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

net/publication/235922824

Evaluation of vehicle fleet maintenance management indicators by

application of DEMATEL and ANP

Article in Expert Systems with Applications · September 2012

DOI: 10.1016/j.eswa.2012.02.159

CITATIONS READS

131 4,285

4 authors:

Davor Vujanović Vladimir Momcilovic

University of Belgrade University of Belgrade
17 PUBLICATIONS 207 CITATIONS 14 PUBLICATIONS 191 CITATIONS

SEE PROFILE SEE PROFILE

Nebojsa Bojovic Vladimir Papić

University of Belgrade University of Belgrade
116 PUBLICATIONS 941 CITATIONS 9 PUBLICATIONS 189 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

“Development of the Model for Managing the Vehicle Technical Condition in Order to Increase its Energy Efficiency and Reduce Exhaust Emissions” View project

Critical infrastructure management for sustainable development in railway, postal and communication sector of Serbia View project

All content following this page was uploaded by Nebojsa Bojovic on 28 April 2019.

The user has requested enhancement of the downloaded file.

 Expert Systems with Applications 39 (2012) 10552–10563

Contents lists available at SciVerse ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier.com/locate/eswa

Evaluation of vehicle fleet maintenance management indicators by application

of DEMATEL and ANP
Davor Vujanović ⇑, Vladimir Momčilović, Nebojša Bojović, Vladimir Papić
University of Belgrade, Faculty of Transport and Traffic Engineering, Vojvode Stepe 305, Belgrade, Serbia

a r t i c l e i n f o a b s t r a c t

Keywords: The paper refers to the importance of maintenance management to increase the vehicle fleet energy effi-
Maintenance management ciency. The fleet maintenance management influences as the vehicle maintenance process itself as well as
Vehicle fleets the primary transport process but also their environment. In order to increase fleet energy efficiency by
Energy efficiency means of a more efficient maintenance management, it is indispensable to observe maintenance process,
Multiple Criteria Decision-Making (MCDM)
transport process and the environment. Since the implementation effects of such measures can be mea-
Decision Making Trial and Evaluation
Laboratory (DEMATEL)
sured by different indicators, this paper analyses the influence of indicators in all three mentioned areas
Analytic Network Process (ANP) on management decision-making. In this sense, appropriate indicators have been defined and subse-
quently used in fleet maintenance management. To determine levels and intensities of interdependence
as well as relative weight of selected indicators two methods have been combined: Decision Making Trial
and Evaluation Laboratory (DEMATEL) and Analytic Network Process (ANP). A model was proposed with
indicators’ interdependence whose relative weights were calculated. The proposed model has been
implemented in several companies with road vehicle fleets. Collected results show the perceived evalu-
ation by company managers in view of maintenance management process influence onto their fleet
energy efficiency. Besides, by proposed model implementation we have obtained equally managers’ eval-
uation upon effectiveness and efficiency of the maintenance management within studied companies.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Metaxiotis, Lekatis, & Assimakopoulos, 2003; Waeyenbergh &

Pintelon, 2002). In these papers different models and systems have
Companies with own road vehicle fleets attain profit by per- been proposed, all with the objective of increasing efficiency and
forming transport services. The amount of profit is significantly productivity of industrial machines.
influenced among other things, by the costs incurred by transport However, vehicle operation and maintenance processes differ
and vehicle maintenance processes. Considered companies seek to from those related to industrial machines. The fact that vehicles
accomplish all the planned transport tasks while minimising are mobile assets, further affected by a large number of external
transport and maintenance costs. environmental factors imposes the need for a different approach
More efficient fleet maintenance management could affect in their maintenance management, compared to other static
rational transport process realisation, i.e. reduction in incurred machines.
costs. Efficient maintenance management facilitates vehicles of In this sense, for an efficient fleet maintenance management it is
best suited construction–operation (CO) groups in a state ‘‘ready necessary to observe jointly: (1) the transport process as a primary
for operation’’ for transport tasks realisation during the required (core) process that brings profit to the company; (2) the vehicle
time periods. This certainly influences the increase in energy effi- maintenance, as logistical support to the core transport process,
ciency of the fleet and transport and maintenance costs reduction, which by means of maintenance interventions transforms vehicle
meanwhile company’s core performance is not jeopardised i.e. all condition from the state of ‘‘unready for operation’’ to the state
planned transport tasks are to be accomplished. ‘‘ready for operation’’; (3) the environment, associated to safety
In order to attain an efficient maintenance management, it is and environmental protection from the maintenance impact, which
necessary to coordinate the primary (core) process with mainte- is monitored via technical inspections. For an integrated approach
nance, which has been researched mostly in the field of industrial to maintenance management an important concept is the ‘‘Process
production (Ashayeri, Teelen, & Selen, 1996; Nikolopoulos, based maintenance’’ (Zhu, Gelders, & Pintelon, 2002) and (Zhu &
Pintelon, 2001). This concept, among other things, involves the def-
⇑ Corresponding author. Tel.: +381 11 309 12 66; fax: +381 11 309 67 04. inition of necessary indicators which allow measuring the
E-mail address: [email protected] (D. Vujanović). implementation effects of specific measures during maintenance

0957-4174/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.eswa.2012.02.159
 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563 10553

management, monitoring the value of indicators in relation to the factors, but its total-influence matrix – T is normalised and incor-
adopted thresholds and management decision making in the event porated into an unweighted supermatrix by the AHP method.
of unauthorized indicator values deviation from threshold values. A combination of DEMATEL and ANP methods has been used in
However, in case of several indicators for measuring the imple- this research with same approach as in the paper by (Lee et al.,
mented measures effects and for management decision making, it 2011). By literature review and based on authors’ personal experi-
is necessary to determine which of the indicators is more signifi- ence, appropriate indicators were defined in the fleet maintenance
cant for achieving a defined objective. Moreover, large number of management. A model which contains three interdependent
observed indicators have interdependent impact. Implementing groups or fields: transport and maintenance processes and their
certain measures within management, could improve an indicator environment has been developed. In each field there are interde-
value, but impact differently a number of other indicators’ values. pendent factors (indicators). Based on a conducted research of per-
The considered issue represents a classic example of Multiple Cri- ceptions of field-related professors and other relevant experts from
teria Decision Making (MCDM). the Faculty of Transport and Traffic engineering in Belgrade, the
Therefore, the problem under consideration in this paper is to interdependences among indicators, as well as interdependences
determine the level of interdependences of indicators and deter- between observed fields have been established. Afterwards, the
mine their significance and their relative weight in the mainte- relative weights of indicators and of each observed field are calcu-
nance management causing an increase in the fleet energy lated within the developed model using the above mentioned
efficiency, provided that planned transport tasks are realised. As methods. By surveys of managers in several transport companies
a solution, a model with ranked indicators upon their impact onto and by the proposed model implementation, an evaluation of per-
the fleet energy efficiency is obtained. The resulting model should ception on maintenance management impact on the enhancement
point out to managers which indicators should be given more of the fleet energy efficiency was made, as well as evaluation of
attention in measuring the implemented measures effects and in managers’ efficiency in the maintenance management within each
maintenance management decision-making. The proposed model studied company.
can be used to evaluate the managers’ perception of the impor- In the following Section 2, the concept of fleet maintenance
tance of maintenance management to increase the fleet energy management has been presented in detail, together with a descrip-
efficiency. Also, managers can be evaluated by means of the model tion of selected indicators. Section 3 describes the DEMATEL and
upon their effectiveness and efficiency in the fleet maintenance ANP methods. Based on survey results, relative weights of indica-
management. tors and observed fields were obtained in the developed model.
To calculate the level of interdependences and determine the le- In the Section 4 the results of proposed model implementation in
vel of significance of indicators in relation to the accomplishment several companies with road fleets were addressed. In Section 5
of a defined objective, a combination of two methods DEMATEL the results were thoroughly analysed, while in the last section
and ANP will be used as tools for Multiple Criteria Decision Making the main conclusions and future research topics were drawn up.
(MCDM).
DEMATEL method has been developed by ‘‘Science and Human
Affairs Program of the Battelle Memorial Institute of Geneva’’ be- 2. Fleet maintenance management
tween 1972 and 1976 and used for research and solving several
groups of complicated and interdependent problems (Fontela & 2.1. Interdependence of transport and maintenance processes and
Gabus, 1974) and (Fontela & Gabus, 1976). This method has been their environment
applied in various fields most recently (Li & Tzeng, 2009; Lin, Chen,
& Tzeng, 2009; Lin, Yang, Kang, & Yu, 2011; Tzeng, Chiang, & Li, Studied transport companies most often own heterogeneous
2007). As a result, total direct and indirect influences of each factor fleets composed of different construction–operation (CO) vehicle
(indicator) are obtained as each factor’s (indicator) influence given groups, especially from the point of view of their available cargo
to other factors, but as well influence received from other factors. capacity. The set of all planned transport tasks that vehicles have
This interdependence is visually depicted by a Network Relation to accomplish in certain time period is defined by the Operation
Map (NRM). Plan (OP) (Milosavljević, Teodorović, Papić, & Pavković, 1996) and
ANP method represents a more developed version of Analytic (Momčilović, Papić, & Vujanović, 2007). In this sense, it can be as-
Hierarchy Process (AHP) method. Proposed by Saaty in (Saaty, sumed that fleets in observed transport companies operate accord-
1996) and (Saaty & Vargas, 1998), in order to avoid hierarchical ing to the predefined OP.
constraints that exist in the AHP method (Saaty, 1980). It is a rela- During transport task realisation, i.e. during the transport pro-
tively new MCDM method used to calculate the interdependences cess, vehicles are undergoing more or less important deterioration
of factors and determine their relative weights. This method has of their technical condition (Fig. 1). As a consequence of such dete-
been applied in many areas (Chung, Lee, & Pearn, 2005). rioration, vehicles are initiating maintenance requests. During real-
However, treatment of factors’ interdependences in ANP meth- isation of maintenance interventions vehicles are in a state
od is not objectively addressed in relation to the actual system. ‘‘unready for operation’’ and such vehicles therefore will not be
This lack is covered by using the DEMATEL method, where interde- available for transport tasks realisation according to OP require-
pendences between groups (sets) of factors are determined more ments. The moments and durations of those periods depend
objectively and based on the NRM form a structure of the observed mainly on fleet maintenance management effectiveness and effi-
system is created, which is subsequently used to calculate the rel- ciency. After completion of required maintenance interventions,
ative weight of factors by using the ANP method (Yang et al., 2008). vehicles turn into the state ‘‘ready for operation’’ and become
Combined use of these two methods has been recently imple- available for further transport tasks realisation.
mented for solving MCDM problems in different fields (Yang & In this sense, vehicle maintenance process represents a logistic
Tzeng, 2011; Wu, 2008; Lin, Hsieh, & Tzeng, 2010). In the paper support for the transport process, which on the other hand should
(Lee, Huang, Chang, & Cheng, 2011), the authors go a step further provide for the transport service in order to satisfy client requests.
and propose a new hybrid method, which is a developed version The objective of vehicle maintenance process is to allow the
of the method compared to the paper by (Ou Yang et al., 2008). accomplishment of Transport Company’s objective through ensur-
According to (Lee et al., 2011), DEMATEL method is used not only ing the required number of vehicles in the state ‘‘ready for opera-
as a more objective view of interdependences of groups (sets) of tion’’ in the exact moment when and the entire period during
10554 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563

Fig. 1. Interdependence of the environment, transport and maintenance processes.

which they are needed with certain level of reliability (Papić, the consumption of non-renewable resources, such as fuel, lubri-
Medar, & Pejčić Tarle, 1999). This illustrates the interdependence cants, tyres, and materials.
and influence of transport and maintenance processes. The result of integrated fleet maintenance management is ra-
On the other hand, maintenance process impacts the environ- tional realisation of transport services with energy efficient vehi-
ment through realisation of maintenance interventions on vehi- cles, and by doing so accomplishing clients’ requests and making
cles. The latter could reflect on vehicle functionality, i.e. its profit. In order to improve fleet maintenance management it is
reliability and safety, as well as on harmful exhaust gas emissions, indispensable to measure continually certain measure implemen-
fuel and oil consumption, etc. (Yamamoto, Madre, & Kitamura, tation effects through appropriate indicators. For that reason,
2004). Therefore, the objective of vehicle maintenance process is based on an extensive set of references and authors’ experience,
to decrease its environmental impact through superior quality of in the subsequent text a set of suitable indicators of efficient fleet
maintenance interventions. From the other side, as to limit the maintenance management was selected.
environmental impact of maintenance, vehicles must remain in
correct technical condition (faultless operation) throughout the en- 2.2. Fleet maintenance management indicators
tire transport process realisation, after completing maintenance
interventions. In order to keep the environmental impact under The authors of the paper (Papić et al., 1999) have used simula-
certain legally prescribed thresholds, vehicles are submitted to tion and the AHP method for determining the most favourable fleet
periodical technical inspections by third parties, specialised for maintenance strategy. As suitable indicator, among others, the
such an activity. If the technical inspection results in a negative authors have chosen the Operation Plan (OP) realisation percentage
finding regarding vehicle condition, it is withdrawn from the trans- (hereafter referred as T1). The indicator T1 represents the percent-
port process and remains ‘‘unready for operation’’ until mainte- age of realised transport tasks in relation to the total number of
nance accomplishes all necessary interventions. This is the planned transport tasks (in tonne-kilometres), according to the re-
interdependence of the ‘‘environment’’, transport and maintenance quested OP in the observed period. The value of this indicator influ-
processes. Accordingly, in the paper by (Samaras & Kitsopanidis, ences company’s income. Paper by (Milosavljević et al., 1996)
2001), the authors recommend a methodology to evaluate the effi- addresses a model based on fuzzy sets theory of vehicle allocation
ciency of alternative short tests applied to inspection and mainte- to transport tasks, according to OP which is explained in detail. In
nance (I/M) to monitor vehicle exhaust gas emissions. the paper are presented two different versions of vehicle allocation
According to (Johnson, 2002), controlled maintenance manage- to transport tasks, and as an indicator for these variants evaluation
ment system consists of organising, planning, supervision, coordi- an indicator (T1) was used. Accordingly (Li, Mirchandani, & Boren-
nation and control of the functions necessary to ensure that stein, 2009) researched the problem of vehicle reallocation to the
equipment perform its designed function within economic possi- tasks in case of vehicle failure, one of the criteria being to maximise
bilities. The system ensures that the company will obtain the most the realisation of transport tasks, i.e. maximisation of indicator T1.
efficient utilisation of labour, equipment and materials in achiev- The paper by (Haghani & Shafahi, 2002) demonstrates several
ing defined objective. forms to solve the vehicle maintenance scheduling problem, where
One of the objectives of fleet maintenance management is in the adopted criterion was to minimise disruption in OP realisation,
providing vehicles from the most adequate CO groups in state i.e. maximisation of the indicator T1.
‘‘ready for operation’’ in required periods, according to OP within As this paper addresses efficient maintenance management in
defined economic possibilities. This objective facilitates rational order to increase fleet energy efficiency, in paper by (Vujanović,
transport service, while accomplishing all planed transport tasks Mijailović, Momčilović, & Papić, 2010) has been indicated that sig-
according to OP requirements, while at the same time enhances nificant increase in freight transport energy efficiency lies in better
vehicle fleet energy efficiency (Vujanović, Momčilović, Papić, & use of vehicle cargo capacity. In this sense, a suitable indicator
Bojović, 2011). Besides the maintenance management objective affecting fleet energy efficiency within the transport process would
of increasing company’s profitability is the objective of limiting be Vehicle Payload Utilisation (hereafter referred as T2), in other
 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563 10555

words cargo capacity utilisation. Indicator T2 is the ratio of the car- (vehicle workspaces, available workers, spare parts, equipment,
ried consignment mass (or volume) and vehicle payload capacity materials, etc.), be acquainted with external specialised mainte-
(or available cargo compartment volume) in observed period. The nance facilities on the market (location, specialisation, quality,
authors (Vujanović et al., 2010) show how improving indicator etc.), monitor the adequate realisation of planned interventions in
T2 leads to a reduction of specific fuel consumption per unit of rea- the MP. As appropriate indicator for measuring the MP realisation
lised transport volume – qt (l/100 tonne-km), namely decrease the efficiency within the maintenance process appears Realisation of
total amount of consumed fuel by the fleet for OP realisation, thus MP (hereafter referred as M3). According to (Arts, Knapp, & Mann,
reducing transport costs and maintenance. Equally (Vujanović 1998) and (Dhillon, 2002), indicator M3 is the ratio of number of
et al., 2011) stress the importance of improving indicator T2 in or- work orders realised according to the MP and total number of
der to increase energy efficiency by fleet maintenance manage- planned work orders from the MP in analysed period. In his paper
ment. In the paper (McKinnon, 1999) it is noted that measures (Norat, 2008) considers indicator M3 as another key performance
improving the vehicle payload utilisation are amongst the most indicators of the maintenance process.
important for increasing fleet energy efficiency. Also (Kamakaté Another suitable indicator of the maintenance process can be
& Schipper, 2009) and (Ruzzenenti & Basosi, 2009) advocate high the Planned Maintenance Percentage (hereafter referred as M4).
potential to increase energy efficiency to better vehicle cargo According to (Arts et al., 1998), indicator M4 is the percentage of
capacity utilisation. labour working hours on planned maintenance work orders from
As well paper (Papić et al., 1999) emphasises significant savings, total labour working hours in a period. According to (Mobley
compared to the current situation, in number of vehicles required et al., 2008), while the percentage of planned maintenance
for OP realisation by application of different strategies leading to increases, it reduces the percentage of unplanned maintenance,
considerable savings in energy and materials for approximately which equally affects the maintenance costs reduction. Growing
the same value of indicator T1. Savings shown in the paper repre- percentage of planned maintenance gives the greater opportunities
sent a good basis for introducing another indicator – Vehicle Fleet for integration of transport and maintenance processes, in this
Utilisation Rate (hereafter referred as T3). Indicator T3 represents manner increasing the fleet energy efficiency (Vujanović et al.,
the ratio of number of vehicles necessary for OP realisation and to- 2011).
tal (inventory) number of vehicles in certain period. The difference Regarding the environmental impact on the fleet maintenance
will represent an additional number of ‘‘spare’’ vehicles, which management the indicator Percentage of Fleet Roadworthiness (here-
could be used as replacement for those undergoing maintenance. after referred as E1). Indicator E1 represents the percentage of vehi-
According to (Haghani & Shafahi, 2002), efficient scheduling of cles complying with minimum requirements regarding technical
vehicles to maintenance would require less replacement vehicles condition checked on technical inspection, i.e. vehicle roadworthi-
in order to succeed in OP realisation, thus improving indicator ness test, in terms of vehicle safety and emissions, compared to
T3, altogether reducing transport and maintenance costs. the total number of vehicles controlled on technical inspection in
In the handbook (Mobley, Higgins, & Wikoff, 2008) it is noted certain period. In the paper (Bin, 2003) is shown the percentage
that when analysing equipment reliability, an indicator of Mean of road vehicle manufacturers not meeting the legal requirements
Time Between Failures (MTBFs) can be used, hereafter referred as (thresholds) on emissions inspection. Those range about 4–10% of
M1. M1 is calculated from the ratio of vehicle working hours (or
kilometres), and number of failures in the observed period. If you
Table 1
get to improve the indicator M1, you can expect higher vehicle reli- Fleet maintenance management indicators.
ability, which equally enhances vehicle safety. Also (Parida &
Interdependent Indicators (factors) Indicator definition
Kumar, 2009) state that M1 is one of the most important indicators
groups
for measuring maintenance productivity, while (Norat, 2008) con-
1. Transport Operation Plan (OP) T1 = (amount of realised tonne-
siders the indicator M1 as a key performance indicator (KPI) of the
process (TP) Realisation kilometres/amount of planned
maintenance process. percentage T1 tonne-kilometres, in OP)  100
Besides previously mentioned, the author (Norat, 2008) consid- Vehicle payload T2 = consignment mass (volume)/
ers as another maintenance process’ KPI – the Mean Time to Repair utilisation T2 cargo (compartment volume)
(MTTR). According to (Mobley et al., 2008), the indicator MTTR is a capacity
Vehicle fleet T3 = number of required vehicles
measure of system’s maintainability. MTTR is calculated from the
utilisation rate T3 for operation/total number of
ratio of total labour working hours required to carry out repairs vehicles
and interventions to the total number of failures in the reporting 2. Maintenance Mean time between M1 = vehicle realised working
period (Parida & Kumar, 2009). However, for the fleet maintenance process (MP) failures M1 hours (or kilometres)/number of
failures
process, a special interest should be given to the indicator of Mean
Mean vehicle M2 = vehicle total hours ‘‘unready
Vehicle Downtime (hereafter referred as M2), i.e. vehicle in the state downtime M2 for operation’’/number of failures
‘‘unready for operation’’. This indicator is obtained by the ratio of Maintenance Plan M3 = number of realised work
total hours when vehicle was unavailable for operation and num- (MP) Realisation M3 orders from MP/total number of
ber of failures in the observed period. Unlike indicator MTTR, when planned work orders in MP
Planned maintenance M4 = (labour working hours on
calculating indicator M2 it is taken into account the required repair
percentage M4 planned maintenance work
or maintenance intervention preparation time e.g. waiting for orders/total labour working
spare parts, free (available) workers, free dedicated workspace in hours)  100
the workshop and so on, as well as eventual time between (to or 3. Environment Percentage of fleet E1 = (number of vehicles
from) the maintenance workshop and transport company. (E) roadworthiness E1 complying with minimum
requirements/total number of
A useful tool in the maintenance management is a Maintenance vehicles controlled on technical
Plan (MP) according to (Momčilović et al., 2007) and (Maróti & inspections)  100
Kroon, 2005). MP is a document used by managers while making Percentage of vehicle E2 = (number of vehicles in
decisions on planned workspaces and timeframe to carry out neces- roadworthiness in accidents that complied with
accidents E2 minimum safety requirements/
sary maintenance interventions on vehicles. To adequately prepare
total number of vehicles in
the MP, managers must: record maintenance requests, have knowl- accidents)  100
edge of OP, be familiar with the capacity of their own workshops
10556 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563

the number of tested vehicles. Meanwhile (Christensen & Elvik, Step 1: Create experts perception matrixes X1, X2, . . . , XH. Assum-
2007) have noted that number of vehicle technical malfunctions de- ing that there are H experts in the observed survey and n factors
crease when increasing the number of performed technical inspec- that are considered, each expert should determine the level of influ-
tions in a given period. In addition, another indicator is also ence of factor i to the factor j. The comparative analysis of couple of
appropriate Percentage of Vehicle Roadworthiness in Accidents (here- factors i and j by the expert k is denoted by the xkij , while: i = 1 , . . . , n;
after referred as E2). Indicator E2 is the percentage of vehicles that j = 1, . . . , n; k = 1, . . . , H. The value of each couple xkij adopts an integer
participated in traffic accident and complied with minimum value with a following meaning: 0 – no influence; 1  low influ-
requirements regarding critical vehicle safety systems (e.g. braking, ence; 2 – medium influence; 3 – high influence; 4 – very high influ-
steering, suspension, lighting system and so on) relative to the total ence. The answer of the expert k is represented by a matrix of rank
number of vehicles that participated in accidents in the observed n  n, while each element k of the matrix in the expression
period. The paper (Randhawa, Miller, Bell, & Montagne, 1998) states X k ¼ ½xkij nn denotes a non-negative number xkij , where k = 1, . . . , H.
that a technical malfunction or failure caused about 5% of the ob- Accordingly, matrixes X1, X2, . . . , XH represent answer matrixes of
served accidents of commercial vehicles. Besides, authors (Rechnit- each of H experts. The diagonal elements of each expert answer ma-
zer, Haworth, & Kowadlo, 2000) provide an overview of the results trix are all set to zero because the factor cannot affect itself.
of studies showing that in 3% of analysed accidents vehicle condi- Step 2: Calculate average perception matrix A. Based on deter-
tion had certain impact on traffic accidents, i.e. technical malfunc- mined answer matrixes X k ¼ ½xkij nn from all of the H experts, it
tion or failure were the principal cause of accident. could be calculated the average answer matrix, A = [aij]nn, which
In the subsequent Table 1 are shown the fleet maintenance represents a medium value of opinions of all H respondents (ex-
management indicators within transport, maintenance and envi- perts) for each element of matrix A in a following way:
ronmental fields, further addressed in this paper. 2 3
a11 a12  a1n
6a a22 ; a2n 7
6 21 7
3. Applied methodology 6 7
A ¼ 6 .. ... .. ... 7 ð1Þ
6. . 7
4 5
DEMATEL method was used in this paper as to determine the le- ..
an1 an2 . ann
vel of interdependences existing between selected indicators of
maintenance management as well as to construct a network rela- where:
tionship map (NRM). Based on constructed NRM and on the calcu- 1P H
lated level of interdependences of indicators, the structure of the aij ¼ xk ð2Þ
H k¼1 ij
model has been developed. By an integrated implementation of
DEMATEL and ANP methods significances are calculated, i.e. rela- Matrix A shows the initial effects caused by a particular factor, but
tive weight of observed indicators in the model related to the de- also the initial effects he receives from other factors.
fined objective, which is to increase fleet energy efficiency by Step 3: Calculate of the average normalised perception matrix D.
maintenance management. The matrix D is calculated from the matrix A, as follows:
In this regard, firstly DEMATEL method for calculating the level Let
of interdependences of selected indicators will be explained. Total !
influence matrix that will be used in the ANP method is shown
P
n P
n
S ¼ max max16i6n aij ; max16i6n aij ð3Þ
afterwards. Next ANP method will be explained and integration j¼1 i¼1

with DEMATEL for calculating the relative weight of chosen indica-

Then
tors. The survey to collect all the necessary data will be explained
then in detail. Based on the methodology presented the relative D ¼ A=s ð4Þ
weight of each of the observed indicators will be calculated.
As the sum of each row i of the matrix A represents total direct ef-
fects that the factor i given to other factors, the expression
P
3.1. DEMATEL method max16i6n nj¼1 aij , represents the most important total direct effects
from the specified factor given to other factors. Similarly, as the
DEMATEL (Decision-Making Trial and Evaluation Laboratory) is sum of each column j of the matrix A represents total direct effects
a comprehensive method for designing and analysing structural that the factor j received from other factors, the expression
P
models of causal relationships between complex factors (Wu & max16j6n ni¼1 aij , represents the most important total direct effects
Lee, 2007). Unlike the Analytical Hierarchy Process (AHP) method that certain factor has received from other factors. The value s (3)
which considers factors that are independent of each other, DEM- takes into account the major value of above mentioned two expres-
ATEL method takes into account interdependent factors and deter- sions. Matrix D is obtained when each element aij of matrix A is di-
mines the level of interdependence between them. The observed vided by expression s. Each element dij of the matrix D takes values
method is based on graph theory, allowing visual planning and between 0 and less than 1.
problem solving so that the relevant factors can be divided into Step 4: Calculate the total relation matrix T. Matrix T is an n  n
causal and consequential for a better understanding of mutual matrix and it is calculated as follows:
relations (Li & Tzeng, 2009). This scientific research method could
improve understanding of the complex structure of the specific T ¼ DðI  DÞ1 ð5Þ
problematique and contribute to identification of relationships be- where I is an n  n identity matrix.
tween factors, workable solutions by a hierarchical structure Let the sums of rows and columns of the matrix T are separately
(Tzeng et al., 2007). The end product of the DEMATEL process is represented by a vector R and vector Cn1, explicitly:
a visual representation – an individual map of the mind – by which
the respondent (manager) organises his or her own action in the T ¼ ½tij nn ; i; j ¼ 1; 2; . . . ; n ð6Þ
world (Lin, Chen, & Tzeng, 2009). " #
The procedure of calculating the level of interdependence of the P
n
R ¼ ½r ij n1 ¼ tij ð7Þ
considered factors with DEMATEL method can be represented by j¼1
n1
the following steps:
 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563 10557

elements of the matrix T, which are larger than the adopted value
p, retain their present value. Should the adopted value p be too low,
the structure of the system will remain complex and difficult to
understand, while if threshold value p is too high, the structure
would be oversimplified and important influences ignored. There-
fore, based on the adopted threshold value p, we can filter minor
effects in matrix T, based on which it will be obtained NRM, as
shown on Fig. 3, which facilitates understanding of the relation-
ships in the considered system.

3.2. Integration of DEMATEL and ANP methods

ANP method was developed to avoid the hierarchical con-

straints that exist in the AHP method (Saaty, 1996). In this paper
ANP, combined with DEMATEL will be used to calculate the relative
weight of factors (indicators) that are used in the fleet maintenance
Fig. 2. Influence relationship map. management.
When using only conventional ANP to calculate the relative
  weight of factors, the levels of interdependences of factors are trea-
P
n
C ¼ ½cj 0n1 ¼ tij ð8Þ ted as reciprocal values. However, according to DEMATEL method,
i¼1 1n
the levels of interdependences of factors do not have reciprocal va-
lue, which is closer to the real system (Yang & Tzeng, 2011). There-
where symbol 0 stands for a transposed matrix. Let ri be the sum of fore, to calculate relative weight of factors it will be used the total
the ith row in matrix T. Then ri represents total direct and indirect relation matrix (matrix T) within a DEMATEL, in order to avoid the
effects that the factor i has given to other factors. Let cj represents shortcomings mentioned in the ANP.
the sum of jth column in matrix T. Then cj shows the total effects, According to (Lee et al., 2011), DEMATEL method is not going to
both direct and indirect received by a factor j given by other factors. be used only to calculate the level of impacts among different
In the case when i = j, then the term (ri + ci) represents a degree of groups of factors, but the normalised total-influence matrix will
importance of the factor, and the term (ri  ci) represents the net ef- be incorporated into un-weighted supermatrix W in the ANP to cal-
fect that the factor contributes to the system in relation to other fac- culate the level of interdependences of different factors.
tors. If the term (ri  ci) is positive, the factor i is net causer, and if In this regard, the proposed methodology of integration of
the previous expression is negative factor i is a net receiver (Tzeng DEMATEL and ANP method consists of four steps. The first step is
et al., 2007). based on the NRM (Fig. 3) and the total-influence matrix T to con-
Step 5: Set a threshold value p and obtain Network Relationship struct the network hierarchical structure of the observed system to
Map (NRM). On the basis of expert opinion is determined the the defined objectives. The second step is to calculate un-weighted
threshold value p, which filters out negligible or small effects in supermatrix W. The sums of influences of each factor in relation to
the matrix T. The value of elements of matrix T, which are smaller the factors of each group in the factor matrix of the total-influence
or equal than the adopted value p, is set to zero, while other in DEMATEL, is shown in (9).

Fig. 3. Network relation map (NRM).

10558 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563

While sums of influences of factors c11, . . . , c1m1 are related to the

group D1 as follows:

11
X
m1
di ¼ t11
cij ; for i ¼ 1; 2;    m1 ð13Þ
j¼1

Elements t 11
cij
represent values of factor influences c11, . . . , c1m1 in
relation to factors from the group D1, and elements t ac11 11
their nor-
malised values.
Calculation procedure for T ac nn , as well as all other normalised
matrixes within matrix T ac is identical. Normalised values of factor
influences in relation to the factors from each group are incorpo-
rated into the calculation of un-weighted matrix W, being used
ð9Þ within the ANP method, according to (14).

While matrix T 11
cis matrix of factors from the group D1 and influ-
ences in respect of the factors from the group D1, shown in (10), and
T 12
c matrix of factors from group D1 and influences in respect of the
factors from the group D2 and so on.
2 3
t 11
c11    t 11
c1j
   t11
c1m1
6 7
6 .. .. .. 7
6. . . 7
6 7
6 7
T 11
c ¼ 6 t 11    t 11    t11
c1m1
7 ð10Þ
6 ci1 cij 7
6. .. .. 7
6 .. . . 7
4 5
t 11
cm11    t 11
cm1j
   t11
cm1m1

Factor total-influence matrix Tc after normalisation becomes matrix

T ac , as shown in (11).
ð14Þ

where the matrix W11 represents the values of factor influences

from the group D1 in relation to factors from the D1 group, accord-
ing to (15). Vector of factor influences with the value zero indicates
the independent effect of this factor with respect to that group. In
the identical way the matrix Wnn is obtained and other matrices
within the supermatrix W.

ð11Þ ð15Þ

where T ac 11 is a normalised matrix of sums of factor influences

c11, . . . , c1m1 related to factors from group D1 and calculated upon
(12).
2 11 11 11
3
t 11
c11 =da ... t 11
c1j
=d1 ... t 11
c1m1
=da
6 7
6 .. .. .. 7 In the third step, it is calculated the weighted supermatrix Ww
6 7
6 . . . 7 as to normalise sums of influences of each group of factors in
6 7
6 11 7 relation to all groups from the matrix of total group influences
T ac 11 ¼ 6 t 11 c i1 =d
11
i . . . t 11
c ij =d
11
i . . . t 11
imi =d i 7
6 c 7 (16).
6 7
6 .. .. .. 7
6 . . . 7
4 5 2 3
11 11 11 11 11 11 t11D . . . t1jD . . . t 1n
D
tcm1 1 =dm1 . . . t cm1 j =di . . . tcm1 m1 =dm1 6.
2 a11 3 6. .. .. 7 7
t c11 . . . tac1j11 . . . t ca1m 11
1
6.
6
. . 7
7
6 7 TD ¼ 6 i1
. . . tijD . . . t in 7 ð16Þ
6 ..
6.
.. .. 7
7 6 tD D 7
6 . . 7 6 7
6 a11 6 .. .. .. 7
a11 a11 7 4. . . 5
¼ 6 t ci1 . . . tcij . . . t cimi 7 ð12Þ
6 7
6. .. .. 7 tn1
D . . . tnj
D . . . t nn
D
6 .. . . 7
4 5 11
a11 a11 a11 While t 11
D determines the sum of all influences from the matrix T c ,
t cm1 1 . . . tcm1 j . . . t cm1 m1 nn
and t nn
D determines the sum of all influences from the matrix T c .
The normalisation procedure is calculated upon (17).
 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563 10559

2 3 2 a11 3
t 11 1j 1n
D =d1 . . . t D =d1 . . . t D =d1 tD ... tDa1j . . . t Da1n and the Average Time Between Failures (M1), while the least impor-
6 . . . 7 6. .. .. 7 tant indicator is Percentage of Vehicle Roadworthiness in Accidents
6 . .. .. 7 6 7
6 . 7 6 .. . . 7 (E2). The indicator influencing the most the change in other indica-
a 6 7 6 7
T D ¼ 6 tD =di . . . t D =di . . . t D =di 7 ¼ 6
6 i1 ij in 7
6 tD
ai1
... tDaij ain 7
. . . tD 7 tors’ values is Planned Maintenance Percentage (M4), with the high-
6 7 6 7 est value of (r  c), followed by the indicator of Mean Time Between
6 .. .. .. 7 6 .. .. .. 7
4 . . . 5 4. . . 5 Failures (M1). Indicator, which is under the major influence of all
t n1 nj nn
D =dn . . . t D =dn . . . t D =dn t dan1 . . . tDanj . . . t Dann other indicators is the Vehicle Payload Utilisation (T2), with the low-
ð17Þ est value of (r  c), followed by OP Realisation Percentage (T1).
When from the total-influence matrix T we sum up all the indi-
where: cators’ effects related to the group, or field, we obtain the matrix of
P
n total group influences TD (Table 4). The Table 4 shows that the most
di ¼ t ijd ; for i ¼ 1; . . . ; n ð18Þ important influence have the maintenance process indicators in
j¼1
relation to the transport process indicators. Due to the selected
Calculation of weighted supermatrix Ww is obtained by integrating threshold value (p = 0.2) set by experts, transport process indica-
the un-weighted matrix W into the normalised matrix of group fac- tors are considered insignificant compared to environmental indi-
tors influences T ac , in accordance with (19). cators influences, and therefore neglected. Table 5 shows that
2 3 overall most important influences come from the maintenance
t aD11 xW 11 taD21 xW 12 ... . . . t aDan1 xW 1n process indicators. The latter for the most part ‘‘give’’ their influ-
6 .. .. 7
6 a12 7 ence to other indicators, since they have the highest value (r  c).
6 t D xW 21 taD22 xW 22 . . 7
Ww ¼ 6
6
7
7 ð19Þ Meanwhile transport process indicators, with the lowest value (r
6
4 ... ...
a
t Dij xW ij . . . tni xW in 7
5
 c) are therefore largely subject to impacts from other indicators,
D
i.e. largely ‘‘receiving’’ the effects and influences from other
t aD1n xW n1 taD2n xW n2 ... ... t aDnn xW nn indicators.
In the fourth step a limited supermatrix is calculated by several According to the results in Tables 2–5, the following Fig. 2 dis-
multiplying of weighted supermatrix WW until the vector values plays the map of interdependence between the observed groups
in a limited supermatrix become stable, in other words: (fields), as well as the influences among the indicators within the
group. Map of interdependences allows better understanding the
lim ¼ W zw ð20Þ
z!1

while the number z tends to infinity. Vectors of the limited superm-

atrix represent relative weights of each factor in relation to the de- Table 2
fined objective. Matrix of total influence of indicators T.

Indicators T1 T2 T3 M1 M2 M3 M4 E1 E2
3.3. Realised data collection T1 0.00 0.28 0.24 0.00 0.00 0.28 0.00 0.00 0.00
T2 0.27 0.00 0.00 0.00 0.23 0.26 0.00 0.00 0.00
On the basis of selected indicators of fleet maintenance man- T3 0.25 0.28 0.00 0.00 0.21 0.24 0.00 0.00 0.00
M1 0.43 0.44 0.36 0.21 0.34 0.41 0.25 0.30 0.26
agement, a survey was carried out as to collect experts’ perceptions
M2 0.43 0.46 0.40 0.27 0.24 0.45 0.23 0.25 0.22
on interdependent influences of the observed indicators of trans- M3 0.47 0.48 0.40 0.24 0.34 0.30 0.24 0.29 0.25
port, maintenance processes and environment. For this purpose a M4 0.46 0.46 0.36 0.35 0.36 0.44 0.00 0.39 0.35
group of professors from the Faculty of Transport and Traffic engi- E1 0.34 0.36 0.28 0.28 0.29 0.29 0.00 0.00 0.30
neering in Belgrade were selected, along with other experts from E2 0.27 0.30 0.00 0.00 0.23 0.26 0.00 0.24 0.00

the field of vehicle fleets operation and maintenance with long-

term working experience. The survey was conducted during the
period from January to May 2011. Experts’ perceptions were ob-
Table 3
tained by a survey. Based on experts’ survey and combined imple- Total sum of effects given and received by indicators.
mentation of DEMATEL and ANP methods, a model was obtained
Indicators r+c rc
with relative weights of maintenance management indicators with
the objective to increase the vehicle fleet energy efficiency. T1 3.70 2.11
T2 3.82 2.31
T3 3.01 0.99
3.4. Results M1 4.35 1.64
M2 5.18 0.73
Based on the conducted surveys of experts and use of DEMATEL M3 5.91 0.08
method, according to (5) the total-influences matrix T was calcu- M4 3.88 2.43
E1 3.59 0.65
lated (Table 2). Besides threshold values were established by ex- E2 2.66 0.07
pert assessment (p = 0.2). Only the levels of interdependences of
indicators larger than the threshold value, established by experts,
were taken into account in continuation, while other interdepen-
dences were neglected i.e. considered independent, and their value Table 4
set to zero. Based on Table 2, it is obvious that there is an indepen- Matrix of total group influences –TD.

dent influence of the transport process in relation to indicators of Groups of Transport Maintenance Environment
the environment and to the indicators M1 and M4 of the mainte- indicators process (TP) process (MP) (E)
nance process. Transport process 1.32 1.22 0.00
Pursuant the calculated total given and received indicators ef- (TP)
fects, according to (7) and (8), in Table 3 can be seen that the most Maintenance 5.13 4.67 2.30
process (MP)
important indicator is Realisation of MP (M3) with the highest value
Environment (E) 1.54 1.34 0.54
of (r + c), followed by the indicator Average Vehicle Downtime (M2)
10560 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563

Table 5 in the Republic of Serbia. To this end, managers of the following

Total group delivered and received effects. companies were involved in the survey: Public Utility Company
Groups of indicators r+c rc (PUC) ‘‘Sanitation’’ Belgrade, Urban Public transport GSP ‘‘Beograd’’
Transport process (TP) 10.53 5.45 PUC ‘‘Belgrade water supply and sewage’’ and a private company
Maintenance process (MP) 19.33 4.87 for sales and distribution of automotive spare parts ‘‘Delmax’’ Ltd.
Environment (E) 6.26 0.58 In the period from June to August 2011 the surveys in form of
interviews were conducted with 19 managers in the selected com-
relationship between indicators and groups within the overall panies in terms of perception of nine indicators. Managers were
structure of the model. tasked to evaluate the indicators from 1 to 5 in relation to two
Pursuant calculation of the total-influence matrix – T within aspects.
DEMATEL and the adopted threshold value (p = 0.2) by experts, At the first evaluation, management determined to what extent
on Fig. 3 below shows the network relations map (NRM), in order was each observed indicator appropriate for measurement of ef-
to facilitate understanding of the model structure used in the ANP fects of the maintenance management, bearing in mind that the
method for calculation of relative weights. lowest score 1 indicates minor advantage (or a higher disadvan-
By incorporating total-influence matrix T in un-weighted tage), meanwhile score 5 designated major advantages.
supermatrix W and by using the total group influence matrix TD At the second evaluation, managers and experts from the Fac-
for calculation of weighted supermatrix WW, it has been obtained ulty of Transport and Traffic engineering have evaluated selected
a limited supermatrix W zw from the ANP, whose vectors represent indicators according to their actual values in the observed enter-
the relative weights of indicators in the model (Table 6). The Table prises (again the score 1 was for the least efficient parameter,
6 below shows that the most important indicators for improving and 5 for the most efficient indicators – highly valued in compa-
energy efficiency in the fleet maintenance management are: Main- nies/vehicle fleets).
tenance Plan Realisation (M3), Operational Plan Realisation Percent- In both above mentioned evaluations an average score was cal-
age (T1), Vehicle Payload Utilisation (T2). Indicator Maintenance culated for each indicator and by the implementation of the pro-
Plan Realisation (M3) with the highest value of relative weight of posed model with relative weights were obtained realised
0.233 belongs to the maintenance process group. However, subse- indicators values, which when summed give an overall score S0
quent two indicators by significance are the Operational Plan Real- and S00 in the observed companies.
isation Percentage (T1) and Vehicle Payload Utilisation (T2) both The total score in the evaluation of indicators benefits show
belonging to the transport process group. The indicators with the some degree of awareness among managers in the companies
lowest level indicators were Planned Maintenance Percentage about the importance of maintenance management for increasing
(M4), with the relative weight of 0.032 and the Mean Time Between the fleet energy efficiency (S0 ). The total score in the evaluation
Failures (M1), with the relative weight of 0.044, belonging to the of indicators in relation to the realised value within the company
maintenance process group, as well as Percentage of Vehicle Road- shows the effectiveness and efficiency of managers in managing
worthiness in Accidents (E2), with relative weight of 0.046, which their fleet maintenance management (S00 ). The results of this survey
belongs to the environment group. are shown in detail below.
When you observe the relative weights of the groups, the most
important for the fleet maintenance management is the transport 4.1. Results of the accomplished evaluation
process, and the least important group is the environment.
Below are shown the results of the model implementation with Pursuant the results shown in Table 7, ‘‘Delmax’’ Ltd. company
the relative weights of indicators within several companies with executives have achieved the best overall score of S0 = 4.129. Exec-
vehicle fleets with the objective of managers’ evaluation in terms utives in this company have a highly developed perception of the
of maintenance management in these companies. importance of fleet maintenance management on profit. In this
sense, managers in this company have easily recognised the poten-
tials to achieve savings and increase profits solely in this field. Com-
4. Implementation of the model within transport companies
pared to managers from other companies, those from this company
have given the best score to the indicator of Maintenance Plan Real-
The Model shown with relative weights of selected indicators
isation (M3), which has the highest relative weight in the model.
has been implemented in several companies with vehicle fleets
With better realisation of the maintenance plan the availability of
the most appropriate vehicles in a state ‘‘ready for operation’’ in
Table 6 the required time periods as required by OP. It will affect the suc-
Model with relative indicator weights. cessful realisation of the OP, as well as better Vehicle Payload Utilisa-
Indicators Relative Rank of Relative tion, thereby increasing fleet energy efficiency, and thus the profit
weights importance weights of of the companies. Executives occupying the second place with an
fields overall score 4.092 were from the Urban Public transport – GSP
Operation Plan (OP) Realisation 0.195 2 TP = 0.472 ‘‘Beograd’’. Managers from the company PUC ‘‘Belgrade water sup-
percentage T1 ply and sewage’’ have achieved the lowest overall score with 3.120.
Vehicle payload utilisation T2 0.169 3
In terms of overall score by S00 dealing with effectiveness and
Vehicle fleet utilisation rate T3 0.108 5
Mean time between failures M1 0.044 8 MP = 0.431
efficiency of maintenance management, the company ‘‘Delmax’’
Mean vehicle downtime M2 0.122 4 managers once again have achieved the best S00 score with 3.968
Maintenance Plan (MP) 0.233 1 (Table 8). This company has achieved higher actual values of those
Realisation M3 indicators, which are significant in the model, compared to other
Planned maintenance 0.032 9
companies. Thus, for example ‘‘Delmax’’ achieved the highest value
percentage M4
Percentage of fleet 0.051 6 E = 0.097 of the indicator Maintenance Plan Realisation (M3), Operational Plan
roadworthiness E1 Realisation Percentage (T1) and Vehicle Payload Utilisation (T2). Effi-
Percentage of vehicle 0.046 7 cient maintenance management has provided more vehicles in the
roadworthiness in accidents state ‘‘ready for operation’’ in the required time periods, which af-
E2
fects better realisation of the Operational Plan, as well as achieving
 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563 10561

Table 7
Results of indicators’ evaluation regarding their adaptability for measuring the maintenance management effects in observed companies.

Relative weights of indicators PUC ‘‘Sanitation’’ GSP ‘‘Beograd’’ PUC ‘‘Belgrade water supply ‘‘Delmax’’ Ltd.
and sewage’’
Average score Realised value Average score Realised value Average score Realised value Average score Realised value
T1 – 0.195 3.000 0.585 4.800 0.936 2.714 0.529 4.000 0.780
T2 – 0.169 4.000 0.676 2.200 0.372 2.429 0.410 3.667 0.620
T3 – 0.108 5.000 0.540 4.400 0.475 3.714 0.401 3.000 0.324
M1 – 0.044 3.000 0.132 4.000 0.176 3.286 0.145 3.333 0.147
M2 – 0.122 4.000 0.488 4.000 0.488 3.429 0.418 4.667 0.569
M3 – 0.233 3.750 0.874 4.600 1.072 3.143 0.732 5.000 1.165
M4 – 0.032 3.250 0.104 4.200 0.134 3.571 0.114 3.667 0.117
E1 – 0.051 4.500 0.230 5.000 0.255 4.286 0.219 4.667 0.238
E2 – 0.046 3.250 0.150 4.000 0.184 3.286 0.151 3.667 0.169
Total score S’ 3.778 4.092 3.120 4.129

Table 8
Results of indicators’ evaluation in view of realised values in observed companies.

Relative weights of indicators PUC ‘‘Sanitation’’ GSP ‘‘Beograd’’ PUC ‘‘Belgrade water supply ‘‘Delmax’’ Ltd.
and sewage’’
Average score Realised value Average score Realised value Average score Realised value Average score Realised value
T1 – 0.195 1.200 0.234 4.000 0.780 2.625 0.512 5.000 0.975
T2 – 0.169 2.400 0.406 1.167 0.197 2.625 0.444 3.667 0.620
T3 – 0.108 3.400 0.367 3.833 0.414 3.375 0.365 4.667 0.504
M1 – 0.044 1.600 0.070 3.667 0.161 3.125 0.138 3.667 0.161
M2 – 0.122 1.400 0.171 3.500 0.427 3.125 0.381 3.000 0.366
M3 – 0.233 3.000 0.699 3.167 0.738 3.250 0.757 3.333 0.777
M4 – 0.032 2.400 0.077 4.000 0.128 2.750 0.088 3.000 0.096
E1 – 0.051 4.000 0.204 4.500 0.230 3.375 0.172 5.000 0.255
E2 – 0.046 1.600 0.074 3.500 0.161 3.500 0.161 4.667 0.215
Total score S00 2.301 3.236 3.017 3.968

greater profit. In addition, the efficient maintenance management the relative weights of maintenance management indicators with
has facilitated greatly the allocation to transport tasks of the most the primary objective to increase fleet energy efficiency.
appropriate vehicle, which increased vehicle payload utilisation, Some conclusions drawn from the developed model could be
thereby increasing their energy efficiency and indirectly improved that the indicators are interdependent in three observed groups/
profitability. Managers from GSP ‘‘Beograd’’ in this case were fields, with varied levels of interdependence and no apparent
ranked second with 3.236. Once again, managers of the Public Util- trade-off. The indicator of Maintenance Plan Realisation (M3) has
ity Company ‘‘Sanitation’’ had the lowest overall score S00 regarding an utmost importance for the efficient fleet maintenance manage-
the effectiveness and efficiency of the maintenance management ment. If all maintenance work orders were realised at the planned
with as low as 2.301. timeframe according to the maintenance plan (MP), it will almost
In all four companies a considerably better overall evaluation of always facilitate the most appropriate vehicle to the transport task
managers was regarding their perception of the importance of during the required time periods as required by the Operation Plan
maintenance management (S’) compared to the overall score for (OP), which will reduce transport and maintenance costs. Improv-
the efficiency of their maintenance management (S00 ). This is ing the value of this indicator is largely affected by the improve-
especially true in the company PUC ‘‘Sanitation’’, where the total ment in value of other interdependent indicators. The second in
score of S’ was 3.778, while the overall score of S00 was just 2.301. rank of significance for the maintenance management is the
Therefore, managers clearly recognise the importance of the fleet Operational Plan Realisation Percentage (T1) while the third in rank-
maintenance management but in matter of implementation they ing is for the Vehicle Payload Utilisation (T2) indicator. Other signif-
are lacking of will or realistic objectives in their own companies. icant indicators are Mean Vehicle Downtime (M2), Vehicle Fleet
Utilisation Rate (T3), Percentage of Fleet Roadworthiness (E1), Per-
centage of Vehicle Roadworthiness in Accidents (E2), Mean Time
5. Conclusions Between Failures (M1), and Planned Maintenance Percentage (M4).
The proposed model with the relative weights of indicators has
This paper analyses the impact of interdependent indicators been implemented for the evaluation of managers in several com-
used to measure the implementation effects of specific measures panies with road vehicle fleets in the Republic of Serbia. Executives
within the maintenance management in order to increase the at ‘‘Delmax’’ Ltd. have achieved the best overall score in terms of
vehicle fleet energy efficiency. Since an efficient fleet maintenance developed awareness of the importance of maintenance manage-
management needs to be observed altogether with the core (trans- ment to increase vehicle fleet energy efficiency (S’), as well as the
port) process and its environment, it is essential to evaluate all best evaluation of the maintenance management efficiency (S00 )
these indicators in these specific fields. Upon literature review compared to managers in other companies. Managers have
and our expert deliberation we have selected nine provisional indi- achieved higher scores regarding the importance of maintenance
cators for evaluation. A combination of DEMATEL and ANP methods management to attain an increase in fleet energy efficiency (S’)
was used for determining the level of interdependence of indicators than the scores in terms of actual efficient maintenance manage-
and to calculate their relative weights. A model was obtained with ment in their companies (S00 ). Based on overall evaluation of
10562 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563

managers in terms of efficient maintenance management, it is con- Lin, Y.-T., Yang, Y.-H., Kang, J.-S., & Yu, H.-C. (2011). Using DEMATEL method to
explore the core competences and causal effect of the IC design service
cluded that in the Public Utility Companies, as PUC ‘‘Sanitation,’’
company: An empirical case study. Expert Systems with Applications, 38(5),
GSP ‘‘Belgrade’’ PUC ‘‘Belgrade water supply and sewerage’’ there 6262–6268. http://dx.doi.org/10.1016/j.eswa.2010.11.092.
is a significant potential to improve a value of indicators with more Maróti, G., & Kroon, L. (2005). Maintenance routing for train units: The transition
important relative weights in the developed model, which will model. Transportation Science, 39(4), 518–525. http://dx.doi.org/10.1287/
trsc.1050.0116.
help them manage their maintenance to become more effective McKinnon, A. (1999). A logistical perspective on the fuel efficiency of road freight
and efficient. transport, Report on the Workshop ‘‘Improving Fuel Efficiency in Road Freight:
The Role of Information Technologies’’. Paris: International Energy Agency,
European Conference of Ministers of Transport.
Acknowledgements Milosavljević, N., Teodorović, D., Papić, V., & Pavković, G. (1996). A fuzzy sets theory
approach to the vehicle assignment problem. Journal of Transportation Planning
and Technology, 20(1), 33–47. http://dx.doi.org/10.1080/03081069608717578.
The research in this paper has been realised within the project Mobley, R.K., Higgins, L.R., & Wikoff, D. J. (2008). Maintenance engineering
‘‘Development of the Model for Managing the Vehicle Technical handbook, 7 Ed., Inc., New York, New York, USA: The McGraw-Hill
Companies. doi: 10.1036/0071546464.
Condition in Order to Increase its Energy Efficiency and Reduce Ex-
Momčilović, V., Papić, V., & Vujanović, D. (2007). Quality of the fleet maintenance
haust Emissions’’ supported by the Ministry of Science and Techno- management: A factor influencing sustainable development. In Proceedings of
logical Development of the Republic of Serbia. the international congress ‘‘transport science & technology congress’’. Prague,
Czech Republic, (pp. 457–464).
Niemira, M. P., & Saaty, T. L. (2004). An analytic network process model for
References financial-crisis forecasting. International Journal of Forecasting, 20(4), 573–587.
http://dx.doi.org/10.1016/j.ijforecast.2003.09.013.
Nikolopoulos, K., Metaxiotis, K., Lekatis, N., & Assimakopoulos, V. (2003). Integrating
Arts, R., Knapp, G. M., & Mann, L. J. (1998). Some aspects of measuring maintenance
industrial maintenance strategy into ERP. Industrial Management and Data
performance in the process industry. Journal of Quality in Maintenance
Systems, 103(3), 184–191. http://dx.doi.org/10.1108/02635570310465661.
Engineering, 4(1), 6–11. http://dx.doi.org/10.1108/13552519810201520.
Norat, G.S. (2008). Maintenance KPI’s, How to start? PdMtech Inc. White Paper.
Ashayeri, J., Teelen, A., & Selen, W. (1996). A production and maintenance planning
Vega Baja, PR: PdMtech Inc.
model for the process industry. International Journal of Production Research,
Papić, V., Medar, O., & Pejčić Tarle, S. (1999). Management of vehicle fleet
34(12), 3311–3326.
maintenance in conditions of sustainable development. Mobility & Vehicle
Bin, O. (2003). A logit analysis of vehicle emissions using inspection and
Mechanics, 25(2–3), 37–46. <http://www.mvm.fink.rs/journal/sajt/image/
maintenance testing data. Transportation Research Part D: Transport and
mobility1999/vol25,num2and3,1999/papic,vladimir,3.pdf> Retrieved.
Environment, 8(3), 215–227. http://dx.doi.org/10.1016/S1361-920903)00004-X.
Parida, A., & Kumar, U. (2009). Maintenance productivity and performance
Brandolese, M., Franci, M., & Pozzetti, A. (1996). Production and maintenance
measurement. In M. Ben-Daya, S. O. Duffuaa, A. Raouf, J. Knezevic, & D. Ait-
integrated planning. International Journal of Production Research, 34(7),
Kadi (Eds.), Handbook of maintenance management and engineering (pp. 17–41).
2059–2075.
London, UK: Springer.
Christensen, P., & Elvik, R. (2007). Effects on accidents of periodic motor vehicle
Partovi, F. Y. (2006). An analytic model for locating facilities strategically. Omega,
inspection on Norway. Accident Analysis & Prevention, 39(1), 47–52. http://
34(1), 41–55. http://dx.doi.org/10.1016/j.omega.2004.07.018.
dx.doi.org/10.1016/j.aap.2006.06.003.
Randhawa, S. U., Miller, S. G., Bell, C. A., & Montagne, P. E. (1998). A study of
Chung, S.-H., Lee, A. H., & Pearn, W. L. (2005). Analytic Network Process (ANP)
commercial vehicle safety alliance’s out-of-service criteria. Accident Analysis and
approach for product mix planning in semiconductor fabricator. International
Prevention, 30(1), 61–67. http://dx.doi.org/10.1016/S0001-457597)00062-6.
Journal of Production Economics, 96(1), 15–36. http://dx.doi.org/10.1016/
Rechnitzer, G., Haworth, N., & Kowadlo, N. (2000). The effect of vehicle
j.ijpe.2004.02.006.
roadworthiness on crash incidence and severity, Report No. 164. Monash
Coulter, K., & Sarkis, J. (2005). Development of a media selection model using the
University, Accident Research Centre, Clayton, Australia. <http://
analytic network process. International Journal of Advertising, 24(2), 193–215.
www.monash.edu.au/miri/research/reports/muarc164.pdf> Retrieved
Dhillon, B. S. (2002). Engineering Maintenance: A modern approach. Boca Raton,
Ruzzenenti, F., & Basosi, R. (2009). Evaluation of the energy efficiency evolution in
Florida, USA: CRC Press.
the European road freight transport sector. Energy Policy, 37(10), 4079–4085.
Fontela, E., & Gabus, A. (1974). DEMATEL, innovative methods. Report no. 2
http://dx.doi.org/10.1016/j.enpol.2009.04.050.
structural analysis of the world problematique. Geneva, Switzerland: Battelle
Saaty, T. L. (1980). The analytic hierarchy process: Planning, priority setting, resource
Geneva Research Institute.
allocation. New York, London: McGraw-Hill.
Fontela, E., & Gabus, A. (1976). The DEMATEL observer. Geneva, Switzerland:
Saaty, T. L. (1996). Decision making with dependence and feedback: The analytic
Battelle Institute. Geneva Research Center.
network process (1st ed.). Pittsburgh, PA, USA: RWS Publications.
Haghani, A., & Shafahi, Y. (2002). Bus maintenance systems and maintenance
Saaty, T. L., & Vargas, L. G. (1998). Diagnosis with dependent symptoms: Bayes
scheduling: Model formulations and solutions. Transportation Research Part A:
theorem and the analytic hierarchy process. Operations Research, 46(4),
Policy and Practice, 36(5), 453–482. http://dx.doi.org/10.1016/S0965-
491–502. <http://www.jstor.org/stable/223127> Retrieved.
856401)00014-3.
Samaras, Z., & Kitsopanidis, I. (2001). Methodology and results of the evaluation of
Johnson, P. D. (2002). Principles of controlled maintenance management. Lilburn,
alternative short tests applied in inspection and maintenance programmes.
Georgia, USA: The Fairmont Press Inc.
Transportation Research Part D: Transport and Environment, 6(2), 111–122.
Kamakaté, F., & Schipper, L. (2009). Trends in truck freight energy use and carbon
http://dx.doi.org/10.1016/S1361-9209(00)00016-X.
emissions in selected OECD countries from 1973 to 2005. Energy Policy, 37(10),
Tesfamariam, D., & Lindberg, B. (2005). Aggregate analysis of manufacturing
3743–3751. http://dx.doi.org/10.1016/j.enpol.2009.07.029.
systems using system dynamics and ANP. Computers and Industrial Engineering,
Lee, W.-S., Huang, A. Y., Chang, Y.-Y., & Cheng, C.-M. (2011). Analysis of decision
49(1), 98–117. http://dx.doi.org/10.1016/j.cie.2005.05.001.
making factors for equity investment by DEMATEL and Analytic Network
Tzeng, G.-H., Chiang, C.-H., & Li, C.-W. (2007). Evaluating intertwined effects in e-
Process. Expert Systems with Applications, 38(7), 8375–8383. http://dx.doi.org/
learning programs: A novel hybrid MCDM model based on factor analysis and
10.1016/j.eswa.2011.01.027.
DEMATEL. Expert systems with Applications, 32(4), 1028–1044. http://dx.doi.org/
Leung, L. C., Lam, K. C., & Cao, D. (2006). Implementing the balanced scorecard using
10.1016/j.eswa.2006.02.004.
the analytic hierarchy process & the analytic network process. Journal of the
Vujanović, D., Momčilović, V., Papić, V., & Bojović, N. (2011). The vehicle fleet
Operational Research Society, 57(6), 682–691. <http://www.jstor.org/stable/
maintenance management’s impact on energy efficiency. In Proceedings of the
4102352> Retrieved.
third regional conference on industrial energy and environmental protection in
Li, J.-Q., Mirchandani, P. B., & Borenstein, D. (2009). Real-time vehicle rerouting
south-eastern europe countries – IEEP 2011. Kopaonik, Serbia: Yugoslav Society
problems with time windows. European Journal of Operational Research, 194(3),
of Heat Transfer Engineers.
711–727. http://dx.doi.org/10.1016/j.ejor.2007.12.037.
Vujanović, D., Mijailović, R., Momčilović, V., & Papić, V. (2010). Energy efficiency as a
Li, C.-W., & Tzeng, G.-H. (2009). Identification of a threshold value for the DEMATEL
criterion in the vehicle fleet management process. Thermal Science, 14(4),
method using the maximum mean de-entropy algorithm to find critical services
865–878. http://dx.doi.org/10.2298/TSCI090719010.
provided by a semiconductor intellectual property mall. Expert Systems with
Waeyenbergh, G., & Pintelon, L. (2002). A framework for maintenance concept
Applications, 36(6), 9891–9898. http://dx.doi.org/10.1016/j.eswa.2009.01.073.
development. International Journal of Production Economics, 77(3), 299–313.
Lin, C.-W., Chen, S.-H., & Tzeng, G.-H. (2009). Constructing a cognition map of
http://dx.doi.org/10.1016/S0925-5273(01)00156-6.
alternative fuel vehicles using the DEMATEL method. Journal of Multi-Criteria
Wu, W.-W. (2008). Choosing knowledge management strategies by using a
Decision Analysis, 16(1–2), 5–19. http://dx.doi.org/10.1002/mcda.438.
combined ANP and DEMATEL approach. Expert Systems with Applications,
Lin, Y.-H., Chiu, C.-C., & Tsai, C.-H. (2008). The study of applying ANP model to assess
35(3), 828–835. http://dx.doi.org/10.1016/j.eswa.2007.07.025.
dispatching rules for wafer fabrication. Expert Systems with Applications, 34(3),
Wu, W.-W., & Lee, Y.-T. (2007). Developing global managers’ competencies using
2148–2163. http://dx.doi.org/10.1016/j.eswa.2007.02.033.
the fuzzy DEMATEL method. Expert Systems with Applications, 32(2), 499–507.
Lin, C.-L., Hsieh, M.-S., & Tzeng, G.-H. (2010). Evaluating vehicle telematics system
http://dx.doi.org/10.1016/j.eswa.2005.12.005.
by using a novel MCDM techniques with dependence and feedback. Expert
Yamamoto, T., Madre, J.-L., & Kitamura, R. (2004). An analysis of the effects of French
Systems with Applications, 37(10), 6723–6736. http://dx.doi.org/10.1016/
vehicle inspection program and grant for scrappage on household vehicle
j.eswa.2010.01.014.
 D. Vujanović et al. / Expert Systems with Applications 39 (2012) 10552–10563 10563

transaction. Transportation Research Part B: Methodological, 38(10), 905–926. Novel%20Hybrid%20MCDM%20Model%20Combined%20with%20DEMATEL%20

http://dx.doi.org/10.1016/j.trb.2004.02.001. and%20ANP%20with%20Applications. paper-3-IJOR-vol5_3_-Yu-Ping_Ou_Yang.
Yang, J. L., & Tzeng, G.-H. (2011). An integrated MCDM technique combined with pdf> Retrieved.
DEMATEL for a novel cluster-weighted with ANP method. Expert Systems with Zhu, G., & Pintelon, L. (2001). Integrated production maintenance management
Applications, 38(3), 1417–1424. http://dx.doi.org/10.1016/j.eswa.2010.07.048. (IPMM) as an enterprise approach to maintenance management. In Proceedings
Ou Yang, Y.-P., Shieh, H.-M., Leu, J.-D., & Tzeng, G.-H. (2008). A novel hybrid MCDM of IIE annual conference, CD-ROM.
model combined with DEMATEL and ANP with applications. International Zhu, G., Gelders, L., & Pintelon, L. (2002). Object/objective-oriented maintenance
Journal of Operations Research, 5(3), 160–168. <http://www.knu.edu.tw/lecture/ management. Journal of Quality in Maintenance Engineering, 8(4), 306–318.
%E8%AA%B2%E7%A8%8B%E8%B3%87%E6%96%99/ELECTRE/Exa.5.13.09/A%20 http://dx.doi.org/10.1108/1355251021048513.

View publication stats