Professional Documents
Culture Documents
net/publication/355203904
CITATIONS READS
4 307
4 authors:
Some of the authors of this publication are also working on these related projects:
Feasibility study to assess the use of waste plastic aggregate in the design and performance of BC-I and BC-II mixtures for sustainable pavements View project
Development of long lasting bituminous mix for flexible pavements View project
All content following this page was uploaded by Akshay Waim on 15 October 2021.
TECHNICAL PAPER
Abstract
Pavement maintenance and rehabilitation to provide the desired level of service for the road users is the most challenging
problem faced by the authorities. The development of a reliable pavement deterioration model is essential to formulate the
appropriate maintenance policies. This exploratory paper presents the network-level pavement performance prediction models
for the district roads of Odisha state, India. Three cycles of pavement distress data were collected for developing the reli-
able pavement deterioration models using a multi-linear regression (MLR) analysis. The developed models can benefit the
authorities in predicting the pavement conditions within the network for building a comprehensive pavement management
system. The pavement condition index (PCI) model was developed using IRI, deflection, cracking, rutting, and raveling as
the distress parameters and validated using the deduct value method mentioned in ASTM D6433-18. The analytical hierarchy
process approach was used for ranking the pavement sections in estimating the maintenance costs consistently. Also, prior-
itization was done using the ‘VlseKriterijumska Optimizacija I Kompromisno Resenje’ (VIKOR) method, and the ranking
was rational compared to the PCI predicted ranking.
Keywords Deterioration models · VIKOR method · Analytical hierarchy process · Pavement management system · PCI ·
Deduct value method
Introduction the available budget [2]. PMS helps in estimating the present
performance and predicting the future condition of the pave-
India has the second largest road network in the world, span- ment [3]. As a part of PMS, various performance prediction
ning a total of 5.89 million km, transporting 64.5% of all models have been developed in the last few decades, of all
goods in the country. Nearly 90% of India’s total passenger the distress, poor riding quality takes the lead [4–7]. Among
traffic uses the road network to commute. The road transpor- the available technique, the use of network survey vehicle
tation in India has gradually increased over the years with an is found to be a suitable equipment for collecting the riding
improvement in connectivity between the cities, towns and quality data along with other distress data. The maintenance
villages [1]. Pavement deterioration is the formation of dis- activity will be scheduled based on a prioritization system
tresses in the pavement as a result of traffic and environmen- that takes into account the current roads condition in the
tal factors which may increase the road user costs. Highway network [8, 9].
authorities are finding it difficult to keep their deteriorating The pavement deteriorates due to traffic and climate, how-
roads in good condition due to budget constraints. There- ever, road geometry, pavement history, material properties,
fore, there is a need for a pavement management system construction methodology and thickness play an important
(PMS) to arrive at an optimal maintenance strategy within role [8, 10]. Hence, pavement age, traffic, and pavement dis-
tresses, viz. rutting, potholes, raveling, and patching, are all
* Sridhar Raju taken into account for developing and an accurate pavement
Sridhar.raju@hyderabad.bits-pilani.ac.in deterioration prediction model [11]. There are two types of
pavement deterioration models: deterministic and probabil-
1
Department of Civil Engineering, BITS Pilani Hyderabad istic models. The deterministic models can be further classi-
Campus, Hyderabad, India
fied into mechanistic, mechanistic-empirical and regression
2
Public Works Department, Government of Odisha State, models. The mechanistic models are developed considering
Bhubaneswar, India
13
Vol.:(0123456789)
28 Page 2 of 17 Innovative Infrastructure Solutions (2022) 7:28
the relationship between the stress, strain, and deflection, the IRI measurement can solely suffice for the develop-
while the mechanistic-empirical models are developed con- ment of a pavement condition index. In order to optimize
sidering the relationship between traffic, roughness, crack- the maintenance strategies and to extend the service life of
ing, and rutting. The regression models are developed con- a pavement, there is a scope for considering the structural
sidering the relationship between the pavement surface and parameter in terms of deflection in addition to the functional
structural condition [11, 12]. parameters.
The major drawbacks of probabilistic models are that The deflection data obtained from the falling weight
it does not accommodate the budget constraints and the deflectometer (FWD) can be considered for assessing the
pavement sections have to be grouped into a large number structural adequacy of the pavement. The deflection values
of homogeneous sections based on the physical characteris- measured using the geophones after normalizing for the
tics [12, 13]. Hence, there will be a large number of mainte- load, after correcting for temperature and subgrade mois-
nance and rehabilitation (M&R) treatments for each homo- ture can be used for back-calculating the layer moduli. The
geneous section. An individual pavement section’s condition static-linear analyses are used for interpreting the deflec-
cannot be evaluated using Markov’s process because the tion bowl results using the back-calculation computer pro-
pavement sections are analyzed as a group. Therefore, a grams. The calculated elastic layer moduli values are used
group of pavement sections rather than an individual section for the assessing the remaining service life and for providing
is used in the optimization programming of M&R strategies an overlay. Though the overlay can be designed using the
from the steady-state probabilities within a given budget. proven and reliable concept of structural number (SN), it
It usually takes many years to reach a steady state as some does not account for the mechanistic analysis. Therefore, the
portion of the pavement sections keep changing every year. FWD deflection values can be used for the design of overlay
Therefore, an objective function that uses steady-state prob- using the concept of mechanistic empirical pavement design
abilities in long transition periods does not fully reflect real- guidelines (MEPDG) [24, 25]. Due to the annual budget
ity [12–14]. limited for the execution of pavement maintenance and
In an empirical model, the dependent variable can be rehabilitation projects, the highway agency has to choose
related to one or more independent variables such as the the alternatives with higher priority first, in order to supply
age of the pavement, applied loads, pavement layer thick- the demands of optimization in the network. The develop-
ness, etc., through a regression analysis. Most of the mod- ment of deterministic deterioration model provides insight
els available in literature are the linear combinations of the into the factors which affects the deterioration process, final
regressions that are already available. In an advanced empir- developed model is very user-friendly—relatively easy to
ical model, the specification forms attempts to simulate the understand. As on date there is no provision for including
physical process of deterioration. The advanced empiri- the deflection values as a part of PCI, though there is scope
cal model can be developed with a reasonable reliability for inclusion. The inclusion of deflection values in the PCI
only after generating the long-term pavement performance models will help in assessing and ranking the pavement on
database. The size of the database that was used in their the basis of both the functional and the structural param-
development limits the implementation of these empirical eters. As a result, the standardization of the PCI method and
models. As a result, they should not be used when the actual its classification-based approach would make it difficult to
conditions are different. One of the best-known examples keep up with future technological advances. However, the
of empirical models is that developed by the World Bank above-mentioned challenge can be addressed by modifying
and given in the HDM-4 manual. The empirical models in the PCI method using a general multi-domain prioritization
line with HDM-4 are the globally accepted models for PMS approach, AHP combined with VIKOR method for the appli-
applications [15]. The World Bank’s HDM-4 model is one cation of the highway engineers.
of the most well-known examples of regression models but This study was focused on developing deterministic pave-
can be used for other conditions after calibration [16–18]. ment deterioration models at a network level for prioritizing
Recently, the PCI model was developed by considering age, the pavement for maintenance. To avoid a tiresome method
traffic, cracking area using the MLR method for the LTPP of arriving at the deduct value, the curves given in ASTM
data by categorizing it into four climatic conditions [19]. D6433-18 were digitized using Engauge digitizer program.
In addition, there is a PCI method which majorly con- The pavement deterioration prediction models were devel-
siders the visual observations with international roughness oped in line with the HDM-4 models (calibrated for the local
index (IRI) as the main criteria for improving the riding conditions). Also, the FWD deflection values as a pavement
quality [20]. Most of the research studies have pointed out structural parameter were considered for the development of
that the progression of distresses mentioned in the PCI pavement condition index (PCI). The moving traffic simulat-
method has direct relationship with the progression of IRI ing FWD deflection values as one of the parameters in the
[8, 21–23]. Therefore, as a part of the functional evaluation deterministic PCI model will help in indirectly predicting
13
Innovative Infrastructure Solutions (2022) 7:28 Page 3 of 17 28
the structural condition of the pavement. The analytical hier- pavement deterioration models followed by ranking of the
archy process (AHP), an advanced pavement maintenance twenty sections. The twenty road sections selected from
rating technique, was adopted in this study for a network Balasore, Sonepur and Sambalpur districts of Odisha were
level PMS. The weightages for distresses, viz. IRI, crack- found to be homogeneous in terms of traffic and climatic
ing, rutting, deflection and raveling, were based on expert conditions, including the bituminous surface layers. Figure 1
opinion, and a pairwise comparison. The VIKOR method depicts the methodology adopted in the present research.
uses the weightages obtained from AHP as an input to give AASHTO cumulative difference approach (CDA) method
a compromised ranking without any conflict but close to an was followed to arrive the homogenous sections [26]. The
ideal solution. The proposed methodology aims to arrive pavement deterioration models were developed in line with
at the decisions for the road network in Odisha, India. The HDM-4 models, but by incorporating changes for suiting
main objectives of this study were: the Indian conditions. IRI, deflection, cracking initiation,
cracking progression, rutting initiation, rutting progression
i. To arrive at the PCI value in accordance with ASTM and raveling progression models were developed by consid-
D6433-18 ering traffic and age of the pavements by fitting nonlinear
ii. To develop the performance prediction models for thin regression models in Minitab statistical software. Apart from
bituminous surfaces this, from the observed distress data to predict the pavement
iii. To determine the weightages of the influencing dis- condition over period of time PCI was calculated by using
tresses using AHP ASTM D6433-18 deduct value method [20]. Later, the pave-
iv. To prioritize the road maintenance within a network ment deterioration models adopted in HDM-4 models were
using VIKOR method calibrated as per the procedure given in the HDM-4 manual
for local conditions. Level-2 calibration method was adopted
since the data collected was for three years. The calibration
Methodology process consists of determining the adjustment factors (ki)
which will attain the good agreement between the HDM-4
This section deals with the methodology followed to develop model predictions and the field data [10, 27].
the pavement deterioration models followed by ranking of The procedure suggested for calibrating surface dis-
the twenty sections. The basis for the methodology was tress initiation factors is based on obtaining the coefficient
explained in this section. The Analytical Hierarchy Process between the observed years of occurrence of the distress
and VIKOR method were used for priority ranking of the to the year of occurrence as predicted by the uncalibrated
road pavements for maintenance within a road network in models. In the case of the progression factors, the calibration
Odisha state, India. The following fundamental parameters is carried out by minimizing the squares of the differences
were used for the prediction of pavement deterioration and of the estimated and observed data or Sum of Squared Dif-
the pavement condition index models as a part of Pavement ferences (SSD). Minimizing the SSD value assists in reduc-
Management System (PMS). The primary objective of the ing the estimated average error, which in turn will make it
PMS is to avoid judgmental bias in the decision-making pro- possible to locate the calibration factor that ensures the best
cess by considering the pavement distresses and responses. adjustment of the distress curve and therefore the calibration
The performance prediction models play the key role in pri- of the analyzed performance model. An analysis of verifi-
oritization and maintenance of pavements. A large amount cation of prediction quality of model is also performed in
of data and resources including advanced algorithms are the study by comparing observed versus predicted distresses
required for prioritization of the pavements. The pavement values for each of the model. The statistical indicators used
condition surveys, deflection and the IRI data along with for comparison of models were R2 and RMSE [10, 18, 27].
pavement history are the essential parameters for the devel- The weightages for distresses, viz. international rough-
opment of pavement deterioration models and pavement ness index (IRI), cracking, rutting, deflection and raveling,
condition index model. The pavement deterioration mod- were obtained by adopting AHP. To determine the weight-
els in line with the widely accepted HDM-4 models (cali- ages of the distresses, 200 PMS experts such were opined
brated for the local conditions) after consistent assumptions through expert opinion survey. Pairwise comparison was
will reflect the qualitative results for implementation of the done between the criteria by considering the matrix. From
PMS. The highway agencies generally adopt the pavement the matrix, Eigenvector values were calculated correspond-
maintenance priority in the form of pavement condition ing to maximum Eigenvalues. Thereafter, random index,
index (PCI) which is computed using an empirical expres- consistency index, and consistency ratio were evaluated.
sion. The pavement deterioration and PCI models have to Later, if consistency ratio is between 0 and 0.1, report the
be simple and easily understood by the road agencies. This weightages of distresses and their ranking. The pavement
section deals with the methodology followed to develop the prioritization for maintenance within the road network was
13
28 Page 4 of 17 Innovative Infrastructure Solutions (2022) 7:28
• Pavement age, pavement distresses as an Input for all the pavements in an excel sheet
• Determination of the distresses using HDM-4 models
• Calculation of RMSE and R2 for kmin to kmax and compare with each increment
Calibration • Arriving at the calibration coefficients (ki) at minimum RMSE and maximum R2
of HDM-4
models
based on the ranking using VIKOR method. The distress vehicles per day (CVPD). This study examined at twenty
weightages were given as input to VIKOR method for pri- separate MDRs’ from three districts of Odisha with a total
oritizing the total 20 sections. Calculation of regret measure length of about 251 km. The data was collected during the
(R), VIKOR index (Q), and utility measure (S) and rank the years 2014, 2016, and 2018. Figure 2 shows the location of
sections in ascending order completes the prioritization of the districts considered for the study analysis.
the road sections integrate with AHP. The methodology is The pavement composition consisted of 25 mm semi-
divided into different parts to arrive the above-mentioned dense bituminous concrete (SDBC) as a surface layer,
objectives. The subsequent sections will explain for each 50 mm bituminous macadam (BM) as a binder layer,
step. 150 mm wet mix macadam (WMM) as a granular base and
100 mm granular sub-base, over a compacted sub-grade
with 4% California bearing ratio (CBR). The pavement
Data collection sections were rehabilitated and upgraded during March
2013 and hence considered as new pavements. Table 1
The data was collected from the major district roads (MDR) gives the road length and the annual average daily traffic
connecting two major districts or district to a highway in (AADT) in terms of number of commercial vehicles per
Odisha state, India. Though the MDRs considered were day (CVPD) was considered. The design traffic in terms
low-volume roads but were significantly contributing to the of cumulative standard axles (CSA) in millions has been
industrial development of Odisha state. MDRs considered considered in the pavement deterioration prediction mod-
were having traffic in the range of 1200 to 1800 commercial els. The design traffic also considers the vehicle damage
13
Innovative Infrastructure Solutions (2022) 7:28 Page 5 of 17 28
factor (VDF) which indirectly takes in to account the axle Pavement deflection data
load spectrum. The design traffic in terms of CSA in mil-
lions has been calculated as suggested by the Indian Roads The Falling Weight Deflectometer (FWD), Dynatest model
Congress the guidelines for the design of flexible pave- 8000 with a 300 mm diameter loading plate, was used for
ments [28]. collecting the deflection data. The geophones were placed at
13
28 Page 6 of 17 Innovative Infrastructure Solutions (2022) 7:28
Table 2 NSV Sample data of km International Rough- Cracking (%) Raveling (%) Pot holes Rutting (mm) Texture
S-1 ness Index (m/km) (no) depth
(mm)
0, 200, 300, 600, 900, and 1200, 1500 mm from the center initiation, cracking progression, rutting initiation, rutting
of the loading plate. Three drops of mass were considered progression and raveling progression models are given
at each testing point with loading time ranging from 0.015 through Eqs. 1 to 7, respectively.
to 0.050 s, while the first drop was ignored, considering it IRI progression model
to be a seating load. When the pavement was found to be in
(IRI)t = 0.156 ∗ (IRI)t−1 − 0.616age + 0.5231 ∗ CSA + 1.185
good condition, the data was collected at 500 m interval,
(1)
but when the pavement was found to be in poor condition,
the number of testing points was increased by reducing the n = 358, R2 = 0.981, RMSE = 0.605.
testing interval. The deflection data obtained were normal- Deflection progression model
ized with respect to a standard load of 40 kN, and corrected 0.0335
to a standard temperature of 35 °C. In addition, the seasonal (Def)t = 0.018 ∗ (Def)t−1 −
Age
+ 0.0179 ∗ CSA + 0.146
correction was considered for sub-grade moisture. The FWD (2)
analysis was carried out in accordance with the guidelines
n = 358, R2 = 0.996, RMSE = 0.005.
set out by the Indian Roads Congress [29].
Cracking initiation model
( )
CSA
Model development CRinitiation = 2.56 ∗ exp ∗ −2.47 ( )−0.63 (3)
3.2 ∗ D0
Homogenous sections
n = 50, R2 = 0.852, RMSE = 0.147.
Cracking progression model
Road agencies have to cope with large amounts of data, and
they need to segment a measurement series automatically. (CR)t = 0.024 ∗ (CR)t−1 + 1.082age + 1.287 ∗ CSA + 0.542
Typical measurement series contain measurements from sev- (4)
eral homogeneous road sections with more or less abrupt n = 358, R2 = 0.987, RMSE = 0.495.
transitions from one to another [30]. The pavement design Raveling initiation model
guide published in 1986 by the American Association of
State Highway Officials (AASHTO) suggests the cumulative Ravinitiation = 1.18 ∗ CSA ∗ (CQ + 1)−0.15 (5)
difference approach (CDA) to segment roads. This is mostly
a graphical method which facilitates the detection of homo- n = 58, R2 = 0.712, RMSE = 0.256.
geneous sections [30]. In the present study CDA is used to Rutting progression model
delineate the pavement sections. The total data was collected 4.746
for 251 km. Based on the IRI from Fig. 3a, b shows along (Rut)t = 0.102 ∗ (Rut)t−1 − + 1.523 ∗ CSA + 0.446
Age
the distance CDA was applied for the total pavements and it (6)
was divided into different segments where the slope changes.
n = 358, R2 = 0.989, RMSE = 0.597.
Raveling progression model
Development of deterioration prediction models
0.0919
(Rav)t = 0.2 ∗ (Rav)t−1 − + 0.0491 ∗ CSA + 0.556
For the development of distress prediction models, Age
inventory data, functional data, and structural data were (7)
collected and compiled. The IRI, deflection, cracking n = 358, R2 = 0.993, RMSE = 0.015.
13
Innovative Infrastructure Solutions (2022) 7:28 Page 7 of 17 28
1500
1000
500
0
0 50 100 150 200 250 300 350 400
Distance (km)
(a) CDF for IRI along the distance for homogeneous sections
20.0
S-1
S-8 S-20
18.0
16.0
14.0
12.0
IRI (m/km)
10.0
8.0
6.0
4.0
2.0
0.0
0 50 100 150 200 250 300 350 400
Distance (km)
Calibration of HDM‑4 deterioration prediction calibration factor, ki, must be calculated for each of the dis-
models for local conditions tresses considered in a study, as well as for each road section
within a network. The calibration factors for each distresses
Figure 1 shows the methodology for arriving at the calibra- and all the road sections are given in Table 3.
tion factors, ki. In the present study, the MATLAB program
developed can be used for the determination of the HDM-4 Pavement condition index (PCI) model
calibration factors by considering the field observed distress
and age data for the pavements considered in the analysis. PCI calculation using deduct value method (ASTM
The MATLAB program shall be used in two steps for cali- D6433‑18)
bration of HDM-4 deterioration models. In the first step, the
program shall be executed for varying calibration factors Considering the distress density, the PCI was calculated
from kmin = 0.1 to kmax = 20, at an increment of 0.1. Once using the deduct value method in accordance with ASTM
the calibration factor pertaining to minimum RMSE and D6433-18. To calculate the distress density, a simple
maximum R2 has been determined from the first step, the MATLAB program was developed as a part of this study.
program should be run again for the second time, using kmin The curves given in ASTM D6433-18 were digitized using
and kmax values within the calibration factor’s closer range Engauge digitizer program to obtain the mathematical
as determined from the first run, and at a smaller increment equations and to calculate the deduct value for all pave-
of 0.01, for refinement (ki) of the calibration factor [15]. The ment distresses considered. The relationship between the
13
28 Page 8 of 17 Innovative Infrastructure Solutions (2022) 7:28
Table 3 Calibration factors for Section name Crack initia- Crack progres- Raveling progres- Rutting progres- Roughness
different road sections tion (Kci) sion (Kcpa) sion (Krap) sion (Krp) progression
(Kgp)
deduct value and the distress density (%) was plotted for distresses at low, medium and high severity were obtained
all the distresses, and Fig. 4 shows an example plot for by digitizing the distresses curves given in ASTM D6433-
alligator crack. The plot for alligator crack distress density 18 and were used for calculating the PCI
is shown in Fig. 4a, while the digitized plot is shown in
y = 36.006x4 − 125.48x3 + 162.4x2 − 83.256x + 20.155, R2 = 0.998
Fig. 4b. The mathematical model to calculate the deduct
(8)
value for medium severity alligator crack density is given
in Eq. 9. Similarly, the mathematical models for all other where y = Deduct value and x = logarithmic value of alligator
cracking density in percentage.
Medium
High Medium Low 100
100
90
90
80
80
70
70
Deduct value
60
Deduct value
60
50
50
40
40
30 30
20 20
10 10
0 0
0.1 1 10 100 0.1 1 10 100
Distress density (%) Distress density (%)
(a) Alligator cracking chart from ASTM D6433-18 (b) Alligator cracking chart from digitizer
Fig. 4 a Alligator cracking chart from ASTM D6433-18. b Alligator cracking chart from digitizer
13
Innovative Infrastructure Solutions (2022) 7:28 Page 9 of 17 28
13
28 Page 10 of 17 Innovative Infrastructure Solutions (2022) 7:28
13
Innovative Infrastructure Solutions (2022) 7:28 Page 11 of 17 28
13
28 Page 12 of 17 Innovative Infrastructure Solutions (2022) 7:28
Table 5 Weightages obtained by AHP method Step 1 The excel sheet consisting of IRI, cracking, deflec-
Category Distresses Weightages Order of tion, rutting, raveling (decision matrix) data is given as an
preference input file in MATLAB.
Step 2 The input data was normalized using the equation,
1 IRI 0.461 1
[ ]
2 Deflection 0.207 2 F = fij m∗n (10)
3 Rutting 0.151 3
4 Cracking 0.145 4 where
5 Raveling 0.036 5 xij
fij = �
∑m (11)
x2
i=1 ij
13
Innovative Infrastructure Solutions (2022) 7:28 Page 13 of 17 28
S− = min Si (19) 0.01 (α = 0.01). From the table it can be found that the high-
i
est signed ranking was 13 and hence considering ‘n = 13’
from the WSR table Wcritical was obtained as 9. The W value
R∗ = max Ri (20) of 41 was greater than the Wcritical of 9 and hence, it can
i
be inferred that the rankings were similar. This shows that
R− = min Ri VIKOR method is equally efficient in prioritizing the pave-
(21)
i
ments sections in addition to its other benefits as mentioned
The term “ v ” is introduced as the weightage of the maxi- above and, hence, can be considered as a robust method for
mum group utility. It ranges between 0 and 1 and is based on ranking of pavements for maintenance.
the level of compromise among the decision-makers. Sensitivity analysis was carried out by varying the risk
Step 6 Rank the pavement sections, by sorting the values factors from 0 to 1 with an increment of 0.1 in the pavement
of S, R, and Q, in ascending order. condition rankings. Table 7 shows the ranking of all the road
sections using VIKOR and PCI methods. The sensitivity
Sensitivity analysis analysis results showed that the priority ranking of the road
sections showed no difference when the risk factors were
To statistically evaluate if the ranking from both the pro- 0.1 to 0.7. The priority ranking marginally varied when the
cesses is similar, the Wilcoxon Signed Rank (WSR) test was risk factors were 0.8 to 1.0, demonstrating the insignificance
carried out and the results are shown in Table 6. For the of risk factor in prioritizing the road sections. Though the
rankings from the two processes to be similar, the W value coefficient of IRI was higher in PCI method and the IRI
shall be greater than Wcritical value. The W value is the mini- weightages were higher in VIKOR method, the ranking was
mum of the W + and W − wherein the W + stands for the sum different (see Table 7). When there are ‘n’ number of sec-
of the positive ranks while the W − stands for the sum of the tions with PCI values either same or very close to each other,
negative ranks. The values of W + and W − were 50 and 41, VIKOR method will be a handy tool for pavement main-
respectively. The Wcritical value was obtained from the WSR tenance as this will make a compromised ranking without
two tailed statistical test table [38] at a significance level of conflicts.
Table 6 Wilcoxon signed rank VIKOR PCI ranking Difference Absolute dif- Ranking Signed ranking
test results ranking ference
3 1 2 2 6.5 6.5
2 2 0 0 n/a N/A
4 3 1 1 2.5 2.5
1 4 − 3 3 10 − 10
5 5 0 0 n/a N/A
6 6 0 0 n/a N/A
7 7 0 0 n/a N/A
8 8 0 0 n/a N/A
19 9 10 10 13 13
13 10 3 3 10 10
10 11 − 1 1 2.5 − 2.5
12 12 0 0 n/a N/A
15 13 2 2 6.5 6.5
11 14 − 3 3 10 − 10
14 15 − 1 1 2.5 − 2.5
9 16 − 7 7 12 − 12
17 17 0 0 n/a N/A
16 18 − 2 2 6.5 − 6.5
20 19 1 1 2.5 2.5
18 20 − 2 2 6.5 − 6.5
W-value 41
W-critical value 17
13
28 Page 14 of 17 Innovative Infrastructure Solutions (2022) 7:28
The methodology in this study was mainly focused on (DBM) with 30 mm bituminous concrete (BC) as surface
arriving at the ranking for pavement maintenance based on layer for the pavement sections S-12, S-18, S-17, S-9, S-8,
the present condition. The ambiguity around pavement sec- and S-16. The maintenance measure predicted using VIKOR
tions with closely spaced PCI ratings can be eliminated by method considers the above-mentioned pavement sections
AHP integrated VIKOR method. Additionally, the rankings as the top priority for maintenance. The PCI predicted value
with VIKOR method were statistically similar to that of was 41.68 in the year 2021 for pavement Sect. 12 (S-12) and
PCI method. Thus, it can be inferred that VIKOR method is it was marginally above the terminal PCI value of 40 (see
qualified in providing the ranking even with one cycle dis- Table 8). Similar tables were created for other pavement
tress data, while precisely sorting them which would have sections, and the rankings predicted by VIKOR method and
otherwise proven to be ambiguous in certain cases. However, those proposed by the Odisha road agency engineers were
a minimum of 3 consecutive cycles’ data is required for the identical. Hence, the road agency may consider the predicted
development of a PCI model. For the practical purpose, it is rankings using VIKOR method as the basis for proposing the
important to rank the sections based on the type of interven- pavement maintenance for budget approvals.
tion. As the data were collected for the first time to arrive
at a network level the PMS, the priority ranking were con-
sidered based on the present pavement condition. With the Conclusions
present condition ranking, the road authority will be able to
demonstrate to the policy makers on the saving in vehicle In this study, the functional and structural behavior of
operation cost by the road user considering various surface pavement has been used for developing the pavement
treatments or overlay or rehabilitation. The road agency will performance models using a deterministic approach. The
be able to arrive at the budget requirement to maintain a developed multi-linear regression method enables the deci-
roadway system at a desirable PCI, where the overall life sion-makers to assess the existing pavement condition and
cycle cost will be reduced. to arrive at the maintenance measures using the PCI as an
Based on their field experience, the road agency engi- intervention criterion. In the present study, AHP has been
neers have prepared and submitted a proposal for budget integrated with the VIKOR method for ranking of the road
approval to the state Government. They have proposed to pavement sections. The AHP method changes the order of
provide an overlay of 50 mm dense bituminous macadam judgement alternatives which is called as a rank reversal and
13
Innovative Infrastructure Solutions (2022) 7:28 Page 15 of 17 28
Table 8 Distress prediction for Year IRI (m/km) Cracking (%) Raveling (%) Rutting (mm) Deflection (mm) PCI
the next 15 years for section
S-12 2013 3.50 0.00 0.00 0.0 0.11 85.14
2014 3.71 0.50 0.27 1.5 0.16 82.08
2015 3.94 1.20 0.97 1.9 0.22 79.34
2016 4.25 2.10 1.87 3.1 0.30 75.13
2017 4.66 3.32 3.09 5.5 0.35 68.83
2018 5.22 5.68 5.45 7.1 0.39 60.30
2019 5.71 7.26 5.82 8.1 0.45 54.98
2020 6.55 8.34 6.11 8.3 0.50 49.66
2021 8.24 8.51 6.54 8.6 0.68 41.68
2022 8.86 8.80 7.24 9.1 0.75 37.73
2023 9.20 9.10 7.68 9.9 0.77 34.90
2024 10.12 9.80 8.26 10.6 0.81 29.00
2025 10.55 10.34 9.86 11.2 0.89 24.78
2026 11.16 12.50 10.25 12.4 0.96 18.08
this could be one of the limitations as the order of ranking 4. The AHP process was used to calculate the weightages
will be changing even with a minor change in the data. When for IRI, cracking, rutting, raveling, and deflection, and
there are more number of levels in the hierarchy, it results it was found that the IRI had the highest weightage of
in higher pair comparisons thereby consuming more process 0.461, followed by the deflection of 0.207. In most of
time and this is another limitation of the AHP method. How- the studies, the IRI is the only parameter considered for
ever, this limitation can be overcome by using the computer determining the maintenance procedure. Though IRI
aided programs for AHP to make it easy, fast and precise. was found to have the highest weightage, similar to other
Though there are limitations in the AHP method, it consid- studies, but considering deflection and other pavement
ers either the quantitative or the qualitative information for distresses in pavement prioritization was a value addi-
decision-making process. As the distresses considered in this tion. As a result, it can be inferred that the AHP method
study were only five in umber and hence, the limitations will be a useful tool for ranking the pavement sections
mentioned above can be overcome and also, with use of in order to reliably estimate the maintenance costs.
computer applications the analysis time was fast with a reli- 5. The sensitivity analysis was conducted on all the twenty
able weightage. Though the uncertainty was not considered roads by changing the risk factor,v from 0.1 to 1.0, and
in this study, the PCI model was validated to be reliable it was found that there was no substantial improvement
through an independent t-test for developing a network-level in the ranking of the road pavements with the change
deterministic model for assessing the variations in PCI with in risk factor. Therefore, it can be concluded that a risk
respect to time in years. The following conclusions were factor of 0.5 can be used for ranking the road pavements
drawn from the present research: in a network. Using the VIKOR method, the priority
ranking was found to be similar to the predicted PCI
1. Since collecting structural and functional data at the net- ranking.
work level of road projects is expensive, the progression 6. The ambiguity around pavement sections with closely
models developed in this study from three cycle data spaced PCI ratings was eliminated by VIKOR method.
can be used to select the optimum maintenance measure Additionally, the rankings with VIKOR method were
for the pavement’s life cycle while avoiding the costs of statistically similar to that of PCI method. Thus, it can
repeated data collection. be inferred that VIKOR method is qualified in providing
2. Using a multi-criteria decision-making approach, AHP the ranking while precisely sorting them which would've
in combination with the VIKOR method was found to otherwise proven to be ambiguous in certain cases.
be a useful tool for priority ranking the pavements in a 7. The pavement sections S-12, S-18, S-17, S-9, S-8, and
network. S-16 were selected for maintenance by the road agency
3. It is highly cost-effective to homogenize the pavement for the year 2021, and the ranking predicted using
sections within the network on basis of IRI, traffic, pave- VIKOR method was identical. Therefore, the road
ment composition, and environment to improve the agency may utilize VIKOR method to rank the pavement
accuracy of prediction in the developed deterioration sections in the network, which may then be adopted for
models. pavement maintenance and budget approvals.
13
28 Page 16 of 17 Innovative Infrastructure Solutions (2022) 7:28
13
Innovative Infrastructure Solutions (2022) 7:28 Page 17 of 17 28
Oper Res 156:445–455. https://doi.org/10.1016/S0377-2217(03) 37. Babashamsi P, Golzadfar A, Yusoff NIM et al (2016) Integrated
00020-1 fuzzy analytic hierarchy process and VIKOR method in the prior-
35. Opricović S (2009) Compromise in cooperative game and the itization of pavement maintenance activities. Int J Pavement Res
VIKOR method. Yugosl J Oper Res 19:225–238. https://doi.org/ Technol 9:112–120. https://doi.org/10.1016/j.ijprt.2016.03.002
10.2298/YJOR0902225O 38. Taheri SM, Hesamian G (2013) A generalization of the Wilcoxon
3 6. Kastratović E, Zimonjić S, Đekić M (2017) Application of VIKOR signed-rank test and its applications. Stat Pap 54:457–470. https://
method in ranking the investment projects. J Econ Law 22:1–10 doi.org/10.1007/s00362-012-0443-4
13