Sei sulla pagina 1di 5

IFAC Conference on Manufacturing Modelling,

IFAC
IFAC Conference
and on
Conference on Manufacturing
Manufacturing Modelling,
Management
IFAC Conference
Management and on Manufacturing Modelling,
Control
Control Modelling,
Management
June 28-30,
Management and
2016. Control
Troyes,
and Control France Available online at www.sciencedirect.com
June
June 28-30, 2016. Troyes, France
June 28-30,
28-30, 2016.
2016. Troyes,
Troyes, France
France

ScienceDirect
IFAC-PapersOnLine 49-12 (2016) 881–885
On
On the
the joint
joint optimization
optimization of
of the
the periodic
periodic
On the joint
preventive optimization
maintenance andofthe
thespare
periodic
parts
preventive maintenance
preventive maintenance and
and the
the spare
spare parts
parts
inventory problem
inventory problem
inventory problem
Anis Mjirda ∗∗
Rachid Benmansour ∗∗ ∗∗ Hamid Allaoui ∗∗∗
Anis MjirdaRachid
∗ Benmansour ∗∗ Hamid Allaoui ∗∗∗ ∗∗∗
Anis
Anis MjirdaRachid
Mjirda ∗ Benmansour
Gilles
Rachid Goncalves ∗∗∗
Benmansour ∗∗ Hamid Allaoui ∗∗∗
∗∗∗
Hamid Allaoui
Gilles
Gilles Goncalves ∗∗∗
Goncalves ∗∗∗
Gilles Goncalves

∗ Institut de recherche technologique Railenium, (e-mail: anis.mjirda@
∗ Institut de recherche technologique Railenium, (e-mail: anis.mjirda@
∗ Institut de recherche technologique Railenium,
Institut de recherche technologique Railenium, (e-mail:
railenium.eu).
railenium.eu). (e-mail: anis.mjirda@
anis.mjirda@
∗∗
∗∗ LAMIH, Universit de railenium.eu).
Valenciennes et
railenium.eu).du Hainaut Cambrsis (e-mail:
∗∗ LAMIH, Universit de Valenciennes et du Hainaut Cambrsis (e-mail:
∗∗ LAMIH, Universit de Valenciennes
Valenciennes et
et du
du Hainaut
Hainaut Cambrsis
rachid.benmansour@univ-valenciennes.fr) Cambrsis (e-mail:
(e-mail:
LAMIH, Universit de
rachid.benmansour@univ-valenciennes.fr)
∗∗∗
∗∗∗
rachid.benmansour@univ-valenciennes.fr)
LGI2A, Universit de l’Artois (e-mail: {hamid.allaoui,
rachid.benmansour@univ-valenciennes.fr)
∗∗∗ LGI2A, Universit de l’Artois (e-mail: {hamid.allaoui,
∗∗∗ LGI2A, Universit
LGI2A, Universit de
de l’Artois (e-mail: {hamid.allaoui,
l’Artois (e-mail:
gilles.gocalves}@univ-artois.fr)
{hamid.allaoui,
gilles.gocalves}@univ-artois.fr)
gilles.gocalves}@univ-artois.fr)
gilles.gocalves}@univ-artois.fr)
Abstract: In this paper, we propose a new model for the joint optimization of the periodic
Abstract:
Abstract: In
In this paper, we propose aa new model for the jointIt optimization of the
an periodic
preventive
Abstract:
preventive In this
this paper,
maintenance
maintenance paper,and we
and
propose
we spare
propose
spare
parts
partsa new
new model
model for
inventory
inventory for the
the joint
problem.
problem. jointIt
optimization
consist to find
optimization
consist to
of
of the
find the
an
periodic
optimal
periodic
optimal
preventive
maintenance
preventive maintenance
schedule of and
maintenance and spare
M machine
spare parts inventory
over ainventory
parts problem.
cyclic horizon
problem. It
composed
It consist to
of τ periods
consist to find
find an
with
an optimal
taking
optimal
maintenance
maintenance
into account the schedule
schedule of
of
inventory M
M machine
machine
management over
over a
a cyclic
cyclic
of aspare horizon
horizon
parts. composed
composed
A computational of
of τ
τ periods
periods
experiment with
wastaking
with taking
made
maintenance
into account schedule
the of
inventory M machine
management over of cyclic
spare horizon
parts. A composed
computational of τ periods
experiment with
was taking
made
into
on
into a account
set
account of the
instances
the inventory
generated
inventory management
to
management show of
the
of spare parts.
efficiency
spare parts. andA
A computational
the limit
computational of our experiment
model.
experiment was
was made
made
on
on a
a set
set of
of instances
instances generated
generated to
to show
show the
the efficiency
efficiency and
and the
the limit
limit of
of our
our model.
model.
on a set of instances generated to show the efficiency and the limit
© 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. of our model.
Keywords: Periodic, Maintenance, Inventory Management, Mathematical Programming.
Keywords:
Keywords: Periodic, Maintenance,
Maintenance, Inventory Management, Management, Mathematical Programming. Programming.
Keywords: Periodic, Periodic, Maintenance, Inventory Inventory Management, Mathematical Mathematical Programming.
1. INTRODUCTION example the Inventory Routing Problem (Mjirda et al.
1. INTRODUCTION
1. INTRODUCTION example the
example the Inventory
Inventory Routing
Routing Problem
Problem (Mjirda(Mjirda et et al.
al.
1. INTRODUCTION (2014)).
example
(2014)). the Inventory Routing Problem (Mjirda et al.
(2014)).
Preventive maintenance is the systematic inspection of Many (2014)).
Preventive
Preventive
incipient maintenance
maintenance
failures, is the
is the systematic
on a pre-determined systematic
schedule, inspection
inspection
that may of Many variants
of variants
of the joint optimization of spare parts
of the
the jointactivity
optimization of spare
spare parts
parts
Preventive
incipientfor maintenance
failures, on is the systematic inspection of Many
inventory
Many variants
and
variants of
of the joint
maintenance optimization
jointactivity of
were presented
optimization of spare in the
parts
incipient
happen
incipient failures,
a givenon
failures, on aa
system
a
pre-determined
pre-determined
(machines, trains,
pre-determined
schedule,
schedule,
schedule,
that etc).
that
vehicles,
that
may inventory and maintenance
may
may inventory
literature and maintenance
several models activity
have were proposed.
were
been presented
presented in
in the
the
These
happen
happen for
for a
a given
given system
system (machines,
(machines, trains,
trains, vehicles,
vehicles, etc).
etc). inventory
literature and
and maintenance
several models activity
have were
been presented
proposed. in the
These
It is
happen known, that
for a that a
givenasystem well-maintained
(machines,system system performs
trains, performs
vehicles, etc). bet- literature
models andand several
approaches models
proposed have been
usually proposed.
are These
of a stochastic
ItIt is known,
isenhance
known, that well-maintained
aa well-maintained bet- literature and several models have been proposed. These
ter,
It known, its efficiency and extendssystem its life performs
cycle. In fact, bet- models and
bet- models and approaches
approaches proposed
proposed usuallyisare
usually are of aatostochastic
of stochastic
ter,isenhance
ter,
analysisenhance that
its
its
well-maintained
efficiency
efficiency
of maintenance and extends
and
costs extends
indicates
system
its life performs
itsthat
life cycle.
cycle.
a repair In fact,
In fact, nature
models
nature
per- nature and
a probabilistic
approaches
a probabilistic
distribution
proposed usuallyisare
distribution
used
used
describe
of atostochastic
describe
ter, enhance
analysis of its efficiency
maintenance and
costs extends
indicates its life
that cycle.
a In
repair fact,
per- the
nature and
failure
and a probabilistic
properties
a probabilistic distribution
(see distribution is
the work ofisHorenbeek used
used to
to describe
et al.
describe
analysis
formed
analysis inof
of maintenance
the reactive or run
maintenance costs
costs toindicates
failure mode
indicates that
that is,aa on
repair per-
average,
repair per- the failure
the
(2013)failure
for aproperties
properties
review). (see the
(see the work
work of of Horenbeek
Horenbeek et al.
et al.
formed
formed in
in the
the reactive
reactive or
or run
run to
to failure
failure mode
mode is,
is, on
on average,
average, the
(2013)failure
for aproperties
review). (see the work of Horenbeek et al.
about
formedthreethree times
in thetimes
reactive higher than
or run the
tothe same
failure repair
mode made
is,made within
on average, (2013) for aa review).
(2013)
this for work,review).
aabout
about
about three
scheduled
three times higher
higher
or preventive
times higher
than
than
than mode the
the
same
same
(Mobley
same
repair
repair
repair made
(2002)).
made
within
Thus, In
within
within In this work,
we consider a deterministic approach and
we aconsider
consider
a scheduled
scheduledgive
acompanies
a scheduled
or preventive
or
or
preventive
an interest mode
preventive
mode
mode (Mobley
(Mobley
of maintaining
(Mobley
(2002)).
(2002)).
their systems.
(2002)).
Thus,
Thus,
Thus,
In
In this
models
this work,
to solve
work, we specificaaa variant
we aconsider
deterministic
deterministic
deterministic
approach
of theapproach
problem.and
approach
and
andIt
companies
companies give
give an
an interest
interest of
of maintaining
maintaining their
their systems.
systems. models
models
concerns to
tothesolve
solve a
periodic specific
specific variant
variant
maintenance of
of
problem the
the problem.
problem.
proposed It
It
by
However,
companies the
giveassociated
an interest costsofwith preventivetheir
maintaining maintenance
systems. models concerns tothesolve a specific
periodic variantproblem
maintenance of the proposed
problem. by It
However,
However,
can the
the
be significant. associated
associated costs
In fact,costs
more with
with
than preventive
preventive maintenance
to be taken Grigoriev
one cost ismaintenance concerns
concerns the
the periodic
et al.
periodic maintenance
(2006)maintenance
Grigoriev et problemproblem proposed
al. (2006)proposed
in wich we by
by
However,
can be be the associated
significant. In fact,
fact,costs
more with preventive
than one maintenance
one costoperation
is to to be taken
taken Grigoriev et al. (2006) Grigoriev et al. (2006) in wich we
can
into
can be significant.
account:
significant. i ) the In costmore
In fact, than
than one cost
of maintenance
more is
is to be
costoperation be takenand Grigoriev
add
Grigoriev
add the
et
et al.
the inventory al. (2006)
theis inventory
inventory (2006) Grigoriev
management
Grigorievof
management of
et al.
etthe
the
(2006)
al.spare
(2006)
spare
in
in wich
parts.
parts.
A new
wich
A new
we
we
new
into
into
ii ) the account:
account: i
i )
) the
the
costs ofi )production cost
cost of
of maintenance
maintenance
losses operation
during theoperation
maintenance and
and add
model
add the presented
inventory management
and solved
management on of
a
of the
set
the ofspare
spare parts.
instances
parts. A
adapted
A new
into
ii )) theaccount:
the costs the
of production
production cost of maintenance
losses during
during the maintenance
maintenance and modelliterature.
is presented
presented and solved solved on aa set set of instances
instances adapted
ii ) the costs
activity
ii of
(Grigoriev
costs of production losses
et al. (2006)).
losses duringMoreover,the in order to model
the maintenance from
model
from
is
is presented and
literature. and solved on on a set of of instances adapted
adapted
activity
activity
restore (Grigoriev
(Grigoriev
the system in et
et al.
al.
such (2006)).
(2006)).
a way that Moreover,
Moreover,
it can in in order
order
perform to
to
its from literature.
from organization
literature. of the paper is as follows: in section 2,
activity
restore the (Grigoriev
the system some et al.
in such
such (2006)). Moreover, in order to The
restore
intended
restore the system
function,
system some in
in such aa
componentway that
a way
that
way that needit can
it can
to be
it can
perform
perform
changed.
perform
its The organization of the paper is as follows: in section 2,
its
its The organization of
intended
intended function,
function, some component
component need
need to
to be
be changed.
changed.
we
The describe
organization
we describe
describe of the
the problem
the problem the paper
problem
treatedis
paper
treated
as
as follows:
is and
and
we present
follows:
we
in
in section
present
a mixed
section 2,
2,
This
intended
This
leadsfunction,
leads
companies
companies someto keep
to component
keep
an inventory
an need
inventory
of be
to
of
spare
spare
parts. we
changed.
parts. integer
we the
formulation.
describe the Next,treated
problem in section
treated and
and 3 we
we show theaa
we present
present a
mixed
mixed
results
mixed
This
It
This leads
is known companies
that inventory
leads companies to keep
to keep an
carryinginventory of
cost influences
an inventory spare
of spare parts. parts.
many integer integer formulation.
formulation. Next,
Next, in
in section 3 we show the results
ItIt isis known
decisions known that
in that inventory
inventory
the tactical, carrying
carrying
analytic, andcost
cost influences
influences
operations many
many
level of of
of the proposed
integer
of the formulation.
the proposed
proposed
model
Next,
model
and
andin section
we discuss
section
wewe
33 we
discuss
show
we its
show
its
the
the results
efficiency
efficiency
for
results
for
It is
decisionsknown in that
the inventory
tactical, carrying
analytic, andcost influences
operations many
level of solving
of the the problem.
proposed model
modelAt and
the
and we
end
we discuss
draw
discuss its
some
its efficiency
conclusions
efficiency for
for
decisions
any business
decisions in
in the
((Van
the tactical,
Horenbeek
tactical, analytic,
analytic, and
et al.,
and operations
2013)).
operations level
level of
of solving
solving the
the problem.
problem. At
At the
the end
end we
we draw
draw some
some conclusions
conclusions
any business
business ((Van((Van Horenbeek
Horenbeek et et al.,
al., 2013)).
2013)). and perspectives.
solving the problem. At the end we draw some conclusions
any
anyis business ((Van Horenbeekactivity et al., 2013)). and perspectives.
perspectives.
It clear that maintenance and inventory man- and and perspectives.
ItIt is
is clear that maintenance activity and inventory man-
is clear
agement
It
agement clearof that
sparemaintenance
ofthat
spare maintenance
parts parts’
activity
activity and
parts are interconnected.
are interconnected.
interconnected. and inventory
Thus, a better
inventory
Thus,
man-
man- 2. PROBLEM DESCRIPTION AND MIP
2. PROBLEM
PROBLEM DESCRIPTION AND AND MIP
agement
organization
agement of
of spare
of spare
spare parts
parts are
are inventory and
interconnected. Thus,
Thus, the aaamainte-
better
better
better 2.
2. PROBLEM DESCRIPTION
DESCRIPTION AND MIP
FORMULATION MIP
organization
organization of spare parts’ inventory and the mainte- FORMULATION
FORMULATION
nance activityof
organization of spare
allows
spare aparts’
significant
parts’ inventory
benefits
inventory and
and forthethe mainte-
companies
mainte- FORMULATION
nanceHorenbeek
nance
Van activity allows
activity allows
et al.aa significant
significant benefits
Thebenefits
problemfor foris companies
companies
known in 2.1 Problem description
nance
Van activity allows
Horenbeek et al. a(2013).
significant
(2013). The benefits
problem foris companies
known in 2.1 2.1 Problem description
description
Van
the
Van
the
Horenbeek
literature
Horenbeek
literature as
et
asetthe al.
al.
the
(2013).
joint
(2013).
joint
The
The problem
optimization
problem
optimization of
is
of isspare known
known
spare
parts
partsin 2.1 Problem
in Problem description
the
inventory
the literature
literature as
and maintenance the joint
as the joint activity. optimization
optimization of
In general, spare
of spare parts
joint We consider the Periodic Maintenance Problem defined
the parts
inventory
inventory
optimization andis maintenance
and maintenance
more interested activity.
activity.
to study In than
In general,
general,to treatthe joint
the joint
each We We Grigoriev
by consider the
consider the Periodic
Periodic
et al. Maintenance
(2006)Maintenance
in which weProblem Problem defined
add the defined
inven-
inventory
optimization and is maintenance
more interested activity.
to studyIn general,
than to the
treat joint
each We
by consider the
Grigoriev et Periodic
al. (2006) Maintenance
in which we Problem
add the defined
inven-
optimization
problem
optimization is
separately. more
is moreSeveral interested
Several
interested to
problems study
to study have than
been
than to treat
studied
to studied each
treat each by by
tory
by Grigoriev
management
Grigoriev et al.
of (2006)
the
et al.of(2006) in
spare which
parts
in which we
needed
we add the
during
add during inven-
the inven-the
problem
problem
the separately.
separately.
OR community Several
and confirmproblems
problems have
have
thathave been
been
improvements studied by
by
can tory
tory management
management
maintenance of
activity. the
the
It spare
spare
consists parts
parts
of needed
needed
determining during
an optimal the
the
problem
the OR separately.
community Several
and problems
confirm that been
improvements studied by
can tory management
maintenance of
activity. the
It spare
consists parts
of needed
determining during
an optimalthe
the
occur
the OR
OR community
oncommunity
the total cost and
and confirm
when
confirm that
we that improvements
consider the integrated
improvements can
can maintenance
scheduling
maintenance activity.
plan It consists
for servicing
activity. It of
a setofof
consists determining
machines over
determining an
an optimal
a plan-
optimal
occur on
occur
problem on rather
the total
the total
than cost
cost when
thewhen we consider
we
sequential consider
one. We thecan
the integrated scheduling
cite for scheduling
integrated ning horizon plan
plan forminimizing
for
while servicing aa the
servicing set of
set of machines
machines
total over aa plan-
over
costs associated plan-
to
occur
problem on rather
the total than costthe when we consider
sequential one. Wethecan integrated
cite for scheduling
ning horizon plan forminimizing
while servicing a the set of machines
total costs over a plan-
associated to
problem rather than the sequential one. We can cite for ning horizon
problem rather than the sequential one. We can cite for ning horizon while minimizing the total costs associated to while minimizing the total costs associated to
Copyright 2016 IFAC
2405-8963 © 2016, 881 Hosting by Elsevier Ltd. All rights reserved.
IFAC (International Federation of Automatic Control)
Copyright
Peer review©
Copyright © 2016
2016 IFAC
IFAC 881
881
Copyright ©under
2016 responsibility
IFAC of International Federation of Automatic
881Control.
10.1016/j.ifacol.2016.07.886
IFAC MIM 2016
882
June 28-30, 2016. Troyes, France Anis Mjirda et al. / IFAC-PapersOnLine 49-12 (2016) 881–885

  
the problem. Let M = {1, . . . , m} be the set of machines + hi Ii,t + zi,t × (di + ci qi,t ) (1)
and τ = {1, . . . , T } is the planning horizon. At each time i∈M t∈τ i∈M t∈τ ∪{0}
period t ∈ τ , at most one machine i ∈ M has to be   
serviced. During the planning horizon, all machines must (B)
be serviced at least once. We assume that if the machine i s.t xi,t+1 ≥ xi,t + 1 − T yi,t+1 , ∀i ∈ M, ∀t ∈ τ − T (2)
is serviced, it needs the replacement of one component (i.e.
one spare parts). Let P = {1, . . . , m} be the set of spare xi,1 ≥ xi,T + 1 − T yi,1 , ∀i ∈ M (3)

parts stored in the depot. We suppose for instance that the yi,t = 1, ∀t ∈ τ (4)
depot have an unlimited capacity to hold the spare parts i∈M
and for each machine i just one spare part (i.e. component)
Ii,t+1 = Ii,t + qi,t − yi,t+1 , ∀i ∈ M, ∀t ∈ τ − T (5)
i is to be replaced. Spare parts must be available at the
depot. In case of an unavailability, spare parts must be Ii,1 = Ii,0 + qi,0 , ∀i ∈ M (6)
ordered. The lead time is equal to one period.  
yi,t = qi,t , ∀i ∈ M (7)
We consider the following costs: t∈τ t∈τ ∪{0}

• Oi = ji (t) × ai : the operation costs where ai is a yi,t ∈ {0, 1}, zi,t ∈ {0, 1}, ∀i ∈ M, ∀t ∈ τ (8)
positive integer and ji (t) is the number of periods Ii,t ≥ 0, qi,t ≥ 0, xi,t ≥ 0, integer ∀i ∈ M, ∀t ∈ τ (9)
elapsed since last servicing machine i.
• bi : servicing cost for the machine i. The objective function (1) minimizes the total costs as-
• hi : holding cost per unit per period. sociated of the problem. The part (A) concern the sum
• di : fixed cost associated for ordering the spare parts of operating costs and servicing costs and the part (B)
i (can be seen as the shipment cost). concern the holding cost and the total costs associated for
• ci : variable cost per unit of the spare parts i. ordering. Constraints (2- 3) ensure the required behavior
At each time period, we need to answer the following of the variable xi,t . Just one machine is to be serviced at a
questions: single period is given by the constraint (4). Constraints (5)
and 16 concerns the inventory balance while constraints
• Which machine is to be serviced? (7) ensure that the total quantity ordered at the end of
• How much spare parts to order? the scheduling cycle is equal to the number of machine
• how much to hold? serviced. Equations (8) and (9) are the integrality and non-
negativity constraints.

2.3 Model linearization


2.2 Mathematical formulation: A quadratic formulation
To linearize the model, we define
The maintenance schedule will be executed in a cyclic way wi,t = zi,t .qi,t
which means that the same machine serviced at period t we add the following constraints to the model:
will be serviced at period k × T + t. We consider that the
inventory level at period 0 Ii,0 = 0 and Ii,T = 0. Keeping wi,t ≤ qi,t , ∀i ∈ M, ∀t ∈ τ (10)
in mind the cyclic nature of the problem(k×T +t), the date
(0 and T ) refers always to the last period of scheduling in wi,t ≥ 0, ∀i ∈ M, ∀t ∈ τ (11)
one cycle. We suppose as well that the shipment time is wi,t ≥ qi,t − T (1 − zi,t ), ∀i ∈ M, ∀t ∈ τ (12)
equal to one period.
wi,t ≤ T zi,t , ∀i ∈ M, ∀t ∈ τ (13)
We introduce then the following variables:
• xi,t number of period elapsed since the last servicing 2.4 Illustrative Example
of machine i at period t.
• yi,t a binary variable which equals to 1 if the machine In order to show the interest of our study, We consider
i is serviced at period t and 0 in other cases. a small instance composed of 10 time periods and 3
• zi,t a binary variable which equals to 1 if the spare machines. For each machine we consider the following
parts i is ordered at period t and 0 in other cases. costs: a1 = 50, a2 = a3 = 1, bi = 0, ∀i ∈ {1, 2, 3}, h1 =
• qi,t an integer variable which determines the quantity h2 = 3, h3 = 5, d1 = 9, d2 = 10, d3 = 8, ci = 5, ∀i ∈
to be ordered from spare part i at each period t. {1, 2, 3}. In this example, we will optimize separately
• Ii,t an integer variable which represents the inventory in a sequential way the maintenance plan and then the
of spare parts i holded at the depot by the end of inventory management of the spare parts. To optimize
period t. the maintenance plan we will use the model provided by
• wi,t = zi,t × qi,t (Grigoriev et al., 2006) and for the inventory management,
we will use the model below:
The model states as follows :
  
min hi Ii,t + zi,t × (di + ci qi,t ) (14)

i∈M t∈τ i∈M t∈τ ∪{0}
min (ai xi,t + bi yi,t )
i∈M t∈τ Ii,t+1 = Ii,t + qi,t − yi,t+1 , ∀i ∈ M, ∀t ∈ τ − T (15)
  
(A) Ii,1 = Ii,0 + qi,0 , ∀i ∈ M (16)

882
IFAC MIM 2016
June 28-30, 2016. Troyes, France Anis Mjirda et al. / IFAC-PapersOnLine 49-12 (2016) 881–885 883

yi,t ∈ {0, 1}, zi,t ∈ {0, 1}, ∀i ∈ M, ∀t ∈ τ (17) ii ai + bi < hi + di + ci .


iii ai + bi <<< hi + di + ci .
Ii,t ≥ 0, qi,t ≥ 0, ∀i ∈ M, ∀t ∈ τ (18)
Data can be found on the demand to authors.
Figure 1 represent the solution of the sequential optimiza-
tion models. Each row represents respectively the machine We report in next tables results for the different instances.
1, 2 and 3 and columns represent the periods from 1 to 10. On each table, we show results for the three different
The blue color filled at each cell indicates which machine configuration mentioned before (i, ii, iii). Each table have
is serviced on a given period. We note that a machine can the following form:
be maintained at more than one period as no constraints The first column of the table denote the instance’s name,
on the original model limit the number of times that a the second column indicates the number of machines m as
machine is maintained. The total cost of the solution is well as the number of period τ . Next, we report the cost
f = 372. of the solution, the computational time in seconds and the
last column indicate if the solution found is optimal or not
Maintenance cost = 190 (which means a Lower Bound for the problem). At the end
Inventory cost= 182 of each configuration, we show the average values of costs
Total = 372 and Time.

t1   t2   t3   t4   t5   t6   t7   t8   t9   t10   Table 1. Results of instances with 3 Machines


M1   for the three given configuration
M2   Instances (M , τ ) Cost Time Status
Inst 3 3 1 i (3, 3) 10,00 0,28 Optimal
M3  
Inst 3 3 2 i (3, 3) 13,00 0,42 Optimal
Inst 3 3 3 i (3, 3) 18,00 0,44 Optimal
Fig. 1. Optimal solution when sequential inventory man- Inst 3 5 i (3, 5) 57,00 0,72 Optimal
agement policy is used Inst 3 10 1 i (3, 10) 306,00 1,37 Optimal
Inst 3 10 2 i (3, 10) 423,00 1,82 Optimal
In the second example, we compute solution using the Inst 3 10 3 i (3, 10) 505,00 1,15 Optimal
model 1-13. The optimal solution is given in figure 2 and Inst 3 24 i (3, 24) 740,00 21,77 Optimal
its objective value is f = 367. Inst 3 50 i (3, 50) 168,00 3,06 Optimal
Inst 3 70 i (3, 70) 235,00 4,94 Optimal
We can easily remark that solving the integrated problem Inst 3 100 i (3, 100) 335,00 7,63 Optimal
can lead to a gain in the total cost. In fact, taking into Average - 255,45 3,96 -
account costs related to shipment and inventory influence Inst 3 3 1 ii (3, 3) 15 0,35 Optimal
the decisions on the optimal solution when solving the joint Inst 3 3 2 ii (3, 3) 19 0,49 Optimal
problem. Inst 3 3 3 ii (3, 3) 23 0,42 Optimal
Inst 3 5 ii (3, 5) 60 0,66 Optimal
Joint optimization cost= 367 Inst 3 10 1 ii (3, 10) 287 0,92 Optimal
Inst 3 10 2 ii (3, 10) 445 1,01 Optimal
Inst 3 10 3 ii (3, 10) 515 1,06 Optimal
t1   t2   t3   t4   t5   t6   t7   t8   t9   t10   Inst 3 24 ii (3, 24) 697 9,60 Optimal
M1   Inst 3 50 ii (3, 50) 252 6,09 Optimal
Inst 3 70 ii (3, 70) 352 11,66 Optimal
M2   Inst 3 100 ii (3, 100) 502 21,76 Optimal
M3   Average - 287,91 4,91 -
Inst 3 3 1 iii (3, 3) 43,00 0,33 Optimal
Fig. 2. Optimal solution for the joint optimization model Inst 3 3 2 iii (3, 3) 65,00 0,464 Optimal
Inst 3 3 3 iii (3, 3) 82,00 0,607 Optimal
Inst 3 5 iii (3, 5) 150,00 0,643 Optimal
3. COMPUTATIONAL RESULTS Inst 3 10 1 iii (3, 10) 367,00 1,354 Optimal
Inst 3 10 2 iii (3, 10) 545,00 0,772 Optimal
The proposed model were solved using CPLEX 12.6 and Inst 3 10 3 iii (3, 10) 515,00 0,866 Optimal
computational tests were run on a CORE i5 with 2.6 Inst 3 24 iii (3, 24) 888,00 10,02 Optimal
Inst 3 50 iii (3, 50) 553,44 3626,847 Feasible
GHz processor and 4GB RAM. In this section, we present Inst 3 70 iii (3, 70) 751,49 3605,177 Feasible
detailed experimental results for the proposed model. Inst 3 100 iii (3, 100) 1066,47 3604,321 Feasible
Computational tests were performed on a set of data that Average - 456,95 986,49 -
were generated based on 19 original instances proposed
by Grigoriev et al. (2006). Each instance is characterized
by the number of machines and the number of periods The first set of computational tests concern instances with
as well as the different costs of the problem. For each 3 Machines. We report these results in Table 1. From Table
instance from the 19, we generated the holding cost hi , 1 we conclude the following remarks:
the fixed cost di and the variable cost ci . To construct
• For the first and the second configuration (i and
these instances, three configurations were settled:
ii), we were able to solve all instances to optimal in
i ai + bi = hi + di + ci . computational time less than 5s in average.

883
IFAC MIM 2016
884
June 28-30, 2016. Troyes, France Anis Mjirda et al. / IFAC-PapersOnLine 49-12 (2016) 881–885

• Tor the third configuration of instances (iii), we were for all instances. Regarding computational time, it tooks
able to solve 7 out of 10 to optimal in a computational less than 3s on Average for the first configurations (i) and
time less than 2s. this time increase for the last 2 configurations.
• The last 3 instances in table 1, we report just the
lower bound. Cplex was unable to find the optimal Table 4. Results of instances with 10 Machines
solution within a time limit equal to one hour (3600s). for the three given configuration
• When the number of period increase and the costs Instances (M , τ ) Cost Time Status
that concern inventory (hi , di and ci ) are larger than Inst 10 18 1 i (10, 18) 817,96 3656,48 Feasible
the maintenace costs, the model is limited and cannot Inst 10 18 2 i (10, 18) 11675,83 2283,85 Optimal
provide optimial solution in a computational time less Inst 10 24 i (10, 24) 13091,59 3610,39 Feasible
than one hour. Average - 8528,46 3183,57 -

Table 2. Results of instances with 4 Machines Inst 10 18 i (10, 18) 874,52 3615,97 Feasible
Inst 10 18 i (10, 18) 11628,36 3604,23 Feasible
for the three given configuration Inst 10 24 i (10, 24) 13014,46 3797,05 Feasible
Instances (M , τ ) Cost Time Status Average - 6251,44 3610,10 -
Inst 4 4 i (4, 4) 27,00 0,50 Optimal
Inst 4 10 i (4, 10) 97,00 2,06 Optimal In the last table (table 4), we provide results for instances
Inst 4 15 i (4, 15) 172,00 3,03 Optimal with 10 machines. From this table we conclude the follow-
Inst 4 22 i (4, 22) 337,00 6,25 Optimal ing remarks:
Inst 4 24 i (4, 24) 1604 19,24 Optimal
Inst 4 30 i (4, 30) 1995,00 47,05 Optimal • Cplex was able to provide optimal solution for one
Inst 4 50 i (4, 50) 3303,96 3601,78 Feasible instance out of four in a large computational time (
Inst 4 70 i (4, 70) 4576,79 3603,90 Feasible 3000s)
Inst 4 100 i (4, 100) 6490,10 3605,90 Feasible • A lower bound is given for the other 3 instances.
Average - 2066,98 1209,97 -
Based on the computational experiments, we remark that:
Inst 4 4 ii (4, 4) 36 0,60 Optimal
Inst 4 10 ii (4, 10) 121 2,54 Optimal • As long as the cycle is large (number of periods)
Inst 4 15 ii (4, 15) 222 3,94 Optimal instances will not be solved to optimality using our
Inst 4 22 ii (4, 22) 399 8,15 Optimal model.
Inst 4 24 ii (4, 24) 1688 319,95 Optimal
Inst 4 30 ii (3, 10) 2102 29,98 Optimal
• The largest instances to solve is within 10 machines
Inst 4 50 ii (4, 50) 3460,30 3608,06 Feasible and 18 periods.
Inst 4 70 ii (4, 70) 4853,14 3610,59 Feasible
Inst 4 100 ii (4, 100) 6843,81 3603,66 Feasible 4. CONCLUSIONS
Average - 2191,69 1243,05 -
Inst 4 4 iii (4, 4) 84 0,63 Optimal We present in this paper, a new model for solving the
Inst 4 10 iii (4, 10) 277 5,74 Optimal joint optimization of maintenance activity and inventory
Inst 4 15 iii (4, 15) 454 10,08 Optimal of spare parts. The model was implemented and tested on
Inst 4 22 iii (4, 22) 677 15,85 Optimal a set of benchmarks generated. We showed as well the limit
Inst 4 24 iii (4, 24) 1772 348,94 Optimal
Inst 4 30 iii (3, 10) 2200 18,49 Optimal
of our model through the computational experiments. The
Inst 4 50 iii (4, 50) 3635,16 3601,00 Feasible perspectives of this work is to propose a hybrid approach
Inst 4 70 iii (4, 70) 5079,81 3602,57 Feasible that can be able to guarant good quality solution for large
Inst 4 100 iii (4, 100) 7236,48 3604,9 Feasible instances.
Average - 910,67 66,62 -
ACKNOWLEDGEMENTS
In Table 2, we report results for instances with 4 machines.
We can remark from Table 2 that all instances were solved This work was supported by the Technological Research
to optimal in an average computational time equal to 10s. Institute Railenium, by the Laboratoire de Génie Informa-
Table 3. Results of instances with 5 Machines tique et d’Automatique de l’Artois (LGI2A) of the Uni-
for the three given configuration versité de l’Artois and by the Laboratoire d’Automatique,
de Mcanique et d’Informatique industrielles et Humaines
Instances (M , τ ) Cost Time Status (LAMIH) of the Universit de Valenciennes et du Hainaut
Inst 5 24 i (5, 24) 84 0,63 Optimal Cambrésis.
Inst 5 24 i (5, 24) 277 5,74 Optimal
Average - 180,50 3,19 - REFERENCES
Inst 5 24 1 ii (5, 24) 436 10,99 Optimal
Inst 5 24 2 ii (5, 24) 5809 141,01 Optimal Grigoriev, A., van de kludert, J., and Spieksma, F.C.
Average - 3122,50 76,00 - (2006). Modeling ad solving the periodic maintenance
Inst 5 24 1 iii (5, 24) 672 13,37 Optimal problem. European Journal of Operational Research,
Inst 5 24 2 iii (5, 24) 6069 140,25 Optimal 172(3), 783–797.
Average - 3370,50 76,81 - Mjirda, A., Jarboui, B., Macedo, R., Hanafi, S., and
Mladenović, N. (2014). A two phase variable neigh-
We present in Table 3, the results of instances with 5 borhood search for the multi-product inventory routing
Machines. We used for this test 6 instances, each one with problem. Computers & Operations Research, 52, 291–
24 periods and CPLEX was able to find optimal solution 299.

884
IFAC MIM 2016
June 28-30, 2016. Troyes, France Anis Mjirda et al. / IFAC-PapersOnLine 49-12 (2016) 881–885 885

Mobley, R.K. (2002). An introduction to predictive main-


tenance. Butterworth-Heinemann.
Van Horenbeek, A., Buré, J., Cattrysse, D., Pintelon, L.,
and Vansteenwegen, P. (2013). Joint maintenance and
inventory optimization systems: A review. International
Journal of Production Economics, 143(2), 499–508.

885

Potrebbero piacerti anche