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

net/publication/325570785

A real options approach for joint overhaul and replacement strategies with
mean reverting prices

Article  in  Annals of Operations Research · June 2018

DOI: 10.1007/s10479-018-2906-z

CITATIONS READS

2 71

3 authors, including:

Alejandro Mac Cawley Rodrigo Pascual

Pontifical Catholic University of Chile University of Concepción
18 PUBLICATIONS   105 CITATIONS    75 PUBLICATIONS   880 CITATIONS   

SEE PROFILE SEE PROFILE

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

Sustainable/Low Emissions Mining Operations View project

Production Planning in high uncertainty enviroments: The Wine Case View project

All content following this page was uploaded by Alejandro Mac Cawley on 24 April 2019.

The user has requested enhancement of the downloaded file.

Ann Oper Res
https://doi.org/10.1007/s10479-018-2906-z

S.I. : CLAIO 2016

A real options approach for joint overhaul and

replacement strategies with mean reverting prices

Alejandro Mac Cawley1 · Maximiliano Cubillos1 ·

Rodrigo Pascual2

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract Due to its significant impact on economic performance, an effective equipment

overhaul and replacement strategy is a key aspect of physical asset management in capital-
intensive industries, such as the mining industry. Classical approaches suggest periodic
interventions based on the physical condition of the equipment, considering factors such
as availability and operational costs. These fixed models generally ignore two important
aspects: first, the flexibility of the decision to overhaul or replace, which may be re-evaluated
within a given period, and second, the uncertainty around economic factors that may affect
future maintenance decisions, such as the product price. This work improves on classical
models by considering the effect of integrated price uncertainty in the definition of joint
overhaul and replacement strategy, using a real options approach and a mean reversion bino-
mial model to represent the uncertainty in price. More specifically, we develop a real options
model and use a backwards recursion algorithm to determine an optimal intervention policy
that maximizes expected profits. We then present a numerical study of the mining industry to
validate the effectiveness of the proposed methodology. Results show that the option-based
decision model economically outperforms the classical periodic strategy approach from with
net present value increments ranging from 36.8 to 8.6%, according to the number of periods
in the maintenance cycle, offering evidence that a new approach to equipment overhaul and
replacement strategy is needed.

B Alejandro Mac Cawley

amac@ing.puc.cl
Maximiliano Cubillos
mecubill@uc.cl
Rodrigo Pascual
rpascual@ing.puc.cl
1 Department of Industrial and Systems Engineering, Pontificia Universidad Católica de Chile,
Santiago, Chile
2 Department of Mining Engineering, Pontificia Universidad Católica de Chile,
Santiago, Chile

123
 Ann Oper Res

Keywords Periodic maintenance · Joint replacement and overhaul policy · Real options ·
Mean reversion

1 Introduction

In asset intensive and resource-based industries, such as gas, petrochemicals, energy, forestry
and mining, investment and maintenance policies can significantly affect firms’ profitability.
As a result, maintenance policies are a part of strategic and operational decision-making
in these industries, and maintenance costs can account for a significant portion of their
operational costs. Breakdowns and downtimes also impact plant capacity, product quality,
and production costs, as well as health, safety and the environment (Parida and Kumar
2006). In addition, commodity-based industries face significant price volatility and cyclical
fluctuations; the rapid (and often unexpected) transitions between price booms and price
slumps are challenging issues for decision makers in these resource-based industries to deal
with Cashin et al. (2002).
Due to price uncertainty and the impact of maintenance decisions on the operating costs,
one might expect managers to devote significant attention to these decisions. But in practice,
most firms base their equipment maintenance strategies on deterministic price scenarios
and aim only to minimize the present value of lifetime equipment costs (Kim and Thomas
2013). Also, most existing maintenance models focus on studying and modeling equipment
failure and repair processes, with little or no attention paid to the effect of product prices on
maintenance policies. Even in the case of power plants, where price is the determinant factor
in the production process, maintenance decisions consider it a production loss cost rather
than part of the income or profit function (Carazas and Souza 2010).
Alsyouf (2007) suggests an explanation for managers’ failure to take product prices into
account in the analysis of maintenance decisions. Using the Swedish paper-mill industry as
a case, he defines profitability as the product of productivity and price recovery (Sumanth
1998). Since price recovery is directly related to the product market price, and in the case
of commodity industries, prices cannot be controlled by the firm; maintenance policies is
the preferred mechanism to improve the productivity and profitability of the firm. With
improvements in profitability which can go up to a as much as 12.5%, managers mainly
focus their attention on maintenance policies rather than considering the product price in
their decision process.
While real options methodology has not been extensively used in the maintenance context,
it has received attention in operations research because it allows for the integration of both
operational and financial considerations within a stylized, but representative, modeling setting
(Ding et al. 2007). The use of a real option approach offers a two-stage framework for a firm’s
decision to do minimal repair, overhaul equipment, or replace equipment. In the first stage,
this decision is made in the presence of price uncertainty; in the second stage, after observing
prices, the firm exercises its option to perform minimal repair, overhaul or replace, and failures
are corrected.
Our primary contribution is the integration of existing Periodic Maintenance (PM) models
with Real Options (RO) methodology, allowing managers to account for price variability in the
maintenance decision process. Combining these two methodologies extends the scope of the
existing models by introducing the following improvements: (1) accounting for uncertainty
in price, (2) considering a flexible, rather than a periodic, policy, (3) allowing for the creation
of contingency plans for different price scenarios. We compare the standard PM model with

123
Ann Oper Res

a version that uses real options, analysing the differences between the maintenance decisions
each prescribes. We determine that the addition of flexibility generates value. Finally, we
perform a sensitivity analysis to examine which variables significantly affect the real-options
decision process.
The rest of this paper is organized as follows. We first present a literature review, focusing
on real options applications, uncertainty in maintenance models and resolution methods.
Section 3 describes two models: the conventional PM model and the proposed RO model.
In Sect. 4, we present the mean reversion binomial model. We then proceed to present a
numerical case study with information from the mining industry, which validates our results
and presents a sensitivity analysis. In Sect. 6, we discuss the implications of our results and
present possible extensions of the model and future work. Finally, Sect. 7 concludes.

2 Literature review

Maintenance activities are usually classified into two categories: corrective maintenance,
which is applied once a system has failed, and preventive maintenance, in which the equipment
is maintained while still operative. Preventive maintenance aims to reduce the probability
of failure or deterioration of the equipment. Normally, this maintenance is either time-based
(performed at defined time intervals), or condition-based (based on predetermined criteria for
the condition of the equipment) (Wu and Zuo 2010). Mathematical models for maintenance
activity vary based on the type of maintenance (minimal, imperfect or perfect), replacement
type (preventive or corrective) and cost structure (Wang 2002). Most of these models account
for uncertainty in the equipment failure process, but in commodity industries such as mining,
forestry, oil and gas, commodity price volatility introduces an additional source of uncertainty.
Real options account for the possibility of making changes in any given project or invest-
ment after previously unknown information is revealed. Mun (2006) defines real options
methodology as: a systematic approach and integrated solution, using financial theory and
economic analysis, in applying options theory in valuing real physical assets in a dynamic
and uncertain business environments. This methodology, initially used to value financial
derivatives, has been extended to the valuation of flexibility in highly uncertain environ-
ments (Gunther McGrath and Nerkar 2004). Usually, this flexibility includes the option to
expand, contract or abandon a project (Dixit and Pindyck 1994). The wide range of potential
applications for real options includes natural resources investments, R&D projects, opera-
tions management, market competition and revenue management (Zeng and Zhang 2011).
Real options research has focused primarily on investment decisions, but operations research
(OR) applications have also proven to be useful (Crasselt and Lohmann 2016). Examples
of real options applications at an operational level include production planning, machine
flexibility, material handling system flexibility and operational flexibility [For a survey on
real options in OR see Zeng and Zhang (2011)].
Real options has been used to model operational planning in volatile price environments.
Dalal and Alghalith (2009) study production decisions under production and price uncer-
tainty. Song (2009) considers the problem of production and preventive maintenance control
in a system subject to multiple uncertainties, such as random customer demands, machine
failure, repair and stochastic processing times. Lim (2013) present a joint optimal pricing
and order quantity model, in which demand is modeled as a function of external price and
cost is modeled as a function of quantity. But little attention has been paid to the use of real
options from a maintenance perspective (Jin and Ni 2013). The maintenance strategy litera-
ture has traditionally considered uncertainty by stochastically modeling internal equipment

123
 Ann Oper Res

parameters such as failure times, operational and maintenance costs, and failures conditions,
among others (Abdel-Hameed 2013). However, some recent contributions have considered
external uncertainties that affect maintenance and replacement decisions. Zambujal-Oliveira
and Duque (2011) studies the optimal moment for asset replacement in a given tax envi-
ronment and depreciation policy using a real-option approach. They develop a two-factor
(Operation/Maintenance cost and salvage value) model with Brownian process to model
uncertainty. Mardin and Arai (2011) consider a system dynamics model for overhaul and
replacement policies subject to the emergence of new technologies that compete with cur-
rently used equipment. Nguyen et al. (2011) propose solving for the optimal overhaul and
replacement policy by modeling profit flows and new technology purchase prices as stochas-
tic processes. Richardson et al. (2013) study the problem of determining when to order
replacements for equipment with long and uncertain delivery lead times using simulations in
a real options framework. No previous work has considered the effects of commodity price
uncertainty in the definition of joint equipment overhaul and replacement strategies.
Andersson (2007) studied almost 300 different commodities over a time period of 36 years
(1970–2006) and found clear evidence that commodity prices are mean-reverting, in contrast
to other financial assets. There are a number of studies have used geometric Brownian motion
to model commodity prices (Brennan and Schwartz 1985; Paddock et al. 1988; Nembhard
et al. 2003). However the mean reverting models are considered to be more appropriate than
the widely used Geometric Brownian motion process for commodity prices as Zhang et al.
(2015) points out. The economic argument to support mean-reverting behaviour is that high
commodity prices stimulate investment to increase production as well as the development of
alternative products, resulting in price decreases. The opposite occurs when prices are low;
as a result, the price tends to revert to a long term mean. Other empirical studies also sup-
port the idea that mean-reverting models can accurately model the evolution of commodity
prices (Schwartz 1997; Bessembinder et al. 1995). It is common that when determining the
value of real options in operations problems, which usually involve interdependence between
decisions across time periods, the preferred modelling approach is a discrete one(Bengtsson
2001). The preferred method to model mean-reverting processes in discrete time are multi-
nomial lattices (Cox et al. 1979) and Monte Carlo simulation (Boyle 1977). However, for
path-dependent options, these approaches present significant limitations due to computational
complexity and practical implementation constraints (Lander and Pinches 1998). In the real
option literature there are a number of papers which consider mutually exclusive projects or
actions (Childs et al. 1996; Dixit 1993; Décamps et al. 2006; Siddiqui and Fleten 2010). The
advantages of a binomial model over alternative valuation methods include its versatility,
easy implementation and precision. Moreover, in contrast to other methods that resemble a
black box, binomial models visually map out decision processes in the optimal decision tree,
allowing for further analysis (Bastian-Pinto et al. 2010). Extensions of this model include the
use of multi-factor mean-reverting processes (Wang and Dyer 2010), non-censored binomial
mean-reverting processes (Bastian-Pinto et al. 2010) and binomial models with changing
volatility (Haahtela 2010).
There is an ample range of studies that use binomial trees to represent mean-reverting
processes in the financial and commodity sectors. In the area of commodity prices, Slade
(2001) applies a mean-reverting process to metal prices in order to value options in a mining
operation. Hahn and Dyer (2008) consider an oil and gas switching option which requires a
binomial model of two correlated one-factor commodity price models. Bastian-Pinto et al.
(2010) extend the model presented by Hahn and Dyer and evaluate the option to expand
considering bio-fuels prices. In the financial arena, Hull and White (1994) uses trees to
represent and value interest rate derivatives and Jaillet et al. (2004) use multinomial trees to

123
Ann Oper Res

value swing options. To represent the mean-reverting behavior of commodity prices, we use
the model proposed by Hahn and Dyer (2008) and analyze how these price changes affect
maintenance and replacement decisions.
One challenge in relying on real options to formulate operational problems is the latter’s
path dependency. A maintenance decision made in any given period depends not only on
the prevailing commodity price, but also on previous maintenance decisions, which directly
affect the equipment failure rate. This augments the recombining structure of the binomial
tree originally presented by Cox et al. (1979); instead of having N+1 nodes at time N, the
non-recombinant tree has 2 N nodes. Backwards deduction heuristics have been used to solve
an optimal decision tree using the binomial lattice approach (Bertocchi et al. 2006; Brandão
and Dyer 2005; Li and Kouvelis 1999; Li et al. 2009; Messina and Bosetti 2006). To determine
the optimal maintenance policy and the value of flexibility in the option tree, we will start at
the end of the tree and work backwards with a dynamic programming approach according to
a risk-neutral valuation and risk-neutral probabilities (Huchzermeier and Loch 2001).

3 Model formulation

Our model is inspired by Zhang and Jardine (1998), who aim to determine the optimal
overhaul and replacement policy for a single asset. We add two characteristics to their model.
First, we include the opportunity for the decision maker to choose between three possible
actions at any given time: minimal repair at failure, overhaul or replacement. Second, we
maximize profit instead of minimizing cost, subject to a mean-reverting binomial process of
product prices. Each maintenance action will have a different outcome, which is modeled as
a reduction in the failure rate, cost and profit loss due to equipment unavailability. Product
prices are modelled and determined externally through a mean reversion binomial model.

3.1 Failure rate reduction model

We consider the equipment failure rate function (failures per unit of time) as a monotonically
increasing function λ(t). Let’s consider λk−1 (t) as the equipment failure rate just before time
(k − 1) and λk (t) as the failure rate just after an overhaul. T corresponds to the time since
the last overhaul and p ∈ (0, 1) is a constant denoting the degree of improvement as a result
of the overhaul. For any time t after an overhaul, the failure rate can be expressed as:

λk (t) = pλk−1 (t − T ) + (1 − p)λk−1 (t) (1)

Note that if the improvement factor p = 0 then λk (t) = λk−1 (t). This corresponds to a
minimal repair, in which the failure rate is not altered. On the other hand, if p = 1 then
λk (t) = λk−1 (t − T ), which returns the failure rate to the condition of the equipment at the
previous overhaul period.

3.2 Failure cost model

The discounted cost due to minimal repairs in a time interval (0, t) depends on the minimal
repair cost Cm and the cost of the lost production while the equipment was in repair (Tm ):

 t
(Cm + x Tm P) λ(t)e−θ t dt (2)
0

123
 Ann Oper Res

where λ(t) corresponds the failure rate, e−θ t is the discount factor, x is the commodity
price and P is the equipment productivity.

3.3 Periodic maintenance model

As a base case, we will use a standard model of periodic overhaul and replacement interven-
tions which minimizes the total discounted cost in an infinite time horizon. We consider the
replacement, overhaul and minimal repair costs (Cr , Co , Cm respectively) as constants. The
decision variables are: the number of overhauls implemented in a replacement cycle (N ) and
the time between overhauls (T ). The replacement is performed in the period t = N T . The
discounted cost function for the first replacement cycle can be expressed as:

N −1  NT
C1 (N , T ) = (Co + x To P)e−iθ T + (Cm + x Tm P) λ̂(t)e−θt dt + (Cr + x Tr P)e−θ N T (3)
0   
i=1   
   Replacement cost
Failure costs
Overhaul costs

The first term represents the discounted sum of the costs of the N − 1 overhauls imple-
mented every T periods, the second term is the total discounted failure costs in the replacement
cycle time N T and the third term is the replacement discounted cost. λ̂(t) is the effective
failure rate at time t due to the periodic policy (N , T ). The discounted cost function for n
production cycles is:

Cn cycles (N , T ) = C1 + C2 e−θ N T + . . . + Cn e−(n−1)θ N T (4)

Since C1 = C2 = . . . = Cn , in a infinite repetition the total discounted cost can be

expressed as:
C1
min (5)
N ,T (1 − e−θ N T )
If P is the equipment productivity and xt is the price of the commodity at period t. The
net present value (N P V ) function for the periodic maintenance model with a single optimal
replacement cycle N̂ and optimal time between overhaul T̂ , can be expressed as:

N̂
N P V1 ( N̂ , T̂ ) = (xt P)e−iθ T̂ − C1 ( N̂ , T̂ ) (6)
  
i=1
   Costs
NPV Income

3.4 Real options model

We consider a finite time horizon that can be divided into T discrete time intervals of equal
length, and the profit in each period depends on price and the equipment performance. We can
now proceed to relax the condition of a periodic maintenance policy, as was the base case, with
all other assumptions unchanged. The main difference between this Real Options (RO) model
and the Periodic Maintenance (PM) model is that the commodity price varies at each period
t. Since Andersson (2007) shows that commodity prices follow a mean reverting process, we
model the logarithm of the commodity price xt as a one-factor Ornstein-Uhlenbeck process
(Schwartz and Smith 2000), defined by the following equation:

dYt = κ(Ȳ − Yt )dt + σ dz (7)

123
Ann Oper Res

We refer to the failure rate before an intervention (overhaul or replacement) as λ̂k−1 (t)
and the failure rate immediately after an intervention as λ̂k (t). Where Yt = log(xt ) is the
logarithm of the commodity price, κ is the mean reversion coefficient, Ȳ is the logarithm of
the long term average price, σ is the volatility and dz is the Wiener standard process.
For each period (t) the decision maker has to take a decision over three possible options:

– Overhaul : The failure rate is reduced by a factor p ∈ (0, 1) at a cost Co . The profit in
period t given this option is:
 t+1
πt = xt P − (Cm + xt Tm P) λ̂(t)e−θ t dt − Co − xt To P (8)
   
Income  t  Overhaul costs
Failure costs

– Replacement: The equipment is replaced at a cost Cr . The failure rate immediately after
replacement is the original rate and the profit in period t given this option is:
 t+1
πt = xt P − (Cm + xt Tm P) λ̂(t)e−θ t dt − Cr − xt Tr P (9)
 t   
Income    Replacement costs
Failure costs

– Minimal repair: The deterioration of the component continues and the profit in period t
given this option is:
 t+1
πt = xt P − (Cm + xt Tm P) λ̂(t)e−θ t dt (10)
 t
Income   
Failure costs

The replacement, overhaul and minimal repair costs are constants over time and Cr >
Co > Cm . In order to determine the best maintenance policy, the decision maker will maxi-
mize expected profit from interval t to the final interval T , which is given by the following
recursive equation:
∗
Vt (xt , Rt−1 ) = max πt + E{Vt+1 (xt+1 , Rt∗ )e−θ } (11)
Rt

∗ corresponds to the real option selected at time interval t − 1, R ∗ is the real option
where Rt−1 t
selected in the current time interval and E{Vt+1 (xt+1 , Rt∗ )e−θ } is the expected value for the
∗ ) is the total expected value from interval t to the final interval.
next period. So Vt (xt , Rt−1
To determine the expected availability for the real option model, we proceed first to obtain
the total expected downtime of the equipment. This value is determined by the sum of the
expected downtimes produced by: equipment failure, overhauls and replacements. Then we
proceed to calculate the percent of the time that the equipment is unavailable, by dividing the
downtime by the total available time (T ), and the reciprocal corresponds to the availability.
The equations are given by:

E(Equipment Downtime) = E(Failure time) + E(Overhaul time) + E(Replacement time)

(12)
E(Equipment Downtime)
Availability = 1 − (13)
T

123
 Ann Oper Res

Fig. 1 Non-recombinant binomial tree

4 Mean reversion binomial model

In order to implement the RO model, we use a binomial lattice process, which allows us to
model price fluctuations in discrete intervals and adds the flexibility of real options. In order to
model the commodity price mean reversion binomial process, we use the approach proposed
by Hahn and Dyer (2008). They propose a binomial sequence of n periods of length t in
a time horizon T . The objective is to find a binomial sequence that converges to a general
differential equation of the form: dYt = μ(Y, t)dt +σ (Y, t)dz, where μ(Y, t) and σ (Y, t) are
the instantaneous drift and standard deviation functions. Note that, in contrast to the general
binomial model, both the drift and deviation are functions of time. As we assume a simple
mean-reverting process, μ(Y, t) = κ(Ȳ − Yt ) and σ (Y, t) = σ . The up and down transition
probabilities which converge to the mean-reverting process can be modeled by the following
expressions:
 
1 κ(Ȳ − Yt )(t)
pt = max 0, min 1, + (Probability of price increase)
2 2σ
1 − pt (Probability of price decrease) (14)

As can be observed, price probabilities are bounded to values between 0 and 1; over time,
commodity prices √
follow a mean-reverting
√
pattern with increases and decreases in each node
given by u = eσ (t) and d = e−σ (t) , respectively.
An important source of complexity in this situation is the fact that maintenance decisions
are path-dependent, since overhaul effects are dependent on previous maintenance decisions.
The replacement decision is the only one that resets the system to its initial state. The current
commodity price is also dependent on the path. In the case of a recombinant decision tree,
the path followed to reach a given node is not affected by past decisions. Figure 1 shows the
path dependency of a binomial tree. In this case there are 2t nodes at each period instead of
t + 1 in the recombinant case, which significantly increases the computational complexity
of the problem.
To construct the decision tree, we start by initializing the root (t = 0) from which two
state-of-nature nodes follow: the up and down price states. Three decision nodes follow

123
Ann Oper Res

from each state of nature, accounting for the three real options possibilities: minimal repair,
overhaul or replace. For each decision node we determine the resulting profit.

Algorithm 1 Decision tree generation

1: initialize P0 , σ, Cr, Co , Cm ,√κ, β0 , β1 , p
2: compute dt = T /N , u = eσ dt , r = e−θdt , Ȳ = log(P0 )
3: initialize root node
4: set node price x0 = P, Y0 = log(x0 )
5: set initial failure rate λ0 (t) = eβ0 +β1 t
6: for t in 1 to T do
7: for i in 0 to 1 do
8: node add child  state-of-nature node in t
i
9: set node price xt = xt−1 u −1 , Yt = log(xt )
10: compute probability q = 21 + κ(Ȳ −Y t )(t)
2σ
11: if i = 0 then
12: Pt = q
13: else
14: Pt = 1 − q
15: end if
16: for j in 1 to 3 do
17: if j = 1 then  Minimal repair option
18: node add child  decision node in t
19: set node failure rate λk (t) = λk−1 (t)

20: set node profit πt = xt P − Cm tt+1 λ(t)r dt

21: else if j = 2 then  Overhaul option
22: node add child  decision node in t
23: set node failure rate λk (t) = pλk−1 (t − T ) + (1 − p)λk−1 (t)

24: set node profit πt = xt P − Cm tt+1 λ(t)r dt − Co

25: else  Replacement option
26: node add child  decision node in t
27: set node failure rate λk (t) = λ0 (t)

28: set node profit πt = xt P − Cm tt+1 λ(t)r dt − Cr

29: end if
30: end for
31: end for
32: end for

Once the decision tree is generated, we can determine the optimal maintenance decisions
by working backwards recursively from the end of the tree (Rubinstein 1994). For each state-
of-nature node at time t, the algorithm selects the child node which maximizes the profit in
time t plus the expected value of the previous nodes:
 
E[Vt+1 (xt+1 , Rt+1 )]
Vt (xt , Rt ) = max πt (xt , Rt ) + (15)
Rt (1 + r )
where,
E[Vt+1 (xt+1 , Rt+1 )] = Ptu · Vt+1 (u · pt ) + Ptd · Vt+1 (d · pt ) (16)
The expected NPV for the RO model is obtained by taking the pondered sum of the profits
over the entire optimal path in the tree, which corresponds to the value of the initial node V0 .

123
 Ann Oper Res

Algorithm 2 Solution algorithm

1: for all node in state-of-nature level in t = T do
2: select the highest profit πt in child nodes and save as node value Vt
3: end for
4: for t = (T − 1) to 1 do
5: for all node in decision level do
6: for all node in child nodes do
7: compute NPV as the sum of node value pondered by node probability
8: NPVt = NPVt + Pt+1 Vt+1
9: end for
10: compute node NPV NPVt = NPVt + πt
11: end for
12: for all node in state-of-nature level do
13: select the highest NPV from child nodes and save as node value Vt
14: end for
15: end for

The PM and RO models consider the following assumptions:

1. A single piece of equipment in an intensive productive system.
2. n decision periods in a fixed time horizon of evaluation T = nt.
3. Equipment is subjected to three types of actions: minimal repair, overhaul, and replace-
ment; each action has its own associated costs (Cm < Co < Cr ) and implementation
times (Tm < To < Tr ).
4. Equipment has an exponential failure rate λ(t) = eβ0 +β1 t .
5. An overhaul improves the equipment by a fixed degree (0 < p < 1), directly affecting
its failure rate.
6. A minimal repair is implemented after every equipment failure and does not modify the
failure rate.
7. The overhaul and replacement options are available for each decision period.
8. The commodity price is uncertain and exhibits mean-reverting behavior.
9. The long-term mean commodity price is x = x̄.
10. Price volatility (σ ) and mean reversion parameters (κ) are constants.

5 Numerical case study

We now present an application of our model to a mining industry case, which uses similar
parameters to those used by Pascual et al. (2016) specifically, a production unit with a single
critical repairable component which operates continously over t = 8. The failure rate is
modeled by λ(t) = eβ0 +β1 t . The overhaul improvement degree is p = 0.7. The minimal
repair, overhaul and replacement costs are Cm = 6, Co = 45 and Cr = 100, respectively.
The discount rate is θ = 0.08 and the failure rate parameters are β0 = 3 and β1 = 0.18. The
long term price, volatility and mean reverting parameter are x 0 = 3, σ = 0.1 and κ = 0.15,
respectively.
To solve the RO model, the algorithms were coded using Python (v2.7) and the ETE 3
(Huerta-Cepas et al. 2016) toolkit for tree-like structure analysis and visualization. The code
was executed using using a cluster with 488GB of RAM with 16 2.3GHz Intel Xeon E7-8880
v3 with Microsoft Windows Server 2016 datacenter operating system.

123
Ann Oper Res

6 Results and sensitivity analysis

In this section we present the results obtained using the binomial lattice model. We compare
the optimal decisions and expected Net Present Value (NPV) under the RO model with
those of the PM model, in order to determine the impact of incorporating commodity prices
uncertainty into maintenance decisions. We also study the effect of the number of time
periods in the optimal maintenance policies and proceed to present the results for cycles of
3 to 10 periods. Finally we characterize the effect on the net present value, of changes in:
maintenance costs, equipment aging factor and the commodity volatility factor.
The results for maintenance time periods which extend from 3 to 10 are presented in
Table 1 and Fig. 2. They show that there is a significant percentage differences in the NPV
between the RO model and the PM model. These difference in the NPV, for an aging factor

Table 1 NPV percentage difference between real option (RO) and periodic model (PM) and execution time
of RO model for different time periods, aging factor and replacement/overhaul cost ratio

Time Aging factor β1 = 0.18 Replacement/overhaul Exec. time (s)

cost ratio Cr /Co = 3.5

VPN RO VPN Per. % Difference VPN RO VPN Per. % Difference

3 151.58 110.83 36.8 149.34 57.31 160.6 0.06

4 174.12 145.55 19.6 133.6 93.84 42.4 0.48
5 198.67 169.11 17.5 146.21 82.6 77 2.52
6 218.79 194.09 12.7 165.72 98.97 67.4 15.17
7 240.10 214.74 11.8 179.93 116.63 46.6 93.12
8 257.69 236.47 8.9 174.16 130.58 33.8 552.43
9 276.21 254.43 8.6 179.56 134.2 33.8 3398.28
10 291.46 256.12 13.8 188.96 141.54 33.5 13,730.93

300 Execution Time 10,000

280 NPV Real Options β 1 = 0.18
NPV Periodic β1 = 0.18
260
NPV Real Options C r /Co = 3.5 1,000
240
NPV Periodic C r /Co = 3.5
220
Log(Time) (Sec.)

200
NPV (mu)

180 100
160
140
10
120
100
80
1
60
40
20
0.1
0
0 1 2 3 4 5 6 7 8 9 10 11
Number of Maintenance Periods (T)

Fig. 2 Net present value for RO and PM models for different time periods. Execution time (seconds) for real
option model for different time periods

123
 Ann Oper Res

of the equipment of β1 = 0.18, range from 36.8 to 8.6%, with an average of 16.2%. For
the case in which the replacement/overhaul cost ratio is Cr /Co = 3.5, those differences are
increased and range from 160.6 to 33.5%, with an average of 64.9%. The average difference
of 16.2 and 64.9% observed in both cases, is a result of the flexibility to adjust maintenance
policies based on price, which is embedded in the real options model, which is also known as
the value of flexibility (Kulatilaka 1995). The increased difference for the case in which the
replacement/overhaul cost ratio is Cr /Co = 3.5, is an indication that there is a significant
effect of maintenance costs in the decisions taken by the RO model. Finally, we can observe
that as the number of maintenance periods is increased the time required to solve the model
grows exponentially, from 0.06 seconds for a 3 period model to 13,730.93 seconds for a 10
period model.
To compare how the optimal decisions change at each time period when price uncertainty
is taken into account, we can observe Fig. 3. We can observe, for each maintenance cycles
which extend from 3 to 10 time periods, the distribution of the actions (Minimal repair,
overhaul and replacement) taken at each time period. For example, for the base case of an
aging factor of the equipment of β1 = 0.18, the decision of the RO model differs from the
PM in just the second period, where a 25% of the times and overhaul is performed and a
replacement is performed in the 75% of the times, depending on the market conditions. For
the case of a 10 period maintenance cycle we can observe that the differences occur in the
6th, 7th and 10th period. When the replacement/overhaul cost ratio is Cr /Co = 3.5, we can
observe that the RO model adjusts the maintenance decision from the 3rd period onward,
making much more use of the overhaul option, due to the increased replacement cost. For the
maintenance cycles of 8 and 10 periods, we can see that the RO model makes a more wide
variety of decisions from the 4th period onward. Finally, when the overhaul improvement
degree is reduced from p = 0.7 to p = 0.6, we can observe that the RO model stops using
the overhaul option at the 2nd period and only in the even numbered cycles uses the overhaul
option, in just the next to the last period.
If we look the moment in which most of the differences between the decisions taken by
the RO and PM model occur, those differences present mostly at the final periods of the
maintenance cycle (All models when T = 10 in above time period 7). This is also the moment
in which the commodity prices can present the highest dispersion from the original level.
Hence, we can infer that as the commodity prices have a greater dispersion, the value of
taking different courses of actions is higher, increasing the value of postponing or changing
maintenance decisions.
The effect of price variability in the decision taken by the RO can be observed in Fig. 4, we
can observe that if we increase the variability from σ = 0.1 to σ = 1.5 the RO model changes
significantly the decisions taken at each time period, with a wider range of possibilities due
to the price fluctuations.
Since the RO model reacts to price changes and prices follow a mean-reverting process,
it is important to analyse the effect of the volatility of prices over the flexibility value. The
effect of and increased volatility in commodity prices is also shown in Table 2 and Fig. 5.
Over this range, the flexibility value for the mean-reverting model increases as the volatility
and the maintenance cycle rises; from 28.67 to 46.36, from 24,51 to 88.01 and 21.02 to
133.64 for 4, 6 and 8 maintenance cycles, respectively. In here we can observe the net present
value difference at different price volatilities, for different lengths of maintenance cycles
(T = 4, T = 6 and T = 8). As the price volatility reaches a value of 1, the net present
value difference between the RO and the PM models is exponentially increased and its more
significant as the length of the maintenance cycles is higher. As a result, including commodity

123
Ann Oper Res

Fig. 3 Percentage of maintenance actions taken for each time period, for an equipment aging factor of
β1 = 0.18, replacement/overhaul cost ratio of Cr /Co = 3.5 and overhaul improvement degree of p = 0.6;
for maintenance cycles ranging from T = 3 to T = 10

price considerations in the definition of overhaul and replacement policy is more valuable in
highly uncertain environments.
Figure 6 also shows the effect of the commodity price in the maintenance decision taken,
for a maintenance cycle of 8 periods (T = 8) and a replacement/overhaul cost ratio is
Cr /Co = 3.5. The lines corresponds to a logistic model fit of the decision (Minimal repair,
overhaul and replace) against the price. The model has a χ 2 = 19.32 ( p value <0.01) and all
parameters are significant at less than 1%. As the price is increased the percentage of minimal
and overhaul actions are reduced while the replacement actions are increased. An increase in

123
 Ann Oper Res

Fig. 4 Percentage of maintenance actions taken according to each time period for a maintenance period of
T = 8 and variability levels of σ = 0.1 and σ = 1.5

the percentage of replacement actions, as the price increases, can be interpreted as a reaction
of the model to increase the equipment availability and capture the benefit associated to
these high prices. This is also confirmed in Fig. 7 which shows the equipment availability
at different price level under increments of 0.75. We can observe that as the price increases,
up to a level of $2.8125, so does the equipment availability. The availability is increased up
to an average level of 74%, over which the marginal cost to improve the availability is not
compensated by the price increase.
The NPV and the value of flexibility are significantly affected by the maintenance cost
parameters, specifically by the replacement and overhaul cost ratio (Cr /Co ) and the minimal
repair cost (Cm ) parameters. The effect of these parameters over the NPV of the PM and
RO model is presented in Table 3. In here the NPV of both PM and RO models is reduced
as the replacement and overhaul cost ratio increases, which comes from the direct effect
of the replacement cost over the income. In the case of the flexibility value, it increases as
the minimal repair cost Cm increases, and its increment depends on the corresponding PM
policy interval. However, the value of flexibility differences remain within the same order of
magnitude with an average value of V = 4.3, 8 and 19.2% for Cm = 4, 5, 6 respectively.
Performing a sensitivity analysis on the aging parameter shows that it has a significant
impact on NPV and flexibility value. For an exponential failure rate, the aging parameter is
defined as β1 in the equation λ0 (t) = eβ0 +β1 t . Table 4 and Fig. 8 show the optimal PM policy
and NPV for different values of β1 . As with the cost structure, different values of β1 result
in different optimal policies for the PM and RO models. For both PM and RO models, an
increase in the aging parameter results in a NPV decrease and the flexibility value depends
on the PM policy. This reduction in the NPV difference, as we increase the aging factor,
can be explained because as the equipment ages faster the overhaul/replacement actions
are taken earlier in time, which reduces the flexibility of the model. For the PM policies
(N , T ) = (2, 3), (2, 2) and (1, 2) we obtained average NPV differences between the PM and
RO models of V = 13.6, 11.5 and 12.7%.

123
 Ann Oper Res

Table 2 Effect of the commodity price volatility (σ ), for an equipment aging parameter of β1 = 0.18, on the NPV of RO and PM models for different time horizons

Sigma T=4 T=6 T=8

Real options Periodic Difference Real options Periodic Difference Real options Periodic Difference

0.001 165.57 136.90 28.67 204.01 179.50 24.51 236.97 215.95 21.02
0.01 165.65 137.10 28.55 204.14 179.60 24.54 237.14 216.12 21.02
0.1 174.12 145.55 28.57 218.79 194.09 24.70 257.69 236.47 21.22
0.5 381.56 352.30 29.26 565.10 539.50 25.60 734.31 711.84 22.48
1 1382.59 1354.70 27.89 2435.53 2404.00 31.53 3480.99 3448.33 32.65
1.5 5322.86 5276.50 46.36 11, 809.41 11,721.40 88.01 19, 009.30 18,875.66 133.64

123
 Ann Oper Res

Fig. 5 Net Present Value difference versus commodity price volatility (σ ) for a maintenance period of T = 8
and equipment aging parameter of β1 = 0.18

Fig. 6 Logistic model fit for the probability of a maintenance decision at each price level for T = 8 and a
replacement/overhaul cost ratio of Cr /Co = 3.5

7 Discussion

Equipment overhaul and replacement strategy has a significant impact on the profitability
of a production system. Typically, these strategies consist of a fixed calendar of periodic
interventions based solely on minimizing total equipment costs; this approach ignores price
changes or the possibility of revising and updating maintenance policies. We have shown
that using real options to introduce the possibility of re-evaluating overhaul and replacement

123
Ann Oper Res

Fig. 7 Availability at each price level (In increments of 0.75) for for T = 8 and a replacement/overhaul cost
ratio of Cr /Co = 3.5. Line between price levels connects the average

decisions and the flexibility to respond to price changes adds significant value, increasing
NPV from 36.8 to 8.6% according to the number of periods in the maintenance cycle.
In the proposed model, price variability and mean-reverting behaviour have a significant
impact on the value of flexibility and the decisions taken by the RO model. This finding results
from the fact that flexibility is more valuable in more volatile environments, and is consistent
with previous works that have applied the binomial model to model commodity prices such
as oil and gas and metals (Hahn and Dyer 2008). Under high price volatility scenarios, the
value of flexibility is increased as the duration of the maintenance cycle increases, because
of the benefit of changing courses of actions according to the price trajectories. But under
a low price variability, the value of flexibility is relatively stable across different lengths of
maintenance cycles.
The finding that a flexible maintenance policy adds value is consistent with previous
work that applies flexible models to other external variables, such as technological change
or replacement lead times (Richardson et al. 2013; Mardin and Arai 2011). When prices
are high, the strategy calls for more overhauls and replacements, decreasing the equipment
failure rate and increasing productivity, which allows the firm to take advantage of the higher
prices for its product. In practice, the RO model allows decision makers to consider a set
of policies that can be considered a contingency maintenance plan based on the observed
commodity price.
The relationship between minimal repair, overhaul and replacement costs has a significant
effect on optimal maintenance policies. Specifically, the ratio between replacement and over-
haul costs determines the strategies recommended by both the PM and RO models. Results
show that a higher cost ratio tends to lengthen replacement cycles for both models, encour-
aging more overhauls to control aging. The value of flexibility is significantly affected by
price volatility, replacement/overhaul cost ratio and the equipment aging parameter. These
variables have a similar effect since option value increases along with how important and
irreversible is the decision in question (Dixit and Pindyck 1994).
The proposed model assumes that maintenance decisions are integrated and the implemen-
tation of a flexible strategy does not incur additional costs. In reality, maintenance on asset

123
123
Table 3 Optimal PM policy (N,T), RO and PM net present value and percentage difference for different replacement/overhaul ratio (Cr /Co ) and minimal repair cost (Cm )

Cr /Co Cm = 4 Cm = 5 Cm = 6

(N,T) RO PM Perc. diff. (%) (N,T) RO PM Perc. diff. (%) (N,T) RO PM Perc. diff. (%)

2 (1,2) 612.26 580.12 5.50 (1,2) 444.94 407.17 9.30 (1,2) 283.39 268.70 5.47
2.5 577.12 546.65 5.60 404.95 373.71 8.40 236.40 200.76 17.75
3 545.78 513.18 6.40 370.39 340.24 8.90 (2,2) 201.80 167.29 20.63
3.5 526.22 479.71 9.70 344.55 306.77 12.30 167.64 133.82 25.27
Ann Oper Res
Ann Oper Res

Table 4 Effect of the equipment aging parameter (β1 ) on the optimal PM policy (N,T) and net present value
for the real options and periodic models for T = 8

β1 (N,T) policy Real options Periodic Percentage difference (%)

0.1 (2,2) 432.14 389.03 11.1

0.12 386.26 349.21 10.6
0.14 338.79 307.71 10.1
0.16 295.71 264.46 11.8

0.18 (1,2) 257.69 236.48 9.0

0.2 227.07 208.29 9.0
0.22 195.78 179.35 9.2
0.24 163.58 149.64 9.3
0.26 130.44 119.13 9.5

Fig. 8 Effect of the equipment aging parameter (β1 ) on the net present value for the real options and periodic
maintenance models for T = 8. The flexibility value is not affected significantly by changes in the equipment
aging parameter

intensive industries is normally performed in an owner-client relationship through mainte-

nance contracts (Pascual et al. 2016). This situation may lead to additional costs (flexibility
costs) associated with changes in maintenance decisions, which are not considered in this
model. Another limitation of our model does not take into account for the possibility of
inaction, which it can be also considered as an optimal decision at some level (Décamps et al.
2006). In addition, production chains and equipment fleets in asset intensive industries usu-
ally work together. The inclusion of more than one productive asset in the proposed model
may present relevant new insights. For industries with more complex price and volatility
patterns, a binomial model might not accurately represent reality and a flexible model may
be less attractive.

123
 Ann Oper Res

8 Conclusion

We have developed and implemented a model that defines an optimal equipment overhaul and
replacement policy in a volatile price environment by integrating a real options methodology
with a traditional periodic maintenance model. We consider a single asset and adopt a discrete
approach with a mean-reverting binomial process to model price behavior. Decision makers
can choose to implement an overhaul or a replacement in every time period, and failures are
corrected by minimal repairs.
Our results show that the economic benefits from of the RO approach with exceed those of
the PM models typically used by managers, due to the addition of flexibility in their decision
process. This added flexibility value stems from two factors: the opportunity to account for
price volatility and the possibility of reevaluating maintenance decisions after they are made.
Sensitivity analysis shows that the optimal decisions suggested by the RO model change based
on the commodity price scenario, which provides managers a maintenance contingency plan.
The proposed methodology does not consider the impact of any costs incurred from the
process of changing maintenance policies. Adding such costs would reduce the flexibility
value, and further research is required to determine the potential effect of this factor. Finally,
the non-recombinant nature of the binomial model imposes a limitation on our work, since the
number of price scenarios increases exponentially as the time horizon lengthens. A possible
extension of the proposed model is to consider algorithms that may reduce computational
complexity. Stochastic programming is an alternative that can reduce the number of scenarios,
reducing computational complexity in comparison with the binomial model (Hu et al. 2012).
This extension would allow for extensions of the proposed model to more complex scenar-
ios that could include multiple equipment suppliers, warranty contracts or other sources of
uncertainty, such as technology improvements or maintenance with incomplete information.

References
Abdel-Hameed, M. (2013). Replacement and maintenance policies of devices: A review. In Stochastic relia-
bility and maintenance modeling (pp. 179–189). Berlin: Springer.
Alsyouf, I. (2007). The role of maintenance in improving companies’ productivity and profitability. Interna-
tional Journal of production economics, 105(1), 70–78.
Andersson, H. (2007). Are commodity prices mean reverting? Applied Financial Economics, 17(10), 769–783.
Bastian-Pinto, C., Brandão, L. E., Hahn, W. J. (2010). A non-censored binomial model for mean reverting
stochastic processes. In Annual international conference on real options, vol 14.
Bengtsson, J. (2001). Manufacturing flexibility and real options: A review. International Journal of Production
Economics, 74(1), 213–224.
Bertocchi, M., Moriggia, V., & Dupačová, J. (2006). Horizon and stages in applications of stochastic program-
ming in finance. Annals of Operations Research, 142(1), 63–78.
Bessembinder, H., Coughenour, J. F., Seguin, P. J., & Smoller, M. M. (1995). Mean reversion in equilibrium
asset prices: Evidence from the futures term structure. The Journal of Finance, 50(1), 361–375.
Boyle, P. P. (1977). Options: A monte carlo approach. Journal of Financial Economics, 4(3), 323–338.
Brandão, L. E., & Dyer, J. S. (2005). Decision analysis and real options: A discrete time approach to real
option valuation. Annals of Operations Research, 135(1), 21–39.
Brennan, M. J., & Schwartz, E. S. (1985). Evaluating natural resource investments. Journal of Business, 58(2),
135–157.
Carazas, F., & Souza, G. F. M. D. (2010). Risk-based decision making method for maintenance policy selection
of thermal power plant equipment. Energy, 35(2), 964–975.
Cashin, P., McDermott, C. J., & Scott, A. (2002). Booms and slumps in world commodity prices. Journal of
Development Economics, 69(1), 277–296.
Childs, P. D., Riddiough, T. J., & Triantis, A. J. (1996). Mixed uses and the redevelopment option. Real Estate
Economics, 24(3), 317–339.

123
Ann Oper Res

Cox, J. C., Ross, S. A., & Rubinstein, M. (1979). Option pricing: A simplified approach. Journal of Financial
Economics, 7(3), 229–263.
Crasselt, N., & Lohmann, C. (2016). Considering real options in short-term decision making. Journal of
Management Control, 27(4), 351–369.
Dalal, A. J., & Alghalith, M. (2009). Production decisions under joint price and production uncertainty.
European Journal of Operational Research, 197(1), 84–92.
Décamps, J. P., Mariotti, T., & Villeneuve, S. (2006). Irreversible investment in alternative projects. Economic
Theory, 28(2), 425–448.
Ding, Q., Dong, L., & Kouvelis, P. (2007). On the integration of production and financial hedging decisions
in global markets. Operations Research, 55(3), 470–489.
Dixit, A. (1993). Choosing among alternative discrete investment projects under uncertainty. Economics Let-
ters, 41(3), 265–268.
Dixit, A. K., & Pindyck, R. S. (1994). Investment under uncertainty. Princeton: Princeton University Press.
Gunther McGrath, R., & Nerkar, A. (2004). Real options reasoning and a new look at the research and
development investment strategies of pharmaceutical firms. Strategic Management Journal, 25(1), 1–21.
Haahtela TJ (2010) Recombining trinomial tree for real option valuation with changing volatility. Annual Real
Option Conference
Hahn, W. J., & Dyer, J. S. (2008). Discrete time modeling of mean-reverting stochastic processes for real
option valuation. European Journal of Operational Research, 184(2), 534–548.
Hu, X., Munson, C. L., & Fotopoulos, S. B. (2012). Purchasing decisions under stochastic prices: Approximate
solutions for order time, order quantity and supplier selection. Annals of Operations Research, 201(1),
287–305.
Huchzermeier, A., & Loch, C. H. (2001). Project management under risk: Using the real options approach to
evaluate flexibility in R... D. Management Science, 47(1), 85–101.
Huerta-Cepas, J., Serra, F., & Bork, P. (2016). ETE 3: Reconstruction, analysis, and visualization of phyloge-
nomic data. Molecular Biology and Evolution, 33(6), 1635–1638.
Hull, J. C., & White, A. D. (1994). Numerical procedures for implementing term structure models II: Two-
factor models. The Journal of Derivatives, 2(2), 37–48.
Jaillet, P., Ronn, E. I., & Tompaidis, S. (2004). Valuation of commodity-based swing options. Management
Science, 50(7), 909–921.
Jin, X., & Ni, J. (2013). Joint production and preventive maintenance strategy for manufacturing systems with
stochastic demand. Journal of Manufacturing Science and Engineering, 135(3), 031016.
Kim, Y. H., & Thomas, L. C. (2013). Repair strategies in an uncertain environment: Stochastic game approach.
In Stochastic reliability and maintenance modeling (pp. 123–140). London: Springer.
Kulatilaka, N. (1995). The value of flexibility: A general model of real options. Real options in capital invest-
ment: Models, strategies, and applications, (pp. 89–107).
Lander, D. M., & Pinches, G. E. (1998). Challenges to the practical implementation of modeling and valuing
real options. The Quarterly Review of Economics and Finance, 38(3), 537–567.
Li, C. L., & Kouvelis, P. (1999). Flexible and risk-sharing supply contracts under price uncertainty. Management
Science, 45(10), 1378–1398.
Li, S., Murat, A., & Huang, W. (2009). Selection of contract suppliers under price and demand uncertainty in
a dynamic market. European Journal of Operational Research, 198(3), 830–847.
Lim, S. (2013). A joint optimal pricing and order quantity model under parameter uncertainty and its practical
implementation. Omega, 41(6), 998–1007.
Mardin, F., & Arai, T. (2011). A system dynamics model for replacement and overhaul policies on capital asset
subject to technological change. In The 29th international conference of the system dynamics society.
Messina, V., & Bosetti, V. (2006). Integrating stochastic programming and decision tree techniques in land
conversion problems. Annals of Operations Research, 142(1), 243–258.
Mun, J. (2006). Modeling risk: Applying Monte Carlo simulation, real options analysis, forecasting, and
optimization techniques (Vol. 347). London: Wiley.
Nembhard, H. B., Shi, L., & Aktan, M. (2003). A real options design for product outsourcing. The Engineering
Economist, 48(3), 199–217.
Nguyen, T., Yeung, T., & Castanier, B. (2011). Impact of maintenance on the replacement investment under
technological improvement. In Advances in safety, reliability and risk management: ESREL, 2011, 139.
Paddock, J. L., Siegel, D. R., & Smith, J. L. (1988). Option valuation of claims on real assets: The case of
offshore petroleum leases. The Quarterly Journal of Economics, 103(3), 479–508.
Parida, A., & Kumar, U. (2006). Maintenance performance measurement (mpm): Issues and challenges.
Journal of Quality in Maintenance Engineering, 12(3), 239–251.
Pascual, R., Santelices, G., Liao, H., & Maturana, S. (2016). Channel coordination on fixed-term maintenance
outsourcing contracts. IIE Transactions, 48(7), 1–10.

123
 Ann Oper Res

Richardson, S., Kefford, A., & Hodkiewicz, M. (2013). Optimised asset replacement strategy in the presence
of lead time uncertainty. International Journal of Production Economics, 141(2), 659–667.
Rubinstein, M. (1994). Implied binomial trees. The Journal of Finance, 49(3), 771–818.
Schwartz, E., & Smith, J. E. (2000). Short-term variations and long-term dynamics in commodity prices.
Management Science, 46(7), 893–911.
Schwartz, E. S. (1997). The stochastic behavior of commodity prices: Implications for valuation and hedging.
The Journal of Finance, 52(3), 923–973.
Siddiqui, A., & Fleten, S. E. (2010). How to proceed with competing alternative energy technologies: A real
options analysis. Energy Economics, 32(4), 817–830.
Slade, M. E. (2001). Valuing managerial flexibility: An application of real-option theory to mining investments.
Journal of Environmental Economics and Management, 41(2), 193–233.
Song, D. P. (2009). Production and preventive maintenance control in a stochastic manufacturing system.
International Journal of Production Economics, 119(1), 101–111.
Sumanth, D. J. (1998). Total productivity management: A systematic and quantitative approach to compete in
quality, price, and time. Florida: St Lucie Press.
Wang, H. (2002). A survey of maintenance policies of deteriorating systems. European Journal of Operational
Research, 139(3), 469–489.
Wang, T., & Dyer, J. S. (2010). Valuing multifactor real options using an implied binomial tree. Decision
Analysis, 7(2), 185–195.
Wu, S., & Zuo, M. J. (2010). Linear and nonlinear preventive maintenance models. IEEE Transactions on
Reliability, 59(1), 242–249.
Zambujal-Oliveira, J., & Duque, J. (2011). Operational asset replacement strategy: A real options approach.
European Journal of Operational Research, 210(2), 318–325.
Zeng, S., Zhang, S., et al. (2011). Real options literature review. IBusiness, 3(01), 43.
Zhang, F., & Jardine, A. K. (1998). Optimal maintenance models with minimal repair, periodic overhaul and
complete renewal. IIE Transactions, 30(12), 1109–1119.
Zhang, K., Nieto, A., & Kleit, A. N. (2015). The real option value of mining operations using mean-reverting
commodity prices. Mineral Economics, 28(1–2), 11–22.

123
View publication stats