You are on page 1of 15

An analytical

The use of an analytical hierarchy


hierarchy process in pavement process

maintenance priority ranking


25
Rezqallah H. Ramadhan, Hamad I. Al-Abdul Wahhab and
Salih O. Duffuaa
King Fahd University of Petroleum and Minerals, Dhahran,
Saudi Arabia

Keywords Analytical hierarchy process, Priority rules, Saudi Arabia


Abstract This paper describes the use of an analytical hierarchy process (AHP) in determining the
rational weights of importance of pavement maintenance priority ranking factors. These weights
were obtained by capturing the local peoples perception towards this vital part of the pavement
management system (PMS). In this regard, different groups of individuals were asked to estimate
the weight of importance in pavement maintenance of different factors for ranking pavement
sections. These factors were road class, pavement condition, operating traffic, riding quality, safety
condition, maintenance cost, and the overall importance of the road section to the community. The
AHP method of pair-wise comparison was employed to get the factor weights, which were compared
with the weights obtained from the direct assignment method. It was concluded that the two
methods were statistically similar which confirms that the results of the direct assignment method
can be used safely with a sound reliability and consistency. This conclusion comes from the fact that
the AHP method has a high reputation and applications, and it uses a high-precision technique for
obtaining the weights (priorities) of alternatives or items. Priority factor weights were used in
developing a pavement maintenance priority ranking procedure for a road network. This procedure
was validated by real case studies, and found to be logically and efficiently able to handle the ranking
of a huge number of pavement sections for maintenance and repair.

Introduction
Application of maintenance actions at the right time results in the saving of
considerable maintenance funds. This is due to the fact that pavements can be
restored, with limited funds, to a near perfect level of service, when of a fair
condition, compared to the cost needed to bring the pavement to the same level,
when the pavement is allowed to deteriorate to a poor condition. This right
maintenance timing can be captured if pavements are evaluated periodically and
managed in a rational systematic manner. One of the main activities of any
pavement management system (PMS) is maintenance priority ranking. The
priority setting or ranking process, as used in PMS, aims to rank pavement
sections in an order of urgency for maintenance and repair. The importance of the
prioritization process in PMS comes from the fact that it is the main step before

The authors wish to acknowledge the Civil Engineering Department and Research Institute of
King Fahd University of Petroleum and Minerals for supporting this research. Acknowledgement Journal of Quality in Maintenance
Engineering, Vol. 5 No. 1, 1999,
is also due to the Ministry of Communication, Municipality of Riyadh, and Municipality of pp. 25-39, MCB University Press,
Dammam in Saudi Arabia, for their cooperation in data collection. 1355-2511
JQME the final decision on maintenance program execution. The quality of priority
5,1 setting can directly influence the effectiveness of available resources that are, in
most cases, the primary judgment of the decision maker (Sharaf, 1993).
Priority analysis is a systematic process that determines the best ranking list
of candidate sections for maintenance based on specific criteria such as
pavement condition, traffic level, pavement function, etc. Various methods and
26 schemes are used for priority analysis ranging from simple listing based on
engineering judgment to true optimization based on mathematical
formulations. Examples of these different methods are detailed in Haas et al.
(1994). These methods can be divided into two broad groups:
(1) ranking methods; and
(2) optimization methods.
The ranking methods are subdivided into those based on a composite index
determination and those based on economics (Karan, 1984).

Composite index ranking method


The composite maintenance priority index was the method selected for this
study. This method is relatively simple and gives close to optimal results, and it
has been applied by many researchers (Chen et al., 1993; KFUPM-RI, 1989;
Nigyuan and Haas, 1994; Sharaf, 1993). Different approaches can be used to
develop a priority-combined index for pavements. These approaches include
the unique sum approach, utility theory, Delphi method, factorial rating
method, and fuzzy set theory (Zhang et al., 1993).
The composite index method considers several factors that affect the priority
rank. Each of these factors has a specific weight of importance to the priority
rank. The priority index (PI) used in this study for pavement maintenance
priority has the following form:
(1)

where:
PI = Priority index for any section (out of 100);
Wj = Factor j weight of importance to priority ranking;
Fj = Factor j value (out of 100); and
Wj = 1.0.

The analytical hierarchy process method


One of the methods used to scale and quantify measurements is the analytical
hierarchy process (AHP). It is an approach, developed by Saaty in the 1970s, for
dealing with complex technological, economic, and sociopolitical problems.
AHP aims at quantifying relative priorities for a given set of alternatives on a
ratio scale (Saaty, 1990; Saaty and Vargas, 1982).
A useful feature of the AHP is its applicability to the measurement of
intangible criteria along with the tangible ones through a ratio scale. In AHP, a
ratio scale between one and nine is used to give the relative preference between An analytical
two alternatives. This scale is able to capture a great deal of information and hierarchy
has proven to be extremely useful due to the fact that the AHP is somewhat process
scale independent (Harker and Vargas, 1987). This scale can be defended for the
following reasons (Saaty, 1990):
(1) The human ability to make qualitative distinctions is well represented by
five attributes: equal, weak, strong, very strong, and absolute. 27
Compromises between adjacent attributes can be made when greater
precision is needed. The totality requires nine values and they may be
consecutive.
(2) There is a practical method often used to evaluate items by classifying
the stimuli into a trichotomy of regions: rejection, indifference, and
acceptance. For finer classification, each of these is subdivided into a
trichotomy of low, medium, and high, resulting in nine shades of
meaningful distinctions.
(3) The human brain has a psychological limit for simultaneous
comparisons of 7 2 items. This mind capacity has something to do
with the number of fingers. Therefore, a nine-point scale will be sufficient
to do the comparisons between items.
The AHP incorporates judgments and personal values in a logical way. It depends
on imagination, experience, and knowledge to structure the hierarchy of the
problem. It also depends on logic, intuition, and experience to provide judgments.
Once accepted and followed, the AHP shows how to connect elements of one part
of the problem with those of another to obtain the combined outcome (Saaty, 1982).
The AHP differs from other conventional scoring methods in the following
(Hagquist, 1994):
(1) AHP uses a set of one-to-one comparisons to evaluate alternatives under
each criterion. These pair-wise comparisons are the smallest in
decisions.
(2) AHP uses one-to-one comparisons to assign criteria importance weights.
(3) AHP does alternative comparisons and criteria weighting in separate
steps.
(4) AHP combines both objective measures and subjective preferences in the
form of criteria weights. Typically, only the objectives (facts) are
quantified.
AHP has many advantages over conventional scoring methods such as an
increase in accuracy and consistency, and that the subjective consideration is
quantified in a structured framework. However, the major drawback in the use of
AHP is the effort required to make all pair-wise comparisons (Millet and Harker,
1990). As the size of the hierarchy increases, the number of required pair-wise
comparisons increases exponentially. Also the AHP is complex in terms of a
higher level of detail required by the evaluators when asked for their preferences
JQME and opinions. In pavement management, usually many factors are considered in
5,1 any modeling process (e.g. prioritization), and in a pavement network there are a
huge number of sections considered for evaluation. These facts make the
sections comparison based on a nine-point scale for several criteria rather a
difficult task. However, for project-level evaluations where a few sections are to
be considered simultaneously, AHP is an effective method for analysis.
28 Another contribution in using AHP in pavement prioritization is the work
done by Cook and Kress (1994). They developed a multiple criteria composite
index for evaluating a set of alternatives relative to a combination of ordinal
(qualitative) and cardinal (quantitative) criteria. This model has been
incorporated in a menu-driven software package called a multi-attribute ranking
system (MARS). The basic feature of this model is that, for a given number of
alternatives N (e.g. projects or sections) to be evaluated, a relative set of K1
ordinal and K2 cardinal criteria are established. In this modeling several factors
were considered such as the pavement condition index (PCI), present traffic,
predicted future traffic, percentage of commercial vehicles, accident level, vehicle
operating costs, and riding disruption. In all but the last two criteria, numerical
values are available. Since roads can be ranked according to riding comfort or
disruption, and since the data regarding vehicle operating costs are difficult to
obtain, the last two criteria are judged to be ordinal (qualitative).

Principles and background of AHP (Saaty, 1990)


When a group of activities (factors) are considered for evaluation by a group of
people, the main objectives of this group are:
(1) to provide judgment on the relative importance of these activities; and
(2) to insure that the judgments are quantified to an extent which also
permits a quantitative interpretation of the judgment among these
activities (factors).
The quantified judgment on pairs of activities ci, and cj are represented by an n-
by-n matrix
(2)
The entries aij are defined by the following entry rules:
Rule 1. If aij = , then aji = 1/, 0.
Rule 2. If C i is judged to be of equal relative importance as C j , then
aij = aji = 1. Obviously aii = 1 for all i.
Thus the matrix A has the form:

(3)
where the aij is the relative importance of activity i to activity j. Having recorded An analytical
the quantified judgments of comparisons on pairs (Ci, Cj) as numerical entries aij in hierarchy
the matrix A, what is left is to assign to the n contingencies C1, C2 , C3 , ..., Cn a set process
of numerical weights w1, w2, w3 , ..., wn that should reflect the recorded judgments.
The eigenvector of the comparison matrix provides the priority ordering
(weight), and the eigenvalue is a measure of consistency. To find the priority
vector or the weight of each factor included in the priority ranking analysis, the 29
eigenvector corresponding to the maximum eigenvalue is to be determined from
matrix analysis. One of the approximation methods to get the weight of each
factor in the pair-wise comparison process is described below.

Weight vector calculation


In mathematical terms, the principle eigenvector is computed, and when
normalized becomes the vector of priorities (weights). To reduce the excessive
computing time needed to solve the problem exactly, and due to the results of
complex numbers, a good estimate of that vector can be obtained by dividing
the elements of each column in the comparison matrix by the sum of that
column (i.e. normalize the column). The elements in each resulting row are
added and the sum is divided by the number of the elements in the row. This is
a process of averaging over the normalized columns. Mathematically, this
process is shown below:

(4)

or in general:

(4a)

where aij is the entry of row i and column j in a comparison matrix of order n.
An approximate calculation method for the principle eigenvalue, max is as
follows: the matrix of comparisons is multiplied on the right by the estimated
solution vector, W (equation 4 or 5) obtaining a new vector W (equation 6). If
the first component of this vector is divided by the first component of the
estimated solution vector W, the second component of W by the second
component of W, and so on, a new vector W is created (equation 7). The sum
of components of W divided by the number of components in this vector gives
an approximation value of max (equation 8). Mathematically max comes from
the following equations:

(5)
JQME (6)
5,1
or

(6a)

30
(7)

(8)

Comparison matrix consistency


Consistency in the pair-wise comparison matrix means that when basic data are
available, all other data can be logically deduced from them. In doing a pair-
wise comparison to relate n activities or factors so that each one is represented
in the data at least once, only n-1 pair-wise comparison judgments are needed.
For example, if activity A1 is threee times more important than factor A2 and
factor A1 is six times more important than factor A3 , then A1 = 3A2, and A1 =
6A3. It should follow that 3A2 = 6A3 or A2 = 2A3 and A3=1/2A2. If the numerical
value of the judgment (comparison) in the (2,3) position is different than 2, then
the matrix would be inconsistent.

Quantification of consistency
It is usually very difficult to identify n-1 comparisons which relate all factors
or activities and of which one is absolutely certain. It turns out that the
consistency of a positive reciprocal matrix is equivalent to the requirement that
its maximum eigenvalue max should be equal to the number of factors n. It is
possible to estimate the departure of consistence (inconsistency) by the
difference max-n divided by n-1, where max n. The value of (max-n)/(n-1)
which is called a consistency index (CI). This index is further used to calculate
the consistency rating (RI) as detailed in (Saaty, 1990). A consistency rating of
0.10 or less is considered acceptable.
In this study, the special spreadsheet files were prepared to calculate the
weights of the priority factors as well as the weight of importance of all
individual groups included in the data collection survey, as shown later.

Evaluation of priority factors


The factors included in this study were road class, pavement condition,
operating traffic, riding quality, safety condition, maintenance cost, and the
overall importance of road section to community. The evaluation of these
factors was done by quantifying them and assigning an index (out of 100) to
each one according to logic and engineering background as detailed by
Ramadhan (1997), and as shown in Tables I and II.
Factor Tool used
An analytical
hierarchy
Road class RC indexa process
Pavement condition PC = 100-PCIb
Operating traffic OT = 100*(ADT/MADT)c
Riding quality RQ = 100* (RI/MRI)d
Safety condition SC = 100-SNe 31
Maintenance cost MC = 100* (1-COST/MCOST)f
Importance to community See g below
Notes:
a RC: road class index (out of 100), arterial without alternatives = 100,
arterial with alternatives = 75, collector = 50, and local = 25.
b PC: condition index (out of 100), PCI: pavement condition index
determined by any condition survey methods (out of 100).
c OT: operating traffic in vehicles per day, ADT: average daily traffic,
MADT: maximum ADT incurred in the study network.
d RQ: riding quality index (out of 100), PR: pavement roughness in cm/km,
measured by any roughness measuring equipment such as bump integrator,
MPR: maximum allowable pavement roughness in cm/km.
e SC: safety condition index (out of 100), SN: skid number (out of 100) measured
by any skid resistance measuring equipment such as Mu meter. Table I.
f MC: maintenance cost index (out of 100), MCOST: maximum incurred
Evaluation of pavement
maintenance cost in the network under study. maintenance priority
g See Table II below. factors

Sub-factor Level Index

Road class See (1) above


Road is passing through CBD? Yes 100
No 0
Road is near official building or used by VIPs? Yes 100
No 0
Operating traffic (ADT) See (3) above Table II.
Availability of alternative roads during maintenance? No 100 Sub-factor values
Yes 0 determination for
Road is near other important public centers? Yes 100 importance to
No 0 community factor

Importance to community road utilization


In this study, this factor is one of the non-measurable maintenance priority
factors. It was quantified by the following sub-factors: road class, distance from
center business district (CBD), operating traffic, VIP usage, alternative roads at
maintenance time, and other public road utilization factors.
The overall importance factor is calculated using similar methodology of
determining the priority index using the following equation:
JQME (9)
5,1
where,
IF = importance to community factor value (out of 100);
SFi = value of sub-factor, i;
32 SWi = weight of importance of the sub-factor, i; and
SWi = 1.0.

Data collection
The information needed for this study was the priority factor weights as
determined by different groups of people using the road network. The
philosophy was that all classes of road users need to be presented in this study
to reflect the wide spectrum of opinions ranging from ordinary road users to the
pavement management specialists.
For this purpose, a specially designed questionnaire was prepared in four
parts. Part I was for collecting the respondents opinion about the weight of
importance of the seven priority factors. Part II was to quantify the
importance of road to community factor, by assigning importance weights for
the six sub-factors. In these two parts, the respondents were asked to fill a blank
column by a value out of 100 that reflect the importance weight of the priority
factors and sub-factors. Part III was designed for the AHP method to pair-wise
compare all the seven factors considered in Part I. Part IV was for a pair-wise
comparison of the individual groups which participated in the survey to
quantify the effect of each person involved in this survey. The procedure
followed in the preparation of this questionnaire was detailed in Ramadhan
(1997).
The individuals included in the survey were: academics in the area of traffic,
highway, pavement maintenance, and pavement management; highway and
pavement maintenance department officials (managers) and engineers;
pavement maintenance supervisors and technicians; qualified non-pavement
individuals; and ordinary road users. Universities, municipalities,
governmental and private pavement agencies, and ordinary people in the
Kingdom of Saudi Arabia were included in this survey.
The data of Parts I and II were entered into spreadsheet files for processing
and analysis. Mean and standard deviation values for all factors and sub-
factors were determined. For Parts III and IV, special spreadsheets were
prepared to calculate the factor weights of importance. Tables III and IV show
typical examples of one data entry for priority factor weight determination.
Table IV shows another typical example for the individual weight of
importance calculation. In these two tables, weights of importance (out of 1.00)
for all factors or individual group, max, consistency index (CI), consistency
rating (CR) are reported. Only one-half of the pair-wise comparison needs to be
entered since the other half is the reciprocal of the entered one.
Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6
An analytical
Road Pavement Operating Riding Safety Maint. Factor 7 hierarchy
class condition traffic quality condition cost Importance Total process
Factor 1
Road class 1.00 2.00 1.00 1.00 2.00 0.50 0.50 8.00
Factor 2 33
Pavement
condition 0.50 1.00 0.50 0.50 1.00 0.25 0.33 4.08
Factor 3
Operating
traffic 1.00 2.00 1.00 1.00 2.00 0.50 1.00 8.50
Factor 4
Riding
quality 1.00 2.00 1.00 1.00 2.00 0.50 1.00 8.50
Factor 5
Safety
condition 0.50 1.00 0.50 0.50 1.00 0.33 0.50 4.33
Factor 6
Maint.
cost 2.00 4.00 2.00 2.00 3.00 1.00 1.00 15.00
Factor 7
Importance 2.00 3.00 1.00 1.00 2.00 1.00 1.00 11.00
Table III.
Eigenvector 0.130 0.248 0.117 0.117 0.225 0.068 0.093 1.000 Typical AHP
Consistency max= C.I. = C.R.= calculation example
7.074 0.012 0.009 of Part III

Group 5
Group 1 Group 2 Group 3 Group 4 Non- Group 6
Academicians Managers Engineers Supervisors pavement Others Total

Group 1
Academicians 1.00 1.00 0.50 0.25 0.11 0.11 2.97
Group 2
Managers 1.00 1.00 0.50 0.25 0.13 0.11 2.99
Group 3
Engineers 2.00 2.00 1.00 0.33 0.20 0.17 5.70
Factor 4
Supervisors 4.00 4.00 3.00 1.00 0.33 0.25 12.58
Group 5
Non-pavement 9.00 8.00 5.00 3.00 1.00 1.00 27.00
Group 6
Others 9.00 9.00 6.00 4.00 1.00 1.00 30.00
Table IV.
Eigenvector 0.329 0.323 0.189 0.092 0.035 0.032 1.000 Typical AHP
Consistency max= C.I. = C.R.= calculation example
6.069 0.014 0.011 of Part IV
JQME The eigenvector column in Tables III and IV corresponds to the principle
5,1 eigenvalue, max which is determined using the equations 5 to 8 presented
before. The eigenvector column, in the two figures, is calculated according to
equation 4. Further, the consistency index CI and the consistency rating CR
determine how consistent and repeatable the AHP pair-wise comparisons are.
Only those entries that had a consistency rating less than 0.10 were considered
34 acceptable, and included in this study.

Data analysis
Tables V-VII show the summary of all collected data of the survey in terms of
average, standard deviation, and replicates. In the direct assignment method,
the pavement condition and safety condition factors had the highest average

Direct assignment method AHP method


Factor Average St. Dev. N Average St. Dev. N

Road class 13.5 6.4 65 13.8 8.8 30


Pavement condition 19.6 11.4 65 16.2 8.5 30
Operating traffic 13.2 6.4 65 14.5 5.2 30
Riding quality 12.4 5.5 65 13.2 7.6 30
Table V. Safety condition 15.5 7.5 65 16.1 8.5 30
Summary of the Maintenance cost 10.6 6.1 65 9.7 5.0 30
collected data of Parts I Importance to
and III community 15.3 8.0 65 16.6 11.4 30

Direct assignment method


Factor Average St. Dev. N

Road class 16.0 7.7 65


Distance from CBD 15.4 6.8 65
Used by VIB 17.4 8.1 65
Operating traffic 21.6 8.1 65
Table VI. Alternative at maintenance 13.6 6.1 65
Summary of the Distance from other
collected data of Part II important building 16.1 6.2 65

AHP method
Factor Average St. Dev. N

Academicians 34.9 5.2 10


Managers 26.3 5.9 10
Engineers 20.4 6.8 10
Table VII. Supervisors 10.7 2.7 10
Summary of the Non-pavement qualified 4.7 1.6 10
collected data of Part IV Other road users 3.0 0.6 10
weightings, considerably above the other factors. However, as Table V indicates, An analytical
there was a substantial disagreement, as measured by standard deviation, hierarchy
among individuals about the pavement condition factor. The maintenance cost process
was the least important factor. AHP method collected data showed that
pavement condition, safety condition and importance to community factors had
almost equal weightings of 16 per cent, and that the least important factor was
the maintenance cost. Generally, in both methods, it was clear that people 35
closely agreed in rating the maintenance cost as the least important factor as
measured by the standard deviations of 6.1 and 5.0 per cent for the direct
assignment method and AHP method, respectively.
Table II shows that the importance to the community is quantified mostly by
operating traffic followed by whether the road is used by VIPs or passes
by important official locations, or not. This part of the data collection was not
duplicated by the AHP method.
The order of importance for the individual groups participating in the data
collection was academics followed by pavement managers, engineers,
supervisors, non-pavement qualified, and ordinary road users. This order is
shown in Table V, which was a result of AHP pair-wise comparisons, as
collected in Part IV of the data collection questionnaire. This preference is due
to the fact that academicians are usually more exposed to new prioritization
techniques, and pavement managers are mostly close to the maintenance
priority decision makers, if they are not taking this responsibility. The results of
Table V were for factor and sub-factor weight adjustment for individuals
estimating the weightings, as discussed later.
The statistical comparison technique has two steps:
(1) checking the equality of the variances of the two methods using the F-
test; and
(2) checking the equality of the means of the two methods using the t-test
(Montgomery, 1984).
Furthermore, the non-parametric statistical Kendall test of independence
(STATISTICA, 1997) was also used for this comparison. These two
comparisons showed that the results of the two methods are statistically similar
with a positive correlation. This agreement strengthens the direct assignment
method since the AHP method is considered one of the pioneering
psychometric-based methods for prioritization.
Another comparison between the two methods is shown in Figure 1, which
demonstrates the individual tendency to prefer assigning numbers ending with
0s and 5s such as 0, 5, 10, 15, etc. Only a few people used full-scale numbers
between 0 to 9. Thus, individuals, in the direct assignment method, are
extremely biased towards numbers ending with 0 and 5. However, in the AHP
method the situation is different. Since the weightings are calculated based on
pair-wise comparisons, the results did not show a numerical bias towards any
specific number. This suggests that the AHP method can present subjective
preferences with greater precision.
JQME 60
5,1
Percentage of total surveyed
50

40

30
36
20

10
Figure 1.
Comparison of last digit 0
bais for the two data 0 1 2 3 4 5 6 7 8 9
collection methods
Last digit of weight assigned

Factor weight adjustment


The weights of importance of all priority factors as well as the sub-factor of
importance to the community, are to be adjusted for the individual experience
factor, as per the results obtained from Part IV of the data collection
questionnaire, in the following manner:

(10)

where
CWi = corrected overall weight of factor i;
Wj = factor weight as estimated by individual j;
Ej = individual experience factor for individual j, obtained from Table V;
i = factor number (one to seven for priority factors, and one to six for
sub-factors); and
j = individual number out of the total number of individuals n.

Final model form


From the results of previous sections, the final form of the developed
maintenance priority ranking procedure is:
(11)
where,
PI = priority index (out of 100),
Corrected weights: CW1 = 0.134, CW2 = 0.205, CW3 = 0.125, CW4 = 0.131,
CW5 = 0.154, CW6 = 0.107, and CW7 = 0.143.
Fi are the priority factors that are evaluated as shown in Table III, and are as An analytical
follows: hierarchy
F1 = Road class; process
F2 = Pavement condition (PC);
F3 = Operating traffic (OP);
F4 = Riding quality (RQ); 37
F5 = Safety condition (SC);
F6 = Maintenance cost (MC); and
F7 = Importance to community (IF) obtained the following sub-factor
weights equation:
(12)

where,
IF = Importance to community factor;
SFi = Sub-factor values obtained from Table III and,
Corrected weights: SW1 = 0.160, SW2 = 0.146, SW3 = 0.187, SW4 = 0.215,
SW5 = 0.139, and SW6 = 0.157.

Model validation
The priority ranking model was generated from the local experience and
opinions of different people representing local perceptions regarding
maintenance priority; therefore, the results obtained from this model should
compare with current adopted procedures based on engineering judgments. To
quantify this comparison, two case studies were considered. Two sets of
pavement sections in need of urgent maintenance were selected from Jubail
industrial city road network and Dammam municipality road network in Saudi
Arabia. The sections in these case studies were selected to represent a complete
range of all maintenance priority factors, and all the situations considered in the
model development. The assumption made, in these two case studies, was that
only those selected sections were in need of urgent maintenance and repair, and
the allocated funds for maintenance were not enough to cover all these sections.
Thus, the priority procedure should be implemented to rank these sections.
In these two case studies, priority ranking was determined by the
engineering judgment as provided by the pavement maintenance management
personnel in consultation with other engineers in the road departments. The
developed procedure was applied for all pavement sections in these two cases
and then the results were compared to those provided by engineering judgment.
The ranking results in the two case studies were also compared statistically
using the non-parametric Spearman rank correlation coefficient (r s ). The
association between the engineering judgment ranking and the one of the
developed procedure was determined. It was found that the two cases had rs
values of 0.77 and 0.95. On average, and considering the limitations in the
engineering judgment, the developed procedure correlated with the engineering
JQME judgment was 86 per cent, which can be considered an acceptable level of
5,1 correlation.
These two case studies showed that the developed procedure for
maintenance priority ranking gives a good match with the engineering
judgment methods being currently adopted by maintenance management
departments. However, the developed procedure was based on objective
38 measurements as well as the local experience parameters that were provided by
both pavement specialists and other road users in the Kingdom of Saudi Arabia.
Generally the road networks consist of a very big number of pavement sections,
and all of these sections cannot be efficiently considered by the engineering
judgment. Therefore, the developed procedure can efficiently, economically, and
systematically cope with any pavement network of any size.

Conclusions
Based on the processing and analysis of collected data, and on the results of the
comprehensive ranking procedure building and validation, the following
conclusions could be drawn:
(1) The direct assignment method of collecting individuals opinions about
the weight of importance of priority factors showed a good agreement
with the analytical hierarchy process (AHP) pair-wise comparison
method. This agreement strengthens the direct assignment method since
the AHP method is considered one of the pioneering psychometric-based
methods of prioritization.
(2) As measured by the AHP method, the surveyed individual groups
academicians, engineers, non-pavement qualified, and road users
showed statistically different results in their opinions regarding the
priority factors weights. Therefore, adjustment for this effect was
necessary.
(3) The AHP method did not show any numerical bias presented by the
direct assignment method of data collection, where numbers ending with
0 and 5 were significantly presented.
(4) Pavement condition had the highest weight of importance in priority
ranking followed by safety condition, importance to community, road
class, riding quality, operating traffic, and finally the maintenance cost.
(5) The developed procedure can adequately and efficiently rank a huge
number of pavement sections for maintenance, unlike the engineering
judgment, which can handle a relatively small number of pavement
sections at the same time.

References
Chen, X., Weissmann, J., Dossey, T. and Hudson, W. R. (1993), URMS: a graphical urban roadway
management system at network level, Transportation Research Record 1397,
Transportation Research Board, Washington, DC, pp. 103-11.
Cook, W.D. and Kress, M. (1994), A multiple-criteria composite index model for quantitative and An analytical
qualitative data, European Journal of Operational Research, No. 78, North Holland,
pp. 367-79. hierarchy
Haas, R., Hudson, W.R. and Zaniewski, J. (1994), Modern Pavement Management, Krieger process
publishing company, Florida.
Hagquist, R.F. (1994), High-precision prioritization using the analytical hierarchy process:
determining state HPMS component weighting factors, Transportation Research Record
1429, Transportation Research Board, Washington, DC, pp. 7-14. 39
Harker, P.T. and Vargas, L.G. (1987), The theory of ratio scale estimation: Saatys analytical
hierarchy process, Management Sciences, Vol. 33 No. 11, pp. 1383-403.
Karan, M.A. (1984), Municipal pavement management system, A Program of Study in
Pavement Management, Training Course Notes by the Center of Transportation Research,
University of Texas at Austin, TX.
KFUPM-RI (1989), Programming Pavement Maintenance for Al-Jubail Road Network, Final
Report, Vol. I, Research Institute, King Fahd University of Petroleum and Minerals (KFUPM),
Dhahran, Saudi Arabia.
Millet, I. and Harker, P.T. (1990), Globally effective questioning in the analytic hierarchy process,
European Journal of Operational Research, No. 48, North Holland, Amsterdam, pp. 88-98.
Montgomery, D.C. (1984), Design and Analysis of Experiment, (second ed.), John Wiley & Sons,
New York, NY.
Ningyuan, L. and Haas, R. (1994), Special implementation of pavement management for large
highway network in developing countries, Proceedings of The Third International
Conference on Managing Pavements, Vol. 1, Transportation Research Board, Texas, pp. 179-
90.
Ramadhan, R.H. (1997), Modeling of Pavement Condition and Maintenance Priority Ranking for
Road Networks, PhD dissertation, Civil Engineering Department, King Fahd University of
Petroleum and Minerals, Dhahran.
Saaty, T.L. (1982), Decision Making for Leaders, Kluwere Nijhoff Publishing, Lifetime Learning
Publications, Belmont, CA.
Saaty, T.L. (1990), The Analytic Hierarchy Process, (second ed.), RWS Publications, Pittsburgh,
PA.
Saaty, T. L. and Vargas, L.G. (1982), Logic of Priorities, Kluwere Nijhoff Publishing, Boston, MA.
Sharaf, E. (1993), Ranking versus simple optimization in setting pavement maintenance
priorities: a case study from Egypt, Transportation Research Record 1397, Transportation
Research Board, Washington DC, pp. 34-8.
STATISTICA (1997), STATISTICA Manual, Release 5, StatSoft, USA.
Zhang, Z., Singh, N. and Hudson, W.R. (1993), Comprehensive ranking index for flexible
pavement using fuzzy sets model, Transportation Research Record 1397, Transportation
Research Board, Washington, DC, pp. 96-102.

You might also like