Soft Computing Applications in Pavement
and Infrastructure Management:
State-of-the-Art
Gerardo W. Flintsch
Assistant Professor, The Via Department of Civil and Environmental Engineering Transportation Fellow, Virginia Tech Transportation Institute
200 Patton Hall, Virginia Polytechnic Institute and State University Blacksburg, VA 24061-0105
voice (540) 231 9748, fax (540) 231 7532 email: [email protected]
Virginia Tech Transportation Institute
Virginia Tech
SOFT COMPUTING APPLICATIONS IN PAVEMENT AND INFRASTRUCTURE MANAGEMENT: STATE-OF-THE-ART
ABSTARCT
Pavement and infrastructure management decisions are often based on data that is uncertain, ambiguous, and incomplete; furthermore, these decisions incorporate engineering judgment and expert opinion. These cases are hard to handle by traditional decision-support tools. Soft computing tools are particularly appropriate because these techniques can handle both numerical (even imprecise, uncertain, ambiguous, and incomplete) and subjective information.
A general modular framework for a pavement management system (PMS) was identified,
considering its relationship with asset management and the different levels of decision-making. The three main soft computing constituents were reviewed: artificial neural networks, fuzzy systems, and
evolutionary computing techniques. The most promising techniques for the different PMS functions were identified based on a comprehensive review of soft computing application in pavement and infrastructure management.
Soft computing techniques hold great promise in supporting pavement management development. Although the soft computing constituents have several advantages when used individually, the
development of practical and efficient intelligent tools is expected to require a synergistic integration of the complementary members into advanced hybrid models. These models should be implemented in user friendly tools that are designed to be easily integrated into existing PMS packages.
INTRODUCTION
Infrastructure management is a very timely issue. On one hand, sound infrastructure systems play a vital role in encouraging a more productive and competitive national economy (1). On the other, increasing demands, shrinking financial and human resources, and increased infrastructure deterioration have made the task of maintaining our infrastructure more challenging than ever before. Infrastructure systems have gradually deteriorated due to environmental action and use that, in many cases, significantly exceeds the design expectations. This has resulted in a decrease in the public level of service, which has been clearly manifested in a recent survey conducted by the American Society of Civil Engineers (2). America’s infrastructure, the nation’s critically important foundation for economic prosperity, only received a cumulative grade of D+. The transportation section is not the exception; the study determined that one-third of the nation's major roads (pavements) are in poor or mediocre condition, costing American drivers an estimated $5.8 billion a year. Furthermore, these poor road conditions contribute to as many as 13,800 highway fatalities annually.
Decision-makers are faced with competing investment demands and must distribute limited resources so that the infrastructure systems are maintained in the best possible condition. Infrastructure management systems have proven successful in bridging the gap between infrastructure condition and user expectations. For this reason they have gained widespread application among infrastructure agencies. Engineering management systems for a specific type of infrastructure (e.g., pavement and bridges) as well as for holistically managing many types of infrastructure assets have been developed by local, state, and national agencies as discussed in the following section.
In many cases, pavement and infrastructure management decisions are based on data that is ambiguous and sometimes incomplete; furthermore, these decisions incorporate engineering judgment and expert opinion. These cases are hard to handle by traditional decision-support tools and require more flexible approaches. Soft computing tools are particularly appropriate because these techniques can handle both numerical (even ambiguous and incomplete) and subjective information. The objective of this paper is to review applications of soft computing techniques in pavement and infrastructure
management, highlighting the advantages over traditional approaches. Although the focus is on pavement management system (PMS) applications, the problem is approached from an asset management
perspective, thus applications for other types of assets are also included. The paper also provides a quick overview of the soft computing techniques that hold the most promises to enhance infrastructure
management process and provides a generic framework for the use of soft computing tools in pavement management.
INFRASTRUCTURE MANAGEMENT SYSTEMS
Public and private agencies have always tried to maintain their infrastructure assets in good and serviceable condition at a minimum cost; thus, they practiced infrastructure management. However, as most of the nation’s infrastructure systems reached maturity and the demands started to rapidly increase in the mid-1960s, infrastructure agencies began to focus on a systems approach for infrastructure management that has evolved into today’s asset management concept, as illustrated in Figure 1. The process started with the development of pavement management systems (3), continued with bridge management systems (4) and infrastructure management systems (5), and has recently evolved into asset management.
One of the key steps in the development of engineering management systems is the concept of
integrated infrastructure management systems (IMS). Hudson et al. (6) defined an infrastructure management system as the operational package that enables the systematic, coordinated planning and programming of investments for expenditure, design, construction, maintenance, rehabilitation and renovation, operation, and in-service evaluation of physical facilities. The system includes methods, procedures, data, software, policies, and decision means necessary for providing and maintaining
infrastructure at a level of service acceptable to the public or owners. The advantages of the integrated approach over the traditional approach of managing each asset type individually are discussed by Zhang and Hudson (7).
The evolution towards asset management has followed the example of the private sector. Industry leaders develop tailored asset management systems to monitor and assess the status and
condition of their physical and financial assets (real estate, physical plants, inventories, and investments) individually and collectively. These systems give them the information and tools they need to retain their competitiveness. Similarly, public sector officials responsible for the nation's infrastructure need tools that allow them to maintain, replace, and preserve the nation’s infrastructure assets in the best possible condition. Decision-makers must make the best use of limited resources and must insure accountability to the public service. This can be achieved effectively using asset management, a systematic process of maintaining, upgrading, and operating physical assets cost-effectively (8).
Asset management combines engineering principles with business practice and economic theory and is today's best approach for balancing growing demands, aging infrastructure, and constrained resources (8, 9). Good asset management also implies, as IMS, a systematic integrated approach to project selection, analysis of tradeoffs, resource optimization, programming, and budgeting (10).
The overall scheme for asset management is presented in Figure 2 (8). As with all engineering management systems, asset management relies on accurate asset inventory, condition, and system performance information. Effective performance modeling is needed to find the best timing for the maintenance and rehabilitation actions as well as to assess the impact of the decisions on the overall performance of the systems being managed. The asset inventory and condition data is used to develop feasible alternatives based on the agency goals and policies, user expectations, and available resources. Alternative investments and funding scenarios are evaluated to determine their impact on system
performance and the compliance with user expectations and the agency’s financial constraints. Decision-makers use this information to prepare short- and long-term plans that are now more systematic, broader in scope, and more supportable by field data than those determined using traditional approaches. Once the plans are implemented, the performance results are monitored to verify the assumptions and
predictions made at the alternative evaluation and planning stages (8). Asset management in its broadest interpretation covers not only physical assets but also human resources, equipment and materials, and other items of values such as right-of-way, data, computer systems, etc.
PMS FUNCTIONAL FRAMEWORK WITHIN AN ASSET MANAGEMENT PERSPECTIVE Pavement management systems are one of the key components of transportation asset management, not only do they provide the framework for their development, but they also account for up to 60% of the total assets in a typical Department of Transportation (DOT). A general functional framework for a pavement management system, considering its relationship with asset management and the different levels of decision-making, is presented in Figure 3. A similar scheme could be used for other types of infrastructure assets, individually or holistically, in an infrastructure management system.
Although the approaches used by agencies differ, the scheme presented in Figure 4 is very flexible and, in general, covers current practices. The foundation of the systems is a database that includes pavement inventory, condition, traffic data, and treatment information. The information in this database is analyzed though a series of modular applications. Some agencies use a "bottom-up"
approach, in which optimal maintenance and rehabilitation (M&R) strategies are determined for each section in the network. The resulting projects are sorted by priority. A prioritized list of the projects is created, and a work program is then prepared based on this list. Other agencies have adopted a "top-down" approach, in which optimal network strategies are first determined and specific treatments are then identified, considering site-specific conditions and administrative policies.
Network-level tools evaluate the condition of the pavement network and predict pavement performance over time; identify appropriate maintenance, rehabilitation, or replacement (M&R) strategies for each roadway section; and evaluate the different alternatives to determine the network needs. Since the needs exceed the available sources in most cases, the process continues by prioritizing or optimizing the allocation of resources to generate plans, programs, and budgets. The systems produce reports and graphical displays tailored to different organizational levels of management and executive levels, as well to the public (11).
Project level analysis tools are then used to design the projects included in the work program. The pavement management cycle continues with the execution of the specified work. Changes in the infrastructure assets as a result of the work conducted, as well as normal deterioration, are periodically monitored by means of nondestructive techniques and fed back into the system.
From an asset management perspective, the network level programs and budgets are defined by higher-level strategic decisions that set the overall goals of system performance and the policies of the agency. These decisions are supported with tools that analyze tradeoffs among competing asset classes and programs so the selected investments provide the expected level of service to the customers at a cost that is compatible with the agency’s budget constraints.
Asset management has brought significant changes to agency practices. Many agencies have focused attention on asset inventory and condition data integration. In addition, there is a trend towards supplementing subjective, policy-based decision-making with objective performance-oriented tools. Many agencies are working on integrating management decisions of existing “stovepipe” management systems (such as PMS), or at least their outputs, with executive-level decisions. They have developed, or are developing, performance-based planning and programming, and executive-level program reviews that explicitly consider program performance and the impact of investment decisions on the service to the users (10). These new approaches require specialized tools for “what if” tradeoff analysis, life-cycle analysis, systems monitoring, and program evaluation. This paper reviews how these tools can be enhanced by the use of soft computing techniques.
SOFT COMPUTING
Soft computing is an umbrella of artificial intelligence techniques that handles both subjective and numerical information and tolerates imprecision, uncertainty, ambiguity, and partial truth. Soft computing includes three principal constituents (12, 13): neural networks, fuzzy mathematical programming, and evolutionary computing (including genetic algorithms). These complementary members can be integrated into hybrid systems that tolerate imprecision, uncertainty, and partial truth to produce stable, tractable, and robust solutions (12).
There are several pavement and infrastructure management characteristics that make these fields particularly attractive for the use of soft computing techniques. These characteristics are described below:
1. Available information may be uncertain, ambiguous (different solutions for the same set of data), and incomplete (the amount of information available for different roadway sections is highly variable).
2. Both objective (numerical) and subjective (linguistic) information may be available and should be considered in the analysis. While some relevant factors are easily quantifiable, other factors or performance objectives (such as environmental effects, comfort, aesthetics, versatility, and mobility considerations) may be better evaluated using subjective terms. 3. Infrastructure management decisions often require a great deal of expert knowledge, and the
resource allocation tradeoffs that support them often involve conflicting asset performance and economical objectives and constraints.
4. The analysis, especially at the network-level, has to consider large amounts of assets as well as several feasible treatments, which can be applied at many different times along the life of the assets, creating very difficult combinatorial optimization problems.
Because of these reasons, the infrastructure management field has been a fertile ground for the application of soft computing and artificial intelligence techniques. Both traditional symbolic, production systems (rule-based) and connectionist or distributed (such as artificial neural networks) models have been used to assist with infrastructure management decisions. The first attempts concentrated on the application of rule-based expert systems, mostly for project and treatment selection (14, 15, 16, 17, 18). However, early infrastructure management applications failed to demonstrate the advantage of expert systems against a traditional rule-based decision tree. Knowledge acquisition has been the most critical aspect of the development of rule-based expert systems.
Soft computing techniques have proven more effective because they allow handling and processing of subjective and ambiguous information, and incomplete data sets. Many soft computing techniques (mainly artificial neural networks, fuzzy systems, and genetic algorithms) have been used in infrastructure management with various degrees of success. Table 1 presents a summary of the main applications that have been reported in the literature. A brief description of the main techniques used is presented following.
Artificial Neural Networks
Artificial neural networks are models structured upon the organization of a human brain and can learn if provided with a range of examples and can produce valid answers from noisy data (58). The idea of computational neural networks came from realizing the incredible capabilities of learning and adaptation of a set of biological neurons. Computational neural networks imitate, to a small extent, some of the operations perceived in biological neurons: they can be trained to assess an observed function when its shape is unknown. Neural networks are able to recognize without defining, which characterizes a highly intelligent behavior (59). This property enables these systems to make generalizations. The architecture of a neural network is characterized by a large number of simple neuron-like processing units
interconnected by a large number of connections. The pattern of connectivity among the processing units and the strength of the connections encode the knowledge of a network. The main advantages of neural networks are their learning capabilities and their distributed architecture that allows for highly parallel implementation. They are particularly appropriate in those cases in which there is a significant amount of examples available.
Fuzzy Logic Systems
Fuzzy logic systems are an extension to the traditional rule-based reasoning (expert systems), and they hold greater potential (60, 61). They incorporate imprecise, qualitative data in the decision-making process by combining descriptive linguistic rules through fuzzy logic. When various rules are activated, the binary rules that define conventional expert systems usually result in discontinuities at the exit of a system. This does not resemble human behavior, where usually a smooth relation exists between cause and consequence. Smooth relationships can be achieved by using fuzzy rules that include descriptive expressions such as poor, fair, or good (Figure 4) to categorize linguistic input and output variables. Fuzzy logic was developed to provide soft algorithms for data processing that can both make inferences about imprecise data and use the data. It enables the variables to partially (up to a certain degree) belong to a particular set and, at the same time, makes use of the generalizations of conventional Boolean logic operators in data processing. The main advantage of this approach is the possibility of introducing and using rules from experience, intuition, and heuristics, and the fact that a model of the process is not required.
A particular type of fuzzy system that holds great potential is the type-2 fuzzy system. The concept of a type-2 fuzzy set was introduced by Zadeh (62, 63) to account for some of the uncertainties in the fuzzy logic systems. Type-2 fuzzy logic systems (64, 65) are those in which at least one of the antecedent or consequent sets is type-2. This type of fuzzy system allows one to consider uncertainty about: the meaning of the words, the consequents, the measurements that activate the fuzzy logic system, and the data that are used to tune the parameters of a fuzzy logic system. The uncertainty about the meaning of the words that are used in the rules is due to the subjective procedure used for determining membership functions. Membership value assignments depend on the given situation as well as on the
person making the assignments. For example, Figure 4a depicts two different membership value
assignments based on two different subjective feelings about the notion of “poor condition,” based on a 1 though 5 condition index. The uncertainty about the consequent that is used in a rule is frequently caused by the experts’ disagreements about possible consequences.
Genetic Algorithms
Genetic algorithms are some of the most common evolutionary computing techniques. These are new developments of computational systems that draw their inspiration from nature. The evolutionary models of computation also include evolutionary strategies, classifier systems, and evolutionary programming. Genetic algorithms (66, 67) represent search techniques based on the mechanics of natural selection used in solving complex combinatorial optimization problems. These algorithms were developed by analogy with Darwin’s theory of evolution and the basic principle of “survival of the fittest.” The search is run in parallel from a population of solutions. New generations of solutions are generated through reproduction, crossover, and mutation until a pre-specified stopping condition is satisfied. The main advantage of this approach is that, in many cases, it is very efficient in producing “good” solutions for difficult
combinatorial optimization problems. Hybrid Systems
Although the soft computing constituents have several advantages when used individually (Table 2), the development of a practical and efficient system often requires a synergistic integration of the
complementary members into hybrid systems. For example, while assessing the feasibility of using fuzzy technologies for life-cycle cost analysis for complex weapon design, Senglaub and Bahill (68) concluded that the technique had potential only in a hybridized environment, not as a stand alone solution. A combined hybrid system makes it possible to achieve tractability, robustness, low solution cost, and better rapport with reality (12). The full potential of soft computing techniques resides in the development of truly “intelligent,” user-friendly, decision-support tools. These are tools that can handle uncertain, subjective, incomplete, and/or ambiguous information; generate knowledge by learning from examples and/or experts; and improve their performance as they are used. These tools can take advantage of hybrid soft computing architectures that cleverly combine several techniques that add to their capabilities and benefits.
SOFT COMPUTING APPLICATIONS IN PAVEMENT MANAGEMENT
The application of soft computing holds great potential for enhancing several pavement and infrastructure management functions as illustrated by the many examples in Table 1. The major implemented and potential pavement management applications are discussed in the following sections.
Condition Assessment
One of the key functions included in most PMS is a pavement condition assessment module that analyzes the condition of the pavement and reduces the various condition measurements (typically distress and roughness) into one of more indices that reflect the pavement’s overall structural and/or functional condition (69). Several soft computing techniques have been used to support and enhance this process.
Backpropagation neural networks (19, 23, 35) and fuzzy systems (37, 38, 41, 43, 45) have been used to combine different pavement condition indicators into a condition index or pavement rating assignment. These techniques are particularly appropriate because they can estimate functions from samples without requiring a mathematical formulation of the dependence of output on input values. They can learn from examples. Furthermore, combinations of neural and fuzzy systems, or neuro-fuzzy models (70, 71), can include the advantages of both techniques. They can be easily trained and have known properties of convergence and stability as do neural networks, and they can provide a certain amount of functional transparency through the rule dependency, which is important to understand the problem solution.
Another area that has received significant contribution from the use of soft computing
technologies is the automatic identification distress. Neural (20, 34), fuzzy (44), and hybrid systems (50, 53) have been used for detecting and classifying distresses from visual images.
Performance Prediction
Pavement performance prediction is another area were subjectivity and uncertainty play a significant role and thus can be enhanced by the use of soft computing techniques. For multi-year programming, the evolution of the pavement condition must be predicted over time based on past condition, physical characteristics (type, structural capacity, etc.), and forecasted usage. The prediction models are based on field and laboratory data, the knowledge and expertise of the organization staff, and research. The most common approaches for performance prediction are regression analyses and Markov chains. However, neural networks and other soft computing techniques are increasingly used instead of the traditional methods. Neural models have been used for predicting roughness (28, 31), cracking development (32, 36), and the pavement Present Serviceability Rating (33). A fuzzy Markov model was formulated that based pavement condition prediction on subjective assessments of the pavement deterioration rates (45). Genetic algorithms have been used as alternatives to the traditional regression methods for developing several pavement deterioration models (48). In addition, genetic algorithms have been used for combining expert knowledge- and performance data-based life expectancy frequencies for the development of Marcov transition matrices (49).
One particular property of soft computing application that may be of great help in this area is adaptability, or the ability to acquire knowledge. These adaptive properties are important because the knowledge base is updated each year as new condition information become available as part of the performance monitoring process. Soft computing-based performance models could potentially be updated automatically as part of the feedback process.
Need Analysis
The identification of pavement sections in need of maintenance, rehabilitation, reinforcement, or replacement, as well as appropriate strategies for the identified sections, involves a great deal of
knowledge about the condition of the assets, effectiveness of the corrective strategies, and impact of the action on the system performance. The procedure for selecting candidate projects for rehabilitation varies from one infrastructure agency to another. In most cases, project selection is conducted based on some type of decision tree. Trigger values are established for various functional and structural condition parameters (11). However, many agencies have gradually been revising their criteria and starting to adopt soft computing-based tools.
Expert systems, neural networks, and fuzzy systems are replacing the traditional decision trees as indicated by some of the examples listed in Table 1. Early on, the performance of rule-based expert systems was compared with artificial neural networks for selecting sections for crack routing and sealing (21). It was concluded that the two techniques exhibited complementary strengths, and thus it was recommended for use in a combined hybrid system. A case study was reported in which a fuzzy logic system was developed for selecting pavement M&R treatments (39). The system also provides the degree
of confidence (certainty) in the proposed treatment. A fuzzy logic, multi-objective, decision-making model has been used for selecting an M&R treatment from a set of feasible treatments prepared using a rule-based expert system (40). A neural network-based project screening procedure has been developed for the Arizona Department of Transportation (25), and prototype neural systems have been developed for recommending appropriate pavement M&R treatments based on the distress types and extend present on the roadway section under evaluation (29). Genetic adaptive neural networks have been used for selecting “optimum” M&R strategies (51, 55).
As can be inferred from the examples discussed, models that allow for the combination of expert knowledge with knowledge acquired from examples may be ideal for this application. A prototype neuro-fuzzy model for pavement treatment selection has recently been proposed (72).
Prioritization
Infrastructure decision-makers often must evaluate different investment alternatives and prioritize the allocation of resources in accordance with user-provided criteria, performance requirements, and budget allocations. Many agencies use simple ranking systems for the prioritization of the competing investment options. Other agencies have already adopted life-cycle cost analysis approaches. Typical project
prioritization decisions deal with evaluating a set of alternatives in terms of a set group of decision criteria (goals, attributes). Attributes represent the different angles from which every alternative can be viewed and studied, and often attributes conflict with each other. For example, a maintenance strategy may maximize the asset performance but it may also have the highest cost. Therefore, a multi-attribute decision-making method is often necessary to weigh the importance assigned to every attribute. These methods can use deterministic, stochastic, fuzzy, or combined models. Combined multi-attribute decision-making methods are very useful for cases that involve combinations of all data types, such as in the case of infrastructure management decisions. The prioritization criteria may include subjective linguistic
evaluations of the relative contribution to each alternative so as to achieve a particular goal, uncertain (and sometimes ambiguous) numerical data, and possibly conflicting objectives. Thus, a fuzzy multi-attribute decision-making method that can handle deterministic, stochastic, and fuzzy multi-attributes emerges as a natural solution. This type of approach has been used for treatment selection (40). Other soft computing-based prioritization schemes have also been reported. Different uses of neural networks for prioritization have been evaluated with positive results (22). A ranking scheme based on analysis of fuzzy condition indexes has been proposed (45).
Optimization
Several pavement management systems currently incorporate mathematical optimization (or near optimization) functions to generate programs and budgets consistent with their performance goals and financial constrains. Linear programming is the most commonly used technique. Nonlinear
programming, dynamic programming, and heuristics have also been used (73). In particular, genetic algorithms provide an efficient heuristic for finding “good” solutions for difficult optimization problems (such as that dealt with in pavement management), and have been use for network level programming of pavement (46) and bridge deck rehabilitation projects (47).
One of the problems with strict mathematical optimization formulations is that the solution is not very stable. A small change in the budget allocation can result in significant changes in the optimized work program. For this reason, many decision-makers have been reluctant to use this type of
programming tools. Furthermore, infrastructure investment decisions are often based on imprecise or incomplete input parameters. There is often uncertainty surrounding specific costs, or some other parameters. Therefore, fuzzy optimization techniques should provide significantly more flexibility while dealing with these imprecise input data. These techniques are able to resolve certain linear or dynamic programming problems when some of the parameters in the model are fuzzy numbers. Depending on the nature of fuzziness, fuzzy numbers can represent coefficients in an objective function and/or set of
constraints. Fuzziness might also appear in the formulation of the constraints (the requirement that quantity x is “approximately less than”). A performance-base network optimization procedure that utilizes a fuzzy objective function has been proposed by Wang and Liu (42). The fuzzy optimization model increased the flexibility and stability of the programming process.
Summary
Soft computing techniques hold great promise in supporting pavement management development because the decisions supported by PMS often require a great deal of expert knowledge and often involve
handling and processing subjective and sometimes ambiguous and incomplete information. The soft computing techniques best suited for enhancing the various pavement management functions discussed are summarized in Table 3.
IMPLEMENTATION ISSUES
There are many recognized advantages for the use of soft computing for supporting pavement and infrastructure asset management, as discussed in the previous sections. Although these applications consist mostly of research efforts rather than implementation projects, these should not lead one to the conclusion that practical benefits from these powerful technologies are still far away. Soft computing has already found its way into mainstream infrastructure management, as it has (almost unnoticed) become part of our daily life. Fuzzy logic systems are used for improving image quality in digital camcorders. Artificial neural networks predict stock prices and detect credit card fraud. In the infrastructure management field, for example, artificial neural networks are an integral part of the new concrete pavement 2002 AASHTO design procedure, fuzzy systems are used for condition evaluation, and genetic algorithms are used for resource allocation optimization.
However, some agencies and transportation practitioners still have reservations about
implementing such techniques. These are usually the result of one or more of the following: resistance to change, difficulty in integrating the principles and techniques with existing practices and legacy systems, lack of understanding of some techniques, lack of quantitative evidence supporting the benefits of using the technologies, and lack of data to develop reliable models. Therefore, a wide implementation of soft computing would require: (i) top management commitment; (ii) a comprehensive education effort to promote an understanding of the principles and algorithms used; (iii) a clear quantification of the benefits of adopting the enhanced decision-support technologies; and (iv) tools that are user-friendly and
compatible with existing asset management tools and agency practices.
One possible approach to facilitate acceptance of these techniques is to provide the soft computing-based tools as alternatives to the traditional analysis tools in available PMS packages. It is expected that when the system is implemented, the user will feel more comfortable with more traditional, proven analysis tools. In addition, the soft computing-based tools may not be fully calibrated to the local conditions at that time. However, as the system is used, these intelligent alternatives will improve their performance (learn) and the user will become more acquainted with the technologies. At this point in time, the user may begin to test the new technologies and then adopt them if they prove to be more effective than the traditional tools for a particular application.
CONCLUSIONS
Pavement and infrastructure management decisions are often based on data that is uncertain, ambiguous, and incomplete; furthermore, these decisions incorporate engineering judgment and expert opinion. These cases are hard to handle by traditional decision-support tools. Soft computing tools are particularly appropriate because these techniques can handle both numerical (even imprecise, uncertain, ambiguous and incomplete) and subjective information.
A general modular framework for a PMS, considering its relationship with asset management and the different levels of decision-making, was identified. The three main soft computing constituents were reviewed: artificial neural networks, fuzzy systems and evolutionary computing techniques. The most promising techniques for the different PMS functions were identified based on a comprehensive review of soft computing application in pavement and infrastructure management. The following conclusions were drawn from this study:
• Neural, fuzzy, and neuro-fuzzy models are recommended for condition assessment and performance prediction. Adaptive hybrid soft computing applications, which are able to acquire knowledge, could be used to develop pavement condition assessment and performance models that are updated automatically as part of the feedback process.
• Models that allow for the combination of expert knowledge with knowledge acquired from examples (numerical data) are ideal for project and treatment selection, and prioritization. Neural networks, fuzzy systems, and combination thereof have been used with successful outcomes in most of the cases. Fuzzy, multi-attribute, decision-making models can be used for developing flexible project selection and prioritization tools, which could handle subjective linguistic evaluations, uncertain (and sometimes ambiguous) numerical data, and possibly conflicting objectives.
• Genetic algorithms, fuzzy mathematical programming, and advanced hybrid systems are best suited for enhancing optimization procedures. Fuzzy optimization techniques, which are able to resolve linear or dynamic programming problems with fuzzy parameters, constraints and/or objective functions, could increase the flexibility and solution stability of the programming process.
Soft computing techniques hold great promise in supporting pavement management development. Although the soft computing constituents have several advantages when used individually, the
development of practical and efficient intelligent tools is expected to require a synergistic integration of the complementary members into hybrid systems. These systems should be implemented in user friendly tools that are designed to be easily integrated into existing PMS packages. Furthermore, the development efforts should be complemented with education and dissemination activities to enhance understanding and highlight the benefits of adopting the enhanced decision-support technologies.
REFERENCES
1. General Accounting Office. U.S. Infrastructure, Funding Trends and opportunities to Improve Investment Decisions. Report to the Congress GAO/RCE/AIMD-00-35. General Accounting Office, Washington, DC, 2000.
2. American Society of Civil Engineers. The 2001 Report Card for America’s Infrastructure.
http://www.asce.org/reportcard/, May 2001.
3. Hudson, W. R., F. N. Finn, F. H. McCullough, K. Nair, and B. A. Vallerga, Systems Approach Applied to Pavement System Formulation, Performance Definition and Materials
Characterization. Final Report National Cooperative Highway Research Program Project 1-10, March 1968.
4. Hudson, S. W., R. F. Charmichael, L. O. Moser, W. R. Hudson, and W. J. Wilkers. Bridge Management Systems. National Cooperative Highway Research Project Report 300, Transportation Research Board, National Research Council, Washington DC, 1987. 5. Hudson, W. R. and S. W. Hudson. Pavement Management Systems Lead the Way for
Infrastructure Management Systems. Proceedings of the Third International Conference on Managing Pavements, San Antonio, Texas, May 1994.
6. Hudson, W. R., R. C. Haas, and W. Uddin. Infrastructure Management. McGraw-Hill, 1997. 7. Zhang and Hudson. Development of an Integrated Pavement Management System. Proceedings
of the Fourth International Conference on Managing Pavements, Vol. 4, May 1998, Durban, South Africa.
8. Federal Highway Administration. Asset Management Primer. Office of Asset Management, FHWA, Washington, DC, 1999.
9. Organization for Economic Cooperation and Development. Asset Management for the Roads Sector. OECD, Paris, 2001.
10. National Cooperative Highway Research Program Draft Phase I Report, Project 20-24(11), Asset Management Guidance for Transportation Agencies. Prepared by Cambridge Systematics, Parsons Brinckerhoff Quade & Douglas, Roy Jorgensen Ass., and P.D. Thomson, NCHRP, Transportation Research Board, Washington, DC, 2001.
11. American Association of State Highway and Transportation officials, Pavement Management Guide. AASHTO, Washington, DC, 2001.
12. Zadeh, L. A. The Role of Fuzzy Logic and Soft Computing in the Conception, Design, and Deployment of Intelligent Systems. In Software Agents and Soft Computing, Springer, 1997. 13. Zadeh, L. A. Applied Sot Computing – Foreword. In Applied Soft Computing, Volume 1, 1-2,
Elsevier, 2001.
14. Hajek, J. J., G. J. Chong, R. C. Haas, and W. A. Phang. Knowledge-Based Expert System
Technology Can Benefit Pavement Maintenance. In Transportation Research Record1145, TRB, National Research Council, Washington, DC, 1987.
15. Ritchie, S. G., Y. Che-I, J. P. Mahoney, and C. J. Newton. Surface Condition Expert System For Pavement Rehabilitation Planning. Journal of Transportation Engineering, Vol. 13, No. 2, American Society of Civil Engineers, New York, NY, 1987.
16. Lee, H. and V. Gadiero. Integrating Expert System in Existing Pavement Management Systems on Microcomputers. In Transportation Research Record 1215, TRB, National Research Council, Washington, DC, 1989.
17. Ross, T., S. Verzi, S. Shuler, G. McKeen, and V. Shaefer. A Pavement Rehabilitation Expert System (PARES) for Preliminary Design, Report FHWA-HPR-NM-88-03, New Mexico State Highway and Transportation Department, Santa Fe, NM, 1990.
18. Clark, G. and P. Mehta. Artificial Intelligence and Networking in Integrated Building Management Systems. Automatization in Construction, Vol. 6, Elsevier, 1997.
19. Pant, D. P., X. Zhou, R. S. Arudi, A. Bodocsi, and A. E. Aktan. Neural-Network-Based Procedure for Condition Assessment of Utility Cuts in Flexible Pavements. Transportation Research Record 1399, TRB, National Research Council, Washington, DC, 1993.
20. Kaesko, M. S. and S. G. Ritchie. A Neural Network-Based Methodology for Pavement Crack Detection and Classification. Transportation Research, Part C: Emerging Technologies, 1C, 1993, pp. 275-291.
21. Hajek, J. J. and B. Hurdal. Comparison of Rule-Based and Neural Network Solutions for a Structured Selection Problem. Transportation Research Record1399, TRB, National Research Council, Washington, DC, 1993.
22. Fwa, T. F. and W. T. Chan. Priority Rating of Highway Needs by Neural Networks. Journal of Transportation Engineering, ASCE, 1993.
23. Eldin, N. N. and A. B. Senouci. A Pavement Condition Rating Model Using Backpropagation Neural Network. Microcomputers in Civil Engineering, Vol. 10, No. 6, 1995, pp. 422-41. 24. Attoh-Okine, N. O. Analysis of Learning Rate and Momentum Term in Backpropagation Neural
Network Algorithm Trained to Predict Pavement Performance. Preprint, 74th Transportation
Research Record Board Annual Meeting, Washington, DC, 1995.
25. Flintsch, G. W., J. P. Zaniewski, and J. Delton. An Artificial Neural Network for Selecting Pavement Rehabilitation Projects. In Transportation Research Record 1524, TRB, National Research Council, Washington, DC, 1996.
26. Razaqpur, A. G., A. O. Abd El Halim, and H. A. Mohamed. Bridge Management by Dynamic Programming and Neural Networks. Canadian Journal of Civil Engineering, Vol. 5, No. 23, 1996.
27. Cattan, J. and J. Mohammadi. Analysis of Bridge Condition Rating Data Using Neural Networks.
Microcomputers in Civil Engineering, Vol. 12, No. 6, Blackwell Publishers, Boston, MA, 1997. 28. Huang, Y. and R. K. Moore. Roughness Level Probability Prediction Using Artificial Neural
Networks. Transportation Research Record 1592, TRB, National Research Council, Washington DC, 1997.
29. Alsugair, A. M. and A. A. Al-Qudrah. Artificial Neural Network Approach for Pavement Maintenance. Journal of Computing in Civil Engineering, Vol. 12, No. 4, ASCE, 1998, pp. 249-55.
30. Heiler, M. and S. McNeil. Pavement Condition Assessment Using Ground Penetrating Radar.
Proceedings of the Second International Conference on Imaging Technologies: Techniques and Applications in Civil Engineering. Davos, Switzerland, 1998.
31. La Torre, F., L. Domenichini, and M. I. Darter. Roughness Prediction Based on the Artificial Neural Network Approach. Proceedings of the Fourth International Conference on Managing Pavements, Vol. 2, May 1998.
32. Owusu-Ababia, S. Effect of Neural Network Topology on Flexible Pavement Cracking
Prediction. Computer-Aided Civil and Infrastructure Engineering, Vol. 13, No. 5, 1998, pp. 349-55.
33. Shekharan, A. R. Effect of Noisy Data on Pavement Performance Prediction by Artificial Neural Networks. Transportation Research Record 1643, TRB, National Research Council, Washington, DC, 1998.
34. Wang, K. C. P., S. Nallamothu, and R. P. Elliot. Classification of Pavement Surface Distress with an Embedded Neural Net Chip. Artificial Neural Networks for Civil Engineers: Advanced
Features and Applications, ASCE, 1998.
35. Van der Gryp, A., S. J. Bredenhann, M. G. Henderson, and G. T. Rohde. Determining the Visual Condition Index of Flexible Pavements using Artificial Neural Networks. Proceedings of the Fourth International Conference on Managing Pavements, Vol. 1, 1998.
36. Lou, Z., M. Gunaratne, J. J. Lu, and B. Dietrich. Application of a Neural Network Model to Forecast Short-Term Pavement Crack Condition: Florida Case Study. Journal of Infrastructure Systems, Vol. 7, No. 4, ASCE, 2001.
37. Elton, D. J. and C. H. Juang. Asphalt Pavement Evaluation Using Fuzzy Sets. Transportation Research Record 1196, TRB, National Research Council, Washington, DC, 1988.
38. Zhang, Z., N. Singh, and W. R. Hudson. Comprehensive Ranking Index for Flexible Pavement Using Fuzzy Sets Model. Transportation Research Record 1397. TRB, National Research Council, Washington, DC, 1993.
39. Grivas, D. and Y. C. Shen. Use of Fuzzy Relations to Manage Decisions in Preserving Civil Infrastructure. Transportation Research Record 1497, TRB, National Research Council, Washington, DC, 1995.
40. Prechaverakul, S. and F. C. Hadipriono. Using a Knowledge-Based Expert System and Fuzzy Logic for Minor Rehabilitation Projects in Ohio. Transportation Research Record1497, TRB, National Research Council, Washington, DC, 1995, pp. 19-24.
41. Shoukry, S. N., D. R. Martinelli, and J. A. Reigle. Universal Pavement Distress Evaluator based on Fuzzy Sets. Transportation Research Record 1592, TRB, National Research Council, Washington, DC, 1997.
42. Wang, K. C. P. and F. Liu. Fuzzy Set-Based and Performance-Oriented Pavement Network Optimization System. Journal of Infrastructure Systems, Vol. 3, No. 4, ASCE, 1997. 43. Fwa, T. F. and R. Shanmugam. Fuzzy Logic Technique for Pavement Condition Rating and
Maintenance-Needs Assessment. Proceedings of the Fourth International Conference on Managing Pavements, May 1998.
44. Cheng, H. D., J. Chen, C. Glazier, and Y. G. Hu. Novel Approach to Pavement Cracking
Detection Based on Fuzzy Set Theory. Journal of Computing in Civil Engineering, Vol. 13, No. 4, ASCE, 1999.
45. Bandara, N. and M. Gunaratne. Current and Future Pavement Maintenance Prioritization Based on Rapid Visual Condition Evaluation. Journal of Transportation Engineering, Vol. 127, No. 2, ASCE, 2001.
46. Fwa, T. F., W. T. Chan, and C. Y. Tan. Genetic-Algorithm Programming of Road Maintenance and Rehabilitation. Journal of Transportation Engineering,Vol. 122, No. 3, ASCE, 1996. 47. Liu, C., A. Hammad, and Y. Itoh. Maintenance Strategy Optimization of Bridge Decks Using
Genetic Algorithm. Journal of Transportation Engineering, Vol. 123, No. 2, ASCE, 1997. 48. Shekharan, A. R. Solution of Pavement Deterioration Equations by Genetic Algorithms.
Transportation Research Record 1669, TRB, National Research Council, Washington, DC, 2000. 49. Hedfi, A. and P. Stephanos. Pavement Performance Modeling: An Applied Approach at the State
of Maryland. Proceedings of the Fifth International Conference on Managing Pavements, Seattle, WA, 2001.
50. Ritchie, S. G., M. Kaseko, and B. Bavarian. Development of an Intelligent System for Automated Pavement Evaluation. Transportation Research Record 1215, TRB, National Research Council, Washington, DC, 1991.
51. Taha, M. A. and A. S. Hanna. Evolutionary Neural Network Model for the Selection of Pavement Maintenance Strategy. Transportation Research Record 1497, TRB, National Research Council, Washington, DC, 1995.
52. Martinelli, D., S. N. Shoukry, and S. T. Varadarajan. Hybrid Artificial Intelligence Approach to Continuous Bridge Monitoring. Transportation Research Record 1497, TRB, National Research Council, Washington, DC, 1995.
53. Chou, J., W. A. O’Neill, and H. Cheng. Pavement Distress Evaluation Using Fuzzy Logic and Moment Invariants. Transportation Research Record1505, TRB, National Research Council, Washington, DC, 1995, pp. 39-46.
54. Chiang, W., K. F. Liu, and J. Lee. Bridge Damage Assessment Through Fuzzy Petri Net Based Expert System. Journal of Computing in Civil Engineering, Vol. 14, No. 2, American Society of Civil Engineers, 2000.
55. Abdelrahim, A. M. and K. P. George. Artificial Neural Network for Enhancing Selection of Pavement Maintenance Strategy. Transport Research Record No. 1669, TRB, National Research Council, Washington, DC, 2000.
56. Chae, M. J. and D. M. Abraham. Neuro-Fuzzy Approaches Sanitary Sewer Pipeline Condition Assessment. Journal of Computer in Civil Engineering, Vol. 15, No. 1, American Society of Civil Engineers, 2001.
57. Liang, M., J. Wu, and C. Liang. Multiple Layer Fuzzy Evaluation for Existing Reinforced Concrete Bridges. Journal of Infrastructure Systems, Vol. 7, No. 4, American Society of Civil Engineers, 2001.
58. Taylor, J. C. Chapter 1: The Promise of Neural Networks. In Neural Networks. Alfred Waller Limited, Henley on Thames, United Kingdom, 1995.
59. Kosko, B. Neural Networks and Fuzzy Systems: A Dynamical Systems Approach to Machine Intelligence. Prentice-Hall, Englewood Cliffs, NJ, 1992.
60. Zadeh, L. A. Fuzzy Sets. Information and Control, Vol. 8, 1965, pp. 338-353.
61. Zadeh, L. A. Outline of a New Approach to the Analysis of Complex Systems and Decision Processes. IEEE Transactions on Systems, Man and Cybernetics, SMC 3, 1973, pp. 28-44. 62. Zadeh, L. A. The Concept of a Linguistic Variable and its Application to Approximate
Reasoning: I. Information Science, Vol. 8, 1975, pp. 199-249.
63. Zadeh, L. A. The Concept of a Linguistic Variable and its Application to Approximate Reasoning: II. Information Science, Vol. 8, 1975, pp. 307-357.
64. Mizumoto, M. and K. Tanaka. Some Properties of Fuzzy Sets Type-2. Information and Control
Vol. 31, 1976, pp. 312-340.
65. Mendel, J. M. Rule-Based Fuzzy Logic Systems. Prentice Hall, Upper Saddle River, NJ, 2001. 66. Holland, J. Adaptation in Natural and Artificial Systems. University of Michigan Press, Ann
Arbor, 1975.
67. Goldberg, D. Genetic Algorithms in Search, Optimization and Machine Learning. Addison-Wesley, Reading, MA, 1989.
68. Sengalaub M. and A. T. Bahill. Application of Fuzzy Technology to Risk-Based Design and Decision Problems. Risk and Safety Assessment: Where is the Balance? American Society of Mechanical Engineers, Pressure Vessel and Piping Division 296, 1995, pp. 191-197.
69. Gramling, W. L. Current Practices in Determining Pavement Condition. National Cooperative Highway Research Program Synthesis of the Highway Practice 203, Washington, DC, 1994 70. Nauck, D., F. Klawonn, and R. Kruse. Foundations of Neuro-Fuzzy Systems. John Wiley & Sons,
New York, 1997.
71. Russo, M., FuGeNeSys - A Fuzzy Generic Neural System for Fuzzy Modeling. IEE Transactions on Fuzzy Systems, Vol. 6, No. 3, 1998.
72. Flintsch, G. W. Soft Computing Applications in Transportation Infrastructure Asset Management.
Application of Advanced Technologies in Transportation 2002, ASCE, August 2002. 73. Zimmerman, Pavement Management Methodologies to Select Projects and Recommend
Preservation Treatments. National Cooperative Highway Research Program Synthesis of the Highway Practice 222, Washington, DC, 1995.
LIST OF TABLES AND FIGURES
TABLE 1. Summary of AI and Soft Computing Applications in Infrastructure Management TABLE 2. Advantage of the Main Soft Computing Techniques
TABLE 3. Soft Computing Technique and Their Applicability for Pavement Management FIGURE 1. Engineering Management System Evolution
FIGURE 2. Generic Asset Management Framework (8) FIGURE 3. Pavement Management System Framework
TABLE 1. Summary of AI and Soft Computing Applications in Infrastructure Management
Asset Performance Needs Analysis Tradeoff Analysis
Technology Reference Condition
Asses. Perform. Prediction Project Selection Treatment Selection Prioriti-zation Optimi-zation Hajek et al. (1987) (14) Ritchie et al. (1987) (15)
Lee & Gadiero (1989) (16)
Ross et al. (1990) (17)
Expert Systems
Clark & Mehta (1997) (18)
Pant et al. (1993) (19)
Kaesko & Ritchie (1993) (20)
Hajek & Hurdal (1993) (21)
Fwa & Chan (1993) (22)
Eldin & Senouci (1995) (23)
Attoh-Okine (1995) (24)
Flintsch et al. (1996) (25)
Razaqpur et al. (1996) (26)
Cattan & Mohammadi (1997) (27)
Huang & Moore (1997) (28)
Alsugair & Al-Qudrah (1998) (29)
Heiler & McNeil (1998) (30)
La Torre et al. (1998) (31)
Owusu-Ababia (1998) (32)
Shekharan (1998) (33)
Wang et al. (1998) (34)
Van der Gryp et al. (1998) (35)
Artificial Neural Networks
Luo et al. (2001) (36)
Elton & Juang (1988) (37)
Zhang et al. (1993) (38)
Grivas & Shen (1995) (39)
Prechaverakul & Had. (1995) (40)
Shoukry et al. (1997) (41)
Wang & Liu (1997) (42)
Fwa & Shanmugam (1998) (43)
Cheng et al. (1999) (44)
Fuzzy Logic Systems
Bandara & Gunaratne (2001) (45)
Fwa et al. (1996) (46)
Liu et al. (1997) (47)
Shekharan (2000) (48)
Genetic Algorithms
Hedfi & Staphanos (2001) (49)
Ritchie et al. (1991) (50)
Taha & Hanna (1995) (51)
Martinelli et al. (1995) (52)
Chou et al. (1995) (53)
Chiang et al. (2000) (54)
Abdelrahim & George (2000) (55)
Chae & Abraham (2001) (56)
Hybrid Systems
TABLE 2. Advantage of the Main Soft Computing Techniques
Technique Main characteristic
Artificial Neural Networks
Excellent pattern recognition capabilities. Can be trained to save, recognize, and search the shapes or elements of databases; solve
combinatorial optimization problems; recognize without definitions; and make generalizations.
Fuzzy Logic Systems
Possibility of introducing and using subjective information (including rules from experience, intuition, and heuristics) and providing functional transparency.
Type-2 Fuzzy Logic Systems are also capable of handling uncertainty. Evolutionary Computation
Can produce “good” solutions for difficult combinatorial optimization problems, can tune Fuzzy Logic Systems, and can be used in the Neural Networks training process.
TABLE 3. Soft Computing Technique and Their Applicability for Pavement Management
Technique Assessment Condition Performance Prediction Analysis Need Prioriti-zation Optimi-zation
Artificial Neural Networks
Fuzzy Logic Systems (type I and II)
Fuzzy Multi-Attribute Decision Making Fuzzy Mathematical Programming
Evolutionary Computation Neuro-Fuzzy Systems Enhanced Hybrid Algorithms
Asset
Management
Systems
Asset
Management
Systems
Pavement
Management
Systems
(60’s)Bridge
Management
Systems
(80’s)Infrastructure
Management
Systems
(90’s)Concepts
Principles
Integration
Business-like
objectives
Non-physical assetsAsset
Management
Systems
Asset
Management
Systems
Pavement
Management
Systems
(60’s)Pavement
Management
Systems
(60’s)Bridge
Management
Systems
(80’s)Bridge
Management
Systems
(80’s)Infrastructure
Management
Systems
(90’s)Infrastructure
Management
Systems
(90’s)Concepts
Principles
Integration
Business-like
objectives
Non-physical assetsShort- & Long-Term Plans (Project Selection)
Short- & Long-Term Plans (Project Selection)
Alternative Evaluation/ Program Optimization
Alternative Evaluation/ Program Optimization
Program Implementation
Program Implementation
Performance Monitoring (Feedback)
Performance Monitoring (Feedback)
Goals & Policies
Goals & Policies
Condition Assessment/ Performance Prediction
Condition Assessment/ Performance Prediction
Asset Inventory
Asset Inventory
Budget/
Allocations
Budget/
Allocations
Short- & Long-Term Plans (Project Selection)
Short- & Long-Term Plans (Project Selection)
Short- & Long-Term Plans (Project Selection)
Short- & Long-Term Plans (Project Selection)
Alternative Evaluation/ Program Optimization
Alternative Evaluation/ Program Optimization
Alternative Evaluation/ Program Optimization
Alternative Evaluation/ Program Optimization
Program Implementation
Program Implementation
Program Implementation
Program Implementation
Performance Monitoring (Feedback)
Performance Monitoring (Feedback)
Performance Monitoring (Feedback)
Performance Monitoring (Feedback)
Goals & Policies
Goals & Policies
Condition Assessment/ Performance Prediction
Condition Assessment/ Performance Prediction
Condition Assessment/ Performance Prediction
Condition Assessment/ Performance Prediction
Asset Inventory
Asset Inventory
Asset Inventory
Asset Inventory
Budget/
Allocations
Budget/
Allocations
NETWORK-LEVEL REPORTS Performance Assessment Network Needs Facility Life-cycle Cost Optimized M&R Program Performance-based Budget CONSTRUCTION DOCUMETS GRAPHICAL DISPLAYS PRODUCTS PAVEMENT DATABASE INV E NTO R Y INV E NTO R Y PAVEMENT CONDITION PAVEMENT CONDITION TRAFFIC TRAFFIC MAINTENANCE STRATEGIES MAINTENANCE STRATEGIES INFORMATION MANAGEMENT INFORMATION MANAGEMENT
Goals & Policies
System Performance
Economic / Social & Env.
Goals & Policies
System Performance
Economic / Social & Env. AllocationsBudget
Budget Allocations
WORK PROGRAM EXECUTION
WORK PROGRAM EXECUTION
PERFORMANCE MONITORING PERFORMANCE MONITORING PROJECT LEVEL ANALYSIS(Design) PROJECT LEVEL ANALYSIS(Design) NETWORK-LEVEL ANALYSIS NEEDS ANALYSIS NEEDS ANALYSIS PRIORITIZATION / OPTIMIZATION PRIORITIZATION / OPTIMIZATION PROGRAMMING PROJECT SELECTION PROGRAMMING PROJECT SELECTION STARTEGIC ANALYSIS STARTEGIC ANALYSIS FEED-BACK FEED-BACK CONDITION ASSESSMENT CONDITION ASSESSMENT PERFORMANCE PREDICTION PERFORMANCE PREDICTION NETWORK-LEVEL REPORTS Performance Assessment Network Needs Facility Life-cycle Cost Optimized M&R Program Performance-based Budget CONSTRUCTION DOCUMETS GRAPHICAL DISPLAYS PRODUCTS NETWORK-LEVEL REPORTS Performance Assessment Network Needs Facility Life-cycle Cost Optimized M&R Program Performance-based Budget CONSTRUCTION DOCUMETS GRAPHICAL DISPLAYS PRODUCTS PAVEMENT DATABASE INV E NTO R Y INV E NTO R Y PAVEMENT CONDITION PAVEMENT CONDITION TRAFFIC TRAFFIC MAINTENANCE STRATEGIES MAINTENANCE STRATEGIES INFORMATION MANAGEMENT INFORMATION MANAGEMENT
Goals & Policies
System Performance
Economic / Social & Env.
Goals & Policies
System Performance
Economic / Social & Env. AllocationsBudget
Budget Allocations
WORK PROGRAM EXECUTION
WORK PROGRAM EXECUTION
PERFORMANCE MONITORING PERFORMANCE MONITORING PROJECT LEVEL ANALYSIS(Design) PROJECT LEVEL ANALYSIS(Design) NETWORK-LEVEL ANALYSIS NEEDS ANALYSIS NEEDS ANALYSIS PRIORITIZATION / OPTIMIZATION PRIORITIZATION / OPTIMIZATION PROGRAMMING PROJECT SELECTION PROGRAMMING PROJECT SELECTION STARTEGIC ANALYSIS STARTEGIC ANALYSIS FEED-BACK FEED-BACK CONDITION ASSESSMENT CONDITION ASSESSMENT PERFORMANCE PREDICTION PERFORMANCE PREDICTION
a. Poor 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 Condition Evaluator 1 Evaluator 2 b. Fair 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 Condition c. Good 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 Condition FIGURE 4. Example of Fuzzy Membership Functions for Three Levels of Pavement Condition