quake
hazard estimates
e in southern California.
C Thhe network of o approxim mately 250 stations prod duces
daily possitions of eacch station in the networkk (Figure 7-6 6 and Figure e 7-7). The resulting
r time se-
ries can be used to identify loca al and regionnal changess in deforma ation and to produce velocity
maps forr the purpos se of inferrin
ng characterristics of faults and activvity. In addittion to identifying
tectonic activity
a the network
n show ws a substantial amountt of ground motion
m assoociated with water
w
changes in the aquife ers.

Figure 7-6. Location


ns of the coontinuous GPS
G stations in the Los
s Angeles region.
r Stattions
considered in this study
s are sh
hown in Fig
gure 7-1.

Page | 12
21
7. Remote Sensing Pilot Study

Figure 7-7. Typical high-precision continuous GPS station. The pipes are drilled 10 me-
ters into the Earth. A ray dome protects the choke ring antenna, which is designed to mi-
nimize noise and ensure a clean radio signal from the GPS satellites.

JPL sees no strong correlation between the GPS time series data, but some features do stand
out in the data (Figure 7-8 and Figure 7-9). Stations LEEP and DSHS show an offset in the time
series in the 2007-2008 year boundary, consistent with the increased blowouts during that time
frame. The increase in blowouts in late summer 2009 is consistent with changes in the north
component for most of the time series. Interestingly, the stations to the west show a northward
motion and the stations to the east show a southward motion, while the stations in the middle
show little north or south motion.

The results need verification, but there is some possibility that a right-lateral creep event oc-
curred across the Los Angeles and San Fernando regions. On May 17, 2009 a magnitude 4.7
earthquake occurred that indicated right-lateral slip near the Newport-Inglewood fault
(http://earthquake.usgs.gov/earthquakes/recenteqsww/Quakes/ci10410337.php). On August 11,
2009, an earthquake occurred in La Crescenta, along the San Gabriel Mountains north of Los
Angeles. This earthquake, while small, also had a right-lateral mechanism and was located at a
depth of about 5 km (Lucy Jones, USGS, personal communication).

It is likely that the water main breaks are most highly correlated with deeper aquifer withdrawal,
which typically peaks in December, before substantial rainfall has occurred. In a study on InSAR
time series analysis in the LA area (Lanari et al., 2004) found that for the Santa Ana basin the
peak recharge time was in March and the maximum subsidence was approximately in Novem-
ber. It may be possible that there is a phase lag between the crustal deformation changes and
actual breaks or detection of leaks. The surface deformation is likely hydrologically induced. The
watering appears to not make a difference in impacts on crustal deformation. It is possible that
stress on the Newport-Inglewood fault or its extension was relieved in the unconsolidated upper

Page | 122
7. Rem
mote Se
ensing Pilot
P Stud
dy
sedimentts due to ground water changes ass observed in the Baldwwin Hills in 1963 from flu
uid in
ejection (Hamilton and
a Meehan, 1971). Mo ore analysis is required to understaand any mo otions
that may have occurred in late su
ummer 2009
9.

Figure 7-8. GPS tim ns distributed across the


me series forr six station t water diistrict. Top time
series in
n blue is north compon nent, middle e green timee series is east
e compo
onent, and bot-
b
tom timee series in yellow
y is ve
ertical or up component

Page | 12
23
7. Rem
mote Se
ensing Pilot
P Stud
dy

Figure 7-9. GPS timme series forr six station


ns distributed across the
t water diistrict plotteed in
a larger format. The e time perio
od is consisstent with th
he water pip
pe data, whiich range frrom
July 200
01 October 2009. Top p time seriess in blue sh
hows north componentt, middle grreen
time serries is east component
c t, and bottom time series in yelloww is verticall or up co
om-
ponent

JPL exammined the summer


s 20009 vertical tiime series at
a Fairfax High
H School (FXHS) in detail
d
(Figure 7-10).
7 There
e is some suggestion
s o offsets in
of n the data associated
a w
with the pip
peline
breaks. However,
H so
ome off the offsetssuc
o h as those occurring
o on
n July 9, 200
09occur prrior to

Page | 12
24
7. Rem
mote Se
ensing Pilot
P Stud
dy
the Auguust 10, 2009 blowout. It is possible that pipes tyypically leak before theyy completely blow
out. Furtther analysiss of this and
d other GPS S time serie ed to verify whether the
es is require ere is
ground motion
m assocciated with leaking
l pipe
es associate
ed with blowouts. While time series from
GPS stattions could be used in some
s casess to search for
f local ground water changes,
c the
e net-
work is not
n dense en nough to be used for rouutine assesssment of pipee breaks. Ho owever, loca
al ob-
served time series changes thatt occur in asssociation with pipe brea aks might provide insigh ht into
the behaavior of the le
eaking pipess and blowouuts.

Figure 7-10. Map vie ew of Fairfa


ax High Schhool, located south of the
t Hollywo ood Hills. Red
R
dates inddicate blow
wouts and blue dates in ndicate pipeeline breaks
s. Locationns are marked
by red trriangles or dots
d respecctively. FXH
HS GPS stattion is indic
cated by the
e yellow box x
and the time series for the perriod May 19, 2009 throuugh November 3, 2009. Blue vertic cal
lines ind
dicate clear offsets in the
t time serries, with th
he dates notted. The red
d vertical lin
nes
show the e times of the pipeline breaks andd blowouts

A view of
o the longerr-term times series sugg gests some correlation with the pip peline break data
(Figure 7-11).
7 FXHS time seriess data show an abrupt changec in the
e east and north
n trend in the
data in th
he 2004.5 timeframe. Th he data werre particularlly noisy duriing 200320004, which was
w a
year of heavy
h rainfall. The offse
ets are fairly large for Ju
uly 16 and August
A 9, 20
009, and reaach 1
cm of verrtical motion
n (Table 7-1)).

Page | 12
25
7. Rem
mote Se
ensing Pilot
P Stud
dy

Figure 7-11. Long-te


erm time se
eries for FX
XHS. Time series
s are fo
or north, ea
ast, and up from
f
top to bo
ottom

Table 7-1. Offsets fo


or each com
mponent of motion for FXHS.

July 16, 2009 offset Augustt 9, 2009 offfset


2 mm so
outh 1 mm north
4 mm ea
ast 2 mm west
w
10 mm down
d 10 mm up

7.4 Co
ovariance
e descrip
ptor based
d analysiis of trans
sient defformation
n
Covariannce descriptor based de etection me ethods are relatively
r reccent technoologies that have
seen connsiderable success
s in th
he field of computer
c e methods are based arround
vission. These
the calcu
ulation of a generalized
g distance me etric for covaariance matrrices (Tuzel 2008), de
efined
as

Page | 12
26
7. Rem
mote Se
ensing Pilot
P Stud
dy

where j is the generalized eig


genvalue fo
or the generalized eige
envector xj defined by

The distaance is so ometimes alsso referred to as the co ovariance diivergence. In n image ana alysis
applicatio
ons, typicallyy a set of filters is appllied to an immage; the ou utputs of theese filters foorm a
feature vector
v for ea
ach pixel. Thhe covariancce matrices of o the featurre vectors foor different im
mage
regions are
a calculate ed (these arre termed th he covariancce descriptors for the re egions), and then
compare ed using the distance metric
m above. The generralized covariance dista ance betwee en re-
gions tha
at resemble one another (in the cho osen feature e space) will be small, while
w the disttance
between regions that are dissimilar will be re elatively larg
ge. In this manner anom malies or mattches
between regions can n be identifie
ed. Of particular note iss the fact that even regions of dissiimilar
size can be compare ed using thiss method, as only their statistics arre compared d, and their cova-
c
riance deescriptors lie
e in the samme dimension nal space. Furthermore e, it is obserrved that misssing
data are not an issue for this ap pproach, as long as suffficient data are a available e in the region to
calculate
e the necessary statisticss.

This metthod can be readily exte ended to GP PS time seriees analysis. In the time series case e, the
regions are
a simply time window ws, while the
e feature vecctors are the e time series measurem ments
themselvves (displaceements in three dimensions), as well as any derived feature es we might wish
to calcula
ate, such ass derivativess, Fourier fe eatures, and so forth. Transient defformation sig gnals
should be anomalous, which is to t say, they should not resemble
r the
e rest of the
e time seriess as a
whole. This
T implies that it is use
eful to calculate the covaariance desccriptor of the
e time seriess as a
whole annd compare it with covariance desccriptors of loccalized time e windows; descriptors
d f lo-
for
calized windows
w thatt are far from
m the descripptor for the entire
e time se
eries are like
ely anomaloous.

JPL applied this approach to th he Phase III, Group A data sets frrom the Southern Califfornia
Earthqua ake Center (SCEC)
( tran
nsient detecction exercisse, calculatin ng for each GPS station the
divergence of a sliding 100 day window with h the covaria ance of the entire time series.
s (Windows
of 10, 500, 200, 300, and 500 dayys were also o tried; smalller windowss generated results that were
too noisyy, while large
er windows made
m little difference
d asside from inccreasing com
mputational time).
t
This app proach forme ed a covariance diverge ence time se eries for each station. A plot of the diver-
d
gence timme series fo or data set 1 (2009.167.001253) for all stations can be seen in Figure 7-12.
The horizzontal striatiions visible in the image e are the ressult of the se easonal sign
nal (not rem
moved
in this an
nalysis).

Page | 12
27
7. Remote Sensing Pilot Study

Figure 7-12. Divergence time series for all stations for 2009.167.001253. The vertical axis
displays time (in days), while the horizontal axis displays stations (in alphabetical order
by station name)

To identify regions of potentially anomalous activity, the average divergence is calculated over
all time for each time series, and then the top twenty most divergent stations are singled out.
Presuming that there was at most a single anomalous region for each data set, and that tran-
sient deformation would occur in relatively small geographical regions, outlier stations are fil-
tered from our candidate group by first fitting a normal distribution to the station locations, and
then by rejecting all stations with a Mahalanobis distance greater than two from the distribution.
The locations of the remaining most divergent stations for each data set are shown in Figure
7-13.

Page | 128
7. Remote Sensing Pilot Study

2009.167.001253 2009.167.002227

2009.167.002307 2009.167.003841

2009.167.003923 2009.167.010157

Figure 7-13. The most anomalous GPS stations as determined by average covariance di-
vergence for each of the Phase II, Group A data sets

Page | 129
7. Remote Sensing Pilot Study
It is clear that strong and significant clusters of anomalous stations are found in data sets 5 and
6 (2009.167.003923 and 2009.167.010157). From visual inspection, it also appears that there
may be clusters of anomalous stations in data sets 1, 2, and possibly even 4 (2009.167.001253,
2009.167.002227, and 2009.167.003841). To determine if these clusters were significant, the
probability density function (pdf) of the spatial distribution of the entire network is first estimated
using a kernel density estimation method (Botev 2009), and then the same method is used to
estimate the pdfs of the spatial distribution of anomalous stations for each data set. The Kull-
back-Liebler (KL) divergence between the pdf of the whole network and the pdf for each group
of candidate stations would then indicate the degree of similarity between the distributions: the
less similar the distributions, the more significant the anomaly. This exercise quantifies the sig-
nificance of these clusters; a Monte Carlo experiment is performed in which JPL repeatedly se-
lected 20 station locations at random, filtered the locations according to the Mahalanobis dis-
tance criterion described above, estimated the pdf for the remaining stations, and calculated the
KL divergence to the pdf of the entire network. Over 1000 Monte Carlo tests, JPL discovered
that groups of 20 stations chosen at random had KL divergences to the network well described
by a normal distribution with mean 0.668 and standard deviation 0.296. In contrast, the clusters
of stations found using the covariance descriptor method had KL divergences (Table 7-2).

Table 7-2. Clusters of stations found using the covariance descriptor method.

Data Set Kullback-Liebler Divergence Significance


2009.167.001253 1.523 > 2
2009.167.002227 1.556 3
2009.167.002307 1.146 >
2009.167.003841 0.814 <
2009.167.003923 3.373 > 9
2009.167.010157 2.772 > 7

From this we can see that the clusters found for data sets 1, 2, 5, and 6 can be considered sig-
nificant. By fitting a normal distribution to these clusters, we can find means and covariances
that describe the locations of the transient signals (Table 7-3). The bulk of the most divergent
stations fall along or in a line north of the Newport-Inglewood fault, again suggesting some type
of activity along this shear zone. As stated earlier, a creep event may have occurred, but it is
likely that the event occurred in the shallow upper unconsolidated sediments. Further analysis
will better define the location and depth of any possible creep that occurred.

Page | 130
7. Rem
mote Se
ensing Pilot
P Stud
dy
3. Means an
Table 7-3 nd covarian
nces that de
escribe the locations of
o the transient signals
s

Data Set
S Location
n Mean Location
n Covariance
2009.1
167.001253 34.1312 -118.0014 0.0938 0.0200
0.0200 0.0328
2009.1
167.002227 34.0458 -118.3348 0.0547 -0.0567
-
-0.0567
7 0.2993
2009.1
167.003923 34.1429 -118.2279 0.0020 0.0005
0.0005 0.0037
2009.1
167.010157 34.0184 -118.2995 0.0053 -0.0002
-
-0.0002 0.0056

Figure 7-14. Top 20


0 most diverrgent GPS stations
s forr the time pe
eriod Novem
mber 2008

Novemb ber 2009

7.5 InS
SAR data
a
As statedd above, JP PL also usess InSAR to monitor cru ustal deform
mation data is
i Interferommetric
Syntheticc Aperture Radar
R (InSARR). This techhnique images motions of the surfacce of the Eaarth in
the line-o
of-site directtion of the sa
atellite. Unlike the GPS network thee repeat inte
erval of the satel-
s

Page | 13
31
7. Remote Sensing Pilot Study
lite is infrequent, but the images are continuous spatially where the data do not decorrelate.
The InSAR data also show significant surface deformation associated with subsurface motions
of fluids. The observations show that the faults in the Los Angeles basin tend to provide boun-
daries along which the bulk of the deformation occurs. InSAR data can be collected by space-
borne or airborne platforms. The United States does not currently have an InSAR satellite, but
European and Japanese spacecraft have collected data infrequently over southern California.

NASA recently put into operation an airborne InSAR platform called Uninhabited Aerial Vehicle
Synthetic Aperture Radar (UAVSAR), and data were first collected over Los Angeles in spring of
2009. Data were collected again in September of 2009, and UAVSAR is again and being flown
in the spring of 2010. Unfortunately the motions are too small to detect anything over that short
of a time frame with UAVSAR. The UAVSAR data will continue to be collected and will form a
composite data set that can be analyzed for both tectonic and hydrologically induced motions.

Analysis of a long time series of radar data indicates that a substantial amount of surface de-
formation occurs in Los Angeles due to discharge, recharge, and migration of fluids in the Los
Angeles basin (Figure 7-15). Sharp gradients occur across the faults indicating that the faults
bound the motions. Faults are typically weaker than the surrounding material and also can have
different rock types with different strengths on either side of the fault. This makes a natural
boundary along which motions are likely to occur. For the most part, these motions are not tec-
tonic.

The bulk of the pipeline breaks occur in the area to the south of the Hollywood Hills (Figure
7-16Figure 7-19). The InSAR data show consistent patterns of deformation in the same region.
It is not clear whether these motions reflect ground water changes from leaking pipes or simply
from changes in the ground water table. Further investigation is required to better understand
the correlation between the surface deformation data and the concentrations of pipeline breaks.

Page | 132
7. Rem
mote Se
ensing Pilot
P Stud
dy

mage from a movie of deformation in California as obserrved by the Eu-


Figure 7-15. Still im
ropean radar
r satelliite. The mottion is clearly bounded
d by the Ne
ewport-Ingle
ewood faultt (Da-
ta provid
ded by Paull Lundgren,, animation created by y Vince Reallmuto, JPL)).

Figure 7-16. Interferrogram fromm the Canaddian RADARSAT for th he time periiod August 24,
2002 September
S 30,
3 2004. Diffferent colo
ors indicate different amounts of motion
m in th
he
direction
n of the sate
ellite. The change
c in co ence in motion across a
olor indicattes a differe
region.

Page | 13
33
7. Rem
mote Se
ensing Pilot
P Stud
dy

Figure 7-17. Pipelinne break datta for the sa


ame time peeriod as the
e interferogram in Figuure
7-16. Thee data sugggest a concentration off breaks south of the Hollywood
H H
Hills, consis
stent
with the substantial deformatio on in that area.
a There is
i a weakerr suggestion n of increas
sed
breaks along
a the so
outhern Sann Fernando Valley

Figure 7-18. Interferrogram for the time pe eriod Septem


mber 12, 20 003 Septemmber 30, 20 004.
Differentt colors ind
dicate different amountts of motionn in the dire
ection of the
e satellite. The
T
change in color ind dicates a diffference in motion
m acro
oss a region n

Page | 13
34
7. Rem
mote Se
ensing Pilot
P Stud
dy

ne break datta for the sa


Figure 7-19. Pipelin ame time pe
eriod as tho
ose shown in
i Figure 7--18.

7.6 Su
ummary
Analysis of crustal deeformation data
d suggessts that there
e is a weak correlation
c b
between cha
anges
ency of pipeline breaks and
in freque a blowoutts and chang ges in GPS time series. There is a weak
w
suggestioon that a rig
ght-lateral slip even mayy have occurrred in the la
ate summer 2009 timefrrame.
onsistent with a linear extent
This is co e of GP
PS stations ono and to thee northwest projection ofo the
Newport--Inglewood fault;
f those stations sho ow divergen nce from norrmal background time series
s
characteristics in thee entire 20009-year timeeframe. Furtther analysiss is requiredd to determine if
there is in
i fact a geoophysical co omponent off motion thatt associatess with an inccrease in pippeline
breaks or whether a creep eventt occurred on n the Newpo ort-Inglewoo
od fault in the
e region.

There is evidence th hat ground-wwater motion ns influence e surface deformation, which


w can bee ob-
served with
w both GP PS and InSA AR. It is nott clear what componentt of the defo ormation is asso-
a
ciated wiith leaking pipes
p and wh hat compone ent is assocciated with annual
a aquife
er discharge e and
rechargee. Analyses ofo data sugggest that we etter years re esult in high
her deformattion rates an
nd an
increase of pipeline breaks. How wever, it is liikely that the ot a result of tec-
e pipeline brreaks are no
tonic defformation of the surfacee of the Eartth, but rathe er are due to
o pressure or
o other cha anges
associateed with the increased
i ra
ainfall, makinng the surface deformattion changess an effect rather
r
than a ca
ause.

Page | 13
35
8 ASSET MANAGEMENT4

In the wake of water pipeline breaks during the summer of 2009, LADWP was criticized for not
having in place an efficient asset management program, which would have ideally replaced wa-
ter pipes before they break. This section reviews the best asset management practices in the
industry and reveals the complex dimensionalities of asset management which involves not only
technical aspects but social and financial considerations. This section makes general recom-
mendations as the authors did not review in details the LADWP asset management program.
The recommendations emphasize a modern integrated risk assessment approach to improve
urban infrastructures.

8.1 Best practices from AWWARF


A number of utilities have developed models for their own prioritization and capital planning. The
LADWP methodology was presented by Mavrakis (2003). LADWP uses an A-F rating scale,
with age-related deterioration scores. For pipelines, the maximum service life was assumed to
vary from 70 to 210 years, depending on pipe material and soil corrosion index. Three soil types
were considered: severely (30%), moderately (50%), and mildly (20%) corrosive. Pipeline
grades were determined separately for steel and cast iron/ductile iron. Trunk lines were eva-
luated, and their lengths were taken from the trunk line hydraulic model. Some 743 miles of
trunk line (16 inches and larger) were evaluated on an overall basis at "B", and with a unit cost
of $13 per inch of diameter per foot. Of the lines, some 70 miles were rated D, F, or critical. The
present value of replacement cost of all trunk lines was $1.67 billion. This analysis yielded an
estimate of $54.5 million needed now for replacement of critical mains and $10.1 million needed
annually to keep up with critical replacements over the next 20 years. Considering all assets
including tanks, reservoirs, treatment plants, and trunk lines a cash flow replacement sche-
dule similar to the Nessie Curve was produced. Total average replacement needs are $16.2 mil-
lion per year.

4
This section of the report is co-authored by Richard Little, USC Keston Institute, and Donald
Ballantyne, MMI, Seattle.

Page | 136
8. Asset Management
Hope (2003) developed an approach for the prioritization of water main replacement based on
assigning a priority action number (PAN) to each main based on agreed-to service level goals.
Several service level criteria were established including pipe age, leak history, water quality,
criticality, hydraulic performance, corrosion, material, system pressure, and location. Once
PANs are determined for each category, categories are weighted according to importance. After
the numbers are assigned, a prioritization model called OWASA (Orange Water and Sewer Au-
thority, Carrboro, NC, and University of North Carolina, Chapel Hill) is created. In it, a variety of
data sets are integrated into one geo-database, such as operational work orders, a sewer-
system hydrologic model, customer service billing, and a water-system hydraulic model. From
this integrated database, a water-main prioritization model can be developed, along with an
overall utility capital improvement plan. With all of these data integrated, sections of pipe are
designated as highest priority for replacement.

8.2 Review of asset management practices at LADWP


In 2004, LADWP began the development of an Asset Management Program to address the
long-term sustainability of its major facilities and infrastructure. The program involves refining
operations and maintenance practices and procedures for each asset type, evaluating the con-
dition and expected life of each asset, and developing a strategy for scheduling the rehabilita-
tion and replacement of these assets. This section of the report briefly describes infrastructure
asset management within the context of the LADWP system and offers some observations for
refining the criteria on which LADWP asset management decisions are made.

Infrastructure asset management is a process by which agencies monitor and maintain facilities
systems with the objective of providing the best possible service to the users, within the con-
straints of available resources (Ben-Akiva et al., 1993). Although this objective appears
straightforward, doing it well has proved both elusive and difficult in practice. Because the best
possible service to users means different things to different stakeholder groups, the effective-
ness of funds spent on the maintenance and repair (M&R) aspect of asset management cannot
be readily measured. As a result, the search for an optimal M&R investment strategy (in the
case of water systems, a large part of this consists of the agencys water main replacement
program) remains something of a Holy Grail and rightly so. Each year, water utilities spend bil-
lions of dollars on main replacement programs in an effort to maintain satisfactory performance
levels for these systems, primarily by replacing sections of water main before they develop un-
acceptable leaks or fail in other ways. For example, LADWP has estimated its water main re-
placement needs at approximately $160 million annually in 2010 dollars over the next 40 years.

Public agencies and private corporations alike grapple with the question of how much they
should spend to maintain their infrastructure assets while at the same time, wonder if they are
spending too much. The desire is, of course, to avoid spending more than necessary while, at
the same time, avoiding the excessive frugality that could bring on calamitous outcomes, (e.g.,
the need for major reconstruction, catastrophic failure, serious injury or loss of life). This dilem-
ma is illustrated conceptually in Figure 8-1, where it can be seen that the optimal main replace-
ment strategy will position the vertical line in the decision table so that the risk of both Type I

Page | 137
8. Asset Management
errors (not replacing mains that need to be replaced) and Type II errors (replacing mains that
are functioning acceptably) is balanced within the risk tolerance of the decision-makers either
for an increased number of failures or higher expenditures for replacing mains with useful life
remaining.

The proper course of action


Dont replace
tion
mains Replace mains

Dont replace
mains
Correct Type I Error
Decision
Asset
Decision

No unnecessary poorly -maintained


expenditures risk of failure
Replace mains

Type II Error Correct


Decision
Unnecessary
Asset
maintenance and
well -maintained
wasted resources

Higher Risk of Lower Rick of


Type I Error Type I Error

Figure 8-1. Main replacement strategies should balance risk and cost

In a landmark study to measure infrastructure performance, the National Research Council


found that, Performance should be assessed on the basis of multiple measures chosen to re-
flect community objectives, which may conflictThe specific measures that communities use to
categorize infrastructure performance may often be grouped into three broad categories: effec-
tiveness, reliability, and cost (NRC, 1995). In essence, a successful asset management pro-
gram would be one that maximizes the effectiveness and reliability of the system while minimiz-
ing cost.

8.3 Asset management as a risk management problem


Asset management can also be thought of as a problem of risk management. Risk (R) can be
quantified as the probability (P) of an adverse event multiplied by the consequences (C) of that
event or

R =P C

and defined by three questions (Kaplan and Garrick, 1981):

1. What can go wrong?

2. What is the likelihood that it would go wrong?

Page | 138
8. Asset Management
3. What are the consequences of failure?

In the case of this assessment of LADWP water infrastructure, major leaks and blowouts are
considered the problem of interest and will be used as a surrogate for the broader range of fail-
ure possibilities, or what can go wrong. The likelihood or probability of failure is influenced by a
number of factors, including the age, type, and size of pipe, soil conditions, and operating
stresses to which the pipe is subjected. The consequences of water main failure are many and
include service disruptions to customers, including critical facilities such as hospitals, damage to
streets, adjacent infrastructure systems, and private property; loss of product; inadequate flows
for fire fighting; traffic diversion and delay; and in extreme cases, possible injury and loss of life.
Most of these consequences can be assigned a dollar value, either as the direct cost of making
repairs and paying damages or the indirect costs of lost productivity, delay, or lower attractive-
ness to business and industry.

Risk management integrates the results of risk assessment with other informationsuch as po-
litical, social, economic, and engineering considerations--to arrive at decisions about the need
and methods for risk reduction. Risk management seeks answers to a second set of questions
(Haimes, 1991):

4. What can be done and what options are available?


5. What are the associated trade-offs in terms of all costs, benefits, and risks?
6. What are the impacts of current management decisions on future options?
As described earlier, the traditional approach for dealing with leaks and blowouts is a proactive
program of water main replacement. Using risk assessment principles can help to refine this
process by identifying high-risk mains (i.e., those with a high probability or high consequence of
failure or both). However, repairing mains as they fail may be an economically rational, if politi-
cally unacceptable, approach to the problem as well. Between these two poles lie other options
such as improved response to main breaks to minimize the consequences and reliance on
commercial or self-insurance to cover the cost of damage. Different strategies have different
cost and benefit profiles and are also influenced by the organizations risk tolerance. One con-
sequence of adopting a strategy with a high tolerance for risk is to defer the inevitable cost of
system renewal well into the future. The accumulated costs of this backlog could seriously ham-
per the future financial viability of the system and this is a situation that now confronts many
older water systems that routinely deferred main replacement programs in an effort to reduce
costs to keep rates at acceptable levels. This is illustrated in Table 8-1.

Page | 139
8. Asset Management
Table 8-1. Risk Options.

Two Risk-based Options for Asset Management

Less Aggressive Strategy More Aggressive Strategy

Replace pipe when the cost of repair (including all Replace all cast iron pipe evenly distributed over
consequential costs) exceeds replacement on an the pipes expected life prioritized on failure
annualized basis. The decision point will be highly rates and consequential costs (e.g., replace all
dependent on what is included in consequential pipe over a 100- year cycle). This scenario should
costs. This approach could result in large avoid sudden large expenditures and minimize
expenditure at some point in the future as large the annual failure rates, but it may result in
quantities of pipe reach their useful life and start to replacing pipe that has many years of useful life
fail in a short period of time. This scenario has the remaining. This scenario should minimize societal
potential of having significant societal impacts and impacts. While few United States utilities are
associated political implications due to system taking this approach, the City of Kobe Japan had
disruption. replaced nearly all of their cast iron pipe with
ductile-iron pipe prior to the 1995 Kobe
earthquake for non-earthquake reasons. In
general, the Japanese have a more aggressive
view of pipe replacement.

8.4 The LADWP Asset Management Program


The LADWP utilizes a multi-factor service life model to guide its main replacement program.
The model combines the size and material composition of the pipe, soil corrosivity, traffic loads,
and operating pressures and follows the general guidelines promulgated by the Water Research
Foundation. The mains in the LADWP system consist of a great deal of cast iron (69%) and
lesser amounts of ductile iron (11%), steel (10.3%), and asbestos cement (9.1%) and LADWP
has compiled a detailed inventory of the system and leak trend data. Through analysis of leak
data, LADWPs main replacement program has targeted those portions of the system that would
be expected to have the highest leak incidence based on experience or prediction, i.e., older
cast iron mains. However, the current analysis of blowouts has identified interesting patterns of
failure that indicate that certain locales, such as La Cienega Boulevard, may be more prone to
this type of failure.

Applying a service-life model of this type to set priorities for main replacement will help to re-
duce the likelihood of failure, and from that standpoint it is a reasonable risk-reduction strategy.
However, as was shown earlier in this section, probability is only half of the risk equation. This
type of model does not account for the consequences of failure, such as the cost of long-term
traffic disruption, loss of water to a hospital or major tourist attraction, or more catastrophic im-

Page | 140
8. Asset Management
pacts such as slope failure. Therefore, it is recommended that LADWPs asset management
program include the consequences of failure in the priority-setting process.

Because many of the secondary impacts of major leaks and blowouts increase rapidly with du-
ration, LADWPs risk management could be improved by reducing response times and by re-
placing valves so that a leaking pipe could be isolated as quickly as possible without causing
operational problems in adjacent parts of the system. This would permit a situation to be norma-
lized as quickly as possible.

Despite the predictions of the service-life model, unexpected failures are inevitable. Therefore,
each new leak or blowout should be reviewed carefully to determine if the incident is merely a
random event or represents the leading edge of a previously unpredicted failure trend. Such
analytical capacity should be housed within the appropriate organizational unit of LADWP. In
order to ensure that the details of all leaks are captured for later analysis, LADWP should also
consider moving to digital field recorders to replace the current practice of written field notes.

Overall, LADWP appears to be following industry practices in its asset management program
and compares favorably with utilities of similar size and complexity. However, this has had little
apparent impact on the frequency or severity of the blowouts experienced during summer 2009,
which suggests that there may be forces at play not accounted for in the service life model. Al-
though it is difficult to determine if the appropriate level of resources is being committed to main
replacement, high-risk sections of the system should be replaced as quickly as practicable. In-
frastructure asset management is an imprecise activity where success may be difficult to define
and measure and research to date has been unable to define a precise relationship between
asset management expenditures and system performance. In order that LADWP customers re-
ceive (or perceive that they are receiving) the best possible service, within the constraints of
available resources, it is critically important that the objectives of the asset management pro-
gram be defined in concert with stakeholders outside of the agency and that performance
measures be adopted that are realistic, reproducible, and transparent.

8.5 Recommendations
To improve its asset management program, LADWP should consider the following actions:

Include the consequences of failure into the priority-setting process for water main re-
placement
Improve response times to major leak incidents and devise less disruptive methods for
isolating leaking mains from the rest of the system
Establish a forensic unit to analyze major leaks within the system to determine if the
leaks fit established patterns or represent a possibly new failure mode
Utilize digital field recorders (rather than written notes) to capture field data in directly
usable form
Increase the level of resources devoted to main replacement so as to reduce the back-
log as quickly as practicable

Page | 141
8. Asset Management
Include a broad range of stakeholders in the setting of objectives for the asset manage-
ment program and in the development of performance objectives
Build a program of asset management that capitalizes on the findings of this study.

The last recommendation implies that additional analysis may be needed to complement the
findings of the present analysis presented in this report.

Page | 142
9 CONCLUSIONS

The main objective of the study is to evaluate why pipeline breaks, including blowouts, oc-
curred in unusually large numbers in Los Angeles during the summer of 2009. A blowout is a
term used by water distribution personnel to describe a pipeline break causing heave and un-
dermining of the street that requires pavement replacement for 100 ft2 or more. An expert review
team examined the data collected by LADWP and identified the main factors contributing to
pipeline breaks and blowouts during the summer of 2009. The team evaluated pipeline perfor-
mance from the perspectives of:

Material sciences
Geotechnical engineering
Statistics and geo-statistics

It also initiated pilot studies in:


Hydraulics to audit what was happening in the water distribution system
Remote sensing using state-of-the-art InSAR and GPS techniques to assess water pipe-
line breaks and identify the effects of:
o tectonics deformation
o subsidence
o earthquakes
Asset management programs for water distribution systems.

9.1 Main findings


LADWP operates a complex water distribution system extending over 7200 miles. Its pipelines
are composed primarily of four major types of materials (cast iron, steel, ductile iron and asbes-
tos cement) and have diameters ranging from 4 inches for distribution mains to 60 inches and
above for trunk lines. Collecting and analyzing information for such a complex system is a time-
consuming exercise, but a prerequisite for understanding the origins of unusual failures.

There was an increase in blowouts in Los Angeles during the summer of 2009101 blowouts
twice as much as during the summer of 2008, and more than occurred during previous sum-
mers (summers include the months of July, August and September). Approximately 65% of
LADWP water pipelines are made of cast iron, a brittle material that becomes further weakened
and more brittle as it corrodes externally. More than 90% of blowouts observed between July 1

Page | 143
9. Conclusions
and October 1, 2009 occurred in cast iron pipelines that were externally corroded. The brittle
failure of cast iron pipes contributed to pipe wall failure through round cracks and longitudinal
cracks, all capable of releasing substantial amounts of water and thereby leading to heaving
and undermining of streets.

The study shows that there are more leaks in the cast iron pipes that make up the majority
(65%) of the LADWP water distribution system. However, when leaks are reported per unit
length of pipelines, steel pipes are found to leak more than other pipes including cast iron pipes.
In terms of diameter and unit length of pipelines, smaller pipes (4 inches in diameter and small-
er) are found to leak more than larger pipes. It is also found that cast iron pipes blowout more
frequently than pipes made up of other materials including steel pipes, and that when reported
in terms of unit length, there are more blowouts in cast iron pipes that are 6 and 8 inches in di-
ameter.

The performance of the LADWP system was compared with that of 6 other U.S. water systems.
The sizes of the other systems vary from 1656 to 8072 miles of pipelines. The LADWP system
was found to perform close to the national average. The annual average leakage rate for
LADWP is 21 pipeline breaks per 100 miles of pipelines, compared with the national average of
21 to 25. The annual average leakage rates for the 6 other systems in the comparative assess-
ment varied from 4 to 42.

The study finds that cast iron pipeline leaks and blowouts occur more frequently in highly corro-
sive soils. A relatively high number of pipeline leaks occur in the area surrounding the Santa
Monica Mountains, where corrosive soils are found. Several LADWP pressure zones in the
Southwest San Fernando Valley and Hollywood experienced an unusually high rate of blowouts.

A simplified geotechnical model was developed for understanding the mechanics of blowouts,
which are characterized by damage to street pavements larger than 100 square feet. This mod-
el suggests that blowouts may form when pipes (1) fail suddenly and release water pressure
that heave the street pavements and burst through them, or (2) leak enough water to erode pipe
bedding and form cavities undermining the street pavements. In both cases, for blowouts to oc-
cur water pipe breaks must be sufficiently large. Only the largest water leaks are likely to create
blowouts, implying that large-diameter pipes and pipes that exhibit brittle failure (e.g., cast iron)
are more likely to generate blowouts.

The analysis of crustal deformation data carried by the Jet Propulsion Laboratory (JPL) sug-
gests that there is a weak correlation between changes in frequency of pipeline breaks and blo-
wouts and changes in GPS time series. Evidence that ground water motions influence surface
deformation can be observed with both GPS and InSAR. It is not clear what component of the
deformation is associated with leaking pipes and what component is associated with annual
aquifer discharge and recharge. JPL analysis concluded that the pipeline breaks are not related
directly to the deformation of the surface of the Earth, but to some other factors.

Portable pressure monitoring instruments were deployed as part of this study and used to
record internal pressures at several locations in the water distribution system. Overall, the pres-
sure monitoring performed as anticipated, with enough accuracy to detect pressure transients.

Page | 144
9. Conclusions
However, it did not measure unexpectedly large pressure transients during the time of monitor-
ing. The absence of high transients does not indicate a lack of adverse pressure; it means simp-
ly that none occurred during the monitoring period. The monitoring program started during the
rainy season in Los Angeles when demand is lowest and pressures are in the lowest ranges. An
extension of the monitoring program to the high demand season may be more revealing in the
future.

The study shows that the most important contributing factors to the failure of cast iron pipelines
are temperature changes and internal pressure. Since 2001, sudden decreases in water tem-
perature during December and January have been systematically related to increasing pipeline
breaks, an effect that is well documented in many other municipalities. However, this thermal
effect during the winter months does not explain the increased water pipeline breaks in the
summer of 2009.

There is evidence that the cyclic amplitudes of internal water pressure increased due to water
rationing and contributed to the failure of cast iron pipes by metal fatigue. At various locations in
the LADWP water distribution system, the water pressure dropped significantly on Mondays and
Thursdays after June 1, 2009, when the water-rationing program began in Los Angeles. Tempo-
rary drops in water pressure on these days were caused by increased water flow during the wa-
tering of lawns. The expert review team study has shown that increased amplitudes of cyclic
daily pressure can have a negative impact on pipe materials with lower fatigue resistance, es-
pecially corroded cast iron pipes. Fatigue effects on cast iron pipes do not appear immediately,
but accumulate gradually. Cycling at increased amplitudes of pressure accumulated gradually in
June through September, 2009 as aging cast iron pipes with low fatigue resistance were sub-
jected to elevated pressure cycles. Fatigue-related pipeline breaks resulted in pipe failures (re-
ferred to as blowouts), which raised concerns from the media and the public because of the
damage they caused to street pavements and properties.

9.2 Recommendations
The LADWP water distribution system is aging and cast iron pipelines within that system are
becoming more susceptible to brittle failure as part of the aging process. An aggressive pipeline
replacement program, as proposed by LADWP, can mitigate the effects of aging and increased
vulnerability over the long term.

LADWP is commended for developing a comprehensive GIS model of its water distribution sys-
tem and for establishing a new program for documenting pipeline repairs. This study, however,
recommends that LADWP minimizes data loss on pipe repairs and starts utilizing digital tech-
niques that streamline the data flow from field observations to its GIS database.

In the immediate future, this study recommends avoiding abrupt variations in water pressure as
much as possible. To avoid increased levels of water main breaks during the summer of 2010, it
is recommended that LADWP alters its water-rationing program so that it evenly distributes the
variations of water pressure over time, and avoids increasing the amplitude of water pressure
cycles. For instance, LADWP could devise water rationing so that properties with odd and even

Page | 145
9. Conclusions
street numbers engage in watering lawns on different days. This solution would reduce sudden
drops of water pressure and would impose less stress on corroded cast iron pipelines.

A water-rationing strategy, which diffuses variations of water pressure, is a short-term measure


to avoid increased levels of breaks and blowouts in the summer of 2010. It is also recommend-
ed that LADWP evaluates an investment in long-term research to better quantify the factors
contributing to pipeline failure, improve field inspection techniques, and develop more efficient
pipe replacement programs and integrated asset management.

LADWP should consider improving its asset management program by:

1. Including the consequences of failure into the priority-setting process for water main re-
placement
2. Improving response times to major leak incidents and devising less disruptive methods
for isolating leaking mains from the rest of the system
3. Utilizing digital field recorders to capture field data in directly usable form
4. Increasing the level of resources devoted to main replacement so as to reduce the back-
log of aging and vulnerable mains as quickly as is practical
5. Including a broad range of stakeholders in the setting of objectives for the asset man-
agement program and the development of performance objectives

Page | 146
10 REFERENCES

1. Agbenowosi, N. 2000. A Mechanistic Analysis Based Decision Support System for


Scheduling Optimal Pipeline Replacement. Ph. D. Dissertation, Virginia Polytechnic Insti-
tute and State University. Blacksburg, VA
2. Andreou, S. A, D.H. Marks, and R. M. Clark. 1987. A new methodology for modeling
break failure patterns in deteriorating water distribution systems: Applications. Advances
in Water Resources. Elsevier Science, B.V. 10(1), 11-20.
3. Andreou, S.A. 1986. Predictive models for pipe break failures and their implications on
maintenance planning strategies for deteriorating water distribution systems. Ph. D. Dis-
sertation. Massachusetts Institute of Technology. Cambridge, MA.
4. Arnold, G.E., E. J. Clark, G. J. Remus, and H.W. Niemeyer. 1960. Experience with
Main Breaks in Four Large Cities. Journal of the American Water Works Association,
52(8):1041-1058.
5. ASCE. 2009. 2009 Report Card for Americas Infrastructure. U.S.A. American Society of
Civil Engineers.
6. Atkinson, K. J.T. Whiter, P.A. Smith and M. Mulheron. 2002. Failure of Small Diameter
Cast Iron Pipes, Urban Water, 4, p. 263.
7. AWWA. 1975. USA Standard for Thickness Design of Cast-Iron Pipe. USAS A21.1-
1967/AWWA H1-67, American Water Works Association, Denver, Colorado.
8. AWWA. 2003. Water:II Stats 2002 Distribution Survey. American Water Works Associa-
tion, Denver, Colorado.
9. AWWA. 2005. Polyethylene Encasement for Ductile-Iron Pipe Systems. American Water
Works Association, Denver, Co. ANSI/AWWA C105/A21.5.
10. AwwaRF. 1997. Main Break Prediction, Prevention, and Control. RFP 461. AwwaRF,
Denver, Colorado.
11. Baracos, A., W.D. Hurst, and R.F. Legget. 1955. Effects of Physical Environment on
Cast-Iron Pipe. Journal of the American Water Works Association, 47(12):1195-1206.
12. Ben-Akiva, M., F. Humplick, S. Madanat, and R. Ramaswamy. 1993. Infrastructure
Management Under Uncertainty: Latent Performance Approach, Journal of Transportation
Engineering, 119 (1): 43-58.
13. Boulos, P.F., K.E. Lansey, and B.W. Karney. 2006. Comprehensive Water Distribution
Systems Analysis Handbook for Engineers and Planners. Pasadena, CA. MWH Soft.

Page | 147
10. References
14. Clark, R. M., C. L. Stafford, and J.A Goodrich. 1982. Water Distribution Systems: A
Spatial and Cost Evaluation. Jour . Water Resources Planning and Management Division,
ASCE, 108(WR3) 243-256.
15. Committee on Public Water Supply Distribution Systems. 2006, Drinking Water Dis-
tribution Systems: Assessing and Reducing Risks. Washington D.C. National Academies
Press.
16. Cullinane M. J. Jr., J. Goodrich, and I. Goulter. 1989. Water Distribution System Eval-
uation. In Reliability Analysis of Water Distribution Systems. Edited by L. W. Mays. New
York, NY: American Society of Civil Engineers.
17. De Rose, P.J. and R.W. Parkinson. 1985, Corrosion of Ductile Iron Pipe, Report
TR241, WRC Engineering, Water Research Centre, Swindon, United Kingdom.
18. Deb, A. K., Y. J. Hasit, H. M. Schoser, and J. K. Snyder. 2002. Decision Support Sys-
tem for Distribution System Piping Renewal. AwwaRF and AWWA, Denver, Colorado.
19. DIPRA. 1984. Handbook of Ductile Iron Pipe Design, 6th ed., Ductile Iron Pipe Research
Association, Birmingham, Alabama, p. 9, 1984.
20. DIPRA. 1990. Cast Iron Pipe Century and Sesquicentury Club records and correspon-
dence, Ductile Iron Pipe Research Association, Birmingham, Alabama.
21. Dolson, F.E. 1955. Factors Causing Main Failures. Journal of the American Water
Works Association, 47(2):465-469.
22. Elnaboulsi, J. and Q. Alexandre. 1997. Determination de la date optimale de renouvel-
lement des reseaux urbains d'eau potable. Canadian Water Resources Journal. Cam-
bridge, CANADA, 22(3):327-350.
23. EPA. 2008. FACTOIDS: Drinking Water and Ground Water Statistics for 2008, US Envi-
ronmental Protection Agency, Office of Water (4601M), EPA 816-K-08-004, November
2008 <<www.epa.gov/safewater/data>>
24. Fitzgerald, J. H. 1968. Corrosion as a Primary Cause of Cast-Iron Main Breaks, Journal
of the American Water Works Association, Vol. 60, No. 8, p.882.
25. Flemming, K.K., J.P. Dugandzic, M.W. LeChevallier, and R.W. Gullick. 2006. Suscep-
tibility of Distribution Systems to Negative Pressure Transients. AwwaRF and AWWA,
Denver, Colorado, and IWA Publishing.
26. Goodrich, J, C. P. S. Frebis, E. J. Read, and Robert Clark. 1987. Management of Water
Distribution Systems Components. Proc. National Conference on Hydraulic Engineering.
Williamsburg, VA.
27. Goodrich, J., L. W. Mays, Y Su and 1 Woodburn. 1989. Data Base Management Sys-
tems. In Reliability Analysis of Water Distribution Systems, Edited by L. W. Mays. New
York, NY: American Society of Civil Engineers.
28. Grigg, N. S. 2005. Assessment and Renewal of Water Distribution Systems. Journal of
the American Water Works Association, 97(2):58-68.
29. Grigg. N. S. 2004. Water Distribution Assessment and Renewal. AwwaRF and AWWA,
Denver, Colorado.
30. Habibian, A. 1994. "Effects of temperature on water main breaks, ASCE Journal of
Transportation Engineering, Vol. 120, No. 2, pp. 312-321.
31. Haimes, Y.Y. 1991. "Total Risk Management," Risk Analysis, 11(2):16971.

Page | 148
10. References
32. Hamilton, D.H., and R.H. Meehan. 1971. Ground Rupture in the Baldwin Hills, Science,
172, 333-334.
33. Hope, R. 2003. Establishing the Estimate of Annual Replacement Rate for Water Mains.
Workshop for Infrastructure: Above and Below Ground. AWWA ACE 2003, Anaheim.
34. Hosoi, Y., H. Murakami, M. Kozai, and S. Marutaka. 1989. Consideration of Water Main
Breakage, Japan Water Works Association, 58(5): 10-16.
35. Ivanova, N. G.1978, Resistance to Fracture of Gray Cast Iron Pipe, Metallovenenie I
Termicheskaya Obrabotka Metallov, No. 7, pp. 25-28.
36. Kane, M. J. and A. Gough. 1996. Database to Prioritize Mains Rehabilitation. (Corpo-
rate author: Severn Trent Water Ltd., Birmingham, England). /WSA Specialized Confe-
rence on New developments in Modeling Monitoring and Control of Water Supply Sys-
tems, Amsterdam, Netherlands, September 24 -5, 1996.
37. Kaplan, S. and B.J. Garrick. 1981. "On the Quantitative Assessment of Risk," Risk Anal-
ysis 1(1):1127.
38. Kettler, A. J. and I. C. Goulter. 1985. An Analysis of Pipe Breakage in Urban Water Dis-
tribution Networks. Canadian Journal of Civil Engineering, 12: 286-293.
39. King, R. A., 1977 A Review of Soil Corrosiveness with Particular Reference to Reinforced
Earth, Transport and Road Research Laboratory, Supplementary Report 16.
40. Kirmeyer, G. J., W. Richards, and C.D. Smith. 1994. An Assessment of Water Distribu-
tion Systems and Associated Research Needs, American Water Works Association Re-
search Foundation, Denver, Colorado.
41. Kleiner, Y and B. Rajani. 2002. Forecasting variations and trends in water-main breaks.
National Research Council. Canada.
42. Kleiner, Y. and R. Balvant. 2001. Comprehensive review of structural deterioration of
water mains: statistical methods. Urban Water, 3, 131-150. .
43. Kleiner, Y., R. Sadiq, and B. Rajani. 2004. Modeling Failure Risk in Buried Pipes Using
Fuzzy Markov Deterioration Process. ASCE International Conference on Pipeline Engi-
neering and Construction. San Diego. August 2-4, 2004.
44. Kulkarni, R. and F. Reid. 1992. Analytical Techniques for Selection of Repair and Re-
place Options for Cast Iron Piping Systems. Final Report, Contract GRI-9210234. Chica-
go, IL: Gas Research Institute (GRI).
45. Kulkarni, R. B., K. Golabi, G. Dugovic and K. Burnham. 1990. Optimal Replacement
decisions in Cast Iron Piping Systems. In Prioritizing and Pipe Replacement Decisions,
Technical Report No XL-8920, American Gas Association, 34-39.
46. LADWP. 2009. LADWP Quick Facts and Figures 2009, December 2009,
<<http://www.ladwp.com/ladwp/cms/ladwp000509.jsp>>
47. LADWP. 2010, A Detailed Analysis of Cast Iron Behavior, Los Angeles Department of
Water and Power Internal Report, received from George Mavrakis via email, 1/07/2010.
48. LADWP. 2010, A Detailed Analysis of Steel Pipe Behavior, Los Angeles Department of
Water and Power Internal Report, received from George Mavrakis via email, 1/07/2010.
49. Lanari, R., P.L. Lundgren, M. Manzo, F. Casu. 2004. Satellite radar interferometry time
series analysis of surface deformation for Los Angeles, California, Geophysical Research
Letters, 31, L23613, doi:10.1029/2004GL021294.
50. LaQue, F. L. 1958. The Corrosive Resistance of Ductile-Iron, Corrosion, 14:485t (1958).

Page | 149
10. References
51. Lim, E. L. 1997. Pipe Evaluation System (PIPES). Environmental Systems Research
Institute (ESRI) User's Conference, San Diego, CA, July, 1997.
52. Loganathan, G.V., H.D. Sherali, S. Park, and S. Subramanian. 2002. Optimal Design
Rehabilitation Strategies for Reliable Water Distribution Systems. Virginia Polytechnic In-
stitute and State University, Water Resources Research Center. Blacksburg, Virginia.
53. Madioc, H., P. Botzung. B. Bremond and P. Eisenbeis. 1996. Diagnosis Methods and
Performance Indicators for Rehabilitation Policies. Water Supply, 12(314)347-358.
54. Makar, J. M. 2000. A preliminary analysis of failures in grey cast iron water pipes, Engi-
neering Failure Analysis 7, pp. 43-53.
55. Makar, J., R. Rogge, S. McDonald, and S. Tesfamariam. 2005. The Effect of Corrosion
Pitting on Circumferential Failures in Grey Cast Iron Pipes. AwwaRF and AWWA, Denver,
Colorado.
56. Marks, D. H., S. Andreou, C. Jeffrey, D. Park and A. Zaslavsky. 1987. Statistical Mod-
els for Water Mains Failures. Contract CR810558, U.S. E.P.A. Drinking Water Research
Division, Water Engineering Research Laboratory, Report 1. EPAl16015-871003108, Cin-
cinnati, OH, 108 pages.
57. Mavrakis, G. 2003. Evaluating Condition of Water Facilities Using Predictive Life Cycles.
Presented at Distribution System Infrastructure Integrity Workshop. April 3-4. Fort Collins,
CO: Colorado State University.
58. Mays, L.W. 1990. Models of Optimal Maintenance of Hydraulic Structures. In Proc. Of
Symposium: Water Resource infrastructure: Needs, Economics, and Financing. J.F. Scott
and R.M. Khanbilvardi, Ed. Ft. Worth, TX: NY: American Society of Civil Engineers.
59. Mohebbi, H., D.A. Jesson, M.J. Mulheron and P.A. Smith. 2009. Characterisation of
the fatigue properties of cast irons used in the water industry and the effects on pipe
strength and performance, 7th International Conference on Modern Practice in Stress and
Vibration Analysis, Journal of Physics, Conference Series 181, 012029.
60. Morris Jr., R.E. 1967. Principal Causes and Remedies of Water Main Breaks. Journal of
the American Water Works Association, 59(7):782-798.
61. NRC. 1995. Measuring and Improving Infrastructure Performance. Washington, D.C.: Na-
tional Academy Press.
62. O'Day, O.K., R. Weiss, S. Chiavari, and D. Blair. 1986. Water Main Evaluation for Reha-
bilitation/Replacement. Denver, Colo.: AwwaRF, AWWA, and Environmental Protection
Agency.
63. Pelletier, G., Mailhot, A. and Villeneuve, J.-P. 2003. Modeling Water Pipe Breaks -
Three Case Studies. Journal of Water Resources Planning and Management, 129 (2),
115-123.
64. Rajani, B., J. Makar, S. McDonald, and C. Zhan. 2000. Investigation of Grey Cast Iron
Water Mains to Develop a Methodology for Estimating Service Life. AwwaRF and AWWA,
Denver, Colorado.
65. Rajani, B.B., C. Zhan, and S. Kuraoka. 1996. Pipe-soil interaction analysis of jointed
water mains, Canadian Geotechnical Journal, Vol. 33, No. 3.
66. Rogers, P. 2006. Failure Assessment Model to Prioritize Pipe Replacement in Water
Utility Asset Management. Ph.D. Dissertation, Colorado State University. Fort Collins,
Colorado.

Page | 150
10. References
67. Romanoff, M. 1957. Underground Corrosion. National Bureau of Standards Circular
579. Reprinted by NACE International, Houston, Tx. 1989.
68. Runge, J. and J. Mann. 2005. State of the Industry Report, Journal of the American Wa-
ter Works Association, 97(10):58-67 .
69. Schiff, M.J. 1993. What is Corrosive Soil? Western States Corrosion Seminar, California
State Polytechnic University at Pomona. May 1993.
70. SEAMS. 2002. Decision Support for Clean Water Infrastructural Planning in UUplc. Shef-
field, U.K.
71. Seica, M.V. and J.A. Packer. 2004. Mechanical Properties and Strength of Aged Cast
Iron Water Pipes, J. Materials in Civil Engineering, Jan/Feb 2004, p. 69-77
72. Shamir, U. and C.O.D. Howard. 1979. An Analytic Approach to Scheduling Pipe Re-
placement. Journal of the American Water Works Association, 3: 45-56.
73. Stroud, T. F. 1985. Infrastructure: Is the problem being blown out of proportion, Ductile
Iron Pipe News, Fall/Winter 1985, p. 9.
74. Stroud, T.F. 1989 Corrosion Control Measures For Ductile Iron Pipe, Corrosion 89, Pa-
per No. 585, (Houston, TX; NACE 1989)
75. Talbot, A.N. 1926. Strength Properties of Cast Iron Pipe Made by Different Processes as
Found by Tests, Journal of the American Water Works Association, Vol. 16, July, pp.1-44.
76. Thames Water Utilities. 1998. Utilnets: model for hydraulic impact consequences of
mains bursts. London, U.K.
77. Vanrenterghem, A. 2003. Modeling the Structural Degradation of an Urban Water Distri-
bution System. Ph. D. Dissertation, Polytechnic University. Brooklyn, N.Y.
78. Wolfe, T. F. 1946. How to Prevent Breaks in Cast-Iron Pipe. Journal of the American
Water Works Association. 38(6) 765.
79. Wood, D. J., S. Lingireddy, and P.F. Boulos. 2005. Pressure Wave Analysis of Tran-
sient Flow in Pipe Distribution Systems. Pasadena, CA. MWH Soft.

Page | 151
11 APPENDICES

11.1 Summary of literature review


AWWA (2007) reviewed the past work on main breaks. Data on reasons for pipe failure began
to appear after the 1940s (Wolfe, 1946; Baracos et al, 1955; Dolson, 1955; and Arnold et al.,
1960). Morris (1967) is one of the earliest comprehensive treatments of cast iron pipe failure
causes. In the last three decades, only a few researchers have reported on main break causes
and prediction. Shamir and Howard (1979) proposed an exponential model for the increase of
breaks with time and to find an optimum year to replace a main. Clark et al (1982) attempted a
statistical analysis of factors expected in water main failures. Their analysis used six factors:
pressure differential, time since first break, land use over main, type of pipe and low and high
corrosivity factors. Using cost data, Wlaski and Pellicia (1982) derived and optimal replacement
time. After reviewing a number of references, Grigg (2004, 2005) compiled a list of nine catego-
ries of predictor variables

Degradation of material, including internal and external corrosion


Weak joints and poor installation procedures
Soil displacements
Frozen soil around pipes
Construction accidents
Water temperature extremes
Externally imposed loads
Manufacturing defects
Pressure surges

Kettler and Goutler (1985) evaluated regression for the number of breaks versus diameter and
time for cast iron and asbestos-concrete main in Canada. They showed a strong inverse linear
correlation between failures and diameter. ODay et al (1986) analyzed data from six utilities
representing diverse geographic, demographic and geologic conditions. They summarized sta-
tistics on failure rates at utilities, types of main deterioration and contributing factors, ways of
assessing main conditions or risk of failure, economic methods for evaluating needed replace-
ment or rehabilitation, needs for break and leak definitions on repair records, mains inventory,
leak detection, condition monitoring and mains testing.

Page | 152
Appendix: Summary of Literature Review
Andreou et al. (1987a and 1987b) and Marks et al. (1987) studied two data sets from large utili-
ties in the Northeast. They applied three different types of models: aggregate models, regres-
sion type models, and probability models. All three approaches were severely limited by data
availability. The analysis shows interesting ways to represent data using histograms of relative
hazard per foot, survivor functions for individual pipes, and change in hazard rate if a variable
changes.

Goodrich et al. (1987) reported on time to first failure. They obtained significant positive linear
relationships to break probability for pressure, length and the multiplicative combination of di-
ameter, period of installation, and installer (utility or contractor). They found an inverse relation-
ship to diameter.

Goodrich et al. (1989) surveyed 12 utilities. They classified 14 mains and break data items in 5
levels of availability. The mains and repairs attributes were: diameter, material age, soil type,
pressure and pipe thickness, plus repair rate and time and clarity of definition for the break
records. Only 4 utilities computerized their data and only one coded a majority of the 14
attributes in the survey.

Boynton (1989) analyzed soil data and correlated it to breaks. He used linear multivariate re-
gressions of age on factors such as soil corrosivity, pH, redox and sulfide measures. Cullinane
et al. (1989) compiles break rate statistics for several utilities. Hosoi et al. (1989) presented
tables and plots of failure rates in Japan and North America versus diameter, type of material,
joint and body pipe, temperature and city. With the exception of few cities in America, Japan
failure rates are over twice as high. Mays (180) demonstrated that there is an economically op-
timal point of time for action in the tradeoff between preventive and reactive maintenance for
any infrastructure component.

The models, named CIMOS (Cast Iron Maintenance Optimization System), were field tested
and validated at five gas companies. Software and methods were found to be useful and appli-
cable to gas companies. This work is also described in a report for the Gas Research Institute
(Kulkarni and Reid, 1992).

Kulkarni et. al. (1990) described an optimization system for decisions between repairs, rehabili-
tation, and inspection/testing for transcontinental gas pipelines. Since failures are infrequent but
consequences are large, the inspection alternatives are important. The transmission model,
named PIMOS, is based on a GIS database of pipeline characteristics, maintenance records,
demographic, and environmental data along the lines.

Kane and Gough (1996) reported on a prioritization system for mains replacements at Severn
Trent Ltd., the company that privatized the 25,000 miles of distribution mains in the Birmingham,
England area. Priorities for District Meter Areas (DMAs) were based on water quality, aesthetics
(discoloration , odor), service interruption, pressure loss, structural condition (age and history of
bursts), maintenance costs, and leakage (damage, cost, supply).

Elnaboulsi and Alexandre (1997) derived a single functional form of the multi-stage proportional
hazards models described by Marks et. al. (1987). They defined and collected data on direct

Page | 153
Appendix: Summary of Literature Review
and indirect economic consequences of breaks and used them to determine the optimum year
to replace mains in the Paris and Strasbourg, France water systems. The considered replace-
ment costs, consequence costs, and discount rates. Direct cost data used were for repairs, ex-
tra pumping, and water lost. Indirect costs used were user damages and loss of profits: Other
possible consequence costs were discussed, to include: traffic delays, street and utility damage,
hospital loss of water, water customer inconvenience, fire protection, electric and computer in-
stallation cooling, work stoppage, flooding, sensitive agricultural operations, and water contami-
nation.

Lim (1997) reported on a mains evaluation system "PIPES" in the City of Seattle. PIPES had
consequence factors in its vulnerability and criticality submodels. These were linked through its
GIS database to the locations of predicted breaks. The vulnerability sub-model considers both
vulnerability of the pipe and of the public. The criticality model considers land use near the main,
such as hospitals, dialysis centers, schools, major water users, community centers, and flood-
prone areas. The system shows implementation of consequence factors via a GIS database to
prioritize and evaluate problem mains,

Lyonnaise des Eaux's (Madioc et. al. 1996) PREDIKASS model is a tool for failure prediction
and condition assessment. PREDlKASS was reported to be require 6 to 10 years of mains data.

Thames Water Utilities, Ltd. (1998) reported "Utilnets" as a hydraulic model to estimate burst
flows with elevation models for route and fate of water and priority data for consequence analy-
sis.

Deb et al (2002) reviewed utility programs to collect data and assess reasons for pipe failure.
They described the 1980s efforts of the Philadelphia Water Department to conduct visual, me-
chanical, and soil tests on pipe failures. Cold weather causes breaks and most breaks are cir-
cumferential. The Severn Trent study found that breaks decrease after cleaning and lining, a
finding that is at odds with the modeling efforts of Andreou et. al. (1987a). They also proposed a
method to prioritize individual pipes. They incorporated two models of pipe behavior: a pipe load
model and a pipe deterioration model. These corresponded to the terms in the equation of An-
dreou et al. (1987a). The pipe deterioration model, corresponding to Andreou's baseline hazard
function, includes information from pipe characteristics and soil samples. Some soil properties,
such as pH, moisture, resistivity and aeration, combined with pipe properties and water quality,
may lead to deterioration over time. The pipe load model takes into account other soil properties
that influence loading such as moisture, bulk density, grain size and expansiveness. Other load-
ing factors such as frost, earth and truck load are also considered, along with internal pressures
like water hammer and thermal changes. These loads are included in the model and used to
determine stresses on the pipe. The two models, along with break history , climate data and
safety factors, predict the expected condition of a pipe and were intended for use to prioritize
main renewal programs. However, lack of validation hinders application of the models.

Canada's National Research Council has modeled risk of individual pipes (Kleiner, et. al., 2004).
They found that "distress indicators" are required and are inherently imprecise and involve sub-
jective judgment. However in addition to these reports, Rogers (2006) reviewed several others
on main failures. For example, Agbenowosi (2000) developed a mechanistic model to predict

Page | 154
Appendix: Summary of Literature Review
pipe failures. Kleiner and Rajani (2001) published a review of statistical methods to model dete-
rioration of water mains. Loganathan, Sherali, Park, and Subramanian (2002) presented a me-
thod to truncate data sets. Pelletier, Mailhot, and Villeneuve (2003) presented three case stu-
dies of water main breaks. Vanrenterghem (2003) studied the New York City water system.
Meanwhile, additional studies and datasets for main failures and models continue to be re-
ported. For example, SEAMS (2002) prepared a model for infrastructure decision support for
U.K. utilities."

Page | 155
11.2 The team of experts

Professor of Civil Engineering


Chair, Sonny Astani Department of Civil and Environmental Engi-
neering
Director, USC Center on Megacities
Education
o Ph.D., 1984, California Institute of Technology, Pasadena
o M. S., 1979, California Institute of Technology, Pasadena
o Engineer, 1978, Ecole Centrale de Lyon, France
Expertise
o Geotechnical engineering
Jean-Pierre Bardet o Earthquake engineering
o Civil Infrastructure systems
o Computational mechanics
o Geo-information

Postdoctoral fellow, Sonny Astani Department of Civil and Envi-


ronmental Engineering
Education
o 2009 Ph.D. Civil Engineering (Structural Mechanics)
o 2007 M.B.S. Architecture (Building Science)
o 2006 M.S. Electrical Engineering (Control Systems)
o 2004 M.S. Civil Engineering (Structural Engineering)
o 2003 B.S. Civil Engineering
(Building Science; minor: Mathematics)
Tat Fu Expertise
o Structural mechanics
o Statistics

Thomas R. Briggs Professor of Engineering


Education
o Ph.D. - University of Illinois at Urbana-Champaign 1975
Expertise
o Teaching and professional practice have covered many
aspects of geotechnical engineering including foundations,
earth retaining structures, slope stability, soil/structure inte-
raction, underground construction, laboratory testing, and
elements of earthquake engineering.
o Authored or co-authored over 280 publications on geo-
Thomas D ORourke technical, underground, and earthquake engineering.
o Elected a member of the National Academy of Engineering
in 1993

Page | 156
Appendix: The Team of Experts
Lecturer/Research Associate, Cornell University
Licensed Professional Engineer in California, New York and
Pennsylvania
Education
o Ph.D. Cornell University 1995
o M.S. SUNNY Buffalo 1989
o B. T. Rochester institute of Technology 1987
Expertise
o GIS
o Geotechnical Engineering
o Water distribution systems
Michael C. Palmer

President and Principal Engineer, Schiff Associates


Education
o Ph.D., 1988, University of California, Los Angeles
o M. S., 1983, University of California, Los Angeles
Expertise
o More than 20 years of experience in the designing and
testing of corrosion control facilities and in the forensic
evaluating of corrosion damage on buried and submerged
structures,
o More than seven years in basic corrosion research at Oak
Ridge National Laboratory.
Graham E. C. Bell o NACE International Certified Corrosion Specialist P and Ca-
thodic Protection Specialist
o Certified by NACE International as both a Corrosion and Ca-
thodic Protection Specialist

President and Principal Consultant of Flow Science


Emeritus Professor of Environmental Engineering Science at the
California Institute of Technology.
Education
o 1965 - Ph.D. California Institute of Technology
o 1962 M.E. (Civil Eng.), University of Auckland, N.Z.
Expertise
o Over 30 years of experience as a consulting engineer, in-
cluding 10 years of litigation support experience.

John E. List

Page | 157
Appendix: The Team of Experts
Vice President of Flow Science
Education
o BSCE - Lehigh University, Bethlehem, Pennsylvania, 1988
Affiliation
o Professional Civil Engineer in California, Washington, Ha-
waii, Arizona, Oregon, Nevada, Utah
o Member of American Society of Civil Engineers
o Member of American Water Works Association
o Consulting Engineers and Land Surveyors of California
Expertise
R. Scott Foster
o Hydraulic analyses and modeling and analysis of pipeline
surges/transients, water hammer and pipeline failure,
including design of surge control devices and pump sta-
tion/turbine operations.

Director, Keston Institute for Public Finance and Infrastructure


Policy University of Southern California
Education
o Rensselaer Polytechnic Institute, Troy, NY
o M.S. Urban-Environmental Studies, 1970
o B.S. Geology, 1969
Professional Certifications and Affiliations
o National Academy of Construction (elected 2008)
o American Institute of Certified Planners (AICP), #044590
o American Planning Association, Society for Risk Analysis
Richard G. Little o Editor: Public Works Management & Policy; Editorial Boards:
ASCE Journal of Performance of Constructed Facilities, The
International Journal of Emergency Management
o Advisory Boards: Institute for Infrastructure and Information
Assurance, James Madison University
o Senior Fellow: University of Southern California
o Adjunct Faculty: James Madison University, Nanyang Tech-
nological University, Singapore

MMI Engineering, Federal Way, WA, Senior Consultant, October


2007 to present
Expertise
o Evaluated and designed civil/mechanical infrastructure for 33
years.
o Conducted 65 assessments of complicated utility systems
many focusing on pipeline performance and failures.

Donald Ballantyne

Page | 158
Appendix: The Team of Experts

Principal Scientist, Jet Propulsion Laboratory, California Institute of


Technology
QuakeSim Principal Investigator
Research Professor, Earth Science, University of Southern Califor-
nia
Education
o Ph.D., 1991, Geophysics, California Institute of Technology
o M.S., 1988, Computer Science, University of Southern Cali-
fornia
Andrea Donnellan o M.S., 1988, Geophysics, California Institute of Technology
o B.S, 1986, Geology, Ohio State University
Expertise
o GPS and InSAR measurement of crustal deformation
o Crustal deformation modeling using web-services and high-
performance computing

President and Principal Engineer, Hushmand Associates, Inc.


Visiting Faculty, California Institute of Technology, 1990-2004
Education
o Ph.D., 1983, California Institute of Technology, Pasadena
o M.S., 1978, California Institute of Technology, Pasadena
o B.S., 1977, Sharif University of Technology, Tehran, Iran
Expertise
o More than 25 years of experience in consulting geotechnical
and earthquake engineering, leading major national and in-
Ben Hushmand ternational design and forensic engineering projects
o Over 30 years of research and teaching experience on
liquefaction; soil-structure interaction; deep foundations;
seismic design of bridges and landfills; loading capacity of
culverts and pipes; base isolation; and ports and offshore
structures for Exxon Mobil, Kajima, USGS, NSF, USEPA, US
Navy, Caltrans, WASHDOT, US Air Force, and DOE.

Page | 159
11.3 Seattle Water System (by Don Ballantyne, MMI)
The Seattle Public Utilities (SPU) water system serves 630,000 people retail, and 720,000
people through wholesale customers. This report focuses on pipe owned by SPU, and not
owned by their wholesale customers.

SPUs water supply comes from two watersheds in the Cascade Mountains, from the Tolt River
to the north, and the Cedar River to the south. The SPU distribution service area has elevations
ranging from 0 to about 500 feet above sea level.

SPUs distribution system is predominantly made up of cast iron pipe as shown in Table 11-1.
Much of the cast iron pipe has leaded joints as shown in Table 11-2. The pipe length is shown
by pipe material and diameter in Table 11-3. The three largest categories are cast iron pipe of 8-
inch, 6-inch, and 12-inch diameters respectively.

Table 11-1. Seattle Water System Pipe Material

Material Feet Miles %oftotal


CastIron 7,160,073 1,356.07 73.316%
DuctileIron 1,214,801 230.08 12.439%
Steel 610,188 115.57 6.248%
ConcreteCylinder 252,676 47.86 2.587%
GalvanizedIron 173,780 32.91 1.779%
RivetedSteel 91,726 17.37 0.939%
LockBarSteel 91,101 17.25 0.933%
Concrete 68,427 12.96 0.701%
WeldedSteel 40,852 7.74 0.418%
Copper 10,488 1.99 0.107%
Polyvinylchloride 10,265 1.94 0.105%
WrappedSteel 8,806 1.67 0.090%
Kalamein 8,788 1.66 0.090%
GalvanizedSteel 8,680 1.64 0.089%
Iron 6,662 1.26 0.068%
WoodStave 3,755 0.71 0.038%
Plastic 2,673 0.51 0.027%
StainlessSteel 2,053 0.39 0.021%
WroughtIron 247 0.05 0.003%
Total 9,766,041 1,849.63 100%

Page | 160
Appendix: Seattle Water System
Table 11-2. Pipe Joint Type

JointType %
LeadJoint(LJ) 55.7%
GasketJoint(GJ) 18.0%
SlipJoint(SJ) 15.8%
RivetedJoint(RVJ) 2.4%
ThreadedJoint(THJ) 2.2%
RestrainedJoint(RJ) 1.3%
MechanicalJoint(MJ) 1.3%
WeldedJoint(WLD) 1.0%

Page | 161
Appendix: Seattle Water System
Table 11-3. Pipe Length by Material and Diameter (miles)

Miles Diam Cast Concrete Ductile Galvan LockBar Riveted Welded


(>2mi) eter Iron Cylinder Concrete Iron izedIron Steel Steel Steel Steel
4 0 4
3 1 0 0 1 0
36 2 1 0 29 0
90 4 69 17 1 3
164 6 149 13 0 1
996 8 863 132 0 0
13 10 12 1 0
184 12 140 0 42 0 0 1
52 16 44 0 6 2
57 20 46 0 9 2
35 24 14 4 4 5 1 0 7 0
29 30 14 6 0 1 1 0 7 1
4 34 4
22 36 0 16 2 0 0 2 1
14 42 2 1 0 0 0 1 10
15 48 3 1 0 4 0 6 0
3 51 2 0 1
8 51.5 8 0
28 54 4 3 8 12 0
18 60 6 2 9 1
36 66 5 1 3 17 9 2
14 78 13 1
6 81 6
4 92 4
4 96 4
1855 Total 1356 48 13 234 33 17 17 116 8
Notes:
Pipe material and diameter categories of less than 2 miles are hidden, so totals may not be cor-
rect using the data shown.
Pipe lengths shown as 0 are rounded, being less than 0.5 miles

Page | 162
Appendix: Seattle Water System
The number and percentage of pipe failures by category for all pipe materials and cast iron are
shown in Table 11-4 and Table 11-5 respectively for pipe data from 1990 through 2002. Corres-
ponding percentages are shown (using a different data set) for all materials from 2000 to 2008.
The significant difference is the reduction in pinhole leaks. This was accomplished through a
pipeline replacement program during that period focusing on thin wall steel pipe.

Table 11-4. Number and Percentage of Pipe Failures by Category for All Materials

H/V EATN BAD EXT H UN LD V PIN Sub


Year BRK AWAY CONN DAMG SPLIT KNOWN JOINT BREAK HOLE total
2002 5 8 6 7 20 40 42 20 148
2001 6 7 6 7 19 44 37 29 155
2000 4 8 3 10 18 27 51 37 158
1999 9 1 9 6 11 22 28 31 117
1998 3 3 8 3 1 30 47 29 124
1997 1 7 9 3 14 12 21 30 39 136
1996 1 16 5 7 8 13 26 36 49 161
1995 3 15 1 9 8 29 36 36 137
1994 5 5 7 2 11 29 28 68 155
1993 6 6 8 9 2 32 42 67 172
1992 1 10 5 6 4 10 60 55 82 233
1991 10 14 10 9 11 43 36 75 208
1990 8 4 20 10 21 32 33 83 211
Total 3 92 90 94 98 157 435 501 645 2115
%1990
2002 0.1% 4.3% 4.3% 4.4% 4.6% 7.4% 20.6% 23.7% 30.5% 100.0%
%2000
2008 0% 6% 5% 7% 6% 7% 24% 27% 16%

Legend:

H/V BRK Horizontal/Vertical Break


EATNAWAY Eaten Away such as from external corrosion
BAD CONN Bad Connection such as at a service tap
EXT DAMG External Damage such as being hit by backhoe
H SPLIT Horizontal Split
UNKNOWN - Unknown
LD JOINT Leaded Joint Failure
V BREAK Vertical Break such as circumferential crack
PINHOLE Pinhole such as from corrosion

Page | 163
Appendix: Seattle Water System

Table 11-5. Number and Percentage of Pipe Failures by Category for Cast Iron Pipe

H/V EATN BAD EXT H UN LD V PIN Sub


Year BRK AWAY CONN DAMG SPLIT KNOWN JOINT BREAK HOLE Total
1990 2 15 9 4 24 28 8 90
1991 3 13 4 8 4 37 33 2 104
1992 1 3 2 4 3 55 51 6 125
1993 4 5 4 7 1 30 39 9 99
1994 1 4 1 2 3 27 26 7 71
1995 14 1 6 1 23 32 3 80
1996 1 1 3 4 7 4 22 34 3 79
1997 1 3 1 13 2 20 26 3 69
1998 1 1 1 3 1 27 43 4 81
1999 1 6 4 5 20 26 4 66
2000 1 7 1 6 8 24 51 7 105
2001 2 6 1 6 9 38 35 2 99
2002 1 7 3 5 6 38 37 3 100
Total 3 14 69 44 80 51 385 461 61 1168
% 1990
2002 0.3% 1.2% 5.9% 3.8% 6.8% 4.4% 33.0% 39.5% 5.2%
Legend see Table 11-4

The corresponding pipe failure rates per 100 miles of pipeline per year are shown in Ta-
ble 11-6 for all pipe types and for cast iron pipe in the time period of 1990 to 2002. Mod-
ified failure rates are shown calculated using a methodology developed by CSIRO Ma-
thematical and Information Sciences Analysis of Hunter Water Mains Data which only in-
cludes: V BREAK, EATNAWAY, PINHOLE, H SPLIT and H/V BRK. This methodology ex-
cludes failures that are not attributed to pipe degradation including: BAD CONN, EXT
DAMAGE, UNKNOWN, and LD JOINT. These modified rates are also shown in Table 6.
SPU did some additional analysis on pipe break rates for the period of 2000 to 2008. The
results are shown in

Table 11-7.

SPU categorizes their breaks into 5 groups, and has categorized breaks into those groups from
2005 to 2007. The groups include:

1. 3 Repairs
2. 4-12 Spot Repairs
3. 4-12 Pipe Segment Replacement
4. 16-42 Spot Repairs
5. 16 42 Pipe Segment Replacement

Page | 164
Appendix: Seattle Water System
Groups 3 and 5, those requiring pipe segments to be replaced, are most likely to have resulted
in major leaks. A description of each of these failure categories over that time period is included
in Table 11-8.

Table 11-6. Pipe Break Rates/100 miles of pipeline/Year for the Period of 1990-2002

AllPipe CastIron CastIron


Year AllPipe Modified Pipe Modified
Method Method
2002 8.0 4.0 5.4 2.5
2001 8.4 4.3 5.4 2.4
2000 8.5 5.5 5.7 3.5
1999 6.3 4.0 3.6 1.8
1998 6.7 4.4 4.4 2.8
1997 7.4 4.9 3.7 2.3
1996 8.7 5.9 4.3 2.5
1995 7.4 4.5 4.3 2.2
1994 8.4 5.6 3.8 1.9
1993 9.3 6.7 5.4 3.2
1992 12.6 8.2 6.8 3.4
1991 11.2 7.0 5.6 2.5
1990 11.4 7.2 4.9 2.4
Ave. 8.8 5.6 4.9 2.6

Table 11-7. Break Rates Using Modified Method for the Period of 2000-2008

Material Diameter(inches) BreakRate


(Breaks/100/miles/year)
All All 4.3
All 4 20.8
CastIron 4 23.3
CastIron 6 7.3
CastIron 8 2.6
CastIron 10 0.9
CastIron 12 0.9
CastIron 16 1.0

Page | 165
Appendix: Seattle Water System
Table 11-8. Water Main Repairs on 4-inch and Larger Pipe Requiring Pipe Segment Re-
placement

DIAM
ETER YEAR NATUREOFFAILURE ADDRESSANDNOTES
CONSTRUCTION
12 2006 REWORK 6400SANDPOINT WAYNE,SOFTFAILUREONLEADEDHB&BLOCKING

12 2005 DAMAGEDBYEXCAV 3500SGRAHAMST,SOUNDTRANSITUNDERMINEDMAIN,ALLOWEDJT


SEPARATION
8 2005 DAMAGEDBYEXCAV SOUNDTRANSIT
8 2005 DAMAGEDBYEXCAV SOUNDTRANSITTWOPOINTLEAKSRESOLVEDW/SHORTSEGREPL

8 2005 DAMAGEDBYEXCAV SOUNDTRANSITTWOPOINTLEAKSRESOLVEDW/SHORTSEGREPL

4 2005 DAMAGEDBYEXCAV SMSPLITATCORPWHENRENEWING


8 2005 DAMAGEDBYEXCAV 5400MLKJRWAYS,SOUNDTRANSIT,EXVCRUSHEDPIPE
8 2005 DAMAGEDBYEXCAV 5300MLK JRWAYS,SOUNDTRANSIT
6 2005 DAMAGEDBYEXCAV SCLAUGEREDINTO6"STUB
4 2005 DAMAGEDBYEXCAV PLATEDROPPEDONMAIN,REPL'DSHORTSEG
12 2005 HORIZONTALSPLIT 36004THAVES,SPLITDURINGCOMPACTIONOF2"SVC
8 2006 DAMAGEDBYEXCAV 3700NW61STST,BNSFRWYCONTRACTOR
8 2006 DAMAGEDBYEXCAV SUBSTANDARDREPAIR
8 2007 HORIZONTALSPLIT 2499FRANKLINAVEE,1928CIP,CRACKFROM2"CORPTO8"GV

8 2007 HORIZONTALSPLIT HAIRLINE@CORP


8 2007 HORIZONTALSPLIT HAIRLINE@CORP
8 2007 HORIZONTALSPLIT HAIRLINE@CORP
8 2007 HORIZONTALSPLIT HAIRLINE@CORP;NEWTAPCOLDPATCHARTERIAL
12 2007 HORIZONTALSPLIT 200N195ST,1942CIP,SHEARSTRESSFRACT@CUT/FILLMARGIN

24 2007 HORIZONTALSPLIT UNIVERSITYBRSOAPPROACH,IMPOSEDLOADSTRESSEDCIP


16 2007 HORIZONTALSPLIT 1040026AVESW,1957MJCIP,CLEANFRACTOFGOODPIPE
HORIZONTAL
8 2005 HORIZONTALSPLIT 944020AVESW,1938CIP
8 2005 HORIZONTALSPLIT 3200SWAVALONWAY,1942CIP,CORPS
8 2005 HORIZONTALSPLIT 1200NW126THST,1968SJCIP,CONCRCULVERTLOADING
8 2005 HORIZONTALSPLIT 1129671AVES,1953CIP,SMALLSPLITBEG'GATBELL
12 2005 HORIZONTALSPLIT 2399WCOMMODOREWY,1972CIP,AT2"CORP
4 2005 HORIZONTALSPLIT SMALLSPLITATBELLJT
8 2005 HORIZONTALSPLIT 1020029THAVNE,MINORSPLITAT1967HEADERCORP
6 2005 HORIZONTALSPLIT SIGNIFICANTSPLITATCORP

Page | 166
Appendix: Seattle Water System
16 2005 HORIZONTALSPLIT 8510SANDPOINTWAYNE,1959CIP,HILLSIDEHOMESCONST'D1978,84

4 2006 HORIZONTALFRACTURE MINORSTRESSFRACTUREEXTENDINGFROMBELL


4 2006 HORIZONTALFRACTURE MINORSTRESSFRACTUREEXTENDINGFROMBELL

DIAM
ETER YEAR NATUREOFFAILURE ADDRESSANDNOTES
HORIZONTALCONTINUED
8 2006 HORIZONTALFRACTURE 2500SAMERICUSST,1968SJCIP
6 2006 HORIZONTALFRACTURE MINORSTRESSFRACTUREEXTENDINGFROMBELL
8 2006 HORIZONTALFRACTURE 13300BAGLEYAVEN,HAIRLINESPLIT,MINORDISCHG,1964SJCIP

12 2006 HORIZONTALFRACTURE 3300WCOMMODOREWY,1944CIP,SPLITAT19752"CORP


8 2006 HORIZONTALFRACTURE 2100N193ST,HAIRLINESPLIT,MINORDISCHG,1953MJCIBELL
6 2006 HORIZONTALFRACTURE 1500SW104ST,HAIRLINESPLIT,MINORDISCHG,1942CIBELL
8 2006 HORIZONTALFRACTURE 650836AVENE,1954CIP
GRAPHITIZATION/CORROSION
8 2005 LONGITUDINAL SPLIT 520052NDAVES,1956CIP,LOWRESISTIVITYSOIL
GRAPHITIZED
12 2005 HORIZONTALSPLIT 200 N INNIS ARDEN WAY, 1979 DIP SPLIT W/ 3 DISTINCT POINTS OF
DAMAGE
8 2006 LONGITUDINAL SPLIT 5100SJUNEAUST,1927CIP(2006CLOSUREOF2005INCIDENT)
GRAPHITIZED
12 2007 HORIZONTALSPLIT 3603GILMANDRW, MAJSPLITDURINGCOMPACTNOFNEW2"TAP

8 2007 DAMAGEDBYEXCAV CONTRACTORHIT1.5"SERVICE/CORPDAMAGINGMAIN


20 2007 DAMAGEDBYEXCAV 300DEXTERAVEN,TRACKHOEPAVBREAKERPIERCEMAIN
OTHER
6 2005 SJ/MJBELLLEAK 1400SW137ST,1965CIP,STEEPSLOPE
12 2005 SJ/MJBELLLEAK 1700 13TH AVE SW, THRUST BLOCK ON 1968 CIP ALLOWED 12" MJ SE
PARATION
12 2006 NICK/ABRASION REWORKEDSOFTFAILUREOFRECENTLYINSTALLED12"DI
10 2007 REWORK 300VIRGINIAST,SOFTFAILUREOF1989DAMAGEREPAIR
LEADEDJOINTFAILURE
12 2005 BELL&SPIGOTLEAK 71008AVES,RECONSTR'DANGLEPOINT&BLOCKING
4 2006 BELL&SPIGOTJTLEAK CUTOUTOBSOLETETEE.OBSOLETEMAINSINCERET'D.
4 2007 STEELPIPE,CORROSION RICHMONDBEACHWATERCOKALPIPE,200LF

VERTICAL
8 2005 VERTICALFRACT SPUBROKEMAINHORIZONTALLYDURINGVERTBREAKEXCAV
12 2005 VERTICALFRACT 2200FAIRVIEWAVEE,FIBEROPTICSVAULT&DUCTLOADING1967CISJ
WATERMAIN,

Page | 167

You might also like