by Matthew Grewe and Tim Medearis
Published in the Florida Water Resources Journal, May 2019 Issue
Ten years ago, there was very little data to prioritize asset rehabilitation properly; today, there is too much data. Utilities are now typically data rich, but information poor, and utilities need to ensure that their data have intent. Finding causal relationships, utilizing real-time data, and completing data-driven decision making—and moving away from the “gut-feel” methodology—is key to maximizing public funds and reducing risk.
Many capital improvement programs spend millions of dollars annually; however, they may do little to lower total system risk if they are focused on infrastructure that is not hydraulically significant to system operations. Inspection, closed circuit television (CCTV), computerized maintenance management systems (CMMS), and enterprise resource planning (ERP) asset management tools collect valuable condition, work order, and asset inventory data, and this information, coupled with analytical tools for developing risk scenarios and logical rehabilitation, will combat these challenges. Plans may be created, and when performed in the geographic information system (GIS) environment, they may be easily bundled, packaged, estimated, and reported on.
InfoAsset Planner software, an ArcGIS Desktop tool, is used to perform batch risk-based modeling and analysis for multiple asset types in water systems to assess the impact of being taken out of service or because of failure. Up to six complex statistical regression models may be employed to quantify the causal relationships and reliability for a utility’s network. These statistical models are based on a utility’s data and not normal industry standards. The statistical models are available to be utilized in risk-based modeling completed by the software.
The likelihood of failure can be based, not only physical characteristics such as age and material, but on other hydraulic, environmental, and inspection data factors, such as velocity, depth/diameter, sea rise, saltwater corrosion, soil type, work history, and sewer CCTV data. Additionally, the consequence of failures can be based on locational and environmental characteristics, such as crossings (railroad, highway, waterways), location to critical facilities (hospitals, schools, military bases, etc.), and demand or flow, to determine the real impact of failure and ensure adequate service to customers.
The software allows for a more detailed, granular approach in which each asset has a risk score calculated as a combination of probability and consequence of failure calculations. Multiple risk scenarios can be modeled with results that can be observed visually in ArcGIS to review high-risk areas, as well as which areas can provide the most risk reduction.
Lastly, utilities need to maximize public funds to meet growing needs for the rehabilitation of their existing networks. As systems near the end of their useful life, utilities run the risk of making costly emergency repairs on pipes that may fail or exceed capacity. By utilizing not only risk scoring, but all available data (CCTV, CMMS work order and inspections, hydraulic modeling results, etc.) to determine a rehabilitation action, as well as prioritize the order of asset rehabilitation, the software’s prioritized rehabilitation plan will ensure that utilities "get the most bang for their buck" and lower their total system risk, as opposed to simply replacing older pipes first.
Asset Management Methodology
With multiple asset management methodologies and frameworks available, such as the International Organization for Standardization (ISO) 55000 and Publicly Available Specification (PAS) 55, utilities have multiple options available to understand the needs of their water systems (potable water, wastewater, and reclaimed water) individually. Additionally, these risk-based methodologies that use consequence of failure (COF) and likelihood of failure (LOF) directly correspond to the American Society of Civil Engineers (ASCE) standards for asset management.
The software was specifically developed for utilities to configure their asset management plan by using any of the previously mentioned methodologies and frameworks, either specifically or even parts and/or combinations of an approach that works best for them. These same methodologies and frameworks can be applied to many of the different GIS-based asset types.
Asset management evaluations are typically based on physical characteristics of an asset; however, there are many other factors (hydraulic, environmental, and inspection) that impact the proper rehabilitation action required for an individual asset. Additionally, many times utilities may have CMMS information stored on one server and CCTV survey inspections on another, but hydraulic models and GIS groups working independently of one another. Understanding how these multiple data sources can be utilized in risk-based evaluation is critical in understanding how each asset performs in its ability to ensure a level-of-service standard. The software is a key tool that allows all these separate data streams to be integrated into one software and be evaluated.
With all of this information tied to the root asset in one database and software, decision makers can make more-informed rehabilitation or capital improvement project (CIP) decisions. By storing the software’s database in the powerful and flexible Environmental Systems Research Institute (ESRI) environment, this data integration is simplified and easy for current ArcGIS users to understand. This also opens the project database and software analysis results for easy access and publishing on ArcGIS Online.
Finding the causal (good and bad) relationships for each asset type is key to identifying the proper rehabilitation method and maximizing the budgets required to proactively maintain level-of-service standards. The software enables a user to evaluate pipe material, age, number of CCTV defects, type and number of associated work orders, and hydraulic capacity side by side. Its ability to integrate with multiple data sources helps to uncover these causal relationships and the multitude of different factors that lead to more easily predicting asset failure. Two examples experienced by existing users include:
These examples of causal relationships can be utilized as a multicriteria likelihood of failure in a risk-based evaluation to prioritize similar assets for faster replacement.
With all of this information in one place, it becomes much easier for organizations to create reports, charts, and graphs showing the relationships among the many factors that might lead to asset failure. Figure 1shows a graph displaying how the overall risk for the network may increase, as a factor of pipe age for example.
Figure 1. Network Risk
A more statistical way to identify trends and causal relationships may be accomplished through statistical deterioration. Standard deviation evaluations can be utilized in the software to define pipe failure criteria based on historical failure records, past work orders, and/or CCTV inspections. Once a set number of failures have been identified and joined to an asset, users can select several variables, such as material, diameter, slope, CCTV scores, or risk results and use a linear regression analysis to determine if the variables are correlated to pipe failure or not. Linear regression analyses are calibrated with a coefficient to either make it more or less lenient in terms of finding variables, which might correlate to overall asset failure.
In Figure 2, material, diameter, and peak score (from CCTV), and the created likelihood of failure (LOF 2) were determined to be significant factors in a particular model, while slope, risk score 1, and length were not significant factors. In this simple but powerful example, software users can get a statistical sense of what variables might be most important to them as they decide how to prioritize and plan rehabilitation projects.
Figure 2. Client Example of Material, Diameter, and Peak Score, and Likelihood of Failure
These basic sensitivity analyses results can, in turn, provide more dynamic reports, such as cohort and regression analyses. These graphical results use the actual failure data from a utility, combined with the sensitivity analysis to produce survival probability, failure rate, and cumulative failure rate curves for linear assets. If utilities do not have the failure data necessary to generate these curves, software users also have the option of manually setting their asset curves based on industry-standard data, such as the American Water Works Association (AWWA) report, “Buried No Longer,” or local observations. Both cohort and regression analyses come with a variety of deterioration equations, which users can choose, depending on their preferences.
Cohort model analyses can be used to predict the service life of groups of facilities. This method will group pipes with similar likeness (e.g., material types, pipe diameter, or soil types) and then create a deterioration model for each group. An example is shown in Figure 3.
Figure 3. Deterioration Model
Regression model analysis can be used to predict the service life or failure probability density of assets based on individual variable values, which are much more data-intensive than cohort models. With this additional data, however, users can generate curves that are independently dependent on any of the variables found to be significant in the sensitivity analysis. An example of failure probability density is shown in Figure 4. Software users can compare a 70-year-old survival curve and ductile iron pipe (with a peak CCTV score of 2) against a 20-year-old cast iron pipe (with a peak CCTV score of 4). This level of granularity is only accessible with large amounts of data from multiple data sources (i.e., GIS pipe feature class and CMMS failure records).
Figure 4. Failure Probability Density
Risk-based modeling (Figure 5) is critical to understanding how to prioritize assets for rehabilitation, as well as being able to quantify the amount of risk that can be reduced for each asset type. The software’s ability to allow users to complete multiple risk-based modeling evaluations is critical to maximizing public funds.
Flexible in nature, but still based on the industry standard and ASCE methodology of COF combined with LOF equaling overall risk (COF x LOF = risk), the software allows users to complete a risk analysis in minutes. Quickly evaluating risk-based evaluations allows users to spend their time modeling how to reduce risk.
When utilities have the ability to weight their criteria, combine different components into a single criterion, and compare risk analyses in ArcGIS, they can better understand their asset prioritization and reduce risk. Risk analyses are especially useful when inspection data do not exist and are not possible due to asset accessibility.
Figure 5. Risk-Based Modeling
For utilities to move away from “gut-feel” rehabilitation planning and move toward planning that is data-driven, it’s important to provide transparent communication throughout an organization, as well as to its customer base. Data-driven rehabilitation planning is only possible by evaluating multiple data sources in one location, allowing different data streams (CCTV, CMMS, GIS, hydraulic models, etc.) and pre-analysis (deterioration modeling, risk analysis, CCTV analysis) to develop a recommended rehabilitation action.
Existing users of the software utilize its rehabilitation plans to not only evaluate risk-based scoring for different asset types but to also process real field data observations that are collected. Real field data, such as CCTV data defect codes, can be utilized to determine if a gravity main should have one of the following rehabilitation actions completed: 1) no action, 2) CCTV again, 3) point repair, 4) cured-in-place pipe (CIPP) lining, or 5) full replacement.
Additionally, just because a risk-based scoring evaluation identifies an asset as “extreme risk” doesn’t mean that it must be replaced. Sometimes the proper rehabilitation action is additional investigated field work to understand and establish a better failure probability. Examples of additional investigative field work completed by existing users are associated with understanding pipe wall deterioration, including completing coupon tests or ultrasonic testing to further understand a rate at which failure may occur.
Figure 6. Rehabilitation Decision Tree
The rehab decision trees (one example is shown in Figure 5) that can be developed in the software allow asset management to come full circle. From data collection to analysis, to rehabilitation recommendations, to work order generation for recommended fieldwork, the results from the new data can be brought in and evaluated during the next asset management cycle.
With the software as an extension within ArcMap, it allows advanced locational and environmental factors to be included when performing risk-based modeling for each individual asset in utility water system networks. Additionally, the software brings multiple data sources together from multiple departments (planning, engineering, operations, information technology, etc.) to statistically evaluate failure determination and/or survival probabilities and identifies a rehabilitation action.
These important features and functions allow utilities to replace the right asset, for the right reason, at the right time—at minimal cost and minimal risk—while at the same time achieve an agreed level of service to all stakeholders.