Professional Documents
Culture Documents
Fuzzy MDMA For Pavement Project Evaluation Using LCC
Fuzzy MDMA For Pavement Project Evaluation Using LCC
Abstract: Project life-cycle cost analysis (LCCA) is a reasonable approach to compare pavement design alternatives, but not taking into
consideration of pavement performance in the most LCCA approaches may lead not to obtain the most desirable alternative for road users and
associated highway agencies. This article introduces a multicriteria decision making method for pavement project evaluation using LCCA
that (1) utilizes fuzzy set theory to model and handle uncertainties; (2) takes into account the extra user costs attributable to inadequate
pavement condition; and (3) as a criterion, considers life-cycle pavement performance that describes overall pavement serviceability
condition. Using a hypothetical project, the paper clarifies the performance of the proposed method for long-term evaluating alternative
pavement design strategies. The alternatives are compared with respect to each criterion of the method, and also using sensitivity analysis,
it is determined that changes in the decision maker’s risk and confidence attitudes can affect the ranking of the alternatives. DOI: 10.1061/
(ASCE)IS.1943-555X.0000170. © 2014 American Society of Civil Engineers.
Author keywords: Decision making; Pavement strategies; Life-cycle analysis; Performance; Fuzzy numbers; Linguistic terms; Risk;
Confidence attitude.
J. Infrastruct. Syst.
in LCCA calculations, and some other DOTs consider these costs Pavement Performance Cost Value
with a weight of lower than 1. Here, considering agency cost as the
Long-term pavement performance is measured using Present
most important criterion, the importance weight of 1 was assigned
Serviceability Index (PSI). Performance curve is produced on
to agency costs and the importance weights of other criteria would
the basis of the equations presented by the American Association
be assigned in comparison with agency costs. Therefore, the fol-
of State Highway and Transportation Officials (AASHTO) for rigid
lowing multicriteria problem would be solved as objective function
and flexible pavements (AASHTO 1993). Based on AASHTO
Minimize ðLCACi þ wLCUC × LCUCi þ wPPCV × PPCVi Þ ð1Þ equations, the drop in PSI in each year is determined on the basis
of accumulated traffic load passed on the road during that year. In
where LCACi , LCUCi and PPCVi are life-cycle agency cost, life- this paper, the jump in the performance curve attributable to
cycle user cost and pavement performance cost value, respectively, improving M&R activities is calculated by using the values used
Downloaded from ascelibrary.org by SOUTHERN CALIFORNIA UNIVERSITY on 04/09/14. Copyright ASCE. For personal use only; all rights reserved.
of design strategy i, and wLCUC and wPPCV are the importance by Chen (2007).
weights of life-cycle user cost and pavement performance cost The criterion of pavement performance, as the term pavement
value, respectively. performance cost value, indicates willingness-to-pay considered by
corresponding highway management official for better pavement
Life-Cycle Agency Cost condition. To calculate this value, the presented method utilizes
the approach presented by Fwa and Sinha (1991). They considered
In addition to initial construction costs, maintenance and rehabili- pavement performance by multiplying average PSI during project
tation (M&R) costs should be included in the LCAC. Moreover, if life-cycle by life-cycle cost. Here, as a constant amount in the cost
the analysis period is shorter than the service lives of alternatives or genre, the average of agency costs of all alternatives has been used
the service lives of different alternatives are different, then remain- to PPCV evaluation of each alternative. So, PPCV for design
ing service life cost value, as another cost component of agency strategy j is calculated as follows:
costs, should be considered in the analysis. This cost component
is calculated based on the remaining life of an alternative at the
end of the analysis period, as a prorated share of the last rehabili- PPCVj ¼ PFj × LCACave ð3Þ
tation cost, and is considered as a negative cost (FHWA 1998;
Lamptey et al. 2005). where LCACave = average of life-cycle agency costs of all compet-
ing alternative design strategies, and PF is Performance Factor of
Life-Cycle User Cost design strategy j which is calculated by
User costs concern highway users in the pavement life-cycle. User
cost components include vehicle operating cost, travel time cost, Pt − PSILC;j
PFj ¼ ð4Þ
crash cost and environmental costs (Lamptey et al. 2005; Li and Pt − Pb
Madanu 2009). Each cost component can be categorized into
two categories: user costs under normal operation conditions
where Pt and Pb are top and bottom limits of the range selected for
and added user costs resulting from work zones (FHWA 1998;
desired amounts of PSI and PSILC;J is life-cycle pavement perfor-
Lamptey et al. 2005; Li and Madanu 2009). LCCA need only con-
mance for design strategy j which is calculated by
sider differential cost among alternatives (FHWA 1998).
Considering work zone operations, little difference between Pn
crash costs of work zones and normal operations is expected PSI · ESALi;j
PSILC;j ¼ Pn i;j
i¼1
ð5Þ
(FHWA 1989; Lamptey et al. 2005). Therefore, user cost differen- i¼1 ESALi;j
tials are an aggregation of excessive travel time costs and vehicle
operating costs. These cost components are calculated by using the
method presented by FHWA (1998). where PSIi;j and ESALi;j are average PSI value and accumulated
During normal operations, crash costs and environmental costs Equivalent Single Axle Load (ESAL) for design strategy j at year i,
are unlikely to vary among alternative pavement designs. Accord- respectively. Average PSI in each year is equal to the average of PSI
ing to various researches (i.e., FHWA 1998 and Lamptey et al. measures at the beginning of the year and at the end of the year.
2005), there is no need to consider crash costs and environmental
costs during normal operations, but vehicle operating costs and Fuzzy Approach Utilized to Model Uncertainty
travel time costs are likely to vary for different design strategies.
Some input parameters of pavement project life-cycle analysis are
The travel time costs should be considered attributable to speed
associated with uncertainties. Because fuzzy set theories facilitate
reduction causes by undesirable pavement conditions. Such extra
modeling uncertainty in the absence of precise and complete infor-
travel time costs are calculated based on a model for speed adjust-
mation, in the developed method, a fuzzy estimating approach
ment factors which is developed by Zaniewski (1982). For such
which was proposed by Shaheen et al. (2007) is used.
vehicle operating costs, the World Bank suggested a general equa-
Uniform, triangular and trapezoidal are the types of fuzzy
tion (Asphalt Institute of Iran 2007) and also conducted several re-
numbers used in this methodology. However, these types of fuzzy
searches on road user costs in New Zealand (FHWA 1998), which
numbers and also crisp numbers can be represented by the trapezoi-
can be used to calibrate the general equation as follows:
dal four-element format (Shaheen et al. 2007). From crisp numbers
VOC to trapezoidal numbers, the uncertainty about the selected number
¼ 1 þ 0.045 × ½maxð0; IRI − 2.5Þ1.406 ð2Þ would increase. If the decision maker (DM) select a crisp number
VOC0
for a variable it means the amount of that variable is completely
where VOC = vehicle operating costs in case the performance certain. On the contrast, selecting a trapezoidal number by the
condition equals to IRI, and VOC0 = vehicle operating costs in case DM indicates even the most likely value for that variable is not
the performance condition is less than 2.5 m=km. The correlation certain. Therefore, a range should be selected even for the most
between PSI and IRI could be obtained from Lay (2009). likely values.
J. Infrastruct. Syst.
Decision Maker’s Risk Attitudes • Step 1: Establish alternative pavement design strategies for the
analysis period.
Certainly, as mentioned by Fenton and Wang (2006), considering
• Step 2: Determine service lives and activity timing. Service lives
risk and uncertainty in multicriteria decision making increases the
of structural activities are determined from pavement
reliability and quality of the results by evaluating the ranking
design calculations. After passing the service life of each
sensitivity to each risk attitude. Moreover, incorporating risk
structural activity, rehabilitation activity is performed to
attitudes will help the DM choose the preferred alternatives from
strengthen the pavement structurally. In the length of time be-
an extreme pessimist to an extreme optimist.
tween two consecutive structural activities, preventive mainte-
Based on the discipline existing in the variables of triangular
nance activities are utilized to correct minor defects, retard
fuzzy numbers to consider the DM’s risk attitudes for benefit
future deteriorations and maintain or improve the functional
parameters (where the higher the value for parameters, the more
Downloaded from ascelibrary.org by SOUTHERN CALIFORNIA UNIVERSITY on 04/09/14. Copyright ASCE. For personal use only; all rights reserved.
J. Infrastruct. Syst.
Downloaded from ascelibrary.org by SOUTHERN CALIFORNIA UNIVERSITY on 04/09/14. Copyright ASCE. For personal use only; all rights reserved.
Therefore, four pavement design strategies were established in the Data for Calculating User Costs
analysis. Activity profiles and corresponding design parameters Unit rates of user costs are gathered from the Iranian government
of each structural activity of the alternatives 1–4 are given in databases and the works done by Matthews and Allouche (2010)
Table 1. and ODOT (2003). These costs are as follows:
• Vehicle operating cost (Cents=km)—passenger cars (14.0,
Data for Calculating Agency Costs 15.5); single-unit trucks (32.5, 35.5); combination trucks
Table 2 shows unit costs of work items gathered from the Iranian (46.5, 49.0)
government databases and expert judgments. These costs are con- • Travel time cost ($=h)—passenger cars (7.50, 8.25); single-
sidered as trapezoidal fuzzy numbers because they are the most unit trucks (9.45, 10.05); combination trucks (11.25,
uncertain input factors of the case. 11.85)
J. Infrastruct. Syst.
Table 1. Alternative Pavement Design Strategies (Activity Timing and Design Parameters)
Alternative Service life (year) Activity type Activity name Timing
No. 1 40 IC Asphalt concrete (5.98)a Base year
RH Thick HMA overlay (5.46)a 25th year
PM HMA crack sealing 4th, 7th, 10th, 13th, 16th, 19th, 22th,
25th, 28th, 31th, 34th, 37th year
No. 2 40 IC Asphalt concrete (5.98)a Base year
RH Thick HMA overlay (5.46)a 25th year
PM HMA crack sealing 8th year
Thin HMA overlay 15th, 33th year
No. 3 42 IC Jointed plain concrete (11.02)b Base year
Downloaded from ascelibrary.org by SOUTHERN CALIFORNIA UNIVERSITY on 04/09/14. Copyright ASCE. For personal use only; all rights reserved.
Table 2. Unit Rates of Material/Work Item Costs • Importance weight of criterion life-cycle user cost—medium,
Material/work item Corresponding fuzzy number Importance weight of criterion pavement performance cost
value—medium.
HMA wearing 19 mm ($=m3 ) (114.20,116.16,118.06,120.02) • Risk attitude—neutral, confidence attitude—neutral.
HMA binder 25 mm ($=m3 ) (104.06,105.98,107.90,109.82)
Granular base ($=m3 ) (20.23,21.02,21.88,22.71)
Granular subbase ($=m3 ) (9.50,9.75,10.26,10.51)
Tack coat ($=m2 ) (0.54,0.56) Results and Discussion
Prime coat($=m2 ) (1.10,1.12)
JPCP ($=m3 ) (94.30,105.50,113.02,124.22) Using PSSFM to evaluate the hypothetical example led to the
Crack sealing ($=m) (7.85,8.35) results shown in Table 3. As shown in the Table, with respect to
Joint cleaning and sealing ($=m) (2.48,2.82,2.98,3.32) agency costs and user costs, Alternatives 1 and 2, as flexible pave-
Mill asphalt surface ($=m2 -cm) (1.52,1.56,1.60,1.64) ment strategies, are the most desirable alternatives, respectively. At
Mill concrete surface ($=m2 -cm) (1.84,1.88,1.92,1.96) the same time, with respect to pavement performance cost value,
Dowel bar ($=m) (21.00,22.00,23.16,24.16)
Tie bar ($=m) (1.52,1.60,1.68,1.76)
Alternatives 4 and 3, as rigid pavement strategies, have taken two
Saw cutting concrete joints ($=dm3 ) (1.18,1.21,1.34,1.37) top orders. Alternative 2, because of not possessing a regular pre-
ventive maintenance policy, has the worst pavement condition in
the analysis period. Eventually, primarily attributable to consider-
ing pavement performance cost value as a decision making cri-
Owing to the uniform ranges of the coefficients used to obtain terion, Alternative 4 is selected as the most appropriate strategy.
these costs, they are considered as uniform fuzzy numbers. To select the most appropriate design strategy and for clear
evaluation, the analysis has been repeated for different DM’s risk
Other Fuzzy Variables Considered in the Analysis and confidence attitudes. The value of objective function and the
• Annual traffic growth rate (triangular fuzzy number)—(1.5, ranking orders of the alternatives are shown in Table 4.
2.0, 2.5%). As shown in Table 4, a DM who is absolutely confident about
• Discount rate (triangular fuzzy number, based on the recom- the fuzzy input values will select Alternative 4 as the most appro-
mended range and rate by FHWA 1998)—(3.0, 4.0, 5.0%). priate strategy, independent of his/her risk attitude. However, a DM
with neutral confidence attitude, under absolutely pessimistic
Other Parameters and Factors Depended on Decision risk attitude will choose Alternative 1, but under absolutely
Maker’s Attitudes optimistic and neutral risk attitude Alternative 4 will be chosen
• Bottom limit of desired range of PSI—2.5, top limit—4.0. for implementation.
Table 3. Values of the Three Criteria and the Objective Function and the Ranking Orders of the Alternatives
Pavement performance
Life-cycle agency costs Life-cycle user costs cost value Objective function value
Standard Standard Standard Standard
Design EV deviation EV deviation EV deviation EV deviation
strategy ($1,000) ($1,000) Order ($1,000) ($1,000) Order ($1,000) ($1,000) Order ($1,000) ($1,000) Order
Alternative 1 16,047 276 1 12,638 1,248 4 2,385 71 3 24,256 1,070 2
Alternative 2 18,025 354 4 7,407 1,399 1 8,516 255 4 26,645 1,424 4
Alternative 3 17,580 626 3 12,548 1,877 3 1,621 49 2 25,222 1,807 3
Alternative 4 16,930 609 2 11,891 1,664 2 1,584 47 1 23,966 1,621 1
J. Infrastruct. Syst.
Table 4. Total Life-Cycle Costs and Ranking Orders of the Alternatives under Different Risk and Confidence Attitudes
Absolutely optimistic Neutral Absolutely pessimistic Risk attitude
Standard Standard Standard
EV deviation EV deviation EV deviation
Design strategy ($1,000) (1,000 $) Order ($1000) ($1,000) Order ($1000) (1,000 $) Order Confidence attitude
Alternative 1 22,959 235 2 23,122 289 2 23,284 235 2 Absolutely confident
Alternative 2 26,186 224 4 26,339 278 4 26,491 224 4
Alternative 3 23,807 388 3 24,076 479 3 24,344 388 3
Alternative 4 22,622 389 1 22,887 483 1 23,152 389 1
Alternative 1 23,554 1,173 2 24,256 1,070 2 25,254 1,179 1 Neutral
Downloaded from ascelibrary.org by SOUTHERN CALIFORNIA UNIVERSITY on 04/09/14. Copyright ASCE. For personal use only; all rights reserved.
Under absolutely nonconfident attitude, cost components of Asphalt Institute of Iran. (2007). Modeling flexible pavement response
concrete pavement associates with more uncertainties (the ranges and performance, Ministry of Roads and Urban Development, Tehran,
given for cost components of rigid pavement are wider than those Iran.
of flexible pavement). Hence, the chance of being selected as the Bonissone, P. P. (1982). “A fuzzy sets based linguistic approach:
supreme alternative for concrete pavement strategies, Alternative 3 Theory and applications.” Approximate reasoning in decision analysis,
and 4, decreases with respect to risk attitude changes from abso- M. M. Gupta and E. Sanchez, eds., North-Holland, Amsterdam,
329–339.
lutely optimistic to absolutely pessimistic. As result, Alternative
Chan, A., Keoleian, G., and Gabler, E. (2008). “Evaluation of life-cycle
1 will be selected as the most appropriate strategy by a DM
cost analysis practices used by the Michigan Department of Transpor-
who is absolutely nonconfident and absolutely pessimistic.
tation.” J. Transp. Eng., 10.1061/(ASCE)0733-947X(2008)134:6(236),
236–245.
Chen, C. (2007). “Soft computing-based life-cycle cost analysis tools for
Conclusions and Recommendations transportation infrastructure management.” Dissertation for the Degree
of Doctor of Philosophy in Civil and Environmental Engineering,
A new methodology was proposed for pavement project life-cycle
Virginia Polytechnic Institute and State Univ., Blacksburg, VA.
analysis that would enable one to consider and handle uncertainties
Federal Highway Administration (FHWA). (1989). Construction costs and
associated with input factors using a fuzzy approach. The proposed
safety impacts of work zone traffic control strategies, U.S. Dept. of
Pavement Strategy Selection Fuzzy Method (PSSFM) also takes Transportation, Washington, DC.
into account pavement performance and its effects on the Federal Highway Administration (FHWA). (1998). Life-cycle cost analysis
user costs. in pavement design—In search of better investment decisions, U.S.
To assess the performance of the presented method, it was Dept. of Transportation, Washington, DC.
utilized in a hypothetical project. The results revealed that paying Federal Highway Administration (FHWA). (2002). Life-cycle cost analysis
attention to long-term costs, in general, decreases total costs of the primer, U.S. Dept. of Transportation, Washington, DC.
projects. Also, considering the effects of pavement performance in Fenton, N., and Wang, W. (2006). “Risk and confidence analysis for
the analysis will help calculate real costs of pavement projects. fuzzy multicriteria decision making.” Knowl. Base. Syst., 19(6),
Moreover, sensitivity analysis of the results, considering DM’s risk 430–437.
and confidence attitudes, can lead to select the most appropriate Fwa, T. F., and Sinha, K. C. (1991). “Pavement performance and life-cycle
pavement design strategy. cost analysis.” J. Transp. Eng., 10.1061/(ASCE)0733-947X(1991)
The presented life-cycle performance calculation technique 117:1(33), 33–46.
which is based on AASHTO equations provides an adequate Labi, S., and Sinha, K. C. (2005). “Life-cycle evaluation of flexible
and convenient approach to estimate pavement performance con- pavement preventive maintenance.” J. Transp. Eng., 10.1061/
dition, especially in the absence of actual pavement condition data. (ASCE)0733-947X(2005)131:10(744), 744–751.
Obviously, the successful application of the presented methodology Lamptey, G., Ahmad, M., Labi, S., and Sinha, K. C. (2005). “Life cycle
cost analysis for INDOT pavement design procedure.” FHWA/IN/JTRP-
depends on utilizing actual estimation of uncertainties associated
2004/28, Purdue Univ., West Lafayette, IN.
with the problem.
Lamptey, G., Labi, S., and Li, Z. (2008). “Decision support for optimal
scheduling of highway pavement preventive maintenance within resur-
facing cycle.” Decis. Support Syst., 46(1), 376–387.
References
Lay, M. G. (2009). Handbook of road technology, Chapter 14, 4th Ed.,
AASHTO. (1993). AASHTO guide for design of pavement structures, Spon Press, New York, 334.
Washington, DC. Li, Z., and Madanu, S. (2009). “Highway project level life-cycle benefit/
AASHTO. (2010). “Life cycle cost analysis for pavement type selection cost analysis under certainty, risk, and uncertainty: Methodology with
responses.” 〈http://www.research.transportation.org/Documents/LCCA case study.” J. Transp. Eng., 10.1061/(ASCE)TE.1943-5436.0000012,
Replies.xls〉 (Dec. 3, 2010). 516–526.
Abaza, K. A. (2002). “Optimum flexible pavement life-cycle analysis Matthews, J. C., and Allouche, E. N. (2010). “A social cost calculator for
model.” J. Transp. Eng., 10.1061/(ASCE)0733-947X(2002)128:6(542), utility construction projects.” NASTT’s 2010 No-Dig Show, North
542–549. American Society for Trenchless Technology (NASTT), Chicago.
J. Infrastruct. Syst.
Oregon Department of Transportation (ODOT). (2003). The value of travel- Tighe, S. (2001). “Guidelines for probabilistic pavement life cycle cost
time: Estimates of the hourly value of time for vehicles in Oregon 2003, analysis.” Transp. Res. Rec., 1769(1), 28–38.
Salem, OR. Wilde, W. J., Waalkes, S., and Harrison, R. (1999). “Life cycle cost
Reigle, J. A., and Zaniewski, J. P. (2002). “Risk-based life-cycle cost analysis of portland cement concrete pavements.” FHWA/TX-00/
analysis for project-level pavement management.” Transp. Res. Rec., 0-1739-1, Univ. of Texas, Austin.
1816(1), 34–42. Zaniewski, J. P., Butler, B. C. Jr., Cunningham, G., Elkins, G. E.,
Shaheen, A. A., Fayek, A. R., and AbouRizk, S. M. (2007). “Fuzzy num- Paggi, M., and Machemehl, R. (1982). “Vehicle operating costs, fuel
bers in cost range estimating.” J. Constr. Eng. Manage., 10.1061/ consumption, and pavement type and condition factors.” FHWA/PL/
(ASCE)0733-9364(2007)133:4(325), 325–334. 82/001, Washington, DC.
Downloaded from ascelibrary.org by SOUTHERN CALIFORNIA UNIVERSITY on 04/09/14. Copyright ASCE. For personal use only; all rights reserved.
J. Infrastruct. Syst.