See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228406818

Integrating dynamic performance prediction models into pavement

management maintenance and rehabilitation programs

Article · January 2001

CITATIONS READS

15 1,973

4 authors, including:

Ningyuan Li Susan Tighe

University of Waterloo University of Waterloo
37 PUBLICATIONS 633 CITATIONS 381 PUBLICATIONS 5,496 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Susan Tighe on 19 June 2014.

The user has requested enhancement of the downloaded file.

 5th International Conference on Managing Pavements (2001)

Integrating Dynamic Performance Prediction Models into Pavement Management

Maintenance and Rehabilitation Programs

Li Ningyuan*, Tom Kazmierowski*, Susan Tighe** and Ralph Haas**

* Ministry of Transportation of Ontario*

Pavements and Foundations Section
1201 Wilson Ave. Downsview,
Ontario, Canada M3M 1J8
Tel: (416) 235-3518 or (416) 235-3512
Email: Li.Ningyuan@mto.gov.on.ca
Tom.Kazmierowski@mto.gov.on.ca

** Department of Civil Engineering

University of Waterloo
Waterloo, Ontario
Canada N2L 3E5
Tel: (519) 885-1211 or (519) 888-4567
Email: Haas@uwaterloo.ca
Sltighe@uwaterloo.ca

Abstract Modeling of pavement performance deterioration in terms of riding quality or pavement

roughness and surface distress is a critical engineering process in pavement management system (PMS).
Integrating the performance prediction models into multi-year network pavement maintenance and
rehabilitation (M&R) program involves several interactive functional processes of the pavement
management, such as database information management, site specific performance models, M&R
treatment alternatives and optimization analysis. Based on review of the pavement management system
developed recently for the Ministry of Transportation of Ontario (MTO), this study presents an integrated
dynamic performance prediction and M&R optimization methodology that may be considered for use in
the future development. In particular, the study discusses the needs for enhancement of the system's
functional ability to integrate pavement deterioration models with multi-year M&R treatments program.
This involves adoption of common deterministic and probabilistic prediction models available for
optimizing the allocations of annual investment in pavement rehabilitation program at network level.

INTRODUCTION

The Ministry of Transportation of Ontario currently has about 18,680 kilometres of provincial highways
consisting of 3,697 km of freeways, 4,460 km of collectors, and 10,522 km of other roads. Most of the
highways are paved with asphalt concrete or are treated with asphalt surface layer. To meet the
requirements of pavement management in terms of database structure and information management, these
roads are broken down into 1710 sections based on regional jurisdiction, pavement structure/materials,
traffic and environmental factors, etc. At present, the Ministry invests about $200 million annually to
ensure that the highway network is maintained above the levels of serviceability required for each
classified highway. The serviceability is assessed and ranked by means of Riding Comfort Index (RCI)
or International Roughness Index (IRI), Distress Manifestation Index (DMI) and their combination
Pavement Condition Index (PCI). Both RCI and DMI range from 0 to 10, while PCI ranges from 0 to
100; new pavements are often rated at 95, and old pavements requiring resurfacing at 45 to 55.

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

Table 1 below gives an example showing the Ministry's current highway network in terms of IRI
measured in 1999 for each classified road and pavement type. Of the 18,680 km of roads, 80 percent is
paved with asphalt concrete and about 17 percent surface is treated by asphalt materials. The total
exposed Portland Cement Concrete (PCC) pavement is about 0.7 percent of roads in the network.

TABLE 1 An Overview of the Ontario Ministry of Transportation’s Road Network and

its Pavement Condition as Surveyed in 1999

Road Class Versus IRI Measured Pavement Type Versus IRI Measured
Road Class Length (km) IRI Pavement Type Length (km) IRI
Freeway 3697 (20%) 1.40 Asphalt Concrete 14,499 (78%) 1.67
Arterial 6084 (32.5%) 1.61 Composite 442 (2%) 1.71
Collector 4460 (24%) 1.94 PCC 138 (0.7%) 2.11
Local 3980 (21%) 3.09 Surface-Treated 3143 (16.8%) 3.31
Gravel 458 (2.5%) N/A Gravel Surface 458 (2.5%) N/A
Note: The total length of roads in 1999 was 18680 km and the weighted average IRI was 1.68

Illustrated in Figure 1 is an overview of pavement types and their riding conditions in terms of the
average IRI values measured on all pavements in each region. The IRI values measured on asphalt
concrete pavements are lower in MTO’s Central and Southwestern (SW) regions (1.41 and 1.50
respectively) in comparison with those in Northern and Northwestern (NW) regions (1.95 and 1.91)
which typically have thin structural sections.

3.5
Asphalt Concrete Composite
Portland Concrete Surface Treated
3

2.5

1.5

0.5

0
Central Southwest Eastern Northern Northwest

FIGURE 1 Distribution of pavement types and roughness condition in each region.

The average IRI values measured on exposed PCC pavements are generally higher than those
measured on asphalt concrete pavements in the same regions, reflecting the increased average age and
lack of maintenance activities on the rigid pavement. The IRI values of all surface-treated pavements,

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

which are mainly located in Northwestern and Northern regions, are nearly two times higher than those of
asphalt concrete pavements.
To facilitate the decision-making process in planning for network pavement maintenance and
rehabilitation program under budget constraints, the Ministry launched in 1998 the development of its
new generation of pavement management system. To meet the technical specifications and requirements,
there are many functional components to be established or modified on the exiting system that was
developed in 1987. This paper describes the pavement performance models built in the new pavement
management system, and then discusses potential application of an integrated dynamic performance
prediction and M&R optimization analysis process to the new system in the future. Previous studies
have concluded that the optimal cost-effective M&R treatment strategies can be obtained by
integrating dynamic performance prediction with M&R programming in the process of pavement
management (1).

SCOPE AND OBJECTIVES

The purpose of this study has twofold: i) to review the pavement performance prediction models and
M&R analysis methodologies applied in the second generation MTO pavement management
system (PMS-II), including database information management, pavement performance
evaluation, maintenance and rehabilitation treatments, 2) to identify the needs for improvement
and introduce a new approach to network pavement M&R multi-year programming integrated with
dynamic performance prediction. The deterioration of pavements is modeled using mechanical-
empirical method with consideration of previous M&R treatment effects on future deterioration
rate of pavement serviceability. Each M&R treatment is defined in terms of structural design,
construction criteria, paving materials, treatment effect on the existing pavement structural and
functional performance. The programming produces maintenance planing for each pavement
section in the network on the basis of annual cost-effectiveness. More specifically, the study
described in this paper covers the following aspects:

• Review and assess the current performance prediction models developed for individual pavement
types used in Ontario provincial highways.
• Introduce a new technical approach to prediction of pavement deterioration integrated with a
number of standardized M&R treatment alternatives and their individual treatment effects.
• Illustrate the basic steps to analyze the impact of funding strategies on network performance with
an example application of integrated dynamic pavement performance deterioration and M&R
treatment effects using a representative database of Ontario road network.

REVIEW OF PAVEMENT PERFORMANCE MODELS

Pavement deterioration in terms of serviceability can be affected by a number of factors, including

pavement age, traffic level, climatic effects, pavement structure, construction, maintenance and
rehabilitation activities. At present, there are various types of techniques available for use in modeling
pavement deterioration (2). Assessment of pavement condition is described in one or more indices related
to riding quality, safety, structural strength and stability. In PMS-II, the following three indices are used
to characterize pavement condition and predict pavement deterioration: 1) Riding Comfort Index (RCI) to
reflect pavement roughness, 2) Distress Manifestation Index (DMI) to measure severity and extent of
several categorized pavement surface distresses which reflect riding safety of a pavement, and 3)
Pavement Condition Index (PCI) to measure an overall pavement serviceability which is a function of
combined components of IRI and DMI. A detailed description of each of the indices is stated as follows:

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

Pavement Roughness Measurement

Prior to 1997, riding comfort index (RCI) was used to measure pavement network roughness by use of
Portable Universal Roughness Device (PURD). The output of PURD measurement for a pavement
section gives an average slope variance computed from profile elevations on 50 m long segments. To
obtain RCI value for a pavement section, a transfer function between RCI and PURD values for each
pavement type has been established through regression analysis.
Since 1997, the Ministry has adopted International Roughness Index (IRI) to measure the riding
quality of its highway network as an evaluation of pavement functional performance. And annual IRI
measurement at the pavement network has been contracted out through open competition tendering
process on a yearly basis. It is required that each year all of the consultants are required to go through
pre-qualification process by submitting the results of IRI measurements conducted at a designated
calibration circuit using the device that is proposed to use throughout the execution of the network
roughness measurements. Because RCI is still used in the PMS-II to reflect pavement roughness
component, it is necessary to establish a proper conversion function to transform IRI to RCI values for
each pavement type. The following conversion functions, as shown in Table 2, were developed using
1996-1997 time frame data.

TABLE 2 Conversion Functions Between IRI and RCI

Pavement Type Transfer Function R2 Standard
Error
Flexible (AC) RCI = 8.52 - 7.49 × log10 (IRI) 0.652 ±0.938
Exposed concrete (PCC) RCI = 9.27 - 6.22 × log10 (IRI) 0.509 ±0.513
Composite (COM) RCI = 8.48 - 3.81 × log10 (IRI) 0.201 ±0.946
Surface Treated (ST) RCI = 15.7 e - 0.307 (IRI) 0.621 ±1.240

Pavement Distress Measurement

Evaluation of pavement distress involves identification and rating of individual surface distresses present
on a given road section. Each individual distress is visually evaluated by aggregating the severity and
extent prescribed in the documents (3 and 4). A systematic method for classifying and assessing visible
consequences of various distresses (i.e., distress manifestations) was developed and implemented through
Ministry-wide in 1970's. To ensure uniformity of interpretation and reporting, a comprehensive manual
was prepared for providing detailed pavement distress assessment guidelines. A series of photographs are
presented to illustrate distress manifestations and explain how to evaluate their severity and density on
ratio scale from 0 to 4.
DMI is used to aggregate the effect of individual distresses present on a given pavement section, and
it is defined as follows:
n
DMI Max − ∑ Wi ( s i + d i )
i =1
DMI = 10 × (Eq.1)
DMI Max

where, i = distress type.

Wi = weighting factor, ranging from 0.5 to 3.0, representing the relative weight or attribute to
overall pavement surface condition of each evaluated pavement section, as listed in Table 2.
si = severity of distress expressed on a 5-point scale, ranging from 0.5 to 4.0.

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

di = density of distress occurrence expressed on a 5-point scale, ranging 0.5 to 4.0.

DMImax = the maximum value theoretically assigned to an individual pavement distress. Based on the
new weighting factors, the DMImax values assigned for AC, PCC, COM and ST pavements
are 208, 196, 216 and 180, respectively.
After re-scaling process included in Equation 1, the output of DMI values changes form 0 to 10, with
10 representing the condition of a newly or rehabilitated pavement, and 0 representing the worst
condition. In the PMS/2, all weighting factors are either newly established or updated from the existing
ones based on each individual distress related to pavement condition assessment. These weighting factors
were reviewed by a MTO work group composed of all regional representatives and discussed at several
workshops held in late 1999. Table 3 lists partial distress weighting factors (Wi) established for individual
pavement types.

TABLE 3 Distress Weighting Factors (Wi) Defined for Individual Pavement Types

AC Pavement Wi PCC Pavement Wi COM Pavement Wi ST Pavement Wi

Raveling 3 Raveling 0.5 Raveling 3 Raveling 3
Flushing 1.5 Polishing 1.5 Flushing 1.5 Flushing 2
Rippling/Shoving 1 Scaling 1.5 Spalling 2 Streaking 1
Wheel Track Rutting 3 Potholing 1 Tenting/Cupping 2.5 Potholing 1
Distortion 3 Joint Cracking/Spalling 2 Wheel Track 3 Rippling & 2
Rutting Shoving
Longitudinal Wheel 1.5 Faulting 2.5 Joint-Failure 3 Wheel Track 3
Track Single- Rutting
Multiple Cracking
Longitudinal Wheel 3 Distortion 1 Distortion & 1 Distortion 3
Track Alligator Settlement
Cracking
Longitudinal 1 Joint-Failure 3 Longitudinal 2 Longitudinal 1
Meaner and Midlane Meander Cracking Cracking
Cracking
Transverse Alligator 3 Longitudinal Meander 2 Transverse 1 Pavement-Edge 2
Cracking Cracking-Multiple Breaking
CL Alligator 2 Transverse-Cracking 2 Transverse Joint 2 Alligator 3
Reflective Cracking Cracking
Pavement Edge 0.5 Sealant loss 0.5 CL Cracking Single 0.5
Single-Multiple
Pavement Edge 1.5 Diagonal Corner/Edge 2.5 CL Cracking 1.5
Alligator Cracking Multiple

Overall Pavement Condition Evaluation

Based on measurements of pavement roughness and distresses, the overall condition for a pavement
section is ranked using the following equation:

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

PCI = 10 × (0.1RCI )1 / 2 × DMI × C i (Eq.2)

In Equation 2, PCI value changes from 0 and 100, and the values defined for RCI and DMI all vary
from 0 to 10, with being a flawless pavement. Ci is a coefficient calibrated for each pavement type based
on regression analysis between the calculated PCI and PCR (pavement condition rating) which is visually
observed and ranked by road condition raters. It should be noted that the regression analysis conducted
for developing PCI calculation formulas was based on the PCR data collected in years 1997 and 1998.
The coefficients calibrated for each of the four pavement types are listed in Table 4.

TABLE 4 Coefficients Calibrated for PCI Calculation

Pavement Type Model Coefficient (Ci) R2
Flexible (AC) 1.088 0.86
Rigid (PCC) 0.998 0.63
Composite (COM) 1.034 0.52
Surface Treated (ST) 0.962 0.51

A further examination was performed to evaluate the impact that will be brought forth by the revised
PCI calculation for asphalt pavements using the 1997 and 1998 performance data. It was found that there
is a strong relationship between the existing PCR and the revised PCI (R2 = 0.86). This analysis is not
performed for other pavement types, since the existing PCI is only available for asphalt pavements. The
assessment of pavement overall performance in terms of PCI is a function of RCI and DMI. In other
wards, IRI has not been directly used in the calculation of an overall pavement condition index PCI.

Built-in Pavement Performance Models

The procedure used by PMS/2 for predicting PCI takes two steps as described below:

1. Predict RCI and DMI individually using their default prediction models. In the PMS/2 pavement
sections are classified in terms of traffic, structural thickness, environment and subgrade type, which
are the main influence factors to the default performance prediction models as summarized in Table 5
below. Provided in Table 6 are some coefficients a, b and c in the default models used to predict RCI
on the basis of the classified performance for each pavement type. These default performance models
can be then modified in terms of updating the coefficients used for individual pavement sections if
site-specific data/performance is available for each section. A common practice is to develop initial
default models using the available information and engineering judgement.

TABLE 5 Influence Factors and Levels Used in Default Prediction Models

Influence Factors Levels of An Influence Factor
Traffic High Medium High
(Total ESALs per year) (> 500,000) (50,000 - 500,000) (< 50,000)
Structural Thickness Thin Medium Think
(Granular Base (GBE < 500 mm) (GBE = 500-750 mm) (GBE > 750 mm)
Equivalency)
Environment (Climate) North Ontario Zone South Ontario Zone
Subgrade Type (Mpa) Weak Medium Strong
(< 30 Mpa) ( 30-45 Mpa) (> 45 Mpa)

The following sigmoidal curves are used to predict RCI and DMI for each individual pavement sections:

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

(a − b × c t )
P = Po − 2e (Eq.3)
Where :

P = performance index, RCI or DMI

Po = P at age o
t = loge(1/Age)
a,b,c = model coefficients

2. Calculate PCI for individual pavement sections by Equation 3 relating PCI with RCI and DMI by
inputting predicted RCI and DMI values. Pavement performance is also related to rehabilitation
activity taken for an existing pavement section. Based on the expected service life span of each
treatment, performance improvements upon completion of treatment and deterioration rate are
developed, individual RCI and DMI prediction models are developed for each of these treatments.
Separate projection models will be developed for each major maintenance and rehabilitation
treatment using the following categories in PMS-II.

TABLE 6 RCI Prediction Model Coefficients

Treatment Average Coefficient
RCI0 a b c
Flexible Pavements
Hot-in-Place Recycling 9.0 1.673 6.786 2.125
1 Lift HM Overlay 8.0 1.673 6.286 2.125
Full Depth Removal/Reclamation, 1 Lift HM Overlay 9.5 3.267 5.869 1.367
Reconstruction to Flexible Pavement 9.5 3.267 6.119 1.367
Pulverize, grading, Double Surface Treatment 8.5 1.673 5.495 2.125
Exposed Concrete Pavements
Concrete Rehab 9.0 1.500 10.977 2.340
Concrete Rehab, Diamond Grinding 9.5 1.500 9.418 2.340
Concrete Rehab, 2 Lift HM Overlay 9.5 1.500 8.520 2.340
Reconstruction with Concrete 9.5 1.500 13.708 2.340
Composite Pavements
Mill, Concrete Rehab , 2 Lift HM Overlay 9.5 1.500 8.520 2.340
Reconstruction to Composite Pavement 9.5 1.500 9.860 2.340
Mill, Rubblize , 3 Lift HM Overlay 9.5 1.500 8.520 2.340
Surface Treated Pavements
Single or Double Surface Treatment 8.0 1.673 5.596 2.125
Reconstruction to Flexible, HM Pavement 9.5 3.267 6.119 1.367

INTEGRATED PERFORMANCE PREDICTION MODELS

Standardizing M&R Treatments

The basic process of developing a number of practical M&R treatment strategies for use in the
management of a road network is illustrated in Figure 2. It starts with a classification of all pavement
sections in a road network. Classification of highway pavements in terms of road function, the minimum
acceptable PCI trigger levels and pavement types are reviewed. Trigger levels must be realistic and
reflect policy and resources of the Ministry. When a pavement reaches the minimum acceptable service

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

level specified, it becomes a need. For each classified pavement type, a number of standardized M&R
treatments are selected from all of the feasible M&R treatment candidates. The M&R treatments range
from do-nothing to major rehabilitation or reconstruction, as detailed in reference (1). In identifying
deficient pavements and estimating future needs of pavement sections, both the existing pavement
performance and treatment effect or influence on the pavement future deterioration are considered by a
comprehensive prediction model, i.e., integrating pavement deterioration with treatment effects.

Road Classification

Factors Considered in Selecting M&R

Treatment Alternatives

• Definition of work contents for each treatment

• Criteria for structural design
• Construction and equipment required
• Expected life and unit cost of each M&R treatment
• Treatment effect and impact on cost-effectiveness

List of M&R List of M&R List of M&R List of M&R

Treatments for Treatments for Treatments for Treatments for
Rigid Concrete Asphalt Concrete Composite Composite
Pavements Pavements Pavements Pavements

FIGURE 2 Process of selecting pavement M&R treatments.

Individual M&R Treatment Effects

If a pavement performance analysis period goes beyond the initial service life of the alternative being
considered, then the performance prediction should include a separate deterioration model to consider the
effects of the selected maintenance or rehabilitation treatment(s) within the program period. In other
words, an integrated performance prediction model with regular treatments should be developed to predict
the pavement serviceability-age relationship over the entire analysis period, including the prediction
model for the initial pavement structure and modified prediction models for the pavement structure after
implementing each of the preservation actions.
Maintenance activities are those used to treat surface defects of a pavement section so as to slow
down deterioration of the pavement condition and to prolong the pavement life. The typical major
maintenance activities carried out by the Ministry are used to correct pavement surface defects and are
determined at the district or regional levels. On the other hand, rehabilitation activities often involve
removing, rejuvenating and recycling of the distressed pavement material in the full length of the
pavement section. The key functions associated with the maintenance and rehabilitation treatments
include the range and average life span, the performance improvement after the treatment in terms of the
index values. In most cases new structure layers are added on top of the existing pavements. For
example, some typical pavement rehabilitation activities used in PMS-II are summarized in Table 7,
including performance improvements or treatment effects on PCI and RCI values as well as expected life
span after completion of the rehabilitation treatments. The DMI values after the rehabilitation treatments

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

were determined by calculation based on the estimated PCI and RCI values. The DMI increments after
the maintenance treatments can not be determined using the same calculation procedure as in the case
with rehabilitation treatments. It is suggested that estimations on how the individual distresses are fixed
with the maintenance treatments be made to determine the DMI increase (5). It should be noted that each
of the treatment strategies, including routine and major maintenance and rehabilitation or reconstruction,
is defined in accordance with Ontario Pavement Design Manual (6).

TABLE 7 List of Some Typical Rehabilitation Activities in PMS-II

Rehabilitation Alternative Expected Performance Index

Life Increase after Treatment
Flexible Pavements PCI RCI DMI
Hot In-Place Recycling 10.5 90 8.5 9
Cold In-Place Recycling 12.5 90 8.5 9
1 Lift Hot Mix Overlay 7 85 8 9
2 Lifts Hot Mix Overlay 10 90 9 9
Mill 1 Lift and 1 Lift Hot Mix Overlay 8.5 90 8.5 9
Full Depth Removal/Reclamation and 1+ Lifts Hot 13.5 95 9.5 10
Mix Overlay
White Topping 11 90 9 9
Reconstruction to Flexible Pavement 16 95 9.5 10
Reclamation, 2 Lifts Hot Mix Overlay 12.5 90 9 9
Full Depth Removal/Mill Reclamation 12.5 90 9 9
Pulverize, Grading, Double Surface Treatment 11 85 8.5 9.5
Exposed Concrete Pavements PCI RCI DMI
Concrete Rehab. 17.5 90 9 9.5
Concrete Rehab. and Diamond Grinding 17.5 95 9.5 9.5
Concrete rehab. and 2 Lifts Hot Mix Overlay 16 90 9 9.5
Rubblizing and 2 Lifts Hot Mix Overlay 13.5 90 9 9.5
Unbonded Concrete Overlay 27.5 95 9.5 10
Reconstruction to Concrete Pavement 27.5 95 9.5 10

Maintenance and Rehabilitation Optimization

Selection of investment alternatives is not easy with mathematical programming because: a) there are
usually multiple investment objectives for each of many individual pavement sections involved in the
network and, b) considerable uncertainty may exist for future funding and each individual pavement
section’s deterioration prediction. Each prioritisation method has certain specific features in terms of
model development, design parameter and economic analysis (5).
The optimisation model proposed for use in PMS-II is a cost-effectiveness based multi-year priority
programming, which was described previously in reference (1). As shown in Equations (4) to (8), the
objective function is, given budget limitations and other constraints, to maximise the total value of cost-
effectiveness (i.e., the total benefit-cost ratios) to compare alternative pavement treatments for a pavement
network with total S sections on yearly basis. Each pavement section with M treatment alternatives
(ranging from do-nothing to reconstruction alternative strategies) every year for T programming years.

Objective function for the optimisation model is:

N
M  ( PCS − A ) × L × AADT × D  
Maximise: ∑ ∑ Xstm ×  Lstm × Wst × C st × (1 + R)st− t st  , ∀t (Eq.4)
s =1  m =1  st st stm 

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

and the objective function is subject to the following conditions or constraints:

M
∑ Xstm = 1, ∀ s, t (Eq.5)
m =1

1 if maintenance alternative m is selected for section s in year t

X stm =  (Eq.6)
0 otherwise
S M
(Eq.7)
∑∑ X stm × ( Lst × Wst × Cstm ) ≤ Bt , for t = 1, 2, 3, . .., T
s =1 m =1
(Eq.8)
PCSs ( t +1) = PCSst + ( Xstm ∆PCSm ) ≤ PCSmax . ∀ s, t , m
where,

PCSst = Generalised Pavement Condition State (such as Pavement Condition Index, i.e. PCI) for
section s (of S total sections in a road network) at year t (of T years of analysis period)
Ast = The minimum acceptable level of PCS required for pavement section s at year t, and
(PCSst - Ast) can be either positive or negative value,
Lst = Length (km) of pavement section s in year t,
AADTst = Annual Average Daily Traffic carried on pavement section s in year t
Dst = Number of service days for traffic flows by pavement section s in year t if treatment
alternative strategy m is selected,
Wst = Width (m) of pavement section s in year t,
Cstm = Unit cost ($ / per square meter) of a standardised M&R treatment alternative strategy m
is applied to pavement section s in year t,
R = Discount rate for calculating present value of future cost,
Bt = Budget limit for all the M&R actions in the network in programming year t
∆PCSm = Treatment effect of a standardised M&R action, which is defined as an amount of PCS
that can be recovered, from the existing Pavement Condition State, by the M&R action
alternative m,
PCSmax. = A maximum value of pavement condition state defined for a pavement. For examples: if
PCS is defined by PSI in one case, which is measured on a scale of 0 to 10, with 10
being perfect, then the highest level of the PCS is 10, i.e., PSImax is 10; if PCS is defined
by PCI in another case, which is measured on a scale of 0 to 100, with 100 being perfect,
then the highest level of the PCS is 100, i.e., PCImax is 100.

A brief description of each above equation is stated as follows:

(Eq.4) is the objective function of the optimisation model, which maximises the value of the total cost-
effectiveness over the entire programming period. It is used to find the optimal M&R action program for
the network in each programming year, as compared to all other alternative M&R action programs.

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

(Eq.5) and (Eq.6) state that the total number of available standardised M&R treatment strategy options
designed for the network is M. In each programming year one and only one of these M&R options for
pavement section s must be chosen, which produces the highest cost-effectiveness from the network
system point of review.

(Eq.7) controls the maximum investments or annual budget available for the network maintenance and
rehabilitation projects of each year. Within the period of multi-year M&R program, available budget of
each programming year can be different from each other.

(Eq.8) indicates that a pavement serviceability level (or pavement condition state, PCI) can not be higher
than its the maximum level at any time. Actually, this constraint plays a role of “penalty function”, which
avoids the optimisation model from selecting projects for those pavements that have a high PCI but
generate low economic benefit or effectiveness.

EXAMPLE APPLICATION

A small road network consisting of 15 pavement sections is used to demonstrate how the optimization
model is applied to develop annual pavement M&R for the network. Table 8 shows the pavement
network information data, including pavement section number or code, section length and width, daily
traffic volume and percentage of trucks, thickness, subgrade soil modulus and PCI in 1999.

TABLE 8 Example Sectional Data for An Asphalt Pavement Network

Highway No. Section No. Section AADT/Truck% PCI Value
(LHRS1) Length/Width (m) (Vehicles) in 1999
79 37320 15200 / 6.7 2150 / 13.0 6.3
25 25480 9000 / 6.8 5700 / 11.7 7.4
28 26420 9300 /6.5 500 / 11.0 6.7
11 17000 2900 /14.6 21250 / 10.0 6.0
132 44720 2500 / 6.5 1000 / 38.0 6.9
1 10014 7500 / 7.1 28000 / 21.0 6.2
7 14421 4900 /17.0 57300 / 6.4 6.4
2 10850 4700 / 15.8 28000 / 3.1 6.8
402 48225 7200 / 14.6 14500 / 35.0 6.6
60 33250 1600 / 6.7 1850 / 11.0 8.3
31 27060 8800 / 7.3 6000 / 11.5 8.5
138 45420 1290 / 7.3 6000 / 13.4 6.3
36 28245 4000 / 6.7 1350 / 5.7 7.1
148 46280 5300 / 6.7 3150 / 7.0 7.5
16 20200 1500 / 14.6 3000 / 12.1 5.2
Note:1.Linear Highway Reference System (LHRS) number coded in the Highway Inventory Management
System (HIMS) of Ontario, Canada
2. Total equivalent granular thickness of pavement
3. Resilient modulus of subgrade soil
4. PCI is equivalent to Pavement Condition Index (PCI) on a scale of 0 -10

The minimum PCI for each pavement in the network is 4.5. The five standardized pavement treatment
alternatives are defined: Do-Nothing, Routine Maintenance, Major Maintenance, Rehabilitation and
reconstruction, as shown in Figure 3. Each of the five treatment strategies is defined by treatment type,
work contents, materials, construction and unit cost. The cost for each of the five treatments in this

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

example is 0, 1.5, 7, 14, and 20 Dollars per square meter. The five treatment strategies, including routine
and major maintenance and rehabilitation or reconstruction are based on the Ontario Pavement Design
Manual for Ontario conditions (6). In addition, the effect of each treatment on existing pavement is
specified. For instance, if the M&R alternative strategy 2 (i.e., Routine Maintenance) is selected for year
t, then a rise of 0.5 units of PCI can be obtained in that year, and there should be a small PCI increase in
that year. Alternatively, if a minor rehabilitation treatment, i.e., strategy 4 is selected in year t, then an
amount of 3.0 units of PCI is recovered in that year. Following the PCI jump point, where a treatment
action is applied, a new deterioration model, which reflects the improved pavement structure by the
treatment, should be established to predict the pavement deterioration in year t+1. The procedure is
repeated in each consecutive year until the entire analysis period is completed for the integrated
performance prediction.

M&R Treatment Alternatives for AC Pavements

(1)
Do-Nothing

(2) Routine (3) Major (4) Minor (5) Major

Maintenance Maintenance Rehabilitation Rehabilitation

Design Criteria, Material Criteria, Construction Criteria, Specifications, Etc.

Spray Patching Rout & Seal Cracks Hot-Mix Resurfacing Hot-Mix Resurfacing
Distortion Repair Hot-Mix Patching Hot-in-Place Recycle (>20 mm Overlay)
Minor Crack Sealing Sealing Asphalt Strip Cold Mix Sealing Full Depth Removal &
Drainage Repair Drainage Surface Treatments Resurfacing Recycle plus
Maintenance Frost Treatment Hot-Mix overlay

Effect of Each M&R Treatment on the Existing Pavement in Term of PCI Increase

Rise of the Existing Rise of the Existing PCI Rise of the Existing PCI Rise of the Existing PCI
PCI by 0.5 Units by 1.5 Units by 3.0 Units up to 9.5
Unit Cost: 1.5 $/ M2 Unit Cost: 7 $/ M2 Unit Cost: 14 $/ M2 Unit Cost: 20 $/ M2

FIGURE 3 An example showing a set of M&R treatment activities.

Table 9 shows the priority programming of the network pavement M&R action plans for 5 years (from
1999 to 2004), together with Pavement Condition Index, PCI, of all pavement sections in each predicted
year, where annual budget of $3 million dollars is assumed. The optimization is formulated, as described
by Equations (4) to (8), to maximize the value of cost-effectiveness of the entire network on an annual
basis. The priority programming was performed by means of GAMS software application.
It is obvious that, from the annual M&R programming outputs, priorities for pavement treatments are
given to those sections which have lower PCI, but higher traffic volume. In other words, if two pavement
section have the same PCI but different traffic volume in a programming year, then a treatment priority
will be given to the one with higher traffic volume if the budget is available in that year.

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

TABLE 9 Example Sectional Data for An Asphalt Pavement Network

Hwy. Section No. 1999 2000 2001 2002 2003 2004 2005
No. (LHRS) PCI & M/R PCI & M/R PCI & M/R PCI & M/R PCI & M/R PCI&M/R PCI &M/R
79 37320 6.3 9.2 8.7 7.9 7.0 5.9 7.4
(4) (1) (1) (1) (3) (1)
25 25480 7.4 7.0 6.6 5.9 5.0 8.0 7.8
(1) (1) (1) (1) (4) (1)
28 26420 6.7 7.9 7.5 7.2 6.7 5.9 5.3
(3) (1) (2) (1) (1) (1)
11 17000 6.0 5.5 4.7 3.8 9.5 8.7 7.6
(1) (1) (1) (5) (1) (1)
132 44720 6.9 6.6 7.1 6.5 5.5 5.0 7.6
(1) (3) (1) (1) (1) (4)
1 10014 6.2 5.9 5.7 5.3 4.8 7.8 7.6
(1) (1) (1) (1) (4) (1)
7 14421 6.4 6.3 5.6 4.7 7.6 7.5 7.0
(1) (1) (1) (4) (4) (1)
2 10850 6.8 6.5 5.9 5.0 4.0 5.5 8.5
(1) (1) (1) (1) (1) (4)
402 48225 6.6 5.8 6.4 8.4 7.5 6.6 6.1
(1) (3) (4) (1) (1) (1)
60 33250 8.3 7.8 6.9 5.6 5.2 8.2 7.9
(1) (1) (1) (1) (1) (1)
31 27060 8.5 8.3 7.5 6.4 9.0 8.6 7.7
(1) (1) (1) (4) (1) (1)
138 45420 6.3 6.0 5.5 7.7 7.0 6.5 6.0
(1) (1) (3) (1) (1) (1)
36 28245 7.1 6.9 6.8 6.5 8.0 7.6 6.9
(1) (2) (1) (3) (1) (1)
148 46280 7.5 7.4 6.8 6.0 5.5 7.2 7.0
(1) (1) (1) (1) (3) (1)

SUMMARY AND CONCLUSIONS

This paper presents a practical optimization approach to programming network pavement maintenance
and rehabilitation treatment activities annually by means of dynamic prediction of pavement performance
with considering individual treatment effects. The optimization of pavement M&R treatment activities is
performed on cost-effectiveness basis. The areas between the performance or deterioration curve and the
minimum acceptable level are used for the calculation of cost-effectiveness. Pavements with different
maintenance and rehabilitation treatment alternatives are considered in the analysis of impact of different
budget scenarios on pavement performance condition. In addition, the integrated performance prediction
and preservation strategy process provides a means of incorporating the link between priority analysis and
budget constraints into the network M&R program optimization process.
Based on the example application described in this study, the most cost-effective maintenance and
rehabilitation program for preservation of a pavement network can be achieved through dynamic process
of pavement performance prediction and optimization programming. Pavement preservation strategy plan
prepared for individual pavement sections provide necessary information for prioritizing and optimizing
pavement preservation strategies.

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).
 5th International Conference on Managing Pavements (2001)

Determination of the optimal maintenance and rehabilitation (M&R) program for preserving a road
network above a certain serviceability level with a limited budget is critically dependent on pavement
performance prediction models. The whole process including analysis of sensitivities to various budget
levels and projected network pavement conditions corresponding to various budget levels is automated.
Application of the optimal M&R strategy selection integrating with time-related pavement performance
prediction models has been proposed for use in the Ministry's PMS-II.

REFERENCES

(1) Ningyuan Li, Ralph Haas, and Michel (1998), "Integer Programming of Maintenance and
Rehabilitation Treatments for pavement Networks", in Transportation Research Board 1632,
TRB, National Research Council, Washington, D.C.
(2) Lytton, R. (1987)"Concepts of Pavement Performance Prediction and Modeling", in the
Proceedings of Second North American Conference on Managing Pavements, Vol. 2, pp.
2.3-2.20, Canada.
(3) Chong, G., W. Phang and G. Wrong, (1989), "Manual for Condition Rating of Flexible
Pavements - Distress Manifestations. Report SP-024, R&D Branch, Ministry of Transportation,
Ontario.
(4) Chong, G. and G. Wrong (1995), "Manual for Condition Rating of Rigid Pavements", Report SP-
026, R&D Branch, Ministry of Transportation, Ontario.
(5) Hajek, J. and W. Phang (1989), "Prioritization and Optimization of Pavement Preservation
Treatments", Transportation Research Record 1216, TRB, National Research Council,
Washington, D.C., pp. 58-68.
(6) Ontario Ministry of Transportation (1990), "Pavement Design and Rehabilitation Manual",
Ministry of Transportation of Ontario, Canada.

TRB Committee AFD10 on Pavement Management Systems is providing the information contained herein for use by individual practitioners
in state and local transportation agencies, researchers in academic institutions, and other members of the transportation research
community. The information in this paper was taken directly from the submission of the author(s).

View publication stats