Water main decision support system seizes advantages

Water main decision support system seizes advantages

Ewe Leng Lim, PE
Senior Systems Analyst
Seattle Public Utilities
710 Second Avenue, 9'hFloor
Seattle, WA 98104

Nick Bodnar, PE
Project Director
Roy F. Weston, Inc.
700 Fifth Avenue, Suite 5700
Seattle, WA 98104-5057

Introduction
Seattle Public Utilities (SPU) has the largest water utility in the state of Washington and one of the largest in the nation. The water utility supplies high-quality water to nearly 600,000 people in the metropolitan Seattle area through direct retail service and another 625,000 people through 28 wholesale customers. The area served by the regional supply system covers 320 square miles. The water pipeline system consists of 160 miles of supply mains, 45 miles of feeder mains, and 1,620 miles of distribution mains. The pipeline system ranges in age from newly installed to over a hundred years old. Various different pipe materials were use throughout the years with varying installation quality and under many different environmental conditions.

As the water distribution infrastructure ages, it becomes increasingly more challenging to assign limited capital expenditures to prioritize the repair, replacement, or rehabilitation of pipelines.

There is a growing need by municipalities to find better ways to prioritize their future projects. Much research has gone into the process of water distribution system rehabilitation planning. The American Water Works Association (AWWA), American Society of Civil Engineers (ASCE), and other professional associations have produced numerous publications on the subject. Results and recommendations from these research have been adopted by water utilities in varying degrees. Most water utilities have adopted some form of subjective ranking system to prioritize pipeline rehabilitation. A smaller number of water utilities have completed statistical analyses to predict pipeline failure and incorporate the results within a cost-benefit analysis. SPU water engineers and managers have seized upon the advantages of accessing water main characteristic information through Seattle’s Geographic Information System (GIS) and have significantly reduced the time spent identi~ing and prioritizing water mains for preventive maintenance, rehabilitation, or replacement. This paper describes how SPU developed and now utilizes GIS-based information in day-to-day decision making and capital improvement program preparation with a focus on the water pipeline infrastructure.

Overview
In order to provide easier access to SPU’S GIS database, the GIS group embarked on a widescale installation of customized ArcView projects on user desktops. ArcView is a relatively user-friendly GIS soflware that is available for the desktop users. The customized ArcView projects allow users to more easily add layers of information with default symbology and other descriptive information. Among the customized tools in the ArcView project is a fairly smart address matching capability that would try to match address to the parcel layer first which will result in a fairly accurate location if the exact parcel is matched. If an exact match cannot be found, the program will then try to match the street lines that will result in a “within-the-block” location, depending upon the accuracy of the street network database. The layers of information available to the users include: water pipeline attributes, such as material, diameter, age, etc.; historical pipeline breaks; water pressure zones; major water users; soil type; land use and zoning, such as industrial, commercial, residential, etc.; critical water service connections such as hospitals, dialysis centers, medical facilities, and others.

With this vast amount of valuable information readily available to the users, they can begin to analyze the water network and ask questions such as:

Which pipelines appear to be most vulnerable or susceptible to failure based on their attributes?
Which pipelines are most critical or would cause most damage in the event of a failure?
What is the correlation between historical pipe breaks and a pipe’s attributes or its relative location to its surrounding conditions, such as soil type or proximity to other features?
How can we best prioritize which pipelines to replace based on the answers to the above questions?

To assist the users in making these and other analyses, a custom decision-support application was developed to leverage the GIS water main characteristic data with statistically based decision models to improve decision making. Correlation of water main characteristics with time-to- failure factors identified statistically significant failure prediction indicators. Based on the identified indicators, decision models were developed for evaluating deterioration, vulnerability, and service criticality factors. Then, based on evaluation results, main segments can be prioritized on the probability of their failure during a given time period. Ratings maybe output to tabular or map reports for further discussion or investigation.

To make replace-versus-repair decisions, SPU staff may apply standard costing data to compare, for each main segment, the present worth of probable future repair costs against current rehabilitation or replacement costs.

Armlication Models
The custom-developed decision-support application or Pipe Evaluation System (PIPES) consists of three main models:

The Deterioration Model is derived from a statistical analysis of the pipelines’ break history and how these data correlates with direct or surrogate parameters associated with each pipeline. Examples of significant parameters that correlate to breaks include the length and diameter of pipe, static water pressure, pipe material, age of pipe, and whether a pipe is located on a steep slope. As a result of the statistical analysis, a number of equations can be generated to calculate the probability of pipe failure. These equations are built into the PIPES application and will be accessed when the deterioration model is selected to evaluate the pipelines. Further discussion of the statistical analysis process is covered in the section on Objective Evaluation.

The Vulnerability Model and Criticality Model consist of a number of parameters that can be selected and ranked subjectively to evaluate the vulnerability or criticality of the pipelines. These two models will provide the answer to questions like “Which pipelines are in need of rehabilitation or replacement and which pipelines will cost the most if a break occurs?’ Examples of vulnerability parameters include soil corrosivity, corrosion protection, pipe material, age of pipe, type of land use zoning the pipe is in, and whether the pipe is located in liquefaction zone or steep slope.

Examples of criticality parameters include whether pipes are directly connected to hospitals, medical facilities, kidney patients, dialysis centers, community centers, schools, major water users, and whether the pipe is located in flood-prone areas.

The application is designed with the flexibility to allow users to rank or prioritize pipelines from different points of interest. For example, the application can be used to rank pipelines for capital improvement program (CIP) improvements, or to prioritize pipelines for corrosion protection, or to prioritize pipelines to improve seismic reliability. This flexibility is achieved by allowing users to create “sessions” that are tailored to a specific point of interest. With each new session, the user can select the models for evaluating the pipelines; within each model, users can select the parameters. All information generated by a “session” created by one user is available to other users who want to review the results or create new “sessions” based on those properties.

Figure 2.1 provides an overview of the different levels that users will need to interact within order to prioritize the pipelines. Figure 2.2 shows the process flow for a user of the PIPES application.

The application does not store session results in the database; instead, the application stores all the properties of a session. This includes information on the models, parameters, categories, and all the ranking points assigned for the particular session. This approach reduces the storage space required since the potential for multiple sessions being created by multiple users to evaluate thousands of pipes in the system could quickly take up a lot of disk space should all the results be saved in a database. The advantage of storing only the session properties is that the same session can be re-executed periodically when enough changes in the pipeline attributes have taken place. The most frequent changes would occur as pipelines are rehabilitated or replaced in the system. Other parameters such as land use zoning, soil type, and neighborhood areas are less likely to change much over time.

Figure 2.1 PIPE Application-Main components

Figure 2.2 PIPES Process Flow Diagram

Development of the evaluation methodology
The use of statistical analysis to predict failure has a lot of benefit in the ranking or prioritizing process, especially since it is an objective evaluation and there is little interaction required by the end user. However, the process of ranking and prioritizing projects may involve much more complex issues such as politics and other areas that require sound engineering judgment. To consider these issues, a more subjective ranking system may be required. A more subjective ranking process is also required to consider events that may not have adequate historical data to support a statistical analysis. The evaluation methodology adopted in the PIPES application combines the benefits of the two approaches—an objective evaluation and a subjective evaluation.

Obiective Evaluation
The purpose of the statistical analysis is to determine if any combination of available data (e.g., pipe age, diameter, leak history, installation date, soil type, zoning, etc.) could be used to predict the time and probability of leaks in water pipes. The final output of the model is a listing of main segments prioritized on the probability of failure for a given time period.

Data Collection and Formatting
Data were collected from Seattle Public Utilities’ GIS and Seattle City Light (SCL) GIS. The data were formatted for use in the statistical software package S-PLUS (Version 3.1 for Windows). Quality control techniques were used to eliminate bad data, such as those pipes with leak dates prior to the installation date, and those pipes which could not be classified into categories of interest for this study

Model Selection
Standard regression analysis, whether linear or non-linear, is inappropriate for projects that contain censored data. Censored data, in the case of water pipes, are data about pipes which have not yet leaked. That is, they have not had a failure within the time period beginning with their installation and ending with the time of this study. Information on censored pipes is as important as information on pipes which have leaked within the study period. In a standard regression analysis, the time until the first leak for these censored pipes would either have to be ignored (i.e., the data discarded) or the failure time estimated as some time in the future. Both methods would result in biased model parameter estimates and poor predictions. For example, if censored data are ignored, the time until the first leak will be grossly underestimated. Estimating the failure time for censored pipes (perhaps as the censored time plus a constant) could lead to bias in either direction. Failure time analysis is a statistical technique for estimating failure times when some data are censored (Weston 1996).

Failure time analysis was used to account for the unobserved or censored leak times, since most of the pipes in the study have never leaked. Standard regression techniques would model the time until the first leak as some function of the independent variables. Failure time analysis models the probability that the pipe will fail before a certain time as some fi.mction of the independent variables. These probabilities for the first and subsequent leaks in water pipes were modeled using various techniques. The Weibull regression model was determined to be the most appropriate model, based on the available data (Weston 1996).

The survival function for the Weibull distribution:

The variables length, epoch, material and press remained significant as predictors of second leak time. No independent variables were found to significantly affect the time of third or subsequent leaks (Weston 1996).

Subjective Evaluation
The basis of a subjective evaluation is to provide the means to determine rational weighings between many attributes, factors, or issues that may influence the outcome of a judgment. Some rational weighings may require sound engineering judgment, while others may be more of a business policy. In order to prioritize or rank pipelines based on numerous different areas, a methodology to transfer subjective factors into a point system was adopted. The point system adapted to rank the pipelines utilizes a multi-level multiplication approach. The first level is in assigning the weight that each modef selected will contribute to the final score. For example if all three models are selected-deterioration, vulnerability, and criticality models—a percentage weight that each model will contribute to the final score may be assigned such as 50 percent, 30 percent, and 20 percent, respectively. The total weights from all models should equal 100 percent.

The second level of the point system is to assign a relative importance factor to each parameter selected form each model. Each model can have one or more parameters selected for evaluation, with the exception of the Deterioration Model. The Deterioration Model is preprogrammed with equations derived from a statistical analysis that has pre-identified a few parameters as being significant in predicting the probability of pipeline failure. For the subjective models, the relative importance factor (from 1 to 10 maximum) will be used as a multiplier to calculate the final score. The most effective way to rank the parameters is to select the most important parameter and assign the maximum point to it. Based on a relative importance scale in relation with the highest assigned parameter, all other parameters are then ranked. For example, four parameters are selected for the Vulnerability Model: (a) age of pipe; (b) soil corrosivity; (c) pipe material and class; and (d) whether a pipe is located upon a steep slope. If the most important parameter that would determine the vulnerability is (d); then a value of 10 is assigned to this parameter. Next, the question is asked “How important are the other parameters in determining the vulnerability of the pipeline with relation to (d)?” If the answer for one of the other parameters is “half as important,” then the relative importance is 5 points. If(d) is ten times more important, then the relative importance is 1. If(d) is a hundred times more important than a certain parameter, then perhaps that certain parameter is too insignificant and it should not be included in this evaluation. The user can easily add or remove parameters from selected models.

The third and final level of the point system is to assign points (1 to 10 max) to each category under each parameter. Categories in the PIPES application refer to the unique values that make up each parameter. For example, the soil corrosivity parameter consist of seven categories - a) Unknown b) Non Corrosive; c) Mildly Corrosive; d) Moderately Corrosive; e) Corrosive; ~ Very Corrosive; g) Severely Corrosive. However, there area number of parameters that have too many unique categories to rank them individually. For these parameters, the categories are grouped into super categories, and the ranking points are assigned to these groupings instead. An example of a parameter with too many unique categories to logically rank is “Age of pipe.” For this parameter, the super categories would consist of pipe ages that are grouped into ranges: 0-20 years, 20-40 years, 40-60 years, etc.

The points system calculations are summarize by the following equations:
Pipe score from one model = S_n[(Relative Importance) x (Category or Super Category Ranking)]
where n = number of parameters selected for this model
Total multi-model score = S_m[(Pipe score from each model) x (Weight of model)]
where m = number of models selected
Total normalize score = [(Total multi-model score)/ (Max score that can be obtained)] x 100

Application Development Process

GIS Analvsis and Data Processing
The backbone of the whole evaluation process rests on the GIS analysis and data processing to provide all the attributes of the pipelines as well as to find spatial relationships to other layers of information. To maximize the performance of the application, the GIS analysis step is done as a preprocessing step. All attributes identified as parameters to be used for statistical analysis as well as for ranking the pipelines are denorrnalized into a single table. The following are some of the GIS analyses and processes performed to compile the information for this application:

Identify and compile essential attributes of the water pipeline iiom the GIS database and various other related tables.
Combine segments of pipelines to produce pipe segments of similar characteristics that are more representative of city blocks. The combined pipe segments are more closely related to pipeline projects. In other words, a project to rehabilitate or replace pipelines will more likely be carried out on a whole city block than on individual pipe segments as captured in the GIS. Pipe segments captured in the GIS are split at all fittings and junctions to provide the most detailed information that would meet all usage requirements of the GIS database.
Perform overlay analysis of the pipelines to capture layers of information that are represented by polygon features such as soil, water pressure zones, land use zoning, steep slope areas, landslide areas, liquefaction zones, flood-prone areas, neighborhood classification, etc.
Perform proximity analysis of the pipelines to capture layers of information that are represented by point-and-line features such as soil corrosivity points, electrical vaults, underground conductors, etc.
Perform calculations of annual water usage by linking GIS pipeline data with an external database that provides information on customer water usage.
Perform calculations of static water pressure by relating pipelines to the elevation contour layer and the pressure zones that the pipelines fall within.
Perform statistical summaries on the number and types of service taps to the pipeline segments.
Perform connectivity analyses of pipelines to critical services such as hospitals, medical facilities, dialysis centers, schools, etc. Compile and create additional data from other independent studies that would be beneficial to this pipeline evaluation process, such as information about the seismic reliability of pipeline segments.

Development Environment
For implementing PIPES, the following development environment was used:

Database: ORACLE 7.2.
Client development: Oracle Developer/2000(Forms, Reports and Graphics).
Server development: Oracle PL/SQL(Packages to store all object behavior and
manage the data in Oracle database).
Desktop GIS: ArcView 3.1 and Avenue.
Data processing: Arc/Info 7.x and AML.

References
Roy F. Weston, Inc., and TerraStat Consulting Group. (December 1996). Pipe Evaluation System (PIPES)-Deterioration Model Statistical Analysis. Internal report from consultants to Seattle Public Utilities.

Acknowledgements
The main developer of the PIPES application database model and graphical user interface design is credited to Ram Pratti (Roy F. Weston, Inc.). In addition, the development of PIPES required considerable input fi-om the engineers, planners, crew chiefs, and other staff members of Seattle Public Utilities whose patience with the process and contributions are much appreciated. The support and contributions from Aziz Alfi, Jim McNerney, Jack Herold, Lionel Sun, Greg McFarland, and Tom Nolan are especially acknowledged.