C. Diallo et al.

An Integrated Approach for Spare Parts Provisioning

An Integrated Approach for Spare Parts Provisioning

C. Diallo1,7 , D. Aı̈t-Kadi2,3,5,6 and A. Chelbi3,4

1
Dalhousie University, Industrial Eng., 5269 Morris Street, Halifax, Nova-Scotia, Canada B3H-4R2
2
Université Laval, Mechanical Eng., 1065, avenue de la Médecine, Québec, Canada G1V-0A6
3
Centre Interuniversitaire de Recherche sur les Réseau d’Entreprise, la Logistique et le
Transport CIRRELT, Université Laval, Canada G1V-0A6
4
Centre de Recherche en Productique CEREP, Ecole supérieure des Sciences et Technique
de Tunis ESSTT, BP 56 Bab Menara 1008, Tunis, Tunisia
5
Centre Interdisciplinaire de Recherche en Réadaptation et Intégration Sociale CIRRIS
Université Laval, Canada G1V-0A6
6
Resident Member of the Académie Hassan II des Sciences et Techniques, Royaume du Maroc
7
Corresponding Author E-mail: [Link]@[Link]

Abstract. This article addresses the problem of spare parts identification and provisioning for
multi-component systems. A framework considering available technical, economical and strategic
information is presented. Mathematical models are proposed to determine, for each spare part,
the required quantity over a given planning horizon. The objective may be to maximize either
the reliability or the availability of the system. Analytical models are proposed to determine the
management parameters.

Key words: maintenance, spare parts, identification, inventory, optimization.

1. Introduction
When acquiring a system, one is often faced with the difficult question of identifying which of its
components may fail to operate on the horizon considered and for which spare parts stocks are to be
supplied. It should be noted that, contrary to standard parts, spare parts are designed for specific use
and they can be acquired only from the manufacturer of the system or its authorized representatives.
Their provisioning leadtime is generally long, if not often unknown. They cannot be resold easily.
Consumption is typically governed by a random process. These parts are, additionally, subject
to obsolescence and deterioration. Because they can be used to perform preventive replacement

Frontiers in Science and Engineering 1

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

(planned replacement) or corrective replacement (random replacement), classical mathematical

models of inventory management [25] cannot be applied directly. Another dimension to consider
stems from the fact that some parts may be remanufactured, or simply repackaged to be used as
spares. Others can only be used once. First, we will deal with the latter category of parts and
then we will tackle the first category in a subsequent section. Inventory management of spare parts
is a strategic and economical issue for all organizations that operate equipment with operating
characteristics deteriorating with age and use. Interruption of service can have high economical
consequences. The unavailability of a component is certainly very detrimental both in terms of
cost and continuity of production or service [30]. On the other hand, maintaining a high stock of
spare parts can be expensive.
Section 2 of this article presents several decision tools for the identification of parts to be kept in
stock. Section 3 deals with the determination of the spare parts quantity required to achieve a pre-
determined performance for non-repairable systems. Section 4 proposes a mathematical model
to find the quantity of repairable spare parts required to maintain a certain level of service. A
conclusion of this work is presented in section 5.

2. Spare Parts Identification

In the absence of data and information to assess the degradation of a component and to estimate
the probability of failure over the economic life of the system, we rely generally on the manufac-
turer’s recommendations. Manufacturers have often provided a list of components to keep in stock
based on their own feedback, on data from accelerated testing conducted according to standard
procedures or from more sophisticated analysis of the failure modes of the main components. To
stay in business, equipment manufacturers are expected not only to meet the needs identified by
the client, but also to anticipate these needs and to demonstrate that the products and services of-
fered are equivalent if not superior to what is available on the market. Nowadays, customers may
claim information that could help them gain maximum advantage from the system being acquired.
More and more customers require their suppliers to provide them with lifetime and degradation
information, analysis of failure modes, effects and criticality (FMEA), etc. Having access to these
data, it becomes possible to identify components for which spare parts are to be kept. Thus, if the
lifetime density function fi (.) or the lifetime cumulative distribution function Fi (.) or the failure
rate hi (.) or the reliability function Ri (.) of component i is known, then spare parts are to be held
if
Fi (t) > F ∗
where F ∗ is the maximum failure risk the buyer (customer) is willing to accept during the mission
duration of the system being bought.
It is also possible to base the decision on the criticality index obtained from the FMEA/FMECA
of the system or by referring to similar equipment or on recommendations from internal or external
experts. Figure 1 presents a generic process for identifying components for which spare parts
are expected. For each equipment under consideration, all available data on lifetimes, repair
times, suppliers, lead-times, etc., are gathered. If partial or no information is available, then the

Frontiers in Science and Engineering 2

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

Figure 1: The spare parts identification process.

Figure 2: Decision tree with filters.

decisions are based on the manufacturers’ recommendations or on information from owners of

similar equipment. When complete information is available, then the decision-maker chooses the

Frontiers in Science and Engineering 3

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

criteria according to which the components will be evaluated to determine if they should be on
the list of spare parts. A list of potential criteria is given in figure 1. According to the criteria
retained, evaluation and final decision-making methods are to be selected next. Each component
of the equipment is then evaluated according to the criteria and a total score is obtained. A list
of potential methods is also provided in figure 1. A ranking of the components based on the final
scores gives an ordered list of potential spare parts. Once the preliminary list of spare parts is
established, it is subjected to filters to select the parts to hold in stock and those to be supplied as
needed. An example of decision tree with filters is depicted in figure 2. This tree takes into account
the cost of acquisition or production, repair costs, delays, whether there are early signs of failure
or not, and if the component is a standard part or not. A standard part is a generic mass-produced
part readily available at reasonable to low cost (e.g., seals, nuts and bolts, high replacement rate
parts). Selected components are then ranked in order of importance. This classification allows
to pay more attention to the components considered as being more important, especially if the
list of spare parts includes a large number of components and the resources available are limited
or scarce. Decision criteria most often used to justify that a spare part must be kept in stock to
serve as a replacement are: criticality, reliability, availability, impacts of failure, failure rate and
maintenance costs incurred in case of failure. In most cases, for simplicity or ignorance of the
analysis tools available, only one or two of these criteria are considered in the analysis. For many
organizations, the total cost of maintenance criterion is often the one used in the decision process.
For each component i, we calculate the critical ratio RGi by
Indirect Costs
RGi =
Direct Costs
For more details on the direct and indirect costs, please refer to reference [9]. Any component i
with RGi ratio greater than 1, is then kept in store as a spare.

The analysis of failure modes, their effects and criticality, is increasingly used in industry to
ensure sustainable use of assets. Each component is associated with a criticality index (C) obtained
by multiplying the severity index (S), the probability of occurrence of the adverse event (O) and
the difficulty of detecting it (D ) (see [12, 6, 36])

C =S×O×D

For a component subject to several failure modes, a criticality index Ck is defined for each
failure mode k (see [16]):
Ck = KA · KE · αkp · βk · λp · T
where
A system or component whose criticality index exceeds a predetermined threshold will be
included in the list of potential spare parts.
In practice, the decision to store a part or not, may involve several criteria (costs, reliability,
frequency of failure, response time, etc.) In the event that more than three criteria must be con-
sidered, multicriteria decision-making methods (MCDM) or tools have to be used. Several studies

Frontiers in Science and Engineering 4

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

Ck : criticality index for failure mode k

KA : failure rate adjustment factor to compensate for actual operating conditions
KE : failure rate adjustment factor to compensate for actual environmental conditions
αk p : proportion of failures of component p due to failure mode k
βk : conditional probability that failure mode k will cause the failure
λp : failure rate of component p
T : mission length

published in the literature have successfully used multicriteria methods for the identification and
classification of equipment and spare parts (for example see [18] and [8]). The same tools can
be adopted to generate a list of components to include in the priority list of potential replacement
parts. Braglia et al. [7] used the AHP multicriteria classification method to classify the parts ac-
cording to the impact of the failure, utilization, inventory problems and characteristics of parts.
Schärlig [32] [31], Roy and Bouyssou [29] propose extensive reviews of multicriteria methods.
Many computer programs and web sites offer the possibility of carrying out a multicriteria clas-
sification. Hammami [22] presents a very detailed review of computer programs dealing with
multicriteria decision support. Eisenhawer et al. [17] propose a method of approximate reasoning
based on fuzzy logic to establish a priority list of items to keep in stock for a nuclear facility. Once
the components to be considered as spare (replacement) parts are identified, one has to determine
the required quantities to be acquired during a given time period in order to achieve the expected
performance levels. The following section will discuss models and methods for calculating the
quantities of parts required during the economic life cycle of the system.

3. Determination of the required quantity of non-repairable

spares
For each component requiring spare parts, it is important to estimate the required amount of spares
needed throughout the economic life cycle of the equipment. To achieve this, one must estimate the
average number of replacements at failure and, where applicable, the average number of preventive
replacements. In this study, the data and information available will primarily be used to determine
the density function f (.), the distribution function F (.), the survivor function or reliability R(.),
the failure rate h(.) associated with the lifetime of the component under consideration. Figure 3
shows how field data is processed to obtain one of the four reliability characteristics. Knowledge
of any one of these four characteristics is sufficient to obtain the other three. Table 1 recalls the
relationships between these four functions.
For a component with lifetime density function f (t) and negligible replacement duration, the
average number of replacements at failure M (t), with replacements carried-out with new spare

Frontiers in Science and Engineering 5

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

Figure 3: Field data processing diagram.

f (t) F (t) R(t) h(t)

Rt R∞ f (t)
f (t) – 0
f (x)dx t
f (x)dx R∞
t
f (x)dx
dF (t) dF (t)
F (t) – 1 − F (t)
dt [1 − F (t)]dt
−dR(t) −dR(t)
R(t) 1 − R(t) –
dtR R(t)dt
− 0t h(x)dx
Rt Rt
h(t) h(t).e 1 − e− 0 h(x)dx
e− 0 h(x)dx
–
Table 1: Basic reliability relationships.

parts during a mission of length t, satisfies the following fundamental renewal equation:
Z t
M (t) = F (t) + M (t − x)f (x)dx.
0

If F (i) (t) denotes the i–fold convolution of F (t) with itself, then
∞
X
M (t) = F (i) (t).
i=1

If at failure the component is minimally repaired without affecting its failure rate h(t), then the
average number of failure during the time interval [0,t] is given by
Z t
M (t) = h(x)dx.
0

Frontiers in Science and Engineering 6

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

When the repair or replacement durations are random, the average number of failures during
the time interval [0,t] is given by
X∞
M (t) = G(i) (t)
i=1

where G(i) (t) denotes the i–fold convolution of G(t) with itself, g(t) = dG(t)
dt
being the convolution
of the lifetime density function f (t) with the repair or replacement duration density function v(t)
such that Z t
g(t) = f (t − x)v(x)dx.
0

Closed-form expressions for the renewal function M (t) are only known to a relatively short list
of distributions used in reliability and maintenance modeling, such as the Uniform, Exponential
and Erlang distributions. However, several numerical methods have been proposed to compute
M (t) (see [2, 24, 10]). Diallo and Aı̈t-Kadi [15] have proposed an approximation based on the
Dirac function to compute g(t) and M (t) when repair or replacement durations are not negligible.
Once M (t) is known, the average number n of spare parts required for the time interval [0,t[ is
obtained by rounding its value to the next integer:

n = dM (t)e.

For a component replaced at failure and after T units of time, according to the age replacement
policy (ARP), the upper bound of the expected number of spare parts na for a mission of length t
is given by [5]: & '
t · [1 − R(T )]
na = RT .
0
R(x)dx
If the component is replaced at failure or at predetermined instants kT (k = 1, 2, 3, · · · ) re-
gardless of its age and state, according to the block replacement policy (BRP), then the expected
number of spare parts nb for a mission of length t is given by [5]:

nb = dk [M (T ) + 1] + M (t − kT )e

with kT ≤ t < (k + 1)T .

For some applications, a spare part can be considered as a stand-by component in the reliability
point of view. The determination of the system reliability RS (t, n) for a stand-by structure with
n components allows to calculate the number of spares n to keep in stock to achieve a desired
predetermined reliability level R∗ for a given mission duration t . This is equivalent to finding the
smallest integer n∗ such that:

RS [t, n∗ (t)] ≥ R∗
Z ∞
∗ (t)+1]
f [n (x)dx ≥ R∗
t

Frontiers in Science and Engineering 7

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

Figure 4: Computation procedure for determining the number of spare parts.

where f (i) (t) denotes the i-fold convolution of f (t) with itself. Once RS (t, n) is known the number
of spare parts to keep can be computed using a simple iterative algorithm from Aı̈t-Kadi et al. [1]
and depicted by figure 4.
Table 2 in Diallo et al. [13] gives the expressions of RS [t, n∗ (t)] for different configurations
of the problem. In the general case, when k components are in operation and (n − k) are kept in
stock, the expression of RS [t, n∗ (t)] becomes:
RS [t, n∗ (t)] = RS [t, n∗ (t), k]
where
n−k
∗ −kλt
X (kλt)j
RS [t, n (t), k] = e .
j=0
j!
Note that the number of spare components can be calculated using a predetermined availability
value A∗ . The approach is to find n∗ (t) such that A[t, n ∗ (t)] ≥ A∗ , where A[t, n ∗ (t)] is the
availability of a stand-by structure consisting of n∗ (t) reserve components and one component in
operation. Diallo et al. [14] proposed a mathematical model for the maximization of the system’s
availability under joint preventive maintenance and spare parts provisioning strategy.

4. Determination of the required quantity of repairable spares

A failed component, according to its degradation state, is repaired and put back in a “as new
condition” or put in a state where it can resume operation. The acquisition cost of these repairable
components is generally high. If the repair is found to be not feasible for technical, economical or
other reasons, the failed component is sent to a recycling or disposal facility. The method based
on the stand-by structure presented in the previous section can also be used here to determine the
number of repairable components to use in a single system.
For a repairable component (as good as new) with failure rate h(t) = λ and repair rate µ(t) = µ,
the expression of RS [t, n∗ (t)] is given by
RS [t, n∗ (t)] = e−zt

Frontiers in Science and Engineering 8

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

where ∗
(1 − γ)2 γ n (t) λ
z=
1 − γ n∗ (t)+1 {1 + [n∗ (t) + 1] (1 − γ)}
and γ = λ/µ.

Inventory management of a fleet of repairable systems is much more complex than that of
standard nonrepairable components. This problem occurs mainly in mining facilities, fleet mainte-
nance, oil industry, civil aviation, military nuclear industry, etc. The complexity of the problem is
usually due to the randomness of the failures, the varying restoration times, the uncertainties about
the state of degradation of the component. Several analytical models, dealing with some variants of
the problem of inventory management for repairable components, were published in the literature.
Figure 5 presents, schematically, a configuration with two bases (operations centers) and a central
depot for the maintenance of components and systems. The objective is to determine systems and
components stock levels, at the bases and at the depot, in order to guarantee a given level of service
at minimum cost. Since the repairs take time and given that any non-availability of the systems
results in high costs, an inventory of systems in operational state is maintained at each base. Once
a failure occurs at a base, a replacement system is taken from the stock and is immediately put
in place to ensure continuity of service. The failed system is either repaired on site (at the base)
with a given probability or sent to the repair center at the depot with a certain shipment leadtime.
According to the level of degradation of the components, the repair can be undertaken or not. If
the repair is not possible, a new one is ordered. When the repair is possible, spare components
are used to carry-out the replacement. If replacement components are not available, an order is
immediately placed. After repair, the system is sent back to the originating base (decentralized
management) or kept in stock at the depot (centralized management). The repair time depends on
the availability of spare parts, the repair capacities and the workload in the repair shops.
For the problem illustrated above, consider that the fleet has N machines (systems) in opera-
tion. These systems are independent and identically distributed each with a failure rate λ(t) and a
repair rate µ(t). A stock of y spare systems is held. A repair shop, consisting of c parallel channels,
is used to repair systems that fail during operation. The production of goods or services is stopped
whenever a total of (N + y) systems have failed. Taylor and Jackson [35] were the first to apply
queueing theory to solve this spare parts provisioning problem. Several studies have subsequently
been devoted to the subject. These include, among others, the work done by Sherbrooke [33, 34]
and others such as [4, 23, 27, 28, 3, 11, 26]. Two cases are distinguished. In the first case, the
capacity of repair stations are assumed sufficient and therefore no waiting line is formed (see Sher-
brooke [33, 34] ). The modeling assumptions adopted lead to lower stocks than would be required
to achieve the specified level of service [3]. In the second case, finite repair capacity is assumed.
The analytical treatment becomes more complex as shown in [27, 3, 19] . A comprehensive review
of articles on inventory management of repairable systems is carried out by Guide and Srivastava
[21]. We will now present the model proposed by Gross et al. [19] to address the problem il-
lustrated in figure 5. This is a Markovian model for a single echelon repairman problem whose
transition diagram is shown in figure 6.
The number y of spare parts to keep in stock should allow to reach the service level N S defined

Frontiers in Science and Engineering 9

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

Figure 5: Flow of parts in the two-echelon repairman problem.

Figure 6: Transition diagram of the repairman model.

as the probability of having at least one spare machine in stock.

Pi is the steady-state probability of having i failed machines awaiting repair or being repaired.
When the failure and repair rates are constant such that λ(t) = λ and µ(t) = µ, then the Pi are
given by Gross et al. [19]:
  i
 Ni λ

 P0 for i = 1, · · · , y
i! µ




  i
N yN ! λ

for c > y Pi = P0 for i = y + 1, · · · , c − 1


 (N − i + y)!i! µ
 i
N yN !


 λ
P0 for i = c, · · · , y + N


i−c
(N − i + y)!c c! µ


Frontiers in Science and Engineering 10

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

  i
i
N λ P

for i = 1, · · · , c − 1
 0
i! µ




  i
Ni λ

for c ≤ y Pi = P0 for i = c, · · · , y − 1


 ci−c c! µ
 i
N yN !


 λ
P0 for i = y, · · · , y + N


i−c
(N − i + y)!c c! µ


The expression of P0 is given by

"
y y+N  i −1
c−1
#

 X i
 i X i
 i X y
N λ N λ N N! λ
1+ + + for c ≤ y




 i! µ ci−c c! µ (N −i+y)!ci−c c! µ
i=1 i=c i=y+1
P0 = "
y−1 y+N  i −1
c
#

 X i
 i X i
 i X y
 N λ N N! λ N N! λ


 1+ i! µ
+ (N −i+y)!c! µ
+ (N −i+y)!ci−c c! µ
for c > y

i=1 i=y i=y+1

The source population (the fleet of machines+spares) is usually finite (N + y), therefore the
probability that a failure is about to occur when there are i broken machines in the repair shop is
given by Qi (see Gross et al. [19]):
N · Pi


 Py+N for i = 0, · · · , y − 1
N − i=y (i − y) Pi


Qi =
(N − i + y) · Pi
for i = y, · · · , y + N


N − y+N
 P
i=y (i − y) Pi


Once the Qi are known, it suffices to find the smallest integer y such that
y−1
X
Qi ≥ N S
i=0

Logistical delays are not taken into account by this model. Moreover, the assumption that fail-
ure and repair rates are constant, sometimes causes decision-makers to question the validity of the
results obtained by this model. Gross [20] studied the sensitivity of the model to the exponentiality
assumption and derived easy rules of thumb to estimate the error induced by such a hypothesis.
Kim et al. [27] present an algorithm to determine the level of the stock of spare parts required in
the different bases in a two-tier system with a central depot. Note that in their model, the depot
does not have a stock of spare parts. It only carries out repairs. The authors propose a mathemati-
cal model to minimize the average total cost of inventory management while satisfying a minimum
service level. They also use the results of queueing theory to determine the probability of shortages
and the likelihood of excess inventory. Logistical delays are included in their model.

Frontiers in Science and Engineering 11

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

5. Conclusion
An integrated approach for the identification and management of spare parts has been proposed.
We have described a methodology for the identification of components for which replacement parts
must be kept. For each spare part, analytical models are presented for determining the quantities
required over a given operating horizon. Models of inventory management were then proposed
for repairable systems. Several factors that affect system performance such that the replenishment
leadtime and random demand are taken into account in the mathematical models presented.
Because the machines and their operating environment tend to change over time, it is wise to
frequently update the management parameters and decision variables to take into account any
technological, economical and strategic change. It should also be worthwhile to implement proper
maintenance procedures and monitoring of spare parts when they are stored for long periods of
time by an appropriate control of their environment (moisture control, greasing, repositioning).
Future research work should focus on the determination of inventory levels at bases and depot
levels when risk-pooling initiatives are possible and lateral transshipments allowed. Setting of
inventory levels of repairable spare parts also called “rotables” in some industries will be stud-
ied when scheduled inspections create random and deterministic demand on top of the demand
generated by random failures.

Acknowledgements
The authors would like to thank the anonymous referees and the editors for their useful suggestions.

References
[1] Aı̈t-Kadi D, Diallo C, Chelbi A. Spare parts identification and provisioning models, Journal
of Decision Systems; 12 (2003), No. 1, 47-65.

[2] Aı̈t-Kadi D, Chelbi A. Procédure numérique de calcul de la fonction de renouvellement,

Technical Report, Université Laval, 1998.

[3] Albright SC. Approximation to the stationary distribution of a multi-echelon repairable-

item inventory system with finite sources and repair channels, Naval Research Logistics; 36
(1989), No. 2, 179-195.

[4] Balania AR, Gross D, Soland RM. Optimal provisioning for single-echelon repairable item
inventory control in a time-varying environment, IIE Transactions; 21 (1989), No. 3, 202-212.

[5] Barlow RE, Proschan F. Mathematical theory of reliability. New-York NY: John Wiley and
Sons, 1965.

[6] Bouti A, Aı̈t-Kadi D. A state-of-the-art review of FMEA/FMECA, International Journal of

Reliability, Quality and Safety Engineering; 1 (1994), No. 4, 515-543.

Frontiers in Science and Engineering 12

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

[7] Braglia M, Grassi A, Montanari R. Multi-attribute classification method for spare parts inven-
tory management, Journal of Quality in Maintenance Engineering; 10 (2004), No. 1, 55-65.

[8] Chelbi A, Aı̈t-Kadi D. Classement des équipements par ordre de priorité pour la maintenance,
Journal of Decision Systems; 11 (2002), No. 1, 91-108.

[9] Chelbi A. Stratégies optimales d’inspection et de remplacements d’équipements sujets á des

défaillances aléatoires [PhD Thesis]. Québec, Canada: Université Laval; 1996.

[10] Cléroux R, McConalogue DJ. A numerical algorithm for recursively defined convolution in-
tegrals involving distribution functions, Management Science; 22 (1976), 1138-1146.

[11] Dada M. A two-echelon inventory system with priority shipments, Management Science; 38
(1992), No. 8, 1140-1153.

[12] Dekker R, Kleijn MJ, Rooij PJ. A spare parts stocking policy based on equipment criticality,
International Journal of Production Economics; 56 (2005), No. 1, 69-77.

[13] Diallo C, Aı̈t-Kadi D, Chelbi A. Integrated Spare Parts Management. In Ben Daya M, Duf-
fuaa SO, Raouf A, Knezevic J, Ait-Kadi D, editors. Handbook of Maintenance Management
and Engineering. London. Springer; 2009.

[14] Diallo C, Aı̈t-Kadi D, Chelbi A.(s, Q) spare parts provisioning strategy for periodically re-
placed system, IEEE Transactions on Reliability; 57 (2008), No. 1, 134-139.

[15] Diallo C , Aı̈t-Kadi D. A simple approximation to compute the mean number of failures for
a repairable system and application to the availability determination in Proceedings of the
Congres International de Génie industriel Trois-Rivieres, Canada, 2007.

[16] Ebeling CE. An Introduction to reliability and maintainability engineering. Long Grove IL:
Waveland Press, Inc.; 2010.

[17] Eisenhawer SW, Bott TF, Jackson JW. Prioritizing the purchase of spare parts using an ap-
proximate reasoning model, Proceedings of the Annual Reliability and Maintenability Sym-
posium; 2002, 20-25.

[18] Gajpal PP, Ganesh LS, Rajnedran C. Criticality analysis of spare parts using the analytic
hierarchy process, International Journal of Production Economics; 35 (1994), No. 1-3, 293-
297.

[19] Gross D, Kahn HD, Marsh JD. Queueing models for spares provisioning, Naval Research
Logistics Quarterly; 24 (1977), No. 4, 521-536.

[20] Gross D. Sensitivity of queueing models to the assumption of exponentiality, Naval Research
Logistics Quarterly; 22 (1975), No. 2, 271-287.

Frontiers in Science and Engineering 13

An International Journal Edited by Hassan II Academy of Science and Technology
C. Diallo et al. An Integrated Approach for Spare Parts Provisioning

[21] Guide VDR, Srivastava R. Repairable inventory theory: Models and applications, European
Journal of Operational Research; 102 (1997), 1-20.
[22] Hammami A. Modélisation technico-économique d’une chaine logistique dans une entreprise
réseau [PhD Thesis]. Québec, Canada: Université Laval - Ecole Nationale Supérieure des
Mines de St-Etienne; 2003.
[23] Hausman WH, Erkip N. Multi-echelon vs. single-echelon inventory control policies for low-
demand items, Management Science; 40 (1994), No. 5, 597-602.
[24] Huiskonen J. Maintenance spare parts logistics: special characteristics and strategic
choices,International Journal of Production Economics; 71 (2001), 125-133.
[25] Kennedy WJ, Patterson JW, Fredendall LD. An overview of recent literature on spare parts
inventories, International Journal of Production Economics; 76 (2002), 201-215.
[26] Kukreja A, Schmidt CP. A model for lumpy demand parts in a multi-location inventory system
with transshipments. Computers and Operations Research; 32 (2005), 20591775.
[27] Kim JS, Shin kc, Yu hk. Optimal algorithm to determine the spare inventory level for a
repairable-item inventory system, Computers and Operations Research; 23 (1996), No. 3,
289-297.
[28] Miller BL. A queueing reward system with several customer classes, Management Science;
16 (2005), No. 3, 234-245.
[29] Roy B, Bouyssou D. Aide multicritere á la décision: méthodes et cas, Paris: Economica;
1993.
[30] Sarker R, Haque A. Optimization of maintenance and spare provisioning policy using simu-
lation, Applied Mathematical Modelling; 24 (2000),751170.
[31] Schärlig A. Pratiquer Electre et Prométhée. Lausanne: Presses polytechniques romandes;
1996.
[32] Schärlig A. Décider sur plusieurs criteres: panorama de l’aide á la décision multicritere.
Lausanne: Presses polytechniques romandes, 1985.
[33] Sherbrooke CC. Optimal Inventory Modeling of systems. New York, NY: John Wiley; 1992.
[34] Sherbrooke CC. Multiechelon inventory systems with lateral supply, Naval Research Logis-
tics; 39 (1992), 29-40.
[35] Taylor J, Jackson RRP. An application of the birth and death process to the provision of spare
machines, Operations Research; 5 (1954), No. 4, 95-108.
[36] Wei BC. A unified approach to failure mode effects and criticality analysis, Proceedings of
the Annual Reliability and Maintenability Symposium; (1991), 260-271.

Frontiers in Science and Engineering 14

An International Journal Edited by Hassan II Academy of Science and Technology