project selection and prioritization of pavement preservation

36

Transportation Research Record: Journal of the Transportation Research Board, No. 2292, Transportation Research Board of the National Academies, Washington, D.C., 2012, pp. 36–44.DOI: 10.3141/2292-05

C. F. Gurganus, Texas Department of Transportation, 205 Northeast Loop 564, Mineola, TX 75773, N. G. Gharaibeh, Zachry Department of Civil Engineering, Texas A&M University, 3136 TAMU, College Station, TX 77843. Corresponding author: C. F. Gurganus, [email protected].

A structured decision-making process allows highway agencies to (a) perform sensitivity analysis (i.e., quantify how, and how much, different variables should be included in the decision-making pro-cess), (b) justify project prioritization decisions through an explana-tion of the decision-making process, and (c) provide new engineers with a decision-support tool that mimics the decision-making process within their organization.

This paper develops an alternative method to develop realistic pavement projects on the basis of competition between pavement management sections (typically 0.5 mi in length).

Literature and Background

Key purposes of a network level pavement management system (PMS) are (a) to identify pavement sections that need improvement, the types of improvement (i.e., preservation, major rehabilitation, reconstruction), and the timing of improvement; and (b) to priori-tize preservation and renewal projects when funds are limited. These tasks normally are accomplished through two types of analysis (1–3):

• Needs analysis (i.e., no budget constraints), which identifies preservation needs and the amount of funds needed to meet them, and• Impact analysis, which answers what-if questions about the

effect of funding on the network condition.

Analytical techniques used to prioritize pavement preservation projects (2, 4–10) include the following:

• Optimization (e.g., dynamic programming, integer program-ming, genetic algorithms);• Incremental benefit–cost analysis;• Ranking on the basis of parameters, such as pavement condi-

tion; and• User-defined heuristic rules.

Final project selection normally is done among stakeholders (e.g., engineers, managers) through negotiation, during which engineering, socioeconomic, political, and practical factors are considered, along with results from network level analyses. The fact that pavement project prioritization decisions involve multiple objectives and con-straints has been recognized in the literature (11, 12). The method pre-sented in this paper offers a different possibility for project selection and prioritization. The method is founded on the analytic hierarchy process (AHP) in which pavement sections are judged on the basis of key decision parameters, which ultimately leads to a prioritized list of candidate preservation projects.

Project Selection and Prioritization of Pavement Preservationcompetitive approach

Charles F. Gurganus and Nasir G. Gharaibeh

Several methods help agencies select and prioritize pavement preserva-tion projects. Often these methods are built within an agency’s pavement management system. Unfortunately, these decision support tools often produce recommendations that do not match actual decisions, particu-larly for project selection of pavement management. Ad hoc selection procedures for preservation projects may be effective for many high-way agencies. Fiscal constraints and pressure from administrators and legislators, however, have forced agencies to justify their use of funds. This paper offers a new method for the selection and prioritization of pavement projects, with the use of the analytic hierarchy process as its multicriteria decision-making platform. The new method uses several parameters and input from decision makers to create a prioritized preser-vation project list. The method was applied in a case study in Texas; proj-ects suggested by the method matched actual decisions 75% of the time. The ability to capture multiple parameters and determine weights for each parameter on the basis of decision-maker input, along with the high level of agreement between the method and actual decisions, indicated that the method could be a viable decision support tool.

Pavement management decisions are made at four levels: strategic, network, project selection, and project. Table 1 defines these levels and the players and capabilities of each.

To manage a pavement network most effectively, levels must act in sync to ensure that decisions made and data used are consistent and support an agency’s mission. This study was particularly con-cerned 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 unrealis-tic projects. Consequently, ad hoc project selection approaches may prevail. Although an ad hoc approach to the selection of pavement preservation projects can be practical, such an approach leaves high-way agencies vulnerable to scrutiny about how effectively money is spent and makes the defense of 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 multicriteria decision making.

Gurganus and Gharaibeh 37

AHP is a multicriteria decision-making method with its origins in the early 1970s when its creator was at work on contingency planning for the Department of Defense (13). AHP’s development stemmed from the need to organize and make decisions to deal with unstruc-tured problems. Not only were the problems unstructured but the components within these problems had no, or several, units of mea-sure. The creator of the method sought to overcome these issues through a hierarchy. The elements on the levels of the hierarchy could be placed in a pairs matrix in which they could be compared against each other. The goal was to mimic how people actually think and decide (14). To compare components with one another created a matrix on the basis of a ratio scale that used matrix calculations to arrive at weights for the competing decision criteria (15).

Sun and Gu studied pavement condition and the prioritization of pavement projects (16). This paper focuses on different parameters included in pavement condition and seeks to make the optimum decision through appropriate aggregation of these parameters with AHP. Sun and Gu prioritized eight roadway sections, although they noted that an actual pavement network would contain thousands of segments (16).

Project SeLection PracticeS

Most agencies tasked with the management of pavement use tools whose stored data describe the system (e.g., condition, inventory, traffic) to generate preservation suggestions. The Texas Department of Transportation (DOT) uses a decision tree approach called the needs estimate. This approach uses information stored in the Texas DOT Pavement Management Information System (PMIS) to gener-ate preservation suggestions. The Texas DOT intends for PMIS to work both as a network level tool used by policy makers and as a project selection level tool used by Texas DOT 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 states are specific in terms of measure, such as Alabama, which suggests an overlay at a score of 55 on its 0 to 100 scale. Others use broader repair categories. For example, Illinois uses the Condition Rating Survey and suggests major rehabilitation on a pavement with a score between 1 and 4.5 (on a 1 to 9 scale). West Virginia gen-

eralizes the need for rehabilitation when a pavement reaches a 2.5 on its 1 to 5 rating scale, on which 5 indicates excellent condition. The components associated with each DOT’s measuring scale con-tribute to how a decision is made. Little information is available, however, on movement from these general preservation categories from a network level perspective to actual project selection (18).

It is not uncommon for these decision support tools to produce recommendations that do not match actual decisions. This dis-connect has been noticed by some DOTs. Arizona overhauled its PMS to eliminate Markovian chain modeling, which decision mak-ers believed did not accurately represent project level decisions (19). The North Carolina DOT attempted to use data mining and knowledge discovery techniques to uncover hidden information within the PMS that would improve decision making. Discrepancies remained, however, between new suggestions and what decision makers would have implemented (20).

aHP for Project SeLection

The multiple variables that must be considered form a stumbling block to decision making on the systematic preservation of pave-ment. To decide which projects to pursue is not as simple as it is to select the roadway with the worst distress. Other factors (e.g., traffic volume, number of trucks, location, number of lanes, regional development, even political pressure) play a part in the decision as to where preservation work should be conducted. In fact, this decision-making process can vary within the same state. The size, location, climate, amount of development, and land type variability across Texas, for example, forces each district within the state’s DOT to consider different variables in different ways. This paper investigates the use of AHP as a way to deal with the multiple criteria considered in pavement project selection decisions.

AHP is constructed on a unique importance rating scale spe-cifically designed to deal with multicriteria decision making. This scale ranges from 1 to 9. The odd numbers represent the primary importance intensity values, whereas the even numbers represent intermediate importance intensity values (Table 2) (13).

The scale is used to compare the input parameters by pairings to determine how much more or less important one parameter is than the other. The parameters are not compared against the decision as a whole. Rather, they are compared with each other to

TABLE 1 Summary of Pavement Management Levels

Pavement Management Level Definition Key Capabilities Key Players

Strategic Analyzes investments and fund allocation across all agency-owned assets

Can show impact of funding options and help justify need for funds

Funding authorities, policy makers, senior management

Communicates these needs and impacts to funding authorities

Network

Analyzes the needs and funding requirements for a specific asset class within an agency

Performs needs analysis to determine what is required and how much it will cost; also performs impact analysis to determine effect of limited funds

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

Indicates most detailed level where planning is complete and detail design and construction occur

Able to consider local constraints and adapt to unforeseen issues at higher levels (usually field issues)

Engineers and technical staff

38 Transportation Research Record 2292

decide how they compete for importance in the decision as a whole. The paired comparison builds an nxn matrix, which consists of the number of parameters included in the decision. The weights associ-ated with the parameters are developed through calculation of the principal eigenvector associated with the maximum eigenvalue for the matrix. This principal eigenvector is normalized to create a rela-tive ratio scale that can be used as the priority vector, or simply, weights associated with each parameter (13).

A ratio is used to assess the consistency of the decision maker in the assignment of the importance intensity values. The consistency ratio (CR) is computed as follows:

CRCI

RI= ( )1

CI =−( )

−( )λmax ( )

n

n 12

where

n = size of matrix,γmax = maximum eigenvalue, CI = consistency index, and RI = random consistency index.

The RI value in the above equation is predefined in AHP literature on the basis of matrix size. The AHP method suggests that the consis-tency ratio should be less than 10%, which implies that the method will allow up to 10% error in human judgment during the paired comparison phase (14).

To bridge the gap between network level PMSs and the project selection process, the parameters associated with pavement pres-ervation 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 can-didate projects. Questions such as the following must be answered: Is a pavement section used by 5,000 vehicles per day (vpd), with one localized failure, 10% alligator cracking, and an international roughness index of 127 in. per mile more in need of preservation work than a section used by 10,000 vpd, with 125 ft of longitudinal cracking and a roughness index of 148 in. per mile? Because this type of question must be answered for every section within the net-work, 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 that require improvement.

AHP requires formulation of the problem in a hierarchical fash-ion. With respect to a decision, there is an ultimate goal with dif-ferent components that contribute to it. The hierarchy of the project selection decision is stratified in three levels as follows:

1. Project selection number. Pavement managers can evaluate each section within the network to determine how important one section is compared with another, and consider the relative weights of each decision parameter. To facilitate this process, decision-maker preferences can be formulated in logic statements, which can be imbedded into computer programs.

2. Decision parameter level. Pavement managers can set the rela-tive weights of the various parameters considered in the decision-making process. AHP is applied to the parameters to determine the weights. This level allows managers to determine how sections rank when each parameter is considered.

3. Section versus section level. For every decision parameter, each section competes against every other one to determine how important it is by comparison.

This hierarchy is illustrated in Figure 1.As mentioned earlier, some network level PMSs store data at the

section level, which typically is 0.5 mi long. A realistic preserva-tion project, however, extends across several miles. To bridge the gap completely 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 to reflect more accurately how decisions are made. Application of this process to network level pavement management data from Texas is discussed in the remainder of this paper.

aPPLication of Project Section ProceSS

The method was applied to preservation decisions within one of the 25 Texas DOT districts. On the basis of interviews with Texas DOT decision makers, it was clear that the available network analysis tool for preservation recommendations was little used at the district level. It is at the district level that the network and project levels interact in the decision-making process.

Through interviews with decision makers from four Texas DOT districts, 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),

TABLE 2 AHP Weighting Scheme

Weight of Importance Definition (13) Explanation (13)

1 Equal importance Two activities contribute equally to the objective.

3

Moderate importance of one over another

Experience and judgment strongly favor one activity over another.

5

Essential or strong importance

Experience and judgment strongly favor one activity over another.

7

Very strong importance

An activity is strongly favored and its dominance is demonstrated in practice.

9

Extreme importance

The evidence favoring one activity over another is of the highest order of affirmation.

2, 4, 6, 8

Intermediate values between two adjacent judgments

Intermediate values are used 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

Reciprocal values follow the same explanation in terms of importance as the initial comparison.


5. Ride quality, and6. Section that receives most routine (in-house) maintenance.

Concurrent research indicated that 3 years of distress data should be evaluated to describe accurately the true condition of a pavement section (21). CS consists of a distress component and a ride com-ponent, which represent overall condition. An agencywide pave-ment performance goal was set on the basis of CS. Therefore, it was important to consider CS in the developed method, despite the risk of a double count of some decision factors.

relative importance across decision Parameters

The parameters at Level 2 of the hierarchy (Figure 1) had to be inserted into a 6 × 6 matrix. A primary decision maker from the case study area was interviewed to complete the paired comparisons. To simplify the interview, only primary weights were used.

The completed matrix is shown in Figure 2, along with the maxi-mum eigenvector and its normalized counterpart, which create the priority vector.

To complete this matrix entails a thought process in which the decision maker moves down a list of parameters on the left and

Vis

ual

Dis

tres

s

Cu

rren

t A

DT

Cu

rren

t T

ruck

AD

T

Co

nd

itio

n S

core

Rid

e Q

ual

ity

Sec

tio

ns

that

rec

eive

mo

st M

ain

t.

Max

Eig

enve

cto

r

Pri

ori

ty V

ecto

r(W

eig

hts

)

VisualDistress 1

0.6711 0.3660

Current ADT 1/70.0546 0.0298

Current TruckADT 1/5

0.0854 0.0466

ConditionScore 1

0.6968 0.3801

Ride Quality 1/70.1839 0.1003

Sections thatreceive mostMaint.

1/7

7

1

3

7

5

3

5

1/3

1

7

5

3

1

1/7

1/7

1

1/7

1/7

7

1/5

1/5

7

1

1

7

1/3

1/3

7

1

1

0.1417 0.0773

FIGURE 2 Project selection matrix with eigenvector and priority vector (maint. = maintenance; max = maximum).

FIGURE 1 Project selection number hierarchy (RM = routine maintenance).

VisualDistress

Section 1 Section 2 Section 3 Section n-1 Section n

CurrentADT

CurrentTruckADT

Project SelectionNumber

ConditionScore

RideQuality

Sectionsreceivingmost RM

Level 1: Overall Goal(Develop ProjectSelection Numberbased on inputs)

Level 2: Criteria used todetermine the overallgoal (Pairwise comparedto determine weights ofeach criterion)

Level 3: Candidate Sections (Each sectionwithin the evaluationnetwork)


compares them with the parameters listed across the top. For exam-ple, visual distress initially is compared with visual distress, and thus explains the 1 located in the upper left box of the matrix. Visual dis-tress then is compared with current ADT and decision-maker values. Visual distress is significantly more important than current ADT, and thus explains the 7 in the second square along the top row. The reciprocal 1/7 is placed in the second square in the first column, where current ADT is compared with visual distress and is signifi-cantly less important. These comparisons continue until the matrix is complete.

The consistency calculations associated with the matrix in Fig-ure 2 are illustrated in the following calculations:

maxeigenvalue = =λmax . ( )6 6399 3

consistency index CI= = −−

= −−

=λmax .n

n 1

6 6399 1

6 100 91574 4. ( )

random index RI predefined for a 6 matr= = ×1 24 6. iix( ) ( )5

consistency ratio CRCI

RI= = = =0 91574

1 240 10

.

.. (66)

relative importance Within each decision Parameter

The previously described application of 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. AHP is applied to determine when and how much more or less important one section is than another. For example, when does a sec-tion become more important in terms of traffic volume? Is a section with 1,500 vpd equally as important as a section with 4,000 vpd? And how much more important is a section with 15,000 vpd than a section with 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.

For each decision parameter, each section must be compared with every other one in a representation of the competition approach. Nor-mally, AHP is applied to a fairly small number of competing alterna-tives, say, 15. A matrix of this size easily is completed in a short period of time. To apply the method, however, to a pavement network with thousands of sections, the number of comparisons would be large, and completion of the matrix through interviews would not be fea-sible. Case in point, the pavement network used in the study included roadways for three counties and consisted of 2,349 sections. These section-by-section paired comparisons were made through the use of logic statements, which resulted in a 2,349 element vector. This vector is the ranking of each section with respect to the specific decision parameter. Ultimately, these vectors will have the decision parameter weights applied and carried through to create a 2,349 ele-ment vector that is the project selection number or ranking of each section. Use of logic statements to compare each section eliminates further need to test consistency. AHP uses consistency calculations to ensure that human comparisons do not deviate to a point that makes the comparisons invalid. Through use of logic statements coded in computational tools, however, consistency can be assumed.

All of the section information on each decision parameter must be stratified so that logic statements can be written. These strati-fications of the data are illustrated in Table 3. Through use of the term “minimum comparison” in Table 3, it is meant that the initial AHP weight of a section is set. For example, if a section with a CS of 40 was compared with one with a CS of 100, the section with a 40 would receive an AHP weight of 7. Not displayed in Table 3 is the weight assigned when a section with a CS of 40 was compared with a section with a CS of 75. This assignment is done with a logic statement through subtraction of the two sections’ weights when each is compared with a minimum, and thus the weight in Table 3 is received. (A 1 must be added to the result of this subtraction to preserve the fact that AHP begins at 1 rather than at 0.)

The visual distress number from Table 3 is a unique parameter, in that the weights had to be applied to the various distresses con-sidered at the district level, and each section had to compete against every other section as to how important more or less distress was and how that importance changed as multiple distresses were evalu-ated at the same time. The hierarchical approach was the same one used with the project selection number, and AHP was applied to determine a distress number that could be used as the visual distress component.

TABLE 3 AHP Weight Associated with Minimum Comparison

AHP Weight Visual Distress Current ADT (vpd)Current FM Truck ADT (trucks/day)

Current Non-FM Truck ADT (trucks/day) Condition Score FM Ride Quality (IRI)a Non-FM Ride Quality (IRI)a Maintenance Cost ($)

1 0.2629 ≤1,000 ≤160 ≤1,225 90 to 100 1 to 119 1 to 59 0

2 0.2629 < DN ≤ 0.433 na na na na na na 0 < cost ≤ 6,000

3 0.433 < DN ≤ 0.603 1,000 < vpd ≤ 2,000 160 < trucks/day ≤ 320 1,225 < trucks/day ≤ 2,450 70 to 89 120 to 154 60 to 119 6,000 < cost ≤ 12,000

4 0.603 < DN ≤ 0.773 na na na na na na 12,000 < cost ≤ 18,000





9 >1.45 >10,000 >640 >4,900 1 to 34 221 to 950 221 to 950 >42,000

Note: na = not applicable; FM = farm-to-market road; DN = visual distress.aIRI = international roughness index score.


relative importance across distresses

The need to include an accurate portrayal of distress is obvious but how to frame the various distresses and their respective densities as matters of importance is not. To determine how the multiple dis-tresses should be aggregated, it had to be determined how impor-tant one distress was, compared with another. To create a uniform method, the best way to determine importance was to apply AHP to distresses in the same way that it was applied to decision parameters, with the use of decision-maker feedback. Calculation of a distress number was performed in a hierarchical way (Figure 3). The paired comparison took place at Level 2 of the hierarchy and resulted in a priority vector (Table 4).

A matrix with each component at Level 2 in Figure 3 was com-pleted and the eigen calculations and consistency calculations were performed to arrive at the importance ranks illustrated in Table 4. Because this exercise was similar to that performed with Figure 2, the details were omitted for brevity.

Development of these weights eventually will help to answer not only whether or not failures are more important than alligator

cracking (for instance) but also whether a section with two fail-ures and 50% alligator cracking is more important than a section with 25% alligator cracking and 15% patching. To answer such questions fully, however, each section must compete against every other one with respect to each distress type illustrated at Level 2 in Figure 3.

In summary, AHP must be applied to failures, deep rutting, alliga-tor cracking, and so on just as AHP 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 2,349 by 2,349 comparison.

relative importance Within each distress type

Table 5 shows the AHP weight assignment to various distress densi-ties. Again, the AHP weight in Table 5 was associated with a com-parison of a section to the minimum, or least important, section. After establishment of this minimum comparison, the sections could be compared with the use of AHP weights to determine importance.

TABLE 3 AHP Weight Associated with Minimum Comparison

AHP Weight Visual Distress Current ADT (vpd)Current FM Truck ADT (trucks/day)

Current Non-FM Truck ADT (trucks/day) Condition Score FM Ride Quality (IRI)a Non-FM Ride Quality (IRI)a Maintenance Cost ($)

1 0.2629 ≤1,000 ≤160 ≤1,225 90 to 100 1 to 119 1 to 59 0

2 0.2629 < DN ≤ 0.433 na na na na na na 0 < cost ≤ 6,000







9 >1.45 >10,000 >640 >4,900 1 to 34 221 to 950 221 to 950 >42,000

Note: na = not applicable; FM = farm-to-market road; DN = visual distress.aIRI = international roughness index score.

FIGURE 3 Distress hierarchy.

Section 1

Failures DeepRutting

BlockCracking

AlligatorCracking

LongitudinalCracking

TransverseCracking

Patching

Section 2 Section 3 Section n-1 Section n

Distress NumberLevel 1: Overall Goal(Develop ProjectSelection Numberbased on inputs)

Level 2: Criteria used todetermine the overallgoal (Pairwise comparedto determine weights ofeach criterion)

Level 3: Candidate Sections (Each sectionwithin the evaluationnetwork)


Common sense suggests that, as distress density increases, so does the importance to perform preservation work on the section. A sec-tion with four failures is more important to repair than a section with only two—but how much more?

A system was established to define the upper limit at which an AHP weight of 9 could be assigned. The upper limit was defined on the basis of actual distress density within the study area. It was assumed that an unacceptable level of all distresses existed, which the district simply would not allow a section of roadway to reach without the performance of some type of maintenance. Any distress above this point could be considered drastically more important than a section in which none of a particular distress was manifest. The study set this upper limit as the point at which 98% of distress density was at or below the limit on the basis of 3 years of distress data.

To fill in the rest of the gaps on AHP weight and distress density, PMIS currently contains curves that describe how distresses affect the distress score for a section of pavement. To use the PMIS utility curves as a starting point, weights can be assigned to the increase in distress density so that a curve can be created for the AHP method that matches the shape of the utility curve. In this way, the method follows current thought at the Texas DOT to determine how distresses affect a section of pavement.

To use these utility curves, logic statements were developed that allowed application of AHP. The method created a ranking of impor-tance independently for each section with respect to all distress types.

The combination of these weight vectors with the weights developed from the application of AHP at Level 2 of the hierarchy (Figure 3) and the aggregation of all components resulted 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 competi-tions performed by each section for all distresses and decision cri-teria and begin to aggregate information to create a final index that can be used to prioritize pavement sections.

With a priority vector for each of the distresses shown at Level 2 of the hierarchy (Figure 3), a distress number for each of the 2,349 sec-tions can be created. The calculation of the distress number is as follows:

DN pzn d nd

k

w==

∑ �1

7( )

where

DNn = distress number for any section within evaluation network, d = distress types considered in decision (k = 7 in current

study), wd = weight associated with a particular distress, and pzn = priority number associated with any section within evalu-

ation network.

This distress number will become the visual distress parameter used in the application of 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 selec-tion number for each section within the network. This calculation is represented with the following equation:

�wn p n

p

k

∑==

PN pz (8)1

TABLE 4 Maximum Eigenvector and Priority Vector for Distress Types

Distress Type Maximum Eigenvector Priority Vector

Failures 0.8484 0.4488

Deep rutting 0.4023 0.2128

Block cracking 0.0500 0.0264

Alligator cracking 0.2537 0.1342

Longitudinal cracking 0.0854 0.0452

Transverse cracking 0.0455 0.0241

Patching 0.2051 0.1085

Sum 1.8905 1

TABLE 5 AHP Distress Weight Associated with Minimum Comparison

AHP Weight Failures (EA)a Deep Rutting (%)

Alligator Cracking (%)

Longitudinal Cracking (ft)

Transverse Cracking (EA)a 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

7 3 11 8 126 to 150 7 22 < patch ≤ 35

8 4 12 9 and 10 151 to 175 8 35 < patch ≤ 44

9 ≥5 ≥13 ≥11 ≥176 9 44 < patch

Note: na = not applicable.aEA = effects analysis.


where

PNn = project selection number for any section within evalua-tion network,

p = decision parameter considered in decision (k = 6 in current study), and

wp = weight associated with a particular parameter.

AHP provides a foundation on which to develop a project selec-tion 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. For a true test of the project selection number as a viable decision support tool, however, preservation projects must be selected and evaluated against actual projects. There must be a move beyond 0.5-mi section information into the actual selection of projects that involve multiple sections. Thus the gap between network and project level pavement management might be bridged.

reSuLtS of caSe Study aPPLication

Through AHP, decision parameters that originally were created with the use of varying measurement units could be accounted for in the project selection number, and variation no longer was an issue. The project selection number from an adjacent section could be added to another section and that number evaluated against other sum-mations of pavement sections. The case study district imposed a 2-mi project minimum. Therefore, sufficient project selection num-bers were added together to create realistic projects. To perform this operation, a logic statement was written to drive through the sections and sum project selection numbers. Project selection num-

bers were sorted. The largest indicated 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. Selection of realistic preservation projects through consideration of various parameters is a positive aspect of the new method. The stated goal of this study, however, was to create a decision support method that might actu-ally be used at the district level. Thus agreement needed to be high between projects selected by the method and actual preservation expenditures.

Of the 20 projects listed in Table 6, the case study district has decided to provide a known preservation action on 15. These preser-vation actions include a construction project performed by a private contractor, a routine maintenance project performed by a private contractor, and use of Texas DOT forces to perform the preservation action. The unknown action status for the remaining five projects does not necessarily mean that the district is not working on those sections or is not preparing a project for those sections; the unknown action status means only that the information was not available to make any definite determination.

Overall, the method matched at least 75% of the district decisions with respect to pavement preservation in the case study.

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 AHP and uses data from a network-level PMS, along with inputs from decision makers.

TABLE 6 Projects Selected Compared with Actual District Decisions

Rank Highway Begin Ref. Marker End Ref. Marker Length (mi) District Action

1 FM 488 0318+00.0 0323+00.5 5.5 Grading, structure, base, surface project let in FY 2009

2 FM 1451 0344+00.5 0349+00.0 4.5 Grading, structure, base, surface project let in FY 2009

3 FM 1644 0389+00.0 0396+00.0 7 Restoration project let in FY 2009

4 I-45 A 0186+00.2 0190+00.0 3.8 Contracted rehabilitation

5 FM 485 0605+00.5 0610+00.0 4.5 Restoration project let in FY 2007

6 FM 1848 0351+00.4 0354+00.2 2.8 Routine maintenance contract



9 FM 416 0625+00.3 0627+00.0 1.7 In-house maintenance forces

10 FM 80 0373+00.0 0376+00.7 3.2 In-house maintenance forces

11 FM 2293 0607+00.0 0610+00.0 3 Routine maintenance contract (FY 2007), then seal coat in FY 2009

12 SH 75 K 0387+00.0 0389+00.5 2.5 Unknown

13 US-190 K 0662+00.0 0665+00.5 3.5 Unknown

14 FM 80 0351+00.0 0354+00.0 3 Grading, structure, base, surface project let in FY 2009


16 FM 2485 0378+00.0 0380+00.0 2 Unknown

17 FM 488 0332+00.5 0335+00.5 3 Routine maintenance contract

18 FM 1940 0624+00.5 0626+00.5 2 In-house maintenance with rehabilitation let in Nov. 2010

19 FM 979 0614+00.7 0616+00.5 1.8 Unknown

20 FM 46 0605+00.0 0607+00.0 2 Unknown

Note: Ref. = reference; SH = state highway; FY = fiscal year.


AHP was used to select and prioritize projects in a way that mimicked how decision makers currently operate. The method was applied to data obtained from the Texas DOT PMS. Decision param-eters included visual distress, current ADT, current truck ADT, CS, ride quality, and maintenance costs. The visual distress parameter was created through the application of AHP to determine how dif-ferent distress types should be weighted in the decision-making process. The distress types considered were localized failures, alli-gator cracking, longitudinal cracking, block cracking, deep rutting, transverse cracking, and patching. The method allowed combina-tion of 0.5-mi pavement sections to create realistic preservation projects, which closely matched preservation decisions made by a Texas DOT district.

Further development of the method should include treatment cost as an additional decision factor. This improvement will not only help prioritize work that should occur but will also enable agencies to per-form a benefit–cost analysis on possible treatments. Further research is needed to incorporate a wider group of decision makers into the method. These techniques allow consideration of input from mul-tiple decision makers, which reflects the real-world decision-making process more accurately than do other methods.

referenceS

1. AASHTO Guidelines for Pavement Management Systems. AASHTO, Washington, D.C., 1990.

2. Haas, R., W. R. Hudson, and J. P. Zaniewski. Modern Pavement Man-agement. Krieger Publication Company, Malabar, Fla., 1994.

3. Smith, R. E. Integrating Pavement Preservation into a Local Agency Pavement Management System. In Transportation Research Record: Journal of the Transportation Research Board, No. 1795, Transpor-tation Research Board of the National Academies, Washington, D.C., 2002, pp. 27–32.

4. Šelih, J., A. Kne, A. Srdic, and M. Žura. Multiple-Criteria Decision Support System in Highway Infrastructure Management. Transport, Vol. 23, No. 4, 2008, pp. 299–305.

5. 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 Transpor-tation Research Board, No. 2119, Transportation Research Board of the National Academies, Washington, D.C., 2009, pp. 27–35.

6. Nuworsoo, C. K., K. Parks, and E. Deakin. Cost Per User as Key Factor in Project Prioritization: Case Study of San Francisco Bay Area, Cali-fornia. In Transportation Research Record: Journal of the Transporta-tion Research Board, No. 1986, Transportation Research Board of the National Academies, Washington, D.C., 2006, pp. 154–161.

7. Madanat, S. Optimal Infrastructure Management Decisions Under Uncertainty. Transportation Research Part C: Emerging Technologies, Vol. 1, No. 1, 1993, pp. 77–88.

8. Papagiannakis, T., and M. Delwar. Computer Model for Life-Cycle Cost Analysis of Roadway Pavements. Journal of Computing in Civil Engineering, Vol. 15, No. 2, 2001, pp. 152–156.

9. 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 1344, TRB, National Research Council, Washington, D.C., 1992, pp. 14–21.

10. Straehl, S. S., and L. Schintler. Montana Secondary Program Reform and Application of Goals Achievement Methodology to Project Prioriti-zation. In Transportation Research Record: Journal of the Transporta-tion Research Board, No. 1895, Transportation Research Board of the National Academies, Washington, D.C., 2004, pp. 85–93.

11. 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. 305–315.

12. 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. 367–374.

13. Saaty, T. L. The Analytic Hierarchy Process. McGraw-Hill Professional Publishing, New York, 1980.

14. Saaty, T. L. Multicriteria Decision Making: The Analytic Hierarchy Pro-cess: Planning, Priority Setting, Resource Allocation. RWS Publications, Pittsburgh, Pa., 1990.

15. Saaty, T. L. How to Make a Decision: The Analytic Hierarchy Process. European Journal of Operational Research, Vol. 48, No. 1, 1990, pp. 9–26.

16. Sun, L., and W. Gu. Pavement Condition Assessment Using Fuzzy Logic Theory and Analytic Hierarchy Process. Journal of Transportation Engineering. Posted ahead of print. Oct. 26, 2010, pp. 1–35.

17. Texas Department of Transportation. Overview of PMIS Needs Estimate. Materials and Pavements Section, Construction Division, Austin, 2009.

18. Papagiannakis, A., N. Gharaibeh, J. Weissmann, and A. Wimsatt. Pave-ment Scores Synthesis. 6386-1. Texas Transportation Institute, Texas A&M University System, College Station, 2009.

19. Li, Y., A. Cheetham, S. Zaghloul, K. Helali, and W. Bekheet. Enhance-ment of Arizona Pavement Management System for Construction and Maintenance Activities. In Transportation Research Record: Journal of the Transportation Research Board, No. 1974, Transportation Research Board of the National Academies, Washington, D.C., 2006, pp. 26–36.

20. Zhou, G., L. Wang, D. Wang, and S. Reichle. Integration of GIS and Data Mining Technology to Enhance the Pavement Management Deci-sion Making. Journal of Transportation Engineering, Vol. 136, No. 4, 2010, pp. 332–341.

21. Gurganus, C. F. Bridging the Gap Between Network and Project Selection Levels in Pavement Management. MS thesis. Texas A&M University, College Station, 2011.

The Pavement Preservation Committee peer-reviewed this paper.

project selection and prioritization of pavement preservation

Documents