ARTICLE IN PRESS

Reliability Engineering and System Safety 94 (2009) 644657

Contents lists available at ScienceDirect

Reliability Engineering and System Safety

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

Beyond its cost, the value of maintenance: An analytical framework for

capturing its net present value$
Karen B. Marais a,, Joseph H. Saleh b
a
b

School of Aeronautics and Astronautics, Purdue University, USA

School of Aerospace Engineering, Georgia Institute of Technology, USA

a r t i c l e in f o

a b s t r a c t

Article history:
Received 27 February 2008
Received in revised form
1 July 2008
Accepted 2 July 2008
Available online 15 July 2008

Maintenance planning and activities have grown dramatically in importance across many industries
and are increasingly recognized as drivers of competitiveness if managed appropriately. Correlated with
this observation is the proliferation of maintenance optimization techniques in the technical literature.
But while all these models deal with the cost of maintenance (as an objective function or a constraint),
only a handful addresses the notion of value of maintenance, and seldom in an analytical or quantitative
way.
In this paper, we propose that maintenance has intrinsic value and argue that existing cost-centric
models ignore an important dimension of maintenance, namely its value, and in so doing, they can lead
to sub-optimal maintenance strategies. We develop a framework for capturing and quantifying the
value of maintenance activities. Our framework is based on four key components. First, we consider
systems that deteriorate stochastically and exhibit multi-state failures, and model their state evolution
using Markov chains and directed graphs. Second, we consider that the system provides a ow of service
per unit time. This ow in turn is priced and a discounted cash ow is calculated resulting in a present
value (PV) for each branch of the graphor value trajectory of the system. Third as the system ages or
deteriorates, it migrates towards lower PV branches of the graph, or lower value trajectories. Fourth, we
conceptualize maintenance as an operator (in a mathematical sense) that raises the system to a higher
PV branch in the graph. We refer to the value of maintenance as the incremental PV between the preand post-maintenance branches of the graphs minus the cost of maintenance. The framework presented
here offers rich possibilities for future work in benchmarking existing maintenance strategies based on
their value implications, and in deriving new maintenance strategies that are value-optimized.
& 2008 Elsevier Ltd. All rights reserved.

Keywords:
Maintenance
Present value
Directed graph
Value trajectory
Multi-state failure

1. Introduction
Maintenance planning and activities have grown dramatically
in importance across many industries. This importance is
manifested by both the signicant material resources allocated
to maintenance departments as well as by the substantial number
of personnel involved in maintenance activities in companiesfor
example over a quarter of the total workforce in the process
industry is said to deal with maintenance work [1]. This situation,
coupled with an increasingly competitive environment, creates
economic pressures and a heightened need to ensure that these
considerable maintenance resources are allocated and used
appropriately, as they can be signicant drivers of competitivenessor lack thereof if mismanaged.
$
This research builds on and extends previous work prepared for the ESREL
2008 conference.
 Corresponding author. Tel.: +1765 494 5117; fax: +1765 494 0307.
E-mail address: karen.marais@alum.mit.edu (K.B. Marais).

0951-8320/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ress.2008.07.004

In response to these pressures, the notion of optimality and

the mathematical tools of optimization and operations research
(OR) have seeped into maintenance planning, and resulted in the
proliferation of optimal maintenance models (see the reviews
by Pham and Wang [2] and Wang [3], for example). In each
optimal maintenance model developed, an objective function is
rst posited, then analytical tools are used to derive a maintenance policy that maximizes or minimizes this objective
function subject to some constraints. For example, the objective
function can be the minimization of cost (cost rate, or life cycle
cost) of maintenance given a system reliability and/or availability
constraint; conversely, the objective function can be the maximization of reliability or availability, given a cost constraint. In
addition to varying the objective function, different optimal
maintenance models are obtained by: (1) varying for example the
system conguration (e.g., single-unit systems versus multi-unit
systems, k-out-of-n systems); (2) by including several degrees of
maintenance (e.g., minimal, imperfect, perfect); (3) by varying the
planning horizon; (4) by using different analytical tools; or (5) by

ARTICLE IN PRESS
K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

positing different types of dependencies between the various

units in a multi-unit system.
Yet, while all these models deal with the cost of maintenance
(as an objective function or a constraint), only a handful of models
touches on the notion of value of maintenance, and seldom in an
analytical or quantitative way (e.g., [24]). Wang [3] highlights a
critical idea for the development of a value-based perspective on
maintenance when he suggests that the cost of maintenance as
well as the resulting system reliability should be considered
together when developing optimal maintenance strategies. Unfortunately, where the benets of maintenance are considered, it
is usually in the sense of avoiding the costs of failure. Interestingly, it is only within the civil engineering community that the
benets in the sense of service delivery are considered and costbenet considerations explicitly taken into account in the
development of maintenance strategies (e.g., [24]).
The argument for dismissing or not focusing on the value of
maintenance, when it is made, goes along these lines: while it is
easy to quantify the (direct) cost of maintenance, it is difcult to
quantify its benets. Other authors wishing to consider the value
of maintenance lament the difculties in quantifying the benets
of maintenance. Dekker [4] for example notes the main question
faced by maintenance management, whether maintenance output
is produced effectively, in terms of contribution to company
prots, [y] is very difcult to answer. Therefore maintenance
planning is usually shifted from a value maximization problem
formulation to a cost minimization problem (see [5,6] for a
discussion of why these two problems are not the same and do
not lead to similar decisions in system design and operation).
Incidentally, in many organizations, maintenance is seen as a cost
function, and maintenance departments are considered cost
centers whose resources are to be optimized or minimized. In
short, as noted by Rosqvist et al. [7] cost-centric mindset prevails
in the maintenance literature for which maintenance has no
intrinsic value.
In this paper, we argue that maintenance has intrinsic value
and argue that existing cost-centric optimizations ignore an
important dimension of maintenance, namely its value, and in
so doing, they can lead to sub-optimal maintenance strategies. We
therefore develop a framework for capturing and quantifying one
important aspect of the value of maintenance activities, their
impact on revenue-generation capability, by connecting an
engineering and OR concept, system state, with a nancial and
managerial concept, the net present value (NPV).1 Here the system
state refers to the condition of the system and hence its ability to
perform and thereby provide a ow of service (hence generate
revenue, or quasi-rent). In order to build this connection, we
rst explore the impact of a systems state on the ow of service
the system can provide over timefor a commercial system, this
translates into the systems revenue-generating capability. Next
we consider the impact of maintenance on system state evolution
and hence value generation capability over time. We then use
traditional discounted cash ow techniques to capture the impact
of system state evolution with and without maintenance on its
nancial worth, or NPV. For simplication, we call the results of
our calculations the value of maintenance. Finally, we discuss the
advantages and limitations of our framework. This work offers
rich possibilities for assessing and benchmarking the value
implications of existing maintenance policies, and deriving new

1
Note that we consider value as the net revenue generated by the system
over a given planning horizon. We do not consider additional dimensions of value
such as the potential positive effects of maintenance on environmental or health
impacts. Such effects can be incorporated in future work, see, for example, Marais
et al. [25] for a discussion of the quantication of environmental and health
impacts of aviation.

645

policies based on maximizing value, instead of minimizing cost of

maintenance.

2. Background
This section provides a brief overview of various maintenance
models. The purpose of this section is to provide context and
background to the model assumptions and analytics we develop
in Sections 3 and 4. The reader interested in extensive reviews of
the subject is referred to the survey papers by Dekker [4], Pham
and Wang [2] and Wang [3]. In the following, we discuss
(1) maintenance classication, (2) maintenance models, and
(3) maintenance policies.
2.1. Types and degrees of maintenance
Maintenance refers to the set of all technical and administrative actions intended to maintain a system in or restore it to a
state in which it can perform at least part of its intended
function(s) [4]. Fig. 1 provides a simple (not comprehensive)
classication scheme of maintenance along three axes: (1) the
type of maintenance; (2) the degree of maintenance; and (3) type
of system to be maintained (system conguration can be
conceived of as a subset of this axis).
Maintenance type can be classied into two main categories:
corrective maintenance and preventive maintenance (PM) [2]. CM,
also referred to as repair or run-to-failure (RTF), refers to
maintenance activities performed after a system has failed in
order to restore its functionality.
PM refers to planned maintenance activities performed while
the system is still operational. Its aim is to retain the system in
some desired operational condition by preventing (or delaying)
failures. PM is further sub-divided into clock-based, age-based,
and condition-based. These sub-divisions refer to what triggers
maintenance activities [8].

 Clock-based maintenance is scheduled at specic calendar




times; its periodicity is preset irrespective of the systems

condition (e.g., every Tuesday).
Age-based maintenance is performed at operating time
intervals or operating cycles of the system (e.g., every 500
on/off cycles, or every 4000 h of ight).
Condition-based maintenance is triggered when the measurement of a condition or state of the system reaches a threshold
that reects some degradation and loss of performance of a
system (but not yet a failure). Condition-based maintenance is
also referred to as predictive maintenance.

Opportunistic maintenance encompasses both corrective and

PM and is relevant for multi-unit systems with economic and
functional dependencies in which the failure of one unit, and
hence its corrective maintenance, offers an opportunity to
perform PM on other still functional units.
Each type of maintenance can be further classied according to
the degree to which it restores the system [2]. At one end of the
spectrum, perfect maintenance restores the system to its initial
operating condition or renders it as good as new. At the other
end of the spectrum, minimal repair returns the system to the
condition it was in immediately prior to failing (in the case of
corrective maintenance), or as bad as old. In between these
extremes lies imperfect maintenance, which in effect returns the
system to a condition somewhere in between as good as new and
as bad as old. Finally, there is also the possibility that maintenance
leaves the system in a worse condition than before the failure,

ARTICLE IN PRESS
646

K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

Degree of
maintenance

Perfect

Imperfect

Minimal

Single-unit
systems

Preventive

Corrective

Opportunistic
(relevant only
to multi-unit
systems)

Type of
maintenance

Multi-unit
systems

Type of system (to

be maintained)
Fig. 1. Classication of maintenance strategies.

through, for example, erroneous actions such as damaging

adjacent parts while replacing a faulty unit.
2.2. Maintenance models
Models used to derive optimal maintenance policies generally
cover four main aspects [4]: (1) a description of the system being
maintained; (2) a model of how the system deteriorates and the
consequences thereof; (3) a description of the available information on the system and the available response options; and (4) an
objective function and an analytical framework (or tools) according to which the optimal maintenance policy is to be derived. This
section reviews the four main classes of maintenance models,
following the reviews in Refs. [2,3,9,10].
The rst class of models developed considered the possibility
only for perfect or minimal repair [2,11,12]. Thus, following
maintenance, the system is returned to as good as new with
some repair probability p, or to as bad as old with probability
(1p). This basic concept is then expanded to take into account
time-dependent repair probabilities, the possibility that maintenance causes the system to be scrapped or to transition to some
intermediate state, and non-negligible repair times (and hence
non-negligible downtime losses).
The second class of models considers maintenance as improving the failure rate or intensity, and thus allows the possibility of
imperfect maintenance [2,13]. It is assumed that maintenance
provides a xed reduction in failure rate, or a proportional
reduction, or that it returns the system to the failure rate curve at
some time prior to the maintenance activity. Perfect maintenance
returns the failure rate to that of a new system, while minimal
maintenance returns it to that of the system immediately prior to
the failure. The degree of improvement of failure rate is referred to
as the improvement factor. The improvement factor is determined
based on historical data, experiment, expert judgment, or by
assuming it correlates with maintenance cost, system age, or the
number of prior maintenance activities [2,14].

The third class of models views maintenance as reducing the

virtual age of the system [15]. It is assumed that maintenance
reduces the age of the system by some proportion (assuming
increasing failure rate, which implies among other things that the
system exhibits no infant mortality). Perfect maintenance returns
the system virtual age to zero, while minimal maintenance
returns the virtual age to the age immediately prior to the failure.
Kijima et al.s [15] original model allowed only a reduction to the
virtual age of the system following the previous repair effort,
though larger reductions in virtual age can be seen as resulting
from more extensive maintenance efforts. Pham and Wang [2]
consider the reduction in virtual age as decreasing over timein
other words, repairs become successively less effective over time.
They further assume that maintenance time increases with
subsequent repairs.
The fourth class of models considers system failures as
manifesting as some level of damage or degradation in response
to a shock. These models are therefore referred to as shock
models. Perfect maintenance then reduces the damage to zero,
minimal maintenance returns the damage level to that immediately prior to the failure, and imperfect damage reduces the
damage by some factor greater than 0 and less than 100%. These
models also allow the possibility for less-effective and more
expensive repairs over time by making the reduction in damage a
decreasing function of time and by successively increasing the
duration of maintenance activities over time [16,17].
In each case these models have been used to derive
maintenance policies that minimize cost or downtime, or
that maximize system availability, as we discuss in the next
subsection.
2.3. Maintenance policies
Maintenance policies describe what types of maintenance
(repair, replacement, etc.) are considered in response to
what types of events (failure, calendar time, machine cycles,

ARTICLE IN PRESS
K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

etc.). In the following, we conne our discussion to maintenance

policies for single-unit systems with increasing failure rates
(IFRs).
One popular maintenance policy is age-dependent PM where a
system is repaired or replaced at a pre-determined age [3]. The
triggering of maintenance in this case may be pre-determined
based on machine time (e.g., every 10,000 cycles) or on time
elapsed since the last maintenance activity. Under a random agedependent maintenance policy, maintenance is performed based
on age and system availability. This policy takes account of the
fact that systems may not be available for maintenance in the
middle of a production run, for example. A further extension of
age-dependent replacement is failure-dependent replacement
where the system is repaired in response to failures and replaced
when a given number of failures has occurred, or at a given time,
whichever occurs rst [18]. Many other variations on the theme of
age-dependent maintenance have been proposed; see [3] for an
extensive review.
An alternative family of maintenance policies, referred to as
periodic PM, is based on calendar time. Here maintenance occurs
on failure and periodically regardless of the failure or operating
history of the system [3]. Variations on this theme are developed
by selecting degrees of repair from minimal to perfect at specic
times or in response to failures. Further variations are developed
by incorporating the failure or operating history of the system. For
example, the level of maintenance may be dependent on the
number of previous repairs [19].
Sequential PM can be seen as a variation of periodic PM where
the interval between PM activities is not constant. For example,
the PM interval may be decreased as the system ages, so that the
system does not exceed a certain operating time without
maintenance [3,20].
This brief overview of maintenance, as mentioned previously,
is intended to provide context and background to our proposed
perspective on maintenance as well as to the model assumptions
and analytics we develop in the following sections.

3. The value perspective in design, operations and maintenance:

a qualitative discussion
The present work builds on the premise that engineering
systems are value-delivery artifacts that provide a ow of services
(or products) to stakeholders. When this ow of services is
priced in a market, this pricing or rent of these systems
services allows the assessment of the systems value, as will be
discussed shortly.2 In other words, the value of an engineering
system is determined by the market assessment of the ow of
services the system provides over its lifetime. We have developed
this perspective in a number of previous publications; for further
details, the interested reader is referred to [5,6,22,23].
In this paper, we extend our value-centric perspective on
design to the case of maintenance. Our argument is based on four
key components:
(i) First, we consider systems that deteriorate stochastically and
exhibit multi-state failures, and we model their state
evolution using Markov chains and directed graphs.
(ii) Second, we consider that the system provides a ow of
service per unit time. This ow in turn is priced and a
2
There are other cases when the ow of services provided by the system is not
priced in a market. This situation is treated in the economic literature under the
heading of un-priced values (see for example [21]). While the value of the
system can still be assessed in this case, its treatment raises additional subtleties
that are beyond the scope of the present work.

647

discounted cash ow is calculated resulting in a PV for each

branch of the graphor value trajectory of the system.
(iii) Third, given our previous two points, it is straightforward to
conceive of the following: as the system ages or deteriorates,
it migrates towards lower PV branches of the graph, or lower
value trajectories.
(iv) Finally, we conceptualize maintenance as an operator (in a
mathematical sense) that raises the system to a higher PV
branch in the graph, or to higher-value trajectory. We refer to
the value of maintenance, or more specically the NPV of
maintenance, as the incremental PV between the pre- and
post-maintenance branches of the graphs minus the cost of
maintenance.
In the following section, we set up the analytical framework
that corresponds to this qualitative discussion.

4. Maintenance and present value branches

In developing our value model of maintenance, we make a
number of simplifying assumptions to keep the focus on the main
argument of this work. These assumptions affect the particular
mechanics of our calculations but bear no impact on the main
results, as will be shown shortly. Our assumptions are the
following:
(i) We consider only the impact of maintenance on revenuegenerating capability.
(ii) We restrict ourselves to the case of perfect maintenance; in
addition we assume that maintenance does not change the
systems deterioration mechanism.
(iii) We restrict ourselves to the case of single-unit systems.
(iv) We only consider systems that exhibit an increasing failure
rate. In other words, as our systems age, they become more
likely to deteriorate in the absence of perfect maintenance.
(v) The systems in our model can be in a nite number of
discrete states, and the current state depends only on the
prior state, though the state transition probabilities may be
time-dependent. This assumption allows us to model the
state evolution of the system as a Markov process.
(vi) The systems in our model have no salvage value at
replacement or end of life.
(vii) Finally, for simulation purposes, we consider discrete-time
steps, and assume that the duration of maintenance
activities is negligible compared with the size of these time
steps.
These assumptions will be relaxed in future work. In the
following, we consider rst how a system deteriorates under nomaintenance and introduce the concept of value trajectories of
the system. Next, we show how maintenance moves the system
onto higher PV trajectories. Finally, we develop expressions for the
NPV of maintenance. We illustrate the framework by means of
numerical examples.
4.1. Deterioration under no-maintenance, and value trajectories
We consider a k-state discrete-time Markov deteriorating
system3 with time-dependent transition probabilities as shown in
Fig. 2, for the no-maintenance case with three states. The states are
numbered from 1 through k in ascending order of deterioration
3
Markov models are frequently used to analyse reliability and maintenance
problems (see, for example [10]).

ARTICLE IN PRESS
648

K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

pnn(i)

p11(2)

p11(1)

B1

p12(2)
p13(2)

New
pnd(i)

p12(1)

p22(2)
p23(2)

p13(1)

p11(0)

New
Deteriorated
Failed

Deteriorated

pnf(i)

pdd(i)

p22(1)

p12(0)

p22(2)
p23(2)

p23(1)
p13(0)

pdf(i)

Bworst

Failed

Fig. 3. Three-state system evolution over time with no maintenance.

1
Fig. 2. Three-state model of system with no maintenance.

where state 1 is the new state and state k is the failed state. The
time-dependence allows us to take account of the fact that a new
(or deteriorated) system will become more likely to transition to
the deteriorated (or failed) state as it ages.4 With no maintenance
the failed state is an absorbing state whence it is not possible to
transition to any of the other states. Further, it is not possible to
transition from the deteriorated state to the new state without
performing maintenance. In other words, the system can only
transition in one direction, from new to failed, perhaps via the
deteriorated state (but the system has no self-healing properties).
The transition matrix for a system with k states and no selfhealing is given by
2
6
6
Pi 6
6
4

p11 i

p12 i



p1k i

0
0

p22 i
..
.


..
.

p2k i 7
7
.. 7
7
. 5



example, mining companies compile detailed failure logs on large

equipment such as drills and heavy trucks.
We represent the evolution of the system over time using a
directed graph, as shown in Fig. 3 for a three-state system. This
representation expands on Fig. 2 and allows in effect an easy read
of the time-dependent transition probabilities, which is difcult
to visualize using the traditional Markov chain representation
(Fig. 2).
We assume that the probability of transitioning to a lower
state increases over time, and correspondingly that the probability
of remaining in a given state (other than the failed state)
decreases over time:
p11 0Xp11 iXp11 i j
pmn 0ppmn ippmn i j

(1)

(2)

If we dene p0 to be the initial probability distribution of the

system, the probability distribution after j state transitions is

pj Pj . . . P2 P1 p0

(3)

For convenience we assume that the system is initially in the

new state, i.e.,

p0 1 0   
where i is the index of the time step considered, and P(i) is in
effect P(ti) in which ti i DT. For simplication purposes, we
retain only the index i in our notation.
Most work in estimating transition probabilities has been done
in civil engineering. Macke and Higuchi [24] demonstrate how
transition probabilities can be derived from failure rates, and also
provide references to several publications demonstrating the use
of fatigue data, condition rating, analytical models of fatigue
accumulation on the structural component level, and global
damage indices to estimate transition probabilities in the civil
engineering domain. Macke and Higuchis approach can also be
applied to systems other than civil engineering structures
assuming that appropriate failure histories are available. For

i; jX1
1pmonpk

Here the time-dependence implies dependence on the virtual age of the

system.

0

(4)

Next, we consider that the system can generate um(t) revenue

per unit time when it is in state m; a degraded system having
lower capacity to provide services (hence generate revenues) than
a fully functional system. This um(t) is the expected utility model
of the system or the price of the ow of service it can provide over
time. We discretize time into small DT bins over which um(t) can
be considered constant.5 Therefore
um i um i DT  um i 1 DT

(5)

To simplify the indexing, we consider that the revenues the

system can generate between (i1) DT and i DT are equal to
um(i) DT.
5
The time step can be set for example according to the organizations
accounting period. For instance, if accounts are paid, or revenue is accounted for,
on a quarterly basis, the time step can be set to 90 days.

ARTICLE IN PRESS
K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

Each branch of the graph represents a particular value

trajectory of the system, as discussed below.
Each time step, the system can remain fully operational, or it
can transition to a degraded or failed state. A branch in the graph
is characterized by the set of states the system can visit for all
time periods considered. For example, the branch B1 {1, 1, 1,
1,y,1} represents the system remaining in state 1 throughout the
N periods considered, whereas Bj {1, 1, 1, 2, 2,y,2} represents a
system starting in state 1 and remaining in this state for the rst
two periods, then transitioning to a degraded state (here state 2)
at the third period and remaining in this particular degraded state.
Notice that the branch Bworst {1, k, k, y, k} represents a new
system transitioning to the failed state at the rst transition.
Since each state has an associated utility um(i), a PV can be
calculated for each branch over N periods as follows:
PN
i1 uBj i
PVN; Bj
(6)
1 r DT i
where uB(i) is shorthand for the revenues along the branch B. rDT is
the discount rate adjusted for the time interval DT. Notice that
Branch 1 has the highest PV since the system remains fully
functional throughout the N periods (in state 1), and branch Bworst
has the lowest PV (PVworst 0) since the system starts its service
life by failing!
The likelihood of the system following a particular branch over
N steps is given by the products of the transition probabilities
along that branch:
pBj

N
Y

pi Bj

(7)

i1

The right-hand side of Eq. (7) is shorthand for the product of the
transition probabilities along the branch Bj. Here the assumption
that the probability of transition to the next state depends only on
the current state allows us simply to multiply the probabilities to
obtain the probability of the branch.
Finally, the expected PV of the system over all the branches is
calculated by weighting the PV of each branch by its likelihood:
X
hPVNi
pBj PVN; Bj
(8)
all branches

Using the Markov chain terminology, this PV can also be

expressed as
hPVNi

N X
k
X
um ipi m
i1 m1

1 r DT i

(9)

The following simple numerical example will help clarify this

discussion and equations.
Numerical example: Consider a three-state system with the
following transition matrix:
2
3
:95 :04 :01
6
7
:9
:1 5
P4 0
0
0
1
The system has a constant revenue model over N periods, and
can generate $100,000/year in state 1, $60,000/year in the
degraded State 2, and $0 in the failed state 3. Assume a yearly
discount rate of 10%. No maintenance is performed on the system.
Next, consider the branches of the system evolution dened by
the transitions in Table 1.
Applying Eq. (5) to this simple case, we obtain the value
trajectories and probabilities shown in Fig. 4.
The expected PV of this system across all branches is given by
Eq. (7):
hPVNi $296; 672

649

Table 1
Branches for a three-state system evolution over four periods
Branch

Transitions

Comment

B1

{1, 1, 1, 1, 1}

B4

{1, 1, 1, 2, 2}

B8

{1, 1, 2, 2, 3}

The system starts in state 1 and remains in this state

throughout the four periods
The system starts in state 1; it remains in state 1 for
two periods, then transitions to the degraded state 2
in the third period and remains in this state 2
The system starts in state 1; it remains in state 1 for
the rst period, then transitions to the degraded
state 2 in the second period; it remains in this
degraded state for the third period, then transitions
to the failed state 3 in the fourth period

In the next subsection, we show how maintenance moves the

system onto higher-value trajectories and therefore increases the
expected PV of the system.
4.2. Deterioration, maintenance, and present value
Maintenance compensates for the lack of self-healing in the
system, and, from a modeling perspective, given a three-state
system, maintenance allows the system to move back, (1) from a
failed state to a deteriorated state (imperfect corrective maintenance), (2) from a failed state to a new state (perfect corrective
maintenance), and (3) from a deteriorated state to a new state
(perfect PM). These maintenance-enabled transitions are shown
in Fig. 5.
In addition to returning the system to a higher-functional state,
maintenance provides another advantage: it modies the timedependent transition probabilities. In particular, the transition
probabilities from the new state after perfect maintenance are
equal to those of a new system. That is, performing perfect
maintenance returns the system to the initial transition probabilities pij(0). Seen from a reliability viewpoint, perfect maintenance returns the system to the initial reliability curve, as
shown in Fig. 6. In the remainder of this work, we focus only on
the case of perfect maintenance.
Qualitatively, we can already foresee the two-sided impact
of maintenance: not only does maintenance raise the system
to a higher-value trajectory, it also increases the likelihood
that the system will remain on higher-value trajectories than
without maintenance (through its impact on the transition
probabilities). These two effects are captured quantitatively in
the following sub-section, and they help us determine the PV of
maintenance.
Fig. 7 shows the Markov model for our simple system under
the option of perfect maintenance. The dotted lines indicate the
new transition probabilities assuming that perfect maintenance
occurred at the end of the previous time step.6
Between maintenance activities the system deteriorates
according to:

pj Pj    P2 P1 p0

(10)

where pj is the vector of probabilities of being in states 1

through k.
The effect of perfect maintenance is to return the system to the
initial probability distribution p0 and to the initial transition
probability matrix P. Thus we can model the transition assuming
6
Given that the duration of maintenance is much smaller than the duration of
the time step considered, we can assume that maintenance occurs quasiinstantaneously during the transition from the end of one time step to the
beginning of the next time step.

ARTICLE IN PRESS
650

K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

$350,000

Present Value

$300,000

PV of Branch 1
PV of Branch 4
PV of Branch 8

p(B1)=81.4%
p(B4)=3.2%

$250,000
$200,000

p(B8)=0.3%

$150,000
$100,000
$50,000
$Period 1

Period 2

Period 3

Period 4

Fig. 4. Illustrative value trajectories for a three-state system (branches dened in Table 1).

Perfect
maintenance

p11(i)
p11(0)

New
New
p12(0)

Perfect
maintenance

p12(i)

Deteriorated

Perfect
maintenance

p13(0) p13(i)

Deteriorated

Imperfect
maintenance

Failed

p22(i)

p23(i)
Failed

Fig. 5. Performing perfect maintenance returns the system to the new state.

Perfect maintenance
shifts the reliability
curve to the right
Reliability

Fig. 7. Performing maintenance introduces new transition possibilities.

maintenance, it restores the initial (more favorable) transition

probabilities, which in effect ensures that the system is more
likely to remain in the new state post-maintenance than premaintenance. The two effects of maintenance on the value of a
system are shown in Fig. 8 for a three-state system. The following
simple examples will help further clarify this discussion.
Example 1: Perfect maintenance and the restoration of the initial
transition probabilities. Consider Branch B1 in the state evolution of
the system (see Fig. 3 and Table 1); B1 as dened previously
corresponds to a system remaining in state 1 (or the new state)
throughout the time periods considered. Under no maintenance,
as the system ages, the likelihood that it will remain on B1
decreases. For example, a system that remained on B1 until the
end of the second period, as shown in Fig. 9, will nd its likelihood
of subsequently remaining on B1 to be:

t
Fig. 6. Impact of maintenance on reliability.

maintenance at time tm4t0 as

pm1 Pk    P2 P1 p0
pm p0
pmk Pk    P2 P1 p0

pnoM_2 B1 p11 2p11 1 . . . p11 n

(12)

(11)

In short, maintenance acts on two levers of value: (1) it lifts the

system to a higher-value trajectory, and (2) in the case of perfect

whereas if perfect maintenance were performed on the system at

the end of the second period, the initial transition probabilities are
restored, and the likelihood that the system will remain for the

ARTICLE IN PRESS
K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

Since maintenance is the only cause of the system accessing

these higher-value trajectories after it has degraded, it is
appropriate to ascribe this incremental PV to maintenance and
term it the PV of maintenance.
In the next subsection, we account for the cost maintenance,
and conceptually, by subtracting it from the PV of maintenance,
we nally obtain what we set out to derive in this work, namely
the NPV of maintenance.

p11(1)

p11(0)

p12(1)
p13(1)

p22(1)
p23(1)

maintenance

p12(0)
p13(0)
p11(0)

New
Deteriorated

4.3. Net present value of maintenance

Failed
Maintenance
p22(1)

p12(0)

p22(2)

p23(1)

nte
mai

n an

ce

p23(2)

p13(0)

651

We model the cost of maintenance and the revenue-generation

capability of the system as being state and time-dependent. The
model can easily be extended to be dependent on the number of
previous repairs. The cost of perfect maintenance is represented
by cm1(i).7 Utility, as discussed earlier, is represented by um(i)
where m is the state of the system. The cost of maintenance is
assumed to increase as deterioration increases, and the revenuegeneration capability decreases as deterioration increases:
c1 iocm1 iock1 i
u1 i4um i4uk i 0

Fig. 8. Maintenance moves the system to a higher PV branch and, in the case of
perfect maintenance, restores the initial transition probabilities.

third period onwards on B1 increases and is given by

pM_2 B1 p11 0p11 1 . . . p11 n  2
and
pM_2 B1 4pnoM_2 B1

(13)

This result is easily extended for perfect maintenance occurring

at the end of any period j:

pM_j B1 4pnoM_j B1

(14)

Fig. 9 illustrates the system evolution over time under the

maintenance and no-maintenance policies.
Example 2: The PV of maintenance. Consider the three-state
system example discussed in Section 4.1, and assume that the
system transitions to a degraded state at the end of the rst
period. If no maintenance is performed, the system will remain on
the lower value branches shown in Fig. 10 (lower box). The
expected PV of these lower branches, calculated at the end of the
rst period, is given by Eq. (8). Using the numerical values
provided in the previous example, we nd:
hPVino_maintenance_1 $165; 600
Next, assume that perfect maintenance is carried out at the end
of period 1 after the system has degraded to State 2. Perfect
maintenance lifts the system up to higher-value trajectories,
shown in the upper box in Fig. 10. The expected PV of these upper
branches, calculated at the end of the rst period, after the
occurrence of perfect maintenance, is given by Eq. (8). Using the
numerical values provided in the previous example, we nd:
hPVimaintenance_1 $242; 500
Finally, consider the difference between the expected PV of the
branches enabled by maintenance and the branches without
maintenance:

DPV hPVimaintenance  hPVino_maintenance

(15)

This incremental PV, which in this case it is calculated at the

end of period 1, is what we dene as the PV of maintenance:

DPV1 hPVimaintenance_1  hPVino_maintenance_1 $76; 900

N X
k
X
um ipi m
i1 m1

and

(16)

We assume that the only costs of maintenance are those of

performing the maintenance. We do not account for costs
associated with the system failure, such as damage to other
equipment. The lost revenue caused by downtime is also
neglected since we assume that downtime is negligible with
respect to our time interval.
The expected PV of the system under no maintenance over N
time steps is given by Eq. (9), repeated here for convenience:
hPVNi

pM_j B1 p11 0p11 1 . . . p11 n  j

2pmok

1 r DT

N
X
Ui  pi
i1

1 r DT i

(17)

This PV of an engineering system under no maintenance policy

(Eq. (13)) will be used as the benchmark against which the
incremental PV of maintenance minus its cost will be compared.
When this value increment is positive, we consider that maintenance added net value to the system (or provided a positive
NPV). Once this framework is established, it is easy to conceive of
maintenance models that are designed to optimize this NPV of
maintenance.
4.4. Numerical examples
In the following, we consider two numerical examples that will
help clarify the previous discussion, and unambiguously capture
and we hope convince the reader ofthe value of maintenance.
We start with a simple case of a single PM intervention: First, we
capture the value of this PM intervention. Second, by varying the
timing of this intervention, we determine different values of
maintenance, which in turn allows us to identify the optimal timing
that maximizes the value of PM. This analysis provides the analytical
groundwork for developing value-optimal clock- and age-based PM.
In the second example, we consider the more reactive case of
corrective maintenance (i.e., no maintenance planning involved)
and demonstrate the value of corrective maintenance and its
dependence on different parameters.
Example 1: Single PM intervention. Consider the simple case
where we allow a single maintenance intervention, at time step
Nj. We assume the system is operated for Nf steps. In this case the
system deteriorates according to its transition matrix until step Nj,
7
Here the subscript m1 represents the transition from state m to the new
state (state 1) brought about by perfect maintenance.

ARTICLE IN PRESS
652

K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

Period 2

Period 1

Period 3

Period 4

Period n
Time

p11(2)

p11(1)

p11(n)

p11(3)

Branch 1
without
maintenance

p11(0)

p11(0)

p11(1)

p11(n-2)

p11(1)

Branch 1
with perfect
maintenance

Perfect
maintenance
occurs here

p11(0)

Fig. 9. Perfect maintenance and the restoration of the initial transition probabilities.

Period 1

Period 2

Period 3

Period 4
Time

Higher value (and more

likely) branches for the
maintained system
Maintenance lifts a
degraded system to
higher value branches

maintenance_1

PV

PV
provided by
maintenance

PV no_maintenance_1
Lower value branches of
the degraded system with
no maintenance

Fig. 10. The incremental present value provided by maintenance.

where we perform perfect maintenance, regardless of the state

the system is in. The PV of the system up to step Nj1 is given by
running Eq. (13) over (Nj1) steps:
Nj 1

hPVN j  1i

k
X

i1 m1

um ipi m
1 r DT i

k
X
cm1 Nj pN1 m
m2

1 r DT

Nj

hPVNj ; Nf i

X Ui  pi
i1

1 r DT i

(18)

Nf N j

u1 N j
1 r DT Nj

Nj 1

k
X X
um Nj ipi m

1 r DT Nj i

i1 m1
Nf N j

u1 N j
Nj

1 r DT

X UNj i  pi
i1

(20)

1 r DT Nj i

Combining the above three equations yields the PV of the

system up to step Nf:

The expected cost of maintenance in step Nm is given by

hC m1 ji

to step Nf is therefore given by

N j 1

(19)

Perfect maintenance returns the system to the transition

probabilities of a new system. The PV of the system from step Nj

hPVNf i

X Ui  pi
i1

1 r DT

1 r DT

k
X
cm1 Nj pN1 m
m2

N f Nj

u1 Nj

1 r DT Nj

Nj

X UNj i  pi
i1

1 r DT Nj i
(21)

ARTICLE IN PRESS
K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

653

Fig. 11. Value of a single maintenance intervention.

Fig. 11 shows the expected PV of our example system (as dened

by Eq. (21)) for the case where maintenance is performed
2 years into the systems operating lifetime. The cost of a
maintenance intervention is set in our numerical example
to be equal to the revenue generated during a single time step.
One can clearly see on Fig. 11 the effect of maintenance, which
lifts the system onto a higher-value trajectory. Furthermore,
the impact of the cost of maintenancein this case, it is set
equal to the revenues generated by system during one time
stepis indicated on Fig. 11 by the leveling of the value trajectory
at year 2.
Using Eq. (21), we can determine the NPV-optimal point
at which to perform a single maintenance intervention
for a system with a given set of transition probabilities and
utility prole u(t). We varied the timing of the maintenance
intervention, and performed a Monte Carlo simulation of the
expected PV of the system in each case. Fig. 12 shows the
relationship between the timing of the single maintenance
intervention and the expected PV of the system for different
initial probabilities of deterioration to a lower state. In each case,
we identify an optimum time at which maintenance should be
performed. The value-optimal timing of PM is shown to increase
as the probability of deterioration decreases; this can be
interpreted intuitively by noticing that the lower the transition
probability to a lower state, the more likely the system will
remain on a higher-value trajectory and for a longer period of
time. Thus, it is not optimal to have a maintenance intervention
early on that can only lift the system onto a slightly higher-value
trajectory.
Having determined the optimal timing of a single maintenance
intervention, we propose to explore in future work, the valueoptimal interval(s) for PM.
In the next example, we turn our attention to the more reactive
case of corrective maintenance and demonstrate its value.
Example 2: Perfect corrective maintenance following any deterioration. We calculate the expected NPV of running the system
with perfect corrective maintenance by determining the cumulative PV of the maintenance strategy in response to different
failure sequences. The expected value of the maintenance strategy
is then found by determining the probability of each failure
sequence and summing over the probability-weighted PV of all

the failure sequences. The result is shown in Fig. 13, and the
following discussion provides the analytics and assumptions for
our result.
Begin with a system which has not experienced deterioration
up to step (N11) and which transitions to a deteriorated or failed
state in step N1. We refer to this trajectory as branch Ba.
Maintenance is accordingly performed and the system is returned
to the new state. The PV of the system up to and including step N1
is found by subtracting the cost of returning the system to the
new state from the revenue generated by the fully functional
system:
hNPVN1 ; Ba i

N1
X
i1

u1 i
1 r DT

k
X
cm1 N 1 pN1 m

1 r DT N1

m2

(22)

To determine the expected PV of all the possible value

trajectories, we must determine the probability of being on
any given value trajectory. The general expression of this
probability is given by Eq. (7). We take advantage of the Markov
notation to express this probability for the particular case of no
deterioration for (N11) steps, followed by deterioration in the
N1th step as:
pdeteriorate at N1 pN1
pN1 1 1

"

k
X

pN1 m

m2

pN1 1 11  pN1 1

(23)

Therefore the probability-weighted NPV for the system up to

and including step N1 is
hNPVN1 i

NX
1 1
i1

u1 i
1 r DT

 pN1

k
X
cm1 N1 pN1 m
m2

1 r DT N1

(24)

We apply the probability only to the second term since under

our assumptions of perfect maintenance and negligible maintenance interval the system is instantaneously (relative to the
length of the time step) returned to its new state revenuegenerating capability. Following perfect maintenance the system
transition matrix returns to the initial value P1.
The system now remains in the new state for another (N21)
steps, before again deteriorating.

ARTICLE IN PRESS
654

K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

Fig. 12. Normalized value of maintenance relative to timing of maintenance intervention.

Fig. 13. Value of perfect corrective maintenance.

The probability-weighted PV for the system up to and

including step N1+N2 is given by
hNPVN 1 N2 i

NX
1 N 2

u1 i

i1

1 r DT

pN 2


i

k 
X

pN1

m2

cm1 N1 N2 pN2 m
1 rN1 N2

cm1 N 1 pN1 m

1 rN1
(25)

For a series of failures (N1, N2, y, Nn) the probability-weighted

PV under corrective perfect maintenance follows:
0
NPV@

n
X
j1

1
Nj A

Pn
N
j1 j
X

k
X

4 pN1 cm1 N1 pN1 m   


i
1 r DT N1
1

r
DT
m2
i0
3
Pn
cm1 j1 N j pNn m
5
Pn
(26)
pNn
N
1 r DT j1 j
u1 i

Eq. (18) allows us to calculate the probability-weighted

PV of corrective maintenance for a given series of failure
events. The expected value of corrective maintenance over N time
steps is then found by averaging over the possible failure
sequences.
We illustrate our argument using a simple hypothetical
example of a single-unit system. The system deteriorates
progressively through four states, from new (state 1) to failed
(state 4). For each state the probability of transitioning to the next
lower state is identical, and given by pdet. The transition matrix is
therefore given by
2
6
6
P6
4

1  pdet

pdet

1  pdet

pdet

0
0

0
0

1  pdet
0

0 7
7
7
pdet 5
1

(27)

ARTICLE IN PRESS
K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

To model system ageing despite maintenance, we allow pdet to

increase over time so that by the end of the planning horizon
there is probability of .5 of deteriorating to a lower state, even if
perfect maintenance has been performed in the previous time
step. We consider those failure sequences where failure occurs on
10% of the state transitions. We further assume that revenue and
maintenance costs are constant over time and use a time step of 6
months. For the nominal case we set the initial probability of
deterioration to 5%.
Assume the system generates $5000 per time step in the new
state, and revenues decrease quadratically through the lower
states, with the system generating no revenue in the failed state.
For the nominal case, maintenance costs eight times the
maximum revenue-generating capability of the system. We also
assess the impact of these assumptions on the value of
maintenance. Since the system proceeds probabilistically along

655

the value trajectories we used a Monte Carlo approach to sample

the trajectory space.
Fig. 13 shows the cumulative PV of a policy of perfect
maintenance in response to system deterioration compared to
that of a policy of no maintenance for the nominal case. The upper
curve shows the net benet of the perfect maintenance policy,
that is, the discounted revenues minus the discounted maintenance costs. The lower curve shows the expected PV of a zero
maintenance policy. The difference between the curves is the
value of corrective maintenance. Maintenance increases the PV of
the system, and delays the time at which cumulative NPV stop
increasing. The PV of maintenance for a given design lifetime can
be read off as the difference between the two curves. For example,
after 5 years of operation, the NPV of maintenance is approximately $13,500, in our example, as indicated by the arrow in the
gure.

Fig. 14. Normalized NPV of maintenance policy.

Fig. 15. Value of maintenance versus probability of deterioration.

ARTICLE IN PRESS
656

K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

Fig. 16. Impact of maintenance cost on value of maintenance. The x-axis shows the ratio of the cost of maintenance to the revenue generated by the system per time step.

Fig. 14 shows the normalized NPV of maintenance, dened as

the quotient of the NPV of the maintenance and PV of the nomaintenance policy:
norm_npv

NPVmaintenance
hPVinomaintenance

(28)

Consider next the drivers of the value of maintenance in our

example. Fig. 15 shows the relationship between the net value of
maintenance and the probability of deterioration. In this case we
assume that the probability of deterioration does not increase
over time. As the probability of deterioration increases, the value
of maintenance increases, achieving a maximum value at a
probability of approximately .84. Why is there a maximum value?
As the probability of deterioration increases, and maintenance
must be performed more often, the total cost of maintenance
increases. Eventually, the total cost of maintenance exceeds the
revenues generated by the system, at which point the system
should be written off as it will no longer be protable to operate.
Fig. 16 shows the impact of the cost of maintenance on the
normalized NPV of maintenance. The x-axis shows the ratio of the
cost of maintenance to the revenue generated by the system per
time step. As expected, the normalized NPV decreases as the cost
of maintenance increases, but with a slope of more than 1. In
other words, the value of maintenance decreases more slowly
than the cost increases.
Thus the NPV of maintenance increases as the probability of
deterioration increases, and decreases as the cost of maintenance
increases relative to the revenue generated by the system.
This example quanties the value of corrective, and illustrates
the functional dependence of this value on various characteristics
of the system (e.g., probability of deterioration, and cost of
maintenance relative to system income).

5. Conclusions
While maintenance optimization techniques abound in the
literature, most of these models focus on minimizing the cost of
maintenance or maximizing system availability, and seldom is
there mention or discussion of the value of maintenance.

In this paper, we argued that maintenance has value and

argued that existing cost-centric models ignore an important
dimension of maintenance, namely its value, and in so doing, they
can lead to sub-optimal maintenance strategies. We argued that
the process for determining a maintenance strategy should
involve both an assessment of the value of maintenancehow
much is it worth to the systems stakeholdersand an assessment
of the costs of maintenance.
We considered value as the net revenue generated by the
system over a given planning horizon. We did not include
additional dimensions of value such as the potential positive
effects of maintenance on environmental or health impacts. Such
effects can be incorporated in future work. Further, we did not
include factors such as safety or regulatory requirements. Such
factors can be easily included as constraints on the optimization
of value in future work.
We developed a framework for capturing and quantifying the
value of maintenance activities by connecting an engineering and
OR concept, system state, with a nancial and managerial concept,
the NPV.
We explored the impact of a systems state on the ow of
service the system can provide over time. Next we considered the
impact of maintenance on system state evolution and hence value
generation capability over time. We used traditional discounted
cash ow techniques to determine the value of maintenance,
which we dened as the difference between the expected NPV of a
system with and without maintenance.
By identifying the impact of system state or condition on NPV,
the framework and analyses developed in this paper provide
(nancial) information for decision-makers to support in part the
maintenance strategy development. Maintenance, as a consequence, should not be conceived as just an operational matter
and guided by purely operational matters but by multi-disciplinary considerations involving the marketing, nance, and operations functions within a company.
We made several simplifying assumptions in order to advance
our main argument. In future work these assumptions will be
validated or relaxed. Our assumption that downtime is negligible
can also be relaxed to incorporate the impact of downtime on for
example production levels. The work can be used as the basis of

ARTICLE IN PRESS
K.B. Marais, J.H. Saleh / Reliability Engineering and System Safety 94 (2009) 644657

the development of optimal maintenance policies in the presence

of constraints (safety, maintenance resources, etc.).
In particular, the work will be expanded to consider multi-unit
systems, in which the possibility of opportunistic maintenance
arises. For example, airlines perform additional PM once specic
panels on aircraft have been opened for corrective maintenance,
or when delays result in the aircraft being on the ground
for longer periods than originally planned. This work can be
used to (1) value opportunistic maintenance interventions and (2)
identify optimal approaches to selecting targets for opportunistic
maintenance.
Is it possible to take limitations in maintenance resources into
account (in the context of multi-unit systems this is relevant)? Is
opportunistic maintenance possible to be taken into account in
the method? How?
One important implication of this work is that the maintenance strategy should be tied to market conditions and the
expected utility prole (core revenue-generating capability of the
system). In other words, a value-optimal maintenance strategy is
dynamic and changes not only in response to environmental and
system conditions but also in response to market conditions.
Finally, we believe that the framework presented here offers
rich possibilities for future work in benchmarking existing
maintenance strategies based on their value implications, and in
deriving new maintenance strategies that are value-optimized.
References
[1] Waeyenbergh G, Pintelon L. A framework for maintenance concept development. Int J Prod Econ 2002;77(3):299313.
[2] Pham Hoang, Wang Hongzhou. Imperfect maintenance. Eur J Oper Res
1996;94:42538.
[3] Wang Hongzhou. A survey of maintenance policies of deteriorating systems.
Eur J Oper Res 2002;139:46989.
[4] Dekker Rommert. Applications of maintenance optimization models: a review
and analysis. Reliab Eng Syst Safety 1996;51:22940.
[5] Saleh JH. Flawed metrics: satellite cost per transponder and cost per day. IEEE
Trans Aerospace Electron Syst 2008;44(1).

657

[6] Saleh JH. Durability choice and optimal design lifetime for complex
engineering systems. J Eng Des 2008;19(2).
[7] Rosqvist T, Laakso K, Reunanen M. Value-driven maintenance planning for a
production plant. Reliab Eng Syst Safety, 2008, in press, doi:10.1016/
j.ress.2007.03.018.
[8] Rausand M, Hyland A. System reliability theory: models, statistical methods,
and applications. 2nd ed. New Jersey: Wiley-Interscience; 2004.
[9] Doyen Laurent, Gaudoin Olivier. Classes of imperfect repair models based on
reduction of failure intensity or virtual age. Reliab Eng Syst Safety
2004;84:4556.
[10] Tan Cher Ming, Raghavan Nagarajan. A framework to practical predictive
maintenance modeling for multi-state systems. Reliab Eng Syst Safety
2008;93(8):113850.
[11] Nakagawa T. Optimum policies when preventive maintenance is imperfect.
IEEE Trans Reliab 1979;28(4):3312.
[12] Nakagawa T. Imperfect preventive-maintenance. IEEE Trans Reliab 1979;28(5)
p. 402.
[13] Block HW, Borges WS, Savits TH. Age-dependent minimal repair. J Appl
Probab 1985;22(2):37085.
[14] Malik MAK. Reliable preventive maintenance scheduling. AIIE Trans
1979;11(3):2218.
[15] Kijima Masaaki, Morimura Hidenori, Suzuki Yasusuke. Periodical replacement
problem without assuming minimal repair. Eur J Oper Res 1988;37(2):
194203.
[16] Wang HZ, Pham H. Optimal maintenance policies for several imperfect repair
models. Int J Syst Sci 1996;27(6):5439.
[17] Wang Hongzhou, Pham Hoang. Optimal age-dependent preventive maintenance policies with imperfect maintenance. Int J Reliab Qual Safety Eng
1996;3(2):11935.
[18] Nakagawa T. Optimal policy of continuous and discrete replacement with
minimal repair at failure. Nav Res Logist Q 1984;31(4):54350.
[19] Wang HZ, Pham H. Some maintenance models and availability with imperfect
maintenance in production systems. Ann Oper Res 1999;91:30518.
[20] Nguyen DG, Murthy DNP. Optimal preventive maintenance policies for
repairable systems. Oper Res 1981;29(6):118194.
[21] Sinden JA, Worrell AC. Unpriced values: decisions without market prices. New
York: Wiley-Inter Science; 1979.
[22] Saleh JH, Lamassoure E, Hastings DE, Newman DJ. Flexibility and the value of
on-orbit servicing: a new customer-centric perspective. J Spacecraft Rockets
2003;40(1):27991.
[23] Saleh JHK, Marais K. Reliability: how much is it worth? Beyond its estimation
or prediction, the (net) present value of reliability. Reliab Eng Syst Safety
2006;91(6):66573.
[24] Macke M, Higuchi S. Optimizing maintenance interventions for deteriorating
structures using cost-benet criteria. J Struct Eng 2007;133(7):92534.
[25] Marais K, Lukachko SP, Jun M, Mahashabde A, Waitz IA. Assessing the impact
of aviation on climate. Meteorologische Zeitschrift 2008;17(2):15772.