Development of the Selection Assistant for Utility Locating Technologies (2011)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

C H A P T E R 2 MethodologyChoice of Decision-Support System The approach to developing decision-support aids for trans- portation professionals was developed with the following issues in mind: • The capability limits of individual utility-locating technolo- gies are poorly defined and may vary from one equipment manufacturer to another even within the same basic class of equipment. • There is little independent testing of utility-locating equip- ment and technologies to provide quantitative informa- tion to users or purchasers of the equipment and hence the principal information available on the range of applica- tion of different equipment comes from the manufacturers of that equipment. Some manufacturers are more optimistic than others. • The ability of specific technologies to find utilities is highly dependent on ground conditions and site-related, error- producing conditions. These problems are described in more detail in the main R01 report (Sterling et al. 2009). This means that if the site or ground conditions can be improved, the success of the utility-locating method can be improved for the same material type, content, diameter, and depth of utility. The issues listed have a strong influence on the type of decision-support system that can be used to provide guidance to the appropriate types of utility equipment. The following are the basic options for decision-support software: a. Deterministic decision-support systems (e.g., the value of each attribute is compared with an acceptable range for each method alternative) (see Matthews et al. 2005). b. Decision support based on fuzzy logic where degrees of membership of a given attribute value for a given method alternative can be expressed as a linguistic description3(e.g., highly probable, probable, neutral, unlikely, highly unlikely) (see Flintsch and Chen 2004). c. Selection based on a case-based selection approach involv- ing comparing equipment capabilities with the character- istics of a particular locating problem (e.g., depth of utility, depth-to-diameter ratio, utility material) based on the best fit with historical cases stored in the software database (see Morcous et al. 2002). d. Expert system analysis based on coding the questions and decisions that an experienced professional would ask and take based on the purpose of the search and the site and util- ity conditions (see Amirkhanian and Baker 1992). e. Decision support provided by synthesis of the choices or preferences that an experienced group of professionals express. These can be analyzed in several ways, including, for example, the Analytical Hierarchy Process l (AHP) (see Al-Barqawi and Zayed 2008). f. Learning-based algorithms using Artificial Neural Network (ANN), generic algorithm (GA), or another classification approach. Site and utility parameters would represent the input variable, and the likelihood of success for each approach could be the output (see Flintsch and Chen 2004). The research team reviewed the possible directions for the decision support for utility-locating technologies and made the following conclusions: a. Insufficient data is available in the public domain on the characteristics of specific equipment or even classes of technologies to make firm selections across all the poten- tial variability of soil and site conditions (thus excluding deterministic methods). b. It is not considered feasible at the current time to assem- ble a sufficient number of cases of successful and unsuc- cessful applications of utility-locating technologies under different site and utility conditions to train an artificial neural network. The large number of input variables and

4possible technologies make ANN a non-optimal approach for the task at hand. c. Many aspects of site and utility conditions are unknown when planning a utility-locating exercise. The process often becomes a series of progressively refined techno- logical approaches—starting with the simplest and cheap- est methods that are likely to work and progressing to more expensive approaches only as necessary. d. Technological approaches are usually combined with off- site and on-site detective work to help define utility loca- tion and characteristics (user knowledge increases as part of the solution convergence process). e. The range of potential utility types and site conditions that would affect the choice of a locating technology makes the creation of a limited set of questions that could reasonably be answered by an expert panel impractical for the type of comparisons used in the AHP process (comparisons made pair-by-pair among all the alternatives available). f. The most feasible way to create a decision-support system at the present time is to build an expert system that follows the decision logic of a highly experienced utility-locating professional. g. The use of fuzzy logic in making decisions would be help- ful but not essential in developing expert system logic for the decision-support process. h. The expert system will only provide guidance based on the range of conditions considered and is not a substitute for direct experience with specific equipment under specific site conditions. Expert System Platform The programming of the logic of expert systems is independent of the specific domain of expertise involved (i.e., the under- lying logic can be used to provide decision support in many application areas). This has led since the 1980s to the develop- ment of a variety of general purpose expert system software platforms (often referred to as expert system “shells”). Some are in the public domain, while others are offered for sale as software development platforms. Expert systems are often based on crisp decision logic (definitive answers to yes-or-no questions or values of decision parameters). A smaller subset of expert system shells provides the ability to use fuzzy param- eters and decision logic in the applications. After a broad survey of the systems available, the suitabil- ity for the utility-locating application and licensing costs and conditions, the team selected Jess, the rule engine for the Java platform, version 7.1p2. Jess is an expert system shell written in Java and uses a ver- sion of the Rete algorithm. An academic license for its usage is available free of cost and a commercial license at negotiable and reasonable prices. The software can be downloaded and installed on a web server; it is small (about 7 MB) and fast.Client computers do not need an installation of Jess. A Java- enabled Internet browser is adequate for executing Jess pro- grams; Jess programs can be compiled and encapsulated within Java applets, which then can be embedded in a web page. This approach has been adopted for the implementation of the expert system shell in the SAULT framework (refer to the next section on software delivery approach). Such an approach eliminates the need to distribute CD-ROMs and makes soft- ware updates at a later date much easier. If desired, access to the Jess program could be controlled by constructing a user- authenticated website. Jess programs can be edited using any text editor. Syntax definition files, which highlight keywords, are available for common text editors. Jess easily interfaces with Java. Jess objects could be used in Java programs, and Java objects could be invoked in Jess programs. This feature renders Jess easily used in conjunction with Java’s wide-ranging applications. Jess supports both forward- and backward-chaining of rules. Fuzzy logic can be implemented using Jess by downloading the FuzzyJess extension available at National Research Council Canada’s website (NRC Canada 2010). A Jess demo is available at the Jess website (Sandia National Laboratories 2010). Software Delivery Approach An important decision in making the software available to the user community is whether the software will be distributed to each user via CD-ROM or web access (download) and then be run locally, or whether the software will be accessed directly via the Internet from a central server. The following are the advantages and disadvantages of each approach: • Software distribution approach  Advantages ▪ Distribution followed by local use of the software does not require Internet access for software operation and using the application does not depend on Internet access speed.  Disadvantages ▪ Software distribution and local installation can create a need for higher levels of user support than simply accessing an application on a central server. • Web access approach  Advantages ▪ Software updates and “bug” fixes can be accomplished more quickly and easily when the software is centrally located. ▪ Software enhancements, such as additional data, case histories, and data corrections to be added to the appli- cation databases, can also be readily accomplished when installed on a central server. ▪ The use of a centralized application greatly simplifies licensing and usage issues and reduces costs.

5▪ With free access to the decision-support website, access- ing and using the software and databases is easy and quick to accomplish.  Disadvantages ▪ Some firewall-related access problems may occur. For the reasons listed, it was decided that the software should be made available over the Internet from a central server. This server is housed at the Trenchless Technology Cen- ter (TTC), Louisiana Tech University, and the software is to be supported by the TTC on behalf of the SHRP 2 program at least through January 15, 2012. Rights to the domain-specific information and logic used by the software are in the public domain. Software development and availability of the SAULT console requires commercial software licenses for the web server. Such licenses need to be available for Jess, Microsoft Access, Java, and others. A diagram of the software architec- ture is shown in Figure 2.1. Knowledge Capture The basis for the expert system is the career-long utility- locating experience of James Anspach (see Appendix C). In preparation for the knowledge capture sessions, a list of theFigure 2.1. SAULT expert systems software architecture.expected influencing parameters for utility-locating technol- ogy choices was created. However, during the preliminary dis- cussions session with Anspach, it was difficult to generate a decision logic starting from the effect of parameters on equip- ment and procedural decisions. The most effective way to capture the decision process was determined to be to follow a job-related decision process in which the nature of the utility- locating task was first identified (e.g., finding a cable or pipe), and then the series of questions that the expert would pose to help define the locating approach required would be captured. The answer to each of these questions triggers a different series of questions concerning items such as the nature of the util- ity material, the conductivity of the soil, and the accessibility of the utility for direct connection of an impressed signal. Because of the iterative nature of many utility-locating exer- cises, it was found necessary in the decision logic to include questions as to whether a particular method had previously been tried. If it had not, then this method could be suggested as the first alternative; if it had previously been tried, then an alternate method would be explored with an additional set of questions. The expected depth of the utility being sought is an important parameter; however, many decisions about the potential of various types of utility-locating approaches can be

6made with only a rough estimate of the utility’s depth of cover. The flowcharts created were grouped into six individ- ual flowcharts for ease of presentation. These flowcharts are provided in Appendix B. The decision logic may shift from one flowchart to another based on the input parameters requested, and, eventually, the decision process will result in one or more recommendations as to suitable utility-locating approaches. When none of the standard choices for utility-locating tech- nologies are expected to be successful, the software will return the answer “Exploratory Test Holes/Prototype Systems.” Otherwise, the technology choices will be one or more of the following technologies: • Magnetic locator; • Metal detector; • Pipe/cable locator (low-frequency conductive mode); • Pipe/cable locator (medium-frequency conductive mode); • Pipe/cable locator (high-frequency conductive mode); • Pipe/cable locator (medium-frequency inductive mode); • Pipe/cable locator (high-frequency inductive mode); • Pipe/cable locator (radio mode); • Pipe/cable locator (60-Hz power mode); • EM sonde and walkover locator; • Noise emission device and receiver (geophone); • Inductive array; • Ground-penetrating radar (GPR); • GPR (multichannel, multifrequency); • Infrared thermography; • Terrain conductivity meter; and • Elastic wave-based techniques. The success of some approaches can be increased by improv- ing the site conditions. These condition improvements are provided as suggestions in connection with the various tech- nology recommendations. The six condition-improvement categories are as follows: • Remove metallic surface obstacles; • Control ambient noise; • Increase thermal difference between ground and utility; • Create a new access point; • Remove snow or leaves from the surface; and • Isolate EM noise/optimize signal. Software Development With the preferred expert system shell identified, the software architecture outlined, and the knowledge capture process ini- tiated, the software development work could commence. This consisted of a series of development stages in which a sim-ple trial application was first tested to ensure that there were no major implementation hurdles to overcome, and then the full application was programmed. Table 2.1 gives the major soft- ware development tasks. Some of these tasks could be under- taken as parallel activities. Three electronic databases related to utility-locating had already been developed as part of Phase 1 of the R01 project. These databases were integrated with the website present- ing the SAULT application and are described in the “Related Searchable Databases” section. Related Searchable Databases Phase 1 of the R01 project included the development of three databases related to utility-locating issues. These databases provide the following: • Examples of utility damage and associated cause(s); • Examples of case studies where the SUE approach had been used for utility mapping together with assessment of its benefit, where available; and • The characteristics of many commonly available utility- locating technologies and equipment based mostly on infor- mation from the manufacturers’ literature.Task Selection of an expert system as the decision-support approach Capture of the decision logic in a series of extended sessions with the “expert” (Jim Anspach) Selection of the expert system shell Design of the software delivery architecture Preliminary testing of the expert system approach and software architecture Development of detailed flowcharts covering all the various utility- locating parameters and site conditions based on the decision logic captured above Customization of Jess console applet: Changing the graphic interface and printout options Development of the Java applet integrating input and output for a pilot Jess program Embedding the Java applet in the web page for a sample portion of the expert system to ensure functionality Integrating the three associated databases into the website Full development of the Jess console and Java applets Testing and validation of the Jess program for all utility-locating modules in accordance with the flowchart logic Table 2.1. Major Software Development Tasks

7In conjunction with the provision of the SAULT over the Internet, it was decided to integrate the three databases with the expert system in a coordinated application. A summary of each database and its searchability via the web application fol- lows. A user’s guide to the website is provided in Appendix A. It is recognized that none of the databases are exhaustive and that many case histories and equipment manufacturers may not be directly represented in the database. An advantage of the web-based application, however, is that additional data- base entries can readily be added and immediately be made available when new information is provided. Utility Strikes Database Utility strikes are frequent events, with a utility strike occur- ring nearly every minute somewhere in the country. Although most utility strikes result in minimal local damages, many others can result in fatalities, injuries, or significant collat- eral damage. The cost of repairing the damaged utility is often overshadowed by costs associated with disruption of services, traffic, and normal life patterns; project delays; contractor claims; and litigation. The 60 case studies currently included are presented in a standard format. Focus is given to the characteristics of the events, a short description, and causes and lessons learned if such were reported. The utility strike incidents represent a small sample of the thousands of utility strikes that occur in the United States each year and have been summarized primarily from incidents reported in the Underground Focus magazine (Planet Underground Media 2010). The incidents selected provide real-world examples of utility damage inci- dents, their causes, and the resulting disruption and finan- cial impact. Utility damage incidents are collected by numerous state agencies responsible for utility safety, and a national reposi- tory has been created by the Common Ground Alliance (CGA 2010). The Planet Underground website provided in the ref- erence also has an accident file archive. Review of the cases included in the utility strikes database suggests that the circumstances of the strike and adequacy of the response could play an equal or greater role than the crit- icality of the utility in determining the degree of damage and losses incurred due to the accident. SUE Case History Database The SUE Case History database presents selected case histo- ries of subsurface utility engineering success stories associated with transportation projects. The SUE process and associated quality level designations for utility information can be foundin CI/ASCE 38-02 Standard Guidelines for the Collection and Depiction of Existing Subsurface Utility Data (ASCE 2002). The database currently contains 59 cases. The case studies were obtained through discussion with practicing professionals, literature search, a survey of SUE projects conducted by the TBE Group Inc., and a research report released by the Univer- sity of Toronto (Osman and El-Diraby 2005; Purdue Univer- sity 1999; Sinha et al. 2007). These cases represent successful applications of SUE technologies and practices in a variety of transportation-related projects. In addition, several non- transportation related projects for which data is available as to the relative cost of the SUE effort or the estimated benefit-cost ratio, or both, for this effort are also briefly described. From a review of the database, SUE mapping surveys seem to consistently have a positive effect when performed early during the design phase of construction projects. It is not uncommon that agencies are driven to undertake SUE investigations following one or more projects that went bad because of multiple utility conflicts or serious utility-related accidents, or both. The benefit-cost ratio to project owners in the cases documented ranged between 2 and 6.6, while the cost of the SUE studies ranged between 0.125% and 2% of the total project budget. The benefit-cost ratios in all cases consid- ered only savings in terms of construction costs and schedule delays; costs associated with possible utility strikes were not considered due to uncertainness associated with the parame- ters involved. The older and more developed the area where construction is scheduled to take place, the greater is the benefit-cost potential; also, the larger the scope of the project the greater the benefit-cost ratio, and the smaller the invest- ment in SUE in terms of percentage of total budget. It can also be noted that a study by Penn State University for PENN- DOT showed a 22:1 cost ratio when it looked at 10 randomly selected PENNDOT projects (Sinha et al. 2007). Utility-Locating Technologies Database The electronic database of the characteristics of utility-locating technologies was assembled primarily using manufacturer data. The information available and the ranges of suggested applicability varied significantly from manufacturer to man- ufacturer even for similar classes of equipment. The web presentation of the database has several areas that either provide information or can be used to refine the selec- tion within the database. The operation of these is described in Appendix A. The utility-locating equipment is divided into various classes of locating technologies. An image illustrating the equipment used is provided for each method class together with a general description of this method class. A list of locat- ing equipment that falls within the method class allows infor-

8mation on specific equipment to be accessed. This information includes performance indicators and other descriptive infor- mation, such as the following: • Whether the equipment is expected to find ferrous or non- ferrous objects, or both; • Equipment applicability to four broad classes of soil type; • Minimum and maximum frequency of operation when applicable; • Effort or training required for data interpretation; • Relative cost indication; • Maximum depth of effectiveness anticipated; and• General application summary and more detailed method description. Conclusion SAULT, though not a replacement for the experiences and expertise of a utility-locating professional, is a valuable source of information and guidance. The software’s expert-system- based decision-support system assists novice users in under- standing the types of utility-locating equipment and their ideal applications. Through its large databases and accessible format, SAULT will aid professionals in understanding and implementing the process of locating underground utilities.