TRANSPORTATION RESEARCH RECORD 1524 137

Prediction of Pavement Remaining Life

T. S. VEPA, K. P. GEORGE, AND A. RAJA SHEKHARAN

The evaluation of remaining life is necessary to make optimal use of the reasonableness of the selected methods will be examined by evalu-
structural capacity of in-service pavements. It simply represents the use- ating the remaining lives of Mississippi global positioning system
ful life left in the pavement until a failure condition is reached. Knowl- (GPS) sections of the Strategic Highway Research Program
edge of remaining life facilitates decision making in regard to strategies
(SHRP)–Long-Term Pavement Performance (LTPP) project by
for reconstruction-rehabilitation of roads, thereby leading to the efficient
use of existing resources. Several methods proposed or used by various each of these methods. The life predicted by each method will sub-
agencies to estimate the remaining lives of pavements are reviewed. They sequently be compared with those predicted by the other methods
are classified under two categories: functional and structural. Making use and discussed. Another objective is to propose a novel method of
of the Mississippi Department of Transportation pavement management calculating remaining life by the survivor curve approach. More
system data base, survivor curves are developed for seven classes of flex- than 3,000 sections of flexible pavements in the Mississippi Depart-
ible pavements with from thin to thick structures. By using these survivor
ment of Transportation (MDOT) road network were used to formu-
curves a novel method for estimating remaining life is proposed. The rea-
sonableness of the selected methods is examined by putting them to use late the survivor curves. With the survivor curves formulated, the
in calculating the remaining lives of each of eight rigid and flexible pave- remaining lives of the SHRP GPS sections are calculated and com-
ment sections, all of them from the Mississippi global positioning sys- pared with those obtained by traditional approaches.
tem sections of the Strategic Highway Research Program–Long-Term
Pavement Performance project (LTPP). With the structural details,
falling weight deflectometer deflection data, and the distress information REVIEW OF REMAINING LIFE
compiled from the LTTP information management system data base, the
authors use two and four methods for rigid and flexible pavements, ESTIMATION PROCEDURES
respectively, to determine the remaining lives. The remaining lives cal-
culated by two methods for rigid pavements are comparable. Three of The failure of a pavement can be categorized as structural or func-
four methods for flexible pavements also generated comparable remain- tional failure. In the functional failure-based approach, the remain-
ing lives. The authors were encouraged by the results and recommend ing life is computed on the basis of the performance of the pavement
that the survivor curve approach be explored further for network-level (for example, serviceability or rideability) and is expressed in terms
remaining life calculations. The reliabilities of various techniques cur-
of years/18-kip equivalent single axle loads (ESALs) (1 kip 5 8167
rently available for the remaining life calculation are discussed.
kg). The reduction of structural capacity, on the other hand, is the
primary concern in the structural failure-based approach. Witczak
With a large network of highways in place, a highway engineer’s (1) noted that remaining life estimates based on those two failure cri-
concern is shifted from construction to maintenance. For facilitating teria will be different. Figure 1 illustrates the structural and func-
the management of the existing network, pavement management tional remaining pavement lives according to Witzcak (1). Figure 1
systems (PMSs) have evolved over the last three decades. Proper is sketched with the premise that the remaining life estimate based
management of the system requires the collection, analysis, and on functional failure is greater than those based on structural failure.
interpretation of factual data relating to construction and mainte- In the following sections, the remaining life estimation methods are
nance activities. Prediction of the future condition of each pavement reviewed under those two categories.
as well as that of the entire network is an essential element of a man-
agement system.
For future planning and budgeting purposes, it is important to Functional Failure-Based Approaches
estimate the remaining life of each pavement section. This is not
only useful for timing a major rehabilitation but it also assists man- The 1986 AASHTO guide (2) describes several approaches for esti-
agers in forecasting the long-term needs of the network. Remaining mating remaining life, including the following three methods. First,
life is the extent of the useful life remaining in a pavement section if reasonably accurate historical traffic, X, is available, remaining
exposed to traffic and environmental forces. Expressed in terms of life (RLx) can be determined from Equation 1:
years/traffic or percentage of life left, it plays a pivotal role in fore-
Nf − X
casting the future rehabilitation needs of the network. RLx = • 100 (1)
Calculating remaining life has been a complex task, to say the Nf
least. Existing methods rely on various concepts from purely empir-
where
ical to truly mechanistic. The lack of adequate performance predic-
tion models has been the major impediment in predicting remaining
RLx 5 remaining life of pavement (in percent),
life. This paper addresses some of the issues of remaining life
Nf 5 total number of traffic applications to failure, and
prediction, including a review of some of the existing methods. The
X 5 cumulative number of 18-kip ESALs.

Department of Civil Engineering, University of Mississippi, University, This is known as the “traffic approach.” Second, if specific traffic
Miss. 38677. information is not available, remaining life can be determined on the
138 TRANSPORTATION RESEARCH RECORD 1524

base, George (3) has developed equations that can be used to fore-
cast the pavement condition rating (PCR) of five types of pave-
ments. The MDOT PMS working group has recently adopted a trig-
ger PCR value of 65 for a “must rehabilitate” level. With the
projected PCR and the trigger value of 65, therefore, remaining life
can be calculated. The same basic approach was used by Weissmann
et al. (4), Sullivan and Scott (5), and Lukanen and Han (6). Sullivan
and Scott (5) made use of an incremental roughness model, whereas
Lukanen and Han (6) used distress prediction models. Prediction
models were used to estimate the remaining lives of continuously
reinforced concrete pavements in a recent study by Weissmann et
al. (4). Ullidtz (7) calculated remaining life using an empirical
expression developed in terms of the present serviceability rating
(PSR) affected by traffic (to a terminal PSR of 2.0) and non-load-
associated degradation. Here, an annual PSR decrease of 0.1 is
assumed for non-load-associated degradation. The equation is

PSR − 2
RL =
FIGURE 1 Structural and functional remaining pavement 2
life (1).
• min[(AASHTO Cumulative Design W18/Yearly W18 ), 20]
basis of the time t that the existing pavement has been in service, the (3)
best estimate of the probable time Tf that the pavement can last
before any overlay is required, and the annual traffic growth rate r. where RL is the remaining life of the pavement (in years), and W18
Mathematically, is the number of 18-kip ESALs.

Tf − t (1 + r )Tf − (1 + r )t Structural Failure-Based Approaches

RLx = = (2)
Tf (1 + r )Tf − 1
According to the AASHTO Guide (2,8) remaining life can be cal-
This method is known as the “time approach.” Third, if the present culated by a unique curve (Figure 2) relating condition factor and
serviceability index of a pavement is known along with the initial remaining life. The condition factor Cx is defined as the ratio of the
structural number or slab thickness, remaining life can be estimated effective structural number to the original structural number or
by graphical procedures, designated the “serviceability approach.” the ratio of the effective slab thickness to the original slab thickness
Yet another approach would be to use a valid performance pre- for flexible and rigid pavements, respectively. AASHTO suggested
diction equation in conjunction with a threshold value that triggers two different methods for estimating the condition factor, which are
maintenance or rehabilitation. Making use of the MDOT PMS data briefly described here.

FIGURE 2 Relationship between condition factor and remaining life (8).

Vepa et al. 139

1. By the nondestructive deflection testing (NDT) approach, the lated modulus and the original thickness of the slab, according to
pavements in service are evaluated by using falling weight deflec- Figure 3. Note that the AASHTO NDT approach for both flexible
tometer (FWD) sensor deflections. Flexible and rigid pavements are and rigid pavements will be invoked in a later section for compari-
evaluated slightly differently, as described here. son studies.
For flexible pavements, by making use of the last sensor deflec- 2. By the visual condition approach, an overall condition factor is
tion, the subgrade resilient modulus is calculated first. The subgrade estimated by modifying the layer coefficient of each layer com-
resilient modulus in conjunction with the first sensor deflection is mensurate with the surface damage.
used to estimate the effective modulus Ep of the pavement layers George (9) developed a graphical procedure to determine remain-
above the subgrade. Ep can be obtained from Figure 5.5 of the ing life on the basis of the effective thickness ratio (ETR) as derived
AASHTO Guide (8) or the corresponding equation. Effective struc- from NDT deflections. Marchionna et al. (10) used mechanistic
tural number (SNeff) can be directly calculated from Ep, by the principles to develop a graphical procedure relating the percentage
following equation: of cracking to the percentage of residual life. Equivalent layer thick-
1
ness and critical strains in pavement layers have been used by Horak
SNeff = 0.0045 • D • ( E p ) 3 ( 4) (11) to compute remaining life through a series of graphical proce-
dures. On the basis of experimental data from field sections, Croney
where D is the total thickness of the pavement structure above the (12) developed deflection and life contours to estimate pavement
subgrade. residual life. With the aid of a fatigue model (considering the rate of
For rigid pavements, by making use of NDT deflections, in situ crack development in Arizona) in conjunction with the backcalcu-
portland cement concrete (PCC) modulus is backcalculated first. lated moduli, Mamlouk et al. (13) computed remaining life. Santha
Second, effective slab thickness is determined from the backcalcu- et al. (14) developed a simple mechanistic rut depth prediction

FIGURE 3 Determination of effective PCC structural capacity (thickness) from

NDT-derived PCC modulus (2) [1 in. 5 25.4 mm; 1 psi (lb/in.2 ) 5 6.9 kPa].
140 TRANSPORTATION RESEARCH RECORD 1524

model that, when used with estimated current traffic, yields the
remaining life. In another popular approach the moduli of the layers
are calculated by a backcalculation procedure, and by using those
moduli in an appropriate equation, remaining life is estimated
(15–17). Not only the NDT data evaluating fatigue life but also the
existing cracks and rutting are factored in calculating the remaining
life, according to the MODULUS program (version 5.0) (18).
In summary, in the functional failure-based approach for esti-
mating remaining life, the decrease in the performance index with
age or traffic is charted in conjunction with a functional failure cri-
terion. Alternately, the structural failure-based approach makes use
of fatigue principles, which requires the effective thickness or mod-
ulus derived from in situ measurements.
FIGURE 4 Survivor curve and Kolomogorov-
Smirnov Dk statistic for Low SNC pavements.
REMAINING LIFE USING SURVIVOR CURVE

All of the methods reviewed in the previous sections make use of subjective. Members of the family of flexible pavements with no
the condition of the pavement in some form and therefore are par- overlay are grouped under seven levels of modified structural num-
ticularly useful at the project-level evaluation. When historical data ber (SNC at the time of construction), which is the AASHO struc-
of pavement performance are available this information can be used tural number modified to account for subgrade. The seven groups are
to compile survivor curves, which can then be used to estimate designated Low (SNC # 2.5), Medium 1 (2.5 , SNC # 3.0),
remaining life. Fundamentally, the survival of a pavement is deter- Medium 2 (3.0 , SNC # 3.5), Medium 3 (3.5 , SNC # 4.0),
mined by the amount of time that it lasts before major maintenance Medium 4 (4.0 , SNC # 4.5), Medium 5 (4.5 , SNC # 5.0), and
or rehabilitation must be performed. Attempts have been made to High (SNC . 5.0). With the pavement lives for each category
estimate the average service life of a particular type of pavement by compiled from the data base, explicit survivor curves are developed
computing total area under survivor curves (19–23). individually for each of the seven groups. The values of q and r deter-
Historical records are used to develop survivor curves, which are mined for each group of pavements depend on the pavement charac-
empirical probability functions that are used to predict the percent- teristics affecting the service life. These include SNC and cumula-
age of pavement length of a specific age that will need rehabilitation tive traffic apart from other inherent variabilities affecting the
in the future. Garcia-Diaz et al. (23) developed survivor curves for pavement group.
flexible pavements at three different performance levels. A method The Kolmogorov-Smirnov test (24) is used to check the goodness
to estimate remaining life that is particularly useful for network- of fit of all survivor curves. Figure 4 is an example of survivor curve
level analysis and that uses survivor curves is proposed here. developed (for the low SNC group) along with the observed
The development of survivor curves follows the work of Garcia- Kolomogorov-Smirnov Dk statistic. The Dk statistic is defined as the
Diaz et al. (23). The performance histories of flexible pavements are maximum difference between actual and predicted survivor curve
obtained from the MDOT PMS data base. Inventory and overlay data. Table 1 lists the parameters of the seven survivor curves along
information is directly used for this purpose. The survivor functions with the observed and critical values of the Kolmogorov-Smirnov
that were developed assume the following form: Dk statistics. From Table 1 it can be seen that the observed Dk is less
than the critical (tabulated) Dk at the 0.05 level of significance for all
−q
v = 1 − exp r  (5) curves indicating the goodness of fit. The survivor curve of a class
W  of pavements can be used to estimate the expected (mean) life of that
class and from that estimate its remaining life.
where
TABLE 1 Parameters of Survivor Curves and Kolmogorov-Smirnov
v 5 percentage of surviving pavement length (mileage), Statistics
q 5 location parameter (i.e., parameter affecting the location of
the survivor curve),
r 5 shape parameter (i.e., parameter affecting the shape of the
curve), and
W 5 cumulative ESALs from the date of construction.
The MDOT PMS traffic data base is used to compile the cumula-
tive traffic until a major rehabilitation or overlay is carried out.
Cumulative ESAL was calculated by using current average daily
traffic, constant traffic growth rate, constant truck percentage, and
constant ESAL factors for each section. All of these data were fur-
nished by MDOT. This traffic or the corresponding years are used as
a surrogate for pavement life. The basic tenet is that even though the
decision to overlay at failure is subjective, in the past engineers con-
sistently estimated the end of life of pavement sections. It should be
noted that the decision to place a rehabilitative overlay is somewhat
Vepa et al. 141

What follows is a proof that the mean life can be estimated with
the area under the survivor curve. No doubt, the time to failure of
nominally identical pavement sections varies over a range because
of the inherent variabilities in materials and construction proce-
dures. Accordingly, the life or time to failure will be treated as a ran-
dom variable and denoted by T. Let the cumulative distribution
function (cdf) of T be defined by the equation

FT (t ) = P(T ≤ t ) (6)

The reliability function RT (t) is

RT (t ) ≡ P(T > t ) = 1 − FT (t ) ( 7)
FIGURE 5 Remaining life using survivor curve.
The reliability function gives the probability that failure does not
occur in (0, t). Since FT (t) is nondecreasing with FT (0) equal to 0 and
FT (t) approaches 1 as t approaches infinity, it follows that RT (t) is REMAINING LIFE CALCULATION
nonincreasing with RT (0) equal to 1 and that R T (t) approaches zero
as t approaches infinity. The random variable T may be either discrete Only a few of the many proposed methods of remaining life calcu-
or continuous. The mean value mT for the case in which T is con- lation are adopted by highway agencies. It is enlightening to make
tinuous and has a probability density function fT (t) is given by the use of some of the methods for calculating the remaining lives of
equation typical pavement sections for the purpose of comparison. Eight rigid
∞
and eight flexible pavement sections of the Mississippi SHRP GPS
∫
m T = E(T ) = t • fT (t )dt
0
(8) program were used for the comparison study.

Data Requirement for Comparison Study

By equation 7
All of the data—distress, roughness, and FWD deflections—for the
fT (t ) dt = − dRT (t ) ( 9)
16 sections are compiled from the SHRP information management
system data base. Similar surface condition data (distresses and
Hence, roughness) and the traffic counts for the same road segments com-
∞ piled by MDOT are made available from the MDOT PMS data base.
∫
m T = − t • dRT (t )
0
(10) SHRP and MDOT PMS data for 1993 afforded a unique opportu-
nity for a side-by-side comparison of the condition survey results.
Note that the SHRP data pertain to 152-m (500-ft) sections, as
A parts integration yields opposed to the longer sections of various lengths in the MDOT
∞
PMS. By making use of the deduct point approach (25,26), PCR val-
mT =
∫ R (t) dt
0
T (11) ues are calculated by using the condition survey data of SHRP and
the MDOT PMS. That the PCR values computed from the two inde-
pendent data acquisition systems are in satisfactory agreement
If T is discrete (Table 2) is an indication of the robustness of the data collection
methodologies.
∞

mT = ∑ R (t ) T (12)
0
Rigid Pavement Remaining Life
From Equation 11 or 12 one concludes that the mean value of the
Eight rigid pavements, four jointed and another four continuously
time to failure is the same as the total area under the cdf curve. This
reinforced SHRP sections, were analyzed for their remaining lives
concept is extended to estimate remaining life (Figure 5). After hav-
by two methods: the 1993 AASHTO NDT method, which is a struc-
ing sustained a traffic volume W (estimated in terms of 18-kip
ESAL) the percentage of life remaining is given by tural failure-based approach, and a procedure based on the condition
prediction equation (PCR equation derived from MDOT PMS data).
A At the outset it should be noted that the pavement sections are
RLx =  1  • 100 (13)
 A0  sparsely distressed, with the PCR being predominantly in the range
of 80. The construction date of each pavement is worthy of note as
where A0 is the total area under the curve or the mean value of the well. As the data indicate all of these sections except Section 4 have
service life of the pavement, and A1 is the remaining area under been in service from 6 to 19 years.
the curve after traffic W passes. A cursory examination of the remaining lives listed in Table 3
By using RLx the remaining life (in years) can be calculated by leads to the following observations:

RLx 1. The remaining lives predicted by AASHTO as well as by the

RL = • Age of pavement (14)
100 − RLx PCR prediction model seem reasonable since the expected lives
142 TRANSPORTATION RESEARCH RECORD 1524

TABLE 2 Comparison of PCR Values Calculated Using SHRP Data and

MDOT PMS Data for SHRP Sections

(current age plus remaining life) of most of the pavements consid- 3. For Section 4, which was in service for only 1 year at the time
ered are found to be in the range of 25 to 30 years, a reasonable range that it was studied, the remaining life predicted by the AASHTO
for the service lives of concrete pavements. Further verification of method is 19 years, whereas that predicted by the PCR method is 38
remaining life models can be carried out by using data from other years. A remaining life of 38 years as predicted by the PCR model
comprehensive data bases, for example, the SHRP-LTPP data base. indicates the model’s inability to forecast so long into the future.
2. By the AASHTO method the FWD deflection values deter-
mine the in situ moduli by using the MODULUS program (version A word of caution is needed here. It is important to exercise engi-
5.0). The moduli range from 2.8 3 107 to 5.5 3 107 kPa (4 million neering judgment in forecasting the remaining lives of relatively
to 8 million lb/in.2). The effective PCC thickness derived from the new pavements. It is comforting to observe that remaining life gen-
AASHTO curve (Figure 3) is practically insensitive to moduli in this erally becomes an issue for moderate to highly distressed pavements
range. Nonetheless, the remaining lives appear to be reasonable. rather than for newly built pavements.

TABLE 3 Comparison of Remaining Lives of SHRP Rigid

Flexible Pavement Remaining Life
Sections
The remaining lives of eight SHRP flexible sections were calculated
by four different methods (Table 4): the AASHTO method, which
is structurally based (method A in Table 4); method B, which is
structural as well as functional failure based because it relies on both
backcalculated values and the surface distresses; method C, which
again includes both structural and functional effects because PCR
depends on distresses and roughness, in that order; and lastly, the
survivor curve method, which in actuality is a functional failure-
based procedure. The data used, for example, the cumulative traffic
and condition data as measured in 1993, along with original struc-
tural details are tabulated in Table 5. A comparison of the remain-
ing lives predicted by the four methods reveals that the MODULUS
program (version 5.0) gives only a gross estimate of the remaining
life. The authors of the program indicate that a refined version of the
remaining life algorithm is forthcoming shortly. The result pertain-
ing to Section 8 is anomalous in that it remains in a superior condi-
Vepa et al. 143

TABLE 4 Comparison of Remaining Lives of SHRP Flexible Sections

tion [as indicated by a PCR of 88 and an international roughness Discounting the anomalous result for Section 8, the following
index (IRI) of 0.99, see Table 5], but it is expected to be in service specific conclusions are offered with respect to the remaining lives
for only 2 to 3 years. The relatively high traffic volume perhaps has computed by the four methods (Table 4):
inflicted excessive structural damage to the pavement over the years
(14 years to be exact), with only a small fraction of its life remain- 1. The remaining lives determined by the AASHTO method, a
ing in 1993. structural failure-based approach, agree with those obtained by the
With the intention of calculating remaining life by the ETR survivor curve approach except for the first section. The exception-
method (9), the moduli of all eight sections are backcalculated by ally low value for the first section can be attributed to the much
using the MODULUS program (version 5.0). The ETR method not reduced structural capacity, as evidenced by FWD deflections.
only requires the current in situ modulus values for the calculation 2. The MODULUS program (version 5.0) predicts consistently
of ETR but it also calls for the original (as-constructed) moduli of high remaining lives because of low-severity cracking and rutting
each layer as well. It should be noted that the backcalculated asphalt in the pavement sections. Augmenting this light distress scenario
concrete moduli are unreasonably high [ranging from 5.1 3 106 are the relatively high backcalculated moduli, another input in the
to 2.2 3 107 kPa (740 to 3,170 kips/in.2), with an average value of MODULUS (version 5.0) remaining life routine.
1.3 3 107 kPa (1,920 kips/in.2)] and that the ETR calculation has not 3. The PCR equation-based and proposed survivor curve meth-
produced remaining lives consistent with the pavement condition ods yielded comparable remaining lives for the eight sections. This
index; accordingly, those results are not tabulated. is due in part to the fact that the PCR equation method depends on

TABLE 5 Pertinent Data for Calculation of Remaining Lives for SHRP

Flexible Sections
144 TRANSPORTATION RESEARCH RECORD 1524

a trigger level of PCR, that is, a PCR level at which rehabilitation is International Conference on Managing Pavements, TRB, National
recommended, and the survivor curve relies on the actual rehabili- Research Council, Washington, D.C., 1994, pp. 63–73.
7. Ullidtz, P. Pavement Analysis. Elsevier, NewYork, 1993, pp. 306–307.
tation times of a class of pavements. 8. AASHTO Guide for Design of Pavement Structures. AASHTO, Wash-
ington, D.C., 1993.
9. George, K. P. Overlay Design and Reflection Cracking Study for Flex-
CONCLUDING OBSERVATIONS ible Pavements. MDOT Report No. MSHD-RD-85-79. Mississippi
Department of Transportation, 1985.
10. Marchionna, A., M. A. Fornaci, and M. Malgarini. Evaluation of Flex-
Despite the advances made in the condition evaluation of in situ ible Pavements and Overlay Design Based on F.W.D. Tests. Proc., Sixth
pavements, the analysis of the resulting data is still elusive, espe- International Conference, Structural Design of Asphalt Pavements,
cially in forecasting how long before a major rehabilitation would Department of Civil Engineering, Michigan University, Ann Arbor,
be needed, designated the remaining life. A comparison of the Vol. 1, 1987.
11. Horak, E. Application of Equivalent-Layer-Thickness Concept in a
prevalent methods confirms the findings of a previous study that Mechanistic Rehabilitation Design Procedure. In Transportation
structural failure-based calculations result in conservative remain- Research Record 1207, TRB, National Research Council, Washington,
ing life values compared with the functional failure-based method- D.C., 1988, pp. 69–75.
ologies. The structural methods, relying on in situ moduli, are sus- 12. Croney, D. Is the Measured Deflection of a Flexible Pavement a Reli-
able Guide to Life Prediction and Overlay Design? Journal of High-
pect in view of the perennial problems of backcalculation. NDT
ways, Vol. 58, 1990, pp. 24–26.
deflection-based approaches probably stand a better chance of being 13. Mamlouk, M. S., J. P. Zaniewski, W. N. Houston, and S. L. Houston.
consistent and reasonably precise in project-level life calculations. Overlay Design Method for Flexible Pavements in Arizona. In Trans-
Another potential method adaptable to both project- and network- portation Research Record 1286, TRB, National Research Council,
Washington, D.C., 1990, pp. 112–122.
level computations would be the use of a sound pavement predic- 14. Santha, B. L., W. Yang, and R. L. Lytton. Microcomputer Application
tion model, for example, the PCR equation used in the present study. to Determine the Load Zoning for Low-Volume Roads. In Transporta-
Nonetheless, the NDT method and the prediction model method call tion Research Record 1260, TRB, National Research Council,
for several inputs for each pavement section. In comparison, the Washington, D.C., 1990, pp. 226–245.
15. Fernando, E. G., D. R. Luhr, and D. A. Anderson. Development of an
survivor curve approach proposed here is simple; is likely to be rel- Overlay Design Procedure for Flexible Pavements. Final Report.
evant to the geographical area, current materials, specifications, and FHWA Report No. FHWA/PA-84/024, Fourth Cycle of Pavement
so forth (because it is developed from historical data for that region); Research, Vol. 2. Pennsylvania Transportation Research Facility, 1984.
and requires only one input, namely, the cumulative traffic that has 16. Richter, C. A., and L. H. Irwin. Application of Deflection Testing to
Overlay Design: A Case Study. In Transportation Research Record
passed on the pavement. The development of survivor curves, how- 1196, TRB, National Research Council, Washington, D.C., 1988,
ever, requires a comprehensive data base, and the evolving PMS pp. 193–200.
data base can be relied on for those data. 17. Kilareski, W. P. Heavy Vehicle Evaluation for Overload Permits. In
Transportation Research Record 1227, TRB, National Research Coun-
cil, Washington, D.C., 1989, pp. 194–204.
18. Scullion, T., and M. Chester. Nondestructive Test Procedures for Ana-
ACKNOWLEDGMENTS lyzing the Structural Condition of Pavements. Texas Transportation
Institute, Texas State Department of Highways and Public Transporta-
This report includes the results of a study entitled “Pavement tion, 1995.
19. Winfrey, R., and F. B. Ferrel. Life Characteristics of Surfaces
Management System—Phase II,” conducted by the Department of Constructed on Primary Rural Highways. HRB Proc., Vol. 20, 1940,
Civil Engineering, University of Mississippi, in cooperation with pp. 165–199.
MDOT and FHWA, U.S. Department of Transportation. The 20. Ferrel, F. B., and H. R. Paterick. Life Characteristics of Highway
authors thank Alfred Crawley and Joy Portera of the MDOT Surfaces. HRB Proc., Vol. 24, 1948, pp. 40–52.
21. Winfrey, R., and P. D. Howell. Highway Pavements—Their Service
Research and Development Division for excellent cooperation and Lives. In Highway Research Record 252, HRB, National Research
valuable technical input in the study. The authors are grateful for the Council, Washington, D.C., 1968, pp. 1–23.
financial support of MDOT as well. 22. Millard, R. S., C. B. White, N. W. Hudson, F. A. Sharman, E. D.
Tingle, R. L. Wilson, and E. H. Woodrow. Lightly Trafficked Roads—
Great Britain. Proc., Permanent International Association Road
Congress, Vol. 2, 1971.
REFERENCES 23. Garcia-Diaz, A., R. L. Lytton, and J. T. Allison. Estimation of Network
Rehabilitation and Maintenance Costs over an Extended Planning Hori-
zon (abridgement). In Transportation Research Record 943, TRB,
1. Witczak, M. W. Determination of Flexible Pavement Life. Executive National Research Council, Washington, D.C., 1983, pp. 9–12.
Summary, Vol. I. FHWA Report No. FHWA/MD/R-79/1. FHWA, U.S. 24. Hays, W. L. Statistics, 4th ed. Holt, Rinehart and Winston, Inc., 1988,
Department of Transportation, 1978. pp. 816–818.
2. AASHTO Guide for Design of Pavement Structures. AASHTO, 25. George, K. P. Pavement Management Information System—Phase II.
Washington D.C., 1986. Interim Report on Composite Index for Pavement Condition Rating.
3. George, K. P. Pavement Management Information System. Final Report. Department of Civil Engineering, The University of Mississippi,
MDOT Report No. MSHD-RD-90-87. Mississippi Department of University, Miss., 1993.
Transportation, 1989. 26. George, K. P., W. Uddin, P. J. Ferguson, A. B. Crawley, and A. R.
4. Weissmann, A. J., B. F. McCullough, and W. R. Hudson. Development Shekharan. Maintenance Planning Methodology for Statewide Pave-
of Performance Prediction Models for Continuously Reinforced ment Management. In Transportation Research Record 1455, TRB,
Concrete Pavements in Texas. Final Report. FHWA Report No. FHWA/ National Research Council, Washington, D.C., 1994, pp. 123–131.
TX-901472-7F. FHWA, U.S. Department of Transportation, 1989.
5. Sullivan, T., and R. Scott. Strategic Road Network Management: An The opinions, findings, and conclusions expressed in this paper are those of
Approach Using Roughness. Proc., Sixth Conference, Road Engineers the authors and are not necessarily those of MDOT or FHWA. This does not
Association of Asia and Australasia, Kuala Lampur, Malaysia, Vol. 1 to constitute a standard, specification, or regulation.
5, March 4–10, 1990.
6. Lukanen, E. O., and C. Han. Performance History and Prediction Mod- Publication of this paper sponsored by Committee on Pavement Manage-
eling for Minnesota Pavements. In Conference Proceedings 1, Third ment Systems.