Pavement Preservation Project Selection and Prioritization: A Competitive Approach

Transcription

1 Pavement Preservation Project Selection and Prioritization: A Competitive Approach Prepared By Charles F. Gurganus (1) and Nasir G. Gharaibeh (2) (1) Corresponding Author, Area Engineer, Texas Department of Transportation, 205 NE Loop 564, Mineola, TX (formerly Graduate Student, Zachry Department of Civil Engineering, Texas A&M University) Charles.Gurganus@txdot.gov (2) Assistant Professor, Texas A&M University, Zachry Department of Civil Engineering, 3136 TAMU, College Station, TX Number of Words: 7,497 equivalent words (5,247 words of text, 6 tables, 3 figures) Prepared for presentation and publication at the 91st Annual Meeting of the Transportation Research Board, January 2012 July 22, 2011 Revised: November 15, 2011

2 Gurganus and Gharaibeh Pavement Preservation Project Selection and Prioritization: A Competitive Approach ABSTRACT Several methods exist to help agencies select and prioritize pavement preservation projects. These methods are often built within an agency s pavement management system. Unfortunately, it is not uncommon for these tools to disagree with actual preservation project decisions, particularly at the project selection level of pavement management. Ad-hoc preservation project selection procedures may be effective for many highway agencies; however current fiscal issues and pressure from administrators and legislators are forcing agencies to provide justification for the use of funds. This paper offers a new pavement project selection and prioritization method, using the Analytic Hierarchy Process as the multi-criteria decision making platform. The new method uses several parameters and input from decision makers to create a prioritized preservation project list. The method was applied to a case study area within Texas and projects suggested by the method matched actual decisions by 75 percent. The ability to capture multiple parameters and determine weights for each parameter based on decision maker input, along with the high level of agreement between the method and actual decisions indicates that the method could be a viable decision support tool that would experience more use at the project selection level. Key words: Project selection, project prioritization, pavement preservation, analytic hierarchy process, pavement management

3 Gurganus and Gharaibeh Pavement Preservation Project Selection and Prioritization: A Competitive Approach INTRODUCTION Pavement management decisions are made at four levels: strategic, network, project selection, and project levels. Table 1 defines these levels and summarizes the players and capabilities of each level. TABLE 1 Summary of Pavement Management Levels Pavement Management Level Definition Key Capabilities Key Players Strategic Network Analyzes investments and fund allocation across all agency owned assets Analyzes the needs and funding requirements for a specific asset class within an agency Can show impact of funding options and help justify the need for funds. Communicate these needs and impacts to funding authorities Perform needs analysis to determine what is required and how much it will cost. Also performs impact analysis to determine the effect of limited funds Funding Authorities, Policy Makers, Senior Management Senior Management and Department Managers Project Selection Identifies constraints not considered at higher levels and refines possible alternatives in accordance with improved cost estimates Selects specific areas for funding and further analysis for project-level design. Department managers Project Most detailed level where planning is complete and detail design and construction occurs Able to consider local constraints and adapt to unforeseen issues at higher levels (usually field issues) Engineers and Technical Staff To most effectively manage a pavement network, levels must act in sync ensuring that decisions made and data used are consistent and support the agency s mission. This paper is particularly concerned with bridging the gap between the network and the project selection levels. A disconnect between the results of network-level analyses and actual project selection decisions can lead to unrealistic projects. Consequently, ad-hoc project selection approaches may prevail. While an ad-hoc approach to selecting pavement preservation projects can be practical; it leaves highway agencies vulnerable to scrutiny about how effectively money is being spent and makes defending project selection difficult. The goal is to formulate the current decision making process in a structured manner; which can help decision makers overcome the stumbling blocks associated with multi-criteria decision making. A structured decision making process allows highway agencies to 1) perform sensitivity analysis (e.g., quantify how and how much different variables should be included in the decision making process), 2) justify project prioritization decisions by explaining the decision making process, and 3) provide new engineers with a decision-support tool that mimics the decision making process within their organization. This paper develops an alternative method for developing realistic pavement projects based on a competition between pavement management sections (typically 0.5-mile in length).

4 Gurganus and Gharaibeh LITERATURE REVIEW AND BACKGROUND A key purpose of a network-level Pavement Management System (PMS) is to identify pavement sections that need improvement, the types of improvement (preservation, major rehabilitation, or reconstruction), and the timing of the improvement, and to prioritize preservation and renewal projects when funds are limited. These tasks are normally accomplished through two types of analysis (1-3): Needs analysis (no budget constraints): This analysis identifies preservation needs and amount of funds needed to complete these projects. Impact analysis: This analysis provides answers to what-if questions regarding the effect of funding on the network condition. Several analytical techniques exist to prioritize pavement preservation projects, including (2, 4-10): Optimization (e.g., dynamic programming, integer programming, genetic algorithms) Incremental benefit-cost (IBC) analysis Ranking based on various parameters, such as pavement condition User-defined heuristic rules Final project selection is normally made through negotiations among stake holders (e.g., engineers, managers, etc.), considering various engineering, socioeconomic, political, and practical factors, along with results from network-level analyses. The fact that pavement project prioritization decisions involve multiple objectives and constraints has been recognized in the literature (11,12). The method presented in this paper offers a different possibility for project selection and prioritization. The method is founded on the Analytic Hierarchy Process (AHP); where pavement sections compete based on key decision parameters ultimately leading to a prioritized list of candidate preservation projects. The AHP is a multi-criteria decision making method that finds its origins in the early 1970s as its creator was working on contingency planning for the Department of Defense (13). Its development stemmed from the need to organize and make decisions dealing with unstructured problems. Not only were the problems unstructured, but the components within these problems had no or several different units of measure. The creator of the method sought to overcome these issues through the creation of a hierarchy. The elements on the levels of the hierarchy could be placed in a pairwise matrix where each element could be compared against the other element. The goal is to mimic how people actually think and decide (14). Comparing components to one another creates a matrix based on a ratio scale that utilizes matrix calculations to arrive at weights for the competing decision criteria (15). A study by Sun and Gu, (16), has been performed that focuses on pavement condition and the prioritization of pavement projects. This paper focuses on different parameters included in pavement condition and seeks to make the optimum decision by appropriately aggregating this parameter with the AHP. Sun and Gu focus on the prioritization of eight roadway sections while mentioning that an actual pavement network will contain thousands of segments (16).

5 Gurganus and Gharaibeh Overview of Project Selection Practices at State DOTs Most agencies tasked with managing pavements have tools that utilize data stored describing the system (condition, inventory, traffic, etc.) to generate preservation suggestions. The Texas Department of Transportation (TxDOT) utilizes a decision tree approach called the Needs Estimate. This approach uses information stored in TxDOT s Pavement Management Information System (PMIS) to generate preservation suggestions. TxDOT intends for PMIS to work both as a network level tool used by policy makers and as a project selection level tool used by TxDOT districts (17). Montana, Nebraska, Nevada, New Hampshire, and Virginia also use decision trees to generate preservation suggestions. Other states use ratings and associated descriptions of rating levels to provide a general idea of what type of preservation action is required. Some are specific in terms of measure, such as Alabama which suggests an overlay at a score of 55 on its scale. Others use broader repair categories. For example, Illinois uses the Condition Rating Survey (CRS) and suggests major rehabilitation on a pavement with a score between one and 4.5 (on a 1-9 scale). West Virginia generalizes the need for rehabilitation when a pavement reaches a 2.5 on its 1-5 rating scale, with five indicating excellent condition. The components associated with each DOT s measuring scale contribute to how the decision is made, however little information is available on moving from these general preservation categories viewed from a network level perspective to actual project selections (18). It is not uncommon for these decision support tools to produce recommendations that do not match actual decisions. This has been noticed by some DOTs. Arizona overhauled its pavement management system to eliminate Markovian chain modeling that decision makers did not feel accurately represented project level decisions (19). North Carolina Department of Transportation (NCDOT) attempted to use data mining and knowledge discovery techniques to uncover hidden information within the PMS that would improve decision making, however there remained discrepancies between new suggestions and what decision makers would have implemented (20). ANALYTIC HIERARCHY PROCESS FOR PROJECT SELECTION A stumbling block in making systematic preservation decisions is the multiple variables that must be considered when making decisions about pavements. A decision about which preservation projects to pursue is not as simple as selecting the roadway with the worst distress. Other factors such as traffic volume, number of trucks, location, number of lanes, regional development, and even political pressure play a part in where preservation work is conducted. In fact, this decision making process can vary within the same state. For example, the size, location, climate, amount of development, and land type variability across Texas forces each district within TxDOT to consider different variables in different ways. This paper investigates the use of the AHP as a way to deal with the multiple criteria considered in pavement project selection decisions. The AHP is constructed on a unique importance rating scale specifically designed to deal with multi-criteria decision making. This scale ranges from 1 to 9. The odd numbers represent the primary importance intensity values, while the even numbers represent intermediate importance intensity values. This is illustrated in Table 2 (13). The scale is then used to compare the input parameters in a pairwise fashion to determine how much more or less important one parameter is compared to the other. The parameters are

6 Gurganus and Gharaibeh not compared to the decision as a whole; rather they are compared with each other to decide how they compete for importance in the decision as a whole. The pairwise comparison builds an nxn matrix consisting of the number of parameters included in the decision. The weights associated with the parameters are developed by calculating the principal eigenvector associated with the maximum eigenvalue for the matrix. This principal eigenvector is normalized to create a relative ratio scale that can be used as the priority vector, or simply, weights associated with each parameter (13). TABLE 2 AHP Weighting Scheme Weight of Importance Definition (13 ) Explanation (13 ) 1 Equal Importance 3 Moderate importance of one over another 5 Essential or strong importance 7 Very strong importance 9 Extreme importance Two activities contribute equally to the objective Experience and judgment strongly favor one activity over another Experience and judgment strongly favor one activity over another An activity is strongly favored and its dominance demonstrated in practice The evidence favoring one activity over another is of the highest order of affirmation 2, 4, 6, 8 Intermediate values between the two adjacent judgments When compromise is needed Reciprocals If activity i has one of the above numbers assigned to it when compared with activity j, then j has the reciprocal value when compared with i A consistency ratio (CR) is used to assess the consistency of the decision maker in assigning the importance intensity values. The CR is computed as follows: CI CR RI...1 max n CI n 1...2

7 Gurganus and Gharaibeh where n size of matrix max maximum eigenvalue The RI value in the above equation is termed the Random Consistency Index and is predefined in the AHP literature based on the matrix size. The AHP method suggests that CR should be less than 10%, implying the method will allow for up to 10% error in human judgment during the pairwise comparison phase (14). To bridge the gap between network-level PMSs and the project selection process, the parameters associated with making pavement preservation decisions must be identified first; then a method must be developed that can use data available in network-level PMSs to determine and unite these decision parameters to help select candidate projects. Questions such as the following must be answered: is a section with 5000 vehicles per day, one localized failure, 10% Alligator Cracking, and an International Roughness Index (IRI) of 127 inch/mile in more need of preservation work than a section with 10,000 vehicles per day, 125 feet of longitudinal cracking, and an IRI of 148 inch/mile? This must be done for every section within the roadway network. Because this type of question must be answered for every section within the network, the easiest way to answer it is to have every section compete against every other section. This competition will result in winners and losers or more accurately a prioritization list of sections requiring improvement. The AHP requires formulating the problem in a hierarchal fashion. In regards to a decision, there is an ultimate goal with different components contributing to that goal. The hierarchy of the project selection decision will be stratified in three levels, as follows: 1. Project Selection Number: At this level, pavement managers can evaluate each section within the network to determine how important one section is compared to another section, considering the relative weights of each decision parameter. To facilitate this process, decision makers preferences can be formulated in logic statements that can be imbedded into computer programs. 2. Decision Parameter Level: At this level, pavement managers can set the relative weights of the various parameters considered in the decision making process. The AHP is applied to the parameters to determine the weights. This level allows managers to determine how sections rank when considering each individual parameter. 3. Section vs. Section Level: At this level, each section competes against every other section to determine how important it is against every other section for every decision parameter. This hierarchy is illustrated in Figure 1.

8 Gurganus and Gharaibeh FIGURE 1 Project selection number hierarchy As mentioned earlier, some network-level PMSs store data at the section level, typically 0.5-mile long. However, a realistic preservation project extends across several miles. To completely bridge the gap between network-level and project-level decision making, these newly established project selection numbers must be aggregated to move from the section level to the project level, more accurately reflecting how decisions are made. The application of this process to network-level pavement management data from Texas is discussed in the remaining parts of this paper. APPLICATION OF PROJECT SECTION PROCESS TO TXDOT DATA The method is applied to preservation decisions within one of TxDOT s 25 districts. Based on interviews with TxDOT decision makers, it is clear that the available network analysis tool for preservation recommendations experiences little use at the district level. It is at the district level where the network and project-level interact in the decision making process. Based on the results of interviews with decision makers from four TxDOT districts, the following six parameters were selected for use in the creation of the project selection number: 1. Visual Distress 2. Current Average Daily Traffic (ADT) 3. Current Truck ADT 4. Condition Score (CS) 5. Ride Quality 6. Section that receive the most routine (in-house) maintenance Rather than merely use the previous year s distress data, concurrent research indicated that three years of distress data should be evaluated to more accurately describe the true condition of a pavement section (21). It is also important to note that Condition Score (CS) consists of a distress component and ride component, representing the overall condition. There is an agency-

9 Visual Distress Current ADT Current Truck ADT Condition Score Ride Quality Sections that receive most Maint. Max Eigenvector Priority Vector (Weigths) Gurganus and Gharaibeh wide pavement performance goal that was set based on CS. Therefore, it was important to consider CS in the developed method, despite the potential for double counting some decision factors. Relative Importance across Decision Parameters The parameters at level two of the hierarchy, in Figure 1, must be inserted into a 6x6 matrix. A primary decision maker from the case study area was interviewed to complete the pariwise comparisons. For purposes of simplifying the interview, only primary weights were used. The completed matrix is shown in Figure 2 along with the maximum eigenvector and its normalized counterpart that creates the priority vector. Visual Distress Current ADT 1/7 1 1/3 1/7 1/5 1/3 Current Truck ADT Condition Score 1/ /7 1/5 1/ Ride Quality 1/ /7 1 1 Sections that receive most Maint. 1/ / FIGURE 2 Project selection matrix with eigenvector and priority vector Completing this matrix would follow a thought process of moving down the parameters listed on the left and comparing them with the parameters listed across the top. For example Visual Distress is initially compared with Visual Distress, thus explaining the 1 located in the upper left box of the matrix. Visual Distress is then compared with Current ADT and the decision maker values Visual Distress as significantly more important than Current ADT, thus explaining the 7 in the second square along the top row. The reciprocal of this, 1/7, is placed in the second square in the first column where Current ADT is compared to Visual Distress and is significantly less important. These comparisons continue until the matrix is complete.

10 Gurganus and Gharaibeh The consistency calculations associated with the matrix in Figure 2 are illustrated in the following calculations: MaxEigenvalue max max n Consistency Index CI n Random Index RI 1.24( predfined fora 6x6 matrix ) CI Consistency Ratio CR 0.10 RI Relative Importance within Each Decision Parameter The previously described application of the AHP establishes how much a particular parameter contributes to the overall decision; it does not explain how varying degrees of presence of each parameter will affect the overall need of a section to receive preservation work. The AHP is applied to determine when and how much more or less important one section is compared to another. For example, when does a section become more important in terms of traffic volume? Is a section with 1,500 vehicles per day (vpd) equally as important as a section with 4,000 vpd, and how much more important is a section with 15,000 vpd compared with a section that has only 750 vpd? The same questions arise for each one of the decision parameters and represent the third level of the hierarchy in Figure 1. To make these comparisons, each section must be compared to every other section for each decision parameter, representing the competition approach. Normally the AHP is applied to a fairly small number of competing alternatives, say 15. A matrix of this size is easily completed in a short period of time, however to apply the method to a pavement network with thousands of sections, the number of comparisons would be large and completing the matrix through interviews is not feasible. Case in point, the pavement network used in the study includes roadways for three counties consisting of 2349 sections. These section-versus-section pairwise comparisons are made through the use of logic statements, resulting in a 2349 element vector. This vector is the ranking of each section in regard to that specific decision parameter. Ultimately, these vectors will have the decision parameter weights applied and carried through to create a 2349 element vector that is the project selection number or ranking of each section. By using logic statements to compare each section, there is no further need to test consistency. The AHP uses consistency calculations to ensure that human comparisons do not deviate to a point that makes the comparisons invalid, however by using logic statements coded in computational tools, consistency can be assumed. All of the section information regarding each decision parameter must be stratified so that logic statements can be written. These stratifications of the data are illustrated in Table 3.

11 Gurganus and Gharaibeh TABLE 3 AHP Weight Associated with Minimum Comparison AHP Weight Visual Distress (DN) Current ADT (veh/day) Current FM Truck ADT (trucks/day Current Non-FM Truck ADT (trucks/day) 1 DN = veh/day 1000 trucks/day 160 trucks/day < DN NA NA NA < DN < veh/day < trucks/day < trucks/day < DN NA NA NA < DN < veh/day < trucks/day < trucks/day < DN NA NA NA < DN < veh/day 10, < trucks/day < trucks/day < DN 1.45 NA NA NA < DN 10,000 < veh/day 640 < trucks/day 4900 < trucks/day AHP Weight Condition Score (CS) FM Ride Quality (IRI) Non-FM Ride Qualtiy (IRI) Maintenance Cost ($) 1 90 to to to 59 Cost = $0 2 NA NA NA $0 < Cost $ to to to 119 $6000 < Cost $12,000 4 NA NA NA $12,000 < Cost $18, to to to 170 $18,000 < Cost $24,000 6 NA NA NA $24,000 < Cost $30, to to to 220 $30,000 < Cost $36,000 8 NA NA NA $36,000 < Cost $42, to to to 950 $42,000 < Cost By using the term minimum comparison in Table 3, it is meant that this sets the initial AHP weight of a section. For example, if a section has a CS of 40 and is compared to a section with a CS of 100, the section with a 40 would receive an AHP weight of seven. What is not displayed in Table 3 is the weight assigned when a section with a CS of 40 is compared to a section with a CS of 75. This is done with a logic statement by merely subtracting the two sections weights when each is compared to a minimum, thus receiving the weight in Table 3. It is important to note that one must be added to the result of this subtraction to preserve the fact that the AHP begins at one rather than zero. The visual distress number from Table 3 is a unique parameter in the sense the weights had to be applied to the various distresses considered at the district level and each section must compete against every other section in regards to how important more or less distress was and how did that importance change as multiple distresses were evaluated at the same time. The same hierarchical approach was used as with the project selection number and the AHP was applied to determine a distress number that could be used as the visual distress component. Relative Importance across Distresses The need to include an accurate portrayal of distresses is obvious, but how to frame the various distresses and respective densities as matter of importance is not. To determine how the multiple distresses should be aggregated, it had to be determined how important one distress is compared to another distress. To create a uniform method, the best way to determine importance is to apply the AHP to distresses in the same way that it was applied to decision parameters, using the decision maker s feedback. Calculation of a distress number was performed in a hierarchical

12 Gurganus and Gharaibeh way as represented in Figure 3. The pairwise comparison took place at level two of the hierarchy and resulted in a priority vector illustrated in Table FIGURE 3 Distress hierarchy A matrix with each component at Level 2 in Figure 3 was completed and the eigen calculations and consistency calculations were performed to arrive at the importance ranks illustrated in Table 4. Because this exercise is similar to that performed with Figure 2, the details were omitted for brevity.

13 Gurganus and Gharaibeh TABLE 4 Maximum Eigenvector and Priority Vector for Distress Types Distress Type Max Eigenvector Priority Vector Failures Deep Rutting Block Cracking Alligator Cracking Longitudinal Cracking Transverse Cracking Patching Sum The development of these weights will eventually help to answer not only the question as to whether or not Failures are more important than Alligator Cracking (for instance), but is a section with two Failures and 50 percent Alligator Cracking more important than a section with 25 percent Alligator Cracking and 15 percent Patching? However, to fully answer questions such as these each section must compete against every other section in regards to each distress type illustrated at level two of Figure 3. In summary, the AHP must be applied to Failures, Deep Rutting, Alligator Cracking, and so on just as it was applied to the six decision parameters. The distress information must be divided in such a way that logic statements can be constructed to perform the 2349 x 2349 comparison. Relative Importance within Each Distresses Type Table 5 shows the AHP weight assignment to various distress densities.

14 Gurganus and Gharaibeh AHP Weight TABLE 5 AHP Distress Weight Associated with Minimum Comparison Failures (EA) Deep Rutting (%) Alligator Cracking (%) Transverse Longitudinal Cracking (ft) Cracking (EA) Patching (%) 1 0 0% to 4% 0% to 2% 0' to 25' 0 to 2 0% Patch 3% 2 1 5% and 6% 3% 26' to 50' 3 3% < Patch 7% 3 2 7% 4% 51' to 75' 4 NA 4 NA 8% 5% NA 5 7% < Patch 11% 5 NA 9% 6% 76' to 100' 6 11% < Patch 15% 6 NA 10% 7% 101' to 125' NA 15% Patch 22% % 8% 126' to 150' 7 22% < Patch 35% % 9% and 10% 151' to 175' 8 35% < Patch 44% % 11% 176' 9 44% < Patch Again, the AHP weight in Table 5 is associated with comparing a section to the minimum, or least important, section. After the establishment of this minimum comparison, the sections can be compared using the AHP weights to determine importance. Common sense suggests that as distress density increases, the importance to perform preservation work on the section also increases. In short, a section that has four Failures is more important than a section with only two Failures, but how much more? A system was established to define the upper limit where the AHP weight of nine can be assigned. The upper limit is defined based on actual distress density within the study area. The assumption was made that there is an unacceptable level of all distresses that the district simply will not allow a section of roadway to reach without performing some type of maintenance. Any distress above this point could be considered drastically more important than a section with none of that particular distress manifested. The study sets this upper limit as the point at which 98 percent of distress density was at or below based on three years of distress data. To fill in the rest of the gaps regarding AHP weight and distress density, PMIS currently contains curves that describe how distresses affect the Distress Score for a section of pavement. To use PMIS s utility curves as a starting point, weights can be assigned to the increase in distress density in such a way that a curve can be created for the AHP method that matches the shape of the utility curve. By doing this, the method follows the current thought used by TxDOT to determine how distresses affect a section of pavement. Using these utility curves, logic statements were developed that allowed for the application of the AHP. The method created a ranking of importance independently for each section in regard to all distress types. The combination of these weight vectors with the weights developed from the application of the AHP at level two of the hierarchy (see Figure 3) and the aggregation of all components together results in the creation of the distress number to be used to finalize the creation of the project selection number. CREATION OF DISTRESS NUMBER AND PROJECT SELECTION NUMBER The next step in the process is to take the results from the competitions performed by each section for all distresses and decision criteria and begin aggregating information to create a final index that can be used to prioritize pavement sections.

15 Gurganus and Gharaibeh With a priority vector for each of the distresses shown at level two of the hierarchy illustrated in Figure 3, a distress number for each of the 2349 sections can be created. The calculation of the distress number follows the calculations below. k n d n d 1 DN w * pz...7 where DN n the distress number for any section within the evaluation network d distress types considered in the decision k=7 in the current study w d pz weight associated with a particular distress n priority number associated with any section within the evaluation network This distress number will become the visual distress parameter used in the application of the AHP to the decision parameters. The final step is to apply the weights for each decision parameter to each of the priority vectors and sum across to create the project selection number for each section within the network. This calculation is represented with the following equation: k n p n p 1 PN w * pz...8 where, PN n the project selection number for any section within the evaluation network p decision parameter considered in the decision k=6 in the current study w p pzn weight associated with a particular parameter priority number associated with any section within the evaluation network The AHP has provided the basis for developing a project selection number that accounts for many variables and does so in a way that assigns realistic weights to the decision parameters in the same way that a particular network manager views these components. However, to truly test the project selection number as a viable decision support tool, preservation projects must be selected and evaluated against actual projects. There must be a move beyond 0.5-mile section information and into the actual selection of projects involving multiple sections, further bridging the gap between network and project-level pavement management. RESULTS OF APPLICATION TO CASE STUDY AREA By utilizing the AHP, decision parameters that were originally on different measurement units were now accounted for in the project selection number and different measurement units were no longer an issue. With this in mind, the project selection number from an adjacent section can be added to another section and that number can be evaluated against other summations of pavement sections. The case study district imposes a two mile project minimum, therefore enough project selection numbers were added together to create realistic projects. To perform this operation, a logic statement was written to drive through the sections and sum project

16 Gurganus and Gharaibeh selection numbers. Project selection numbers were sorted with the largest indicating the most important pavement preservation project. Table 6 lists the top 20 projects created for the case study area along with the centerline miles for each project. While selecting realistic preservation projects by considering various parameters is a positive aspect of the new method, in order to accomplish the stated goal of this study and create a decision support method that might actually be used at the district level, the agreement between the projects selected by the method and actual preservation expenditures must be high. Of the 20 projects in Table 6, the case study district has made decisions to provide a known preservation action on 15. These preservation actions include a construction project performed by a private contractor, a routine maintenance project performed by a private contractor, or the use of TxDOT forces to perform the preservation action. The five projects with unknown actions do not necessarily mean that the district is not working on those sections or is not preparing a project for those sections; it only means the information was not available to make any definite determination.

17 Gurganus and Gharaibeh Table 6 Projects Selected Compared with Actual District Decisions Rank HWY Begin Ref. Marker End Ref. Marker Length (mi) District Action 1 FM FM Grading, Structure, Base, Surface Project Let in FY 2009 Grading, Structure, Base, Surface Project Let in FY FM Restoration Project Let in FY IH 45 A Contracted Rehab 5 FM Restoration Project Let in FY FM Routine Maintenance Contract 7 FM Routine Maintenance Contract 8 FM Routine Maintenance Contract 9 FM In House Maintenance Forces 10 FM In House Maintenance Forces 11 FM Routine Maintenance Contract (FY 2007), then Seal Coat FY SH 75 K Unknown 13 US 190 K Unknown 14 FM Grading, Structure, Base, Surface Project Let in FY FM Routine Maintenance Contract 16 FM Unknown 17 FM Routine Maintenance Contract 18 FM FM Unknown In House Maintenance with Rehab let in Nov FM Unknown Overall, the method created matches at least 75 percent of district decisions regarding pavement preservation decisions in the case study area.

18 Gurganus and Gharaibeh SUMMARY AND CONCLUSIONS This paper describes a new decision support method for pavement preservation project selection that accounts for both quantitative and qualitative variables considered by pavement managers. This method is founded on the AHP and uses data from a network-level PMS along with inputs from decision makers. The AHP was used in an attempt to select and prioritize projects in a way that mimicked how decision makers currently operate. The method was applied to data obtained from TxDOT s PMS. Decision parameters included visual distress, current ADT, current truck ADT, Condition Score, ride quality, and maintenance costs. The visual distress parameter was created by applying the AHP to determine how different distress types should be weighted in the decision making process. The distress types considered were localized failures, alligator cracking, longitudinal cracking, block cracking, deep rutting, transverse cracking, and patching. The method allowed for combining 0.5-mile pavement sections to create realistic preservation projects that closely matched preservation decisions currently being made by a TxDOT district. Further development of the method should include treatment cost as an additional decision factor. This improvement will not only help prioritize where work should occur, but will also enable agencies to perform a benefit to cost analysis on possible treatments. In addition to this improvement, further research is needed to incorporate a wider group of decision makers into the method. These techniques allow for considering input from multiple decision makers; which reflects real-world decision situations more accurately. REFERENCES 1. American Association of State Highway and Transportation Officials (AASHTO). AASHTO Guidelines for Pavement Management Systems. Washington, DC, Haas, R., W. R. Hudson, and J. P. Zaniewski. Modern Pavement Management. Krieger Pub Co, Malabar, FL, Smith, R. E. Integrating Pavement Preservation into a Local Agency Pavement Management System. In Transportation Research Record: Journal of the Transportation Research Board, Vol. 1795, 2002, pp Šelih, J., A. Kne, A. Srdić, and M. Žura. Multiple-Criteria Decision Support System in Highway Infrastructure Management. Transport, Vol. 23, No. 4, 2008, pp Guerre, J. A., and J. Evans. Applying System-Level Performance Measures and Targets in the Detroit, Michigan, Metropolitan Planning Process. In Transportation Research Record: Journal of the Transportation Research Board, Vol. 2119, 2009, pp Nuworsoo, C., K. Parks, and E. Deakin. Cost Per User as Key Factor in Project Prioritization: Case Study of San Francisco Bay Area, California. In Transportation Research Record: Journal of the Transportation Research Board, Vol. 1986, 2006, pp Madanat, S. Optimal Infrastructure Management Decisions Under Uncertainty. Transportation Research Part C: Emerging Technologies, Vol. 1, No. 1, 1993, pp Papagiannakis, T., and M. Delwar. Computer Model for Life-Cycle Cost Analysis of Roadway Pavements. Journal of Computing in Civil Engineering, Vol. 15, 2001, pp Smith, R. E., and K. M. Fallaha. Developing an Interface between Network-and Project- Level Pavement Management Systems for Local Agencies. In Transportation Research Record, Vol. 1344, 1992, pp

19 Gurganus and Gharaibeh Straehl, S. S., and L. Schintler. Montana Secondary Program Reform and Application of Goals Achievement Methodology to Project Prioritization. In Transportation Research Record: Journal of the Transportation Research Board, Vol. 1895, 2004, pp Wu, Z., and G. W. Flintsch. Pavement Preservation Optimization Considering Multiple Objectives and Budget Variability. Journal of Transportation Engineering, Vol. 135, No. 5, 2009, pp Fwa, T. F., W. T. Chan, and K. Z. Hoque. Multiobjective Optimization for Pavement Maintenance Programming. Journal of Transportation Engineering, Vol. 126, No. 5, 2000, pp Saaty, T. L. The Analytic Hierarchy Process. McGraw-Hill Profession Publishing, New York, NY, Saaty, T.L. Multicriteria Decision Making: The Analytic Hierarchy Process: Planning, Priority Setting, Resource Allocation. RWS publications, Pittsburgh, PA, Saaty, T.L. How to make a Decision: The Analytic Hierarchy Process. European Journal of Operational Research, Vol. 48, No. 1, 1990, pp Sun, L., and W. Gu. Pavement Condition Assessment using Fuzzy Logic Theory and Analytic Hierarchy Process. Journal of Transportation Engineering, Posted ahead of print, October 26, 2010, pp Texas Department of Transportation (TxDOT). Overview of PMIS Needs Estimate. Materials and Pavements Section of the Construction Division, Austin, TX, Papagiannakis, A., N. Gharaibeh, J. Weissmann, and A. Wimsatt. Pavement Scores Synthesis , Texas Transportation Institute The Texas A&M University System, Li, Y., A. Cheetham, S. Zaghloul, K. Helali, and W. Bekheet. Enhancement of Arizona Pavement Management System for Construction and Maintenance Activities. In Transportation Research Record: Journal of the Transportation Research Board, Vol. 1974, 2006, pp Zhou, G., L. Wang, D. Wang, and S. Reichle. Integration of GIS and Data Mining Technology to Enhance the Pavement Management Decision Making. Journal of Transportation Engineering, Vol. 136, 2010, pp Gurganus, C. F. Bridging the Gap between Network and Project Selection Levels in Pavement Management. Masters of Science Thesis, Texas A&M University, College Station, TX, 2011.