UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Data assessment and utilization for improving asset management of small and medium size water utilities 2007

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_2007-268223.pdf [ 29.15MB ]
ubc_2007-268223.pdf
Metadata
JSON: 1.0063271.json
JSON-LD: 1.0063271+ld.json
RDF/XML (Pretty): 1.0063271.xml
RDF/JSON: 1.0063271+rdf.json
Turtle: 1.0063271+rdf-turtle.txt
N-Triples: 1.0063271+rdf-ntriples.txt
Citation
1.0063271.ris

Full Text

DATA ASSESSMENT AND UTILIZATION FOR IMPRO VING ASSET MAN A GEMENT OF SMALL AND MEDIUM SIZE WATER UTILITIES b y A N D R E W W O O D B . A . S c , The U n i v e r s i t y o f B r i t i s h C o l u m b i a , 1987 M . E n g . , The U n i v e r s i t y o f B r i t i s h C o l u m b i a , 1998 A T H E S I S S U B M I T T E D I N P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T F O R T H E D E G R E E O F D O C T O R O F P H I L O S O P H Y i n T H E F A C U L T Y O F G R A D U A T E S T U D I E S ( C i v i l Engineer ing) T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A January 2007 © A n d r e w W o o d , 2007 A B S T R A C T D a t a regarding water m a i n breaks are essential for undertaking informed and effective infrastructure asset management. Th i s thesis reports o n the f indings o f a survey regarding water m a i n break data co l lec t ion practices across N o r t h A m e r i c a and develops an approach for constructing databases and integrating the data w i t h break predic t ion models to improve the asset management practices o f a ut i l i ty . The survey determines the amount and type o f data col lected b y water ut i l i t ies , the leve l o f comfort w i t h the amount o f data col lected and the ava i lab i l i ty o f alternate sources o f data. The responses p rov ide insight into the strategies and data co l l ec t ion practices o f sma l l to mid- s i ze uti l i t ies and show that the amount o f data col lected b y uti l i t ies can be class i f ied b y the degree o f data richness and defined as either an expanded, intermediate, l imi t ed or m i n i m a l data set. U t i l i t i e s can implement recommended practices to increase the amount o f data they col lect , increase effectiveness o f data co l l ec t ion and processing and consider addi t ional sources o f data for water m a i n breaks to improve their data sets. The thesis also introduces an approach for constructing a water m a i n break and general ne twork database that relates data from mul t ip le sources to augment the amount o f data avai lable for asset management analysis w h i l e main ta in ing exis t ing data warehousing practices. W h e n used, managers m a y gain insight into current and future performance o f the dis t r ibut ion network and develop future asset management strategies. The approach is f lexib le , uses c o m m o n l y avai lable software tools and anticipates the evolu t ion o f data co l lec t ion , ver i f ica t ion and storage capabil i t ies w i t h i n the ut i l i ty . F i n a l l y , a f ramework is presented that guides sma l l to m e d i u m water ut i l i t ies i n ident i fy ing k e y data to be used i n asset management and pipe break predic t ion m o d e l i n g i i and i n selecting appropriate water m a i n break predic t ion models . The framework m a y be used to identify the magni tude o f a uti l i ty 's pipe burst problems today and i n the future, enhance the development o f pipe replacement priori t ies based on forecasted breaks and identify k e y data to col lect i n future data acquis i t ion programs. Water ut i l i t ies w i t h va ry ing amounts o f data can easi ly implement it w i t h their exis t ing data management and analysis tools. i i i TABLE OF CONTENTS Abstract i i Table of contents i v List of tables . . . . . . . . v i i List of figures v i i i Acknowledgements x Co-authorship statement x i i 1 Introduction 1 1.1 B a c k g r o u n d 2 1.2 Thesis objectives and organizat ion 7 1.3 Asse t management 9 1.4 Pred ic t ing water m a i n breaks 13 1.5 Research methods 18 1.5 References 29 2 Assessment of water main break data for asset management 37 Preface 38 2.1 Introduction '. 40 2.2 D e s i g n o f the survey '• 42 2.3 Survey results 43 2.4 D i s c u s s i o n and recommendat ions 51 2.5 Conc lu s ions 56 2.6 Acknowledgemen t s 58 iv 2.7 References 59 3 Constructing water main break databases for asset management 75 Preface 76 3.1 Introduction 78 3.2 State o f data i n uti l i t ies .' 80 3.3 Water m a i n break data for asset management 81 3.4 Const ruc t ing water m a i n databases 84 3.5 A water m a i n break database for M a p l e R i d g e , B C 92 3.6 D i s c u s s i o n 98 3.7 Conc lu s ions 100 3.8 Acknowledgemen t s 102 3.9 References 103 4 Using water main break data to improve asset management for small and medium utilities 119 Preface , 120 4.1 Introduction 122 4.2 Water m a i n breaks 123 4.3 A framework for us ing predic t ion models to improve asset management 128 4.4 Break predic t ion models for L a i t y V i e w , M a p l e R i d g e , B C 132 4.5 Conc lus ions 139 4.6 A c k n o w ledgem ents 140 v 4.7 R e f e r e n c e s . . . 141 5 Conclusions and recommendations 153 5.1 S u m m a r y o f research goals . .154 5.2 Conc lus ions 155 5.3 Observations 157 5.4 C l o s i n g remarks 165 References 167 Appendices 179 A p p e n d i x A C o m p a r i s o n o f survey questions for D e b et al. (2002) and W o o d and Lence (2006) '. 179 A p p e n d i x B Statistical analysis for guiding the interpretation o f survey results 183 A p p e n d i x C Desc r ip t ion o f M a p l e R i d g e water system 187 A p p e n d i x D Water m a i n break survey form 193 A p p e n d i x E L i s t o f organizations that were sent a c o p y o f the water m a i n break survey 203 A p p e n d i x F Survey responses 215 A p p e n d i x G Water m a i n break and system data and statistics o f models 393 A p p e n d i x H Degree o f accuracy o f predict ions for p ipe groups 437 v i LIST OF TABLES i Table 2.1 Service popula t ion o f respondents 62 Table 2.2 Percentage o f respondents that record loca t ion data 63_ Table 2.3 A d d i t i o n a l sources o f break-related phys ica l data for respondents .64 Table 2.4 Suggested sources and approaches for co l lec t ing phys i ca l data on water m a i n breaks 66 Table 3.1 Water m a i n break data ava i lab i l i ty for M a p l e R i d g e 107 Table 3.2 Water m a i n breaks for a g iven year o f instal la t ion 109 Tab le 3.3 P ipe breaks for pipes o f a g iven diameter (1983-1999) 110 Table 4.1 T y p i c a l data used i n models and factors for w h i c h they are a surrogate 146 v i i LIST O F FIGURES Figure 1.1 Twenty-year total per household infrastructure cost estimates for different s ized water systems 36 F igure 2.1 Percentage o f respondents that col lect general informat ion ..68 F igure 2.2 Percentage o f respondents that record phys i ca l data 69 F igure 2.3 Percentage o f respondents that record failure causes 70 F igure 2.4 Percentage o f respondents that record repair activit ies 71 F igure 2.5 Percentage o f respondents that col lect different types o f environmental data : 72 F igure 2.6 Percentage o f respondents that expressed confidence i n data col lec ted 73 F igure 2.7 Classes o f data richness among water ut i l i t ies 74 F igure 3.1 A schematic for constructing and us ing water m a i n break data for knowledge d iscovery I l l F igure 3.2 A process for d ig i t i z ing and creating data f rom arch iva l geographical data 112 F igure 3.3 L i n k i n g hydrau l ic m o d e l data w i t h network data 113 F igure 3.4 Buf fe r ing data to create data relationships us ing G I S 114 F igure 3.5 M a p l e R i d g e water m a i n break analysis data web 115 F igure 3.6 C u m u l a t i v e breaks i n L a i t y v i e w area: 1983-1999 116 F igure 3.7 N u m b e r o f years i n service w h e n break occurred i n pipes (1983-1999) 117 F igure 3.8 N u m b e r o f breaks for pipes i n a g iven so i l type instal la t ion (1983-1999) 118 v i i i Figure 4.1 Improv ing asset management us ing pipe break predic t ion models 147 F igure 4.2 Degree o f accuracy o f t ime-l inear and t ime-exponent ia l predict ions for mater ial groups 148 F igure 4.3 Degree o f accuracy o f t ime-l inear and t ime-exponent ia l predict ions for material and diameter groups 149 F igure 4.4 Degree o f accuracy o f t ime-l inear and t ime-exponent ia l predict ions for material , diameter and age groups 150 F igure 4.5 Degree o f accuracy o f t ime-l inear and t ime-exponent ia l predict ions for material , diameter, so i l and age groups 151 F igure 4.6 Degree o f accuracy o f t ime-l inear and t ime-exponent ial predict ions for material , diameter, age and surface mater ial groups 152 i x A C K N O W L E D G E M E N T S I be l ieve that beh ind any w o r k are people and events that supported, inf luenced and shaped it and this is indeed true for m y thesis adventure and journey. ' I a m grateful to Barbara Lence , m y academic supervisor w h o encouraged and supported m y ideas and work , gave me important advice and was patient throughout m y program. Y o u demonstrated b y many long hours, a dedicat ion to mentor ing and I thank y o u for that and for ign i t ing m y passion for research, academia and success. M y other commit tee members w h o mentored and gave me t imely , succinct, gracious and k i n d advice and guidance were A l a n R u s s e l l and J i m Atwater . M y journey was successful because y o u pointed out or ro l l ed stones for me to step on as I crossed the doctoral r iver . Thank y o u for insp i r ing me to connect work , research and c o m m u n i t y service. I can say o f m y supervisory committee, as Isaac N e w t o n penned, " I f I have seen further it is b y standing o n the shoulders o f giants." M y f a m i l y has been instrumental , b y caring, p ray ing and encouraging throughout the journey. Cla re , I c o u l d not have done this without your endless support, patience and be l i e f i n me. I love you . Y o u shared the weight o f the studies i n so many ways and also freed m e to journey through un imagined waters. M a r k and Ju l i a , w h o have o n l y k n o w n their father as a graduate student, hav ing y o u both a long w i t h me has been a j o y , and I a m truly blessed. I have also been supported b y m y mother, Be t ty b y her steadfast prayers and confident and patient hope. I am grateful for the support o f m y employers , M a p l e R i d g e and C o q u i t l a m w h o tasked and then entrusted me w i t h the quest o f i m p r o v i n g their asset management practices. x I trust that this w o r k w i l l support their efforts to sustain their water infrastructure, protect pub l i c health and support their loca l economy. F i n a l l y , I have had the pleasure o f w o r k i n g w i t h m a n y colleagues and fe l low practitioners. O f these, I w i s h to spec i f ica l ly acknowledge W i l s o n L i u . x i CO-AUTHORSHIP S T A T E M E N T Andrew Wood was the lead and principal researcher of the work contained in the thesis titled "Data Assessment and Utilization for Improving Asset Management of Small and Medium Size Water Utilities". Dr. Barbara Lence, the Research Supervisor of the thesis, provided inspiration and supported the writing of the papers and Wilson Liu assisted with the preparation of figures used in the paper submitted as Chapter 3. xii C H A P T E R 1 INTRODUCTION 1.1 B A C K G R O U N D Rel i ab le , efficient and effective water dis t r ibut ion systems are c ruc ia l to pub l i c health and safety. These systems are also essential to the economic we l l -be ing o f m a n y munic ipa l i t i es since manufacturing, industry and commerce re ly to a large degree on obta ining rel iable and economica l ly -p r iced water del ivered through a network o f water pipe l ines, more c o m m o n l y referred to as Water mains . M a n y o f these water mains instal led over the decades i n N o r t h A m e r i c a are n o w beg inn ing to break and fa i l . The tradit ional pub l i c works emphasis on managing water m a i n breaks has been directed toward m i n i m i z i n g the loss o f water to k e y businesses and cr i t ica l facil i t ies and the damage to bui l t and natural infrastructure. H o w e v e r , breaks are also potential gateways to contaminat ion o f the water dis t r ibut ion system ( A W W A and E E S , 2002) and are identif ied as a h igh pr ior i ty i n the assessment o f water supply health r isks b y the N a t i o n a l Research C o u n c i l o f A c a d e m y Sciences (2005). B y replac ing the pipes just before they fa i l , ut i l i t ies reduce their r isks and costs o f water m a i n breaks. T h i s thesis is focused on he lp ing smal l and m e d i u m size uti l i t ies b y p r o v i d i n g informat ion o n water m a i n break data co l lec t ion , construct ion, compi la t ion , and management i n support o f asset management. The research presented includes an approach for these uti l i t ies to use water m a i n break predic t ion models o n a p ipe b y pipe basis for p r io r i t i z ing the replac ing o f water mains. The w o r k is presented as a series o f three manuscripts that are publ i shed i n peer-reviewed journals or are i n the process o f be ing peer- reviewed. W h e n I embarked on this research program m y goal was to assist sma l l and m e d i u m size uti l i t ies w i t h managing their pipe networks to address the r isks arid costs associated w i t h replac ing water mains . Spec i f i ca l ly , I wanted to assist them w i t h ident i fying, 2 . co l lec t ing and construct ing relevant pipe break data to analyze their p ipe network, p r io r i t i z ing their water m a i n replacements b y us ing break predict ions and gu id ing their data acquis i t ion and analysis programs. D u r i n g m y almost twenty years o f experience as a professional engineer i n pub l i c ut i l i t ies I observed that m y colleagues i n other ut i l i t ies were not us ing break predic t ion models i n their practice. W i t h i n the organizations w i t h w h i c h I was famil iar , co l l ec t ion o f water m a i n break data was undertaken o n an ad-hoc basis and storage o f such informat ion was not t yp i ca l ly comprehensive. A m o n g those uti l i t ies, there were no c o m m o n practices or standards for data co l l ec t ion and ve ry few uti l i t ies , i f any, were aware o f the data col lected b y other uti l i t ies. Bes t practices for Canad ian water uti l i t ies regarding w h i c h data to col lect were introduced i n 2002 ( N S G M I , 2002). A s a manager, I also wanted to compare the informat ion I was co l l ec t ing w i t h that o f other ut i l i t ies , to understand h o w data were in fo rming us and to share this knowledge w i t h m y colleagues. D u r i n g this t ime, I also became aware o f the general b e l i e f he ld b y m a n y u t i l i ty managers that asset management is a panacea for s o l v i n g our ag ing water system problems. W e needed to k n o w when , where and h o w m a n y pipes to replace; but w e were focusing on the entire network, not on specific pipes, to obtain funding for programs. I found that among m y colleagues, once funding was obtained, decisions o n p r io r i t i z ing pipe replacements were based o n his tor ica l data, experiences and management po l ic ies rather than o n the expected performance o f specif ic pipes though there are models for predic t ing water m a i n breaks. In fact, many o f m y colleagues were not aware o f these models and none, i f any, were us ing them. Thus , I was conv inced that w e needed an approach that appl ied pipe break models to in fo rm our decisions regarding the pr ior i t iza t ion and schedul ing o f specif ic pipe replacements. 3 A s a result o f m y observations, m y thesis identifies data that are col lected and avai lable for analysis across uti l i t ies i n N o r t h A m e r i c a , develops a methodology for creating and l i n k i n g data across various data sources and develops a f ramework for assisting i n the pr ior i t i za t ion o f water m a i n replacements and data acquis i t ion based on predict ions o f future water m a i n breaks w i t h i n a g iven water dis t r ibut ion system. The framework m a y be appl ied across the range o f data avai lable i n typ ica l water ut i l i t ies, acknowledges exis t ing industry needs and practices and W i l l help managers identify and use k e y avai lable break data. Our water mains are aging and failing. The water mains o f m a n y munic ipa l i t i es are predominant ly o f the 1960s to 1990s vintage because o f extensive urban development dur ing those years, w i t h some smaller amounts pr ior to that per iod. H o w e v e r , the replacement needs o f aging pub l i c works infrastructure has become the focus o f governmental and academic attention i n recent years. Because few water ut i l i t ies have ac t ive ly pursued aggressive water m a i n rehabil i ta t ion or replacement programs, they are n o w facing the p rob lem o f replac ing or rehabil i tat ing their aging systems (Deb et al., 2002). The U n i t e d States Env i ronmen ta l Protect ion A g e n c y ( U S E P A ) identifies that w h i l e U n i t e d States ( U S ) communi t ies spent one t r i l l i o n U S dollars i n 2001 o n d r ink ing water treatment, supply and waste water treatment and disposal , this expenditure l eve l m a y not be sufficient to keep pace w i t h future infrastructure needs ( U S E P A , 2002). T h e y also identify i n a 2003 survey that the 53,000 c o m m u n i t y water systems and 21,400 not-for-profit non-communi ty water systems i n the U S w i l l need an estimated $276.8 b i l l i o n to continue to p rov ide their services ( U S E P A , 2005). S i m i l a r l y , the Ontar io M i n i s t r y o f P u b l i c Infrastructure R e n e w a l (PIR) identifies that over the next 15 years, $25 b i l l i o n w i l l be needed for capital renewal o f Ontario 's $72 b i l l i o n water and waste water assets (Ontario P I R , 2005). The Federat ion o f 4 Canadian M u n i c i p a l i t i e s ( F C M ) state that k e y investments must be made i n core pub l i c infrastructure to manage waste and water systems, to meet pressing environmental and air qual i ty needs and to main ta in the economic health o f Canada's communi t ies ( F C M , 2001). G i v e n the current l eve l o f expenditures and the increasing age o f systems, ut i l i t ies need to predict breaks and use this informat ion to pr ior i t ize and p lan w h e n pipes should be replaced. The ab i l i ty to respond to the challenge o f k n o w i n g w h e n to replace a p ipe varies among water ut i l i t ies . S m a l l to m e d i u m size uti l i t ies m a y f ind it par t icular ly cha l lenging due to their l imi t ed resources and technical capaci ty and there are m a n y o f these ut i l i t ies across N o r t h A m e r i c a . In the U S , smal l water systems make up 90 percent o f those 53,000 c o m m u n i t y water systems ( A m e r i c a n Socie ty o f C i v i l Engineers , A S C E , 1999). N o analogous data exist for Canada, al though the Canad ian N a t i o n a l Research C o u n c i l ( C N R C ) identifies that there are 3500 munic ipa l i t ies serving fewer than 5000 people and that there are o n l y 63 munic ipa l i t i es serving populat ions greater than 100,000 (Van ie r and R a h m a n , 2004). M o s t sma l l to m e d i u m size water uti l i t ies and munic ipa l i t i es share s imi la r characteristics. T h e y purchase bu lk water f rom a regional or larger supplier and require some treatment. The treatment processes are s imple processes such as ozonat ion, ultra- viole t ( U V ) or chlor ine dis infect ion. S m a l l u t i l i ty systems are t yp i ca l l y compr ised o f smal l diameter pipes and.as noted b y Ket t le r and Goul te r (1985), smal ler pipes break have been observed to break more frequently than larger ones. T h i s m a y be because their beam strength and w a l l thickness are general ly less than those o f larger pipes. 5 Smal le r ut i l i t ies have scarce resources and technical expertise is t yp i ca l l y riot avai lable because staff are general ly expected to undertake a number o f different functions and do not have the luxury o f deve lop ing in-house expertise. C o m p o u n d i n g this is the fact that w i t h i n these systems, there is lit t le or no rel iable documentat ion o f the locat ion, capacity, cond i t ion and adequacy o f p ipe network elements for meet ing present or future needs ( M y e r s , 2001). These uti l i t ies m a y also lack the resources, f inanc ia l ly and organizat ional ly , to implement a complex informat ion management system program, nor the h is tor ica l data or tools to fu l ly analyze their system. T h e y require capital to rehabilitate, upgrade and instal l infrastructure, but face an economic challenge i n pay ing for these costs g iven a smal l revenue base. F o r example, al though the costs o f a sma l l water system m a y be modest compared w i t h those o f large systems, the per household costs are s igni f icant ly higher than those o f larger systems ( U S E P A , 2001). The data shown i n F igure 1.1 are the per household costs for different s ized water systems for meet ing anticipated needs. These data show that households i n smal l systems face costs that are over three and one h a l f t imes those o f households serviced b y large systems. Ut i l i t i e s are v i e w i n g asset management as an approach for addressing this d i l e m m a o f p lann ing pipe replacements b y understanding h o w m u c h rehabil i ta t ion w i l l cost, what to do first and w h e n to rehabilitate their systems. Asse t management is a business administrat ion approach to dec i s ion-making that covers an extended t ime hor i zon , draws from economics as w e l l as engineering science and considers a broad range o f assets. The approach incorporates the economic assessment o f alternative investment options and uses this information, to help make cost-effective investment decisions (Uni ted States Federal H i g h w a y Admin i s t r a t ion , U S F H W A , 1999). The benefits o f app ly ing an asset management 6 approach inc lude the ab i l i ty to l i n k user expectations and needs and to identify the means o f assessing value, system condi t ion , performance, service l ife and management and investment strategies. H o w e v e r , for smaller uti l i t ies to adopt asset management, they need portable, read i ly useable approaches that require l i t t le modi f ica t ion , are l i k e l y to be met w i t h litt le organizat ional resistance and serve to incrementa l ly improve water m a i n replacement analysis and planning. The need for p lann ing is urgent and the need to improve the p lann ing approaches w i l l o n l y increase w i t h t ime. 1.2 THESIS O B J E C T I V E S AND O R G A N I Z A T I O N T h i s thesis: • assesses the state o f data avai lable to smal l arid m e d i u m size water ut i l i t ies , • develops an approach for constructing and c o m p i l i n g water m a i n break data for analysis and • develops a f ramework that guides the ident i f icat ion o f k e y data for asset management and the select ion o f the most appropriate data and models for predic t ing water m a i n breaks. Organization of thesis. The thesis is presented as a series o f three manuscripts that are publ i shed i n peer-reviewed journals or are i n the process o f be ing publ i shed or peer- reviewed. E a c h o f the three manuscript chapters (Chapters 2 to 4) contains a preface, introduct ion, ove rv i ew o f the literature, body, conc lus ion and references. Tables and figures referenced i n each chapter are inc luded at the end o f that chapter and a l l appendices are p rov ided at the end o f the thesis. F i n a l l y , Chapter 5 summarizes the major conclus ions o f this work . 7 There are m a n y benefits o f a manuscript-based thesis over the tradit ional thesis and the p r imary benefit is that research f indings are publ i shed and shared w i t h the research commun i ty m u c h earlier than i n the product ion o f a tradit ional thesis. H o w e v e r , there are also significant disadvantages. Literature reviews are often repeated, some cont inui ty is lost and m u c h o f the data, results, and observations that w o u l d be i n a t radit ional thesis are condensed or e l iminated. T h i s results i n an abbreviated thesis and a percept ion b y the reader o f a lesser effort than was the case i n the research. T h i s thesis attempts to address these l imita t ions b y i nc lud ing a section i n this chapter on the approach, methodology and context o f the research. E a c h chapter preface provides the context for the w o r k presented i n that chapter (paper) w i t h respect to the overa l l research and thesis and the preceding chapter. D a t a not p rov ided i n the papers are also inc luded i n the appendices. Chapter 2 provides the results o f a survey that investigates the state o f data co l lec t ion practices, the c o m m o n l y avai lable water m a i n break data and the data sources w i t h i n N o r t h A m e r i c a n water uti l i t ies. The paper was publ i shed i n the J u l y 2006 issue o f the Journal o f the A W W A . A compar ison o f the survey questions and those o f D e b et al. (2002) is presented i n A p p e n d i x A . Chapter 3 presents an approach for b u i l d i n g and relat ing data from various sources for analysis, based on observations d rawn from the survey responses reported i n Chapter 2. Th i s w o r k demonstrates h o w to construct water m a i n data and databases w h i c h are based on l i n k i n g , relat ing, extracting and c o m p i l i n g data f rom sources internal and external to a u t i l i ty for the purpose o f analysis, or i n other words relating relational databases. Issues related to creating, l i n k i n g , t ransforming, cleansing, scrubbing and integrating data are identif ied and approaches for addressing them are presented. A number o f databases were created i n this por t ion o f the research. These are summar ized i n A p p e n d i x G . T h i s paper 8 was submitted to the Journal o f the A W W A i n February 2006, revised i n October 2006 and is scheduled for pub l ica t ion i n January 2007. Chapter 4 develops a framework that guides uti l i t ies i n ident i fy ing k e y data to be used i n asset management i n general and spec i f ica l ly i n pipe break predic t ion mode l ing and i n selecting the most appropriate mode l for predic t ing water m a i n breaks. It provides the u t i l i ty w i t h a method for consider ing future p ipe breaks i n the analysis o f p ipe pr ior i t iza t ion strategies. It incorporates exis t ing tools for data management and analysis that are w i d e l y avai lable and easy to implement b y smal l to m e d i u m size uti l i t ies w i t h va ry ing amounts o f data. T h i s paper was submitted to the A S C E Journal o f Infrastructure Systems i n Ju ly , 2006. The fifth and f inal chapter summarizes the conclus ions o f and discussions ar is ing from the research. A d iscuss ion o f future research is also presented. 1.3 ASSET M A N A G E M E N T S i x components o f an asset management p rogram have been suggested (Vanie r , 2001). These are inventory, asset value, deferred maintenance, condi t ion , service l ife predic t ion, and pr ior i t iza t ion o f rehabil i tat ion, replacement and renewal . Inventory. The first b u i l d i n g b l o c k o f asset management is to determine the type, compos i t ion , quantity and extent o f the assets. The p r imary capi tal assets that typ ica l water ut i l i t ies o w n inc lude water supply reservoirs, dams and support ing hydrau l ic structures, water treatment plants, water dis t r ibut ion systems inc lud ing pipes, valves , fire hydrants, d is t r ibut ion reservoirs, pump stations, pressure reducing valves and meters to measure distributed and purchased water. U t i l i t i e s are t yp i ca l ly main ta in ing inventories us ing tools such as Geographic Information Systems (GISs) , Compute r A i d e d Draf t ing ( C A D ) systems and relat ional databases. These systems require significant effort and resources to P implement and mainta in but for smal l systems, a s imple spreadsheet m a y be a l l that is required. W o o d and Lence (2G06) found that a rchiva l records are s t i l l the predominant sources o f data for ut i l i t ies that serve populat ions fewer than 50,000. In recent years, the use o f Compute r i zed Main tenance Management Systems ( C M M S s ) has also g r o w n i n popular i ty and these systems usua l ly interface w i t h or use G I S data. These systems are expensive to implement and require significant resources to mainta in . M a n y uti l i t ies s t i l l do not have C M M S s . Asset value. T o p lan for replacement, water u t i l i ty managers must k n o w the total value o f their assets. Managers must have a basic understanding o f system wor th and h o w m u c h is required to replace it. A c c o r d i n g to V a n i e r (2001), the s ix c o m m o n l y used terms to describe asset value are his tor ical , appreciated his tor ical , capital replacement, performance- in-use, depr iva l cost and market value. H i s t o r i c a l value is defined as the or ig ina l or book value, appreciated h is tor ica l value is the his tor ica l value calculated i n current dollars and capital replacement value is the cost o f the asset i n current dol lars . The performance-in-use value is the prescribed value o f the actual asset to the user. D e p r i v a l cost is the cost that w o u l d be incurred i f depr ived o f the asset but s t i l l required to de l iver the service. F i n a l l y , market value is the value i f the asset is so ld on the open market. U t i l i t i e s m a y not use a l l o f these values. F e w munic ipa l i t i es have specific asset values. In most cases, ut i l i t ies inherit their new infrastructure from developers who instal l it as part o f a hous ing subd iv i s ion and then transfer the ownership and maintenance o f the assets to the munic ipa l i ty . O n l y the book value o f the assets constructed di rect ly b y the mun ic ipa l i t y is t yp i ca l ly recorded. F o r many uti l i t ies , the value o f their assets are s i m p l y calculated b y m u l t i p l y i n g an average cost o f construct ion per metre o f p ipe b y the total length o f the network. Condition. Asse t management requires knowledge o f the cond i t ion o f the assets. In the past, a significant amount o f data were captured for evaluat ing operational values and objectives and a s imple spreadsheet was used to record these data. In recent years, C M M S have been used to warehouse and process informat ion and the management o f maintenance act iv i ty p lanning . H o w e v e r , asset managers face the p rob lem o f determining h o w and what to evaluate, o f def in ing what constitutes condi t ion and indices , o f p r o v i d i n g for data storage and o f determining h o w to use condi t ion data. N o t o n l y are there are no standard cond i t ion indices for water mains ( G r i g g , 2004) , but the challenge for m a n y water ut i l i t ies is that condi t ion assessments have to be on-going. T h e y can be cost ly and o n l y p rov ide informat ion for that asset at a specific t ime. There is also a need for a significant improvement i n the level o f accuracy i n p ipe l ine condi t ion assessment and accurate predic t ion o f p ipe l ine failures ( D e S i l v a et al, 2002). Deferred maintenance. Deferred maintenance is maintenance that has not been performed or has been deferred. F o r ut i l i t ies , managing deferred u t i l i ty maintenance requires ident i fy ing and managing the maintenance that is to be deferred and the compound ing effect o f maintenance that has not been performed such as not c leaning or repair ing water mains. The challenge for water u t i l i ty asset management l ies i n managing and integrating the amount o f deferred maintenance between the various components that make up a system w i t h the subjective nature o f h o w a maintenance b a c k l o g is calculated. F o r example, most pipe networks are segmented and o f differ ing age, condi t ion , value and consequence o f failure. De te rmin ing quanti tat ively the maintenance b a c k l o g o f that 11 network is ve ry dif f icul t wi thout accounting for reservoirs, pump stations and other assets. Thus there is ve ry litt le informat ion regarding deferred maintenance o f water dis t r ibut ion systems and few i f any munic ipa l i t ies can determine their deferred maintenance. Remaining service life. M a n a g i n g water u t i l i ty assets and p lann ing replacement requires estimates o f technical and economic service l i fe . Asse t managers face funding constraints and compet ing needs, thus op t im iz ing the expenditures f rom technical and economic perspectives is important. W h i l e technical l i fe m a y be diff icul t to predict, determining economic service l ife is m u c h s impler since it compares the immediate costs o f repairs w i t h the costs o f renewal . So the question that is often raised as part o f the understanding and analysis o f p ipe deterioration is w h e n does a p ipe reach the end o f its useful l ife? One def ini t ion o f w h e n a pipe reaches the end o f its useful l i fe is w h e n the pipe is replaced b y a ut i l i ty . T h i s can be mis lead ing since a u t i l i ty ' s dec i s ion to replace a p ipe m a y be based o n other factors such as po l i t i ca l choice , perception o f the re l i ab i l i ty o f the p ipe or the need to increase the hydraul ic capacity o f the pipe. Rajan i and M a k a r (2000) define the t ime o f death o f a pipe as the t ime at w h i c h its mechanica l factor o f safety falls b e l o w an acceptable value set b y the ut i l i ty . K l e i n e r and Rajani (1999) propose that the useful l i fe o f a p ipe is a function o f the economic costs o f deterioration rate and replacement and suggest that p ipe death coincides w i t h the op t imal t ime o f replacement. Statistics Canada defines the service l ife o f an asset as its useful l i fe at the t ime o f its acquis i t ion w h i c h general ly ends at demol i t ion (Statistics Canada, 2006). A c c o r d i n g to D e b et al. (1998), there are no standardized methods for predic t ing the l ife o f dis tr ibut ion systems. 12 Prioritizing. F i n a l l y , effective management o f water u t i l i ty assets requires the pr ior i t i za t ion o f replacement needs. That is , managers must be able to determine what to replace and w h e n to replace it. T h i s issue is c lose ly t ied to r e so lv ing f inancia l and technical challenges as to whether to mainta in , repair or renew an asset or to choose an alternative ) such as t w i n n i n g a water ma in . Because u t i l i ty managers are usua l ly required to p lan annual capital projects, this component o f asset management is usua l ly performed. In do ing so, they have to overcome obstacles to effective pr ior i t i za t ion such as h o w to address uncertainty w h e n longer-term p lann ing hor izons are considered or h o w to balance the various needs among an organizat ion. Tradi t iona l ly , ut i l i t ies have pr io r i t i zed p ipe replacements based on a combina t ion o f current management practices and his tor ical pipe breakage data. Management practices include directives based on general guidel ines, consequence assessments, legislat ive requirements and other u t i l i ty priori t ies . Rudimenta ry analyses that interpret h is tor ica l pipe break data, i nc lud ing locat ion, t ime and date o f break and pipe diameter and material have p rov ided informat ion regarding where and h o w m a n y breaks are occur r ing and what pipes are exper iencing breaks (K le ine r and Rajan i , 1999). 1.4 PREDICTING W A T E R M A I N B R E A K S A p r imary goal o f p r io r i t i z ing pipe replacements is to identify investment strategies that, o n one hand, avo id premature replacement o f pipes (i.e., unnecessary pre-investment o f funds), and o n the Other hand, avo id water m a i n breaks, interruptions i n service, potential contaminat ion o f water and the costs o f damage. I f ut i l i t ies can predict when and where water mains m a y break, this informat ion is also useful for assisting w i t h op t im iz ing 13 crew efforts and m i n i m i z i n g the results o f loss o f water to k e y businesses and cr i t ica l facil i t ies. Thus asset management decisions can be i m p r o v e d b y an abi l i ty to determine the future performance o f water mains b y predic t ing water m a i n breaks and poss ib ly ident i fy ing h o w such breaks m a y occur. Causes of breaks. A number o f authors have analyzed the causes o f breaks, * i nc lud ing O ' D a y (1982), M a l e et al. (1990), Sav ic and Wal ters (1999), Rajan i and M a k a r (2000), Rajan i and K l e i n e r (2001) and D i n g u s et al. (2002). A c c o r d i n g to Rajani and Tesfamar iam (2005), a combina t ion o f circumstances leads to pipe failure i n most cases and different factors cause failure i n different pipe networks. The causes o f breaks inc lude deterioration as a result o f use (e.g., internal corrosion) , phys i ca l loads appl ied to the pipe (e.g., traffic, frost), l im i t ed structural resistance o f the p ipe because o f construct ion practices dur ing instal la t ion and dec l in ing resistance over t ime (e.g., cor ros ion , aging factors). D ingus et al. (2002) surveyed the 46 largest A m e r i c a n Water W o r k s A s s o c i a t i o n Research Founda t ion ( A W W A R F ) member uti l i t ies i n 1997 and noted mul t ip le c o m m o n failure modes for cast i r o n p i p i n g systems. Cor ros ion , improper instal la t ion and ground movement were the three most c o m m o n causes o f pipe failure. Break prediction models. B reak predic t ion models have been developed to help the water industry understand h o w pipes deteriorate and w h e n pipes w i l l break i n the future. These models are typ ica l ly grouped into two classes - statistical and phys ica l - mechanica l models (K le ine r and Rajani , 2001). Statist ical models use his tor ica l p ipe break data to identify break patterns and extrapolation o f these patterns to predict future pipe breaks, or degrees o f deterioration. Phys ica l -mechan ica l models predict failure b y 14 s imula t ing the phys i ca l effects and loads o n pipes and the capaci ty o f the p ipe to resist failure over t ime. Statist ical models have been used to analyze large dis t r ibut ion systems (e.g., K l e i n e r and Rajan i , 1999) and are t yp i ca l ly characterized as either statistical determinist ic or statistical probabi l i s t ic equations (Kle ine r and Rajani , 2001). U n d e r the statistical determinist ic models , the p ipe breakage is estimated based o n a fit o f pipe breakage data to various time-dependent equations, w h i c h m a y represent the cumula t ive pipe breaks as a function o f t ime from date o f instal la t ion or f rom the earliest date o f avai lable break data, and most c o m m o n l y are t ime-l inear (Ket t ler and Goul ter , 1985) or t ime-exponent ial functions (Shamir and H o w a r d , 1979; W a l s k i , 1982 and K l e i n e r and Rajan i , 1999). Determinis t ic models were developed as early as 1979. Statist ical probabi l is t ic models predict not o n l y the failure potential , but the dis t r ibut ion o f failure. These models are more complex than determinist ic models and require more data. E x a m p l e s o f these include cohort su rv iva l , such as K A N E W (Deb et al, 2002), B a y e s i a n diagnost ic , break clustering, s e m i - M a r k o v C h a i n and data f i l ter ing methods. S u r v i v a l analysis has been demonstrated b y M a i l h o t et al (2000) as useful i f there are adequate histories o f pipe failures. Statist ical models can use avai lable h i s to r i ca l ' data on past failures to identify breakage patterns and are useful i f the data are l imi ted . Statist ical methods require some technical expertise i n deve lop ing the models and interpreting results, but not to the degree o f expertise required o f staff i f phys i ca l - mechanica l models are used. K l e i n e r and Rajani (2001) suggest that statistical models based on fewer data m a y be used to gain insights for future performance. 15 Phys ica l -mechan ica l models t yp i ca l ly fa l l into one o f two classes: determinist ic models w h i c h estimate pipe failure based o n s imula t ion o f the phys i ca l condi t ions affecting the p ipe (Doleac et al., 1980 and Rajani and M a k a r , 2000), and probabi l i s t ic models that use a dis t r ibut ion o f input condi t ions, such as rate o f corros ion, to predict the l i k e l i h o o d and dis t r ibut ion o f p ipe failure ( A h a m m e d and Me lche r s , 1994). P h y s i c a l models have been developed p r i m a r i l y for cast i ron and cement pipes and have significant data needs. K l e i n e r and Rajani (2001) suggest that o n l y larger diameter mains w i t h cos t ly consequences o f failure m a y jus t i fy the required data co l lec t ion efforts for these models . r Other developments i n recent years inc lude the use o f A r t i f i c i a l N e u r a l N e t w o r k s ( A N N s ) b y Sac lu t i (1999) and fuzzy-based techniques b y K l e i n e r et al. (2004). The A N N m o d e l predicts the number o f water m a i n breaks based o n a seven day weather forecast and is appl ied o n l y to homogeneous groups o f water mains for short-term maintenance work . The fuzzy-based techniques are appl ied to large t ransmiss ion water mains . In addi t ion, an agent-based system for predic t ing water m a i n breaks is proposed b y D a v i s (2000). W h i l e m u c h w o r k has been undertaken toward deve lop ing deterioration models , the use o f these models is not c o m m o n among uti l i t ies. U t i l i t i e s face obstacles such as a lack o f k e y data, l imi t ed inst i tut ional capaci ty w i t h i n their organizat ion to understand and use models and the complex i ty i n managing w o r k that i n v o l v e s a number o f groups. There is no c o m m o n m o d e l that m a y be appl ied to every water system. Because the literature suggests that breaks and causes o f breaks for any part icular water dis t r ibut ion system are system- specific (Rajani and Tesfamar iam, 2005), a u t i l i ty must create its system-specific mode l based o n factors o f deterioration that are relevant for the ut i l i ty . 16 S m a l l and m e d i u m size uti l i t ies typ ica l ly have the capaci ty to use statistical determinist ic models , but the implementa t ion o f more sophisticated models is not pract ical due to the data co l l ec t ion efforts and m o d e l maintenance required. A s this research demonstrates, statistical determinist ic models show promise for use b y these uti l i t ies because they can be appl ied us ing c o m m o n l y used software, they do not require specia l ized tools or expertise to operate, and m a y b y their degree o f accuracy prov ide managers w i t h insights into their system o n future pipe breaks. U l t ima te ly , predic t ion models need data and i n particular, data that are relevant for the models and are explanatory for accurate predict ions. Da ta co l l ec t ion regarding water m a i n breaks is not a s imple exercise nor is the practice consistent across ut i l i t ies . F o r a l l ut i l i t ies data co l lec t ion can have significant costs i f performed at a comprehensive leve l . These costs inc lude not o n l y out o f pocket costs but also organizat ional effort and human resources. W h i l e a co l l ec t ion o f best practices have been recommended b y the N a t i o n a l G u i d e to Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R F (Deb at al., 2002), the current state o f knowledge o f data col lected b y sma l l and m e d i u m size uti l i t ies is l imi ted . Recent water u t i l i ty surveys include those init iated b y the N a t i o n a l Water and Wastewater B e n c h m a r k i n g Initiative (Earth Tech , 2004), the A m e r i c a n Water W o r k s Assoc i a t i on , ( A W W A , 2004) and A W W A R F (Deb at al., 2002). The N a t i o n a l Water and Wastewater B e n c h m a r k i n g Initiative, a partnership o f a number o f Canad ian uti l i t ies, examines water m a i n breaks as a performance measure but does not examine water m a i n break data i n detail . The A W W A database ( A W W A , 2004), c o m p i l e d as a jo in t effort between A W W A and A W W A R F , focuses on general ized data for dis t r ibut ion systems, 17 ut i l i ty revenue, treatment practices and finances and does not inc lude data o n water m a i n breaks except the number o f breaks. The survey b y D e b et al. (2002) reports on responses f rom five ut i l i t ies serving populat ions fewer than 100,000 and 32 w i t h populat ions greater than 100,000. Thus knowledge o f what smal l and m e d i u m size uti l i t ies have is needed. ( S m a l l and m e d i u m size uti l i t ies need help w i t h m a k i n g predict ions o f the remain ing service l ife o f water mains and guidance for integrating data acquis i t ion and analysis programs to improve the pr ior i t i za t ion o f their water m a i n replacements. T h e y lack knowledge regarding h o w to develop databases that support their efforts i n predic t ing future breaks and what to do w i t h break predict ions. In particular, research is needed to determine the extent o f data col lected, h o w the data should be c o m p i l e d and what can be done to support the use o f models w i t h i n these uti l i t ies. 1.5 R E S E A R C H M E T H O D S Institutional factors affecting asset management. The research goals were in i t i a l l y focused o n des igning and conduct ing experiments to determine k e y data avai lable i n smal l and m e d i u m size uti l i t ies and to assist these data w i t h us ing break predic t ion models that are appl ied o n a pipe b y pipe basis. A s the research progressed, it became apparent that an understanding o f u t i l i ty organizat ional structures, behaviors and practices, as w e l l as a wi l l ingness o n the part o f organizat ion was needed i f this research is to influence practice. T h i s is because water ut i l i t ies are complex organizations w i t h many staff that have var ious roles, responsibi l i t ies and accountabil i t ies. In addi t ion, these organizations face a significant amount o f staff retirements and "insti tutional m e m o r y loss" 18 i n the c o m i n g years. Pract ices that are easi ly taught and documented w i l l be important for business continuity. Furthermore, ut i l i t ies also cannot re ly so le ly on ora l corporate knowledge and need data that are shared among the organizat ion as w e l l as systems that facilitate data sharing. M o s t ut i l i t ies are t yp i ca l ly compr ised o f operations, f inancial , informat ion technology and engineering departments. Operations departments are responsible for operation and maintenance activit ies, f inancial departments manage the f inancia l aspects o f the u t i l i ty business, informat ion technology departments main ta in and manage data systems, w h i l e engineering departments are responsible for the expansion or construct ion o f new network assets. In m a n y uti l i t ies, operating departments are responsible for determining w h i c h pipes should be replaced. C o m m o n l y activit ies, roles and responsibi l i t ies evolve to fit a part icular person, their strengths and interests, though these m a y be inconsistent w i t h the general organizat ional structure. Da ta are usual ly col lected and managed o n an ad-hoc basis throughout the organizat ion b y those w h o are i n v o l v e d i n a specific ac t iv i ty or w h o have specific interest i n the data. N o one single person t yp i ca l l y coordinates or manages a l l the data ( typ ica l ly asset management ini t iat ives stall i n m a n y organizat ions due to the inab i l i ty to dedicate or create a pos i t ion for this role). Furthermore, w i t h i n any u t i l i ty , data m a y be ora l or documented and the format and storage o f such data m a y differ throughout that ut i l i ty . In addi t ion, budgets constrain activities i n these organizations. Dec i s ions and changes i n these organizations are usua l ly based o n dialogue among and acceptance b y those affected. W h i l e practices require approval b y senior managers, pub l i c p o l i c y decisions are referred to m u n i c i p a l counci ls . The pr ior i t iza t ion and schedul ing o f asset replacement projects are usua l ly performed b y groups o f ind iv idua l s or i n some cases b y 19 one i n d i v i d u a l and submitted to their m u n i c i p a l c o u n c i l for approval . These factors required deve lop ing research methods that cou ld be appl ied to a variety o f organizat ional structure, technical d i sc ip l ine , knowledge , s k i l l , ab i l i ty and size. N The District o f Maple Ridge. The Dis t r ic t o f M a p l e R i d g e , the case study for the research is a typ ica l water ut i l i ty . M a p l e R i d g e was incorporated as a mun ic ipa l i t y i n 1874 and experienced significant urban development over the past three decades. It was used to test the practicali t ies o f app ly ing the framework and approaches developed herein. M a p l e R i d g e was interested i n learning and i m p r o v i n g its practices. B u t it was a challenge as a researcher to k n o w , not o n l y for M a p l e R i d g e but for uti l i t ies i n general, what data were avai lable, to ident ify potential sources o f data, to determine h o w to access the data, to develop approaches to create useful data from various sources, and to develop a framework for associating, accessing, and managing the data across a l l functional l ines respecting responsibi l i ty , management and territorial issues. F o r example, i n M a p l e R i d g e , data are "owned" and managed b y various departments. Water m a i n break data are col lected, stored and managed b y those operating crews that respond to breaks. Other p ipe related data such as p ipe mater ial , diameter and age are stored i n the Eng inee r ing department. The operations department manager w h o oversees the superintendents and operating crews prepares and submits a list o f replacement projects to the ind iv idua l tasked as the corporate capital p rogram coordinator who then schedules those projects a long w i t h a l l other capital projects (that are submitted b y a l l other departments). The schedul ing respects the avai lable annual funding and is based o n the submitted informat ion and the coordinator's v i e w o f the future needs o f the organizat ion. The scheduled projects are then rev iewed b y the senior managers and the C h i e f Admin i s t r a t ive Off icer (more c o m m o n l y k n o w n as " C i t y Manage r " i n many munic ipa l i t ies ) 20 pr ior to submiss ion to m u n i c i p a l counc i l for approval . Thus the l i se o f M a p l e R i d g e as a case study required the research methods that c o u l d identify data i n mul t ip le sources, val idate his tor ic projected and estimated data, determine the basis o f col laborated decisions and dis t inguish p o l i c y and practice. Research time frame and tools. The research i n M a p l e R i d g e began i n early 2002 and was completed b y mid-year 2006. It is based o n three components: f irst ly, on conduct ing a survey o f N o r t h A m e r i c a n water ut i l i t ies, secondly, o n creating, l i n k i n g and m i n i n g data and f inal ly , o n developing the framework for us ing predic t ion models . The data used i n the experiments are actual water system data from the Dis t r i c t o f M a p l e R i d g e . A n a l y t i c a l software used i n the research include M i c r o s o f t Acces s® and Exce l®, (Mic rosof t Off ice 2000 versions) , A d o b e Acroba t Reader v 6 , K A N E W b y W e s t o n Solut ions (based on Mic roso f t A c c e s s ® Off ice 97 version), S-Plus v6 and V7 (Insightful Corporat ion) and A r c v i e w G I S 3 . 2 A and A r c m a p v 9 b y E S R I . Survey of North American utilities. There were a number o f challenges i n v o l v e d i n determining the data co l l ec t ion practices o f N o r t h A m e r i c a n water ut i l i t ies . The best practices indicated b y D e b et al. (2002) and N S G M I (2002) are fa i r ly recent observations and there are no publ i shed reports regarding industry awareness or use o f the recommended best practices. Because the best practices recommended i n these documents are not exact ly the same, it was necessary to reconci le the differences and develop the questions i n the survey questionnaire to reflect both sets o f best practices. Ano the r challenge i n developing the survey was the need to balance the degree o f detail i n w h i c h I was interested w i t h the effort required o f the respondent to vo lun ta r i ly provide such informat ion. I soon recognized that i n order for the survey to fu l f i l l m y research goal o f ident i fy ing the data avai lable to mid- s i ze ut i l i t ies , the questionnaire w o u l d 21' have to be detailed and that the informat ion w o u l d not be readi ly avai lable i n one place or w i t h any one person w i t h i n any organizat ion except w i t h i n very sma l l organizations. The survey w o u l d have to be easy to read, and not discourage uti l i t ies f rom responding to it. It w o u l d have to create interest and encourage persistence w i t h i n those organizations attempting to complete it. The survey w o u l d need a layout and format that was easy to read and complete (regardless o f the ab i l i ty o f the person comple t ing it), and yet easy for me to compi l e the responses. W h i l e a few electronic formats were examined, it became apparent that the survey w o u l d be most eas i ly completed i f the forms were sent out as a spreadsheet w o r k b o o k i n order that they cou ld be printed, completed b y hand and faxed back or that they cou ld be completed and sent back e lect ronical ly to me b y users (because spreadsheet software is c o m m o n l y used b y uti l i t ies) . A draft vers ion o f the survey was sent to four colleagues, two i n M a p l e R i d g e , one i n the C i t y o f Wes t V a n c o u v e r and one i n the C i t y o f B u r n a b y for their comments and this a l l owed me to test methods o f e lect ronical ly c o m p i l i n g the results. M i n o r edi torial and c lar i f ica t ion changes were made and incorporated i n the f inal vers ion. The f inal vers ion inc luded text control (for c o m p i l i n g and ana lyz ing the data) us ing drop d o w n boxes w h i c h also made the survey easier for users to complete. Ano the r challenge was h o w to distribute the survey and h o w to encourage people to complete the survey. I considered g i v i n g honorar ia but decided against that due to costs. T o encourage responses, I noted i n the dis tr ibut ion that, the results w o u l d be shared. I selected emai l dis t r ibut ion to m i n i m i z e the dis tr ibut ion costs and de l ivery t ime. N o t a l l professional associations I approached to help me distribute the survey were helpful , though most were. O v e r 400 surveys were di rect ly sent out and about thirty personal requests and specific f o l l o w up telephone cal ls were made. In a l l , three mai l ings o f the same survey were sent out. The first m a i l i n g was sent to members o f the Canad ian Water and Waste A s s o c i a t i o n 22 ( C W W A ) . The second to the members o f the A m e r i c a n P u b l i c W o r k s A s s o c i a t i o n ( A P W A ) and th i rd to the C i t y Engineers i n the Greater V a n c o u v e r R e g i o n a l Dis t r ic t . The responses were received gradual ly and the in i t i a l deadline for rece iv ing surveys was extended because o f a l o w number o f t i m e l y responses. The first set o f responses was from the C W W A members. The A P W A distr ibut ion was made through the A P W A to its members w h o were water resources professionals and this m a i l i n g generated a large response f rom a number o f U . S . ut i l i t ies. The C W W A dis t r ibut ion inc luded the spreadsheet fi le w h i l e the A P W A emai l dis t r ibut ion contained a l i n k to the fi le. The l i n k affected the submiss ion o f some uti l i t ies as it added another step to the response process and it created confusion on the part o f the respondents. In many A P W A cases, o n l y a por t ion o f the survey was returned in i t i a l l y and the uti l i t ies had to re-send the survey. W h i l e a majori ty o f surveys were completed and sent back electronical ly , a number o f responses were returned b y fax. The faxed responses were then inputted manua l ly into spreadsheets for processing. W h e n the responses were received, they were rev iewed for errors and text control . The raw data were exported into a M i c r o s o f t A c c e s s ® database. The database was queried and then inputted to a M i c r o s o f t Exce l® fi le for data manipula t ion , analysis and graphing. M o s t respondents spent about 45 minutes comple t ing the survey. O f the 70 responses received, 11 responses were rejected as incomplete . These usua l ly had o n l y one or more o f the seven spread sheets completed due to the confusion caused b y the l i n k i n g requirement or d i d not real ize that the fi le contained several spreadsheets. Those surveys were sent back and f o l l o w up telephone cal ls were made where appropriate. The design and development o f the survey began i n 2003 and the surveys were f ina l ly distributed, c o m p i l e d and analyzed i n 2004. A manuscript report ing on the survey, Chapter 2, was submitted to Journal A W W A i n 2005 and was publ i shed i n J u l y 2006. 23 Data creation. The creation o f data described i n Chapter 3 was inspired b y the survey results that showed a gap between data presently col lected and those avai lable (see F igure 2.2 o f Chapter 2), as w e l l as b y the fact that ut i l i t ies noted the locations o f different data w i t h i n their various departments. It was also d r iven b y the d i scovery that the data set purchased f rom the C i t y o f Seattle w h i c h was thought to have breadth and his tory was actual ly quite incomplete . Af te r repeated correspondences and a site v is i t to Seattle, it was determined that the data f rom Seattle c o u l d not be used for the experiments. I then contacted four munic ipa l i t i es i n the Greater V a n c o u v e r R e g i o n a l Dis t r ic t (Burnaby, R i c h m o n d , Wes t V a n c o u v e r and C o q u i t l a m and rev iewed the breadth and his tory o f their data. Y e t again I determined that the data p rov ided b y these munic ipa l i t ies was insufficient. Thus I decided to use M a p l e R i d g e data, though l imi ted , for input for the experiments because o f m y ab i l i ty to access these data, and to col lect data f rom other internal and external sources for this w o r k and for other m u n i c i p a l projects beyond the focus o f m y research. Because o f its l imi t ed data, M a p l e R i d g e is representative o f m a n y uti l i t ies and is ideal for demonstrating the concepts o f f inding and gathering data w i t h i n and external to an organizat ion and o f constructing and l i n k i n g data for asset management. The L a i t y V i e w area o f M a p l e R i d g e was selected to be a case study area because it represents an urban area and experienced construct ion practices typ ica l for the munic ipa l i ty . It comprises 13 percent o f the uti l i ty 's p ipe network. De te rmin ing i f data were avai lable required meet ing w i t h and exp lor ing activit ies o f various w o r k units throughout the organizat ion. The research effort o f creating data was significant and the approach and techniques used i n do ing so are presented i n this thesis. These tools are general and m a y be used b y a l l ut i l i t ies. N e w data constructed for M a p l e R i d g e i n this research include so i l , surface 24 material , water pressure and f low i n pipes, bedding and b a c k f i l l mater ial and traffic loading. A d d i t i o n a l activit ies were carr ied out to ver i fy data i n c l u d i n g f ie ld surveys o f so i l and p re l imina ry p ipe sampl ing and analysis. W i t h different departments co l lec t ing and storing data, the process o f scrubbing data was not a straightforward exercise. D u r i n g the process o f l i n k i n g the data, inconsistencies and m i s s i n g data i n the various databases were d iscovered (e.g., m i s s i n g material or diameter data or discrepancies between the hydraul ic m o d e l and G I S data). T y p i c a l l y , I had to delve into the details o f those databases to see w h i c h data were more current or i n some cases, to ver i fy those data us ing another source. F o r example , i f a pipe had different ages i n the different databases (e.g., the hydraul ic m o d e l and the G I S ) , I . checked the age i n each database to ensure that an input error had not been made, and checked the capital works program to see i f the p ipe had been replaced but that the age o f the new pipe had not been updated i n one o f the databases. I also rev iewed the data for consis tency w i t h the knowledge I have about M a p l e Ridge ' s practices (e.g., po l ic ies such as that there should not be asbestos cement pipes instal led after 1985). Where discrepancies existed, as-built drawings were rev iewed to determine i f data were correct. It was unfortunate that data were sometimes lost or deleted (e.g., at M a p l e R i d g e when a pipe is replaced, a l l records pertaining to that pipe are t yp i ca l ly el iminated). In some cases, I was able to ver i fy some data for these pipes, but I was not successful i n most cases. A s w e l l , the research required the development o f a technique to facilitate creating on-demand databases for analysis and updating. T h i s was required to address the need to obtain data for experiments and to be able to update the o r ig ina l database but also to respect the ownership o f the data and the organizat ional structure. 25 An example of how creating data. H o w pipe break records were created for this research and h o w they are planned to be managed i n the future b y M a p l e R i d g e is an example o f the activities and efforts described above. The break records for M a p l e R i d g e are not stored i n any one locat ion but i n var ious files and phys i ca l locat ions and w i t h different staff o f the operations department i n the operations b u i l d i n g ( w h i c h is separate f rom C i t y H a l l and the engineering department). Because o f m y experiences w i t h other uti l i t ies, I was certain that M a p l e R i d g e had some records o n water m a i n breaks. In the course o f conversations w i t h various staff over a number o f months, I made m a n y inquir ies and requests o f var ious operations and engineering staff (wi th suggestions o f w h o m a y have the data or where they m a y be stored, e.g., the archive files, i n someone's previous files or i i n w o r k request logs) , I was able to obtain a l l water m a i n break data for the per iod 1983- 2004. Information regarding water m a i n breaks for certain years was o n l y avai lable f rom f ie ld staff break forms, i n paper form, and for other years, it was avai lable o n printed computer reports stored i n separate files. O n c e a l l the records were obtained, the data were rev iewed and scrubbed to separate out records o f service connect ion failures (e.g. saddle failures, l eak ing copper service lines) from those o f pipe breaks. Subsequently, an electronic database o f the breaks was then created b y input t ing the data b y hand into spreadsheets, ident i fy ing the specific ident i f icat ion number (see Chapter 3) o f the pipes that broke b y locat ing those pipes on maps us ing the loca t ion data recorded, and adding the G I S pipe ident i f icat ion number as a p ipe attribute. The spreadsheets were then exported to a M i c r o s o f t Acces s® database where a new input fo rm was created to a l l ow the addi t ion o f future break data from revised f ie ld record forms (as discussed i n Chapter 3). The storage locat ion o f the new water m a i n break database was ident i f ied and mapped b y those responsible i n the engineering department for data warehousing. T h e database is avai lable 26 to the operations department staff for searching, f ind ing and sharing the raw data. The break data for the L a i t y V i e w case study was then extracted f rom the new water m a i n break database for the experiments described i n Chapter 4. W h i l e the creation o f the database became more mechan ica l once the paper records were obtained, creating the database required d iscuss ion and resolut ion regarding i ) future data co l lec t ion resources, efforts, qual i ty control and training, and i i ) database b u i l d i n g , ownership , management and maintenance. A number o f discussions and negotiations transpired between the engineering, operations and informat ion technology departments regarding the goals and potential results resul t ing from pipe break data co l lec t ion and analysis, the roles and responsibi l i t ies for data co l lec t ion , storage, management and access, and the contr ibut ion b y each department o f staff t ime and finances. These discussions and m y desire to improve data warehousing practices, yet respecting data ownership issues w i t h i n M a p l e R i d g e , inspi red the development o f the distributed but related databases for on-demand analysis described i n Chapter 3. Th i s por t ion o f the research took f rom early 2004 un t i l summer 2005. Framework for using prediction models. Th i s por t ion o f the research experiments was in i t i a l l y based o n ana lyz ing the created data to determine i f there were k e y data that are current ly avai lable to most smal l and m e d i u m size uti l i t ies that c o u l d be used i n predic t ion models for i m p r o v i n g asset management. The L a i t y V i e w data set w i t h break his tory from 1983 to 2004 was used for the experiments and is described i n Chapter 4. ^ The experiments were designed to use statistical determinist ic t ime-exponent ial and t ime-l inear regression models , su rv iva l analysis models and K A N E W (Deb et al., 1998). T ime- l inea r and t ime-exponent ia l models can be used b y smal l to m e d i u m size uti l i t ies because these uti l i t ies t yp i ca l ly have the inst i tut ional capaci ty to use spreadsheets or s imple 27 statistical software packages. The surv iva l analysis and K A N E W (Deb et al, 1998) techniques were appl ied to determine the sui tabi l i ty o f those applications for more adept ut i l i t ies , even though most uti l i t ies m a y not have the ab i l i ty to use these more sophisticated techniques or operate the more complex software required. W h i l e not reported i n Chapter 4, the appl ica t ion o f su rv iva l analysis was not successful due to the l o w number o f breaks i n the L a i t y V i e w case w h i c h is a young system, the l imi t ed breaks i n older pipes and the loss o f data regarding pipes that were replaced. S u r v i v a l functions were der ived for a l l the p ipe groups described i n Chapter 4 but were incomplete for the most part because surv iva l analysis is dependent i n k n o w i n g the failure his tory across a large range o f pipe ages. There were not enough failures i n the older pipes to determine a complete surv iva l function for any type o f pipe. A s such, it was conc luded that su rv iva l functions cannot be expected to be c o m m o n l y used b y smal l and m e d i u m size uti l i t ies w i t h l imi t ed data. S i m i l a r l y , the use o f K A N E W is constrained by ; the l im i t ed break his tory to derive cohort specif ic survivals . K A N E W , a Mic roso f t Acces s® 97-based program was developed based on interviews w i t h u t i l i ty staff regarding their experience and can be ve ry conservative. I had a number o f discussions w i t h the developers o f K A N E W regarding the program, its development, use, l imitat ions and applicat ion. A g a i n , because o f the incomplete failure his tory o f pipes i n M a p l e R i d g e , K A N E W c o u l d not be appl ied w i t h success and was not reported o n i n Chapter 4. The strength o f K A N E W is i n a l l o w i n g a u t i l i ty to develop replacement budgets. U s i n g expected pipe l ives f rom the experiences o f other ut i l i t ies has some value for deve lop ing network replacement budgets but is not useful for pipe b y pipe replacements. 28 The experiments reported o n i n Chapter 4 were conducted dur ing summer and fa l l 2005 , and the analysis was completed i n 2006. Contribution of research. Th i s thesis offers a number contributions for sma l l and m e d i u m size uti l i t ies and future researchers. It assesses the data col lected b y uti l i t ies, it develops and demonstrates h o w to create and l i n k data for asset management i n general and it explores the under ly ing causes o f water m a i n breaks. F i n a l l y , the research develops a framework for uti l i t ies to use break predic t ion models on a pipe b y pipe basis to improve their method o f p r io r i t i z ing the replacing o f water mains for asset management and to in fo rm their future data acquis i t ion and storage programs. Managers m a y gain insights from the lessons learned from conduct ing this research o n organizat ional considerations such as inst i tut ional memory , staff t raining, engineering and operations responsibi l i t ies and strategic th ink ing for asset management. In particular, researchers w i l l benefit from this thesis because it identifies w h i c h data are avai lable for deve lop ing future asset management tools and h o w to access and construct water m a i n break data. 29 1.6 R E F E R E N C E S A h a m m e d , M . and M e l c h e r s , R . E . , 1994. R e l i a b i l i t y o f underground pipel ines subjected to corrosion. Journal of Transportation Engineering, 120:6: 989-1003. . A S C E ( A m e r i c a n Socie ty o f C i v i l Engineers) , 1999. A m e r i c a n Soc ie ty o f C i v i l Engineers report card and issue briefs. Public Works Management and Policy, 4 :1 : 58-76. A W W A ( A m e r i c a n Wate r W o r k s Assoc ia t ion) , 2004. WATER:\STATS - The Water Utility Database. 2002 version [CD-ROM]. A m e r i c a n Water W o r k s Assoc i a t i on , Denver , C O . 80235. A W W A and E E S , Inc., 2002. New or repaired water mains. A v a i l a b l e on- l ine at http://www.epa.gov/safewater/tcr/pdf/maincontam.pdf. , A c c e s s e d February 12, 2006. Canada. Statistics Canada, 2006. The age of public infrastructure in Canada, b y V a l e r i e Gaudreault and Pat r ick L e m i r e . January 2006. (Catalogue N o . 11 -621-MIE2006035) . Ot tawa, M i n i s t e r o f Industry. D a v i s , D . N . , 2000. Agent-based dec i s ion support f ramework for water supply infrastructure rehabi l i ta t ion and development. Computers, Environment and Urban Systems, 24:2000: 173-190. 30 D e b , A . R . , Grablu tz , F . M . , Has i t , Y . J . , Synder, J . K . , Longanathan, G . V . and A g b e n o w s k i , N . , 2002. P r i o r i t i z i n g Water m a i n Replacement and Rehabi l i ta t ion . A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. D e b , A . K . , Hans i t , Y . J . and Grabul tz , F . M . , 1998. Quant i fy ing Future Rehabi l i t a t ion and Replacement Needs o f Watermains . A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. D e S i l v a , D . , D a v i s , P . , B u r n , L . S., Ferguson, P . , M a s s i e , D . , C u l l , J . , E i sw i r fh , M . and Heske , C , 2002. C o n d i t i o n Assessment o f Cast Iron and Asbes tos Cement Pipes b y In-pipe Probes and Select ive S a m p l i n g for Es t ima t ion o f R e m a i n i n g Service L i f e . Proceedings of the 20th International No-Dig Conference, Copenhagen, Denmark . D ingus , M . , H a v e n , J . and R u s s e l l , A . , 2002. Nondestructuve, N o n i n v a s i v e Assessment o f Underg round Pipe l ines . A W W A Research Foundat ion , 6666 W e s t Q u i n c y A v e n u e , Denver , C O . 80235. Do leac , M . L . , L a c k e y , S. L . and Brat ton, G . N . , 1980. P red ic t ion o f t ime-to-failure for bur ied cast i r on pipe. Proceedings of A WW A Annual Conference, Denver , C O . Ear th Tech , 2004. N a t i o n a l Water and Wastewater B e n c h m a r k i n g Init iat ive F i n a l Resul ts - June 2004. Ear th Tech , Burnaby , B C . 31 F C M , 2001 . Backgrounder: Protecting Canada's Quality of Life. Federat ion o f Canad ian M u n i c i p a l i t i e s , (http:/ /www.fcm.ca/infra/throne-a.html). A c c e s s e d December 2002. G r i g g , N . S., 2004. Assessment and renewal o f water dis t r ibut ion systems. A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. Ket t ler , A . J . and Goul ter , I. C , 1985. A n analysis o f p ipe breakage i n urban water dis t r ibut ion networks. Canadian Journal of Civil Engineering, 12:2: 286-293. K l e i n e r , Y . and Rajan i , B . B . , 2001 . Comprehens ive rev iew o f structural deterioration o f water mains : statistical models. Urban Water, 3:3: 131-150. K l e i n e r , Y . and Rajan i , B . B . , 1999. U s i n g l imi ted data to assess future needs. Journal AWWA, 9 1 : 7 : 4 7 - 6 2 . K l e i n e r , Y . , Ra jan i , B . B . and Sadiq , R . , 2004. Management o f failure r i sk i n large diameter bur ied pipes us ing fuzzy-based techniques. 4th International Conference on Decision Making in Urban and Civil Engineering, Porto, Por tugal . M a i l h o t , A . , Pel let ier , G . , N o e l , J . -F . and V i l l e n e u v e , J . -P. , 2000. M o d e l i n g the evolu t ion o f the structural state o f water p ipe networks w i t h b r i e f recorded pipe break histories: M e t h o d o l o g y and appl icat ion. Water Resources Research, 36:10: 3053-3062. 32 M a l e , J . W . , W a l s k i , T . and Slutsky, A . H . , 1990. A n a l y z i n g Water M a i n Replacement Po l i c i e s . Journal of Water Resources Planning and Management, 116:3: 362-374. M y e r s , C , 2001 . R u r a l Areas gain h igh tech infrastructure p lan through partnership, Utility Executive, 3:2: 5-8, Water Env i ronment Federat ion, 601 W y t h e Street, A l e x a n d r i a , V A , 22314-1994. N G S M I (Nat iona l G u i d e to Sustainable M u n i c i p a l Infrastructure - Inffaguide), 2002. Deter iora t ion and Inspection o f Water Dis t r ibu t ion Systems. Infraguide - Potable Water . Ot tawa, O N , Canada. N a t i o n a l Research C o u n c i l o f the Na t iona l A c a d e m i e s - Commi t t ee o n P u b l i c Water S u p p l y Dis t r ibu t ion Systems: Assess ing and R e d u c i n g R i s k s , 2005 . Public Water Supply Distribution Systems: Assessing and Reducing Risks - First Report. O ' D a y , D . K . , 1982. O r g a n i z i n g and ana lyz ing leak and break data for m a k i n g m a i n replacement decisions. Journal AWWA, 74:11: 588-594. Ontar io P I R . 2005. WATERTIGHT: The case for change in Ontario's water and wastewater sector. Repor t o f the Water Strategy Exper t Panel . Ontar io M i n i s t r y o f P u b l i c Infrastructure R e n e w a l . Ra jan i , B . B . and K l e i n e r , Y . , 2001. Comprehens ive rev iew o f structural deterioration o f water mains : p h y s i c a l l y based models . Urban Water, 3:3: 151-164. 33 Rajani , B . B . and M a k a r , J . , 2000. A methodology to estimate remain ing service life o f grey cast i ron water mains. Canadian Journal of Civil Engineering, 27:6: 1259-1272. Rajani , B . B . and Tesfamar iam, S., 2005. Es t imat ing t ime to failure o f ageing cast i r on water mains under uncertainties. Water Management for the 21st Century, U n i v e r s i t y o f Exeter , U K . , U n i v e r s i t y o f Exeter , U K . , 1-7. Sac lu t i , F . , 1999. M o d e l i n g Water Dis t r ibu t ion P ipe Fai lures U s i n g A r t i f i c i a l N e u r a l Ne tworks , Mas t e r o f Science, U n i v e r s i t y o f Albe r t a , E d m o n t o n , Albe r t a . Sav ic , D A . and Wal ters , G . A . , 1999. Hydro informat ics , D a t a M i n i n g and Main tenance o f U K Water Ne tworks . Anti-Corrosion Methods and Materials, 46:6: 415-425. Shamir , U . and H o w a r d , C , 1979. A n analytic approach to schedul ing pipe replacement. Journal AWWA, 71:5: 248-258. U S E P A , 2005. 2003 D r i n k i n g Water Infrastructure Needs Survey and Assessment , U n i t e d States Env i ronmenta l Protect ion A g e n c y , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . U S E P A , 2002. C l e a n Water and D r i n k i n g Water Infrastructure G a p A n a l y s i s , Env i ronmen ta l Protect ion A g e n c y , Off ice o f Water , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . 34 U S E P A , 2001 . D r i n k i n g Water Infrastructure Needs Survey. Second Repor t to Congress. U n i t e d States. Env i ronmen ta l Protect ion A g e n c y , Off ice o f Water , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . U S E P A , 1999. D r i n k i n g Water Infrastructure Needs Survey. U n i t e d States Env i ronmen ta l Protect ion A g e n c y , Of f ice o f Water , D r i n k i n g Wate r D i v i s i o n , Wash ing ton , D C . t J S F H W A , 1999. Asse t Management Pr imer , Federal H i g h w a y Admin i s t r a t ion , U S Department o f Transportat ion, Off ice o f Asse t Management , 400 7th Street, S . W . Wash ing ton D . C . 20590. Van ie r , D . J . , 2001 . Asse t Management : " A " to " Z " , American Public Works Association Annual Congress and Exposition - Innovations in Urban Infrastructure Seminar, Phi lade lph ia , U . S . September, 2001 . 1-16. Van ie r , D . J . and R a h m a n , S., 2004. M u n i c i p a l Infrastructure Investment P l a n n i n g ( M I I P ) M l f P Report : A P r imer o n M u n i c i p a l Infrastructure Asse t Management . Report B-5123.3, N a t i o n a l Research C o u n c i l Canada, Ottawa, O N . W a l s k i , T . M . , 1982. E c o n o m i c A n a l y s i s o f Water M a i n Breaks . Journal of Water Resources Planning Management Division, 108:3: 296-308. W o o d , A . and Lence , B . J . , 2006. Assessment o f Water M a i n Break D a t a for Asse t Management . Journal AWWA, 98:07. 35 Figure 1.1 Twenty-year total per household infrastructure cost estimates for different sized water systems $ 3 , 0 0 0 L a r g e M e d i u m S m a l l s y s t e m s s y s t e m s s y s t e m s Source: USEPA 1999 Drinking Water Infrastructure Needs Survey 36 C H A P T E R 2 ASSESSMENT OF W A T E R M A I N B R E A K D A T A F O R ASSET M A N A G E M E N T A version of this paper was published in the July 2006 issue of Journal A WW A. The paper is titled as Assessment of Water Main Break Data for Asset Management by A. Wood and B.J. hence. 37 P R E F A C E M u c h research has been conducted over the years regarding the development o f water m a i n break models . B u t as a manager, I found that none o f the m e d i u m size uti l i t ies w i t h w h o m I was famil iar , used break predic t ion models to pr ior i t ize their pipe replacements. Instead o f developing more models , I wanted to f ind a w a y to use break predic t ion models and exis t ing data i n asset management and i n p r io r i t i z ing specif ic pipes and to help other ut i l i t ies to do l ikewise . A s a manager, I k n e w that the u t i l i ty that I w o r k e d for d i d not have a large amount o f data w h i c h w o u l d l imi t their ab i l i ty to use models and I wanted to k n o w i f the amount o f data available was also a hindrance for other s imi la r s ize ut i l i t ies . In addi t ion, as a manager, I was interested i n compar ing our data co l lec t ion performance w i t h other ut i l i t ies and i n i m p r o v i n g our co l l ec t ion practices. T h o u g h best practices for water m a i n break data co l lec t ion were recommended i n 2002 (see N a t i o n a l G u i d e T o Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R F (Deb et al, 2002), I c o u l d not f ind reports regarding the use o f the best practices, a l though D e b et al. (2002) had surveyed some large uti l i t ies (see A p p e n d i x A for a summary o f the size o f ut i l i t ies surveyed and the questions posed). Therefore, I developed a survey, the results o f w h i c h w o u l d serve as a foundation for m y research and distributed it to smal l and m e d i u m size uti l i t ies i n order that I might use it to guide m y research i n deve lop ing tools and techniques for i m p r o v i n g u t i l i ty asset management. I have received feedback on the results such as it is t i m e l y and provides managers w i t h a benchmark o f what is currently be ing done b y uti l i t ies against best practices. Some o f the results were presented at the 2006 B r i t i s h C o l u m b i a P u b l i c W o r k s A s s o c i a t i o n Conference, Q u a l i c u m Beach , B C and questions b y attendees inc lude: w h i c h data are most 38 important to col lect , what methods should uti l i t ies emp loy to col lec t data and h o w should data be stored and analyzed for asset management? The raw data regarding survey responses are appended to this thesis. 39 2.1 I N T R O D U C T I O N Water m a i n breaks can result i n loss o f water to key businesses and cr i t ica l faci l i t ies, and lead to damage o f infrastructure. Such events highl ight the deterioration, o f aging infrastructure, w h i c h is at the forefront o f p o l i c y and program discussions between nat ional and p rov inc i a l or state governments. The need for aging water m a i n rehabil i ta t ion is increasing, the costs o f repairs and replacement can be h igh , and the impact to customers potent ia l ly significant ( U S E P A 2001). F o r u t i l i ty managers, co l lec t ing , recording and mon i to r ing water m a i n breaks is important, not o n l y because such events m a y result i n significant pub l i c impact and destruction o f private property, but also because the data obtained f rom breaks m a y provide insights for management o f the system as a whole . T h i s informat ion is important i n developing the tradeoffs between expenditures and the leve l o f service p rov ided , and i n managing rehabil i tat ion programs to achieve a desired leve l o f service. T h i s paper reports o n the results o f a 2004 survey o f water ut i l i t ies i n the U . S . and Canada that determines the degree and types o f field informat ion that are currently col lected and other data avai lable w i t h i n different departments o f the u t i l i ty regarding water m a i n breaks. Queries were posed w i t h reference to the best practices recommended b y the Na t iona l G u i d e to Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R F (Deb et al, 2002). The survey results identify the data that are deemed important and the approach undertaken for recording and storing these data b y different sizes o f ut i l i t ies . The leve l o f confidence that respondents have i n some o f the data be ing col lected and their l eve l o f comfort w i t h the amount and types o f data col lected, is also identif ied. The survey p rov ided the u t i l i ty managers w i t h a l ist o f data recommended as best practices against w h i c h they cou ld assess their o w n data co l lec t ion practices. The survey 40 f indings are important for the broader infrastructure management communi ty , because knowledge o f the data avai lable underpins the development o f relevant data acquis i t ion and storage strategies and the advancement o f asset management approaches. T h e y also prov ide u t i l i ty managers w i t h a basis for gauging their practices relat ive to those o f other ut i l i t ies . The survey was developed consider ing other recent water u t i l i ty surveys, i nc lud ing those ini t iated b y the N a t i o n a l Water and Wastewater B e n c h m a r k i n g Init iat ive (Earth Tech , 2004), the A m e r i c a n Water W o r k s Assoc i a t i on , ( A W W A , 2004), and A W W A R F (Deb et al, 2002). The N a t i o n a l Water and Wastewater B e n c h m a r k i n g Init iat ive (Earth Tech , 2004), a partnership o f Canad ian uti l i t ies, considers water m a i n breaks as a benchmark ing measure but does not examine water m a i n break data i n detail . The A W W A ( A W W A , 2004) database, c o m p i l e d as a jo in t effort between A W W A and A W W A R F , is based o n a 2002 and 2003 survey o f 337 smal l , m e d i u m , and large U . S . and Canad ian uti l i t ies and focuses o n a broad range o f potable water dis t r ibut ion characteristics i nc lud ing general informat ion regarding service popula t ion, pipe material , valves , fire hydrants and f lushing, f inished water storage faci l i t ies , corros ion control , customer metering, water, supply audit ing, and leakage management. It includes the reported number o f water m a i n breaks o f the uti l i t ies surveyed but does not inc lude data recorded o n water m a i n breaks. The survey reported on i n this paper bu i lds on previous A W W A w o r k (Deb et al, 2002), but queried i n greater detail the data that are col lected b y uti l i t ies, such as failure causes and the general phys i ca l characteristics o f the p ipe and so i l , and had more responses f rom sma l l to m e d i u m size uti l i t ies. In this survey, 37 uti l i t ies serving populat ions be low 100,000 and 22 uti l i t ies serving populat ions greater than 100,000 responded, compared 41 w i t h five ut i l i t ies serving populat ions be low 100,000 and 32 uti l i t ies serving populat ions greater than 100,000 that responded to the 2002 survey. 2 . 2 DESIGN O F T H E S U R V E Y The survey was designed for ease o f comple t ion and compi l a t i on into a data base for analysis. Users were able to print the survey and complete it b y hand, or complete it e lec t ronical ly us ing personal d ig i ta l assistants and laptops. M i c r o s o f t Exce l® w o r k sheets were chosen due to their c o m m o n use. Design of the questionnaire. Based on a rev iew o f the literature ( N G S M I , 2002; and D e b et al, 2002), the questions posed i n the survey were grouped into the f o l l o w i n g s ix Categories: • Genera l informat ion such as t ime o f break, customers affected, response personnel and equipment, and cost o f repairs; • L o c a t i o n data such as nearest property address and geographical coordinates; • P h y s i c a l data such as pipe mater ial and depth o f cover; • Env i ronmen ta l data such as depth o f frost and s o i l and air temperature; • Fa i lu re informat ion such as type and cause o f failure; and • Repa i r informat ion such as components replaced, repaired or instal led. The survey respondents were queried regarding h o w they current ly store data related to water m a i n breaks, their estimated level o f confidence i n selected data, their comfort w i t h the amount o f data and the percentage o f events for w h i c h data are recorded. T h e y were also g iven the opportuni ty to list the addi t ional data elements that they collect . 42 Survey distribution. The survey was distributed b y direct ema i l to members o f the Canad ian Water and Waste A s s o c i a t i o n ( C W W A ) , and members o f the A m e r i c a n P u b l i c W o r k s A s s o c i a t i o n ( A P W A ) w h o identif ied in,their member prof i le that they were either di rect ly responsible for, or interested i n , water dis t r ibut ion and treatment. 2.3 S U R V E Y R E S U L T S Seventy responses were received and after a rev iew o f each response for completeness, 59 surveys were deemed complete and analyzed herein as respondents. The general feedback f rom the respondents was that, w h i l e water m a i n break data co l lec t ion has been ident i f ied as important and a strategic ini t ia t ive ( G r i g g , 2004), ut i l i t ies are just beg inn ing to col lect data comprehensively , co l l ec t ion practices va ry w i d e l y , and most uti l i t ies v i e w their efforts as evo lv ing . The sizes o f the respondents range from one u t i l i ty serving 2,600 people to two uti l i t ies serving a popula t ion o f over one m i l l i o n . M o s t o f the respondents serve populat ions o f between 50,000 and 500,000. The service populat ions o f the respondents are shown i n Table 2 .1 . General water main break information. Genera l water m a i n break informat ion includes the date and t ime that a break is reported, the response t ime and resources used to address the break, and the impact on customers. The percentages o f a l l respondents that col lect different general informat ion are also shown i n F igure 2 .1 . A l l the respondents indicated that they record the date and t ime w h e n a break was reported. The data related to resources expended on the repair are general ly reported i n terms o f the amount o f crew hours spent responding to a reported break. H o w e v e r , o n l y about 70 percent o f respondents 43 record labor, materials, and equipment costs and less than 30 percent record the cost o f property damage or the effect o f the break o n customers. F r o m a management perspective, water m a i n breaks are a l i ab i l i t y and thus clear documentat ion o f a report o f the break or the request for assistance, and the u t i l i ty response to the informat ion, are required should c la ims be f i led against the ut i l i ty . L e n g t h y break response t imes m a y result i n significant private or pub l i c property damage and can be cos t ly and embarrassing for u t i l i ty managers. O n l y 70 percent o f respondents record costs. W h i l e this is surpris ing, it m a y be due to the d i f f icul ty associated w i t h recording some costs (e.g., some costs m a y o n l y be determined at a later t ime, such as costs from a c l a i m or for re-paving a road). Some costs m a y be determined us ing other sources o f data such as through pay ro l l or w o r k order systems. In addi t ion, w h i l e managers need to track expenditures for cost contro l , accountabi l i ty o f resources and p lann ing for future management decisions, they m a y use different report ing systems and m a y be required to report the costs for responding to water m a i n breaks o n l y on an aggregate basis. The survey responses show that few respondents record the effect o f the breaks o n customers i n terms o f damages incurred and the number or type o f customers that experience a water stoppage as a result o f the break or repair. H o w e v e r , the data m a y be avai lable or calculated i f necessary b y determining i f and where valves were operated to shut o f f water to sections o f a water m a i n and i f services were turned off, h o w long it took to restore the water service. T h i s w o u l d a l l o w a u t i l i ty to calculate the number o f customers affected and the length o f the interruption as a surrogate for the direct impact on customers. In addi t ion, c la ims data were also referenced as a potential source o f informat ion. 44 L o c a t i o n d a t a . L o c a t i o n data include the nearest property address, cross street, and geospatial coordinates. The percentage o f respondents that record var ious locat ion data is shown i n Table 2.2. L o c a t i o n data a long w i t h maps can a id response crews i n determining the t ime and effort o f response, the size o f m a i n to be repaired, the supplies needed to repair the break and the loca t ion o f valves that need to be c losed i n order to facilitate repairs. In addi t ion, recording the loca t ion o f the break a l lows future managers to determine i f there are repeated breaks i n a part icular section o f water m a i n and m a y help to assess i f rehabil i ta t ion or replacement o f the m a i n is required. M o s t respondents record the nearest property address and cross street name, but few record addi t ional informat ion that can identify the exact locat ion o f the break. The survey responses suggest that address and cross street data come f rom the in i t i a l report o f the break (i.e., the customer) but are not revised w h e n the actual break loca t ion is determined. The fact that data are col lected suggests that dispatching a response is the p r imary mot iva t ion for the co l lec t ion , but that co l lec t ing loca t ion details that w o u l d assist future break analyses is not as h igh a pr ior i ty . The lack o f actual break locat ion data m a y suggest that either respondents f ind it diff icul t to record and store spatial data or do not consider the specif ic locat ion o f a break important. The fact that few respondents record whether i so la t ion valves are operated can . make it diff icul t for u t i l i ty managers to determine the number o f customers affected b y a break and a repair. T h e use o f Geographic Information Systems (GISs ) for storing informat ion and v i s u a l i z i n g break data, and the l o w cost o f G l o b a l Pos i t i on ing Sys tem ( G P S ) survey equipment, m a y change this practice i n the future. GIS-s tored locat ion informat ion m a y u l t imate ly enhance dec i s ion-making regarding rehabi l i ta t ion and replacement. 45 Physical data. The data typ i ca l ly classif ied as phys i ca l include: the age, mater ial and characteristics o f the pipe, and surrounding so i l . The results reported i n F igure 2.2 show that most respondents col lect a narrow variety o f phys i ca l data. M o s t respondents record data that inc lude p ipe diameter, pipe mater ial , water service type, cover depth, and whether the surface is a roadway or other surface. The lack o f phys i ca l data col lected b y respondents m a y be attributed to the fact that - water m a i n breaks are typ ica l ly treated as emergency situations i n w h i c h the goals are to contain damage, repair the break, and restore lost water service to customers or that some informat ion on the phys ica l characteristics o f water mains m a y be avai lable elsewhere i n the data warehouse o f the ut i l i ty , such as i n as-built records, or i n G I S form. W h e n other data sources are accounted for, the percentage o f respondents that possess the various data elements increases. A n a l y s i s o f the survey results suggests respondents and uti l i t ies can be classif ied, according to the richness o f the data recorded, into four groups. These groups include uti l i t ies possessing: i . basic data consis t ing o f p ipe size and material ; i i . basic mater ial and diameter data plus l im i t ed informat ion such as age o f the pipe, use o f the surface at ground leve l , operating pressure, and type o f pipe jo int . I f pipes i n a network are fa i l ing at jo ints , details about jo in ts can be used to develop a strategy to anticipate, prevent, and repair breaks; i i i . i n addi t ion to the data described i n (i) and ( i i ) , data o n construct ion o f the water m a i n ; and i v . i n addi t ion to data described i n (i), ( i i ) , and ( i i i ) , detai led pipe informat ion such as pipe w a l l thickness and pipe fracture toughness ( typ ica l ly as a result o f pipe testing). 46 Failure causes and modes. A n important aspect o f water m a i n break data is the nature o f the failure. T h o u g h the mode o f failure cannot be def in i t ive ly correlated w i t h specif ic causes o f the failure, they m a y indicate a failure mechan i sm and suggest a cause o f failure for analysis b y asset managers. T h e survey queried the respondents regarding eleven c o m m o n failure modes and the responses to the survey show that 85 percent o f respondents record leak ing jo ints , valves , hydrants and service connections and between 7Q and 83 percent record the remain ing seven failure modes. These failures modes are: l eak ing jo in t , l eak ing service connect ion, l eak ing va lve , l eak ing hydrant, longi tudinal break, b low-out , split b e l l , corros ion pit hole, curb stop failure, tap failure and fai led b l o w - o f f (i.e., air release va lve) . Respondents cou ld also select an "other failure modes" category i n the event that the failure is different from the list o f failure modes provided . N o other modes were reported. W h i l e ut i l i t ies are able to determine and record the failure mode, o n l y 25 percent o f respondents record the cause i n 100 percent o f their records, 37 percent o f the respondents record the cause i n 75 percent o f their records, and o n l y 40 percent o f respondents record a cause i n at least 50 percent o f their records. Th i s suggests that few respondents have a consistently h i g h l eve l o f informat ion on the causes o f breaks to their water mains. The different causes o f failure and the percentage o f respondents that record a specific cause o f failure are shown o n F igure 2.3. The analysis o f data shown i n F igure 2.3 suggests that it is diff icul t for respondents to determine the cause o f failure. Managers m a y f ind that in ' their organizat ion, breaks are treated as emergency situations and the u t i l i ty staff at the scene o f the break focus o n cont ro l l ing the extent o f col lateral damage from the break. A l s o , some specia l ized 47 engineering background is required to confident ly determine the cause o f breaks under response situations. Repair activities. W h i l e it m a y be diff icul t to determine and record the cause o f breaks, the response regarding the recording o f repair activit ies is h igh . The data i n F igure 2.4 show the percentage o f respondents that record a specif ic repair act ivi ty. Repa i r activit ies t yp i ca l ly inc lude repair ing clamps and jo in ts or replac ing sections o f pipes, valves , hydrants, and connections. T h i s cou ld be expla ined b y the fact that ut i l i t ies can more readi ly record response actions i n the f ie ld than determine the cause o f a water m a i n break. Environmental data. Env i ronmen ta l data related to water m a i n breaks inc lude informat ion such as air temperature, so i l acidi ty , and moisture, and other antecedent condi t ions. The environment is important for water mains as the N a t i o n a l Research C o u n c i l o f the N a t i o n a l A c a d e m i e s (2005) identifies that water m a i n breaks and their repairs are also potential gateways to contaminat ion o f the water dis t r ibut ion system. The survey responses show that ve ry litt le environmental informat ion is col lected b y respondents. O n l y 12 percent o f the respondents record environmental data for a l l o f their water m a i n breaks and o n l y 27 percent record any environmental data at a l l . In fact, environmental data is the least recorded group o f informat ion for a l l o f the respondents. I f a respondent col lects environmental informat ion, most l i k e l y it is informat ion on the depth o f frost (27 percent o f respondents). Less than 10 percent o f the respondents take so i l samples. T h i s is surprising as it is w e l l pub l i c i zed that there is a significant relationship between external cor ros ion o f water mains and s o i l condi t ions. F igure 2.5 shows the percentage o f the respondents that record a g iven type o f environmental data, or can obtain the data elsewhere either w i t h i n the ut i l i ty or f rom sources external to the u t i l i ty such as 48 other agencies. In some cases, a s ignif icant ly greater percentage o f respondents indicated that more data are avai lable from other sources than are recorded. Hydraulic models. W h e n asked whether the u t i l i ty has a water m o d e l and whether it is used, 78 percent o f the respondents report hav ing a hydrau l ic mode l . O n l y 71 percent o f the respondents use these models for some purpose. The most popular uses o f models i n descending order o f popular i ty (i.e., based on the percentage o f respondents) are for capital p lann ing (69 percent), development p lanning (61 percent), operations (56 percent), and maintenance purposes (42 percent ). Confidence in data. A k e y survey question regards the leve l o f confidence that a u t i l i ty places o n the phys i ca l data that it collects. U t i l i t i e s were asked i f their confidence i n the data col lected is h igh . The intent was to determine not o n l y the l eve l o f confidence but also i f this l eve l varies b y data type. The confidence o f the respondents i n selected parameters is summar ized i n F igure 2.6. A s shown i n F igure 2.6, at least 72 percent o f uti l i t ies have h igh confidence i n pipe diameter and material , but o n l y 48 percent are h i g h l y confident about the year o f instal lat ion. Level of comfort with the amount of data collected. U t i l i t i e s were asked to indicate their l eve l o f comfort w i t h the amount o f data col lected. S ix ty- four percent o f the respondents were comfortable w i t h the amount o f data col lected. H o w e v e r , 22 percent were not comfortable and 14 percent had no op in ion regarding their l eve l o f confidence i n the data col lected. The leve l o f comfort w i t h the amount o f data col lec ted varies w i t h u t i l i ty size. Be tween 50 and 60 percent o f respondents w i t h service populat ions between 10,000 and 100,000 and approximate ly 70 percent o f those serving between 100,000 and 500,000 are comfortable w i t h the amount o f data col lected. B o t h respondents w i t h service populat ions greater than one m i l l i o n were also comfortable w i t h their data. 49 Sources of data available to utilities. The addi t ional sources o f phys i ca l data for the respondents are summar ized i n Table 2.3. The number o f respondents that indicated that data were avai lable from the other sources is also listed. F o r example , 46 percent o f respondents ident i f ied other sources for obtaining normal operating pressures. Managers m a y use this table to consider s imi la r data sources w i t h i n their organizat ion. T h e y m a y also be encouraged to determine i f this informat ion is avai lable f rom other sources that are not as yet identif ied. Storage of data. Different sizes o f uti l i t ies have different means o f data storage. The use o f a G I S as a data management system is greatest for respondents that serve a popula t ion o f between 50,000 and 100,000 (four o f these ten communi t ies use G I S s ) , and archiva l records are the predominant source o f data for ut i l i t ies that serve a popula t ion o f between 10,000 and 50,000 (eight o f these fourteen communi t ies use archiva l records). O f the fifteen respondents that serve a popula t ion o f between 100,000.and 500,000, seven use archiva l records (such as as-builts and other paper-based h is tor ica l records) as a source o f water m a i n break data, and four use G I S s . S t a t i s t i ca l conf idence of the survey results. A l t h o u g h the number o f respondents is l o w , the survey results can be used to draw some inferences regarding the practices o f water uti l i t ies i n general. A statistical analysis o f the s ignif icance o f the survey sample for p r o v i d i n g observations for the general C a n a d i a n - U . S . u t i l i ty popula t ion was undertaken. The accepted measure range o f confidence l imi t s is calculated us ing W i l d and Seber (2000), to be between 11 percent and 13 percent for this survey. T h i s is to say that i f 63 percent o f the respondents indicate that they record w h e n water service is restored, w e m a y expect that more than 50 percent ( i . e . , the lower confidence l imi t ) but less than 75 percent 50 (i.e., the higher confidence l imi t ) o f a l l water ut i l i t ies i n the general popula t ion w o u l d record the same. A p p e n d i x B summarizes the statistical confidence o f these results. 2.4 DISCUSSION AND R E C O M M E N D A T I O N S W h i l e a l l respondents are co l lec t ing data, they c lear ly do not col lect a l l data suggested b y best practices recommended b y the Na t iona l G u i d e to Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R F (Deb at al, 2002). H o w e v e r , based on the responses it is apparent that data acquis i t ion is evo lv ing and that there is a strong interest i n compar ing their practices w i t h others and w i t h best practices. Feedback f rom the respondents indicates that the average respondent spent about 30 to 45 minutes comple t ing the survey. Th i s is i n addi t ion to the t ime that respondents spent determining the appropriate people w i t h i n the organizat ion to complete the survey. S u c h decis ions are c o m m o n among larger organizations where responsibi l i t ies for data co l lec t ion , management and analysis are shared. The amount o f data that respondents and uti l i t ies i n general have can be categorized into general classes o f data richness. Based o n the survey results, it is suggested that four classes exist. These are: expanded, intermediate, l imi t ed and m i n i m a l data set classes. A s shown i n F igure 2.7, the proposed data set classes are based on the breadth and record length o f data irrespective o f the tradit ional data categories such as break, pipe inventory, and operational data because the survey results indicate that m a n y respondents col lect data i n va ry ing amounts i n each o f the tradit ional data categories. Expanded data sets are compr ised o f informat ion, and records over a long per iod o f t ime. These m a y include: an inventory o f pipes that is correlated to pipe informat ion such as diameter, material , year o f instal lat ion, type o f jo in t , surface cover, type o f failures, probable cause o f failure, type o f 51 repair, pipe testing informat ion such as pipe w a l l thickness at t ime o f break, type o f p ipe l i n ing , so i l testing informat ion such as corros iv i ty , and p H . Intermediate data sets are compr i sed o f informat ion on inventory and pressure zones, a break his tory o f some length o f t ime, and some amount o f informat ion regarding pipe diameter, mater ial , age, exterior surface condi t ion , and instal lat ion, surface cover, and pipe protection. A u t i l i ty possessing informat ion w i t h this degree o f data richness has general ized descriptions regarding the type o f failures encountered and probable cause o f those failures. T h i s case is s imi la r to the expanded data case but does not include pipe and so i l testing data al though instal lat ion details m a y be avai lable . Limited data sets have o n l y l imi t ed informat ion o n the pipe network, surface uses and loads. Th i s w o u l d be an inventory, informat ion regarding pressure zones and general ized p rob lem areas, a break his tory o f a m i n i m a l length o f t ime, and a n o m i n a l amount o f informat ion regarding pipe diameter, pipe mater ial , year o f instal lat ion, and surface.cover. U n d e r this case, the informat ion compr i s ing the intermediate data set is s ignif icant ly reduced. Minimal data sets are compr i sed o f p ipe lengths, ident i f icat ion o f pressure zones, general ized p rob lem areas, and n o m i n a l informat ion regarding pipe diameters. G i v e n that asset management is becoming important for u t i l i ty managers, more attention and effort should be g iven to i m p r o v i n g data co l lec t ion i n the f o l l o w i n g areas: Collect for the future. Managers need to consider the long v i e w w h e n assessing their data co l l ec t ion strategy. Water m a i n break data w i l l be useful for future asset managers w h o w i l l have to make diff icul t investment decis ions regarding w h i c h pipes to replace. Estimates for the U . S . ( U S E P A , 2002) suggest that the capi tal needs for d r ink ing water i n the U . S . for the per iod between 2000 and 2019 range f rom 154 to 446 b i l l i o n U . S . dollars. There are no s imi l a r estimates for Canada. Future managers w i l l need to make wise 52 decisions to take into account not o n l y the ab i l i ty o f the pub l i c to afford replac ing these pipes but also the scarcity o f human resources and capital to actual ly perform the work . Because the practice o f infrastructure asset management w i l l evolve , data requirements for m a k i n g c r i t i ca l decisions are important but can also be expected to grow. Thus , the data co l l ec t ion strategy for ut i l i t ies should be to gather the data recommended as best practices keeping i n m i n d that m u c h o f the data w i l l be useful for future management decisions rather than for today 's decisions. It w i l l l i k e l y take some t ime to establish a data record o f sufficient length to support such decisions. M a n y managers m a y resist this approach because they m a y prefer to k n o w what is needed and w h y before embark ing o n data co l lec t ion programs. In some cases, such a strategy m a y be undertaken w i t h l i t t le addi t ional effort. F o r example , a u t i l i ty m a y be able to accumulate more informat ion b y co l lec t ing addi t ional informat ion as part o f exis t ing tasks, such as a descr ipt ion o f the failure mode, bedding, b a c k f i l l and depth o f cover. T o facilitate asset management, data co l lec t ion efforts should also inc lude , where possible , the types and causes o f breaks, and the phys ica l data o f a l l mains so as to provide a deeper understanding o f the failure modes and mechanisms. R e c o r d d a t a effect ively. F o r co l l ec t ion practices to be effective, ut i l i t ies need to record data cont inuously , consistently, and accurately. D a t a should be col lected on a l l - breaks to ensure that the records portray the state o f the w h o l e system. M i s s i n g records m a y create a l eve l o f uncertainty regarding the records that are col lected and undermine the confidence i n conclus ions d rawn from technical analyses and lead to organizat ional frustration. In order to record data i n a consistent manner, the use o f standard operating practices, procedures, and forms as w e l l as t ra ining for those w h o col lect data are important. I f a 53 ut i l i ty is already co l lec t ing data, a r ev iew o f a l l data co l lec t ion forms is useful to ensure consis tency i n methods as w e l l as to expand the data col lected i f appropriate. A n audit o f the records for qual i ty and accuracy should be undertaken per iod ica l ly . A suggested m o d e l for an audit exercise is the formation o f a qual i ty control group w i t h i n the u t i l i ty to audit the data and ensure qual i ty across the organizat ion. Th i s group cou ld also develop guidelines o f practice for educating staff. The most efficient approach for recording data w i l l vary among uti l i t ies and should reflect business and w o r k f low processes and organizat ional structure o f the ut i l i ty . M a n y uti l i t ies m a y have different departments, such as laboratory, maintenance, technical , and f inancia l departments, and response teams that col lect or generate informat ion. In such cases, it m a y be benef ic ia l to bundle or group data based on organizat ional departments and develop a process for ensuring that the data are col lected, assessed for relat ional l inks related to other data, recorded and stored i n an effective manner. F o r example , data forms cou ld be circulated among selected departments w i t h i n the organizat ion to col lect input data comprehens ive ly and to improve the sharing o f informat ion among departments, or depending o n the data and organizat ion, the data co l l ec t ion process between departments c o u l d be independent but l i nked or related v i a an asset index. The storage o f data can be undertaken i n various ways (see Table 2.3), or i n a comprehensive data warehouse. Some uti l i t ies prefer a comprehensive data warehouse to facilitate the synthesis and v i sua l representation o f data and v i e w G I S technology as the ideal tool for this purpose. Alternate sources of data and relating data. The survey results c lear ly show that alternate sources o f data related to water m a i n breaks exist i n many organizations. Table 2.4 summarizes the potential range o f informat ion sources identif ied i n this w o r k for different 54 data elements. F o r example , current and future capi tal project designs, as-builts, or asset pro-formas can be used to capture informat ion such as specifications, test pit logs, and inspect ion records. U t i l i t i e s m a y w i s h to consider the concept o f a data web, w h i c h m a y be thought o f as a relat ional structure o f the u t i l i ty ' s data sources, i nc lud ing break records, maintenance reports, pipe and so i l samples, customer informat ion, a rchiva l systems, hydraul ic m o d e l output, capital rehabil i ta t ion p lann ing data, G I S s , and maintenance management systems. The l inks between these data sources, support the web, and are keyed to some index o f the asset, for example , a pipe ident if icat ion number. The use o f the l inks reduces the need to convert data when a new informat ion management system is developed and implemented and facilitates the synthesis o f exis t ing corporate data for specif ic analyses. A data web is not a data storage appl ica t ion nor software, but a concept o f relat ing or l i n k i n g data and separate data sources i n an organizat ion to each other. W h i l e the development o f a data web for a u t i l i ty reduces the need to convert and store data, the addi t ion o f data elements i n each data system, such as the asset index (e.g., corresponding to a p ipe ident i f icat ion number) m a y be required. A wel l -des igned data web w o u l d enhance u t i l i ty management, par t icular ly for sma l l communi t ies w h o cannot afford wholesa le data management system implementat ion, conversions and upgrades. D e s i g n considerations m a y include ident i f icat ion o f the architecture o f the web, the characteristics and number o f required l inks , and the appropriate connections between the elements to be "webbed" . Taci t informat ion or knowledge that is currently not recorded can also be l i nked w i t h i n the web. Decision support tools. In deve lop ing dec is ion tools for support ing asset management, ut i l i t ies should consider a range o f systems that are f lex ib le and easi ly accessible for both present and future uses. The systems should be rel iable and mainta ined 55 over t ime. D a t a that have been col lected are useless i f the dec i s ion makers w i t h i n the organizat ion cannot access them or are unaware o f their existence. A n a l y s i s o f the survey results suggests that dec i s ion support tools for p r io r i t i z ing the replacement or rehabi l i ta t ion o f water mains should be tai lored to the degree o f richness o f the data avai lable to a ut i l i ty . U t i l i t i e s that have a m i n i m a l amount o f data cannot use sophisticated tools such as phys ica l (Rajani and M a k a r , 2000) or statistical pipe deterioration models (Shamir and H o w a r d , 1979; Jacobs and Karney , 1994; A n d r e o u et al, 1987; K l e i n e r and Rajan i , 1999), or l i fe cyc le cost ing. These tools, and most recent research i n water dis t r ibut ion asset management have focused o n uti l i t ies that possess large amounts o f data and more research is required to develop robust approaches for uti l i t ies w i t h l imi ted or m i n i m a l data records. These tools should also be f lexible enough to adapt as uti l i t ies increase the amount and types o f data col lected. 2.5 CONCLUSIONS Studies, even as recent as i n 2004, have identif ied the need for standardized m a i n break databases and te rminology and continued research regarding database development as strategic for informed infrastructure management (e.g., G r i g g , 2004; O ' D a y et al, 1986). The results o f the survey reported herein indicate that water m a i n break data co l lec t ion is evo lv ing , that industry practices do not match best practices recommended b y the Na t iona l G u i d e to Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R F (Deb et al, 2002) at this t ime, and that most respondents recognize the need for a strategy for data qual i ty improvement . The data-related challenges that a l l ut i l i t ies face inc lude di f f icul ty i n m o b i l i z i n g f inancia l and human resources, absence o f h is tor ica l data, lack o f knowledge o f 56 current organizat ional practices, l o w re l iab i l i ty o f p rev ious ly col lected data, d i f f icu l ty i n p r io r i t i z ing data co l lec t ion , and the need to develop effective data storage programs. W h i l e it is generally accepted that the use o f re l iable data regarding asset inventory and cond i t ion w i l l enhance the management o f m u n i c i p a l infrastructure (Vanie r , 2001), the feedback from the respondents is that data co l l ec t ion practices regarding both inventory and cond i t ion vary w i d e l y . Genera l data regarding customer locat ion, t ime o f break, and emergency response actions are t yp i ca l l y avai lable, but informat ion regarding specif ic pipe loca t ion and phys ica l attributes is inconsistent. M o s t respondents do not have a consistently h i g h l eve l o f informat ion regarding causes o f failure and so i l and pipe sampl ing are general ly not undertaken. Conf idence i n and comfort w i t h the amount o f data col lected varies; mid - s i ze respondents expressed the least l eve l o f comfort w i t h the amount o f data col lected. M a n y respondents identif ied addi t ional sources o f informat ion inc lud ing archival , operation and maintenance and G I S informat ion, and hydraul ic models . M o r e o v e r , hav ing a.hydraulic mode l does not guarantee that it is used. T h i s m a y depend o n the staffs' ab i l i ty to operate and main ta in the mode l and on the re l i ab i l i ty and currency o f the input data. W h i l e both phys i ca l and statistical models have been developed for predic t ing p ipe deterioration and for deve lop ing water m a i n rehabil i ta t ion plans, it is evident that the choice and appl ica t ion o f these models are l imi ted b y the data that ut i l i t ies have regarding water m a i n breaks (Rajani and K l e i n e r , 2001 ; K l e i n e r and Rajan i , 2001). In general, uti l i t ies can be classif ied as those possessing expanded, intermediate, l im i t ed or m i n i m a l data. Character izat ion o f these classes m a y be used to in form the development o f new asset management techniques such as condi t ion models or heuristics and new ways o f effectively 57 col lec t ing , storing, combin ing , and representing water m a i n break data. T h i s is a subject o f a for thcoming paper b y the authors. 2 . 6 A C K N O W L E D G E M E N T S The authors gratefully acknowledge those uti l i t ies w h o invested their t ime and effort i n comple t ing the survey and Professors A . D . R u s s e l l and J . A twa te r at the U n i v e r s i t y o f B r i t i s h C o l u m b i a and D r . D . V a n i e r at the N a t i o n a l Research C o u n c i l Canada, w h o guided the development o f the survey. M r . s G . P h i l l i p s (ret.), W . L i u and G . I r w i n o f the Dis t r i c t o f M a p l e R i d g e assisted i n survey development, d is t r ibut ion and compi la t ion ; M s . K . Feh rman o f the Canad ian Water and Waste A s s o c i a t i o n and M r . A . G a l l o f the A m e r i c a n P u b l i c W o r k s A s s o c i a t i o n promoted and distributed the survey to their respective organizations. W e are also grateful to D r . N . G r i g g at the Co lo rado State U n i v e r s i t y for his thoughtful r ev iew o f this manuscript . c 58 2.7 R E F E R E N C E S A n d r e o u , S., M a r k s , D . and C l a r k , R . , 1987. A new M e t h o d o l o g y for M o d e l i n g Break Fa i lure Patterns i n Deter iorat ing Water Di s t r ibu t ion Systems: Theory . Advances in Water Resources, 10:1:2-10. E l sev ie r Science, B . V . A S C E ( A m e r i c a n Socie ty o f C i v i l Engineers) , 1999. A m e r i c a n Soc ie ty o f C i v i l Engineers report card and issue briefs. Public Works Management and Policy, 4 :1 : 58-76. A W W A ( A m e r i c a n Wate r W o r k s Assoc ia t ion) , 2004. WA TER: \STA TS - The Water Utility Database. 2002 version [CD-ROM]. A m e r i c a n Water W o r k s Assoc i a t i on , Denver , C O . 80235. D e b , A . R . , Grab lu tz , F . M . , Has i t , Y . J . , Synder, J . K . , Longanathan, G . V . and A g b e n o w s k i , N . , 2002. P r i o r i t i z i n g Water m a i n Replacement and Rehabi l i ta t ion . A W W A Research Foundat ion , 6666 W e s t Q u i n c y A v e n u e , Denver , C O . 80235. Ear th Tech , 2004. N a t i o n a l Water and Wastewater B e n c h m a r k i n g Init iat ive F i n a l Results - June 2004. Ear th T e c h , Burnaby , B C . Ear th Tech , 2003 . N a t i o n a l Water and Wastewater B e n c h m a r k i n g Init iat ive- Water U t i l i t y Def in i t ions 2003 Release B . , Ear th Tech , Burnaby , B C . G r i g g , N . S., 2004. Assessment and renewal o f water dis t r ibut ion systems. A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. 59 Jacobs, P . and K a m e y , B . , 1994. G I S development w i t h appl icat ion to cast i ron water m a i n breakage rates. 2nd International Conference on Water Pipeline Systems, Ed inburgh , Scot land, B H R Group . 53-62. K l e i n e r , Y . and Rajan i , B . B . , 2001 . Comprehens ive rev iew o f structural deterioration o f water mains : statistical models. Urban Water, 3:3: 131-150. K l e i n e r , Y . and Rajan i , B . B . , 1999. U s i n g l imi t ed data to assess future needs. Journal AWWA, 91:7: 47-62. N G S M I (Nat iona l G u i d e to Sustainable M u n i c i p a l Infrastructure - Infraguide), 2002. Deter iora t ion and Inspection o f Water Di s t r ibu t ion Systems. Infraguide - Potable Water . Ot tawa, O N , Canada. N a t i o n a l Research C o u n c i l o f the N a t i o n a l A c a d e m i e s - Commi t t ee o n P u b l i c Water S u p p l y Dis t r ibu t ion Systems: Asses s ing and R e d u c i n g R i s k s , 2005 . Public Water Supply Distribution Systems: Assessing and Reducing Risks - First Report. O ' D a y , D . K . , W e i s s , R . , C h i a v a r i , S., and B l a i r , D . , 1986. Wa te rma in Eva lua t ion for Rehabi l i t a t ion / Replacement . A W W A Research Founda t ion ( A W W A R F ) and U S Env i ronmen ta l Protect ion A g e n c y ( U S E P A ) , A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. 60 Rajani , B . B . and K l e i n e r , Y . , 2001 . Comprehens ive rev iew o f structural deterioration o f water mains: p h y s i c a l l y based models . Urban Water, 3:3: 151-164. Rajan i , B . B . and M a k a r , J . , 2000. A methodology to estimate remain ing service l ife o f grey cast i ron water mains. Canadian Journal of Civil Engineering, 27:6: 1259-1272. Shamir , U . and H o w a r d , C , 1979. A n analytic approach to schedul ing pipe replacement. Journal AWWA, 71:5: 248-258. U S E P A , 2002. C l e a n Water and D r i n k i n g Water Infrastructure G a p A n a l y s i s , Env i ronmen ta l Protect ion A g e n c y , Off ice o f Water , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . U S E P A , 2001 . D r i n k i n g Water Infrastructure Needs Survey. Second Repor t to Congress. U n i t e d States. Env i ronmen ta l Protect ion A g e n c y , Off ice o f Water , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . V a n i e r , D . J . , 2001 . Asse t Management : " A " to " Z " , American Public Works Association Annual Congress and Exposition - Innovations in Urban Infrastructure Seminar, Phi lade lph ia , U . S . September, 2001 . 1-16. W i l d , C . J . and Seber, G . A . F . , 2000. Chance Encounters - A First Course in Data Analysis and Inference. John W i l e y and Sons Inc., N e w Y o r k , N . Y . 61 Table 2.1 Service population of respondents Service Population >1,000 but <5,000 <10,000 <50,000 <100,000 <500,000 <1 million >1 million Total Canadian utilities a l b 2 8 9 6 3 1 30 U.S. Utilitiesc 1 13 3 11 1 29 Total responses 1 3 21 12 17 3 2 , 59 a) Canad ian respondents were from A l b e r t a (5), B r i t i s h C o l u m b i a (9), M a n i t o b a (2), N e w B r u n s w i c k (2), N e w f o u n d l a n d (1), N o v a Scot ia (1), Ontar io (6) and Saskatchewan (4). A n incomplete survey was received f rom one Quebec mun ic ipa l i t y so its response was not inc luded i n the analysis. A d m i t t e d l y , the survey was i n E n g l i s h and m a y be a reason for the l o w number o f responses from Quebec. b) Popu la t ion served is approximate ly 2600. c) U . S . respondents were from A l a s k a (1), A r i z o n a (1), C a l i f o r n i a (3), Co lo rado (1), F l o r i d a (1), I l l ino is (1), Kansas (2), Massachusetts (1), M a r y l a n d (1), M i c h i g a n (1), M i n n e s o t a (1), M i s s i s s i p p i (1), N o r t h C a r o l i n a (1), N e v a d a (1), Pennsy lvan ia (1), South D a k o t a (1), Tennessee (1), Texas (3), U t a h (1), Wash ing ton (4) and W i s c o n s i n (1). 62 Table 2.2 Percentage of respondents that record location data Location Data Percentage of respondents that record location data Nearest property address 9 3 % Cross street name 7 8 % Dis tance f rom cross street 4 4 % Isolat ion va lve operated 3 6 % Dis tance from nearest property l ine 3 4 % Coordinates (northing and easting) 10% 63 Table 2.3 Additional sources of break-related physical data for respondents Number of respondents having additional sources Archival O & M 7 GISs2 Other Normal operating pressure 27 X X X Models , . f i r e f low tests Traffic classification or type of road usage 23 X X X Exte rna l sources and traffic management systems Year of installation 22 X X X Typical flow in area of break 21 X X X M o d e l s Pipe wall thickness / classification 19 X X X Externa l sources Type of pipe lining 16 X X X Exte rna l sources Length of pipe segment containing the repair 14 X X X3 Pipe protection (wrapped / anodes) 13 X X X Under boulevard or roadway 12 X X Exte rna l sources Surface material 11 ' X X Pavement management systems Depth of cover 9 X X Type of joint 8 X X X Bedding material 8 X X Type of water service 7 X X Category of native soil 7 X X X Pipe fracture toughness 4 X X Externa l sources Backfill material 4 X X Pipe modulus or rupture 3 X Externa l sources Pipe sample collected 2 X M a i n tappings 64 Condition of cement lined pipe interior 2 X 4 1 M o d e l s Pipe material 1 X Pipe diameter 1 X Condition of unlined pipe interior 1 X Condition of bedding 1 X Condition of pipe exterior 1 X 1 Operation and Maintenance records (O&M) 2 Geographic Information Systems (GISs) 3 GISs are the most popular source of length ofpipe segment containing repair data. 4 Evaluated with swabbing. 65 Table 2.4 Suggested sources and approaches for collecting physical data on water main breaks i Physical data Suggested sources of information P i p e diameter Bes t captured i n the f ie ld dur ing break repairs and avai lable f rom as-builts D e p t h o f cover Best captured i n the f ie ld dur ing break repairs and avai lable f rom as-builts P ipe mater ial C o u l d be captured i n the f ie ld dur ing break repairs, avai lable f rom as-builts and analyzed off-site, e.g., i n a laboratory for determining the different types o f cast i r on pipes C o n d i t i o n o f bedding C o u l d be captured i n the f ie ld dur ing break repairs and also obtained w i t h extra f ie ld w o r k Category o f native s o i l C o u l d be captured i n the f ie ld dur ing break repairs, determined from other records and also c o u l d be analyzed i n a laboratory C o n d i t i o n o f un l ined pipe interior C o u l d be captured i n the f ie ld dur ing break repairs, and also cou ld be analyzed i n a laboratory C o n d i t i o n o f p ipe exterior C o u l d be captured i n the f ie ld dur ing break repairs and slo cou ld be analyzed i n a laboratory T y p e o f water service C o u l d be captured i n the f ie ld dur ing break repairs or f rom other sources o f informat ion (e.g. l and use plans) Surface mater ial C o u l d be captured i n the f ie ld dur ing break repairs or from other sources o f informat ion (e.g. models) Surface use (under boulevard or roadway) C o u l d be captured i n the f ie ld dur ing break repairs or from other sources o f informat ion (e.g. mode l ) Traff ic c lass i f icat ion or type o f road usage C o u l d be captured i n the f ie ld dur ing break repairs or from other sources o f informat ion (e.g. transportation p lan , traffic models) " 1 L e n g t h o f p ipe segment conta ining the repair C o u l d be captured i n the f ie ld dur ing break repairs Or obtained w i t h addi t ional f ie ld work , or f rom as-builts or G I S B e d d i n g mater ial C o u l d be captured i n the f ie ld dur ing break repairs, m a y be avai lable f rom as-builts and other sources and/or bundled w i t h informat ion gained through addi t ional f ie ld w o r k P ipe protect ion (wrapped/ anodes) C o u l d be captured i n the f ie ld dur ing break repairs, m a y be available f rom as-builts and other sources and/or bundled w i t h informat ion gained through addi t ional f ie ld w o r k 66 Physical data Suggested sources of information T y p e o f jo in t C o u l d be captured i n the f ie ld (bundled w i t h informat ion gained through addi t ional f ie ld w o r k or from other sources o f informat ion (e.g., archives or construct ion inspect ion records) Y e a r o f instal lat ion M a y be avai lable f rom as-builts or other a rchiva l records such as construct ion inspect ion reports B a c k f i l l material M a y be avai lable f rom as-builts or other a rchiva l records such as construct ion inspect ion reports T y p i c a l f l ow i n area o f break Techn ica l tools (e.g., models) N o r m a l operating pressure Techn ica l tools (e.g., models) C o n d i t i o n o f cement l ined pipe interior M a y be avai lable from as-builts and better analyzed i n a laboratory for current condi t ion P ipe modulus o f rupture M a y be avai lable f rom as-builts and better analyzed i n a laboratory for current cond i t ion T y p e o f p ipe l i n i n g M a y be avai lable from as-builts and better analyzed i n a laboratory for current condi t ion P ipe w a l l thickness/classif icat ion M a y be avai lable from as-builts and better analyzed i n a laboratory for current cond i t ion P ipe fracture toughness M a y be avai lable from as-builts and better analyzed i n a laboratory for current condi t ion 67 Figure 2.1 Percentage of respondents that collect general information X X Mfy. fc \ \ \ \ s0 'fc «&. ^ v i>3 °o- % \ •\ x So <5- w ft ^ V '-fc x . %3 ^p ^ p ^p ^ p s O ^p cv*" o"̂  cf̂  cT̂  cr^ cf̂ D O O O O O O O O O O : O ) O O N ( D I T ) ^ C O C M T - So 68  Figure 23 Percentage of respondents that record failure causes s 70 Figure 2.4 Percentage of respondents that record repair activities 1 0 0 % 9 0 % 8 0 % 7 0 % 6 0 % 5 0 % 4 0 % 3 0 % 2 0 % 1 0 % 0 % i 4 # ^ y / 71 Figure 2.5 Percentage of respondents that collect different types of environmental data 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% • Data recorded • • Data recorded and/or available from other sources <0* , 0 s I I 72 Figure 2.6 Percentage of respondents that expressed confidence in data collected Figure 2.7 Classes of data richness among water utilities Increasing data collected or available for analysis (regardless of data categories, e.g., pipe inventory, break data, operational data) 74 CHAPTER 3 CONSTRUCTING WATER MAIN BREAK DATABASES FOR ASSET MANAGEMENT A version of this paper has been accepted and scheduled for publication in the January 2007 issue of Journal AWWA. The paper is titled as Constructing Water Main Break Databases for Asset Management by A. Wood, B. J. Lence and W. Liu. 75 P R E F A C E The survey results as reported i n Chapter 2 conf i rm that a number o f sources o f data exist for use i n ana lyz ing water m a i n breaks. In fact, the figure shown b e l o w (reference Figure 2.2) inspi red m y interest i n deve lop ing an approach for obta in ing and m a x i m i z i n g data from different sources internal and external to the ut i l i ty . I f ut i l i t ies can obtain and integrate data from a wide variety o f sources for analysis, they should be able to improve the breadth and richness o f their water m a i n break data for asset management. Getting data from other sources can help utilities ggo/ o _ JLg^B^^ • Data available from break records ( I n n H Data available from break records and other sources 8 0 % " ~ ~ H ~ ~ 6 0 % - J J - | - 50% - J 40%- I I I I I n n 0 A k e y considerat ion for uti l i t ies is storage and management once the data are created. The focus o f h o w to manage data i n the Asse t Management Systems current ly being marketed is on consol idat ing a l l data into one large database (e.g., M A X I M O and 76 Hansen) . Conve r t i ng and consol idat ing data into one database is expensive, resource intensive^ and creates data ownership confl icts . It is unrealist ic to expect sma l l and m e d i u m water ut i l i t ies to make the f inancia l and organizat ional investment to achieve such as database artd thus this is a significant barrier to i m p r o v i n g asset management. The research presented i n this chapter describes an approach for accompl i sh ing the acquis i t ion and integration o f data f rom other sources for analysis and m a y be used to foster d iscuss ion between var ious departments on h o w asset management data can be managed and coordinated w i t h i n an organizat ion. The research provides techniques and approaches for creating data, constructing databases and relat ing those databases to each Other. M o s t important ly, this w o r k can be appl ied to practice, and has been the case o f the L a i t y V i e w area o f M a p l e R i d g e . F o r M a p l e R i d g e , the research resulted i n the construct ion o f a water m a i n break database, the compi l a t i on o f so i l type data w i t h w h i c h to analyze the relat ionship between s o i l and water m a i n breaks, a r ev iew o f the current water m a i n replacement p o l i c y (o f so le ly replac ing asbestos cement pipes), a survey o f s o i l and cor ros ion potential and a renewed staff interest i n i m p r o v i n g the process and framework o f dec id ing w h y , w h i c h and w h e n pipes should be replaced. In addi t ion, the approach o f constructing databases from different sources and relat ing databases has been adopted b y staff for deal ing w i t h a l l water system data and is be ing explored for use w i t h M a p l e Ridge ' s sewer system. 77 3.1 I N T R O D U C T I O N The tradit ional pub l i c w o r k s emphasis o n managing water m a i n breaks has been directed toward m i n i m i z i n g the loss o f water to k e y businesses and cr i t i ca l facil i t ies (such as hospitals and industr ial plants) and m i n i m i z i n g the damage to bui l t and natural infrastructure. H o w e v e r , breaks are also potential gateways to contaminat ion o f the water dis t r ibut ion system and have been identif ied as a h i g h pr ior i ty i n the assessment o f water supply health r isks b y the Na t iona l Research C o u n c i l o f A c a d e m y Sciences (2005). Pred ic t ing water m a i n breaks to reduce such r isks and op t imize the investment i n aging infrastructure requires rel iable p ipe data. These data include age, diameter and material for the subject dis t r ibut ion system and the number and nature o f breaks that occur i n the water mains. A recent study identifies a need for standardized m a i n break databases and cont inued research regarding database development ( G r i g g , 2004). B a s e d o n a 2004 survey o f N o r t h A m e r i c a n uti l i t ies, W o o d and Lence (2006) observe that water m a i n break data co l l ec t ion is e v o l v i n g and industry practices do not match best practices recommended b y the N a t i o n a l G u i d e to Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R F (Deb et al, 2002). F o r u t i l i ty managers, co l lec t ing , recording and mon i to r ing water m a i n breaks is also important because such events m a y be used to ga in insights for the management o f the entire network. T h i s informat ion is important i n developing the tradeoffs between expenditures and leve l o f service provided , and i n managing rehabil i ta t ion programs to achieve a desired leve l o f service. F o r example, many ut i l i t ies focus their water m a i n replacement program on tolerating a certain leve l o f breaks i n the water dis t r ibut ion system, or perhaps through targeting replacement o f water mains o f a part icular vintage, or o f a mater ial that is prone to breaks. 78 W o o d and Lence (2006) suggest a number o f sources o f data that m a y be used for i m p r o v i n g a uti l i ty 's data breadth and richness for ana lyz ing water m a i n breaks. Th i s paper introduces an approach for constructing a water m a i n database w h i c h is based o n l i n k i n g or " re la t ing" data f rom sources internal and external to a u t i l i ty for the purpose o f knowledge discovery , or i n other words relating relational databases. Issues related to creating, l i n k i n g , t ransforming, c leansing, scrubbing and integrating data are ident i f ied and approaches for addressing them are presented. T h i s approach m a y assist ut i l i t ies i n deve lop ing data acquis i t ion and management strategies and gu id ing knowledge discovery . R e c o g n i z i n g that data co l l ec t ion and storage is organizat ional ly dr iven , the approach is designed to be easi ly adapted. The approach l inks informat ion among mul t ip le databases and respects decentral ized data input i n order to mainta in ease o f data storage and management. T h i s reduces the need for wholesale convers ion o f data to a central database and is l i k e l y to be cost effective for sma l l to mid- s i ze uti l i t ies. The decentral ized approach to database construct ion requires coordinat ion o f data acquis i t ion, personnel and a clear strategy for encouraging departmental cooperation. T o demonstrate the eff icacy o f this approach for use i n ana lyz ing and predic t ing breaks i n a water dis t r ibut ion system, a data schematic is created for ana lyz ing water mains and predic t ing future breaks for the L a i t y V i e w area o f the Dis t r i c t o f M a p l e R i d g e , B C . The development o f a data schematic is not mere ly a data storage exercise. Rather it is the creation o f l inkages that can be used to aggregate informat ion for analysis and that a l l ow data to be updated over t ime. The l inkages are created among p r imary and secondary data sources. F o r example, a mun ic ipa l i t y m a y not have a transportation p lan that specifies traffic vo lumes , but i f knowledge o f the v o l u m e o f vehic le traffic over a water m a i n is 79 desired, it m a y be able to use v o l u m e data f rom a traffic management system database. The use o f data f rom different sources provides f l ex ib i l i t y for ut i l i t ies to focus o n co l lec t ing data for future analyses without hav ing to commi t to specific appl ica t ion software. 3.2 S T A T E O F D A T A IN UTILITIES Water m a i n break data co l lec t ion practices va ry across ut i l i t ies and for many uti l i t ies , data co l lec t ion can have significant costs i f performed at a comprehensive leve l . These costs inc lude direct f inancia l costs as w e l l as organizat ional effort and human resources. The challenges o f co l lec t ing data include d i f f icu l ty i n m o b i l i z i n g f inancia l and human resources, absence o f h is tor ica l data, lack o f knowledge o f current organizat ional practices, poor re l i ab i l i ty o f p rev ious ly col lected data, compl ica t ions due to the emergency- oriented co l l ec t ion condi t ions w h e n breaks occur, d i f f icul ty i n p r io r i t i z ing co l l ec t ion efforts, and the need to develop effective data storage programs ( W o o d and Lence , 2006). In recent years, m a n y uti l i t ies have been deve lop ing improved data acquis i t ion and management strategies for water m a i n breaks and i n some cases us ing third parties for analyt ical tasks for obta ining specia l ized data such as so i l conduct iv i ty . M o s t munic ipa l i t i es do have some informat ion regarding their water pipes and condi t ions, but few have been main ta in ing records o f pipe breaks for longer than a decade, and ve ry li t t le informat ion is available regarding i n d i v i d u a l pipes i n a g iven network (Pellet ier et al, 2003). In a case study o f three munic ipa l i t i es i n Quebec, they found that o n l y s ix parameters (diameter, length, type o f material , year o f instal lat ion, type o f s o i l and land use above the pipe) were available for analysis. In recent years, best practices have been identif ied for water m a i n break data co l lec t ion (Deb et al, 2002; N S G M I , 2002). W o o d and Lence (2006) surveyed N o r t h A m e r i c a n uti l i t ies and conducted detailed 80 interviews to determine the richness o f data avai lable to uti l i t ies for ana lyz ing water m a i n breaks. T h e y found that ut i l i t ies i n general can be categorized into general classes o f data richness. These are the: expanded, intermediate, l imi t ed and m i n i m a l data set class. The data set classes are based o n the breadth and record length o f data irrespective o f the tradit ional data categories such as break, p ipe inventory, and operational data because the survey results indicate that m a n y respondents col lect data i n va ry ing amounts i n each o f the tradit ional data categories. The results o f the survey also showed that w h i l e m a n y uti l i t ies do not t yp i ca l ly have a c o m m o n break and water m a i n database or the appropriate data, they m a y f ind relevant informat ion avai lable elsewhere i n their organizat ion and use this informat ion to expand their database. 3 . 3 W A T E R M A I N B R E A K D A T A F O R ASSET M A N A G E M E N T T w o cha l lenging asset management issues are the determination o f remain ing service l ife and the pr ior i t i za t ion o f rehabil i tat ion efforts. B o t h o f these issues re ly o n knowledge discovery . T o determine remain ing service l i fe , one must be able to assess the cond i t ion o f a pipe and the expected remain ing l ife or some measure o f h o w long a g iven p ipe can be expected to last f rom the date o f instal lat ion. It is diff icul t to determine the exact c o n d i t i o n e d bur ied pipes because they are diff icul t to comprehens ive ly inspect. A s a m i n i m u m , uti l i t ies should have a database o f water m a i n breaks because the occurrence o f breaks m a y reflect the condi t ion o f a pipe and typ ica l ly the number o f annual water m a i n breaks is used as a surrogate for the condi t ion o f the network. H o w e v e r , breaks do not necessari ly reflect pipe condi t ion as there are many causes o f breaks such as damage from adjacent construct ion and frost heave. 81 Once pipe condi t ion is determined, the ca lcula t ion o f remain ing service l ife can be made us ing deterioration models that predict when failure or future breaks w i l l occur. These models m a y be either statistically or p h y s i c a l l y based. The concept o f remain ing service l ife is that, g iven use and t ime, a l l pipes w i l l reach a point w h e n they are replaced for reasons such as poor condi t ion , perception o f poor re l i ab i l i ty or the need to increase hydrau l ic capacity. Rajan i and M a k a r (2000) define the t ime o f death o f a p ipe as the t ime at w h i c h its mechanica l factor o f safety falls b e l o w an acceptable value. K l e i n e r and Rajani (1999) propose that the useful l i fe o f a p ipe is a function o f the economic costs o f deterioration and replacement and suggest that pipe death coincides w i t h the op t imal t ime o f replacement. P r io r i t i za t ion o f rehabil i tat ion efforts i nvo lve the t i m i n g and schedul ing o f repairs or replacement o f pipes. T h i s is c lose ly t ied to reso lv ing f inancia l and technical challenges as to whether to mainta in , repair or renew an asset or to choose an alternative such as tw inn ing a water m a i n or constructing an alternative water ma in . P r io r i t i za t ion is compounded b y uncertainty regarding avai lable funds and organizat ional trends w h e n longer-term p lann ing hor izons are considered. M a n y engineers have faced circumstances where short-term solutions m a y not be the most economica l i n the long term but are the most expedient, for instance w h e n a m a i n is repeatedly repaired instead o f replaced because capital replacement funds are di f f icul t to obtain, but emergency operating funds are avai lable . D e b et al. (2002) describe four general approaches for p r io r i t i z ing pipes for replacement: the Deter iora t ion Poin t Ass ignment method ( D P A ) , break-even analysis, failure p robabi l i ty and regression methods, and mechanis t ic methods. K l e i n e r and Rajani (2001) suggest that o n l y larger diameter mains w i t h cos t ly consequences o f failure m a y jus t i fy the data co l lec t ion efforts and costs required to calibrate mechanis t ic models . D e b et 82 al. (2002) suggests that break-even analyses be augmented w i t h predic t ive techniques for pipe breaks, such as failure probabi l i ty , regression and mechanis t ic methods. Other considerations for p r io r i t i z ing water m a i n replacements also include re l i ab i l i ty ( X u and Goul ter , 1998), consequence o f failure (Cooper et al, 2000), considerat ion o f other assets ( G r i g g , 2004; V a n i e r , 2001) , and on-going engineering and management processes. D a v i s (2000) suggests that for ana lyz ing the impact o f changes i n a water m a i n rehabil i ta t ion strategy, an agent-based approach is p romis ing . Agents are defined as information-processing systems and are based on ar t i f ic ial intel l igence approaches. D a v i s proposes a loose ly coupled generic agent-based dec i s ion support f ramework for water uti l i t ies. In this framework, agents are used to extract data from infrastructure, Geographic Information Systems ( G I S ) and strategic databases. Agents are also used to cleanse data, interface w i t h other databases, predict pipe deterioration and assist i n deve lop ing rehabil i ta t ion strategies. S m a l l to m e d i u m size uti l i t ies often do not have staff w i t h the sk i l l s to implement agent-based approaches. F o r these ut i l i t ies , an approach that employs engineering expertise and c o m m o n data processing systems is l i k e l y to be more feasible. K n o w l e d g e d i scovery is the process o f ident i fying v a l i d , useful and ul t imate ly understandable patterns i n data (Torra et al, 2004). The analysis o f water m a i n breaks is l imi t ed b y the challenges faced i n constructing databases such as l im i t ed personnel and resources, m i s s ing and conf l ic t ing data, and non-computer ized informat ion (Pellet ier et al, 2003; H a b i b i a n , 1992; and O ' D a y , 1982). Furthermore, baby-boomer staff o f many uti l i t ies w i l l retire over the c o m i n g years and data m i n i n g w i l l be overshadowed b y the issue o f data creat ion and storage because m u c h water dis t r ibut ion data are ora l ly recorded or are stored i n discrete departmentally managed databases rather than i n a central database. 83 3.4 C O N S T R U C T I N G W A T E R M A I N D A T A B A S E S The approach for sourcing, constructing and l i n k i n g water m a i n break data proposed herein is to create a connected data schematic and undertake knowledge d i scovery based o n these data as s h o w n as F igure 3.1. Identifiers that are shared among databases are used to loose ly associate and relate data across databases, as represented b y the w a v y l ines i n the figure. D a t a f rom different databases such as p ipe break data files, s o i l characteristics maps, ortho-photographs, as-builts and o ra l ly recorded informat ion are l i n k e d b y creating identifiers i n each data source that are c o m m o n to one or more sources. F o r example, specific p ipe segments, surface mater ial and break data can each be assigned a c o m m o n pipe ident i f icat ion number (pipe ID) . Information that are o n l y o ra l ly recorded and transmitted between staff or no longer documented w i t h i n an organizat ion m a y be used to "create" data that can be ver i f ied and integrated for analysis. Cons t ruc t ion standards and practices and manufacturer data can be identif ied through interviews. Da ta from external sources such as the U S Department o f Agr i cu l tu re ( U S D A ) and Canad ian S o i l Survey ( C S C ) can also be related. The database is created us ing processed and b lended data. The process ing and b lend ing o f data is performed b y technical specialists us ing tools such as G I S , spreadsheets and databases. A n a l y s i s o f the data can then be performed to ident ify ne twork performance inc lud ing the occurrence o f water m a i n breaks, complaints and pressure deficiencies. These can be accompl i shed w i t h var ious approaches such as statistical, phys i ca l and neural network mode l ing and r i sk analysis. F r o m the analysis and knowledge discovery , uti l i t ies can develop strategies to predict the remain ing service l ife and pr ior i t i ze rehabil i tat ion efforts. 84 Identifying the Purpose of Analysis. The first step is to identify the purpose for w h i c h the data are to be analyzed. In addi t ion to asset management, objectives such as , performance measurement and improvement , research and development, cost recovery o f services, transparency and pub l i c accountabi l i ty m a y be identif ied. Af t e r the purpose is established, the data required m a y be determined, i.e., i f the purpose o f the analysis is to determine factors associated w i t h breaks that lead to service l ife estimates, the data should include p ipe material , age and diameter. Asse t management analyses o f pump stations require horsepower rating, pump run times, v ibra t ion levels and maintenance details. Pavement asset management requires data o n pavement thickness, sub-base material , A n n u a l Ave rage D a i l y Traff ic ( A A D T ) and type o f axle loading . The purpose o f the analysis affects the design o f the schematic, the re l i ab i l i ty o f the relat ionships between the elements and the degree o f informat ion transferred across a l ink . Developing a data schematic. A data schematic identifies the data required, their potential sources and i f and h o w they are related. The schematic should be constructed cons ider ing the characteristics o f the data elements i nc lud ing the ava i lab i l i ty o f data, whether sources are p r imary or secondary or require interpretation, whether data can be obtained f rom para l le l or through a series o f sources, the confidence i n and explici tness o f the data, and h o w data m a y be l inked . In addi t ion, data can be obtained f rom mul t ip le sources and some sources m a y be more read i ly avai lable than others. Pa ra l l e l sources contain the same or related data and data can be d rawn i n para l le l , w h i l e sources that are i n series contain data that i n some w a y can be related to each other and are d rawn i n series. Once sourced, data can be l i nked across var ious databases us ing identifiers that are unique for specif ic databases or are c o m m o n across a number o f databases (e.g., a c o m m o n pipe I D number). L i n k i n g data f rom various sources also a l lows for the updat ing o f the various 85 . sources o f data as a result o f analysis, knowledge d iscovery or further research. F o r instance, i f the purpose o f the analysis is to determine i f water mains under heav i ly traveled roads experience higher rates o f water m a i n breaks, data o f interest inc lude the surface material , road function and traffic vo lumes . I f current network pipe data do not inc lude these attributes, surface ortho-photographs m a y be used to determine the surface mater ial . The road funct ion o n the surface over a water m a i n m a y be determined us ing a transportation p lan . Al te rna t ive ly , i f the transportation p lan does not contain road use data, but traffic counts or a traffic network p lan exist, this informat ion m a y be used to determine traffic loading . I d e n t i f y i n g Sources o f da ta . W o o d and Lence (2006) provide a l ist o f alternate secondary sources for break-related data based o n survey respondents ' wri t ten observations and personal interviews. O n their o w n , these data are not ve ry useful for ana lyz ing breaks, but w h e n combined , they can provide useful informat ion for deve lop ing rehabi l i ta t ion and operation and maintenance strategies. The f o l l o w i n g is a detai led d iscuss ion o f the alternative sources that ut i l i t ies should consider i n b u i l d i n g a water m a i n break database. N e t w o r k data. N e t w o r k data m a y be available f rom G I S s , Asse t Management ( A M ) / Fac i l i t i es Management ( F M ) systems, databases, spreadsheet systems and other analyt ical models , e.g., hydrau l ic models . Other sources o f data inc lude Superv isory C o n t r o l and D a t a A c q u i s i t i o n ( S C A D A ) systems and moni to r ing systems. A M / F M and S C A D A systems typ i ca l l y are enterprise systems, used b y m a n y departments throughout an organizat ion, store large amounts o f data on large server databases and operate o n a client- server environment. A M / F M systems col lect and store maintenance, repair and f inancial informat ion w h i l e S C A D A systems col lect and store operating data such as pump-run times, pressure and hydrau l ic data. 86 I f G I S s are be ing used for network mapping , each pipe i n the system is g iven an i nd iv idua l p ipe I D and m a y have corresponding informat ion such as year o f instal lat ion, length o f pipe segments, diameter and material . The length o f p ipe segment is typ ica l ly obtained f rom the geometry o f the spatial f i le and are usua l ly col lec ted f rom as-builts as part o f the construct ion o f the G I S . M a n y uti l i t ies use hydraul ic models for their water, sewer and drainage infrastructure. H y d r a u l i c models contain informat ion such as f low, pressure and pipe layout that can be useful for analysis o f water m a i n breaks and asset management. F o r example, roughness, mass balance discrepancies and hydrau l ic losses calculated i n a m o d e l can be indicat ions o f pipe condi t ion . A s - b u i l t drawings, previous construct ion standards, staff experience, purchasing records, manufacturer specifications, inspector and surveyor f ie ld notes, m a i n tappings, f lushing records, swabbing , laboratory, hydrant fire f l ow testing, inspections, water qual i ty testing results, and customer complaints m a y be used to indicate informat ion such as the bedding and b a c k f i l l material , depth o f cover, type o f jo ints , p ipe protection, o r ig ina l p ipe w a l l thickness, current w a l l thickness, cond i t ion o f pipe interior and leve l o f internal corros ion. Surface data. Sources for surface material , land use and traffic load ing data include h igh resolut ion (e.g., 0.15 metre / 6 inches) ortho-photographs and as-built drawings. L a n d use and transportation plans, G I S maps o f roads and o ra l ly recorded knowledge o f types and ratios o f road users, and traffic patterns provide informat ion about surface use, traffic loading, and potential sources o f damage from surface activit ies. Ortho-photographs m a y be loaded into G I S as a "background theme" and over la in w i t h the p ipe network to generate drawings to determine surface mater ial data for each pipe. T h i s m a y reduce the need for f ie ld surveys. V 87 S o i l data. S o i l data are avai lable from both p rov inc i a l and federal government databases i n Canada and analogous sources i n the U S . The C S C N a t i o n a l S o i l Database contains the s o i l types for a l l o f Canada and is the national reposi tory for survey informat ion f rom the broad (1:1 m i l l i o n scale) to the detailed leve l (1:10,000 to 1:250,000 scale). The U S D A Na tu ra l Resources Conserva t ion Service is responsible for the N a t i o n a l Coopera t ive S o i l Survey w h i c h includes the efforts o f federal, state, and academic institutions (Rossiter, 2005). The Wash ing ton Suburban Sanitary C o m m i s s i o n has developed so i l cor ros iv i ty maps based o n U S D A so i l survey results (Habibian,1992) . W h i l e agricul tural surveys m a y be l imi t ed i n terms o f their usefulness i n cor ros ion analysis, w h e n combined w i t h a so i l sampl ing or survey program, they m a y be used to p rov ide a general understanding o f the so i l condi t ions. U t i l i t i e s should also r ev iew boreholes and logs f rom previous studies, w e l l d r i l l i n g data and informat ion gathered on construct ion projects. S o i l permeabi l i ty data and water table informat ion can be useful for drainage and sanitary sewer i n f l o w and inf i l t ra t ion analysis. Other sources o f data. U t i l i t i e s should explore the data col lected b y others or used w i t h i n the organizat ion such as mapp ing o f envi ronmenta l ly sensitive areas, pavement management systems and infrastructure plans. Env i ronmenta l informat ion such as water and ambient temperature and the frequency o f frost, m a y be avai lable f rom water qual i ty testing, weather and environment agencies. Gas p ipel ine , electr ical , t e lecommunica t ion or other ut i l i t ies operating i n the area m a y have informat ion o n s o i l data, stray currents and the cor ros iv i ty o f soi ls . Information that are not recorded but avai lable o ra l ly from staff can be obtained b y us ing interviews o f staff to determine, for example, the type o f bedding and b a c k f i l l material , p ipe l i n i n g and pipe protection. Whenever possible , ver i f ica t ion o f observations 88 is recommended. B r o a d surveys and spot testing are other techniques to consider w h e n obta ining data. F o r example, a program that samples s o i l for cor ros iv i ty and relates the agricul tural classif icat ions to soi ls can y i e l d insights into potential corros ion hot spots and focus moni to r ing efforts. L i n k i n g da ta . The objective o f connect ing or l i n k i n g data is b y relat ing relat ional databases to support on-demand data analysis. Da ta from mul t ip le sources m a y be l i nked i n series, or para l le l or remain un l inked . Th i s approach can facilitate the updating o f the analysis data set as data sources are updated and thereby support future knowledge discovery . Information processing tools, such as G I S s , spreadsheets and databases can be used to create electronic records, manipulate and analyze data. Output f rom these tools is also easi ly exported to other, comprehensive databases. G I S shape files can conta in a m y r i a d o f informat ion that m a y be v i e w e d and exported for analysis us ing other analyt ical software such as spreadsheets and databases. Depend ing o n the expertise and sk i l l s o f those ana lyz ing the data for knowledge discovery , ut i l i t ies m a y choose to use a spreadsheet fi le as the m a i n data analysis f i le . A n example o f w h i c h is described i n the case study presented i n this paper. J I f break records are stored o n l y o n paper, an electronic break database should be created. Spreadsheets are easy to use and the data can be easi ly transferred to G I S format for v i e w i n g w h i l e be ing retained as the k e y database. P i p e I D numbers that are assigned for G I S ne twork maps can be used as the connect ion between the break records and pipe characteristic data. I f records do not specify the exact loca t ion o f the break a long a pipe segment, some extrapolat ion m a y be required to identify the damaged pipe. F o r example , a house number m a y need to be used to define the locat ion a long a p ipe where the break occurred. . •• 89 A process for conver t ing geographical a rchiva l informat ion into electronic data for relat ing data is shown i n F igure 3.2. In the first step, informat ion i n the fo rm o f paper maps is scanned to create T a g Image F i l e Format ( T I F F ) images. T h e n the images are d ig i t i zed to create p o l y l i n e drawings and G I S are used to b u i l d p o l y g o n coverage. Attr ibutes are attached for each p o l y g o n and then related to a c o m m o n identifier, such as a pipe I D . U s i n g the intersect funct ion o f the G I S , the attributes o f the ne twork pipes and the attributes o f the po lygons are combined and exported as a spreadsheet f i le . A l l the files fo rm a por t ion o f the data warehouse that is used to create the m a i n analysis f i le . F o r example , s o i l image maps m a y be d ig i t i zed to create a c losed p o l y l i n e d rawing o f so i l types, fitted to the cadastral drawings o f the water network i n A u t o D e s k M a p ® us ing the 2 D transformation process, and then used to b u i l d p o l y g o n coverage o f the so i l condi t ions w i t h A r c M a p ® . I f each p o l y g o n is assigned a so i l type ident i f icat ion number that corresponds to a so i l type, then us ing the intersect funct ion o f A r c M a p ® , the pipe and corresponding s o i l attributes can be generated and exported to a shape f i le , and u l t imate ly exported as a spreadsheet fi le o f pipe IDs and corresponding so i l type. H y d r a u l i c models m a y be used to calculate the f l o w through m o d e l l inks , and the pressure and heads at nodes w i t h i n the water dis t r ibut ion system. A process for est imating pressure data a long each l i n k w i t h i n the network is shown i n F igure 3.3. H y d r a u l i c mode l outputs (e.g., f l ow and fr ic t ion coefficient estimates i n l inks and pressure estimates at nodes) are determined for the corresponding pipe I D . The pipe l i n k and node shape files are combined us ing G I S tools, exported to a spreadsheet f i le , and thereafter combined w i t h the other pipe network data, such as pipe material , diameter, date o f instal lat ion, and length. Often, the hydrau l ic m o d e l and the pipe network do not have a one to one relat ionship, and c o m m o n l y the hydraul ic m o d e l is a skeleton o f the network where a l i n k 90 i n the m o d e l m a y actual ly represent a number o f pipes i n the real system. F o r example , the C i t y o f Toronto recently skeletonized their 307,956 pipe network to a 76,989 l i n k hydraul ic m o d e l (Sch ick , 2005). In this case, the pressure and f low m a y be related b y buffering the hydraul ic m o d e l output to the pipe network data. Buf f e r ing creates a relat ionship between the l inks i n the m o d e l and the w e l l documented pipes i n the p ipe network; a process shown i n F igure 3.4. Here , the pipe network data (a . D W G file) is converted to a shape fi le w i t h ident i f icat ion numbers (usual ly the pipe ID) and the hydraul ic m o d e l data are prepared as a shape fi le . The two shape files are buffered us ing G I S analysis tools. Buf fe r ing the two shape files creates a relat ionship between each pipe and the m o d e l pressure and f low i n the l inks . U s i n g this process, a l l pipes can be associated w i t h corresponding f lows and pressures. T h i s approach m a y also be used for c o m b i n i n g sewer or drainage hydrau l ic models or other spatial data. P r o c e s s i n g D a t a . P r io r to the analysis stage, data should be processed to reduce errors and inconsistencies and to produce a coherent data set. D a t a process ing activit ies inc lude s imple transformations (e.g., detecting and r emov ing outliers), c leaning and scrubbing, b l end ing data f rom the various sources and summar i z ing the data to reduce the number o f records (Torra et al, 2004). In the process o f relat ing the data, analysts m a y d iscover inconsistencies i n the databases and m a y have to determine the re l i ab i l i ty o f the data and h o w to treat incomplete data, e.g., pipes that do not have an instal lat ion date. Ana lys t s m a y also f ind pipes that are mi s s ing material or diameter data or that hydrau l ic m o d e l output and G I S data differ. It falls o n the analyst and data schematic bui lder to careful ly consider the impacts o f data use approaches and h o w these approaches affect the ab i l i ty to achieve the purpose o f the analysis. 91 Knowledge discovery. Once data processing is complete , u t i l i ty managers m a y analyze and extract knowledge and gain insights regarding their system. T y p i c a l knowledge d i scovery techniques include spatial and statistical analyses us ing G I S , spreadsheets, databases, phys i ca l and statistical p ipe deterioration models and art i f ic ial intel l igence techniques. Here , patterns i n the data m a y be determined such as break patterns, and rates for differ ing p ipe materials, diameters, and ages and the spatial dis t r ibut ion o f the breaks over t ime. A s an example o f the use o f such observations, k n o w i n g the propor t ion o f pipe materials o f a dis t r ibut ion system and the failure rates w i t h i n each mater ial class can help determine the vulnerabi l i t ies and failures that m a y be expected o f the system. De te rmin ing the past, present and future failure rates for pipes o f different vintages, diameters, so i l condi t ions, and surface loadings is important for gu id ing a u t i l i ty ' s rehabi l i ta t ion program. 3.5 W A T E R M A I N B R E A K D A T A B A S E S F O R M A P L E RIDGE, B C The construct ion o f a data schematic for ana lyz ing water m a i n breaks is demonstrated here for the area o f L a i t y V i e w i n the Dis t r ic t o f M a p l e R i d g e . Th i s area comprises 13 percent o f the 335 k i lometer dis t r ibut ion system for the Dis t r i c t and represents an urban area. The L a i t y V i e w area experienced construct ion practices and has , so i l types typ ica l for M a p l e R i d g e . The Dis t r i c t has informat ion res id ing i n various formats (e.g., electronic, a rch iva l , and oral) that are distributed and managed across the organizat ion. F o r example , the Operations Department has paper copies w i t h l imi t ed informat ion on the break his tory o f water mains f rom 1983 to 2004, and no database, w h i l e the Eng inee r ing Department has a skeletonized hydraul ic m o d e l o f the system constructed i n 2001 i n a . D W G format and a spatial representation o f the pipe network i n G I S . 92 The purpose o f this analysis was to determine relationships among water m a i n breaks and factors such as pipe age, mater ial , so i l and depth o f cover so as to a id i n asset management ( N S G M I , 2003). The informat ion processing tools used were Mic roso f t E x c e l ® , M i c r o s o f t A c c e s s ® , A r c V i e w ® G I S and A u t o d e s k ® M a p . These software tools were selected because they are available in-house, and Dis t r ic t staff are trained to use them. The data set for the L a i t y V i e w area o f M a p l e R i d g e represents an intermediate water m a i n break data set w i t h a 20 year break his tory (1983 to 2004). The data inc luded pipe material , p ipe diameter, whether the pipe is under a boulevard or roadway, the year o f instal lat ion, depth o f cover, length o f pipe segment, surface mater ial , no rmal operating pressure, bedd ing mater ial , whether the pipe is wrapped or anode protected (wrapped/anodes), b a c k f i l l material , type o f road function, traffic c lassi f icat ion, type o f pipe l i n i n g and typ ica l f l ow i n the pipe. Constructing and linking data. The ava i lab i l i ty and loca t ion o f the data, l eve l o f confidence i n them, and informat ion regarding h o w they were created are ident i f ied and l is ted i n Table 3.1. Da ta associated w i t h a l o w leve l o f confidence are candidates for a data ver i f ica t ion program. B a s e d on the var ious sources o f the data and the goal o f a id ing asset management, the schematic o f h o w data can be related as shown i n F igure 3.5 was created. The schematic identifies potential l inks and inputs for analysis and major data o f interest. P r i m a r y sources o f data t yp i ca l l y aggregate data and are read i ly avai lable but do not contain a l l the data deemed o f interest for asset management. Secondary sources that contain addi t ional informat ion were identif ied. M a n y o f these use var ious data formats (e.g., databases and paper records) and are distributed across the organizat ion. Some data are avai lable f rom paral le l sources and one or a l l o f the sources m a y be used for analysis. F o r example , traffic vo lumes m a y be obtained f rom a transportation p lan or f rom a traffic 93 network p lan , traffic v o l u m e database and pavement management system. The transportation plan , traffic network p lan and pavement management system use A A D T for def ining the v o l u m e o f traffic w h i l e the traffic network p lan contains 24-hour vehic le vo lumes w h i c h m a y be transformed to equivalent A A D T . Where possible , data were l i nked us ing the p ipe I D as the c o m m o n l i n k across the databases i n order to facilitate updating o f a l l data sources us ing the results o f the knowledge d i scovery process and addi t ional f ie ld invest igat ion. L i k e m a n y uti l i t ies, the electronic water dis t r ibut ion network informat ion for M a p l e R i d g e includes year o f instal lat ion, diameter and material and these are avai lable from the G I S and d rawing files. The length o f the p ipe segment was obtained from the geometry o f the pipes i n these files. Au todesk M a p ® tools were used to define and create the shape f i le for the region. The study area was defined for extraction from the water network map, and exported to a shape file w i t h the object data for each pipe. The shape file was v i e w e d and plot ted i n A r c V i e w ® G I S 3 . 2 A and exported as a spreadsheet file as the m a i n data file for analyses. The Dis t r i c t ' s hydrau l ic m o d e l data are stored i n a . D W G format but the m o d e l uses l inks and nodes w h i c h do not comple te ly relate to specific pipes. In m a n y cases, l inks represent a number o f pipes i n series and required buffering. The break his tory for M a p l e R i d g e is l imi t ed i n detai l , but is re la t ively long . Breaks are recorded b y operat ion and maintenance staff and stored i n paper form. The recording o f break locat ions as w e l l as the amount o f environmental and break informat ion varies among the records w i t h i n a g iven year and dur ing the per iod o f record. F i e l d crews typ ica l ly record locat ions w i t h respect to the nearest cross street but do not record the exact locat ion o f the break. A s a result, i f a pipe experienced mul t ip le breaks but the exact locations o f these breaks are not identif ied, it cannot be determined i f the p ipe fai led at the same 94 locat ion or at mul t ip le sites a long the pipe. H o w e v e r , ident i f icat ion o f the exact loca t ion was not necessary because the purpose o f this analysis was to determine i f and w h e n a p ipe broke and not the exact loca t ion o f the break a long each pipe. T o create the electronic break database, the data f rom the field c rew reports were entered into an electronic spreadsheet. P ipe I D f rom the water ne twork map were added to the break record. The data worksheet was then converted to a database that was then exported as a . D B F fi le to G I S where it was plotted for v i sua l iza t ion . The Dis t r i c t does not have detailed so i l informat ion. T o construct these data, a number o f secondary sources inc lud ing federal, p r o v i n c i a l and loca l government reports (Golder , 2002; Lu t tmerd ing , 1981; Lut tmerd ing , 1980) were reviewed. W h i l e the p rov inc i a l reports use agricul tural so i l classifications, so i l groups based o n the parent mater ial (i.e., the upper stratigraphic unit) better represent the so i l types at instal lat ion depth. The three m a i n classes o f parent material are: marine c lay , marine sand or eol ian silt. The p rov inc i a l s o i l maps were converted into an electronic fo rm and intersected w i t h the water ne twork maps as described p rev ious ly i n the l i n k i n g data section. T o obtain informat ion regarding surface material , the ortho-photographs o f surface features were plotted as background to the water dis t r ibut ion network. The possible surface material types are asphalt, gravel ( typ ica l ly representing a road shoulder) and landscaping. Whether the water m a i n is under a boulevard or roadway was s i m i l a r l y determined and recorded. A GIS-based map o f the roads, over la id o n the water pipe network and knowledge based o n the Dis t r i c t o f M a p l e R i d g e Transportat ion P l a n was used to classify the traffic on the ground surface above the water ma in . Categories o f traffic loads that were used include those attributed to loca l , col lector , arterial, and commerc i a l roads, and lanes and those 95 exper iencing no traffic, such as boulevards. In the case where a p ipe is under two or more road classif icat ions, the c lass i f icat ion corresponding to the heavier traffic road function was selected. A s is the case for many uti l i t ies, the qual i ty and amount o f construct ion details p rov ided i n as-builts varies depending on the designer, contractor and t ime per iod dur ing w h i c h the m a i n was constructed. Thus , the confidence i n these details also varies. H o w e v e r , i n spite o f the var iab i l i ty o f records, ut i l i t ies m a y f ind that construct ion standards, i n place for years, m a y provide rel iable information. F o r example, the standard cover used for construct ion i n M a p l e R i d g e is general ly 900 mi l l imet res (three feet) and has been ver i f ied b y interviews w i t h staff (Thain , 2005). Da ta regarding bedding and b a c k f i l l material , and pipe l i n i n g and protect ion practices were determined through interviews w i t h staff. Rela t ionships between the bedding and pipe material were established and a data set o f bedding mater ial was created. S i m i l a r l y , whether p ipe protect ion exists for pipes i n the study area was determined us ing the s o i l type for each pipe. Where the so i l type indicates c l ay and the p ipe mater ial is ducti le i ron , the pipe was considered on the basis o f practice to be wrapped. A l l other pipes are considered to be unprotected. T h i s assumption is identif ied as one w h i c h m a y require further ver i f ica t ion. W h i l e some data were obtained us ing secondary sources, it was not possible to obtain or construct a l l data. A l s o , w h i l e m i n i m a l scrubbing o f the data was required, there were some gaps i n data. F o r example , when the Dis t r i c t replaces a pipe or segment o f a pipe, it does not retain any his tory o f the pipe. Consequent ly , there is no informat ion regarding the former his tory o f the age, material or diameter o f pipes that have been ' replaced unless the data are recoded on the break record fo rm ( typ ica l ly this w o u l d o n l y be diameter and mater ial data). 96 Preliminary analysis and knowledge discovery. K n o w l e d g e d i scovery results for this case are reported i n Figures 3.6 to 3.8 and Tables 3.2 and 3.3. In the study area, 69 percent o f the pipes are duct i le i ron , 26 percent are asbestos cement, f ive percent are cast i ron and less than one percent is steel pipe. The number o f breaks over time,, plotted i n F igure 3.6, m a y help managers develop strategies and budgets. W h i l e data indicate a stable number o f annual breaks over the past years, a predic t ion m o d e l w o u l d be useful to estimate i f and w h e n the number o f breaks w i l l increase. O f the 47 breaks that occurred, there were 32 first-time breaks. D u r i n g 1983-1999, ten breaks occurred i n nine pipes and these pipes were replaced. Because M a p l e R i d g e does not retain data on pipes that are replaced except what is o n the break records ( w h i c h m a y record the p ipe diameter and material) , o n l y 37 breaks had associated pipe diameter, material and age data. The usefulness o f knowledge d iscovery is i n determining patterns o f p ipe breakage and insights based o n these patterns. One example pattern is i l lustrated i n F igure 3.7 and Table 3.2 w h i c h show that the major i ty o f first breaks occur when pipes are between 15 to 19 years o l d , and that pipes o f the 1960-1974 vintage have a s igni f icant ly higher rate o f failure than others, and thus attention should be g iven to these pipes. In addi t ion, other relationships such as that between breaks and pipe size (as shown i n Table 3.3) and breaks and so i l condi t ions (as shown i n F igure 3.8) can prov ide insights, such as the amount o f failures i n duct i le i ron pipes that are instal led i n c lay soils . Current ly , the water m a i n replacement p rogram o f M a p l e R i d g e is targeted at rep lac ing asbestos cement pipes. A s a result o f this project, M a p l e R i d g e ini t iated a so i l cor ros iv i ty survey and plans to correlate those results w i t h the so i l data created i n this study. T h e y are also examin ing ways to improve pr ior i t iza t ion o f the rehabil i tat ion o f water mains us ing the n e w l y gained knowledge . Practices have also been revised inc lud ing : data 97 o n replaced pipes are n o w retained, f ie ld data are n o w be ing col lected w h e n new pipes are instal led, and new water m a i n break forms have been implemented . C o m m u n i c a t i o n on data co l l ec t ion is i m p r o v i n g as departments w o r k together to col lect and share data and have a better understanding o f each other 's role i n ach iev ing the c o m m o n goal o f effective asset management. 3.6 DISCUSSION The process o f creating a data schematic among mul t ip le databases for data compi la t ion , analysis and knowledge d iscovery m a y y i e l d valuable insights for i m p r o v i n g asset management practices. It expands the data avai lable for analysis for both present and future applicat ions and respects decentral ized data input and management. The use o f decentral ized data storage reduces problems related to data ownership among units w i t h i n organizations regarding data co l lec t ion , management and disseminat ion o f water, sewer, drainage or other infrastructure information. A s i n any business process, it is important to document the purpose for constructing, structuring, sourc ihg and l i n k i n g the data schematic. A c lear ly documented outl ine o f the relationships between the k e y data sets is important for future managers and those u t i l i z i n g the data. N S G M I (2003) suggests some k e y relat ing informat ion such as asset number, asset locat ion and w o r k order numbers. The w o r k undertaken herein suggests that those co l lec t ing data w i t h i n an organizat ion should share informat ion regarding the purpose for co l lec t ing and selecting the amount and type o f informat ion to avo id dupl ica t ion and focus efforts. D a t a managers must also be aware o f disclosure control i n w h i c h important or confident ial data m a y be inadvertently connected and unintent ional ly released to th i rd parties (Torra et al., 2004). F o r pub l i c ut i l i t ies, the issue o f private and proprietary 98 i informat ion is important and should be considered (e.g., cons ider ing the F reedom o f Information A c t i n B r i t i s h Co lumb i a . ) . Process ing o f data is a significant step i n data schematic construct ion and relat ing data. F o r example , i n the M a p l e R i d g e case study, use o f agricul tural so i l classif ications resulted i n too many s o i l zones for analysis but reducing the number o f classif icat ions and us ing the parent mater ial categories improved the understanding o f the relat ionship between s o i l type and the number o f water m a i n breaks. S u m m a r i z i n g and aggregating data requires expertise and knowledge o f the intended analyses. In seeking data sources, ut i l i t ies should consider data that other agencies are co l lec t ing even i f they ate do ing so for different purposes and determine whether the data are useful as a secondary source. A p p l i c a t i o n o f new technologies, such as infrared, electro- magnetic surveys and L i g h t Detec t ion and R a n g i n g ( L I D A R ) m a y be employed to capture relevant data. U t i l i t i e s face the challenge o f capturing undocumented inst i tut ional informat ion and knowledge over the next decades as their baby-boomer staff retire. A s demonstrated herein, in terviews are useful i n capturing these data, but the processes o f in te rv iewing staff can be awkward . The authors found that staff shared more informat ion w h e n interviewers began the session b y m a k i n g observations and ask ing about the v a l i d i t y o f these observations, and then ask ing specific questions. M a p s and other v i sua l cues usual ly triggered the in terviewee 's memory . It is important for interviewers to pose questions and present the exercise as so l i c i t ing informat ion to a id i n knowledge d iscovery rather than c r i t i c i sm o f past practices. It is equal ly important to, where appropriate, val idate the observations and experiences o f staff. It is noted that retired staff seem uncomfortable w i t h 99 general observations unless they cou ld also provide exceptions. T h o u g h there m a y be confidence i n o ra l ly recorded data, ver i f ica t ion should be undertaken where possible . A l imi ta t ion o f constructed databases is the poss ib i l i ty that created data are more correlated than data that are col lected independently. F o r example , i n the case study, a l l duct i le i ron pipes are defined as cement mortar l ined and the p ipe mater ial determined the b a c k f i l l and bedding data. Th i s m a y affect the amount o f explanatory variables i n an analysis. Other l imita t ions include the fact that data m a y not a lways be accurate or quantifiable. The l imita t ions m a y be overcome b y a program to ver i fy the data over t ime, w h i c h is a long-term exercise. F o r predic t ing future water m a i n breaks, it m a y be diff icul t to determine i f variables are significant i f there is a l imi t ed his tory o f breaks. It is important that the data are stored and that future break data be added to increase the his tory and data set for further analysis. 3.7 CONCLUSIONS G r i g g (2005) has ident i f ied the need for standardized m a i n break databases and cont inued research regarding database development as strategic for in formed infrastructure management though the topic o f databases has been the subject o f other authors such as D e b et al, (2002); H a b i b i a n (1992), and O ' D a y , 1982). H o w e v e r , surveys and literature indicate that data are t yp i ca l l y scarce. C o m p o u n d i n g this, water m a i n break data i n central ized databases are not c o m m o n and approaches and techniques to relate and manage data for analysis are needed. The approach presented herein expands the sources ( inc luding , i n particular, oral transmission) and the amount o f data for asset management. Researchers and managers m a y gain insight into a system b y systematical ly sourcing, relat ing, processing, b lend ing 100 and ana lyz ing data. The framework is f lexible , anticipates the evolu t ion o f data co l lec t ion , b u i l d i n g , ver i f ica t ion and storage and a l lows for a variety o f users. It does not abruptly disrupt data co l lec t ion and warehousing practices. A k e y benefit o f this approach o f relat ing data is that it a l lows managers to continue to expand data co l lec t ion because databases are decentralized. It also fosters a dialogue o f data development, knowledge d i scovery and informat ion processing across an organizat ion. It is f lex ib le and can be adapted to a l l ut i l i t ies, whether they are smal l , m e d i u m or large, and regardless o f the uniqueness o f the data col lected and organizat ional framework. The recording o f future breaks and ver i f ica t ion tasks are important for b u i l d i n g confidence i n the data and for augmenting it. A long-term strategy should be developed that verifies and improves the breadth o f data and confidence i n the database. The tasks can be opportunist ic , undertaken w h e n other repairs o f the system are occur r ing , or systematic. W h e n pipes are replaced or new service connections are instal led, opportunities arise for obtaining phys i ca l samples and ver i fy ing the nature and condi t ion o f p ipe protection, bedding, b a c k f i l l and the pipe exterior. A n example systematic program is a u t i l i ty -wide survey o f s o i l cor ros iv i ty i n var ious so i l classif ications and an exercise that improves the understanding o f the relat ionships among so i l classifications, cor ros iv i ty and breaks. T h i s approach can also be used for sewerage, drainage and other systems for i m p r o v i n g asset management, operation and maintenance analysis , performance and pub l i c accountabil i ty. F o r example, b y relat ing and ana lyz ing grease bui ld-up areas, complaints , slope, f l ow and m o d e l f lushing veloci t ies , a sewer system manager can analyze performance and resource efforts and develop plans to address p rev ious ly "un-connected" p rob lem areas. A s w e l l , for a road network, constructing relat ionships between complaints , ( 101 traffic vo lume , speed and crash data, l igh t ing leve l , signage, road condi t ion and geometry, managers m a y be able to improve road safety and predict h igh crash locations. A s observed i n the L a i t y V i e w area case study, the process and results improve communica t ion and the basis for dec is ion m a k i n g . N e w insights can assist managers i n focusing effort and resources and d i scover ing unanticipated issues and challenges. W h i l e the focus o f ut i l i t ies has been on co l lec t ing and storing data, the next step w i l l be i n the appl ica t ion o f data m i n i n g and knowledge discovery . A s more tools become avai lable to analyze data, managers need to consider the business intel l igence that it should emp loy to w i s e l y invest resources to meet future demands. 3.8 A C K N O W L E D G E M E N T S N The authors gratefully acknowledge the Dis t r ic t o f M a p l e R i d g e for p rov id ing data and support i n the fo rm o f employee resources, Mess rs . D . T h a i n (ret.), H . Y o u (ret.) and J . Scherban f rom the Dis t r i c t for p r o v i d i n g tacit informat ion, and Professors A . D . R u s s e l l , J . A twa te r and H . Schr ier at the U n i v e r s i t y o f B r i t i s h C o l u m b i a ( U B C ) for their insights and suggestions. M r . A . M a l y u k o f the Dis t r ic t o f M a p l e R i d g e assisted i n data processing and compi l a t i on and Mess r s B . K a m p a l a , M . A . S c . Candidate , U B C , and W . Johnstone, P r i n c i p a l , Spat ial V i s i o n Group assisted i n r ev i ewing the manuscript . 102 3 . 9 R E F E R E N C E S Cooper , N . R . , B l a k e y , G . , Sherwin , C , T a , T. , Whi te r , J. T . and W o o d w a r d , C . A . , 2000. The use o f G I S to develop a probabi l i ty-based trunk mains burst r i sk mode l . Urban Water, 2:2000: 97-103. D a v i s , D . N . , 2000. Agent-based dec i s ion support f ramework for water supply infrastructure rehabil i ta t ion and development. Computers, Environment and Urban Systems, 24:2000: 173-190. D e b , A . R . , Grab lu tz , F . M . , Hasi t , Y . J . , Synder, J .K. , Longanathan, G . V . and A g b e n o w s k i , N . , 2002. P r i o r i t i z i n g Water m a i n Replacement and Rehabi l i ta t ion . A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. G o l d e r Associa tes L t d . , 2002. Geotechnica l Input to the seismic vu lnerab i l i ty assessment for the Dis t r i c t o f M a p l e R i d g e , B . C . 500-4260 S t i l l Creek D r i v e , Burnaby , B C , V 5 C 6 C 6 . G r i g g , N . S . , 2005. Assessment and R e n e w a l o f Water D i s t r ibu t ion Systems. Journal AWWA, 97:2: 58-67. G r i g g , N . S., 2004. Assessment and renewal o f water dis t r ibut ion systems. A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. 103 Hab ib i an , A . , 1992. D e v e l o p i n g and u t i l i z i n g databases for water, m a i n rehabil i tat ion. Journal AWWA, 84:7: 75-79. K l e i n e r , Y . and Rajani , B . B . , 1999. U s i n g l imi t ed data to assess future needs. Journal AWWA, 91:7: 47-62. K l e i n e r , Y . and Rajan i , B . B . , 2001 . Comprehens ive rev iew o f structural deterioration o f water mains : statistical models. Urban Water, 3:3: 131-150. Lut tmerd ing , H . A . , 1981. Soils of the Langley-Vancouver map area: Report NoJ5, Volume 6. B . C . M i n i s t r y o f Env i ronment Assessment and P l a n n i n g D i v i s i o n , K e l o w n a , B C . Lut tmerd ing , H . A . , 1980. Soils of the Langley-Vancouver map area Report No.15 Volume 1. B . C . M i n i s t r y o f Env i ronment Assessment and P l a n n i n g D i v i s i o n . K e l o w n a , B C . N G S M I , (Nat iona l G u i d e to Sustainable M u n i c i p a l Infrastructure - Infraguide), 2003. Bes t Practices for U t i l i t y - B a s e d Data . Infraguide - Potable Water . Ot tawa, O N , Canada. N G S M I (Nat iona l G u i d e to Sustainable M u n i c i p a l Infrastructure - Infraguide), 2002. Deter iora t ion and Inspection o f Water Di s t r ibu t ion Systems. Infraguide - Potable Water . Ot tawa, O N , Canada. I 104 N a t i o n a l Research C o u n c i l o f the N a t i o n a l A c a d e m i e s - Commi t t ee o n P u b l i c Water S u p p l y Dis t r ibu t ion Systems: Assess ing and R e d u c i n g R i s k s , 2005 . Public Water Supply Distribution Systems: Assessing and Reducing Risks - First Report. O ' D a y , D . K . , 1982. O r g a n i z i n g and ana lyz ing leak and break data for m a k i n g m a i n replacement decisions. Journal A WW A, 74:11: 588-594. O ' D a y , D . K . , W e i s s , R . , C h i a v a r i , S., and B l a i r , D . , 1986. Wa te rma in Eva lua t i on for Rehabi l i t a t ion / Replacement . A W W A Research Founda t ion ( A W W A R F ) and U S Env i ronmen ta l Protect ion A g e n c y ( U S E P A ) , A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235." Pellet ier , G . , M a i l h o t , A . and V i l l e n e u v e , J . -P. , 2003 . M o d e l i n g water pipe breaks - three case studies. Journal of Water Resources Planning and Management, 129:2: 115-123. Rajan i , B . B . and M a k a r , J . , 2000. A methodology to estimate remain ing service life o f grey cast i ron water mains . Canadian Journal of Civil Engineering, 27:6: 1259-1272. Rossi ter , D . G . , 2005. A compendium of on-line soil survey information - Soil survey institutes and activities. A c c e s s e d M a y 31 , 2005, h t tp : / /www. i tc .n l / rossiter/research/ rsrch ss sources.html. 105 Sch ick , S., 2005. Toronto bridges data needs i n water network project. ITBusiness.ca July 21, 2005. Acces sed J u l y 22, 2005 http:/ /www.itbusiness.ca/index.asp7theaction =61&sd i= 59540. T h a i n , D . , 2005. Water mains i n the Dis t r i c t o f M a p l e R i d g e . A . W o o d , ed., M a p l e R i d g e , Interview notes. Tor ra , V . , Domingo-Fer re r , J . and Torres, A . , 2004. Da ta m i n i n g methods for l i n k i n g data c o m i n g f rom several sources. 3rd Joint Un/ECE-Eurostat Work Session on Statistical Data Confidentiality, Monographs in Official Statistics, L u x e m b o u r g , Eurostat. 143-150. V a n i e r , D . J . , 2001 . Asse t Management : " A " to " Z " , American Public Works Association Annual Congress and Exposition - Innovations in Urban Infrastructure Seminar, Phi lade lph ia , U . S . September, 2001 . 1-16. W o o d , A . and Lence , B . J . , Assessment o f Water M a i n Break D a t a for Asse t Management . Journal AWWA, 98:07. X u , C . and Goul ter , I . C , 1998. Probabi l i s t ic mode l for water dis t r ibut ion re l iabi l i ty . Journal of Water Resources Planning and Management, 124:4: 218-228. 106 V Table 3.1 Water main break data availability for Maple Ridge Data element Availability for "off the shelf analysis Secondary source Notes Confidence/reliability Pipe material Available in AutoCad in .DWG format High Pipe diameter Available in AutoCad in .DWG format and GIS High Type of water service Available in Ortho- photographs (0.5metre/1.6 feet * and0.15metre/0.5 feet intervals) Requires interpretation of data available Medium Under boulevard or roadway Available in 1 Ortho- photographs (0.5metre/1.6feet and0.15metre/0.5 feet intervals) Requires interpretation of data available High Year of installation /age Available in AutoCad in .DWG format and GIS Medium - high Depth of cover Orally recorded information Based on standards of the day Medium - but could be verified over time Length of pipe segment Available in AutoCad in .DWG format and GIS Medium Normal operating pressure Available in model and in .DWG format Need buffering to assign pipe with node data Medium - High Type of joint Not available Field investigation commenced for future pipe breaks Low Condition of pipe exterior Not available Field investigation commenced for future pipe breaks Bedding material Orally recorded information Data inferred from known pipe material, future field verification required Medium - but could be verified over time 707 Data element Availability for "off the shelf analysis Secondary source Notes Confidence/reliability Pipe protection (wrapped/anodes) Orally recorded information Use of soil maps to infer information, future field verification required Medium - but could be verified over time Backfill material Orally recorded information Data inferred from known pipe material, Future field verification required Medium - but could be verified over time Traffic classification or type of road usage Transportation plan designation and maps Need better understanding of traffic loading Type of pipe lining Orally recorded information Data inferred from known pipe material, future field verification required Medium - but could be verified over time Typical flow in area of break Available in hydraulic model and in .DWG format Some more modeling information (typical flow) Medium-high (from model confidence) 108 Table 3.2 Water main breaks for a given year of installation (1983-1999) Year of Installation Number of Pipes Length in metres (feet) Number of breaks as an age group Percentage of pipes in age group with breaks 1955-1959 14 1,687 (5,535) 4 29% 1960-1964 32 • 2,968 (9,738) 8 25% 1965-1969 12 1,511 (4,958) 7 58% 1970-1974 52 4,791 (15,719) 11 • 21% 1975-1979 56 4,030 (13,222) 3 5% 1980-1984 56 4,122 (13,524) 1 2% 1985-1989 56 4,584 (15,040) 3 5% 1990-1994 96 7,452 (24,450) 0 0% 1995-1999 51 3,003 (9,853) 0 0% Unknown 4 29 (95) 0 0% Total 429 34,177 (112,135) 37 9% 109 Table 3.3 Pipe breaks for pipes of a given diameter (1983-1999) Pipe diameter in millimeters Number of Pipes Number of breaks Percentage of pipes for each diameter group with breaks 100 2 0 0.0% 150 246 26 10.6% 200 117 7 6.0% 250 62 4 6.5% 300 2 0 0.0% 350 0 0 0.0% 429 37 9.0% 110 Figure 3.1 A schematic for constructing and using water main break data for knowledge discovery Verificalion"and:;upaating;of;aata';:;™s Soil maps transportation I plans , Ortho- photographsl reports AM/FM systerrf data Break data files Linking or relating data using identifiers common to two or more databases Processing Application Software: -GIS -CAD - spreadsheets Human analyst and application software 4 Data Analysis and Application Analysis tools -GIS -CAD - Spreadsheets - Databases Knowledge Discovery Strategies and Plans ill Figure 3.2 Process for digitizing and creating data from archival geographical data Scan hard copy image and create TIFF image Create closed polyline drawing Use GIS to create polygon coverage (closed polylines) and identify attributes to be linked Apply GIS processing tools (e.g., intersect function) to pipe network and the polygon to generate shape file Export attribute data of the shape file to analysis file (e.g., spreadsheet file) XZ1 = 5 Sde ndte%Mo%> 1 § for titisshfidkd I f = : :: 1 I i Note: Title of data columns Number Month Years Pipe of Pipe Material Year of Year Of Of of Years In length breaks ID Size Material Code installation Failure Failure Service Ground (metres) to 1999 Has pipe been Soil Number Under replaced since zone of soil boulevard Surface Traffic Pipe break? type zones or road material classification lining Bedding Backfill Q V C_number HGL Max demand day Pipe protection C factor Roughness (l/s) (m/s) (Hazen) (metres) pressure _(PSI) 112 Figure 3.3 Linking hydraulic model data with network data Determine hydraulic model input and output data and relationships of data. For example, are pressures calculated only for nodes or are they also assigned to model links? Do the pipe link data include all the data being sought for analysis? Combine hydraulic model output node pressures and link flow data for a given pipe ID. Determine pipe network information such as material, diameter, date of installation, length, etc. How does the pipe network relate to the model, i.e., are there one-to-one relationships or is the model skeletonized? Combine hydraulic model data with pipe network data. Export combined attributes to analysis file 113 Figure 3.4 Buffering data to create data relationships using GIS Link ID 1 is a link that defines connectivity in the hydraulic model. Each linkjs-~ identified as unique in the> model. Buffering the Link ID 1 data to Pipe ID 1 and Pipe ID 2 gives the attributes of Link ID 1 to the network pipes Pipe ID 1 and Pipe ID 2. Pipe ID 1 and Pipe ID 2 are network pipes in GIS. Each pipe identified as unique. ID 1 Pipe ID 1 with Link ID1 data Pipe ID 2 with Link ID1 data Steps to create buffered data 1. Create shape files of model data and network data. 2. Buffer data using capabilities of GIS. 3. Export the attributes of the buffered shape file as a spreadsheet file. 114 Figure 3.5 Maple Ridge water main break analysis data schematic Tacit data: bedding, construction standards, pipe protection, cover, etc. Project report: pipe materia], installation date, pipe length, etc. Hydraulic model: flow, pressure, pipe material, friction coefficient, etc. Geological Survey of Canada (GSC) data Ministry of Environment (MOE) soil data Pipe inspection record: pipe material, etc. Road as-built: surface material, road function, etc. Break report form: break data, date, costs and effort of repairs Primary sources Secondary sources Potential link (e.g. pipe ID) Potential input to analysis Accounting system: break activity costs, annual break expenditure, etc. Pavement management system: surface material, road function, etc. Traffic network plan: road function, traffic volume, etc. Traffic volume data base: traffic volume, road function, etc. 115 Figure 3.6 Cumulative breaks in Laity View area (1983-1999) 116 Figure 3.7 Number of years in service when break occurred in pipes (1983-1999) 10 Number of years in service Note: Of the 47 breaks in the area, no pipe age data were available for 10 of the 47 breaks. 117 Figure 3.8 Number of breaks for pipes in a given soil type installation (1983-1999) Asbestos cement pipes Cast iron pipes Ductile iron pipes • Marine Clay • Eolian Silt • Silty Clay 2 O Marine Clay a Marine Clay • Eolian Silt 118 C H A P T E R 4 USING W A T E R M A I N B R E A K D A T A TO I M P R O V E ASSET M A N A G E M E N T F O R S M A L L AND M E D I U M UTILITIES A version of this paper has been submitted for publication to ASCE Journal of Infrastructure Systems as Using Water Main Break Data to Improve Asset Management for Small and Medium Utilities by A. Wood and B. J. Lence. 119 PREFACE In the previous chapter, I created data and a number o f databases to demonstrate some techniques that ut i l i t ies can emp loy to enr ich their asset management data sets. In Chapter 4 ,1 demonstrate that once data are created and l i nked for analysis, uti l i t ies can use these data w i t h i n a f ramework that I developed for i m p r o v i n g their asset management practices. The intent o f this research is not to create a new m o d e l , but to develop a framework that uses break predic t ion models that smal l and m e d i u m size uti l i t ies can apply. The so i l and surface mater ial data that were created i n Chapter 3 were used i n the experimental appl ica t ion o f this f ramework and give 's ignif icant insights into the factors that influence pipe breaks. W h i l e the data created p rov ided a demonstration case, there was not sufficient informat ion to apply a l l the models that I i n i t i a l l y proposed to investigate. The use o f statistical determinist ic t ime-l inear and t ime-exponent ial models c o u l d be suff icient ly demonstrated w i t h the data created i n Chapter 3, but there were insufficient data to obtain meaningful results f rom an appl icat ion o f the surv iva l analysis and K A N E W (Deb at al, 1998) models . Thus , a l l four types o f models were appl ied i n an effort to obtain p r o o f o f concept, but o n l y two types o f models are reported i n the manuscript that comprises Chapter 4. It is important for ut i l i t ies to continue to create and mine data. T h i s research reinforces the no t ion that not a l l data m a y be useful for app ly ing sophisticated models but that unsophist icated pipe break models can provide insights into the performance o f a water m a i n system, be used i n ident i fy ing system specific factors that m a y cause breaks, guide the development o f a water m a i n break data co l l ec t ion strategy, be used to identify groups 120 o f pipes, their h is tor ica l and predicted break frequency to further investigate and pr ior i t ize for rep lac ing and thus be a benefit to communi t ies . 121 4.1 INTRODUCTION Water uti l i t ies have aging and deteriorating infrastructure and must pr ior i t ize the replacement o f their water mains to m i n i m i z e pipe breaks. Breaks result i n loss o f water to k e y businesses and cr i t i ca l facil i t ies, m a y lead to damage o f other infrastructure, and have been ident i f ied as a pa thway for m i c r o b i a l contaminat ion o f dis t r ibut ion systems ( A W W A and E E S , 2002). The need for rehabil i tat ing aging water mains is increasing, the costs o f repairs and replacement can be h igh , and the impact o n customers potent ia l ly significant ( U S E P A , 2001). Asse t management practices are general ly used to pr ior i t ize p ipe replacements and thereby identify investment strategies that, on one hand, avo id premature replacement o f pipes (i.e., unnecessary pre-investment o f funds), and o n the other hand, avo id water m a i n breaks, commensurate interruptions i n service and the costs o f damage. A n effective asset management dec is ion is dependent on the ab i l i ty to determine the future performance o f water mains b y predic t ing water m a i n breaks, and ident i fy ing h o w such breaks m a y occur. M u c h research has focused o n the development o f models for predic t ing water m a i n breaks and pipe deterioration, but the use o f such models is not c o m m o n among uti l i t ies. In addi t ion, the amount and qual i ty o f water m a i n break data avai lable for deve lop ing or implement ing these models varies among uti l i t ies ( W o o d and Lence , 2006) . M a n y uti l i t ies lack data and are not confident i n the data they have and this is general ly an impediment to their invest ing i n pipe predic t ion models . H o w e v e r , they can create and relate data that can be useful for asset management ( W o o d et al, 2007). T h i s paper develops a f ramework that guides uti l i t ies i n ident i fy ing k e y data to be used i n asset management i n general and spec i f ica l ly for p ipe break predic t ion mode l ing v and selecting the most appropriate mode l for predic t ing water m a i n breaks. T h i s 122 informat ion m a y then be used to enhance the development o f replacement priori t ies based on forecasted breaks, the maintenance o f the database, and the ident i f icat ion o f future data acquis i t ion programs. It provides the u t i l i ty w i t h a method for cons ider ing future p ipe breaks i n the analysis o f pipe pr ior i t iza t ion strategies, and it incorporates exis t ing tools for data management and analysis that are w i d e l y avai lable and easy to implement b y smal l and m e d i u m size uti l i t ies. The framework is appl icable to ut i l i t ies w i t h va ry ing amounts o f data, and it is demonstrated here w i t h a case study based on the L a i t y V i e w area o f M a p l e R i d g e , B . C and constructed data. The f o l l o w i n g sections rev iew the avai lable techniques for predic t ing pipe breaks, the factors that influence break predict ions, the framework developed for assisting i n asset management o f pipe networks and the results o f the M a p l e R i d g e example implementa t ion o f this framework. The framework can be appl ied wi thout creating and constructing data, but the usefulness wi thout such efforts is l imi ted . 4 . 2 W A T E R M A I N B R E A K S A number o f authors analyze and report o n the causes o f breaks, i nc lud ing O ' D a y (1982), M a r k s et al. (1987), M a l e et al. (1990), Sav ic and Wal ters (1999), Rajan i and M a k a r (2000), Rajan i and K l e i n e r (2001) and D i n g u s et al. (2002). A c c o r d i n g to Rajan i and Tesfamar iam (2005), a combina t ion o f circumstances leads to p ipe failure i n most cases and different factors cause failure i n different pipe networks. The causes o f breaks include deterioration as a result o f use (e.g., internal corrosion) , phys i ca l loads appl ied to the pipe (e.g., traffic, frost), l im i t ed structural resistance o f the pipe because o f construct ion practices dur ing instal la t ion and dec l in ing resistance over t ime (e.g., corros ion, aging factors). D i n g u s et al. (2002) surveyed the 46 largest A m e r i c a n Wate r W o r k s A s s o c i a t i o n Research Founda t ion ( A w w a R F ) member uti l i t ies i n 1997 and note mul t ip le c o m m o n 123 failure modes for cast i ron p ipe systems. Cor ros ion , improper insta l la t ion and ground movement are the three most c o m m o n causes o f pipe failure. A c c o r d i n g to L e v e l t o n (2005), cor ros ion is dependent o n a number o f factors i nc lud ing mater ial , s o i l type, chemica l characteristics o f so i l , so i l bacteria and stray electr ical currents. Prediction modeling of water main breaks. B reak predic t ion models have been developed to help the water industry understand h o w pipes deteriorate and w h e n pipes w i l l break i n the future. These models are t yp i ca l l y grouped into two classes - statistical and phys ica l -mechanica l models (K le ine r and Rajan i , 2001). Statist ical models use h is tor ica l pipe break data to identify break patterns and extrapolat ion o f these patterns to predict future p ipe breaks, or degrees o f deterioration. Phys ica l -mechan ica l mode ls predict failure b y s imula t ing the phys i ca l effects and loads on pipes and the capaci ty o f the pipe to'resist failure over t ime. Statist ical models are t yp i ca l ly characterized as either determinist ic or probabi l is t ic equations ( K l e i n e r and Rajan i , 2001). U n d e r the determinist ic models , the p ipe breakage is estimated based o n a fit o f pipe breakage data to var ious time-dependent equations, w h i c h m a y represent the cumula t ive p ipe breaks as a function o f t ime from date o f instal lat ion or f rom the earliest date o f avai lable break data, most c o m m o n l y are t ime-l inear (Ket t ler and Goul ter , 1985) or t ime-exponent ia l functions (Shamir and H o w a r d , 1979; W a l s k i , 1982 and K l e i n e r and Rajan i , 1999). P r io r to fi t t ing these functions, pipes are part i t ioned into groups that have s imi la r characteristics, and the functions are evaluated for these groups. The ' characteristics used to sort the pipes are based o n the factors that are assumed to influence breaks such as p ipe age, pipe material , diameter, or so i l type. Probabi l i s t ic models predict not o n l y the failure potential , but the dis t r ibut ion o f failure. These models are more complex than deterministic models and require more data. E x a m p l e s o f these include 124 cohort su rv iva l , such as K A N E W (Deb et al. 1998), Bay es i an diagnostic, break cluster ing, s e m i - M a r k o v C h a i n and data f i l ter ing methods. Phys ica l -mechan ica l models t yp i ca l l y fa l l into one o f two classes: determinist ic models w h i c h estimate p ipe failure based on s imula t ion o f the phys i ca l condit ions affecting the pipe (Doleac et al, 1980, and Rajani and M a k a r , 2000), or probabi l i s t ic models that use a dis t r ibut ion o f input condi t ions, such as rate o f corros ion, to predict the l i k e l i h o o d and dis t r ibut ion o f p ipe failure ( A h a m m e d and Melche r s , 1994). These models have been developed p r i m a r i l y for cast i ron and cement pipes. P h y s i c a l models have significant data needs. K l e i n e r and Ra jan i (2001) suggest that o n l y larger diameter mains w i t h cos t ly consequences o f failure m a y jus t i fy the required data co l lec t ion efforts for these models , and that statistical models based on fewer data m a y be used to gain insights for future performance. Rajan i and Tesfamar iam (2005), i n us ing a probabi l is t ic approach, suggests that breaks and causes o f breaks for any part icular water dis t r ibut ion network are system-specific and that a u t i l i ty must create its system-specific m o d e l based o n the deterioration factors that are relevant for that ut i l i ty . W h i l e smal l and m e d i u m uti l i t ies typ ica l ly have the capaci ty to use statistical determinist ic models , the implementa t ion o f phys i ca l models is not pract ical due to the data co l l ec t ion efforts and m o d e l maintenance required. F o r smal l and m e d i u m uti l i t ies, it is often most important to gain insights about the rate o f pipe breakage, that is , whether the quanti ty o f expected breaks is increasing l inear ly or exponential ly. K n o w i n g the rate o f change for u t i l i ty managers is important because budgets and performance are based i n part o n future needs such as the number and rate o f breaks i n the system. Factors for predicting water main breaks. A number o f studies identify factors for predic t ing water m a i n breaks, though what is considered relevant data appears to be 125 specific to the system investigated. O ' D a y (1982) reviews break studies i n Manhat tan and B inghamton , N e w Y o r k , and cites a number o f studies that use age as an indicator for predic t ing break rates for cast i ron pipes. H e notes, however , that age alone is a poor predictor o f m a i n break patterns and identifies the major determinants o f water m a i n break rates, as l oca l i zed factors such as cor ros ion condit ions, construct ion practices and external loads. H e also finds that so i l type affects external forces on water mains , such as shrink- swe l l , frost penetration and external corrosion. A c c o r d i n g to Jacobs and K a r n e y (1994), pipe age range is an effective basis for models because pipes o f a g iven age range are typ i ca l ly un i fo rm w i t h respect to manufacture, instal la t ion and to a large extent, operating condi t ions. M o r e o v e r , pipe instal led i n geographical ly contiguous sections often share s imi la r so i l condi t ions , instal lat ion condit ions and pressure regimes. In their study, Jacobs and K a r n e y group pipes based on material , diameter and fa i r ly broad age ranges and develop regression relationships for pipe breakage versus age and versus p ipe length. Sav ic and Wal ters (1999) suggest that the causes o f water m a i n failures m a y be spli t into pipe qual i ty and age, type o f environment, qual i ty o f construct ion workmansh ip and service condi t ion and f ind that age, length, and diameter are the most important variables i n inf luencing p ipe bursts. Ket t l e r and Goul te r (1985) f ind that break rate, age and material are related for asbestos cement and cast i ron pipes. In their study, no single type o f failure o f asbestos cement pipes exhibi ted a marked change i n the rate o f failure w i t h t ime, w h i l e there were distinct changes i n the failure rate w i t h t ime for some types o f failure i n cast i ron pipes. A c c o r d i n g to M a l e (1990), different manufacturing processes o f cast i r on pipes can account for differences i n durabi l i ty . Coope r et al (2000) apply a probabi l i s t ic approach to estimate trunk m a i n failure probabi l i ty , based on four k e y variables: number o f buses per hour, pipe diameter, s o i l cor ros iv i ty and density o f pipes i n a g iven area. T h e y f ind that 126 \ pipe age and mater ial are important factors contr ibut ing to the break probabi l i ty . Rajan i and Tesfamar iam (2005) show that long-term performance o f bur ied cast i ron is dictated b y pit g rowth rate, unsupported length, fracture toughness and temperature differential. Da ta t yp i ca l l y used i n models are surrogates for factors that can expla in breaks. F o r example , as shown i n Table 4 .1 , the age o f a pipe m a y represent the method o f p ipe manufacture or part icular construct ion standards, as w e l l as deterioration over t ime. B e d d i n g mater ial m a y be an indicator o f a part icular construct ion practice that induces phys ica l stress, o f the structural resistance o f the pipe, or o f the so i l type. F o r example, i n some uti l i t ies where native so i l is used as b a c k f i l l , the so i l m a y not screened for rocks and other objects or proper ly leveled. A s a result, this construct ion practice results i n circumstances where a stress is induced o n pipes and ul t imate ly causes failures. In some cases, fines migra t ion o f corrosive native s o i l through part icular bedd ing types can occur and create the potential for external corrosion. S o i l type can represent cor ros iv i ty and potential for external p ipe corros ion, this is also dependent o n the pipe mater ial . Availability of water main data within utilities for models. The amount o f water m a i n break data needed for extensive m o d e l development is not c o m m o n l y avai lable i n uti l i t ies ( W o o d and Lerice , 2006), i n spi te ,of best practices recommended b y the N a t i o n a l G u i d e to Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R F (Deb et al, 2002). M o s t munic ipa l i t ies o n l y have l imi t ed recorded pipe breakage histories and do not have m u c h data for analysis (Pellet ier et al, 2003). H o w e v e r , i n m a n y instances uti l i t ies m a y have more avai lable data than they realize. T h e y can apply approaches such as constructing and relat ing avai lable data from archives, models and other such sources ( W o o d and Lence , 2006) to construct and l i n k databases for analysis. 127 A k e y to any data management strategy is ident i fy ing the purpose for w h i c h one is co l lec t ing and ana lyz ing the data, whether it is for asset management, c o m p i l i n g an inventory o f assets or d i scover ing the magnitude and nature o f p ipe breaks. There is g r o w i n g interest i n us ing K n o w l e d g e D i s c o v e r y techniques such as data m i n i n g for water m a i n break data (Savic and Wal ters , 1999). K n o w l e d g e D i s c o v e r y is the process o f ident i fy ing v a l i d , nove l , potent ia l ly useful and ul t imate ly understandable patterns i n data (Torra et al, 2004). Such patterns m a y help to identify factors that are related to breaks. 4 . 3 A F R A M E W O R K F O R USING D A T A AND PREDICTION M O D E L S T O I M P R O V E ASSET M A N A G E M E N T The framework developed herein m a y be used to guide a u t i l i ty i n ident i fy ing the magnitude o f its water m a i n break problems today and i n the future, and thereby enhance the development o f strategies for p r io r i t i z ing pipe replacements and data co l lec t ion . Its salient feature is that it integrates break predic t ion or deterioration models that provide an ind ica t ion o f future pipe condit ions w i t h exis t ing data, and thereby uses enhanced estimates o f vu lnerab i l i ty for each pipe. It is designed to accommodate systems w i t h l imi t ed data but is suff iciently f lex ib le to adapt for addi t ional informat ion that m a y be acquired over t ime. Other des ign considerations include ease and transparency o f use and faci l i ta t ion o f a dec i s ion-making process that is repeatable and defensible. Trad i t iona l ly , ut i l i t ies pr ior i t ize pipe replacements based on a combina t ion o f current management practices and his tor ical pipe breakage data. Management practices inc lude directives based on general guidelines, consequence assessments, legislat ive requirements, and other u t i l i ty priori t ies . Rud imenta ry analyses employed interpret h is tor ica l p ipe break data, i nc lud ing locat ion, t ime and date o f break, and pipe diameter and 128 material , and t yp i ca l l y has p rov ided informat ion regarding where and h o w many breaks are occurr ing , and what pipes are exper iencing breaks ( K l e i n e r and Rajani , 1999). Cons ide r ing this informat ion, the pr ior i ty o f the u t i l i ty m a y be to replace water mains o f a certain mater ial or size, those i n a certain area due to previous failures, those under roads that are to be re-paved, those that are current ly undersized, or those that have significant consequences i f failures were to occur, such as mains that serve hospitals. Some uti l i t ies m a y use a mul t ip le objective approach, we igh t ing each o f a number o f pr ior i t i za t ion cri teria, and assigning points to each pipe that describe the degree to w h i c h it meets a g iven cri teria (Deb et al, 2002; Sargeant, 2003). F o r each pipe, the sum o f the product o f the weight and assigned points for each pr ior i t iza t ion cri teria is obtained and used to pr ior i t ize candidate pipes. The framework developed i n this research is s h o w n i n F igure 4 .1 . In order to forecast p ipe breaks, the h is tor ica l data set m a y need to be expanded w i t h data available f rom other sources w i t h i n the u t i l i ty and from external agencies. In addi t ion to h is tor ica l p ipe breakage data, data for factors that m a y be important for predic t ing water m a i n breaks as p rev ious ly described m a y need to be obtained, i nc lud ing so i l type, surface, bedding, and b a c k f i l l material , type o f road usage, or typ ica l f l ow i n area o f break. T h i s informat ion m a y be consol idated b y creating a schematic o f data, w h i c h does not establish a new database per se, but draws from avai lable data for analysis when required, as described b y W o o d and Lence (2006) and W o o d et al, 2007. These data m a y be used direct ly i n the pr ior i t i za t ion process and as input to break predic t ion and deterioration models . The input to the models is developed b y grouping pipes i n w h i c h breaks have occurred based on factors that contribute to breaks. M a t e r i a l and diameter data are avai lable to most uti l i t ies and should be considered as the m i n i m u m 129 factors o n w h i c h to base pipe groups. The dec is ion o f whether to use a phys i ca l - mechanica l or statistical break predic t ion mode l m a y be made at this point , because the p ipe material determines whether a phys i ca l m o d e l exists for a g iven pipe and the diameter influences the pract ica l i ty o f app ly ing such a mode l . F o r sma l l and m e d i u m size uti l i t ies, the pract ical starting point is determinist ic statistical models , w h i c h m a y be developed w i t h read i ly avai lable commerc i a l software, i nc lud ing spreadsheets. M o r e capable uti l i t ies m a y consider more complex statistical or even phys ica l -mechanica l models , however , the l ong term use and maintenance o f these models is a serious considerat ion for those w h o choose these models . T o evaluate the accuracy o f a g iven statistical mode l , a por t ion o f the break data should be used to develop the equations and the most recent por t ion o f the break data should be retained as a holdout sample for compar ison . F o r example , i f a u t i l i ty has twenty years o f break history, it m a y choose to develop models based o n the first fifteen years o f data, and compare the m o d e l predict ions w i t h the remain ing five years o f actual breaks to assess the accuracy o f the predict ive mode l . W h i l e five years is a reasonable length o f holdout sample, this is a function o f the length o f record, and data required to generate the statistical models . In deve lop ing and us ing statistical models , one must determine the amount o f data that are required, the l eve l o f detail to be modeled , and the knowledge that w i l l be gained. In order to determine the length o f record required to develop a credible statistical mode l , the p ipe break record used to mode l the system m a y be var ied to evaluate the sensi t ivi ty o f the m o d e l accuracy to the length o f record used. In order to evaluate the important factors for predic t ing p ipe breaks, the pipe break data m a y be subdiv ided into different sub-groups and models for each o f these sub-groups 130 m a y be developed and compared i n terms o f their relative accuracy. T h i s process natural ly reduces the number o f breaks w i t h i n each sub-group used i n per forming statistical analyses, but m a y y i e l d more credible models . Cons ide r ing data that are t yp i ca l l y avai lable to uti l i t ies ( W o o d and Lence , 2006), potential sub-groups o f pipes for these models , inc lude those o f a specif ied i) p ipe material and diameter, w h i c h indicate pipe strength; i i ) p ipe material , diameter, and age w h i c h indicate pipe strength and age effects such as deterioration and construct ion practices; and i i i ) p ipe material , diameter, s o i l type, and age w h i c h indicate p ipe strength, interaction o f the p ipe mater ial and the so i l , and age effects. Shou ld the u t i l i ty have access to informat ion regarding surface condi t ions, this m a y also be considered i n fo rming the pipe sub-groups. W i t h the knowledge gained from the mode l results, managers can then target pipes that have the highest predicted breaks or rates o f breaks for pr ior i t iza t ion. T h i s informat ion is also useful for ident i fy ing future investigative programs such as s o i l and pipe condi t ion assessments and data acquis i t ion strategies such as changes i n f ie ld co l l ec t ion practices. The u t i l i ty m a y also choose to ver i fy the data or conduct invest igat ive assessments to understand the deterioration o f pipes that have significant breaks but cannot be accurately modeled . F r o m these activit ies, new data can be created to improve the understanding o f p ipe deterioration factors. P ipe network management practices m a y also be altered based o n the m o d e l results. E x a m p l e s o f such changes include ident if icat ion o f new design specifications such as the type o f jo ints required for certain pipes i n a part icular so i l , and in t roduct ion o f corrective measures such as cathodic protection programs. In order to main ta in relevance, it is recommended that models be rout inely rev iewed and updated as part o f the detailed capital p lan o f the u t i l i ty and to account for changes i n the rate at w h i c h breaks are occur r ing as a 131 result o f the changes i n pipe management practices. F i n a l l y , break data should be kept current. 4 . 4 BREAK PREDICTION MODELS FOR LAITY VIEW, MAPLE RIDGE, BC The appl ica t ion o f the framework is demonstrated us ing the L a i t y V i e w area o f M a p l e R i d g e , B C , Canada. T h i s area comprises 13 percent o f the 335 k i lometer dis t r ibut ion system for M a p l e R i d g e , is representative o f the urban area, experienced the same construct ion practices and has so i l types found i n the rest o f the munic ipa l i ty , and is home to a popula t ion o f approximate ly 6,000. The pipe materials found i n the area are asbestos cement, cast i ron , duct i le i r o n and steel, and i n diameters o f 150, 200 and 250 m m . P ipe insta l la t ion records began i n 1959 and few pipes i n M a p l e R i d g e were instal led before this date. The s o i l types found i n the L a i t y V i e w area are c lay , s i l ty-c lay , silt and sand. B reak data are avai lable f rom 1983 to 2004. A total o f 54 breaks occurred i n this per iod, and seven o f these occurred after the year 2000. P re l im ina ry analysis o f these data indicates that breaks are occur r ing i n asbestos cement, cast i r on and duct i le i ron pipes, i n pipes that are greater than 15 years o ld , and i n c l ay and s i l ty -c lay type soi ls ( W o o d et al, 2007). G i v e n the 20-year his tory o f record, the f inal f ive years f rom 2000 to 2004 was selected as the holdout sample. T o investigate the important factors for predic t ing p ipe breaks, pipes i n the area were grouped based o n the four types o f sub-groups p rev ious ly described. Information for surface mater ial w h i c h inc luded asphalt, concrete, and gravel or grass, is avai lable for this reg ion and thus another sub-grouping was examined that inc luded pipes o f a specif ied pipe material , diameter, age and surface material . The 132 combina t ion o f k n o w i n g w h i c h factors are important for predic t ing breaks and the c o m m o n failure types for a ne twork can provide insight on p ipe deterioration behavior. P ipe age sub-groups were created b y examin ing the data and ident i fy ing t ime periods i n w h i c h a meaningful number o f breaks occurred. F o r asbestos cement and cast i r on pipes, these sub-groups were compr ised o f pipes w i t h instal la t ion dates before 1959, between 1960 and 1974, and between 1975 and 1984. Asbestos cement and cast i ron pipes were not instal led i n M a p l e R i d g e after 1984. D u c t i l e i ron pipes were sub-grouped into pipes w i t h instal la t ion dates between 1970 and 1979, 1980 and 1989, 1990 and 1999, and subsequent to 1999. The o n l y steel pipes were instal led i n 1978 and are approximate ly 24 metres i n length. These have not broken. Statist ical determinist ic equations for each group o f L a i t y V i e w pipes were developed for t ime-l inear and t ime-exponent ia l functions. Statist ical determinist ic equations were selected as most appropriate for M a p l e R i d g e because they do not have a sufficient amount o f pipes (e.g. grey cast i ron) and data (such as remain ing p ipe w a l l thickness) to use phys ica l -mechanica l models or the resources to main ta in compl ica ted models ( M a p l e R i d g e has o n l y three engineers on staff and relies on technical support staff for m u c h o f the engineering department duties and responsibi l i t ies) . Statist ical determinist ic models can be easi ly taught to and appl ied b y technical staff and require fewer data. The results should provide insights for future performance and improve M a p l e Ridge ' s current practices o f p r io r i t i z ing water m a i n replacements w h i c h are based o n experience. The t ime-l inear equations for the cumulat ive number o f breaks at year t are based on Equa t ion 1. N(t) = A(t-to) + C (1) 133 Where N( t ) is the cumula t ive number o f breaks for the year t, t 0 is the reference year, i n the case o f L a i t y V i e w , 1983, A is a coefficient and C is a constant. T ime-exponent ia l equations for the cumulat ive number o f breaks at year t are based o n Equa t ion 2. N( t ) = A e k ( t " t 0 ) (2) Where A and k are coefficients and a l l other variables are as described above. A s noted earlier i n this chapter, these equations and their coefficients are specif ic to the L a i t y V i e w area pipes and their respective sub-groups. U t i l i t i e s should develop their o w n equations us ing their system-specific data, selected sub-groups and estimated coefficients (though they m a y choose to also use t ime-l inear and t ime-l inear regression). F o r each sub-group w h i c h had sufficient data, equations were der ived us ing S-Plus® and , spreadsheet software to solve for the coefficients. A m i n i m u m o f two breaks is required i n order to estimate these equations, and thus equations c o u l d not be der ived for a l l sub- groups analyzed. F o r each sub-grouping analysis, the percent o f a l l pipes for w h i c h an equation cou ld be der ived was estimated, as this is an ind ica t ion o f the extent o f the network that m a y be modeled . The accuracy o f the der ived equations, henceforth referred to as models , was calculated as the percent error o f mode l predict ions relative to the cumula t ive breaks i n 2004. F i n a l l y for t ime-l inear models , R - squared estimates are reported. Resu l t s o f b r e a k p r e d i c t i o n mode l s . The accuracy o f the predic t ion results for both t ime-l inear and t ime-exponent ia l models for the various sub-groups are shown i n Figures 4.2 through 4.6. The different amount o f breaks and the rate o f breaks among the 134 various groups suggest that there are differences i n behavior for deterioration and breakage. F o r the mater ial sub-groups, three sub-groups cou ld be modeled ; those for asbestos cement, cast i ron and ducti le i ron pipes and these represent approximate ly one hundred percent o f the pipe length i n the network. A s shown i n F igure 4.2, the t ime-l inear models are more accurate than the t ime-exponent ial models for asbestos cement and duct i le i r on pipes; the percent error for the t ime-l inear models was 29 and 34, and the percent error for the t ime- exponent ial models was 210 and 136, for the asbestos cement and ducti le i r on pipes, respectively. The range o f R-squared statistic for a l l o f the t ime-l inear models was 0.81 to 0.92 and the average R-squared value was 0.84. The results for the cast i r on pipes indicate that w h i l e few breaks have occurred i n these pipes they are occur r ing at an increasing rate. The accuracy o f predict ions for mater ial and diameter sub-groups is shown i n F igure 4.3. Here , seven sub-groups had sufficient number o f breaks to be mode led and these represent 99 percent o f the p ipe length i n the network. A g a i n , w i t h the except ion o f the cast-iron pipes, the t ime-l inear models are more accurate than the t ime-exponent ia l models . W h i l e the most accurate mode l is the t ime-l inear m o d e l for the duct i le i ron pipe w i t h a diameter o f 150 m m , i n general the performance o f t ime-l inear models for the asbestos cement and duct i le i ron pipes is s imi lar . The average R : s q u a r e d statistic for a l l o f the t ime-l inear models was 0.84 and the R-squared statistic was between 0.75 and 0.95. W h e n age was considered, seven sub-groups contained sufficient number o f breaks to be modeled , and these represent 56 percent o f the pipe length i n the network. A s shown i n F igure 4.4, the accuracy o f the t ime-l inear models for ducti le i ron pipes improved dramat ical ly , ind ica t ing that age effects are important i n predic t ing break rates for these pipes, and should be investigated. W i t h respect to the asbestos cement pipes, the age delineated sub-groups suggest that different age groups o f 150 m m asbestos cement pipes 135 are behaving differently w i t h respect to breaks, the abi l i ty to accurately predict breaks differ and that the number o f breaks for pipes instal led between 1975 and 1984 are increasing. The R-squared statistic for a l l o f the t ime-l inear models was between 0.75 and 0.94 and the average R-squared statistic was 0.84. W h e n p ipe - so i l interactions were considered, eight sub-groups cou ld be modeled , but this represented o n l y 3 8 % o f the pipe length i n the network. The accuracy o f predict ions for the mater ial , diameter, so i l and age sub-groups is shown i n F igure 4.5. These results suggest that the accuracy o f predict ions for pipes o f the same material differs i n different soi ls , even w h e n they are instal led at the same t ime. W h e n c l ay is considered as a factor i n the analysis , the accuracy o f the t ime-l inear models stayed the same or improved relative to analyses that considered o n l y material , diameter and age. The average R-squared statistic for a l l o f the t ime-l inear models for these sub-groupings was 0.78 and ranged between 0.63 and 0.94. ' The accuracy o f the models for material , diameter, age, and surface mater ial sub- groups is shown i n F igure 4.6. Here , eight sub-groups cou ld be mode led but these represent o n l y 40 percent o f the pipe length i n the network. W h i l e the average R-squared statistic for a l l t ime-l inear models for this case was 0.84 and ranged from 0.72 to 0.95, the accuracy o f these models is no better than the accuracy o f the t ime-l inear models for the material , diameter, and age sub-groups alone. Th i s indicates that, i n contrast to so i l type, surface mater ia l m a y not be an important factor to consider i n predic t ing pipe breaks for M a p l e R i d g e . O b s e r v a t i o n s . It is important to note that some o f the data that were created as part o f an earlier water m a i n break database constructed (see Chapter 3) contributed to increasing the accuracy o f the applications o f the break predic t ion models . In particular, the 136 • s o i l data p rov ided insights for engineering staff b y indica t ing that s o i l characteristics m a y be an influence on the rate o f cor ros ion i n certain pipes or bedding and b a c k f i l l used i n ins ta l l ing the pipes and thus conf i rmed the value o f creating, relat ing and processing data. A s a result, though the effort was significant, the data creat ion and m i n i n g p rov ided insights and is useful for management decisions and asset management. Wi thou t created data, the analysis and use o f predic t ion models w o u l d be l imi ted . F o r M a p l e R i d g e , the p ipe groups associated w i t h models that accurately predict h igh break rates are: 250 mi l l ime te r diameter asbestos cement pipes instal led i n c lay so i l between 1960 and 1969 and 150 mi l l ime te r diameter asbestos cement pipes instal led i n c lay s o i l between 1960 and 1969. A s a result o f these analyses, considerat ion o f the break rates o f var ious pipe groups and discussions w i t h operations and maintenance staff,, asbestos cement pipes w i l l be p r io r i t i zed for replacement (along w i t h cast i ron pipes w h e n opportunities arise) and further attention w i l l be g iven to co l l ec t ing data o n duct i le i ron pipes. M o r e important ly, because so i l type was identif ied as an important factor i n m o d e l i n g breaks, a so i l s ampl ing program was undertaken to improve the u t i l i ty ' s informat ion regarding s o i l resist ivi ty, p H , chlorides and s o i l type. A p re l imina ry p ipe sampl ing program was implemented at the same t ime to col lect informat ion on asbestos cement pipes and duct i le i ron pipes i n the area. A s a result o f the sampl ing programs, the importance o f bedding and back f i l l i ng practices and construct ion inspections was identif ied and changes i n construct ion specifications and inspect ion practices are be ing developed. Plans are underway to apply this f ramework to the rest o f the M a p l e R i d g e network. U l t ima te ly , schedul ing o f the p ipe replacements and budget estimates w i l l be undertaken i n conjunct ion w i t h other management considerations such as road rehabil i tat ion. 137 A n ongoing p rob lem for u t i l i ty managers is the a l locat ion o f scarce resources for both data co l lec t ion and analysis. One approach for determining the value o f the framework and a support ing data co l lec t ion program is to evaluate the value o f the addi t ional informat ion obtained. The value o f addi t ional informat ion m a y be estimated b y compar ing the dec is ion that w o u l d be undertaken wi thout the addi t ional informat ion w i t h the dec i s ion that w o u l d be undertaken w i t h the addi t ional informat ion (Schuyler , 2001) . F o r example , for L a i t y V i e w , the value o f deve lop ing statistical models that incorporate so i l data m a y be determined b y compar ing the cost o f rep lac ing the group o f pipes that has the highest break rates based o n data for material , diameter and age (i.e., 250 m m diameter asbestos cement pipes) w i t h that o f a replacement strategy that considers replac ing o n l y those 250 m m diameter asbestos cement pipes i n c l ay soils . I f a l l 250 m m diameter asbestos cement pipes, w i t h a total length o f 646 meters were to be replaced, the total cost o f replacement w o u l d be $193,800 (assuming a replacement cost o f $300 per meter). B y app ly ing the f ramework it was determined that a l l the breaks i n these pipes occurred i n c l ay so i l . I f it is assumed that a l l future breaks o f this pipe type w i l l occur i n c l ay soi ls , replacement o f these pipes, w i t h a total length o f o n l y 258 meters w o u l d cost $77,400. Thus the value o f the analysis and the so i l informat ion is approximate ly $193,800 - $77,400 = $116,400. W h i l e this addi t ional informat ion m a y not a lways lead to savings i n terms o f reducing the cost o f p ipe replacement, for example i n cases where the a l l pipes o f a certain material , diameter and age were instal led i n the same so i l type, the informat ion regarding so i l type m a y s t i l l be o f benefit. T h i s informat ion c o u l d be used to improve the instal lat ion practices or jus t i fy corrective measures. 138 4 . 5 CONCLUSIONS Predic t ive m o d e l i n g is useful for ident i fy ing replacement needs over t ime. H o w e v e r , ut i l i t ies do not c o m m o n l y use predict ive m o d e l i n g as part o f their asset, management practices. There are no c o m m o n databases for break analysis or c o m m o n condi t ion indices , and few uti l i t ies undertake cond i t ion assessment ( G r i g g , 2004), a l l o f w h i c h hinders indust ry-wide use o f predict ive mode l ing . The framework presented i n this paper improves upon the tradit ional pipe pr ior i t iza t ion approaches that o n l y l ook at the past his tory i n aggregate and do not necessari ly take into account trends and t i m i n g o f future breaks. A n advantage o f the framework is that it can be appl ied b y smal l to m e d i u m size uti l i t ies w i t h l imi ted informat ion and c o m m o n l y used analyt ical tools. F o r example , for M a p l e R i d g e , w h i l e useful informat ion was gained b y invest igat ing so i l type, reasonable insights m a y have been d rawn f rom analyses that considered o n l y material , diameter and age, informat ion that is t yp i ca l ly avai lable to most ut i l i t ies. Because factors that cause breaks va ry among ut i l i t ies , ut i l i t ies m a y f ind i n d i v i d u a l l y that creating more data (such as b y co l lec t ing traffic loading , b a c k f i l l and pressure data) and relat ing data for m i n i n g and analysis is wor thy o f the effort and expands the use o f this framework. W i t h this i n m i n d , the f ramework is f lex ib le and a l lows for considerat ion o f any avai lable data. In addi t ion to gu id ing water m a i n replacements, the f ramework m a y also be used to identify the k e y data for predic t ing water m a i n breaks. Because there is va r iab i l i ty i n the causes o f pipe breaks among different ut i l i t ies, i n order to understand the performance o f their system, 'uti l i t ies should col lect data as' ident i f ied i n recommended Best Practices; see N a t i o n a l G u i d e to Sustainable M u n i c i p a l Infrastructure ( N G S M I , 2002) and A W W A R P (Deb et al, 2002). A d d i t i o n a l informat ion m a y often be obtained eff iciently at the t ime o f the break repair b y rev i s ing forms to col lect 139 more informat ion, such as bedding or b a c k f i l l material ( W o o d and Lence , 2006) . T ra in ing w i l l often be required, and it is prudent to ver i fy data. C o n v i n c i n g staff to col lect data m a y be an obstacle, but i n v o l v i n g them i n dec is ion m a k i n g can be a w a y to gain support. B y us ing models to predict future breaks, r ev i ewing the accuracy o f the predict ions and updating the models , a u t i l i ty can improve its asset management practices. 4.6 A C K N O W L E D G E M E N T S The authors gratefully acknowledge the Dis t r i c t o f M a p l e R i d g e for p rov id ing data arid support i n the fo rm o f employee resources, and Professors A . D . R u s s e l l , J . W . Atwa te r at the U n i v e r s i t y o f B r i t i s h C o l u m b i a ( U B C ) for their insights and suggestions. M r . W . L i u assisted w i t h the preparation o f break data and M r . A . M a l y u k o f the Dis t r i c t o f M a p l e R i d g e assisted i n data processing and compi l a t i on . ' 140 4 . 7 R E F E R E N C E S A W W A and E E S , Inc., 2002. N e w or repaired water mains . A v a i l a b l e on- l ine at http:/ /www.epa.gov/safewater/tcr/pdf/maincontam.pdf. , A c c e s s e d February 12, 2006 Cooper , N . R . , B l a k e y , G . , She rwin , C , T a , T. , Whi te r , J . T . , and W o o d w a r d , C . A . , 2000. The use o f G I S to develop a probabi l i ty-based trunk mains burst r isk m o d e l . Urban Water, 2 :2000 :97 -103 . D e b , A . K . , Hansi t , Y . J . and Grabul tz , F . M . , 1998. Quant i fy ing Future Rehabi l i t a t ion and Replacement Needs o f Watermains . A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. D e b , A . R . , Grablu tz , F . M . , Hasi t , Y . J . , Synder, J . K . , Longanathan, G . V . and A g b e n o w s k i , N . , 2002. P r i o r i t i z i n g Water m a i n Replacement and Rehabi l i ta t ion . 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235, A W W A Research Founda t ion : 200. D i n g u s , M . , H a v e n , J . , and R u s s e l l , A . , 2002. Nondestructuve, N o n i n v a s i v e Assessment o f Underground Pipe l ines . AWWARFReport 90873, A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O 80235. Do leac , M . L . , L a c k e y , S. L . , and Brat ton, G . N . , 1980. P red ic t ion o f t ime-to-failure for bur ied cast i r on pipe. Proceedings of A WW A Annual Conference, Denver , C O . 141 G r i g g , N . S., 2004. Assessment and R e n e w a l o f Water Di s t r ibu t ion Systems. 6666 Wes t Q u i n c y A v e n u e , Denver , C O , A W W A Research Foundat ion . Jacobs, P . , and K a r n e y , B . , 1994. G I S development w i t h appl ica t ion to cast i ron water m a i n breakage rates. 2nd International Conference on Water Pipeline Systems, Ed inburgh , Scot land, 53-62. Ket t ler , A . J . and Goul ter , I. C , 1985. A n analysis o f pipe breakage i n urban water dis t r ibut ion networks. Canadian Journal of Civil Engineering, 12:2: 286-293. K l e i n e r , Y . , and Rajan i , B . B . , 2001 . Comprehens ive rev iew o f structural deterioration o f water mains : statistical models . Urban Water, 3:3: 131-150. K l e i n e r , Y . , and Rajani , B . B . , 1999. U s i n g l imi t ed data to assess future needs. Journal AWWA, 91:7: 47-62. L e v e l t o n Consultants L t d . , 2005. M a p l e R i d g e Water W o r k s C o r r o s i o n Investigation. Report 2705-0380, L e v e l t o n Consultants L t d . 102-19292 60th A v e n u e , Surrey, B C . M a l e , J . W . , W a l s k i , T . , and Slu tsky, A . F L , 1990. A n a l y z i n g Water M a i n Replacement Po l i c i e s . Journal of Water Resources Planning and Management, A S C E , 116:3: 362-374. 142 M a r k s , D . H . , A n d r e o u , S., Jeffrey, L . , Park , C , and Zas l avsky , A . , 1987. Statist ical M o d e l s for Water M a i n Fai lures . EPA/600/5-87/003, Wate r Eng inee r ing Research L a b , U S E P A . C inc inna t i , O H . N G S M I , 2002. Deter iora t ion and Inspection o f Water D i s t r ibu t ion Systems. Infraguide - Potable Water . Ot tawa, Canada, N a t i o n a l G u i d e to Sustainable M u n i c i p a l Infrastructure. O ' D a y , D . K . , 1982. O r g a n i z i n g and ana lyz ing leak and break data for m a k i n g m a i n replacement decisions. Journal A WWA, 74:11: 588-594. Pellet ier , G . , M a i l h o t , A . , and V i l l e n e u v e , J . -P. , 2003. M o d e l i n g water p ipe breaks - three case studies. Journal of'Water Resources Planning and Management, A S C E , 129:2: 115- 123. Ra jan i , B . B . , and M a k a r J . , 2000. A methodology to estimate remain ing service life o f grey cast i ron water mains . Canadian Journal of Civil Engineering, 27: 1259-1272. Rajan i , B . B . , and K l i e n e r , Y . 2001 . Comprehens ive rev iew o f structural deterioration o f water mains : p h y s i c a l l y based models . Urban Water, 3:3: 151-164. Rajan i , B . B . , and Tesfamar iam, S., 2005. Es t imat ing t ime to failure o f ageing cast i ron water mains under uncertainties." Water Management for the 21st Century, U n i v e r s i t y o f Exeter , U K . , 1-7. 143 Sargeant, D . , 2003. Water M a i n Rehabi l i t a t ion Pr ior i t i za t ion . AWWA 2003 Seminar - Infrastructure: above and below ground, A n a h e i m , C A . Sav ic , D . A . , and Wal ters , G . A . , 1999. Hydro informat ics , D a t a M i n i n g and Main tenance o f U K Water Ne tworks . Anti-Corrosion Methods and Materials, 46:6: 415-425. Schuyler , J . , 2001 . Risk and Decision Analysis in Projects. Project Management Institute, F o u r Campus B l v d . N e w t o w n Square, Pennsy lvan ia . Shamir , U . and H o w a r d , C . D . D . , 1979. A n analytic approach to schedul ing pipe replacement. Journal AWWA, 71:5: 248-258 Torra , V . , Domingo-Fer re r , J . , and Torres, A . , 2004. D a t a m i n i n g methods for l i n k i n g data c o m i n g f rom several sources. 3rd Joint Un/ECE-Eurostat Work Session on Statistical Data Confidentiality, Monographs in Official Statistics, L u x e m b o u r g , Eurostat, 143-150 U S E P A , 2001 . D r i n k i n g Water Infrastructure Needs Survey. Second Repor t to Congress. EPA 816-R-01-004, U . S . Env i ronmenta l Protect ion A g e n c y Off ice o f Water , Washington , D C . W a l s k i , T . M . , 1982. E c o n o m i c A n a l y s i s o f Water M a i n Breaks . Journal of Water Resources Planning Management Division, A S C E , 108:3: 296-308. •J 144 W o o d , A . and Lence , B . J . , 2006. Assessment o f Water M a i n Break D a t a for Asse t Management . Journal AWWA, 98:07. W o o d , A . , Lence , B . J . and L i u , W . , 2007. Cons t ruc t ing Water M a i n Break Da ta for Asse t Management . Journal AWWA. 99:01. 145 Table 4.1 Typical data used in models and factors for which they are a surrogate Surrogate Factor Age Method of pipe manufacture, construction standards, deterioration over time Pipe material Construction practice, method of manufacture, failure mechanisms and causes, joint failures Pipe diameter Wall thickness and resistance to beam loading, pipe use, method of pipe manufacture, construction standards Type of pipe lining Method of pipe manufacture, resistance to corrosion Bedding and backfill material Physical stress on pipes caused by construction practices, structural resistance, soil type, fines migration Pipe protection (wrapped/anodes) Structural resistance, life expectancy, construction practice, method of pipe manufacture Pipe condition Remaining life Soil type Soil corrosivity, physical loading on the pipe such as swelling and frost, level of pipe protection, ground water effects such as draining ability or corrosion, construction practice, bedding and/or backfill material Under a boulevard or roadway Physical loading from surface loads such as traffic, road salt effects Depth of cover Physical loading on the pipe from the weight of soil Surface material/type Physical loading from surface use Normal operating pressure Internal pressure on pipe structure Typical flow in area of break Physical impact from factors such as accelerated internal corrosion from low flow mains Traffic classification Physical loading from surface loads such as traffic volumes and wheel loads Road/surface usage Physical loading from surface loads 146 Figure 4.1 Improving asset management using pipe break prediction models Management considerations Management strategies Master growth plans, construction of other infrastructure, risk assessments, legislative requirements, etc. "'Prioritization of pipe replaccmenl'for asset ^Ifl^wiagernent. Inputs'indude^he:,' ^consideration of managemehfrhistbiical and future predictions • [Historical data \ Stratify pipes into groups based on factors Data construction, linking and updating of data • Investigative programs '.- - • Verification program of constructed data • Pipe network practices assessments Predictions of , future breaks Apply models, compare results with holdout samples 'and evaluate . model performance Target pipes with critical or highest break rates as part of asset management replacement strategy 147 Figure 4.2 Degree of accuracy of time-linear and time-exponential predictions for material groups 250% 200% 150% 100% 50% 0% -50% -100% Asbestos cement pipes Cast iron pipes • Time-linear • Time-exponential Ductile iron pipes Note In 2004, there were a total of 32 breaks in asbestos cement pipes, 5 in cast iron pipes and 17 in ductile iron pipes. 148 Figure 4.3 Degree of accuracy of time-linear and time-exponential predictions for material and diameter groups 300% Notes AC denotes asbestos cement pipes, CI denotes cast iron pipes and DI denotes ductile iron pipes. Pipes are grouped by material and diameter (in millimeters). For example, AC 150pipes are asbestos cement pipes of 150 millimeters diameter. In 2004, there were 24 breaks in AC 150pipes, 5 in AC 200, 3 in AC 250, 4 in CI 150, 9 in DI 150 and 2 in DI 250 pipes. 149 Figure 4.4 Degree of accuracy of time-linear and time-exponential predictions for material, diameter and age groups 250% 200% 150% 100% 50% 0% -50% • Time-linear • Time-exponential AC 150 AC 150 AC 150- AC 200 AC 250 DI 150 1980- DI 200 1980- 1960-1969 1970-1974 1975-1984 1970-1974 1960-1969 1989 1989 Notes AC denotes asbestos cement pipes, CI denotes cast iron pipes and DI denotes ductile iron pipes. Pipes are grouped by material, diameter (in millimeters) and age. For example, AC 150 1960-1969 pipes are asbestos cement pipes of 150 millimeters diameter installed between 1960 and 1969. In 2004, there were 12 breaks in AC 150 1960-1969pipes, 7 in AC 150 1970-1974, 5 in AC 150 1975-1984, 3 in AC 200 1970-1974, 3 in AC 250 1960-1969, 4 in DI 150 1980-1989 and 3 in DI 200 1980-1989 pipes. 150 Figure 4.5 Degree of accuracy of time-linear and time-exponential predictions for material, diameter, soil and age groups 400% 350% 300% 250% 200% 150% 100% 50% 0% • Time-linear • Time-exponential AC 150 AC 150 AC 150 A C 250 Dl 150 A C 150 AC 200 DI200 Clay Clay Clay Clay Clay Silt 1960- Silt 1970- Silt 1980- 1960- 1970- 1975- 1960- 1980- 1969 1974 1989 1969 1974 1984 1969 1989 Notes AC denotes asbestos cement pipes, CI denotes cast iron pipes and DI denotes ductile iron pipes. Pipes are grouped by material, diameter (in millimeters), soil type and age. For example, AC 150 Clay 1960-1969 pipes are asbestos cement pipes of 150 millimeters diameter installed in clay soil between I960 and 1969. In 2004, there were 7 breaks in AC 150 Clay 1960-1969pipes, 7 in AC 150 Clay 1970- 1974, 2 in AC 150 Clay 1975-1984, 2 in AC 250 Clay 1960-1969, 4 in DI 150 Clay 1980- 1989, 3 in AC 150 Silt 1960-1969, 3 in AC 200 Silt 1970-1974 and 2 in DI200 Silt 1980- 1989 pipes. 151 Figure 4.6 Degree of accuracy of time-linear and time-exponential predictions for material, diameter, age and surface material groups 200% 150% 100% 50% 0% -50% • Time-linear • Time-exponential *-' v." v.- ^ ^ ^ ^ ^2 ^ %, 'S>± '<°> <9> ^ *3 \ \ V V V V \ \ \ \ \ * \ \ \ % \ % \ c?o °;o Notes AC denotes asbestos cement pipes, CI denotes cast iron pipes and DI denotes ductile iron pipes. Pipes are grouped by material, diameter (in millimeters), age and surface material. For example, AC 150 1970 - 1974 Asphalt pipes are asbestos cement pipes of 150 millimeters diameter installed between 1970 and 1974 under an asphalt surface. In 2004, there were 5 breaks in AC 150 1970-1974 Asphalt pipes, 7 in AC 150 1960-1969 Gravel/grass, 5 in AC 150 1970-1974 Gravel/grass, 3 in AC 150 1975-1984 Gravel/grass, 2 in AC 200 1 970-1974 Concrete, 2 in AC 250 1960-1969 Gravel/grass, 4 in DI 150 1980- 1989 Gravel/grass and 2 in DI 200 1980-1989 Asphalt pipes. 152 C H A P T E R 5 CONCLUSIONS AND R E C O M M E N D A T I O N S 153 5.1 S U M M A R Y O F R E S E A R C H G O A L S Asset management o f water systems involves assessing w h e n to replace aging and deteriorating pipes. The goal o f this research is to assist sma l l to m e d i u m size uti l i t ies w i t h ident i fying, co l lec t ing and constructing relevant pipe break data to analyze their pipe network, and us ing break predict ions to in fo rm their water m a i n replacement strategy and guide their data acquis i t ion programs. Firs t , the data that are col lec ted and avai lable for analysis across ut i l i t ies i n N o r t h A m e r i c a are identif ied. N e x t , a methodology to create and l i n k data obtained f rom various data sources that can be used i n uti l i t ies o f a l l sizes to construct databases is developed. F i n a l l y , a f ramework to assist the pr ior i t iza t ion o f water m a i n replacements and data acquis i t ion based o n predic t ing future water m a i n breaks w i t h i n a g iven water dis t r ibut ion system is presented. The framework is appl icable for the range o f data avai lable i n typ ica l water ut i l i t ies , acknowledges exis t ing industry needs and practices and m a y help managers to acquire and use avai lable data. M u c h o f the focus o f water m a i n break research has been on the data r i c h and technica l ly sophisticated larger uti l i t ies. H o w e v e r , sma l l and m e d i u m size uti l i t ies need the techniques developed i n this research because they have scarce resources for asset management w i t h i n their organizations. Compared w i t h larger ut i l i t ies, they do not have the staffing expertise or the capaci ty for t ra ining, moni to r ing or deve lop ing back-up systems (Ontario P I R , 2005). W i t h i n these uti l i t ies, there is lit t le or no rel iable documentat ion regarding the locat ion, capacity, condi t ion and adequacy o f p ipe network elements for meet ing present or future needs ( M y e r s , 2001). These uti l i t ies m a y also lack the f inancial and organizat ional resources to implement a complex asset management p rogram or lack the h is tor ica l data or tools to fu l ly analyze their system. T h e y need re la t ive ly inexpensive techniques. T h i s research provides adaptable approaches for eff ic ient ly acquir ing data 154 regarding water m a i n breaks, c o m p i l i n g , ana lyz ing and us ing the data to predict future water m a i n breaks and i m p r o v i n g pr ior i t iza t ion o f pipe replacements. Se lec t ing the case s tudy . The research i n this thesis uses real data from the Dis t r i c t o f M a p l e R i d g e , B C . M a p l e R i d g e was selected because it is a m e d i u m size water u t i l i ty , is s imi la r to munic ipa l i t ies that have undergone urbanizat ion over the past decades and possesses data that were made available to me. M o r e o v e r , M a p l e R i d g e was interested i n us ing break predic t ion models to improve its asset management practices and w h i l e it l acked a comprehensive water m a i n break database, it was receptive to deve lop ing databases for analysis. It also needed a strategy to manage and main ta in the data after they were used to predict water m a i n breaks. A p p e n d i x C is a summary descr ipt ion o f the M a p l e R i d g e water system. 5.2 C O N C L U S I O N S Water m a i n break data co l lec t ion is evo lv ing and industry practices do not match best practices recommended b y N G S M I (2002) and D e b et al (2002) at this t ime. Ut i l i t i e s need a strategy for data qual i ty improvement that w i l l help them deal w i t h challenges such as d i f f icu l ty i n m o b i l i z i n g f inancia l and human resources, absence o f h is tor ica l data, l ack o f knowledge o f current organizat ional practices, l o w re l iab i l i ty o f p rev ious ly col lected data, d i f f icu l ty i n p r io r i t i z ing data co l lec t ion , and the need to develop effective data storage programs. In general, ut i l i t ies can be classif ied as those possessing expanded, intermediate, l imi t ed or m i n i m a l data. W h i l e both phys ica l and statistical models have been developed for predic t ing pipe deterioration and for deve lop ing water m a i n rehabil i ta t ion plans, it is evident that the choice and appl icat ion o f these models are l imi t ed b y the data that uti l i t ies have regarding water m a i n breaks (Rajani and K l e i n e r , 2001 ; K l e i n e r and Rajan i , 2001). 155 F o r practit ioners and researchers a l ike , characterizat ion o f these data classes m a y be used to in fo rm the development o f new asset management techniques that are tai lored to the data l imita t ions that ut i l i t ies face. U t i l i t i e s m a y also improve their data co l lec t ion b y m o d i f y i n g their current practices and more s ignif icant ly , they can seek out alternative data sources f rom w h i c h to analyze breaks and u l t imate ly predict future breaks. The alternate sources that are identif ied i n this research can y i e l d informat ion for researchers and managers al ike. The process o f creating the data schematic as developed i n this research is useful for l i n k i n g mul t ip le databases i n order to compi l e and analyze data. It expands the data avai lable for analysis for both present and future applicat ions and a l lows decentral ized data input and management. M o r e important ly, it is a f lexible approach that a l l ut i l i t ies can employ wi thout significant resource commitments . It reduces problems related to data ownership among units w i t h i n organizations regarding data co l lec t ion , management and disseminat ion o f infrastructure information. The framework developed i n this research for l i n k i n g data is f lex ib le , anticipates the evolu t ion o f data co l lec t ion , bu i ld ing , ver i f ica t ion and storage and a l lows for a var iety o f users. It does not abruptly disrupt data co l lec t ion and warehousing practices and it a l lows managers to continue to expand data co l lec t ion because databases are decentral ized. It is f lex ib le and can be easi ly adapted to a l l ut i l i t ies, whether they are smal l , m e d i u m or large, and regardless o f the uniqueness o f the data col lected and organizat ional framework. Th i s technique can also incorporate tacit data. The importance o f capturing tacit data w i l l increase over the next decades as baby boomer staff retire. T h i s thesis develops an approach for us ing break predic t ion models for ident i fy ing replacement needs over t ime and uses improvement i n m o d e l accuracy as means o f 156 ident i fy ing the k e y data for predic t ing future water m a i n breaks and in fo rming future data acquis i t ion strategies. Trad i t iona l approaches for p r io r i t i z ing p ipe replacements do not incorporate break predict ions. The framework developed i n this research a l lows for the construction, assessment and use o f any available data b y any size ut i l i ty . 5.3 O B S E R V A T I O N S G e n e r a l obse rva t ions . A number o f observations arise from this research. W h i l e this research p r i m a r i l y focuses on engineering science, a number o f the observations relate to management science and the relationships organizations and people have w i t h data u t i l iza t ion . W h i l e these topics arise from the research, they are not addressed i n this thesis and future research i n these areas w i l l be valuable. 1. A s noted i n the thesis, data are col lected throughout an organizat ion b y var ious departments and staff for various purposes. F o r data to be used corporately, executives must acknowledge and address data ownership among departments and managers. 2. Because asset management is a corporate responsibi l i ty , proprietary issues and organizat ional compartmental izat ion can pose major challenges to implement ing an asset management program. Corporate objectives o f knowledge management (e.g., database development and maintenance) should be established and most important ly, be accepted b y those responsible for co l lec t ing , managing and analysing the data. These objectives should be established and promoted b y u t i l i ty executives throughout the u t i l i ty because i n some organizations knowledge m a y be v i e w e d as power and data m a y be interpreted as a surrogate for knowledge . 157 3. A n important task for managers is to establish and encourage a culture o f knowledge sharing (Conne l ly , 2000). M o t i v a t i n g and coaching staff to share knowledge can be a major challenge. A s w e l l , managers themselves are not immune to the habit o f hoarding informat ion and knowledge , though this tendency m a y be reflective o f the corporate culture and the individual ' s relationship w i t h their subordinates, peers and supervisor. 4. A significant need exists i n uti l i t ies for capturing inst i tut ional m e m o r y and data that currently exist and w h i c h w i l l be lost as employees retire. C o m m o n l y , large amounts o f inst i tut ional m e m o r y are not recorded. F o r most ut i l i t ies, there is a need to capture the m e m o r y o f operating departments since they are more often characterized as "act ion oriented" and "hands on" rather than "paper l o v i n g " . Managers must f ind solutions to address this important issue. 5. Water ut i l i t ies also face recrui t ing, t ra ining and capaci ty b u i l d i n g challenges. The v i e w that staff w i l l j o i n an organizat ion early i n their career and spend thirty years i n that organizat ion is becoming extinct. F o r example , Dis t r ic t o f M a p l e R i d g e recruiters consider obta ining five to ten years o f service f rom a non-un ion manager (before he or she leaves the organizat ion) as a valuable experience. C o m p o u n d i n g this, currently, there is significant compet i t ion among B C employers to recruit and retain engineers. U t i l i t i e s also require t raining and capaci ty b u i l d i n g programs for new staff as w e l l as ex is t ing staff. One challenge for m a n y uti l i t ies is that t raining budgets are often the first to be cut i n t imes o f f inancial restraint (because budgets are p r i m a r i l y v i e w e d as support ing expenditures and t ra ining as a discret ionary expense). T h i s was the case i n M a p l e Ridge ' s history. The success o f t ra ining programs is dependent on the market ing and implementa t ion o f the t raining as w e l l as the qua l i ty o f the training. Th i s aspect o f organizat ional development is 158 also important because accord ing to Jacobson and Prusak (2006), organizations w i l l receive greater value f rom informat ion b y deve lop ing strategies and t ra ining staff to help them use what they have rather than b y searching for more data. 6. U t i l i t y managers also need to inspire and motivate posi t ive change. D u r i n g the course o f the research, the author observed that managers f ind that support for change (from their supervisor or employees) is not automatic, and factors that influence the support or change inc lude the career goals o f their supervisor, their l eve l o f influence w i t h i n the organization's pol i t ics , staff motivators and h o w the u t i l i ty is governed. F o r example , changes i n data co l l ec t ion practices because o f legis la t ion tend to be more rap id ly implemented than other reasons for changes ( N S G M I , 2003) . 7. A f inal observation is that organizat ional leadership, structure and behaviour have a major influence o n a uti l i ty 's focus and practices. Based on what I have observed o f var ious ut i l i t ies , the manner i n w h i c h engineering and maintenance departments function and i n w h i c h responsibi l i t ies are distributed w i t h i n a u t i l i ty can lead to dupl ica t ion o f w o r k (because o f ambigui ty or the desire to possess the informat ion or responsibi l i ty) or the incomple t ion o f w o r k (because each department m a y deny responsib i l i ty and assume that the other is addressing the issue). Furthermore, u t i l i ty executives need to provide strategic th ink ing and leadership o f the organizat ion, w h i l e ba lanc ing corporate and i nd iv idua l goals and strengths. Wi thou t this, staff efforts m a y be misp laced or frustrated or m a y flounder. P r o f e s s i o n a l a p p l i c a t i o n s . Th i s thesis has many pract ical applications for smal l and m e d i u m size uti l i t ies. It provides informat ion o n data co l lec t ion and presents an approach that a water u t i l i ty m a y adopt should they decide to use break predic t ion models . G i v e n the w o r k reported i n Chapter 2, ut i l i t ies can assess their practices against those o f their peers and those recommended as best practices. T h e y can ident ify and 159 improve their data co l l ec t ion practices and use the survey results to facilitate discussions w i t h i n their organizat ion regarding the ava i lab i l i ty and storage o f data. T h e y m a y f ind addi t ional data avai lable from other sources such as those ident i f ied i n this thesis. I f there are insufficient data for analysis, they m a y compi l e and l i n k data and construct addi t ional databases to extend the breadth o f data as demonstrated i n Chapter 3. F o l l o w i n g that example , they can construct a data schematic for their organizat ion to analyze their system for asset management i n general and to explore the under ly ing causes o f water m a i n breaks. Once they have constructed and analyzed the data w h i c h is useful despite the effort required, they can then use the framework developed and demonstrated i n Chapter 4 to . incorporate the use o f break predic t ion models to improve asset management and to in fo rm their future data acquis i t ion and storage programs. The approaches developed i n Chapters 3 and 4 address the problems faced b y uti l i t ies, and are designed to be adaptable to their needs and the examples used are based on real data. Researchers w i l l also benefit from the w o r k reported i n Chapters 2 and 3 w h i c h identifies the data that are available for deve lop ing future asset management tools and h o w uti l i t ies can access and construct data for research purposes. L e s s o n s l e a r n e d . W h i l e the research has a number o f applicat ions, a number o f lessons were learned. W h e n app ly ing the research as described i n Chapter 3, a u t i l i ty requires staff and external help i n two areas. F i r s t ly , they require technical assistance i n the f o l l o w i n g tasks: • D e c i d i n g o n the objective o f the exercise. A suggested objective is presented i n the thesis and other purposes are also identif ied. These should be developed b y managers and engineers. 160 • Crea t ing data b y r e v i e w i n g different sources o f data (internal and external to the ut i l i ty) . The thesis presents a number o f techniques such as buffering a co l l ec t ion o f data, in te rv iewing people to capture tacit informat ion, examin ing previous practice standards and conduct ing surveys. Some o f this w o r k can be undertaken b y consultants and b y training exis t ing staff. • D e c i d i n g upon the databases to be related and the nature o f the relationships. Some databases and relationships are suggested i n this thesis, but these decisions should be determined j o i n t l y b y u t i l i ty and data managers. Secondly , the u t i l i ty w i l l require more technical resources to: • R e v i e w the extent o f data ava i lab i l i ty throughout the organizat ion. The thesis suggests sources to explore. • Create data from paper records. . • Es tab l i sh l inks among various databases. • M a p the l inks and various databases. T h i s can be undertaken b y l ine staff and this thesis provides guidance for this task. • Per form some exploratory analyses. It should be noted, that the technical w o r k requires some understanding o f the data and some ab i l i ty to read design drawings. In addi t ion, the project manager app ly ing the research should have strong communica t ion and interpersonal sk i l l s to determine where data m a y be avai lable and to obtain consensus on data sharing among var ious departments. T h e y should be tactful, d ip lomat ic , persuasive, patient and persistent and should not be 161 easi ly discouraged because app ly ing the w o r k w i l l take t ime w i t h i n any ut i l i ty . Sponsorship and support f rom the u t i l i ty executive should be obtained w h i c h w i l l help w h e n w o r k i n g through departmental and corporate issues. F i n a l l y , it is not unrealist ic to expect that a project such as described i n Chapter 3 w o u l d take a 1.2 person years to two person years o f effort and a project such as described i n Chapter 4 m a y take as m u c h as one person year. Validation of the work. W h i l e the p r o o f o f this w o r k lies i n the arguments and demonstrat ion o f the approaches developed, the va l ida t ion o f the contr ibut ion w i l l be the acceptance o f the methods proposed i n practice and whether they are employed b y uti l i t ies to improve asset management and data product iv i ty . In us ing their o w n system-specific data and the thesis as guidance, this research m a y be appl ied and adapted to uti l i t ies o f various sizes possessing a range o f abil i t ies for i m p r o v i n g their asset management practices. U l t ima te va l ida t ion m a y be undertaken b y evaluat ing the success o f such applications. Further research. T o make the appl ica t ion o f this research more useful, further research is needed i n the f o l l o w i n g areas. 1. The survey was developed i n E n g l i s h and forwarded to those communi t ies l is ted i n A p p e n d i x E . O n l y one response was received f rom the A n g l o - Q u e b e c munic ipa l i t ies . Investigation o f French-Canadian water m a i n break data co l l ec t ion practices c o u l d in fo rm these communi t ies and the larger water u t i l i ty industry, al though the data avai lable to Pel le t ier et al. (2003) suggest that French-Canadian practices are no different than those identif ied i n this research. Regardless, translating and dis t r ibut ing the survey throughout Quebec is a potential future project. 2. Tes t ing the f ramework o f us ing water m a i n break predic t ion models i n a u t i l i ty that has more water m a i n breaks, a longer his tory o f breaks or different c l imate w o u l d add 162 further conf i rmat ion o f the appl icabi l i ty o f the research to other ut i l i t ies . The framework has been designed and demonstrated to be robust, though further conf i rmat ion w o u l d be useful. Furthermore, w i t h a larger number o f breaks to analyze, it is possible that other k e y data m a y be identif ied as important for determining the accuracy o f break predic t ion models . 3. A p p l i c a t i o n o f a combina t ion o f statistical and phys ica l -mechanica l models should be explored w i t h i n the framework p rov ided herein. The accuracy o f predict ions for the different types o f models c o u l d also be assessed to g ive further guidance for choos ing models . In addi t ion, c o m b i n i n g the models i n a case study c o u l d prov ide insights that can be used to potent ia l ly demonstrate or improve the robustness and f l ex ib i l i t y o f the f ramework developed herein. 4. F i n a l l y , the topics ident i f ied i n the general observations section should be further explored to assist ut i l i t ies i n app ly ing the research i n practice. These issues include h o w uti l i t ies can increase knowledge sharing (o f data and practices) w i t h i n their organizat ion and ident i fy ing k e y organizat ional structures, leadership attributes and motivators that are needed to improve water u t i l i ty management. W h i l e smal l and m e d i u m size uti l i t ies are able to make operating p o l i c y and practice changes q u i c k l y because dec i s ion m a k i n g typ ica l ly rests w i t h fewer staff than is the case i n larger organizat ions, they are l ike most organizations. T h e y are l im i t ed b y resources i nc lud ing leadership and mot iva t ion sk i l l s and training to execute large changes. Research i n determining and assessing the required sk i l l s , personali ty aptitudes, abi l i t ies , t raining and human resource needs for ins t i tu t ional iz ing the approaches presented i n Chapters 3 and 4 w o u l d be useful i n extending the impact o f this work . 163 Future research interests of the author. Three areas that I am consider ing for further research ar is ing f rom the w o r k i n this thesis inc lude app ly ing addi t ional approaches to ana lyz ing break data, app ly ing the data construct ion approach and framework developed herein to other infrastructure assets, and enhancing u t i l i ty knowledge management practices. 1. A p p l y i n g K n o w l e d g e D i s c o v e r y tools such as ar t i f ic ia l intel l igence to analyse large water m a i n break databases. T h i s has potential to assist those water ut i l i t ies fortunate enough to have a breadth o f data and a l ong his tory o f breaks and to reduce the necessity for h a v i n g m a n y experienced staff to manage infrastructure assets. 2. A p p l y i n g the concept o f l i n k i n g and relat ing data (as shown i n the water m a i n break schematic shown i n Chapter 3) and us ing statistical determinist ic and other predic t ion models to sewerage and drainage systems. Ex tend ing the approach presented i n this thesis to these infrastructure systems has several advantages. These systems are usua l ly o w n e d and managed w i t h i n the same manageria l unit and typ ica l ly have s imi la r replacement values as water systems. Nonetheless , there are differences among the systems. F o r example, sewer systems are easier to inspect us ing v ideo cameras and a c o m m o n condi t ion rat ing system for them exists. T h o u g h drainage systems c o u l d use the same condi t ion rat ing system, they are rarely evaluated and wear out at a faster rate. 3. E x a m i n i n g h o w to manage knowledge o f a system w i t h i n an organizat ion. The data schematic and concept o f relat ing relat ional databases as discussed i n this thesis has potential to p l ay a significant role i n connect ing data for var ious services among the members o f an organizat ion, but research is needed to assist ut i l i t ies i n capturing, managing, main ta in ing and us ing this knowledge . E x a m p l e research questions inc lude: what are the best techniques for in te rv iewing and recording tacit data from ret ir ing 164 employees, what is the op t ima l organizat ional structure for sharing knowledge among engineers and maintenance staff, and what are the most efficient and effective approaches that ut i l i t ies can emp loy to make data accessible across an organizat ion? 5 . 4 C L O S I N G R E M A R K S The approaches developed herein are easi ly appl ied. There are s ignif icant ly more sma l l and m e d i u m size uti l i t ies than larger uti l i t ies (Van ie r and R a h m a n , 2004) and these uti l i t ies can benefit from apply ing the work . The approaches have been demonstrated w i t h real data and m a y prov ide uti l i t ies w i t h insights and guidance for a l locat ing their scarce resources, consider ing their current l imita t ions and future needs. The appl ica t ion is pract ical and can be implemented incremental ly . The thesis chapters are at various stages o f pub l ica t ion or review. Chapter 2 was publ i shed i n the J u l y 2006 issue o f the Journal o f the A W W A . Chapter 3 is publ i shed i n the January 2007 issue o f the Journal o f the A W W A . Chapter 4 was submitted to the A S C E Journal o f Infrastructure Systems i n Ju ly , 2006. A n invi ted presentation on the results o f the first paper was made at the September 2006 A n n u a l P u b l i c W o r k s A s s o c i a t i o n o f B C (a chapter o f the A m e r i c a n P u b l i c W o r k s Assoc ia t ion) Conference i n Q u a l i c u m Beach , B C . In addi t ion, the w o r k w i l l be presented as part o f a water system asset management workshop to engineers i n the Greater V a n c o u v e r area. In c los ing , as a result o f the w o r k conducted i n this thesis, M a p l e R i d g e has implemented a number o f the findings to improve its water m a i n break data co l lec t ion practices. It has undertaken so i l surveys, tested a number o f p ipe samples and ver i f ied some o f the l i nked and related data. It is also developing a u t i l i ty w i d e data schematic that w i l l form the basis o f a data warehouse strategy that w i l l be appl ied to the entire water system 165 and sanitary sewer system. F i n a l l y , the framework for determining the data that contribute to p ipe failures w i l l be implemented beyond the case study area to improve M a p l e Ridge ' s asset management practices. \ 166 R E F E R E N C E S 167 R E F E R E N C E S A h a m m e d , M . and Me lche r s , R . E . , 1994. R e l i a b i l i t y o f underground pipel ines subjected to corrosion. Journal of Transportation Engineering, 120:6: 989-1003. A n d r e o u , S., M a r k s , D . and C la rk , R . , 1987. A new M e t h o d o l o g y for M o d e l i n g Break Fa i lure Patterns i n Deter iorat ing Water Di s t r ibu t ion Systems: Theory . Advances in Water Resources, 10:1:2-10. E l sev ie r Science, B . V . A S C E ( A m e r i c a n Socie ty o f C i v i l Engineers) , 1999. A m e r i c a n Soc ie ty o f C i v i l Engineers report card and issue briefs. Public Works Management and Policy, 4 :1 : 58-76. A W W A ( A m e r i c a n Water W o r k s Assoc ia t ion) , 2004. WATER: \STATS- The Water Utility Database. 2002 version [CD-ROM]. A m e r i c a n Water W o r k s Assoc i a t i on , Denver , C O . 80235. A W W A and E E S , Inc., 2002. New or repaired water mains. A v a i l a b l e on- l ine at http://www.epa.gov/safewater/tcr/pdf/maincontam.pdf. , A c c e s s e d February 12, 2006. Canada. Statistics Canada, 2006. The age of public infrastructure in Canada, b y V a l e r i e Gaudreault and Pat r ick L e m i r e . January 2006. (Catalogue N o . 11 -621-MIE2006035) . Ot tawa, M i n i s t e r o f Industry. 168 C o n n e l l y , C . E . 2000. Predictors o f knowledge sharing i n organizations. Mas te r o f Science Thesis , Queens Un ive r s i t y , K i n g s t o n , Ontar io , Canada. Cooper , N . R . , B l a k e y , G . , She rwin , C , T a , T . , Whi te r , J . T . and W o o d w a r d , C . A . , 2000. The use o f G I S to develop a probabi l i ty-based trunk mains burst r i sk mode l . Urban Water, 2:2000: 97-103. D a v i s , D . N . , 2000. Agent-based dec is ion support f ramework for water supply infrastructure rehabil i ta t ion and development. Computers, Environment and Urban Systems, 24:2000: 173-190. D e b , A . R . , Grablu tz , F . M . , Has i t , Y . J . , Synder, J . K . , Longanathan, G . V . and A g b e n o w s k i , N . , 2002. P r i o r i t i z i n g Water m a i n Replacement and Rehabi l i ta t ion . A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. D e b , A . K . , Hans i t , Y . J . and Grabul tz , F . M . , 1998. Quant i fy ing Future Rehabi l i t a t ion and Replacement Needs o f Watermains . A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. D e S i l v a , D . , D a v i s , P . , B u r n , L . S., Ferguson, P . , M a s s i e , D . , C u l l , J . , E i s w i r t h , M . and Heske , C , 2002. C o n d i t i o n Assessment o f Cast Iron and Asbestos Cement P ipes b y In-pipe Probes and Select ive S a m p l i n g for Es t imat ion o f R e m a i n i n g Service L i f e . Proceedings of the 20th International No-Dig Conference, Copenhagen, Denmark . 169 Dingus , M . , H a v e n , J . and R u s s e l l , A . , 2002. Nondes t ruc t ive , N o n i n v a s i v e Assessment o f Underg round Pipe l ines . A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. Do leac , M . L . , L a c k e y , S. L . and Brat ton, G . N . , 1980. P r e d i c t i o n o f t ime-to-failure for bur ied cast i ron pipe. Proceedings of A WW A Annual Conference, Denver , C O . Ear th Tech , 2004. N a t i o n a l Water and Wastewater B e n c h m a r k i n g Init iat ive F i n a l Resul ts - June 2004. Ear th Tech , Burnaby , B C . Ear th Tech , 2003. N a t i o n a l Water and Wastewater B e n c h m a r k i n g Init iat ive- Water U t i l i t y Def in i t ions 2003 Release B . , Ear th Tech , Burnaby , B C . F C M , 2001 . Backgrounder: Protecting Canada's Quality of Life. Federat ion o f Canad ian M u n i c i p a l i t i e s , (http:/ /www.fcm.ca/infra/throne-a.html). A c c e s s e d December 2002. \ G o l d e r Associa tes L t d . , 2002. Geotechnica l Input to the seismic vulnerabi l i ty assessment for the Dis t r i c t o f M a p l e R i d g e , B . C . 500-4260 S t i l l Creek D r i v e , Burnaby , B C , V 5 C 6 C 6 . G r i g g , N . S . , 2005. Assessment and R e n e w a l o f Water Di s t r ibu t ion Systems. Journal AWWA, 97:2: 58-67. G r i g g , N . S., 2004. Assessment and renewal o f water dis t r ibut ion systems^ A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. 170 Hab ib i an , A . , 1992. D e v e l o p i n g and u t i l i z i n g databases for water m a i n rehabil i tat ion. Journal AWWA, 84:7: 75-79. Jacobs, P . and K a r n e y , B . , 1994. G I S development w i t h appl ica t ion to cast i ron water m a i n breakage rates. 2nd International Conference on Water Pipeline Systems, Ed inburgh , Scot land, B H R Group . 53-62. Jacobson, A . and Prusak, L . , 2006. The cost o f knowledge . Harvard Business Review, 84:11:34 Ket t ler , A . J . and Goul ter , I. C . , 1985. A n analysis o f p ipe breakage i n urban water dis t r ibut ion networks. Canadian Journal of Civil Engineering, 12:2: 286-293. K l e i n e r , Y . and Rajan i , B . B . , 2001 . Comprehens ive rev iew o f structural deterioration o f water mains: statistical models. Urban Water, 3:3: 131-150. K l e i n e r , Y . and Rajan i , B . B . , 1999. U s i n g l imi t ed data to assess future needs. Journal AWWA, 91 :7 :47 -62 . K l e i n e r , Y . , Ra jan i , B . B . and Sadiq , R . , 2004. Management o f failure r isk i n large diameter bur ied pipes us ing fuzzy-based techniques. 4th International Conference on Decision Making in Urban and Civil Engineering, Porto, Por tugal . 171 L e v e l t o n Consultants L t d . , 2005. M a p l e R i d g e Water W o r k s C o r r o s i o n Investigation. Report 2705-0380, L e v e l t o n Consultants L t d . 102-19292 60th A v e n u e , Surrey, B C . Lut tmerd ing , H . A . , 1981. Soils of the Langley-Vancouver map area: Report No. 15, Volume 6. B . C . M i n i s t r y o f Env i ronment Assessment and P l a n n i n g D i v i s i o n , K e l o w n a , B C . Lut tmerd ing , H . A . , 1980. Soils of the Langley-Vancouver map area Report No. 15 Volume 1. B . C . M i n i s t r y o f Env i ronment Assessment and P lann ing D i v i s i o n . K e l o w n a , B C . M a i l h o t , A . , Pel let ier , G . , N o e l , J . -F . and V i l l e n e u v e , J . -P. , 2000. M o d e l i n g the evolu t ion o f the structural state o f water pipe networks w i t h b r i e f recorded pipe break histories: M e t h o d o l o g y and appl icat ion. Water Resources Research, 36:10: 3053-3062. M a l e , J . W . , W a l s k i , T . and Slutsky, A . H . , 1990. A n a l y z i n g Water M a i n Replacement Po l i c i e s . Journal of Water Resources Planning and Management, 116:3: 362-374. M a r k s , D . H . , A n d r e o u , S., Jeffrey, L . , Park, C . and Zas l avsky , A . , 1987. Statist ical M o d e l s for Wate r M a i n Fai lures . Water Eng inee r ing Research L a b , U S E P A . C inc inna t i , O H . M y e r s , C , 2001 . R u r a l Areas gain h igh tech infrastructure p l an through partnership, Utility Executive, 3:2: 5-8, Water Env i ronment Federat ion, 601 W y t h e Street, A l e x a n d r i a , V A , 22314-1994. 172 N G S M I , (Nat iona l G u i d e to Sustainable M u n i c i p a l Infrastructure - Infraguide), 2003. Bes t Practices for U t i l i t y - B a s e d Data . Infraguide - Potable Water . Ot tawa, O N , Canada. N G S M I (Nat iona l G u i d e to Sustainable M u n i c i p a l Infrastructure - Infraguide), 2002. Deter iora t ion and Inspection o f Water Dis t r ibu t ion Systems. Infraguide - Potable Water . Ot tawa, O N , Canada. N a t i o n a l Research C o u n c i l o f the Na t iona l A c a d e m i e s - Commi t t ee o n P u b l i c Water S u p p l y Di s t r ibu t ion Systems: Assess ing and R e d u c i n g R i s k s , 2005 . Public Water Supply Distribution Systems: Assessing and Reducing Risks - First Report. O ' D a y , D . K . , 1986. Wate rma in Eva lua t ion for Rehabi l i t a t ion / Replacement . A W W A Research Founda t ion ( A W W A R F ) and U S Env i ronmen ta l Protect ion A g e n c y ( U S E P A ) , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. O ' D a y , D . K . , 1982. O r g a n i z i n g and ana lyz ing leak and break data for m a k i n g m a i n replacement decisions. Journal A WWA, 74:11: 588-594. O ' D a y , D . K . , W e i s s , R . , C h i a v a r i , S., and B l a i r , D . , 1986. Wa te rma in Eva lua t i on for Rehabi l i t a t ion / Replacement . A W W A Research Founda t ion ( A W W A R F ) and U S Env i ronmen ta l Protect ion A g e n c y ( U S E P A ) , A W W A Research Foundat ion , 6666 Wes t Q u i n c y A v e n u e , Denver , C O . 80235. 173 Ontar io P I R . 2005. WATERTIGHT: The case for change in Ontario's water and wastewater sector. Report o f the Water Strategy Exper t Panel . Ontar io M i n i s t r y o f P u b l i c Infrastructure R e n e w a l . Pel let ier , G . , M a i l h o t , A . and V i l l e n e u v e , J . -P. , 2003. M o d e l i n g water p ipe breaks - three case studies. Journal of Water Resources Planning and Management, 129:2: 115-123. Rajan i , B . B . and K l e i n e r , Y . , 2001 . Comprehens ive rev iew o f structural deterioration o f water mains: p h y s i c a l l y based models . Urban Water, 3:3: 151-164. Rajan i , B . B . and M a k a r , J . , 2000. A methodology to estimate remain ing service l ife o f grey cast i ron water mains . Canadian Journal of Civil Engineering, 27:6: 1259-1272. Ra jan i , B . B . and Tesfamar iam, S., 2005. Es t ima t ing t ime to failure o f ageing cast i ron water mains under uncertainties. Water Management for the 21st Century, U n i v e r s i t y o f Exeter , U K . , U n i v e r s i t y o f Exeter , U K . , 1-7. Rossi ter , D . G . , 2005. A compendium of on-line soil survey information - Soil survey institutes and activities. A c c e s s e d M a y 31 , 2005 h t tp : / /www.i tc .n l / rossiter/research/ rsrch_ss_sOurces.html. Sac lu t i , F . , 1999. M o d e l i n g Water Di s t r ibu t ion P ipe Fai lures U s i n g A r t i f i c i a l N eu r a l Ne t works , Mas t e r o f Science Thesis , U n i v e r s i t y o f Albe r t a , E d m o n t o n , A lbe r t a . 174 Sargeant, D . , 2003. Water M a i n Rehabi l i t a t ion Pr ior i t i za t ion . AWWA 2003 Seminar - Infrastructure: above and below ground, A n a h e i m , C A . Sav ic , D . A . and Wal ters , G . A . , 1999. Hydro informat ics , Da ta M i n i n g and Main tenance o f U K Water Ne tworks . Anti-Corrosion Methods and Materials, 46:6: 415-425. Sch ick , S., 2005. Toronto bridges data needs i n water network project. ITBusiness.ca July 21,2005. A c c e s s e d J u l y 22 , 2005 http:/ /www.itbusiness.ca/index.asp7theaction =61&sd i= 59540. Schuyler , J . , 2001. Risk and Decision Analysis in Projects. Project Management Institute, F o u r C a m p u s B l v d . N e w t o w n Square, Pennsylvania . Shamir , U . and H o w a r d , C , 1979. A n analytic approach to schedul ing p ipe replacement. Journal AWWA, 71:5: 248-258. Tha in , D . , 2005. Water mains i n the Dis t r ic t o f M a p l e R i d g e . A . W o o d , ed., M a p l e R i d g e , Interview notes. Torra , V . , Domingo-Fer re r , J . and Torres, A . , 2004. Da ta m i n i n g methods for l i n k i n g data c o m i n g f rom several sources. 3rd Joint Un/ECE-Eurostat Work Session on Statistical Data Confidentiality, Monographs in Official Statistics, L u x e m b o u r g , Eurostat. 143-150. 175 U S E P A , 2005. 2003 D r i n k i n g Water Infrastructure Needs Survey and Assessment , U n i t e d States Env i ronmen ta l Protect ion A g e n c y , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . U S E P A , 2002. C l e a n Water and D r i n k i n g Water Infrastructure Gap A n a l y s i s , Env i ronmen ta l Protect ion A g e n c y , Off ice o f Water , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . U S E P A , 2001 . D r i n k i n g Water Infrastructure Needs Survey. Second Repor t to Congress. U n i t e d States. Env i ronmen ta l Protect ion A g e n c y , Off ice o f Water , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . U S E P A , 1999. D r i n k i n g Water Infrastructure Needs Survey. U n i t e d States Env i ronmen ta l Protect ion A g e n c y , Of f ice o f Water , D r i n k i n g Water D i v i s i o n , Wash ing ton , D C . U S F H W A , 1999. Asse t Management Pr imer , Federal H i g h w a y s Admin i s t r a t i on , U S Department o f Transportat ion, Off ice o f Asset Management , 400 7th Street, S . W . Wash ing ton D . C . 20590. V a n i e r , D . J . , 2001 . Asse t Management : " A " to " Z " , American Public Works Association Annual Congress and Exposition - Innovations in Urban Infrastructure Seminar, Phi lade lph ia , U . S . September, 2001 . 1-16. 176 V a n i e r , D J . and R a h m a n , S., 2004. M u n i c i p a l Infrastructure Investment P l a n n i n g (MID?) M U P Report : A P r i m e r o n M u n i c i p a l Infrastructure Asse t Management . Report B-5123.3, N a t i o n a l Research C o u n c i l Canada, Ottawa, O N . W a l s k i , T . M . , 1982. E c o n o m i c A n a l y s i s o f Water M a i n Breaks . Journal of Water Resources Planning Management Division, 108:3: 296-308. W i l d , C . J . and Seber, G . A . F . , 2000. Chance Encounters - A First Course in Data Analysis and Inference. John W i l e y and Sons Inc., N e w Y o r k , N . Y . W o o d , A . and Lence , B . J . , 2006. Assessment o f Water M a i n Break D a t a for Asse t Management . Journal AWWA, 98:07. W o o d , A . , Lence , B . J . and L i u , W . , 2007. Cons t ruc t ing Water M a i n Break D a t a for Asse t Management , Journal AWWA, 99:01. X u , C . and Goul ter , I .C . , 1998. Probabi l i s t ic mode l for water dis t r ibut ion re l iabi l i ty . Journal of Water Resources Planning and Management, 1 2 4 : 4 : 218-228. 177 P A G E L E F T B L A N K 178 APPENDICES APPENDIX A 179 A p p e n d i x A summarizes the uti l i t ies surveyed and the data queried b y Deb et al. (2002) and b y the author as part o f the survey reported o n i n Chapter 2. A compar i son o f the survey responses b y service popula t ion is shown i n Table A . l . A compar i son o f the questions asked b y the two surveys is shown i n Table A . 2 . Table A.l Comparison of North American utility responses by service populations Survey <100,000 >100,000 Survey b y D e b et al (2002) 5 32 Survey b y W o o d and L e n c e (2006) 37 22 Table A.2 Comparison of questions posed by Deb et al. (2002) and Wood and Lence (2006) Survey question by Deb et al (2002) Question addressed by Wood and Lence Notes Utility size yes Water production no Total length of pipe yes Main failures Formal program for control of failure n/a Main inventory n/a Computerized main inventory of total n/a Failure records yes Wood and Lence survey focused on failure records Computerized failure records of total n/a General Information yes Date, address yes Temperature yes Wood and Lence also queried air, water, change in water temperature and soil temperature Time of detection and arrival yes includes repair date Impact on surroundings Services affected yes Wood and Lence survey includes type of services and length of outage 180 Blocks affected no Wood and Lence survey includes quantityof parcels and number of customers affected and property damage Hydrants affected no Proximity to buried objects no Water main information Material yes Location in street yes Wood and Lence survey includes surface use Diameter yes Depth yes Cover depth Installation date yes Cathodic protection yes Wood and Lence survey queried pipe protection and /or anode installation Joint type yes Type of repair yes Pressure Range yes Operating pressure Pump station status no Wood and Lence survey queried flow in area Failures Type of failure yes Wood and Lence survey queried on suite of types (11), Debet al.just queried i f type of failure was recorded Probable cause yes Wood and Lence survey queried on suite of causes (13) Type of repair yes Wood and Lence survey queried on suite of repairs (7) and additional treatments Exterior/interior of pipe yes Wood and Lence survey queried on both exterior and interior, lining condition and other details Bell condition no Wood and Lence survey queried on Joint type Condition of valves Required for isolation yes Condition no 181 Reporting bedding conditions yes Wood and Lence survey queried on material and condition separately Reporting of seismic/geotechnical conditions Soil description yes Wood and Lence survey queried on native soil, soil pH, soil Moisture content while Deb at al. identified i f soil description was recorded Geologic unit description no Groundwater depth no Seismic hazard unit no Collecting field samples Pipe samples yes Soil samples yes Use of automated systems Wood and Lence survey explored other sources of data Field portable computers see'note GPS see note GIS see note D B M S see note Formal renewal program Main replacement n/a Main rehabilitation n/a Costs records Direct labour yes Wood and Lence survey queried on crew hours Indirect labour see note Wood and Lence survey queried on all labour costs Materials yes Equipment, yes Surface repairs see note Wood and Lence survey queried on Property damage costs Damage yes- Wood and Lence survey queried on Property damage costs 182 APPENDIX B 183 s Append ix B provides the statistical analysis for guiding the interpretation o f the survey results and their applicabili ty to the general population o f water utilities. G i v e n a specified percentage o f respondents that collect a particular type o f data, one may wish to determine the percentage o f a l l water utilities that are l ike ly to collect the given data. Such questions are addressed using the confidence range o f observations regarding the particular type o f data for utilities i n the general population, and these are based on the standard error o f the sample population. W i l d and Seber (2000) suggest that an accepted measure o f the confidence in the behaviour o f the general population (e.g., a l l water utilities) is equivalent to two standard errors o f the sample proportion (e.g., the respondents). The standard error o f the sample proportion is calculated using the equation: Se(p) = ( p * ( l - p ) / n ) 0 5 ( B l ) Where Se (p) = the standard error o f a sample proportion, p = the proportion o f the sample that collects a given data type or element, n = the sample size, in this case 59 utilities. For example, using Equation B l , i f 63 percent o f the respondents (i.e., 37 responses o f the 59 survey respondents), indicate that they record when water service was restored, we may expect that more than 50 percent (i.e., the lower confidence l imit) but less than 75 percent (i.e., the upper confidence l imit) o f a l l water utilities in the general population wou ld record the same. The lower confidence l imi t o f 50 percent is calculated by the response proportion minus two times the standard error o f the sample population, i.e., 0.50 = 0.63 - 0.13, where 0.13 = 2*Se(p) = 2 (0.63*0.37/59) ° ' 5 . 184 Because the standard error is calculated using the sample proportions, the standard error varies wi th the number o f responses. For example, the higher the number o f responses for ' recording a given type o f data, the more certain we are o f the general population recording that type o f data. The range o f standard error is between 5.5 percent and 6.5 percent for response proportions o f between 33 percent (20 responses) and 76 percent (45 responses) A summary o f values for the standard error for various levels o f responses and the general population confidence l imits corresponding to those responses is shown in Table B . l . Table B.l Confidence limits for the general population0 based on proportion of responses Upper limit of Lower limit of the confidence the confidence level for the level for the Number of Percentage of Standard error of general population (within 2 standard general population (within 2 standard responses respondents response - % errors) errors) 20 34% 6.2% 46% 22% 30 51% 1 6.5% 64% 38% 37 63% 6.3% 75% 50% 39 66% 6.2% 78% 54% 44 75% 5.7% 86% 63% 45 76% . 5.5% 87% 65% a) Based on 59 responses (i.e., the sample size). 185 P A G E L E F T B L A N K 186 APPENDIX C 187 A p p e n d i x C describes the mun ic ipa l i t y o f M a p l e R i d g e , its water u t i l i ty and the L a i t y V i e w area. MAPLE RIDGE, B.C. General description. The Dis t r i c t o f M a p l e R i d g e ( B C ) is a mun ic ipa l i t y located w i t h i n the Greater V a n c o u v e r region. O f the 75,000 people that resided i n the mun ic ipa l i t y i n 2005, approximate ly 65,000 residents were served b y the water u t i l i ty and 12,000 were served b y on-site private we l l s . The mun ic ipa l i t y is i n transit ion from be ing predominant ly rural to be ing a suburban communi ty . It has a t own centre and surrounding urban area. Outs ide o f the urban boundary are lands that are zoned rural and agricul tural . Distribution system. The dis t r ibut ion system o f M a p l e R i d g e has over 15,000 connections and is compr i sed o f 5 pump stations, 7 reservoirs and approximate ly 350,000 metres o f water mains . A summary o f the u t i l i ty ' s pipe inventory b y mater ial types for 2001 is shown i n Table C . l A summary o f the v o l u m e o f water purchased i n 2001 from the Greater V a n c o u v e r Wate r Dis t r ic t ( G V W D ) and distr ibuted da i l y is presented i n Table C . 2 . A compar i son o f the age o f M a p l e R i d g e water mains w i t h a number o f munic ipa l i t i es across Canada that are part o f the Ear th T e c h benchmark ing ini t ia t ive (Earth T e c h 2004) is shown i n Table C . 3 . The table shows the dis t r ibut ion o f pipes b y age cohorts o f M a p l e R i d g e compared w i t h the entire set o f communi t ies i n the study. The participants inc lude R i c h m o n d H i l l ( O N ) , C i t y o f D e l t a ( B C ) , C i t y o f Wate r loo ( O N ) , C i t y o f Ca lga ry ( A B ) , C i t y o f Ot tawa ( O N ) , Reg iona l M u n i c i p a l i t y o f H a l t o n ( O N ) , C i t y o f Saskatoon ( S K ) , C i t y o f L o n d o n ( O N ) , C i t y o f Toronto ( O N ) , C i t y o f H a m i l t o n ( O N ) , C i t y o f Thunder B a y ( O N ) , C i t y o f St. Catherines ( O N ) and C i t y o f V i c t o r i a ( B C ) . 188 F i n a n c i a l measures . F inanc i a l l y , the u t i l i ty operates as a se l f - l iquidat ing u t i l i ty - i.e., annual revenues and expenditures must balance. A five year business p l an is submitted to M a p l e R i d g e C o u n c i l each year and rates are established b y C o u n c i l by l aw . In 2003, residential water customers were charged $230 per household and metered customers were charged $0,395 per cubic metre. A f inancia l summary o f the 2001 major expenditure categories is shown i n Table C .4 . L a i t y V i e w . The L a i t y V i e w area o f M a p l e R i d g e represents approximate ly ten percent o f the water dis t r ibut ion system o f M a p l e R i d g e and serves a popula t ion o f approximate ly 6,000. The total length o f water mains i n this area is 36,300 metres. A regional water m a i n forms the boundary o n one side o f the study